Biosynthetic processing of the Pro-α1(V)Pro-α2(V)Pro-α3(V) procollagen heterotrimer

Type V collagen is a quantitatively minor fibrillar collagen comprised of different chain compositions in different tissues. The most widely distributed form, an α1(V)2α2(V) heterotrimer, regulates the physical properties of type I/V heterotypic collagen fibrils via partially processed NH2-terminal globular sequences. A less characterized α1(V)α2(V)α3(V) heterotrimer has a much more limited distribution of expression and unknown function(s). We characterized the biosynthetic processing of pro-α1(V)2pro-α2(V) procollagen previously and showed it to differ in important ways from biosynthetic processing of the major fibrillar procollagens I–III. Here we have successfully produced recombinant pro-α1(V)pro-α2(V)pro-α3(V) heterotrimers. We use these, and mouse embryo fibroblasts doubly homozygous null for the Bmp1 gene, which encodes the metalloproteinase bone morphogenetic protein-1 (BMP-1), and for a gene encoding the closely related metalloproteinase mammalian Tolloid-like 1, to characterize biosynthetic processing of pro-α1(V)pro-α2(V)pro-α3(V) heterotrimers, thus completing characterization of type V collagen biosynthetic processing. Whereas pro-α1(V) and pro-α2(V) processing in pro-α1(V)pro-α2(V)pro-α3(V) heterotrimers is similar to that which occurs in pro-α1(V)2pro-α2(V) heterotrimers, the processing of pro-α3(V) by BMP-1 occurs at an unexpected site within NH2-terminal globular sequences. We also demonstrate that, despite similarities in NH2-terminal domain structures, pro-α2(V) NH2-terminal globular sequences are not cleaved by ADAMTS-2, the metalloproteinase that cleaves the N-propeptides of the major fibrillar procollagen chains.

The major fibrillar collagen types I-III are synthesized as procollagens with N-and C-propeptides 1 that are proteolytically removed to yield mature triple helical monomers capable of forming fibrils (1,2). Impairment in the ability to cleave the N-propeptide of procollagen type I results in abnormal fibril morphology and forms of Ehlers-Danlos syndrome type VII (2), whereas failure to remove major fibrillar procollagen C-propeptides may be incompatible with fibrillogenesis (3). The N-propeptides of procollagens I-III are cleaved by the metalloproteinase ADAMTS-2 (4,5) and perhaps by the closely related ADAMTS-3 and -14 as well (6,7). The C-propeptides of procollagens I-III are cleaved by the metalloproteinase bone morphogenetic protein 1 (BMP-1) (8,9) and by other members of a small family of metalloproteinases closely related to BMP-1 (10,11). BMP-1 and related proteinases also process the prodomains of a number of other precursors to produce the mature functional forms of a variety of proteins involved in formation of the exracellular matrix, (12)(13)(14)(15)(16). BMP-1-like proteinases also process chordin (10,11,17), an extracellular antagonist of signaling by transforming growth factor-␤-like BMPs, such as BMP-4 (18), and process within propeptide sequences to activate the transforming growth factor-␤-like protein growth differentiation factor 8/myostatin (19). Thus, BMP-1-like proteinases may coordinate extracellular matrix formation with signaling by a subset of transforming growth factor-␤-related proteins in morphogenesis and homeostasis.
Monomers of the minor, low abundance fibrillar collagen types V and XI are incorporated into growing type I and type II collagen fibrils, respectively, and play roles in regulating the shapes and diameters of the resultant heterotypic fibrils (20 -25). Type V collagen is distributed broadly in type I collagencontaining tissues as a heterotrimer of the chain composition ␣1(V) 2 ␣2(V) (26) and is found in a limited number of cell types and tissues as a rare ␣1(V) 3 homotrimer (26 -29). In addition, a relatively uncharacterized ␣1(V)␣2(V)␣3(V) heterotrimer has been isolated from human placenta (30 -32) and has also been reported in uterus, skin, and synovial membranes (26,(33)(34)(35). Detection of ␣3(V) expression in nascent ligamentous attachments of developing joints, membranous linings of developing skeletal muscle, and in developing and regenerating peripheral nerves in mouse and rat (36,37) suggests roles for the ␣1(V)␣2(V)␣3(V) heterotrimer in these tissues as well. Although type XI collagen was first described as an ␣1(XI)␣2(XI)␣3(XI) heterotrimer confined to cartilage (38), it is now apparent that the separately discovered collagen types V and XI may be viewed as constituting a single collagen type in which different combinations of chains associate in a tissue-specific manner. This conclusion is based on the finding of type XI chains in non-cartilaginous tissues (39), of type V chains in cartilage (40), and of cross-type heterotrimers composed of both type V and XI chains (41,42) and on the extreme similarities in sequence and domain structure shown by certain of the type V and XI collagen chains. Unlike fibrillar collagens I-III, ␣1(V) 2 ␣2(V) and ␣1(XI)␣2(XI)␣3(XI) heterotrimers retain partial NH 2 -terminal globular sequences (29,28,(43)(44)(45)(46)(47). These protrude beyond the surface of heterotypic fibrils and may regulate fibrillogenesis by hin-dering addition of collagen monomers to the fibril surface (45).
