Bone Morphogenetic Protein-1 Processes the NH2-terminal Propeptide, and a Furin-like Proprotein Convertase Processes the COOH-terminal Propeptide of pro-α1(V) Collagen*

Bone morphogenetic protein-1 (BMP-1) plays key roles in regulating the deposition of vertebrate extracellular matrix; it is the procollagen C-proteinase that processes the major fibrillar collagen types I–III, and it may process prolysyl oxidase to the mature enzyme necessary to the formation of covalent cross-links in collagen and elastic fibers. Type V collagen is a fibrillar collagen of low abundance that is incorporated into and helps regulate the shape and diameter of type I collagen fibrils. Here we show that, in contrast to its action on procollagens I–III, BMP-1 does not cleave the C-propeptide of pro-α1(V) homotrimers. Instead, the single BMP-1-specific cleavage site within pro-α1(V) chains, lies within the large globular N-propeptide. This cleavage site is immediately upstream of a glutamine, thus redefining the specificity of cleavage for BMP-1-like enzymes. It also produces an NH2 terminus that corresponds to an equivalent NH2 terminus on the processed matrix form of the similar α1(XI) chain, thus suggesting physiological significance. Cleavage of the C-propeptide occurs efficiently in recombinant pro-α1(V) homotrimers produced in 293-EBNA human embryonic kidney cells, and this cleavage is shown to occur immediately downstream of the sequence RTRR. This is similar to sites cleaved by subtilisin-like proprotein/prohormone convertases and is shown to be specifically cleaved by the recombinant subtilisin-like proprotein/prohormone convertase furin.

mers capable of forming fibrils (1)(2)(3). In particular, failure to remove the C-propeptide seems incompatible with fibrillogenesis (4). The C-propeptides of procollagens I-III are cleaved by procollagen C-proteinase (5)(6)(7), an activity of bone morphogenetic protein-1 (BMP-1) and mammalian tolloid (mTld), two proteins encoded by alternatively spliced mRNAs of the BMP1 gene (8 -10). BMP-1 is the prototype of a subfamily of astacinlike proteases involved in embryogenetic patterning in diverse organisms (11), in some cases by liberating transforming growth factor-␤-like morphogens from latent complexes (12)(13)(14). Thus, identification of BMP-1 as procollagen C-proteinase provided a link between enzymes involved in matrix deposition and genes involved in pattern formation and suggested that such enzymes may be involved in coordinating various molecular events underlying morphogenesis.
Monomers of the low abundance or minor fibrillar collagen types V and XI are incorporated into the fibrils of the much more abundant collagen types I and II, respectively, and act as regulators of the sizes and shapes of the resultant heterotypic fibrils (15)(16)(17)(18)(19)(20)(21). Type V collagen is most widely distributed in tissues as a heterotrimer of the chain composition ␣1(V) 2 ␣2(V) (22) but is also found, almost exclusively in placenta, as the heterotrimer ␣1(V)␣2(V)␣3(V) (23), and in certain cell types and tissues as an ␣1(V) 3 homotrimer (22, 24 -26). Type XI collagen, in the form of an ␣1(XI)␣2(XI)␣3(XI) heterotrimer (27), was first characterized as a minor collagen of cartilage. However, findings of type XI chains in noncartilaginous tissues (28), of type V chains in cartilage (29), and of cross-type heterotrimers composed of both type V and XI chains (30,31) now suggest that type V and XI chains constitute a single collagen type in which different combinations of chains associate in a tissue-specific manner. Unlike the major fibrillar collagens I-III, collagens V and XI retain N-propeptide sequences (22,24,26,(32)(33)(34)(35)(36). These appear to be of functional importance, since, as shown for type V collagen, they protrude beyond the surface of heterotypic fibrils and may directly control fibrillogenesis by sterically hindering the further addition of collagen monomers to the fibril surface (34).
In the present study, we have produced recombinant pro-␣1(V) homotrimers and subjected them to cleavage with BMP-1. Surprisingly, BMP-1 did not cleave the pro-␣1(V) Cpropeptide but, instead, cleaved at a single specific site within the pro-␣1(V) N-propeptide. Cleavage at this site produces an NH 2 terminus that corresponds to the NH 2 terminus of the processed matrix form of the similar ␣1(XI) chain deposited by chick chondrocytes (35), thus suggesting physiological signifi-cance. NH 2 -terminal sequencing of the pro-␣1(V) C-propeptide showed it to have been cleaved, in 293-EBNA embryonic kidney cell cultures, by a subtilisin-like proprotein convertase (SPC) (37,38). Implications of the data for collagen deposition and morphogenesis are discussed.

