Transforming Growth Factor-β Regulation of Bone Morphogenetic Protein-1/Procollagen C-proteinase and Related Proteins in Fibrogenic Cells and Keratinocytes*

Transforming growth factor-β1 (TGF-β1) induces increased extracellular matrix deposition. Bone morphogenetic protein-1 (BMP-1) also plays key roles in regulating vertebrate matrix deposition; it is the procollagen C-proteinase (PCP) that processes procollagen types I–III, and it may also mediate biosynthetic processing of lysyl oxidase and laminin 5. Here we show that BMP-1 is itself up-regulated by TGF-β1 and that secreted BMP-1, induced by TGF-β1, is either processed to an active form or remains as unprocessed proenzyme, in a cell type-dependent manner. In MG-63 osteosacrcoma cells, TGF-β1 elevated levels of BMP-1 mRNA ∼7-fold and elevated levels of mRNA for mammalian tolloid (mTld), an alternatively spliced product of the BMP1 gene, to a lesser extent. Induction of RNA was dose- and time-dependent and cycloheximide-inhibitable. Secreted BMP-1 and mTld, induced by TGF-β1 in MG-63 and other fibrogenic cell cultures, were predominantly in forms in which proregions had been removed to yield activated enzyme. TGF-β1 treatment also induced procollagen N-proteinase activity in fibrogenic cultures, while expression of the procollagen C-proteinase enhancer (PCPE), a glycoprotein that stimulates PCP activity, was unaffected. In contrast to fibrogenic cells, keratinocytes lacked detectable PCPE under any culture conditions and were induced by TGF-β1 to secrete BMP-1 and mTld predominantly as unprocessed proenzymes.

teogenic bone extracts with transforming growth factor-␤ (TGF-␤)-like proteins BMP-2 through -7 (1). Thus, it was suggested that BMP-1, by structure an astacin-like protease, may function in morphogenesis by activating TGF-␤-like molecules (1). Consistent with this possibility, BMP-1 has a domain structure similar to, but shorter than, that of tolloid, a Drosophila protein that appears to act in patterning of embryos by potentiating the activity of decapentaplegic, a TGF-␤ family member (2,3). The mammalian BMP1 gene is now known to produce alternatively spliced mRNAs for BMP-1 and for a longer protein, mammalian tolloid (mTld), which has a domain structure identical to that of Drosophila tolloid (4).
Fibrillar collagen types I-III are synthesized as procollagens, precursors containing N-and C-terminal propeptides that are cleaved extracellularly to yield mature triple helical monomers capable of associating into fibrils (for a review, see Ref. 5). Recently, BMP-1 was shown to be identical to procollagen C-proteinase (PCP) (6,7), the activity that cleaves the C-propeptides of procollagen types I-III (8 -10), and mTld has also been found to have PCP activity (7). 2 Demonstration of PCP activity, however, does not preclude the possibility that BMP-1 and/or mTld may also activate TGF-␤-like proteins. In fact, it is becoming increasingly apparent that products of the BMP1 gene play multiple roles in matrix deposition. These include proteolytic activation by fibrogenic cells of lysyl oxidase (11), an enzyme necessary to formation of covalent cross-links in fibrillar collagens and elastin, and biosynthetic processing by keratinocytes of laminin 5 (12), a major basement membrane component of skin. PCP activity of BMP-1 is stimulated ϳ10-fold by the procollagen C-proteinase enhancer (PCPE), a glycoprotein that binds the type I procollagen C-propeptide (10). However, possible involvement of PCPE in other biological activities of BMP-1 and mTld has not been examined.
TGF-␤1, prototype of the TGF-␤ superfamily, induces net increases in the deposition of insoluble matrix by cells. This is accomplished by effecting decreased production of proteases that degrade matrix and increased production of (i) inhibitors for such proteases, (ii) structural matrix components such as procollagen types I-III (13), and (iii) lysyl oxidase (14). Expression of the genes for the three polypeptide chains of laminin 5 is also up-regulated by TGF-␤ in keratinocytes (15). The induction by TGF-␤1 of gene products involved in deposition of matrix in general, and of known and potential substrates of BMP-1/mTld in particular, prompted us to examine whether TGF-␤1 also regulates BMP-1/mTld expression and/or the expression of PCPE. Here we document the effects of TGF-␤1 on BMP-1, mTld, and PCPE expression in fibrogenic cells and keratinocytes and on levels of cleavage of type I procollagen Cand N-propeptides. Mechanisms for the post-translational regulation of BMP-1 and mTld activity are also noted, and implications of the various data for the regulation of matrix deposition are discussed.

