Biosynthetic Processing of the Pro-α1(V)2Pro-α2(V) Collagen Heterotrimer by Bone Morphogenetic Protein-1 and Furin-like Proprotein Convertases*

The low abundance fibrillar collagen type V is incorporated into and regulates the diameters of type I collagen fibrils. Bone morphogenetic protein-1 (BMP-1) is a metalloprotease that plays key roles in regulating formation of vertebrate extracellular matrix; it cleaves the C-propeptides of the major fibrillar procollagens I–III and processes precursors to produce the mature forms of the cross-linking enzyme prolysyl oxidase, the proteoglycan biglycan, and the basement membrane protein laminin 5. Here we have successfully produced recombinant pro-α1(V)2pro-α2(V) heterotrimers, and we have used these to characterize biosynthetic processing of the most prevalent in vivo form of type V procollagen. In addition, we have compared the processing of endogenous pro-α1(V) chains by wild type mouse embryo fibroblasts and by fibroblasts derived from embryos doubly homozygous null for the Bmp-1 gene and for a gene encoding the closely related metalloprotease mammalian Tolloid-like 1. Together, results presented herein indicate that within pro-α1(V)2pro-α2(V) heterotrimers, pro-α1(V) N-propeptides and pro-α2(V) C-propeptides are processed by BMP-1-like enzymes, and pro-α1(V) C-propeptides are processed by furin-like proprotein convertases in vivo.

The major fibrillar collagen types I-III are synthesized as procollagens with N-propeptides 1 and C-propeptides that are cleaved to produce mature triple helical monomers capable of forming fibrils (1)(2)(3). In particular, failure to remove C-propeptides seems incompatible with fibrillogenesis (4). The major procollagen C-propeptides are cleaved by the metalloprotease bone morphogenetic protein 1 (BMP-1) and by the closely related metalloproteases mammalian Tolloid-like (mTLD) and mammalian Tolloid-like 1 (mTLL-1) (5-7). 2 These enzymes also process the prodomains of various other precursor proteins involved in formation of the extracellular matrix (8 -10) and cleave Chordin (7,11), an extracellular antagonist of signaling by certain transforming growth factor-␤-like BMPs such as BMP-4 (12). Thus, BMP-1 and related proteases may orchestrate deposition of matrix components with BMP signaling in morphogenetic events.
The minor fibrillar collagen type V is broadly distributed in tissues as an ␣1(V) 2 ␣2(V) heterotrimer (13,14) that is incorporated into type I collagen fibrils and acts to regulate the diameters of the resulting heterotypic fibrils (15)(16)(17)(18)(19). A lower abundance ␣1(V)␣2(V)␣3(V) form of type V collagen has been isolated from placenta (20,21), whereas a rare ␣1(V) 3 homotrimer has been reported in a limited number of cell types and tissues (22)(23)(24). Type XI collagen, a minor fibrillar collagen of cartilage with the chain composition ␣1(XI)␣2(XI)␣3(XI) (25), appears to interact with the major cartilage collagen type II in a manner analogous to the interaction of types V and I in other tissues (26,27). Findings of type XI chains in noncartilaginous tissues (28), type V chains in cartilage (28), and heterotrimers comprising both type V and XI chains (30,31) now suggest V and XI as a single collagen type in which different chain combinations form in a tissue-specific manner. Unlike major fibrillar collagens, collagens V and XI retain partial N-propeptide sequences (22,24,(32)(33)(34)(35)(36) that appear to protrude beyond the surface of heterotypic fibrils and may regulate fibrillogenesis by hindering addition of monomers to the fibril surface (34) (for reviews of collagens V and XI see Refs. 37 and 38).