EXPERIMENTAL PROCEDURES
pro-␣1(V) Expression Construct-A 3310-base pair NaeI-EcoRI fragment from cDNA clone CW334 (39), corresponding to the 5Ј portion of human pro-␣1(V) sequences, was subcloned between the EcoRV and EcoRI sites of pBluescript II KSϩ (Stratagene) and re-excised with EcoRI and SalI, such that 19 base pairs of pBluescript polylinker were added to the 5Ј of the pro-␣1(V) fragment. A 926-base pair EcoRI-PstI fragment of cDNA CW197 (39), corresponding to the 3Ј portion of pro-␣1(V) sequences, was separately subcloned between the EcoRI and PstI sites of pBluescript II KSϩ, and the EcoRI-SalI fragment derived from CW334 was added to this construct between the CW197 EcoRI site and the SalI site of the pBluescript polylinker. A 1811-base pair EcoRI fragment of cDNA CW32 (39), corresponding to the middle portion of pro-␣1(V), was inserted into the EcoRI site of the preceding construct to reconstitute a full-length pro-␣1(V) coding sequence. The full-length pro-␣1(V) cDNA, extending from nucleotide 224 to nucleotide 6276 of the original pro-␣1(V) sequence (39), was then excised with HindIII and NotI, which cut within the pBluescript polylinker, and inserted between the HindIII and NotI sites of expression vector pCEP-Pu (40).
Production of pro-␣1(V) Procollagen-293-EBNA human embryonic kidney cells (Invitrogen) were maintained in growth medium consisting of Dulbecco's modified Eagle's medium (DMEM) supplemented with 1 mM L-glutamine, 0.1 mM nonessential amino acids, 250 g/ml G418 (Life Technologies, Inc.), and 10% fetal bovine serum (HyClone). Cells at 90% confluence were transfected with 12 g of human pro-␣1(V) expression vector or empty pCEP-Pu vector, per 100-mm tissue culture dish, using Lipofectamine according to the manufacturer's protocol (Life Technologies). After 36 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 twice with phosphate-buffered saline and switched to serum-free DMEM with or without 100 mM L-arginine (Sigma). After 24 h, cells were switched to fresh serum-free DMEM containing 40 g/ml soybean trypsin inhibitor (Sigma) with or without 100 mM L-arginine. Conditioned media were harvested 24 h later, and protease inhibitors were added to final concentrations of 0.2 mM phenylmethylsulfonyl fluoride, 1 mM N-ethylmaleimide, 1 mM p-aminobenzoic acid, and 10 mM EDTA. Samples were centrifuged to remove cell debris, and the supernatants were stored at Ϫ70°C.
Preparation of Type I Procollagen-Confluent human neonatal foreskin fibroblasts, grown to confluence in DMEM containing 10% fetal bovine serum and 1 mM L-glutamine, were washed three times with phosphate-buffered saline and then incubated for 24 h in serum-free DMEM containing 50 g/ml ascorbic acid. Conditioned medium was made 1 mM in phenylmethylsulfonyl fluoride, p-aminobenzoic acid, and N-ethylmaleimide; cell debris was removed by centrifugation; and collagens were precipitated in 25% saturated (NH 4 ) 2 SO 4 for 16 h at 4°C. Precipitates were resuspended in storage buffer (100 mM Tris-HCl, pH 7.5, 400 mM NaCl) and cleared by ultracentrifugation (30 min at 40,000 ϫ g), and the supernatant was dialyzed against DE buffer (75 mM Tris-HCl, pH 7.8, 2 M urea, 0.01% NaN 3 ). Dialyzed supernatant was applied to a 1.5 ϫ 10 cm DEAE-cellulose column (DE-52 Whatman) previously equilibrated with DE buffer. After washing with 50 ml of DE buffer, the column was eluted with a linear gradient of 0 -0.12 M NaCl in 400 ml of DE buffer, and fractions were analyzed by SDS-PAGE. Fractions containing type I procollagen were pooled, dialyzed against storage buffer, and concentrated with a pressure ultrafiltration membrane (Amicon YM 10).