Pro-␣2(V) closely resembles the major fibrillar procollagen chains in domain structure (50,51), as does pro-␣3(XI), which is a modified product of the type II collagen pro-␣1(II) gene (40). In contrast, the pro-␣1(V), pro-␣1(XI), pro-␣2(XI), and pro-␣3(V) chains, which have triple helical and C-propeptides that resemble those of the other fibrillar procollagen chains, form a subgroup with a shared NH 2 -terminal globular protein domain structure that differs in size and configuration from the NH 2terminal globular sequences of the other chains (36). We reported previously the surprising findings that C-propeptides of pro-␣1(V) chains in either ␣1(V) 3 or ␣1(V) 2 ␣2(V) trimers are cleaved by furin-like proprotein convertases, whereas BMP-1 cleaves at a specific site within pro-␣1(V) NH 2 -terminal globular sequences (48,49). Processing at similar sites and by similar proteinases appears to hold for pro-␣1(XI) as well (11,52). Putative furin recognition sites in pro-␣2(XI) and pro-␣3(V) telopeptides suggest that the C-propeptides of these chains may also be cleaved by furin-like proteinases, although the question has not been addressed experimentally. However, although residues flanking the BMP-1 cleavage site in pro-␣1(V) NH 2 -terminal sequences are conserved in the NH 2 -terminal sequences of pro-␣1(XI) and pro-␣2(XI), such conservation is lacking in pro-␣3(V) (49). Thus, it remains to be determined whether pro-␣3(V) NH 2 -terminal globular sequences are processed and, if so, at which site(s), and by what type of proteinase(s).
Here we address experimentally in vitro and in vivo processing of the pro-␣3(V) chain and the question of whether the pro-␣2(V) N-propeptide is susceptible to cleavage by ADAMTS-2.
For production of recombinant pro-␣1(V)pro-␣2(V)pro-␣3(V) heterotrimers, the full-length human pro-␣1(V) cDNA (48) was excised from pBluescript II KSϩ with HindIII and SpeI and was inserted between the HindIII and XbaI sites of expression vector pBudCE4.1 (Invitrogen). A full-length human pro-␣3(V) cDNA was then assembled from previously described human pro-␣3(V) cDNAs (36) by separately subcloning cDNA fragments corresponding to the 5Ј-and 3Ј-halves of pro-␣3(V) cDNA into pBluescript II KSϩ, excising the 5Ј-cDNA with NotI and KpnI and the 3Ј-cDNA with KpnI and BssHII, and inserting the two pieces together between the NotI and MluI sites of the pro-␣1(V) cDNA-containing pBudCE4.1 vector. The resulting vector thus expresses both pro-␣1(V) and pro-␣3(V) mRNAs, transcribed via two separate promoters. The pro-␣2(V) cDNA-containing pcDNA4/TO/myc-His Zeo construct described above was cleaved with AgeI, then filled with dCTP using the Klenow fragment, and subsequently cut with AflII. The resulting full-length pro-␣2(V) cDNA was then inserted into a pcDNA5/TO vector (Invitrogen) that had been cut with Bsp120I, filled with dGTP using the Klenow fragment, and subsequently cut with AflII. These vectors were cotransfected into T-REx-293 cells, as described below, and then selected for zeocin and hygromycin resistance.
T-REx-293 cell cultures maintained in complete medium (Dulbecco's modified Eagle's medium supplemented with 1 mM L-glutamine, antibiotic-antimycotic (Invitrogen), and 10% fetal bovine serum (Hyclone)) were transfected with the constructs described above, using Lipo-fectAMINE according to the manufacturer's protocol (Invitrogen). In each case, cells were transfected first with the appropriate pro-␣2(V) expression construct, followed by selection for resistance to the corre-sponding antibiotic, ring cloning, and screening of individual clones, after induction with tetracycline, for pro-␣2(V) expression by immunoblot, using anti-␣2(V) polyclonal antibodies (Santa Cruz Biotechnology). Clonal lines expressing the highest levels of pro-␣2(V) were then transfected either with the pC1-neo/pro-␣1(V) expression vector or with the pBudCE4.1/pro-␣1(V)/pro-␣3(V) expression vector, followed by selection for resistance to the corresponding antibiotic, ring cloning, and screening of individual clones for pro-␣1(V) expression by immunoblot, using previously described (49) anti-␣1(V) polyclonal antibodies. Clonal lines expressing the highest levels of pro-␣1(V) chains were then tested for production of either pro-␣1(V) 2 pro-␣2(V) or pro-␣1(V) pro-␣2(V)pro-␣3(V) heterotrimers, after induction with tetracycline, by SDS-PAGE and Coomassie Blue staining.