Low levels of the minor fibrillar collagens in tissues and cell cultures have limited their characterization. To begin characterizing biosynthetic processing, we previously produced recombinant pro-␣1(V) 3 homotrimers in 293-EBNA cells. Unexpectedly, pro-␣1(V) 3 C-propeptides were found to be efficiently cleaved at a furin consensus cleavage site by 293-EBNA cells and by recombinant furin in vitro (39), whereas in vitro assays showed BMP-1 to cleave within the pro-␣1(V) 3 N-propeptide at a site predicted to produce that portion of pro-␣1(V) thought to protrude beyond type I/V heterotrimer surfaces (39). Such as-says did not detect BMP-1 cleavage of pro-␣1(V) 3 C-propeptides (39). More recently, Kessler et al. (40), using a similar system, found BMP-1 capable of cleaving pro-␣1(V) 3 C-propeptides in vitro, albeit less efficiently than N-propeptides (40).
Here recombinant pro-␣1(V) 2 pro-␣2(V) heterotrimers, representing what is by far the most prevalent in vivo form of type V procollagen, are successfully produced, and their processing, by cellular proteases in 293-EBNA cultures and by recombinant BMP-1 and furin in vitro, is characterized. In addition, biosynthetic processing of endogenous pro-␣1(V) chains produced by fibroblasts derived from mouse embryos homozygous null for the genes encoding BMP-1, mTLD, and mTLL-1 is analyzed. Results support the conclusion that pro-␣1(V) N-propeptides and pro-␣2(V) C-propeptides are processed by BMP-1-like enzymes and pro-␣1(V) C-propeptides are processed by furin-like enzymes, in vivo. Implications of the data for ECM formation and morphogenesis are discussed.

EXPERIMENTAL PROCEDURES
Production of Pro-␣1(V) 2 Pro-␣2(V) Heterotrimers-Transfected human embryonic kidney 293-EBNA cells expressing ϳ1 g/ml recombinant pro-␣1(V) 3 homotrimers from an episomal vector were derived as described previously (39) and 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, 214 g/ml G418 (Invitrogen), 5 g/ml puromycin, and 10% fetal bovine serum. Cells at ϳ80% confluence were retransfected with 10 g of the expression construct pGGH31 described previously (41), comprising full-length human pro-␣2(V) cDNA downstream of human cytomegalovirus promoter/ regulatory sequences. After 17 h, cells were switched to fresh growth medium, and 48 h later cells were harvested by trypsinization. Cells were dilution-cloned in 96-well culture dishes (Costar), and media of 144 clones were analyzed for pro-␣1(V) and pro-␣2(V) chains by precipitating with trichloroacetic acid at 50 M and visualization by SDS-PAGE and Coomassie Blue staining. Chains were tested for triple helicity by bringing media to a final concentration of 0.5 M acetic acid and incubating overnight with 100 g/ml pepsin at 4°C, followed by precipitation in 1.7 M NaCl, washing of the pellet with 70% ethanol, and analysis by SDS-PAGE and Coomassie Blue staining. The clonal line producing highest levels of pepsin-resistant ␣1(V) 2 ␣2(V) heterotrimers was amplified for production of recombinant material used in this study.
To obtain recombinant pro-␣1(V) 2 pro-␣2(V) heterotrimers, cell cultures were rinsed twice with phosphate-buffered saline and placed in serum-free DMEM with or without either 100 mM L-arginine or varying concentrations of decanoyl-RVKR-chloromethyl ketone (Bachem), as indicated in the text. After 24 h, cells were switched to fresh serum-free DMEM containing 40 g/ml soybean trypsin inhibitor (Sigma) and either decanoyl-RVKR-chloromethyl ketone, 100 mM L-arginine, or neither. 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 debris, and supernatants were dialyzed against 50 mM Tris-HCl, pH 8.6, containing 0.1 mM phenylmethylsulfonyl fluoride, 1 mM N-ethylmaleimide, 1 mM p-aminobenzoic acid, and 5 mM EDTA at 4°C to precipitate type V procollagen, as described (39,42). Pepsin digestion was with 20 g/ml pepsin (Sigma) in 0.5 M acetic acid (pH 2.0) for 4 h at 4°C.