EXPERIMENTAL PROCEDURES
Cell Culture-MG-63 human osteosarcoma cells were purchased from the American Type Culture Collection (Rockville, MD), MC3T3-E1 murine osteoblastic cells were obtained from Dr. Richard Wenstrup (Children's Hospital Research Foundation, Cincinnati, OH), and human AH1F neonatal foreskin fibroblasts and primary keratinocytes were obtained from Dr. Lynn Allen-Hoffman (University of Wisconsin, Madison, WI). MG-63, MC3T3-E1, and AH1F cells were maintained in Dulbecco's modified Eagle's medium with 10% heat-inactivated ( Northern Blots-A 471-bp probe for RNA sequences shared by human BMP-1 and mTld has been described (4). The 1474-bp insert of full-length human cDNA clone KT11 (16) was used as a probe for PCPE RNA. Probe for human ␣1(I) collagen RNA was a 4.3-kb EcoRI fragment from plasmid pHUC (17), and a 2.0-kb human ␤-actin probe was purchased (CLONTECH). Poly(A ϩ ) RNA was prepared with the FastTrack kit (Invitrogen). 2 g of poly(A ϩ ) RNA/lane was electrophoresed on 1.2% agarose, 2.2 M formaldehyde gels and transferred to Hybond-Nϩ membranes (Amersham Corp.). Probes were radiolabeled to a specific activity of 4 -6 ϫ 10 9 cpm/g by random priming (18) and hybridized to blots in QuikHyb (Stratagene) at 65°C for 1 h. Blots were washed twice in 2 ϫ SSC, 0.1% SDS for 10 min at room temperature and then twice in 0.1 ϫ SSC, 0.1% SDS for 20 min at 65°C and exposed to Kodak X-Omat AR film with intensifying screens at Ϫ70°C. Autoradiograms were quantitated by scanning densitometry (Biomed Instruments) of autoradiograms exposed for varying lengths of time.
Antibodies-All antibodies were derived from polyclonal rabbit antisera. Antisera LF-40, LF-41, and LF-67 for the human pro-␣1(I) N-and C-propeptides and C-telopeptide, respectively, have been described (19) and were kindly provided by Dr. Larry Fisher (National Institutes of Health, Bethesda, MD).
Antibodies to the C termini of BMP-1 and mTld were raised against peptides CPHQLKFRVQKRNRTPQ and CLRYTSTKFQDTLHSRK, representing the final 16 residues of each respective protein plus an additional cysteine for coupling to keyhole limpet hemocyanin. Rabbits injected with the mTld peptide were boosted with the same peptide conjugated to ovalbumin to increase titers. BMP-1/mTld proregion antibodies, kindly provided by Dr. Mitch Brenner (FibroGen Inc.), were raised against peptide DLAEEDDSEPLNYKDPC, corresponding to residues 31-47 of the BMP-1 sequence (1), linked to keyhole limpet hemocyanin. Peptide antibodies were affinity-purified on columns of the appropriate peptide coupled to TC gel (Quality Controlled Biochemicals) via the cysteine thiol. They were eluted with 3 M MgCl 2 , 25% ethylene glycol and dialyzed against 10 mM sodium phosphate (pH 7.4), 20 mM NaCl.
Western Blots-Protease inhibitors (2.5 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 1 mM p-aminobenzoic acid, 1 mM N-ethylmaleimide) were added to harvested media, and proteins were precipitated by adding trichloroacetic acid to 10%. Pellets were washed with ice-cold acetone and then washed twice with 75% ethanol, 12.5 mM Tris (pH 7.5), dried, and resuspended in SDS sample buffer with 5% ␤-mercaptoethanol. Cell layers were scraped into hot SDS sample buffer, as described (23). Media and cell layer samples, equivalent to 5 ϫ 10 5 fibrogenic cells or keratinocytes per lane, were subjected to SDS-PAGE and transferred to Immobilon-P membranes (Millipore) by electroblotting in 25 mM Tris, 192 mM glycine, 10% methanol at 4°C. Blots were incubated ϳ14 h with primary antibody diluted 1:5000 in PBS, 1% bovine serum albumin, 0.05% Tween 20. After washing three times with wash buffer (PBS, 0.05% Tween-20), blots were incubated 1 h with horseradish peroxidase-conjugated donkey anti-rabbit IgG (Amersham) diluted 1:4000. Blots were then washed four times with wash buffer, incubated with SuperSignal CL-HRP substrate (Pierce), and exposed to film. Apparent molecular weights of bands were estimated by comparison to electrophoretic mobilities of prestained standards (Bio-Rad).
To quantitate relative amounts of BMP-1, gel lanes were loaded with sample corresponding to the medium of 2.5 ϫ 10 5 untreated MG-63 cells or to the medium of 1.25 ϫ 10 4 , 5.0 ϫ 10 4 , or 2.5 ϫ 10 5 TGF-␤1-treated MG-63 cells. After SDS-PAGE, a blot was prepared, treated with antibody to the BMP-1 C terminus (as above), and then incubated with 125 I-labeled protein A (Amersham) for 1 h in PBS, 1% bovine serum albumin, 0.05% Tween-20, 1 mM dithiothreitol. The blot was washed 8 h with wash buffer and exposed to film, and the autoradiograph used as a template for excising radioactive BMP-1 bands, which were counted in an A 800CD ␥ counter (Packard).