T-REx-293 clones expressing highest levels of pro-␣1(V) 2 pro-␣2(V) or pro-␣1(V)pro-␣2(V)pro-␣3(V) heterotrimers were incubated with 50 g/ml ascorbate for 20 h before induction of collagen production. Cells were washed in phosphate-buffered saline (PBS), then incubated in serum-free Dulbecco's modified Eagle's medium for 45 min, rewashed with PBS, and then incubated in serum-free Dulbecco's modified Eagle's medium containing 100 g/ml soybean trypsin inhibitor, 50 g/ml ascorbate, and 1 mM tetracycline (Sigma) in the absence or presence of 100 mM L-arginine or 20 M highly specific furin inhibitor decanoyl-RVKR-chloromethyl ketone (Bachem), as noted in the text. Conditioned media were recovered 24 h later and replaced with fresh media for a second harvest after an additional 24 h. Protease inhibitors were added to collected media to final concentrations of 0.2 mM phenylmethylsulfonyl fluoride, 1 mM N-ethylmaleimide, and 1 mM p-aminobenzoic acid; media were centrifuged to remove debris and were then stored at Ϫ70°C.
Stored samples were thawed and dialyzed against 50 mM Tris-HCl, pH 8.6, 0.1 mM phenylmethylsulfonyl fluoride, 1 mM N-ethylmaleimide, and 1 mM p-aminobenzoic acid at 4°C, as described previously (48), for precipitation of type V procollagens. Precipitates were collected by centrifugation at 65,000 ϫ g for 30 min, and pellets were resuspended in buffer A (50 mM Tris-HCl, pH 7.5, 150 mM NaCl).
Pepsinization-Pro-␣1(V) 2 pro-␣2(V) and pro-␣1(V)pro-␣2(V)pro-␣3(V) heterotrimers were subjected to digestion with 100 g/ml pepsin (Sigma) in 0.5 N acetic acid, pH 2.0, for 6 h at 4°C. The reaction was stopped by neutralizing with a final concentration of 0.5 M NaOH and by adding pepstatin to a final concentration of 30 g/ml. Pepsin-resistant collagen chains were precipitated with ethanol and analyzed by SDS-PAGE.
Immunoblots-Samples were subjected to SDS-PAGE and subsequently electrotransferred to Immobilon-P membranes (Millipore) as described previously (54). Blots were blocked with 3% bovine serum albumin in T-PBS (PBS and 0.05% Tween) for 2 h and incubated with primary antibody at a 1:6,000 dilution for anti-␣1(V) antibodies, 1:5,000 for anti-␣2(V) antibodies, and 1:4,000 for anti-␣3(V) antibodies. Blots were washed with T-PBS and blocked with 3% bovine serum albumin in T-PBS followed by incubation with secondary antibodies at 1:5,000 dilution. After eight washes with T-PBS, blots were incubated for 5 min in SuperSignal West Pico substrate (Pierce) and exposed to Scientific Imaging film (Kodak).
Affinity-purified polyclonal antibodies raised against the variable domain of pro-␣1(V) have been described previously (49), as has the anti-␣3(V) antibody raised against the rat ␣3(V) homolog, P200 (53). The latter antibody was the generous gift of David J. Carey (Sigfried and Janet Weis Center for Research).
Immunoprecipitation-Pro-␣1(V)pro-␣2(V)pro-␣3(V) heterotrimers were immunoprecipitated with anti-␣3(V) antibody using essentially the same methodology as described by Scott et al. (17). Briefly, 500 ng of recombinant protein, precipitated under low salt conditions from conditioned media (see above), was incubated either with protein A-Sepharose beads (Amersham Biosciences) or with protein A-Sepharose beads prebound to antibody against the rat P200 ␣3(V) homolog. Proteins bound to the beads were eluted by boiling in SDS-PAGE sample buffer and analyzed by SDS-PAGE on 5% acrylamide gels followed by Western blot analysis with anti-␣1(V), anti-␣2(V), and anti-␣3(V) antibodies.