Mouse Embryo Fibroblasts (MEFs)-Fibroblasts isolated from 13.5day post-conception embryos, as described (43), and passaged no more than five times were grown to ϳ80% confluency and treated overnight in growth medium containing 50 g/ml ascorbate. The next day, cells were switched to serum-free medium containing 10 ng of transforming growth factor-␤, 40 g/ml soybean trypsin inhibitor, and 50 g/ml ascorbate, with or without 20 M decanoyl-RVKR-chloromethyl ketone. Media were collected after 48 h, made 30% saturated with (NH 4 ) 2 SO 4 , and nutated overnight at 4°C. Samples were centrifuged for 15 min at 38,581 ϫ g and pellets were transferred to new tubes in Ϫ20°C acetone. Pellets were washed 3 times in 12.5 mM Tris-HCl, pH 7.5, 75% ethanol at 4°C, resuspended in 4ϫ SDS-PAGE loading buffer, and boiled 5 min.
Immunoblots-Subsequent to SDS-PAGE, proteins were transferred to Immobilon-P membranes (Millipore) as described (44). Blots were blocked in phosphate-buffered saline, 0.05% Tween 20 (T-PBS) and incubated 15 h with primary antibody diluted 1:10,000 in T-PBS, 3% bovine serum albumin. After six 10-min washes with T-PBS, blots were blocked 1 h in T-PBS, 3% bovine serum albumin, and then incubated 30 min with peroxidase-conjugated donkey anti-rabbit IgG (Amersham Biosciences) diluted 1:5000 in T-PBS. Blots were washed six times, for 10 min each, with T-PBS, incubated 5 min with SuperSignal West Pico substrate (Pierce), and exposed to film. Apparent molecular weights of bands were estimated by comparison to electrophoretic mobilities of prestained standards (Bio-Rad).
Polyclonal antibodies to the pro-␣1(V) variable subdomain were raised in rabbits against the peptide CADDLEGEFTEETIRNLD, corresponding to residues 397-413 of the published human pro-␣1(V) sequence (45) plus an additional cysteine for coupling to keyhole limpet hemocyanin. Antibodies were affinity-purified on columns of the same peptide coupled to TC gel (BIOSOURCE) via the cysteine thiol (44).
Amino Acid Sequence Analysis-Subsequent to SDS-PAGE, proteins were electrotransferred as described (44) onto Sequi-Blot polyvinylidene difluoride membrane (Bio-Rad). Bands were excised, and NH 2terminal sequencing was performed by automated Edman degradation at the Harvard Microchemistry Facility.
Pepsin digestion of recombinant material from media yielded resistant bands equal in size to the ␣1(V) and ␣2(V) chains of pepsinized type V collagen from human placenta (Fig. 1B), thus demonstrating the recombinant type V procollagen to have a native triple helical configuration. Moreover, because pro-␣2(V) chains are pepsin-sensitive unless in heterotrimers together with pro-␣1(V) chains (41), the pepsin-resistant nature of the material in Fig. 1B demonstrates all recombinant pro-␣2(V) chains to be contained in heterotrimers also containing pro-␣1(V) chains. The roughly 2:1 ratio of pro-␣1(V) to pro-␣2(V) chains in material either treated or untreated with pepsin (e.g. Fig. 1B, lanes 1 and 2) indicates the majority of recombinant type V procollagen produced in the current study to be in the form of pro-␣1(V) 2 pro-␣2(V) heterotrimers.