A recombinant truncated/secreted form of furin produced in a baculovirus system and a wild type baculovirus control were the kind gifts of Dr. Claire M. Dubois (Université de Sherbrooke, Sherbrooke, Quebec, Canada). High Five insect cells (ϳ5 ϫ 10 6 ) (Invitrogen) in HYQ-CCM3 serum-free medium (HyClone) were infected at a multiplicity of infection of 5 with baculovirus containing an insert for a soluble COOHterminal truncated form of furin (42,43). Control cells were similarly infected with wild type baculovirus. After 72 h at 27°C, conditioned media were collected, concentrated 4-fold in a Centricon-30 concentrator (Amicon), and made 15% in glycerol for storage at Ϫ70°C. Furincontaining samples had ϳ350 units/ml of furin activity, with 1 unit defined as the amount of enzyme that digests 1 pmol/h of the boc RVRR-aminomethylcoumarin substrate (42,43). Incubations were performed with 3 l of furin (ϳ1 unit) or control preparation and either ϳ2 g of type I procollagen or ϳ1.5 g of pro-␣1(V) collagen in a total reaction volume of 40 l of 50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 1 mM CaCl 2 , 1 mM ␤-mercaptoethanol, 0.1 mM phenylmethylsulfonyl fluoride, 0.1 mg/ml soybean trypsin inhibitor, 10 g/ml leupeptin at 30°C for 5 h.
For collagenase treatment, ϳ2 g of recombinant pro-␣1(V) ϩ pN␣1(V) collagen in buffer A (see above) was made 2 mM in CaCl 2 , 1 mM in N-ethylmaleimide and incubated for 3 h at 37°C in a total reaction volume of 50 l with 1 unit of bacterial collagenase (Advanced Biofactures).
Pepsinized type V collagen (used as a size marker in Fig. 1B) was prepared from human placenta as described by Miller and Rhodes (44).
Amino Acid Sequence Analysis-Proteins on 5 or 12% acrylamide SDS-PAGE gels, with 3.5% stacking gels, were electrotransferred, as described (46), to Sequi-Blot polyvinylidene difluoride membrane (Bio-Rad). 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.

RESULTS
To examine the ability of BMP-1 to process pro-␣1(V) homotrimers, a full-length pro-␣1(V) cDNA was inserted into the episomal expression vector pCEP-Pu (40) for production of recombinant pro-␣1(V) in 293-EBNA human embryonic kidney cells. Mass cultures of 293-EBNA cells, transfected with the pro-␣1(V)/pCEP-Pu construct produced ϳ1 g/ml of a ϳ230-kDa recombinant protein and lesser amounts of a ϳ40-kDa recombinant protein, not produced by 293-EBNA cells transfected with an empty pCEP-Pu vector (Fig. 1A, lanes 2 and 3) or by parental 293-EBNA cells (not shown). The appearance of a single high molecular weight recombinant protein band and a second ϳ40-kDa band, similar to the expected size of cleaved pro-␣1(V) C-propeptide (22), suggested that the large band represented a pN form, lacking the C-propeptide but retaining the N-propeptide. Recently, others using a system similar to that described here, also found that the C-propeptide of recombinant pro-␣1(V) is rapidly cleaved in 293-EBNA cultures (41). The previous report (41) also described rapid processing of the pN form to a form similar in size to pepsin-treated ␣1(V) chains, although such a form was not apparent as a significant band in our system. Pepsin treatment of the recombinant material from our system produced a pepsin-resistant band identical in size to the ␣1(V) band of type V collagen from human placenta (Fig. 1B), showing the presumptive pN␣1(V) form to have a native triple-helical configuration.