Enzyme Cleavage Assays-A human BMP-1 cDNA, which is fulllength except that signal peptide sequences have been replaced by sequences encoding the BM40 signal peptide, for optimization of secretion (10) was inserted between the Acc65I and NotI sites of the pcDNA4/TO/myc-His expression vector. This construct was transfected into T-REx 293 cells, which were selected in Zeocin, and a clonal line expressing high levels of tetracycline-inducible BMP-1 was identified. Recombinant BMP-1 produced by this line was purified, as described previously (10), and 110 ng was incubated with 4 g of pro- To test the ability of ADAMTS-2 to cleave the pro-␣2(V) chain, recombinant pro-␣1(V)pro-␣2(V)pro-␣3(V) heterotrimers were metabolically radiolabeled with L-[2,3-3 H]proline (PerkinElmer Life Sciences), using conditions described previously (55). The radiolabeled procollagen, purified by low salt precipitation, as described above, was treated with ADAMTS-2 for 20 h at 37°C, using conditions described previously (5), followed by SDS-PAGE on a 5% acrylamide gel. Coomassie Bluestained gels were treated with En 3 Hance (PerkinElmer Life Sciences) and exposed to Hyperfilm (Amersham Biosciences).
Mouse Embryo Fibroblasts (MEFs)-MEFs were prepared from 13.5day postconception wild type and Bmp1;Tll1 doubly homozygous null embryos, as described previously (11), and were then serially passaged in complete medium to produce immortal cell lines. These lines were then electroporated with an expression construct comprising full-length pro-␣3(V) cDNA inserted between the AflII and NotI sites of expression vector pcDNA3.1/Hygro (Invitrogen). For electroporation, 7 ϫ 10 6 cells in suspension were washed once in ice-cold PBS, resuspended in 500 l of ice-cold PBS, transferred to a GenePulser cuvette (Bio-Rad), and incubated 10 min on ice with 5 g of expression construct linearized by cutting with SspI within vector sequences. Electroporation was performed using the GenePulser II system (Bio-Rad) at 950 microfarads/ 226 V, after which cells were allowed to recover by incubating for 15 min on ice and then transferred to 10-cm tissue culture plates and grown to confluence in complete medium before selection with hygromycin. Resistant colonies were allowed to grow to 75% confluent mass cultures. The latter were pretreated overnight with 50 g/ml ascorbate and then were treated with 2 ng/ml recombinant human transforming growth factor-␤1 (R&D Systems) (for the induction of endogenous pro-␣1(V) and pro-␣2(V) chains) in the presence of 100 g/ml soybean trypsin inhibitor and 50 g/ml ascorbate in the presence or absence of 20 M decanoyl-RVKR-chloromethyl ketone. Conditioned media were harvested 48 h later and were subjected to precipitation with 30% saturated ammonium sulfate overnight at 4°C in the presence of 0.1 mM phenylmethylsulfonyl fluoride, 1 mM N-ethylmaleimide, and 1 mM paminobenzoic acid. Samples were centrifuged at 40,000 ϫ g and pellets were transferred to fresh Microfuge tubes in cold acetone, washed three times in 75% ethanol and 12.5 mM Tris-HCl, pH 7.5, resuspended in 4 ϫ SDS-PAGE loading buffer, and boiled for 10 min before loading onto a 4% acrylamide SDS-polyacrylamide gel.
Amino Acid Sequence Analysis-After separation on either 5% or 10% SDS-polyacrylamide gels, proteins were electrotransferred to Sequi-Blot polyvinylidene difluoride membranes (Bio-Rad). Bands were excised, and NH 2 -terminal amino acid sequencing by Edman degradation was performed at the Harvard Microchemistry Facility.

Production of Stable Recombinant Pro
Heterotrimers-Attempts to produce either recombinant pro-␣3(V) 3 homotrimers or heterotrimers composed solely of pro-␣3(V) and pro-␣1(V) chains failed to yield detectable pro-␣3(V) chains (data not shown), suggesting that pro-␣3(V) 3 homotrimers and pro-␣3(V)pro-␣1(V) heterotrimers are not formed, or are not stable. In contrast, cotransfection of cells with expression vectors for producing pro-␣3(V), pro-␣1(V), and pro-␣2(V) chains was successful in yielding stable pro-␣1(V)pro-␣2(V)pro-␣3(V) heterotrimers (Fig. 1). In addition to producing bands identical to those produced by cells transfected with expression constructs for pro-␣1(V) and pro-␣2(V) alone, clonal lines transfected with expression vectors for all three chains produced a novel ϳ190 kDa band as well (Fig. 1A). Because we have found previously that 293 cells transfected with pro-␣1(V) cDNA vectors primarily produce the pN form of pro-␣1(V)-derived sequences (Refs. 48 and 49 and Fig. 1A), the novel 190 kDa band in Fig. 1A was, by analogy, provisionally designated pN␣3(V). Immunoblots using antibodies specific for pro-␣1(V)-derived sequences (49) or antibodies specific for pro-␣3(V)-derived sequences (53) confirmed that cells transfected either with expression constructs for pro-␣1(V) and pro-␣2(V) alone or with constructs for expression of all three type V chains had similar banding patterns for pro-␣1(V)-derived proteins ( Fig. 1B) but that only cells transfected with constructs for expression of all three type V chains expressed the 190-kDa species, which was pro-␣3(V)-derived (Fig. 1C). Pepsin digestion of recombinant material from cells transfected either with constructs for pro-␣1(V) and pro-␣2(V) alone, or with constructs for expression of all three type V chains, demonstrated that the latter cells produced a novel pepsin-resistant chain and a pattern of three resistant chains identical with that of authentic ␣1(V)␣2(V)␣3(V) collagen heterotrimers isolated from tissues ( Fig. 1D and Refs. 30 and 32). The ratio of pepsin-resistant ␣1(V), ␣2(V), and ␣3(V) chains in the sample suggests that the great majority of type V procollagen produced by cells transfected with constructs for expression of all three type V chains is in the form of stable pro-␣1(V)pro-␣2(V)pro-␣3(V) heterotrimers, possibly admixed with lesser amounts of pro-␣1(V) 2 pro-␣2(V) heterotrimers and/or pro-␣1(V) homotrimers.