In Vitro Procollagen Cleavage Assays-14 C-labeled procollagen substrate prepared from chick embryo tendon cultures (8) was kindly provided by Dr. Darwin Prockop (Allegheny University, Philadelphia, PA). Media tested for PCP activity contained 1 mM each phenylmethylsulfonyl fluoride, p-aminobenzoic acid, and N-ethylmaleimide; was concentrated 50-fold in Centriprep concentrators (Amicon); and then was dialyzed against 50 mM Tris-HCl, 0.15 NaCl, 5 mM CaCl 2 . 5 l of concentrated medium was added to 0.125 g of substrate in 5 l of 0.1 M Tris-HCl, 0.1 M NaCl, 5 mM CaCl 2 , 0.02% Brij 35 (pH 7.6) and incubated for 4 h at 35°C. Reactions were stopped by adding 5 l of 3 ϫ SDS sample buffer and heating for 4 min at 100°C prior to SDS-PAGE on nonreduced 6% gels. Gels were fixed in 20% MeOH, placed in Amplify (Amersham) for 30 min, dried, and exposed to film. To quantitate PCP activity, percentage of cleavage was calculated for each lane of an autofluorogram by comparing signal remaining in the pro-␥ band (procollagen in which cleavage has not occurred in any of the three PCP cleavage sites) to total signal (pro-␥ ϩ pro-␤ ϩ pro-␣ ϩ pN␣1 ϩ ␣1 ϩ pN␣2 ϩ ␣2 ϩ free C-propeptide bands) (see Fig. 5). The percentages of cleavages achieved with media from cultures treated with TGF-␤1, ascorbate, or TGF-␤1 plus ascorbate were then compared with the percentages of cleavage achieved with media from untreated controls to calculate -fold increases in PCP activity. Signals on autofluorograms were quantitated with an IS1000 digital video imaging system (Alpha Innotech).
RNase Protection Assays-To generate BMP-1/mTld-specific riboprobe, a 1619-bp ApaI-EagI fragment was excised from a baculovirus transfer vector containing full-length human BMP-1 cDNA (6). The fragment, extending from nucleotide 615 of the BMP-1 cDNA sequence (1) to an EagI site in the vector polylinker, was inserted between pBluescript II KSϩ (Stratagene) ApaI and EagI sites and linearized at a HincII site at nucleotide 1655 of the BMP-1 sequence (1). This template produced a 617-base riboprobe of which 575 bases are BMP-1specific and 481 bases are specific for alternatively spliced mTld RNA. For PCPE-specific riboprobe, a 329-bp fragment of PCPE cDNA clone KT3 (16) was excised with EcoRI and SacII and inserted between pBluescript II KSϩ EcoRI and SacII sites. This template, linearized at the pBluescript XhoI site, produced a 391-base riboprobe, 329 bases of which correspond to nucleotides 557-885 of the human PCPE cDNA sequence (16). Control riboprobe was generated with template pTRI-␤actin-Human (Ambion). Uniformly 32 P-labeled riboprobes were gener-ated by transcription with RNA polymerase T7, hybridized to total RNA isolated with TRIzol (Life Technologies, Inc.), and analyzed by digestion with RNases A and T1 and electrophoresis on 6% denaturing gels, as described (24).

TGF-␤1 Selectively Increases Levels of BMP-1 and mTld, but
Not PCPE, mRNA in MG-63 Osteosarcoma Cells-To determine whether levels of PCPE, BMP-1, and/or mTld mRNA are influenced by treatment with TGF-␤, just-confluent MG-63 cultures were treated for 24 h with 2 ng/ml TGF-␤1. Levels of the ϳ3.0-kb BMP-1 mRNA and ϳ5.5and ϳ4.1-kb mTld mRNAs (4) were all increased by TGF-␤1 treatment (Fig. 1). Interestingly, mRNAs for the two proteins did not increase to the same extent, with BMP-1 mRNA increasing ϳ7-fold and both mTld mRNA forms increasing only ϳ4-fold. By comparison, levels of the 4.8-and 5.8-kb pro-␣1(I) collagen mRNAs (25) increased ϳ9-fold, while those of the ␤-actin mRNA control did not change. In contrast to the induction of BMP-1 and mTld mRNA, levels of PCPE mRNA were not increased by TGF-␤1.