BMP-1 Can Efficiently Cleave Pro-␣1(V) N-propeptides and Pro-␣2(V) C-propeptides and Less Efficiently Cleave Pro-␣1(V)
C-propeptides in Vitro-Previously, we observed in vitro cleavage by BMP-1 of the N-but not C-propeptides of recombinant pro-␣1(V) 3 homotrimers (39). More recently, however, Kessler et al. (40), although confirming our observation of the specific cleavage by BMP-1 of pro-␣1(V) 3 N-propeptides, observed that BMP-1 is capable of processing the C-propeptides of these molecules as well. An obvious difference between the two previous studies was the higher specific activity of BMP-1 preparations used by Kessler et al. (40), which completely processed pro-␣1(V) 3 N-propeptides after a 2-h incubation, compared with those used by Imamura et al. (39), which processed only ϳ50% of pro-␣1(V) 3 N-propeptides after a 17-h incubation, under similar reaction conditions. Another difference between the two studies is that, prior to incubation with BMP-1, pro-␣1(V) 3 homotrimers used as substrate by Kessler et al. (40) had been protected with decanoyl-RVKR-chloromethyl ketone from cleavage by the furin-like activity of 293-EBNA cultures, whereas pro-␣1(V) 3 homotrimers used as substrate by Imamura et al. (39) had been protected from C-propeptide cleavage by use of arginine. Thus, in the current study we utilized relatively high specific activity recombinant BMP-1, recently prepared in 293 cells (7), so that processing of pro-␣1(V) C-propeptides in pro-␣1(V) 2 pro-␣2(V) heterotrimers might be detected if BMP-1 is indeed capable of such cleavage. In addition, two different batches of pro-␣1(V) 2 pro-␣2(V) heterotrimeric substrate were prepared in which cleavage by furin-like activities in 293-EBNA cells was inhibited either by arginine or by decanoyl-RVKR-chloromethyl ketone. As can be seen (Fig. 3A), BMP-1 processing of pro-␣1(V) 2 pro-␣2(V) heterotrimers, produced in the presence of either arginine or decanoyl-RVKR-chloromethyl ketone, yields three bands. The largest (Fig. 3A, lanes 2 and 3) corresponds to pC-␣1(V), in which only the PARP subdomain of the N-propeptide has been removed, and which consistently migrates on SDS-PAGE gels with a mobility faster than that of pN-␣1(V) (Fig. 3A, lanes 1  and 3, also see Fig. 5). The other 2 bands are mature ␣1(V) and pN-␣2(V) chains (Fig. 3A, lanes 2 and 3). The former of these lacks the N-propeptide PARP subdomain and the C-propeptide.  Fig. 3A, lane 3). Therefore, in vitro BMP-1 processing of pro-␣1(V) N-propeptides is more efficient than processing of pro-␣1(V) C-propeptides, consistent with the observation of Kessler et al. (40) of a 4-fold more efficient in vitro cleavage by BMP-1 of pro-␣1(V) 3 N-propeptides than of C-propeptides.
Unlike the processing of pro-␣1(V) C-propeptides, in vitro processing by BMP-1 of the C-propeptides of pro-␣2(V) chains within the same pro-␣1(V) 2 pro-␣2(V) heterotrimers is efficient and complete (Fig. 3A). Subsequent to incubations with BMP-1, only pN-␣2(V) forms are observable, with no residual pro-␣2(V) chains and no evidence of processing within the N-propeptide or other regions of the pro-␣2(V) chain (Fig. 3A). The cleavage site for proteolytic removal of the pro-␣2(V) C-propeptide has not previously been determined experimentally. To determine this site, aliquots of the same BMP-1-treated samples analyzed by SDS-PAGE on the 4% polyacrylamide gel of Fig. 3A were electrophoresed on a 10% polyacrylamide gel for separation of the low molecular weight cleavage products (Fig. 3B) and subsequent determination of their NH 2 -terminal amino acid sequences by automated Edman degradation. The NH 2 -terminal sequence DQAAPDDKXK (X designates an undetermined residue) identified an ϳ42-kDa band (C-␣2(V), Fig. 3B, lanes 2 and  3) as the pro-␣2(V) C-propeptide, also identifying the peptide bond between Glu-1250 and Asp-1251 of the prepro-␣2(V) amino acid sequence (Refs. 46 -48; GenBank TM accession number Y14690) as the BMP-1 cleavage site. Alignment of fulllength pro-␣ chains shows pro-␣2(V) Asp-1251 to correspond to aspartates conserved at similar positions in the C-telopeptides