Basement Membrane Zone Type XV Collagen Is a Disulfide-bonded Chondroitin Sulfate Proteoglycan in Human Tissues and Cultured Cells*

Type XV collagen has a widespread distribution in human tissues, but a nearly restricted localization in basement membrane zones. The α1(XV) chain contains a highly interrupted collagenous region of 577 residues, and noncollagenous amino- and carboxyl-terminal domains of 530 and 256 residues, respectively. Cysteines are present in each domain and consensus sequences forO-linked glycosaminoglycans are situated in the amino terminus and in two large, noncollagenous interruptions. We now report that type XV collagen is a chondroitin sulfate proteoglycan in human tissues and cultured cells, and that the α chains are covalently linked by interchain disulfide bonds only between the two cysteines in the collagenous region. Western blotting of tissue extracts revealed a diffuse smear with a mean size ≥400 kDa, which after chondroitinase digestion resolved into a 250-kDa band in umbilical cord, and 250- and 225-kDa bands in placenta, lung, colon, and skeletal muscle. The latter two bands were also directly visualized by alcian blue/silver staining of a purified placenta extract. In a human rhabdomyosarcoma cell line, almost all of the newly synthesized type XV collagen was secreted into the medium and upon chondroitinase digestion just the 250-kDa α chain was generated. Chondroitinase plus collagenase digestion of tissue and medium proteins followed by Western blotting using domain-specific antibodies revealed a 135-kDa amino-terminal fragment containing glycosaminoglycan chains and a 27-kDa fragment representing the intact carboxyl terminus. However, a truncated carboxyl peptide of ∼8-kDa was also evident in tissue extracts containing the 225-kDa form. Our data suggest that the 225-kDa form arises from differential carboxyl cleavage of the 250-kDa form, and could explain the ∼19-kDa endostatin-related fragments (John, H., Preissner, K. T., Forssmann, W.-G., and Ständker, L. (1999)Biochemistry 38, 10217–10224), which may be liberated from the α1(XV) chain.

Basement membrane zones (BMZs) 1 can be operationally defined as morphological entities consisting of a basement membrane plus its closely associated matrix components, which extend into or originate from the sub-basal lamina (1). The BMZ contains the molecules responsible for attaching basement membranes to their contiguous stroma and/or epithelium. Those components integral to basement membranes include type IV collagen, laminin, entactin/nidogen, and perlecan (for review, see Ref. 2), whereas a plethora of other matrix proteins, glycoproteins, and proteoglycans have been assigned to the BMZ using different methods. Many of these constituents have been studied in depth biochemically and ultrastructurally, while several newer and less abundant ones are not yet well characterized. Within this latter category are three more recently discovered nonfibrillar collagens: types XV, XVIII, and XIX (3)(4)(5)(6). Independently identified from DNA clone isolation, they are considered members of a unique collagen subclass because of their widespread distribution in BMZs of many tissues (7). Immunohistochemical light microscopy demonstrated that these three collagens co-localize in some BMZs, but are differentially expressed in others (7)(8)(9)(10)(11)(12).
Type XV and XVIII, but not type XIX, collagens were also shown to exhibit major similarities by primary structure alignment. Comparison of domain arrangement, restricted sequence homology, as well as intron/exon organization indicated that ␣1(XV) and ␣1(XVIII) evolved from a common ancestral gene (4,(13)(14)(15)(16). Both collagens, but especially type XV, contain extensive interruptions in their collagenous regions such that the majority of the residues in each chain are found within the amino-and carboxyl-terminal noncollagenous domains. In particular, the carboxyl-terminal domain of type XVIII collagen has become a focal point in tumor biology upon finding that the terminal 20-kDa fragment is identical to the potent anti-angiogenic factor, endostatin (17). Pursuant to this discovery is more current research in which the analogous peptide of type XV collagen, displaying by far the highest degree of sequence conservation with type XVIII, is being investigated for related properties (18,19).