Incubation of MG-63 cells with TGF-␤1 concentrations ranging from 0.01 to 10 ng/ml, showed induction of BMP-1 and mTld mRNAs to be dose-dependent ( Fig. 2A), with substantial increases first noted at 0.5 ng/ml and maximum induction at 2 ng/ml. Induction of mTld and BMP-1 mRNAs was somewhat less at 10 ng/ml than at 2 ng/ml, possibly due to the toxicity of TGF-␤1 at higher concentrations (26). When kinetics of induction of BMP-1/mTld mRNAs were examined by incubating MG-63 cultures in 2 ng/ml TGF-␤1 for varying times (Fig. 2B), substantial increases first occurred at 12 h, with maximal levels attained at 24 h post-treatment. Thus, induction was delayed compared with induction of RNAs encoding a variety of extracellular matrix proteins, which generally occurs within 3-5 h of TGF-␤ treatment of various cell types (13). This delay suggested that induction of BMP-1/mTld mRNAs may occur secondarily to an earlier event triggered by TGF-␤1, a possibility supported by experiments in which treatment of confluent cultures with cycloheximide prior to TGF-␤1-treatment showed protein synthesis to be required prior to induction of BMP-1/mTld mRNAs (Fig. 2C).
Levels of Secreted BMP-1 and mTld, but Not PCPE, are Up-regulated by TGF-␤1 in Fibrogenic Cell Cultures-To determine whether induction of steady state levels of BMP-1 and mTld mRNA in MG-63 cells was paralleled by increases in the cognate proteins, levels of BMP-1 and mTld protein secreted into culture media were examined by Western blots. Antibodies specific for the BMP-1 C terminus detected bands of ϳ88 and ϳ77 kDa, both of which were strongly up-regulated by TGF-␤1 ( Fig. 3A). Antibody to the mTld C terminus detected a ϳ130-kDa band that was up-regulated by TGF-␤1 to a lesser extent than the anti-BMP-1-reactive bands. The relative extent of induction of BMP-1-and mTld-specific bands, thus appeared similar to the relative extent of induction of the cognate mRNAs.
To quantitate the induction of secreted BMP-1, Western blots were incubated with antibody to the BMP-1 C terminus and then with 125 I-protein A, to allow subsequent excision and quantitating of counts in BMP-1 bands. This method, which can provide good quantitation of relative amounts of proteins (27), showed levels of the 88-kDa BMP-1 band to be ϳ8-fold higher in TGF-␤1-treated MG63 cultures than in untreated cultures (not shown). Thus, the level of induction of secreted BMP-1 was similar to the level of induction of cognate mRNA. The 77-kDa band was not detected on these blots (Fig. 3D, lane  3) and was determined to be unrelated to BMP-1 (see below).
Ascorbate may stimulate production of type I procollagen at transcriptional and posttranscriptional levels (28 -30). Since we were interested in mechanisms that might contribute to co-expression of BMP-1 and type I procollagen, we examined whether levels of BMP-1 were also up-regulated by ascorbate. Although the intensity of BMP-1 and mTld protein bands did not appreciably increase in the presence of ascorbate alone, intensities seemed somewhat higher in the presence of TGF-␤1 plus ascorbate than in the presence of TGF-␤1 alone (Fig. 3A). In contrast to BMP-1 and mTld, levels of secreted PCPE did not increase in MG-63 cultures treated with TGF-␤1, ascorbate, or ascorbate plus TGF-␤1 (Fig. 3A). Thus, as with BMP-1 and mTld, levels of secreted PCPE paralleled levels of cognate mRNA.
To ascertain whether results like those obtained with MG-63 cells were common to other fibrogenic cells, analyses were done on MC3T3-E1 osteoblastic mouse cells and human dermal fibroblasts (Fig. 3, B and C). Results obtained with these cells were similar to those observed with MG-63 cells in that levels of secreted BMP-1 and mTld were elevated in the presence of TGF-␤1 and appeared to be slightly more elevated in the presence of TGF-␤1 plus ascorbate. As with MG-63 cells, levels of PCPE were not dramatically changed in a consistent way by the addition of TGF-␤1 and/or ascorbate.
Some Western blots using BMP-1 C terminus antibody detected a ϳ77-kDa band, in addition to the 88-kDa BMP-1 band, in MG-63 media (Fig. 3, A and D, lane 2). This band, without counterpart in MC3T3-E1 or fibroblast cultures, was not always detected in MG-63 cultures treated with TGF-␤1 or TGF-␤1 plus ascorbate and was absent upon culturing in the presence of 0.1% serum (Fig. 3D, lane 1) or when protein A was used instead of secondary antibody (Fig. 3D, lane 3). The 77-kDa MG-63 band was detected on blots when preimmune sera or mTld C terminus antibody was used as primary antibody and by secondary antibody in the absence of primary antibody (not shown). We therefore conclude that the 77-kDa MG-63 band is unrelated to BMP-1.