of the pro-␣1(I), pro-␣2(I), pro-␣1(II), and pro-␣1(III) major procollagen chains (Fig. 3C). Such sequence comparisons previously prompted Myers et al. (46) to predict the peptide bond between Glu-1250 and Asp-1251 as the probable site for cleavage of the pro-␣2(V) C-propeptide. We have noted previously (9) the presence of residues with aromatic side chains and/or Met residues upstream of various BMP-1 cleavage sites with P 1 Ј aspartates. As can be seen (Fig. 3C), a Phe is in the P3 position of the pro-␣2(V) cleavage site. Neither residues with aromatic side chains nor methionines are located directly upstream of the site at which BMP-1 is capable of cleaving pro-␣1(V) chains in vitro, perhaps contributing to the lower in vitro efficiency of cleavage at this site. NH 2 -terminal sequencing of an ϳ35-kDa band in the BMP-1-treated material (N-␣1(V), Fig. 3B, lanes 2 and 3) showed it to correspond to the NH 2 -terminal portion of the pro-␣1(V) Npropeptide previously shown to be cleaved in pro-␣1(V) 3 homotrimers by BMP-1 (39), whereas NH 2 -terminal sequencing of an ϳ38-kDa band (C-␣1(V) BMP-1 , Fig. 3B, lanes 2 and 3) showed it to correspond to pro-␣1(V) C-propeptides cleaved between Asp-1594 and Asp-1595. The latter bond was previously predicted by this laboratory (45) as the probable pro-␣1(V) C-propeptide cleavage site, based solely on nucleic acid sequences and the assumption that minor fibrillar procollagen C-propeptides would be cleaved by the same enzymes that process the major fibrillar procollagen C-propeptides. This same site was more recently shown by Kessler et al. (40) to be the site where BMP-1 is capable of in vitro cleavage of pro-␣1(V) 3 homotrimer C-propeptides. An ϳ40-kDa band observed in pro-␣1(V) 2 pro-␣2(V) samples from 293-EBNA cultures in which furin-like activity had been partially inhibited by incubation with arginine (e.g. C-␣1(V) furin , Fig. 3B, lane 1) was identified by NH 2 -terminal sequencing as pro-␣1(V) C-propeptide cleaved adjacent to the furin recognition sequence RTRR. The absence of this band in similar samples subsequent to treatment with BMP-1 (e.g. Fig. 3B, lane 2) is consistent with  1 and 2) or with 20 M decanoyl-RVKR-chloromethyl ketone (lanes 3 and 4) were incubated for 17 h in the absence (lanes 1 and 4) or presence (lanes 2 and 3) of BMP-1. Samples were analyzed by SDS-PAGE on a 4 (A) or 10% (B) acrylamide gel and stained with Coomassie Blue. C, alignment of the site where BMP-1 cleaves the C-propeptides of pro-␣2(V) chains contained within pro-␣1(V) 2 pro-␣2(V) heterotrimers, with the in vivo cleavage sites of procollagen I-III C-propeptides (5,50,51). Conserved aspartates, residues with aromatic side chains, and a methionine residue are in boldface. the observation of Kessler et al. (40) that free pro-␣1(V) Cpropeptides cleaved by furin-like activities remain potential substrates for further processing by BMP-1.

BMP-1-like Proteases Cleave Pro-␣1(V) N-propeptides and Furin-like Proteases Cleave Pro-␣1(V) C-propeptides in
Vivo-We next sought to correlate the patterns of processing obtained by cleavage of recombinant type V procollagen with processing of endogenous type V procollagen chains in fibrogenic cells. Toward this end, we developed high affinity polyclonal antibodies directed against sequences in the pro-␣1(V) N-propeptide variable subdomain that in vitro cleavage assays in this and our previous study (39) predict to lie within that portion of the N-propeptide retained on mature processed ␣1(V) chains (see Fig. 4). As can be seen (Fig. 5A), cultures of normal human diploid fibroblasts produce endogenous processing intermediates corresponding exactly in size to the pN-␣1(V), pC-␣1(V), and mature ␣1(V) forms derived from partial cleavage of recombinant pro-␣1(V) 2 pro-␣2(V) heterotrimers with furin and BMP-1. It should be noted that most or all endogenous pro-␣1(V) chains detected in the human fibroblast cultures are likely to be contained within pro-␣1(V) 2 pro-␣2(V) heterotrimers, as this is the predominant form of type V procollagen produced by normal adult human dermal fibroblasts.