Another parallel between types XV and XVIII collagen can be drawn by comparing their noncollagenous amino-terminal domains. Among the common features are a number of consensus sequences for attachment of O-linked glycosaminoglycans (4,5,13,14). (Several additional sequences are present within interruptions in the collagenous region (see Refs. 3, 4, and 6).). Until recently, it was not known whether any of these sites were occupied. It has since been described that ␣1(XVIII) chains are sensitive to heparitinase, but not to chondroitinase ABC digestion (12). Thus, type XVIII collagen is a heparan sulfate proteoglycan with a core protein with molecular mass of 180 kDa.
The first immunochemical studies of type XV collagen to determine its tissue distribution were reported using antibodies directed at the carboxyl terminus of the protein (9). In human placenta and colon tissue extracts, our antibody, derived from a recombinant protein antigen, recognized a 116-kDa collagenase-sensitive protein and a 27-kDa collagenaseresistant fragment (9). The latter was in accord with the size expected for the 256-residue carboxyl terminus, whereas the former appeared considerably smaller than would be predicted for the 1388-residue intact protein. In a separate analysis, other investigators showed that their type XV antibody, generated from a synthetic carboxyl peptide, reacted with a 110-kDa band in human heart extract and with 110-and 70-kDa bands in kidney samples (10). Subsequent use of antibodies from both sources, however, revealed a similar pattern of BMZ localization (9,10).
To conduct a comprehensive biochemical characterization of type XV collagen, we embarked upon a series of purification steps beginning with human placenta tissue. Type XV was identified using both carboxyl-terminal antibodies and a new antibody prepared against a peptide sequence located in the amino-terminal noncollagenous domain. The results presented here surprisingly show that type XV collagen exists as a chondroitin sulfate proteoglycan in the five human tissues examined. In four of these tissues, type XV is present as 250-and 225-kDa core protein forms, which differ at their carboxyl terminus. In vitro studies of type XV production in cultured human cells showed that almost all of the newly synthesized collagen is secreted into the medium, consists of only the 250-kDa core protein form, and is modified by the addition of chondroitin sulfate chains. Further analysis of type XV collagen by differential use of the domain-specific antibodies revealed that the trimer is linked by interchain disulfide bonds and these involve only two of the eight cysteines in the molecule. Taken together, our data provide crucial new insight into the structure and expression of this complex BMZ collagen.

MATERIALS AND METHODS
Affinity Purification of Type XV Antibodies-Preparation of the type XV collagen recombinant protein (corresponding to the first 120 residues of the noncollagenous carboxyl-terminal domain) and the original type XV-COOH antibody (COOH-Ab) have been described previously (9). Sera from the three additional rabbits injected (Berkeley Antibody Co., Richmond, CA) with the carboxyl recombinant protein (500 g/ rabbit) were precipitated by addition of 50% saturated ammonium sulfate, dialyzed against phosphate-buffered saline and affinity-purified using Affi-Gel 10 resin according to the manufacturer's instructions (Bio-Rad). The type XV NH 2 -antibody (NH 2 -Ab) was prepared using a 15-amino acid peptide (GPGDEEDLAAATTEE) synthesized by the Protein Chemistry Laboratory of the University of Pennsylvania School of Medicine. The peptide was coupled to keyhole limpet hemocyanin and injected into two rabbits. Serum was affinity-purified as above using Affi-Gel 15 resin.
Cell Culture-The CCL136 human rhabdomyosarcoma cell line was obtained from the American Type Culture Collection and grown in a humidified atmosphere of 5% CO 2 at 37°C. Cells in T75 flasks were grown for 2 days in RPMI medium 1640 containing 10% fetal bovine serum (Sigma) to 90% confluence and used to seed five T75 flasks at a density of ϳ7.5 ϫ 10 6 cells/flask (ϳ100,000 cells/cm 2 ) in growth medium pre-equilibrated in the CO 2 incubator. After 22 h, the cells at ϳ75% confluence were washed twice with medium to remove the serum, and were then incubated in medium containing 0.1% serum and 50 g/ml ascorbate. The following day, the cultures had reached ϳ90% confluence. Ascorbate was again added to a final concentration of 50 g/ml (to the preexisting medium), and cultures were maintained for an additional 20 h. The respective media and cell layer/matrix fractions from the five flasks were pooled and processed as described in the following section.