Arginine has previously been found to inhibit the normal proteolytic removal of the C-propeptides of procollagens I, II, III, and V in cell-free systems and in tissues (26). We therefore cultured transfected 293-EBNA cells in the presence of 100 mM arginine, resulting in the appearance in these cultures of a ϳ260-kDa form with a concomitant reduction in amounts of 230-and 40-kDa forms detected (Fig. 1A, lane 4). The appear-ance of the 260-kDa form at the expense of the 230-and 40-kDa forms was consistent with their identities as pro-␣1(V), pN␣1(V), and C-propeptide, respectively, suggesting that arginine had partially inhibited C-propeptide cleavage. Treatment of this material with bacterial collagenase (Fig. 1C, lane 1) produced disappearance of the 260-and 230-kDa bands, demonstrating the collagenous nature of these forms. The 40-kDa band remained after collagenase treatment, demonstrating its noncollagenous nature. In addition, an ϳ85-kDa band appeared in the collagenase-treated material, corresponding in size to that of the intact pro-␣1(V) N-propeptide (22). A ϳ42-kDa collagenase-resistant form also appeared, consistent in size to the C-propeptide plus additional collagenase-resistant C-telopeptide sequences. The above results were, thus, consistent with the presence of intact pro-␣1(V) collagen chains in the media of arginine-treated cultures. Since samples prepared from arginine-treated cultures contained presumptive pro-␣1(V) chains, in addition to pN␣1(V) chains, this material, after the removal of arginine, was incubated with BMP-1 to assay for possible cleavage of pro-␣1(V) C-propeptides.
It was expected that if BMP-1 cleaves the C-propeptides of pro-␣1(V) chains, as it does the chains of procollagens I-III (8,9), that the relative intensity of the ϳ260-kDa band would decrease and the relative intensities of the ϳ230and ϳ40-kDa bands would increase, as pro-␣1(V) chains were cleaved to produce additional pN␣1(V) forms and C-propeptides. Surprisingly, incubation of the substrate with BMP-1 resulted in the appearance of two new bands of ϳ200 and ϳ35 kDa ( Fig. 2A,  lane 2). SDS-PAGE under nonreducing conditions was performed to ascertain which forms contained the C-propeptide, the only pro-␣1(V) domain capable of intermolecular disulfide bonding. Under nonreducing conditions, the 40-kDa band, whether derived from substrate that had been incubated with BMP-1 or from substrate not incubated with BMP-1, disappeared from gels ( Fig. 2A, lanes 3 and 4). This was presumably due to the formation of intermolecular disulfide bonds between the 40-kDa form and other species to form high molecular weight bands found near the top of the gel. Such behavior was consistent with identity of the 40-kDa band as the C-propeptide. The 260-kDa form also disappeared under nonreducing conditions, consistent with its identity as full-length, C-propeptide-containing, pro-␣1(V) chains. Interestingly, the ϳ200and ϳ35-kDa bands did not disappear under nonreducing conditions, but the ϳ35-kDa form did show an increase in mobility to ϳ25 kDa. The increase in mobility of the 35/25-kDa form was consistent with its identity as a proteolytic cleavage product of the N-propeptide, containing cysteine residues capable of affecting electrophoretic mobility through formation of intramolecular disulfide bonds. The fact that the 200-kDa form was smaller than the presumptive pN␣1(V) band and the fact that it did not disappear under nonreducing conditions suggested that this form lacked both a C-propeptide and some portion of the N-propeptide, presumably that portion corresponding to the 35/25-kDa fragment. The 230-kDa band, when derived from substrate not incubated with BMP-1, was not diminished in intensity on gels under nonreducing conditions. In contrast, when derived from substrate incubated with BMP-1, the 230-kDa band was reduced in intensity by about one-half under nonreducing conditions (Fig. 2, A and B). This was interpreted to mean that the 230-kDa band from BMP-1-digested material was a doublet containing pN␣1(V) forms, incapable of forming intermolecular disulfide bonds, and a pC␣1(V) form of similar size that retained the C-propeptide but from which a portion of the N-propeptide had been removed by BMP-1.
The various electrophoretic results described above suggested that BMP-1 cleaved at a single specific site within the pro-␣1(V) N-propeptide. Moreover, since the 40-kDa band did not increase in intensity upon cleavage of substrate with BMP-1 ( Fig. 2A, lanes 1 and 2), these results also suggested that BMP-1 does not cleave pro-␣1(V) C-propeptides. Samples previously run on 4 -15% acrylamide gradient gels ( Fig. 2A) were electrophoresed on 5% acrylamide gels (Fig. 2B) for better resolution of high molecular weight forms, and the predicted identities of the various forms, based on the electrophoretic data described above, are shown in Fig. 2C.