To investigate directly the in vivo roles of BMP-1-like proteases in biosynthethic processing of pro-␣1(V) chains, we compared processing of endogenous pro-␣1(V) chains in cultures of fibroblasts derived from wild type mouse embryos to that within cultures of fibroblasts derived from genetically altered mouse embryos doubly homozygous null for the Bmp1 and Tll1 genes, which together encode the closely related proteases BMP-1, mTLD, and mTLL-1. Use of the doubly null cells removed possible functional redundancy, as we have shown previously that products of the two genes overlap in their substrate specificities 2 (7,9,10,43,49). Processing was also compared for the two types of mouse embryo fibroblasts (MEFs) grown in the presence or absence of decanoyl-RVKR-chloromethyl ketone, to assay for possible roles of furin-like proteases in pro-␣1(V) processing in these cells. These comparisons were performed using the polyclonal antibodies described above, which recognize the same area of the pro-␣1(V) variable subdomain in human and mouse.
As can be seen (Fig. 5B), two bands are detected in conditioned media of wild type MEFs. The smaller band corresponds in size to mature ␣1(V) chains from which both the C-propeptide and the PARP subdomain of the N-propeptide have been removed, whereas the larger band corresponds in size to pN-␣1(V), which retains the complete N-propeptide. Thus, as in normal adult human fibroblasts, pro-␣1(V) chains are proteolytically processed at both ends in wild type MEF cultures, yielding mature ␣1(V) chains similar in size to those observed upon in vitro cleavage of recombinant pro-␣1(V) chains with a combination of BMP-1 and furin. Interestingly, when decanoyl-RVKR-chloromethyl ketone is added to wild type MEF cultures, both ␣1(V) and pN-␣1(V) forms disappear, to be replaced by full-length pro-␣1(V) chains and by pC-␣1(V) forms, which lack N-, but retain C-propeptides. As noted above, pC-␣1(V) has a faster mobility than pN-␣1(V), as is evident in the blot of Vertical arrows mark the sites of cleavage by BMP-1-and furin-like proteases, as indicated by data presented in the present study and by Imamura et al. (39). A horizontal black bar over the pro-␣1(V) variable subdomain marks the approximate position of amino acid residues that correspond to a peptide used in the production of anti-pro-␣1(V) antibodies described in the text. 5B, derived from an SDS-PAGE gel run for an extended period of time. The observed replacement of ␣1(V) and pN-␣1(V) with pro-␣1(V) and pC-␣1(V) chains upon addition of decanoyl-RVKR-chloromethyl ketone to wild type MEF cultures indicates that processing of C-, but not N-propeptides, is blocked by this specific furin inhibitor.
In contrast to wild type MEFs, only one band is detected in media of Bmp1Ϫ/Ϫ;Tll1Ϫ/Ϫ doubly null MEFs not treated with furin inhibitor. This band corresponds in size to pN-␣1(V) (Fig.  5B). Thus, absence of BMP-1-like proteases in these mutant MEF cultures blocks processing of N-but not C-propeptides. As expected from results obtained with the wild type MEFs, addition of decanoyl-RVKR-chloromethyl ketone to doubly null MEF cultures blocks cleavage of the C-propeptide, resulting in the appearance of only unprocessed pro-␣1(V) chains (Fig. 5B). Results obtained from characterization of MEF cultures are thus consistent with the conclusion that BMP-1-like enzymes process the N-propeptide and furin-like enzymes process the C-propeptide of pro-␣1(V) chains in vivo, at least in MEFs.