Preparation of Small Scale Samples from Cell Cultures and Human Tissues-Medium from the cultures (10 ml/flask) was quick-chilled in an ice slurry and adjusted to a final concentration of 5 mM EDTA, 0.2 mM phenylmethylsulfonyl fluoride (PMSF), and 10 mM N-ethylmaleimide (NEM). The medium was clarified for 10 min at 7500 ϫ g and concentrated 25-35 fold at 2800 ϫ g at 4°C using Centriplus XM-100 filters (Amicon/Millipore). Preparation of cell/matrix samples has been described previously (20). Cell/matrix and medium samples were aliquoted in small volumes, quick-frozen, and stored at Ϫ75°C.
Normal human tissue samples, obtained from the Hospital of the University of Pennsylvania and the Cooperative Human Tissue Network, were frozen at Ϫ75°C or in liquid nitrogen, normally within 60 min after excision. Tissue (ϳ0.25-0.5 g) was homogenized for 5 ϫ 1 min at 30,000 rpm in an ice-chilled buffer consisting of 50 mM Tris-HCl, pH 7.5, 100 mM NaCl, 1 mM EDTA, 1% Nonidet P-40, 0.2 mM PMSF, 22 mM NEM, 0.0162 trypsin inhibitor units/ml aprotinin, and 20 g/ml leupeptin. The mixture was placed on a rocker platform at 4°C for 30 min and centrifuged for 10 min at 1500 ϫ g. Small aliquots of the supernatant were stored at Ϫ75°C.
Partial Purification of Type XV Collagen-Fresh human placenta (60 g wet weight) was washed several times in sterile ice-chilled 50 mM Tris-HCl, 4.5 M NaCl, 20 mM EDTA, pH 7.5, 1 mM PMSF, 2 mM NEM, 1 g/ml pepstatin A and stored at Ϫ75°C. Frozen tissue was ground to a powder in a mortar and pestle chilled in liquid nitrogen. The tissue powder was added to a urea-containing extraction buffer in a 5-10:1 (v/w) ratio of buffer to tissue under non-reducing conditions. The buffer consisted of 7 M urea, 50 mM Tris-HCl, pH 8.5, 1 mM EDTA, 50 mM NaCl, 2% CHAPS (Sigma) (21) and the protease inhibitors 100 mM ⑀-amino-n-caproic acid, 10 mM NEM, and 1 mM PMSF. (Extraction in 4 M guanidine HCl, pH 5.8, buffer was equally effective.) The tissue suspension was stirred at 4°C for 24 h and centrifuged at 32,000 ϫ g for 20 min. The solubilized protein (100 mg) was incubated with 15 ml of Q-Sepharose Fast Flow resin (Amersham Pharmacia Biotech) preequilibrated in 7 M urea, 50 mM Tris-HCl, pH 8.5, 1 mM EDTA, 50 mM NaCl, plus the above protease inhibitors. The sample was incubated with the resin on a rocker overnight at 4°C. The suspension was centrifuged at 480 ϫ g and the resin washed first with binding buffer containing 0.2% CHAPS; second, with the same buffer containing 0.5 M NaCl; and third, with the same buffer containing 1 M NaCl. To measure what remained bound, an aliquot of the resin was subsequently boiled in SDS-polyacrylamide gel buffer. All fractions were analyzed by Western blotting. Type XV collagen eluted in 1 M NaCl buffer and was concentrated by ultrafiltration using Centriplus XM-100 filters. The concentrated sample was brought to a final concentration of 100 mM DTT and applied to a 1 ϫ 115-cm column of Sephacryl S-500 Superfine (Amersham Pharmacia Biotech) equilibrated in 50 mM Tris-HCl, pH 8.0, 0.25 M sodium sulfate, 20 mM EDTA, and 4 M guanidine HCl (22). Fractions were eluted and assayed by dot-blot screening, and the peak fractions were pooled and concentrated by ultrafiltration.