To ascertain directly that heterotrimers with the composition pro-␣1(V)pro-␣2(V)pro-␣3(V) heterotrimers are formed in the expression system, immunoprecipitations were performed with antibody specific for the pro-␣3(V) chain, and it was determined whether or not pro-␣1(V)-and pro-␣2(V)-derived chains were coimmunoprecipitated. As can be seen (Fig. 2), anti-pro-␣3(V) antibody coimmunoprecipitated both pro-␣1(V) and pro-␣2(V) chains from samples derived from cells transfected with expression vectors for all three chains. The specificity of the anti-pro-␣3(V) antibody is demonstrated by the observation that neither pro-␣1(V)-nor pro-␣2(V)-derived bands were detectable upon immunoprecipitation using this antibody, and samples derived from cells transfected only with pro-␣1(V) and pro-␣2(V) expression vector. The most straightforward interpretation of the various data presented above is that heterotrimers with the composition pro-␣1(V)pro-  (48,49). To determine whether the pro-␣3(V) C-propeptide is or is not cleaved by a similar activity, a clonal line of cells producing recombinant pro-␣1(V)pro-␣2(V)pro-␣3(V) heterotrimers was incubated either in the presence of 100 mM arginine, which partially blocks such activity (48,49), or in the presence of decanoyl-RVKR-chloromethyl ketone, a potent and highly specific furin inhibitor. As can be seen (Fig. 3), the shift of pN␣1(V) to full-length pro-␣1(V) chains is evident in the presence of arginine or decanoyl-RVKR-chloromethyl ketone, whereas there is no shift in the mobility of pro-␣2(V) chains, which have no furin recognition sites. In contrast, a slight but reproducible shift is detectable in the mobility of the novel 190 kDa band, consistent with its identity as the pN␣3(V) chain and consistent with the possibility that the pro-␣3(V) C-propeptide is cleaved by furin-like proprotein convertase activity in cell culture.
Pro-␣3(V) Sequences Are Cleaved by BMP-1 in Vitro-Previously, we have demonstrated that C-propeptides of pro-␣2(V) chains and NH 2 -terminal globular sequences of pro-␣1(V) chains are cleaved by BMP-1 in vitro (48,49). To determine whether pro-␣3(V) chains are processed by BMP-1, recombinant collagenous material from the media of a cell line expressing all three type V procollagen chains was incubated 20 h in the presence or absence of BMP-1. As can be seen (Fig. 4A), BMP-1 not only cleaves pN␣1(V) chains to produce mature ␣1(V) and cleaves pro-␣2(V) chains to produce pN␣2(V), but also induces a mobility shift in pN␣3(V) chains to produce a faster mobility band. The most straightforward interpretation of the data is that the pN␣3(V) chain is cleaved within NH 2terminal globular sequences to produce what may correspond to the mature ␣3(V) chain. The band corresponding to the putative ␣3(V) chain was isolated and subjected to NH 2 -terminal sequencing via automated Edman degradation and was indeed found to have been cleaved within NH 2 -terminal globular sequences. The NH 2 -terminal sequence of AQAQAVLQQT identified the peptide bond between Gln-463 and Ala-464 of the human prepro-␣3(V) amino acid sequence (Ref. 36; GenBank accession number AF177941) as the BMP-1 cleavage site. This site is within a short non-triple-helical region that lies between the major collagenous domain (COL1) and a small collagenous domain (COL2) just COOH-terminal of the variable subdomain of pro-␣3(V) NH 2 -terminal sequences (Figs. 5 and 8).