To ascertain whether the identities of the various SDS-PAGE bands were as inferred above, 260-, 230-, and 40-kDa bands derived from substrate that had not been incubated with BMP-1 were isolated from a reducing gel; 40-and 35-kDa bands derived from substrate that had been incubated with BMP-1 were isolated from a reducing gel; and 230-and 200-kDa bands derived from substrate that had been incubated with BMP-1 were isolated from a nonreducing gel (see Fig. 3) and subjected to automated Edman degradation for determination of NH 2 -terminal amino acid sequences. As expected, NH 2 -terminal sequences of both the 260-kDa form and the 230-kDa form (from either reducing or nonreducing gels) corresponded to the NH 2 terminus of the pro-␣1(V) N-propeptide ( Fig. 3), beginning with the first amino acid after the predicted cleavage site (39) of the prepro-␣1(V) signal peptide. This confirmed the identities of the 260-and 230-kDa forms as pro-␣1(V) and pN␣1(V) chains, respectively. In addition, the 35-kDa product resulting from cleavage with BMP-1, was found to have the same NH 2 -terminal sequences as pro-␣1(V) and pN␣1(V) chains, thus confirming it as a cleavage product of the pro-␣1(V) N-propeptide.
The that BMP-1 had cleaved the pro-␣1(V) N-propeptide between residues 254 and 255 of the published prepro-␣1(V) amino acid sequence (39,47) (Fig. 3). This site occurs immediately downstream of a pair of cysteines that divide the pro-␣1(V) Npropeptide into a somewhat basic upstream domain and a downstream domain rich in acidic residues and tyrosines (39). The upstream domain is fairly conserved between pro-␣1(V), pro-␣1(XI), and pro-␣2(XI) chains and has been designated the PARP domain, whereas the downstream domain, which is not conserved among the same three chains, has been designated the variable region (48) (see Fig. 3). The NH 2 terminus remaining on the 200-kDa form corresponds to an equivalent NH 2 terminus of the processed matrix form of the very similar ␣1(XI) collagen chain (35). Pro␣1(V), pro-␣1(XI), and pro-␣2(XI) chains form a subgroup among fibrillar procollagen chains on the basis of sequence similarities, structures of cognate genes, and size and configuration of N-propeptides (39,(47)(48)(49)(50)(51). Thus, an alignment of pro-␣1(V), pro-␣1(XI), and pro-␣2(XI) sequences, from various species, is shown for the region in which BMP-1 cleaves human pro-␣1(V) and in which chicken pro-␣1(XI) is cleaved to produce the ␣1(XI) chain of chondrocyte matrix (Fig. 4). Glutamines at the P1Ј and P2 positions, and a proline at the P3Ј position of the pro-␣1(V) cleavage site are conserved in the pro-␣1(XI) and pro-␣2(XI) chains. Comparison of these residues with residues flanking previously described BMP-1/mTld cleavage sites (8,52,53) (data not shown) found prolines also located at the P3Ј position of the pro-␣1(III), pro-␣2(I), and prolysyl oxidase sites. Prolines two or three amino acids from cleavage sites can strongly affect the activity of astacin-like proteases (11). However, prolines are not found at similar positions at other previously identified BMP-1/mTld cleavage sites. Other features common to the pro-␣1(V) BMP-1 cleavage site and previously described sites were not immediately apparent.
The NH 2 -terminal amino acid sequence of the 40-kDa form shows it to be cleaved C-propeptide (Fig. 3). However, the cleavage site is not at either of the two previously predicted pro-␣1(V) C-propeptide cleavage sites, which were based on the placement of aspartate residues in the pro-␣1(V) C-telopeptide region (39,47). Instead, cleavage occurs immediately down-  (39,47), hamsters (39), and chickens (34); from pro-␣1(XI) chains of humans (49,57), mice (58), rats (57,59), and chickens (60); or from pro-␣2(XI) chain of humans (48,61) and mice (62), were manually aligned in the two regions of interest. The arrows denote sites of pro-␣1(V) cleavage by BMP-1 and by a furin-like SPC. Residues found in peptide sequences from the NH 2 termini of BMP-1-cleaved human pro-␣1(V) and the matrix form of chicken ␣1(XI) (35) are underlined. Residues conserved in all three chains for all species in the area of BMP-1 cleavage of pro-␣1(V) are in boldface type, as are residues corresponding to a consensus sequence for cleavage by furin-like SPCs.

FIG. 3. Comparison of peptide sequences of recombinant pro-␣1(V)-and pro-␣1(V)-derived forms with the human pro-␣1(V) sequence deduced from cDNA.