Cleavage of Endogenous Type I Procollagen C-and N-propeptides Is Induced in MG-63 Cultures by TGF-␤1 plus Ascorbate-We next examined whether induction of secreted BMP-1 and mTld in MG-63 cultures was paralleled by induction of PCP activity, as evidenced by increased cleavage of endogenous type I procollagen C-propeptides. Immunoblots using antibody to the pro-␣1(I) N-propeptide (Fig. 4A, part a) did not detect pN␣1(I) chains in media of untreated cells or, surprisingly, in media of cells treated with TGF-␤1 alone, although the latter contained relatively high levels of procollagen (Fig. 4A), BMP-1, and mTld (Fig. 3A). In contrast, pN␣1(I) chains were found in media of ascorbate-treated cells, in a pro-␣1(I):pN␣1(I) ratio of ϳ1.0:0.3, and at higher levels in media of cells treated with TGF-␤1 plus ascorbate, with a pro-␣1(I):pN␣1(I) ratio of ϳ1.0:0.8 (Fig. 4A, part a).
Pro-␣1(I) C-telopeptide antibody, capable of detecting ␣1(I), pN␣1(I), and pC␣1(I) chains (Fig. 4A, part b), revealed relative levels of pro-␣1(I) and pN␣1(I) chains in the various media, similar to those detected by the first antibody. In addition, pC␣1(I) and mature ␣1(I) chains were detected in media of cells treated with TGF-␤1 plus ascorbate. In the latter sample, the ratio of unprocessed pro-␣1(I) chains to chains from which the C-propeptide had been cleaved (pN␣1(I) ϩ ␣1(I)) was ϳ1.0:2.7, a value ϳ10-fold greater than the pro-␣1(I):pN␣1(I) ratio (ϳ1.0: 0.3) detected in the ascorbate-treated sample by the same antibody. Thus, although cultures treated with TGF-␤1 plus ascorbate contained substantially more substrate than cultures treated with ascorbate alone, a greater fraction had undergone removal of C-propeptides, indicating significantly increased PCP activity. A diffuse band, about the size of pC␣1(I), faintly detected by C-telopeptide antibody in samples treated with TGF-␤1 alone (Fig. 4A, part b) was not detected by C-propeptide-specific antibody (see below) and probably represents partially degraded pro-␣1(I) chains.
The appearance of quantities of pC␣1(I) and ␣1(I) chains only in media of MG-63 cells treated with TGF-␤1 plus ascorbate, indicated that procollagen N-proteinase (PNP) activity is also elevated under these conditions. Highlighting this induction, pro-␣1(I) C-propeptide antibody detected pC␣1(I) chains only in media of cells treated with TGF-␤1 plus ascorbate (Fig.  4A, part c). Since the N-propeptides of procollagens I and III  2 and 3). Antibodies to the C termini of mTld and BMP-1 or to full-length PCPE were primary antibodies for the various blots. For the blot in D, lane 3, 125 I-labeled protein A was used instead of secondary antibody. Thin and thick arrows denote 101-kDa latent and 88-kDa processed forms of BMP-1, respectively. Arrowheads denote the 77-kDa artifactual band. SDS-PAGE gels for mTld, BMP-1, and PCPE blots were 7.5, 10, and 12% polyacrylamide, respectively.

FIG. 4. Cleavage of endogenous type I procollagen C-and Npropeptides in MG-63 cultures. MG-63 media (A) or cell layers (B)
were examined by Western blot analysis using primary antibodies directed against the pro-␣1(I) N-propeptide (parts a), ␣1(I) C-telopeptide (parts b) or pro-␣1(I) C-propeptide (parts c). Lanes C, T, A, and TA are from untreated control cultures or from cultures treated with TGF-␤1, ascorbate, or ascorbate plus TGF-␤1, respectively. SDS-PAGE gels were 7.5% polyacrylamide. are cleaved by two different PNP enzymes (31,32), and since MG-63 cells produce procollagen III (33), Western blots using antibody against the type III procollagen C-propeptide were performed and showed that procollagen III PNP activity is also elevated in the presence of TGF-␤1 plus ascorbate (not shown).