DISCUSSION
Type V collagen occurs predominantly as ␣1(V) 2 ␣2(V) heterotrimers in vivo (37,38), whereas ␣1(V) 3 homotrimers have been detected in tissues in a very limited number of instances (22,24). Nevertheless, we previously began study of type V procollagen biosynthetic processing using recombinant pro-␣1(V) 3 homotrimers (39), production of which was more straightforward than production of recombinant pro-␣1(V) 2 pro-␣2(V) heterotrimers. In that study (39), recombinant pro-␣1(V) 3 C-propeptides were efficiently and exclusively cleaved at a C-telopeptide furin consensus site by transfected 293-EBNA cultures and were cleaved at the same site by recombinant furin in vitro. Cleavage in 293-EBNA cultures did not occur adjacent to any of 3 Asp residues within the pro-␣1(V) C-telopeptide region, despite cleavage of all major fibrillar procollagen C-propeptides by BMP-1-like proteases at sites with P 1 Ј Asp residues (50,51). These indications that furin-like proteases rather than BMP-1-like proteases process pro-␣1(V) 3 C-propeptides were surprising, as the overall similarity of domain structures had suggested that all fibrillar procollagen C-propeptides might be cleaved by the same enzymatic activity. Indeed, upon determining the pro-␣1(V) sequence (45), this lab predicted a C-propeptide cleavage site between Asp residues 1594 and 1595 based solely on the observation that P 1 Ј Asp residues had invariably marked the C-propeptide cleavage sites of the major fibrillar procollagen chains. However, we also noted that proposal of this putative cleavage site was somewhat arbitrary (45), as three different Asp residues exist within the pro-␣1(V) C-telopeptide region, none of which are adjacent to residues resembling those adjacent to P 1 Ј Asp residues at C-propeptide cleavage sites of the major fibrillar chains. In fact, this apparent lack of a conserved consensus sequence prompted us to suggest (45) that different enzymes might be responsible for cleaving pro-␣1(V) C-propeptides and those of the major fibrillar procollagen chains.
Imamura et al. (39) observed BMP-1 to readily process pro-␣1(V) 3 homotrimer chains at a single site, within the N-propeptide. Biological significance was suggested for this cleavage, as it creates an NH 2 terminus that corresponds to the NH 2 terminus of the chondrocyte matrix form of the highly similar ␣1(XI) chain (35) and leaves retained sequences corresponding in size to retained ␣1(V) N-propeptide sequences observed in other studies (34,37). However, the pro-␣1(V) N-propeptide cleavage site was unusual in that, unlike previous BMP-1 cleavage sites, it lacked a P 1 Ј Asp (39). Subsequently, Kessler et al. (40) demonstrated that, at higher levels of activity than those used by Imamura et al. (39), BMP-1 is capable of in vitro cleavage between Asp residues 1594 and 1595 to cleave the C-propeptides of pro-␣1(V) 3 homotrimers, previously protected from 293-EBNA furin-like proteases by decanoyl-RVKR-chloromethyl ketone (40).
The present study addresses and resolves issues of in vivo relevance raised by earlier studies. We have successfully produced recombinant pro-␣1(V) 2 pro-␣2(V) heterotrimers representing what is, by far, the predominant in vivo form of type V procollagen. We demonstrate that in pro-␣1(V) 2 pro-␣2(V) heterotrimers, as in pro-␣1(V) 3 homotrimers, pro-␣1(V) N-propeptides are cleaved in vitro by BMP-1 between residues 254 (Ser) and 255 (Gln), whereas pro-␣1(V) C-propeptides are cleaved exclusively between residues 1585 (Arg) and 1586 (Asn) at a furin consensus sequence by a furin-like activity in 293-EBNA cultures and by recombinant furin in vitro. In contrast, pro-␣2(V) chains within the same pro-␣1(V) 2 pro-␣2(V) heterotrimers are impervious to cleavage by the furin-like activity of 293-EBNA cells or by recombinant furin, whereas BMP-1 efficiently cleaves pro-␣2(V) C-propeptides at a single site between residues 1250 (Glu) and 1251 (Asp). These data, plus other data here and in a previous report (39) showing procollagens I and II to be impervious to cleavage by furin, indicate cleavage of pro-␣1(V) C-propeptides by furin-like proteases to be highly specific.