To determine NH 2 -terminal amino acid sequences, 260-(V7), 230-(V5 and V6), and 200-kDa (V4) bands were isolated from a 5% acrylamide SDS-PAGE gel, and 40-(V2 and V3) and 35-kDa (V1) bands were isolated from a 12% acrylamide SDS-PAGE gel for automated Edman degradation. Bands were derived from substrate either incubated or not incubated with BMP-1 and were from samples either reduced or not reduced with ␤-mercaptoethanol (␤-ME), as indicated. Residues derived from NH 2 -terminal sequencing are aligned with the published human pro-␣1(V) sequence deduced from cDNA (39,47), and positions of residues in the published sequence (39,47) corresponding to the first residues of the various peptides are given. Residues corresponding to a consensus sequence for cleavage by furin-like SPCs are in boldface type. Positions of BMP-1 and furin cleavage sites are shown in relation to a schematic of the pro-␣1(V) chain. C-pro, C-propeptide; proline/ arginine-rich protein (PARP) and Var, PARP and variable subdomains of the N-propeptide, respectively. stream of the sequence RTRR, thus predicting that the pro-␣1(V) C-propeptide is cleaved in 293-EBNA cultures by a mammalian SPC of which furin is the prototype (37,38). Interestingly, alignment shows similar (R/K)XRR sequences, suitable for cleavage by furin-like SPCs, in the C-telopeptide regions of the pro-␣1(V), pro-␣1(XI), and pro-␣2(XI) chains of various species (Fig. 4). As noted above, intensity of the 40-kDa band did not increase upon incubation of substrate with BMP-1, thus suggesting that BMP-1 cleavage of the pro-␣1(V) C-propeptide did not occur. Similarly, NH 2 -terminal sequences were the same whether the 40-kDa form had been derived from substrate incubated, or not incubated, with BMP-1, (Fig. 3, bands V2 and V3, respectively), with no evidence of any additional NH 2 -sequences in the 40-kDa form from sample incubated with BMP-1. Thus, cleavages by BMP-1 did not occur at additional sites within the C-telopeptide region.
To further explore the apparent cleavage of the pro-␣1(V) C-propeptide by furin-like SPCs, substrate containing pro-␣1(V) and pN␣1(V) chains was incubated with a recombinant truncated/secreted form of furin (42) produced in a baculovirus system. As can be seen (Fig. 5, lane 2), furin cleaves pro-␣1(V) chains to produce additional pN␣1(V) chains and C-propeptides, with the absence of any nonspecific cleavages elsewhere in the pro-␣1(V) molecule. Interestingly, just as 100 mM arginine inhibited cleavage of about 50% of pro-␣1(V) C-propeptides by endogenous SPC in our culture system, 100 mM arginine also inhibited cleavage of about 50% of pro-␣1(V) C-propeptides by furin in this in vitro assay (data not shown).
As a control for the specificity of the furin and BMP-1 preparations used in this study, both preparations were incubated with type I procollagen substrate. As expected, BMP-1 cleaved type I procollagen only at C-propeptide cleavage sites, producing the C1(I) and C2(I) C-propeptide subunits of the pro-␣1(I) and pro-␣2(I) chains, respectively (Fig. 5, lane 4). In contrast, there was an absence of observable cleavages of type I procollagen upon incubation with furin (Fig. 5, lane 5). In addition, we have found an absence of cleavage of type II procollagen by furin. 2 These results strengthen the conclusion that furin and BMP-1 cleavages at particular sites within the pro-␣1(V) Cand N-propeptides, respectively, are highly specific, similar to the degree of specificity with which BMP-1 cleaves type I procollagen C-propeptides. DISCUSSION BMP-1 and mTld play multiple roles in regulating matrix deposition including provision of procollagen C-proteinase activity for procollagens I-III (8,9), activation of lysyl oxidase (52), and biosynthetic processing of laminin 5 (53), a component of skin basement membranes. Here, we describe a new BMP-1 activity, processing of the pro-␣1(V) N-propeptide, suggesting an additional role in matrix deposition. Biological significance for this cleavage is suggested, since it creates an NH 2 terminus corresponding to the NH 2 terminus of the chondrocyte matrix form of the highly similar ␣1(XI) collagen chain (35). In fact, previous rotary shadowing analysis of ␣1(V) chains in chick cornea (34) estimated an NH 2 terminus only 2 amino acids upstream of the cleavage site demonstrated here, while a number of other studies (22) show retention of ␣1(V) N-propeptide sequences consistent in size with cleavage at the BMP-1cleaved site. Some of the latter studies (22) show retained sequences to contain most or all sulfated tyrosines found in full-length pro-␣1(V) N-propeptides, also consistent with cleavage at the site described here. In contrast, some studies (24,32,36) show ␣1(V) chains extracted from tissues to contain Npropeptide sequences shorter than would be produced by cleavage at the BMP-1 site. These latter findings may suggest either additional processing of the ␣1(V) N-propeptide in some tissues or artifactual proteolysis of ␣1(V) chains during tissue extraction.