Since procollagen processing may help regulate incorporation of monomers into fibrils (34), levels of processed ␣1(I) chains incorporated into MG-63 cell layers were also examined. Consistent with results obtained from media, C-telopeptide antibody found high levels of ␣1(I) chains in cell layers treated with TGF-␤1 plus ascorbate, while untreated cell layers or cell layers treated with TGF-␤1 alone did not contain detectable ␣1(I) (Fig. 4B, part b). Cell layers treated with ascorbate or with ascorbate plus TGF-␤ were enriched for ␣1(I) chains compared with media from the same cultures (Fig. 4, A and B, parts  b). This enrichment may reflect preferential incorporation of mature monomers into growing fibrils, rather than increased processing in cell layers compared with media. Consistent with this likelihood, pN␣1(I), but not pC␣1(I), chains were found in cell layers treated with ascorbate or with TGF-␤1 plus ascorbate (Fig. 4B, parts a, b, and c), appearing to confirm results previously obtained with in vitro fibrillogenesis systems (34), suggesting that pN␣1(I) but not pC␣1(I) chains are incorporated into growing fibrils. Unprocessed pro-␣1(I) in MG-63 cell layers (Fig. 4B) was likely intracellular, since procollagen is not thought to be incorporated into fibrils (34). As in media, a diffuse band about the size of pC␣1(I), detected in some cell layers by C-telopeptide and N-propeptide antibodies but not by C-propeptide antibody, probably represents partially degraded pro-␣1(I) chains.

PCP and PNP Activity against Exogenous Procollagen Substrate Is Increased in Media of TGF-␤1-treated MG-63
Cultures-It seemed paradoxical that although secreted BMP-1 and mTld were elevated to only slightly higher levels in MG-63 cultures treated with TGF-␤1 plus ascorbate than in cultures treated with TGF-␤1 alone (Fig. 3A), processing of endogenous procollagen was detectable only in the former (Fig. 4). Ascorbate is a necessary cofactor for enzymic hydroxylation of collagen prolyl residues, leading to a more stable triple helix (35). Thus, to examine whether differences in endogenous substrate might contribute to differences observed in levels of processing, MG-63 media were assayed for PCP and PNP activity using exogenous radiolabeled procollagen substrate. As with endogenous substrate, PCP and PNP activities against exogenous substrate were highest in media of cells treated with TGF-␤1 plus ascorbate, as evidenced by the generation of mature ␣-chains, processing intermediates, and free C-propeptides (Fig. 5). However, in contrast to results with endogenous substrate, PCP and PNP activities against exogenous substrate were increased in media of cells treated with TGF-␤1 alone, to levels approaching those of cultures treated with TGF-␤1 plus ascorbate. Thus, in seven independent experiments, one of which is shown in Fig. 5, the mean increase of PCP activity against exogenous substrate was 2.4 Ϯ 0.2-fold for cultures treated with TGF-␤1 plus ascorbate and 1.7 Ϯ 0.1-fold for cultures treated with TGF-␤1 alone (values are mean Ϯ S.E.). PCP activity in media of cultures treated with ascorbate alone was virtually unchanged, 1.0 Ϯ 0.1-fold, compared with untreated controls. These results suggest that observed differences in processing of endogenous substrate in cultures treated with TGF-␤1 alone, compared with cultures treated with TGF-␤1 plus ascorbate, were at least partly due to differences in the hydroxylation state and conformation of the substrate.
Most BMP-1 and mTld Secreted by Fibrogenic Cells in Response to TGF-␤1 Is Processed to Mature, Active Forms-Proteases of the astacin family are synthesized as proenzymes with N-terminal proregions that must be removed for activation (36,37). Thus, the lesser induction of PCP activity against exogenous substrate (ϳ1.7-fold for TGF-␤1-treated cultures), relative to levels of induction of secreted BMP-1 detected by quantitative Western blots (ϳ8-fold for TGF-␤1-treated cultures), might have been due to secretion of most BMP-1 and mTld as inactive precursors. To ascertain whether this was the case, antibody to proregion sequences common to BMP-1 and mTld precursors was prepared and used to examine immunoblots of MG-63 medium samples. This antibody did not recognize the major 88-kDa BMP-1 band but did recognize a ϳ101-kDa band that also appears as a minor band on immunoblots using the BMP-1 C terminus antibody (Figs. 6A and 3A). Thus, although a small proportion is unprocessed, the majority of TGF-␤1-induced BMP-1 secreted by the fibrogenic cells is in the processed active form. In addition, the proregion antibody did not recognize the ϳ130-kDa band detected by mTld C terminus antibody but did recognize a ϳ143-kDa band (Fig.  6A). Thus, the majority of mTld secreted by TGF-␤-treated fibrogenic cells is also processed to the mature form, while a small proportion, at levels undetectable with the mTld C terminus antibody, is unprocessed. Interestingly, in repeated experiments of the type shown in Fig. 6A, media of MG-63 cells treated with TGF-␤1 alone consistently had lesser amounts of activated, and greater amounts of unprocessed, BMP-1 and mTld than did cultures treated with TGF-␤1 plus ascorbate. Thus, additional processing of BMP-1 and mTld into mature forms seems to occur in the presence of ascorbate, perhaps accounting for the higher levels of PCP activity against exogenous substrate in media of cultures treated with TGF-␤1 plus ascorbate than in media of cultures treated with TGF-␤1 alone.