To examine more closely the nature of processing of the pro-␣3(V) chain, BMP-1-cleaved samples, similar to those of Fig. 4A, were electrophoresed on a 10% SDS-polyacrylamide gel for separation of the low molecular mass cleavage products (Fig. 4B) and subsequent determination of their NH 2 -terminal amino acid sequences by automated Edman degradation. As can be seen (Fig. 4B), four bands of approximate molecular masses of 52, 46, 41, and 38 kDa were visible by Coomassie Blue staining. It should be noted that molecular masses for cleaved pro-␣3(V) N-and C-propeptides, pro-␣1(V) N-and Cpropeptides, and pro-␣2(V) C-propeptides are 45.377, 27.301, 24.326, 28.166, and 27.198, respectively, based on amino acid sequences. However, each of these peptides is thought to be glycosylated, and in fact, we have previously estimated molec-  lanes 1 and 2) or pro-␣1(V)pro-␣2(V)pro-␣3(V) (lanes 3-5) heterotrimers. Clonal lines of cells were grown in the absence (Ϫ) or presence (ϩ) of 100 mM Larginine (Arg) and furin inhibitor (FI) decanoyl-RVKR-chloromethyl ketone. Samples were subjected to SDS-PAGE on a 3.5% acrylamide stacking and 5% acrylamide running gel, which was stained with Coomassie Blue. both support the conclusion that BMP-1 cleaves pro-␣3(V) NH 2 -terminal sequences solely at the peptide bond between Gln-463 and Ala-464, to produce the largest N-propeptide yet described for a fibrillar procollagen. The minor sequence derived from the 52 kDa band corresponds to the pro-␣3(V) Cpropeptide and demonstrates cleavage to have occurred between Arg-1501 and Arg-1502, immediately downstream of the furin recognition site RRRR, consistent with the probability that the pro-␣3(V) C-propeptide is cleaved by a furin-like activity. The larger than expected size of the pro-␣3(V) C-propeptide may be related to the fact that the human pro-␣3(V) C-propeptide contains three potential sites for Asn-linked glycosylation, rather than the two found in other type V procollagen C-propeptides (36). Both the extra glycosylation site and apparent retarded mobility on SDS-PAGE are consistent with the possibility that the human pro-␣3(V) C-propeptide has relatively high levels of glycosylation.

Cells Are Dependent on BMP-1-and Furin-like Proteinases for Cleaving Pro-␣3(V) N-and C-propeptides, Respectively-
We next sought to determine directly the possible involvement of endogenous BMP-1-like and furin-like proteinases in the processing of pro-␣1(V)pro-␣2(V)pro-␣3(V) heterotrimers by cells. We have previously assayed for roles for BMP-1related proteinases in the in vivo processing of various substrates by comparing processing of those substrates in fibroblasts derived from wild type embryos to processing in fibroblasts derived from embryos doubly homozygous null for the Bmp1 gene, which encodes both BMP-1 and alternatively spliced mRNA for the related proteinase mammalian Tolloid, and for the Tll1 gene, which encodes the BMP-1-related proteinase mammalian Tolloid like-1 (mTLL-1) (11,13,14,16,49). Use of the doubly null cells removes possible functional redundancy, as we have shown previously that products of the two genes overlap in their substrate specificities (10,11,13,14). However, although we have demonstrated previously that such cells produce readily detectable levels of pro-␣1(V) chains (49), presumably in the context of pro-␣1(V) 2 5) heterotrimers. Samples were incubated in the absence (Ϫ) or presence (ϩ) of BMP-1 and were subjected to SDS-PAGE on a 3.5% acrylamide stacking and 5% acrylamide running gel (A) or on a 3.5% acrylamide stacking and 10% acrylamide running gel (B), which were stained with Coomassie Blue. Lane 3 in A contains size markers, whose approximate molecular masses are 250, 150, and 100 kDa. Bands in B were shown by NH 2 -terminal sequencing to represent the cleaved pro-␣3(V) N-propeptide (N-␣3(V), pro-␣3(V) C-propeptide (C-␣3(V)), pro-␣2(V) C-propeptide (C-␣2(V)), pro-␣1(V) C-propeptide (C-␣1(V)), or pro-␣1(V) N-propeptide (N-␣1(V)).
heterotrimers, attempts in the present study to detect endogenous pro-␣3(V)-derived chains in such cells by immunoblot were unsuccessful (data not shown). To overcome the inability to detect endogenous pro-␣3(V)-derived chains in MEFs, a pro-␣3(V) expression vector was electroporated into wild type MEFs and into MEFs derived from an embryo littermate doubly homozygous null for the Bmp1 and Tll1 genes, and mass cultures were selected that secreted stable pro-␣3(V) chains. Based on results from our attempts to express recombinant pro-␣3(V) chains in 293 cells (see above) it seems highly probable that the stable recombinant pro-␣3(V) chains secreted by MEF mass cultures are in the context of heterotrimers composed of recombinant pro-␣3(V) chains bound to endogenous pro-␣1(V) and pro-␣2(V) chains.