The pro-␣1(V) BMP-1 cleavage site identified here differs from previously described sites (8,52,53), 3 most notably in lacking an otherwise invariant aspartate at the P1Ј position. It should be stressed, however, that even the limited conservation found in residues flanking previously described BMP-1/mTld cleavage sites is somewhat misleading, since (i) conserved residues flanking procollagen I-III chain sites may reflect the similar evolutionary origin of these chains rather than an absolute functional requirement for such residues and (ii) prolysyl oxidase and laminin 5 were first identified as possible BMP-1/mTld substrates based on similarities of in vivo cleavage sites, to those of procollagens I-III (52,53). The relative lack of similarity between residues immediately flanking the pro-␣1(V) site and previous BMP-1/mTld cleavage sites indicates that BMP-1 and mTld, like other astacin-like proteases (11), are not highly specific for such residues and suggests a reappraisal of features that influence cleavage of substrates by BMP-1 and mTld. Certainly, future candidates for BMP-1/ mTld substrates should not be limited to proteins with cleavage sites resembling those of procollagen I-III C-propeptides.
Original predictions of potential pro-␣1(V) C-propeptide cleavage sites (39,47) assumed cleavage by an activity similar to that which cleaves procollagens I-III and were based on the positions of aspartates in the C-telopeptide region. Nevertheless, sequence differences in this region between pro-␣1(V) and procollagen I-III chains and differential processing of pro-␣1(V) and pro-␣1(I) C-propeptides in a given cell type had suggested that different enzymes might cleave the C-propeptides of these different procollagen chain types (39). Data presented here indicate that the pro-␣1(V) C-propeptide is not cleaved by BMP-1 but that it is cleaved by cells at a site consistent with cleavage by a furin-like SPC and by furin itself. Moreover, alignment shows conservation of the sequence (R/K)XRR, suitable for cleavage by furin-like SPCs (37,38), in the C-telopep- tide regions of pro-␣1(V), pro-␣1(XI), and pro-␣2(XI) chains (Fig. 4), suggesting a common mechanism of cleavage. In support of this possibility, preliminary data indicate that endogenous pro-␣1(XI) C-propeptides are cleaved at the predicted (R/K)XRR site in the A204 rhabdomyosarcoma cell line. 2 SPC processing of some procollagen V/XI C-propeptides adds another level of regulation to the fibrillogenesis of type I/V and type II/XI heterotypic fibrils and may link fibrillogenesis with other processes, since the same types of furin-like SPCs process growth factors, receptors, matrix metalloproteinases (37,38), and probably BMP-1/mTld-like proteases (11).
Of the known mammalian SPCs, only furin (SPC1), PACE 4 (SPC4), PC5/PC6 (SPC6), and PC7/PC8 (SPC7) have the broad ranges of expression (37,38) necessary for processing the widely distributed pro-␣1(V), pro-␣1(XI), and pro-␣2(XI) chains. Although furin and SPCs 4, 6, and 7 all typically cut at RX(K/R)R or similar sites, studies have found that most substrates cleaved at correct sites by recombinant furin are not correctly processed by the other enzymes (38). Thus, demonstration here of pro-␣1(V) cleavage by furin and furin's role as a major proprotein processing enzyme of the constitutive secretory pathway suggest furin as a reasonable candidate for cleavage of pro-␣1(V), pro-␣1(XI), and pro-␣2(XI) C-propeptides in vivo. In its native form, furin has a transmembrane domain and is primarily localized in the trans-Golgi but cycles between the trans-Golgi and the plasma membrane (37,38). Thus, pro-␣1(V), pro-␣1(XI), and pro-␣2(XI) C-propeptide cleavage may occur in the trans-Golgi or at the cell surface.