Previously, recombinant human BMP-1 produced in a baculovirus system by this laboratory was shown to have high PCP activity and to be processed to the mature form (6). Thus, BMP-1 secreted by TGF-␤1-treated fibrogenic cells should have the same electrophoretic mobility as the recombinant BMP-1, if in an active processed form. Electrophoretic mobilities of BMP-1 from TGF-␤1-treated fibroblast media and from the baculovirus system were compared after treatment of samples with peptide-N-glycosidase F to control for mobility differences due to differences in Asn-linked glycosylation between insect An autofluorogram is shown of 14 C-labeled procollagen substrate incubated with concentrated MG-63 medium samples. Lanes C, T, A, and TA are from untreated control cultures or from cultures treated with TGF-␤1, ascorbate, or ascorbate plus TGF-␤1, respectively. Lane S is undigested substrate at T ϭ 0; lane R is substrate incubated with ϳ3 ng of recombinant human BMP-1 prepared in a baculovirus system and purified chromatographically, as described (6). Pro␥, intact type I procollagen heterotrimer; pro␤, two pro-␣ chains disulfide-bonded to a free C-propeptide chain; pro␣, one pro-␣ chain disulfide-bonded to two free C-propeptide chains; C3, three disulfide-bonded free C-propeptide chains. and mammalian cells (6,38). Upon removal of Asn-linked carbohydrates, the major form of BMP-1 secreted by TGF-␤1treated fibroblasts had the same mobility as recombinant active BMP-1 (Fig. 6B), bolstering the conclusion that most BMP-1 secreted by fibrogenic cells in response to TGF-␤1 is processed to the mature form.
The Major Secreted Product of the BMP1 Gene in TGF-␤1treated Human Keratinocytes Is Unprocessed mTld-It was of interest to determine whether the patterns of expression and regulation of BMP-1 and mTld described above are also found in non-fibrogenic cells. Keratinocytes were chosen for study, since they do not produce readily detectable fibrillar collagens but do appear to utilize BMP-1 and/or mTld for processing laminin 5 (12). Consistent with previous reports that keratinocytes do not produce type I collagen (39), type I collagen was not detected in media of keratinocytes treated or untreated with TGF-␤1 and/or ascorbate by Western blots using pro-␣1(I) Npropeptide, C-propeptide, or C-telopeptide antibodies (not shown). In addition, and in contrast to results obtained with fibrogenic cells, PCPE expression by keratinocytes was undetectable by Western blot, regardless of culture conditions (Fig.  7A). Also in contrast to results with fibrogenic cells, antibodies against BMP-1 or mTld C termini did not clearly detect specific bands on Western blots of medium samples from keratinocytes treated or untreated with TGF-␤1 and/or ascorbate (not shown). This indicated that if keratinocytes secrete BMP-1 and/or mTld, then steady state levels in media are significantly lower than in fibrogenic cell media.
We had noted in previous experiments that the proregion antibody was of higher affinity than antibodies against the C termini of BMP-1 and mTld. The proregion antibody was able to detect ϳ143and ϳ101-kDa bands, corresponding in size to unprocessed mTld and BMP-1 bands observed in MG-63 cultures, in media of keratinocytes treated with TGF-␤1 or with TGF-␤1 plus ascorbate (Figs. 6A and 7B). The ϳ143-kDa band was barely detectable, and the ϳ101-kDa band was not detected, in media of untreated or ascorbate-treated keratinocytes. Consistent with the observation that keratinocytes secrete predominantly unprocessed mTld and BMP-1 in response to TGF-␤1, PCP activity against exogenous procollagen sub-strate was at very low levels in media of keratinocytes treated or untreated with TGF-␤1 and/or ascorbate (not shown).
An RNase protection assay showed keratinocytes and MG-63 cells to contain comparable levels of mTld RNA but MG-63 cells to contain higher levels of BMP-1 RNA (Fig. 7C). Thus, secretion of predominantly mTld by keratinocytes and predominantly BMP-1 by MG-63 cells is reflected, at least in part, in levels of cognate RNA transcripts. RNase protection also demonstrated that absence of PCPE secretion by keratinocytes is reflected in a virtual absence of PCPE RNA (Fig. 7C). DISCUSSION The necessary action of BMP-1 and mTld in processing of matrix components (6,7,12) and lysyl oxidase (11) implies that these proteins play key roles in controlling the deposition of matrix in developmental and homeostatic processes. Previously, however, mechanisms for regulating functional expression of these key proteins have not been explored. In this study we have demonstrated that TGF-␤1 elevates levels of BMP-1 and mTld in fibrogenic cells and keratinocytes. PNP activity against procollagens I and III was also elevated by TGF-␤1 in fibrogenic cells. Thus, in addition to increasing the deposition of matrix through induction of matrix components, lysyl oxidase, and metalloprotease inhibitors, TGF-␤1 also influences net matrix deposition by inducing the proteases that process procollagens, laminin 5, and lysyl oxidase into mature forms.