As reported by Kessler et al. (40) for pro-␣1(V) 3 homotrimers, we show BMP-1 is capable of in vitro cleavage of the C-propeptides of pro-␣1(V) chains within pro-␣1(V) 2 pro-␣2(V) heterotrimers, albeit with less efficiency than the in vitro cleavage by BMP-1 of pro-␣1(V) N-propeptides or pro-␣2(V) C-propeptides. However, we also address directly the in vivo relevance of this observation, using a genetic approach that employs MEFs derived from mouse embryos doubly homozygous null for the Bmp1 gene, which encodes alternatively spliced mRNAs for BMP-1 and mTLD (52), and the Tll1 gene, which encodes the closely related protease mTLL-1 (7,53). Results from the MEF study are inconsistent with the probability that BMP-1-like enzymes cleave pro-␣1(V) C-propeptides in vivo. In contrast, they strongly support in vivo relevance for pro-␣1(V) N-propeptide processing by BMP-1-like enzymes. Moreover, use of the highly specific furin-inhibitor decanoyl-RVKR-chloromethyl ketone demonstrates furin-like enzymes to be responsible for cleaving the C-propeptides of endogenous pro-␣1(V) chains, likely to be contained in pro-␣1(V) 2 pro-␣2(V) heterotrimers. Thus, cleavage of pro-␣1(V) C-propeptides exclusively by furinlike enzymes in transfected 293-EBNA cultures is not an artifact of this recombinant system.
In summary, data presented here indicate that BMP-1-like enzymes process N-propeptides and furin-like enzymes process C-propeptides of pro-␣1(V) chains in vivo. Although Kessler et al. (40) suggest that pro-␣1(V) C-propeptides may be cleaved by BMP-1-like enzymes in some cells and furin-like enzymes in others, this seems unlikely, as various furin-like proteases are broadly expressed and furin itself, a major processing enzyme of the constitutive secretory pathway, is expressed in all tissues and cell lines examined to date (54,55). Conservation of sequences suitable for cleavage by furin-like proteases within C-telopeptide regions of pro-␣1(V), pro-␣3(V), pro-␣1(XI), and pro-␣2(XI) chains (39,56) suggests C-propeptides of this subset of fibrillar procollagen chains to be cleaved by furin-like enzymes. This proposed similarity in processing fits the observation that these 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 (56,57). However, although residues at the P 1 Ј, P 3 Ј, and P 2 positions of the pro-␣1(V) chain N-propeptide BMP-1 cleavage site are conserved at the same positions in pro-␣1(XI) and pro-␣2(XI) N-propeptides in various species, such conservation is not evident in the pro-␣3(V) N-propeptide, despite an overall conservation of N-propeptide domain structure (56). Thus, future studies will address the issue of which proteases may be involved in processing the pro-␣3(V) N-propeptide.
Cleavage of C-propeptides of pro-␣1(V)-like chains by furinlike proteases may link fibrillogenesis to other processes governed by these enzymes (54,55). Based on data presented here, the C-propeptide of pro-␣2(V) chains, which by various criteria are more similar to procollagen I-III chains than to pro-␣1(V)like chains (47,48,57,58), is likely cleaved by BMP-1-like enzymes. Such cleavage, and likely cleavage by BMP-1-like proteases of the C-propeptide of pro-␣3(XI) chains, a modified product of the type II collagen pro-␣1(II) gene (29), may coordinate deposition of minor and major fibrillar collagen monomers within heterotypic fibrils.