As can be seen (Fig. 6A) two pro-␣3(V)-derived bands are detectable in conditioned media of wild type MEFs. Although the gel for this blot was run longer than the SDS-polyacrylamide gels shown in Figs. 1-3, producing an electrophoretic pattern that differs from those of Figs. 1-3, mobilities of the two pro-␣3(V)-derived bands from wild type MEF media corresponded to the mobilities of stained recombinant pN␣3(V) and mature ␣3(V) chains run on the same gel (not shown). When the wild type MEFs were cultured in the presence of the furin inhibitor decanoyl-RVKR-chloromethyl ketone, a new band, the size of intact pro-␣3(V), was observed, concomitant with the disappearance of the putative mature ␣3(V) form. These results are consistent with a role for furin-like proprotein convertase activity in the processing of pro-␣3(V) in MEFs. The fact that all traces of the putative ␣3(V) band disappear but that a strong band is still observable with about the same mobility of the pN␣3(V) form is most consistent with the probability that the blockage of C-propeptide cleavage has resulted in replacement of ␣3(V) forms with pC␣3(V) forms that have a mobility similar to that of pN␣3(V).
In contrast to wild type MEFs, only a single pro-␣3(V)-derived chain is detectable in conditioned media of MEFs doubly null for the Bmp1 and Tll1 genes (Fig. 6A). This difference between wild type and Bmp1/Tll1-null MEFs indicates that endogenous BMP-1-like proteinases play a role in processing the pro-␣3(V) chain in MEFs. Culturing the Bmp1/Tll1-null MEFs in the presence of the furin inhibitor decanoyl-RVKRchloromethyl ketone results in disappearance of the single band and its replacement by intact pro-␣3(V) chains. This latter result is consistent with the probability that the single chain in untreated Bmp1/Tll1-null MEF media is the pN␣3(V) form and that blocking cleavage of the C-propeptide with furin inhibitor results in replacement of pN␣3(V) chains with intact pro-␣3(V) chains.
Resulting in part from low levels and limited distribution of expression, the recently cloned and sequenced pro-␣3(V) chain (36) is the least characterized type V/XI chain, and ␣1(V)␣2(V)␣3(V) is the least characterized form of type V collagen. Studies have shown ␣1(V) 2 ␣2(V) heterotrimers to be incorporated into type I/V heterotrimers and to retain partial NH 2 -terminal globular sequences that regulate the shape and diameter of the resultant heterotypic fibrils (20,21,28,29,45,47), whereas ␣1(V) 3 homotrimers may be localized to fibril surfaces and do not appear to regulate the geometries of heterotypic type I/V fibrils (64). In contrast, the nature of ␣1(V)␣2(V)␣3(V) macromolecular associations is relatively uncharacterized, and the nature of possible processing of a pro-␣1(V)pro-␣2(V)pro-␣3(V) precursor is totally unknown. The somewhat challenging prospect of producing a heterotrimer composed of three different recombinant proteins has contributed to delay in elucidating the biology of the ␣1(V)␣2(V)␣3(V) heterotrimer.
Here we have successfully produced recombinant pro-␣1(V)pro-␣2(V)pro-␣3(V) heterotrimers and have used biochemical, cell culture, and genetic means to characterize its proteolytic processing. Initial attempts at expressing the pro-␣3(V) chain in T-REx 293 cells transfected only with a pro-␣3(V) expression vector or with pro-␣3(V) and pro-␣1(V) expression vectors in the absence of a pro-␣2(V) expression vector were unsuccessful. Thus, pro-␣3(V) chains do not appear capable of forming stable homotrimers or heterotrimers composed solely of pro-␣1(V) and pro-␣3(V) chains.
We noted previously that the pro-␣3(V) C-propeptide might be processed by a furin-like activity, based on the relatedness of pro-␣3(V) and pro-␣1(V) chains and on the conservation of a consensus site for cleavage by furin-like proprotein convertases in the pro-␣3(V) telopeptide (36). Here, we provide evidence that pro-␣3(V) C-propeptides are cleaved by furin-like activities in cultures of T-REx-293 cells and MEFs. These results and recent results showing cleavage of the pro-␣1(XI) C-propeptide by a furin-like activity (11) support the conclusion that Cpropeptides of the entire subfamily of pro-␣1(V), pro-␣1(XI), pro-␣2(XI), and pro-␣3(V) chains are likely to be cleaved by this class of proteinases. In contrast, residues flanking the BMP-1 cleavage site in pro-␣1(V) NH 2 -terminal globular sequences, and conserved at the same positions in pro-␣1(XI) and pro-␣2(XI), are not conserved at similar positions in pro-␣3(V) (36). This has previously raised the issue of whether pro-␣3(V) NH 2terminal globular sequences are cleaved by BMP-1-like proteinases and, if so, at what site.