The induction of BMP-1 and mTld by TGF-␤1, described here, is particularly intriguing in the context of previous suggestions that BMP-1-and tolloid-like proteins may activate TGF-␤-like molecules (1)(2)(3). TGF-␤1 is itself secreted as a latent form (40), and the possibility, therefore, exists of a positive feedback loop in which BMP-1 and/or mTld activate, and are in return induced by, TGF-␤1. There is a precedent for such positive feedback loops in matrix deposition, since the effects of TGF-␤1 on matrix deposition are amplified and prolonged through autoinduction by TGF-␤1 of its own expression (41). Preliminary studies with recombinant proteins suggest that BMP-1 may indeed be capable of directly activating the TGF-␤1 small latent complex. 3 Removal of the proregion from astacin-like proenzymes is thought necessary for the production of active forms of these enzymes (37). Thus, persistence or removal of the proregion represents another potential control point for regulating BMP-1 and mTld activities. In the present study, the degree of processing of secreted BMP-1 and mTld is shown to be cell type-specific, with predominantly processed forms produced by TGF-␤1-treated fibrogenic cells and predominantly unprocessed forms produced by TGF-␤1-treated keratinocytes. Clearly, the production of large amounts of activated BMP-1 and mTld would aid fibrogenic cells in their highly specialized roles of producing large quantities of fibrillar collagen matrix, especially in response to TGF-␤. It is less clear why keratinocytes produce predominantly unprocessed, inactive forms of mTld and BMP-1 in response to TGF-␤. Consistent with this finding, however, is the observation that processing of ␥2, the laminin 5 chain cleaved by mTld and/or BMP-1, is delayed following secretion by cultured keratinocytes (12). Thus, extracellular processing of mTld and BMP-1 may be a rate-limiting step in the deposition of keratinocyte extracellular matrix, of which laminin 5 is a major component (42,43). In vivo, such processing may be regulated by epithelial-mesenchymal interactions that influence the production of matrix by basal keratinocytes (44). Keratinocytes were found not to produce detectable amounts of PCPE, suggesting that PCPE may not play a role in laminin 5 processing. However, the possibility that, in vivo, PCPE may be provided by dermal fibroblasts for this purpose has not been precluded.
Interestingly, the relative amounts of BMP-1 and mTld produced by cells were also found to be cell type-specific: fibrogenic cells produced relatively large amounts of BMP-1, while keratinocytes produced predominantly mTld, at both RNA and protein levels. The possible functional significance for the production of differing ratios of BMP-1 and mTld by different cell types remains to be determined, however, since a functional difference has yet to be discerned for these two protein products of the same gene.
Previously, the low levels of PCP activity detectable in tissues and in cell culture systems have led to suggestions that the enzyme is either secreted as an inactive precursor or coexpressed with an endogenous inhibitor (8). Either of these possibilities might have explained the discrepancy observed in the present study between the induction by TGF-␤1 in MG-63 cultures of an ϳ8-fold increase in secreted BMP-1 and a ϳ2fold increase in medium PCP activity against exogenous substrate. However, since secreted BMP-1 and mTld were both found predominantly as processed forms, this leaves the interesting possibility of an endogenously produced inhibitor. Pre-viously, localized control over the activities of various proteases, including degradative matrix metalloproteases such as stromelysins and collagenases, has been shown to involve not only the processing of proenzymes to mature forms but also the balance between levels of activated enzyme and levels of specific inhibitors co-expressed by the same cell types (45). Studies to determine the possible existence of inhibitors for BMP-1, mTld, and related proteases (46) seem warranted by results presented in the current study. Clearly, such inhibitors, should they exist, could play roles in morphogenetic processes as important as those of the proteases with which they interact.
A final point of interest relates to the observed processing of endogenous procollagen in MG-63 cultures treated with TGF-␤1 plus ascorbate and the absence of such processing in cultures treated with TGF-␤1 alone. Both cultures secreted high levels of similarly processed BMP-1 and mTld. Moreover, the TGF-␤1 plus ascorbate-treated cultures had only slightly higher PCP activity against exogenous substrate, corresponding to a slightly higher ratio of activated to unprocessed BMP-1. One interpretation of these data is that fully hydroxylated procollagen produced by ascorbate-treated cultures, or supplied as exogenous substrate, is much better substrate for BMP-1 than is underhydroxylated procollagen produced in the absence of ascorbate. Since underhydroxylated procollagen is not likely to be in a compact triple helical form at 37°C (35), these data might indicate some conformational requirement for the cleavage of procollagen by PCP. However, this would be in contrast to an earlier report (8) that found PCP to cleave heat-denatured procollagen with about the same efficiency as native procollagen.