Here we demonstrate that BMP-1 cleavage releases a pro-␣3(V) N-propeptide constituting essentially the entirety of NH 2 -terminal globular sequences, including PARP and variable subdomains, plus the small COL2 collagenous domain. The site of BMP-1 cleavage was surprising for several reasons. First, residues flanking the site show essentially no homology with the residues flanking any previously characterized site of cleavage by BMP-1-like proteinases (65). The pro-␣3(V) BMP-1 cleavage site thus furthers the view that BMP-1-like proteinases, like other members of the astacin proteinase family (66), are not necessarily highly specific for residues immediately flanking the scissile bond and that other features must influence the recognition of such sites. A second reason for surprise regarding the pro-␣3(V) BMP-1 cleavage site is that it is relatively far removed from the conserved position, between PARP and variable subdomains, at which pro-␣1(V) and pro-␣1(XI) chains are cleaved by BMP-1 (11,48,49,52) despite similar NH 2 -terminal globular domain structures and sequences in these chains (Fig. 8). Despite a string of basic residues at the junction of the pro-␣3(V) PARP and variable subdomains (36), there was no evidence for cleavage by a furin-like activity at this site.
We demonstrate here that ␣1(V) chains in ␣1(V)␣2(V)␣3(V) heterotrimers have the same NH 2 termini as do ␣1(V) chains in ␣1(V) 2 ␣2(V) heterotrimers and that mature ␣2(V) chains are likely to retain all NH 2 -terminal globular sequences. Thus, ␣1(V) 2 ␣2(V) heterotrimers are likely each to have two ␣1(V) variable domains and one ␣2(V) intact NH 2 -terminal globular domain protruding from the surfaces of heterotypic type I/V collagen fibrils, whereas ␣1(V)␣2(V)␣3(V) heterotrimers, if incorporated into heterotypic fibrils, are likely to have one ␣1(V) variable domain and one ␣2(V) NH 2 -terminal domain protruding from fibril surfaces. The latter fibrils would thus differ from the former in having a lower density of ␣1(V) PARP domains protruding from their surfaces. This difference could clearly have profound effects on physical characteristics of heterotypic fibrils, such as shape and diameter, and could also affect their interactions with other extracellular molecules. It should be noted, however, that a conserved lysine residue located NH 2terminal to the COL1 domain in ␣1(V), ␣1(XI), and ␣2(XI) chains that forms homotypic covalent cross-links within type I/V or type II/XI heterotypic fibrils (67,68) is conserved at the same position in pro-␣3(V) (36) (see Fig. 5). This lysine is NH 2 -terminal to the pro-␣3(V) BMP-1 cleavage site. Thus, it is possible that, at least in some circumstances, cleaved pro-␣3(V) NH 2 -terminal globular sequences might be retained in heterotypic fibrils via cross-linking to other type V collagen chains.
Carey and colleagues (37) have reported a rat type V procollagen chain that they have designated pro-␣4(V). This chain has 95 and 82% identity to mouse and human pro-␣3(V) chains, respectively, previously cloned by authors of the current study (36).
Carey and colleagues (37) have suggested that although the human sequences reported by us are undoubtedly pro-␣3(V), the mouse sequences are likely the mouse ortholog of the rat pro-␣4(V) chain. However, we have reported previously that the genes that encode the human and murine procollagen chains map to homologous positions in their respective genomes, supporting the contention that they correspond to the same gene, rather then to genes for related but genetically distinct procollagen chains. More importantly, searches of the recently available human, mouse, and rat genome data bases by authors of the current report have found evidence for only one pro-␣3(V)-like sequence in each genome. Thus, it must be concluded that the reported pro-␣4(V) sequences actually correspond to the rat pro-␣3(V) chain. Interestingly, Carey and colleagues have noted that the rat chain, secreted by Schwann cells, colocalizes with type I collagen in fibrillar structures within sciatic nerve and that they retain some NH 2 -terminal noncollagenous sequences (37,53,69). Thus, ␣3(V) chains may be incorporated into heterotypic fibrils and retain some NH 2terminal noncollagenous sequences in some tissues, such as sciatic nerve. Nevertheless, MEF data presented in the current study indicate that essentially all pro-␣3(V) NH 2 -terminal noncollagenous sequences will be cleaved by cells that produce both pro-␣3(V) chains and BMP-1-like proteinases. Thus, it is possible that differential processing of pro-␣3(V) NH 2 -terminal sequences is a mechanism for tailoring the properties of ␣3(V)containing structures to the different needs of distinct tissues. In tissues where essentially all pro-␣3(V) NH 2 -terminal sequences are cleaved as a single polypeptide, it is interesting to speculate whether this polypeptide might persist and serve some functional role(s) because persistence of large amounts of cleaved NH 2 -terminal globular sequences in tissues has already been demonstrated for the pro-␣2(XI) chain (70).