Collagenase and Glycosidase Digestions-Bacterial collagenase (Advance Biofactures, Lynbrook, NY) digestions have been described previously (7,9). Chondroitinase ABC, protease-free (Proteus vulgaris) and heparitinase (Flavobacterium heparinum) were purchased from Seikagaku Corp. (Rockville, MD). Chondroitinase digestions (15 l) were performed for 90 min at 37°C in 100 mM Tris-HCl, 30 mM sodium acetate buffer, pH 8.0, using 20 milliunits of enzyme. In those reactions where collagenase digestion was sequentially performed, 4 l of 50 mM calcium acetate was added, followed by 3 l (3 units) of bacterial collagenase. Reactions were then incubated for another 60 min at 37°C. Heparitinase digestions (15 l) were carried out for 90 min at 37°C in a pH 7.0 buffer consisting of 100 mM sodium acetate, 10 mM calcium acetate, using 5 milliunits of enzyme. In those reactions where bacterial collagenase digestion was sequentially performed, 4 l of 5ϫ collagenase buffer (250 mM Tris-HCl, pH 7.2, 50 mM calcium acetate) was added, followed by 3 l (3 units) of bacterial collagenase. Reactions were then incubated for another 60 min at 37°C. In all procedures, the control digestion(s), indicated in the specific figure legend, was incubated in the identical buffer and for the identical time and temperature except without the respective enzyme(s).
Alcian Blue and Silver Staining of Type XV Collagen-The procedure was modified from the 20 steps listed in the published report (23). The major differences involved changes in the following steps (S): S1, overnight; S4, 45 min; S5-S7 10 min each; S8 and S9, 1 h each (or until the background cleared); S10, 10 min; S15, 20 min. Additionally, S15 to S20 were conducted at room temperature.

Western Blot Analysis of Type XV Collagen in Human Placenta Using Carboxyl-and Amino-terminal Domain-specific
Antibodies-Following preparation of our initial type XV antibody (9), three additional carboxyl-derived antibodies (COOH-Abs) were generated using the same recombinant protein as antigen. Immunostaining of human tissues using the new type XV antibodies (7,24) revealed the identical pattern of BMZ staining reported earlier (9). Western blot analysis using these COOH-Abs was consistent except in one respect. Only the original type XV COOH-Ab, prepared from the early bleeds of one rabbit, reacted with a 116-kDa collagenase-sensitive protein 2 found in extracts from several human tissues (9). However, all four type XV COOH-Abs identified a 27-kDa collagenase-resistant fragment (Fig. 1, lane 2, and Ref. 9), which was the size expected for the 256-residue carboxyl-terminal noncollagenous domain (see Fig. 9). In addition, all four COOH-Abs reacted with very high molecular mass, collagenase-sensitive material, which remained in a 5% stacking gel of an 8% separating gel (used in prior SDS-PAGE, Ref. 9), but which appeared as a diffuse area Ն400 kDa upon migration through a 3% stacking gel into a 5% separating gel (Fig. 1, lane 3).
To address the questions presented by these different protein forms, a peptide sequence in the type XV amino terminus was targeted for preparation of another polyclonal antibody (see Fig. 9 and "Materials and Methods"). By immunohistochemistry, this purified antibody (NH 2 -Ab) showed the same BMZ staining as the COOH-Abs (24). In Western blot analyses of untreated placenta extracts, the NH 2 -Ab reacted minimally with the high molecular mass collagenase-sensitive form identified with the COOH-Ab. After collagenase treatment, however, there was no evidence of any size cleavage fragment (expected to be Ն50 kDa, see Fig. 9) corresponding to the type XV amino-terminal domain (data not shown).
Purification of Type XV Collagen by Ion Exchange and Gel Filtration Chromatography-It therefore appeared that purification steps would be required to further characterize the type XV protein. To this end, fresh frozen placenta was pulverized in liquid nitrogen, and the protein extracted in a 7 M urea buffer under non-reducing conditions (see "Materials and Methods"). Following centrifugation, the solubilized protein was bound to Q-Sepharose and eluted stepwise using increasing concentrations of NaCl. As shown in Fig. 2A, Western blot analysis of the various fractions using the collagenase-resistant, 27-kDa carboxyl-terminal fragment as a marker (9), showed that all of the type XV collagen bound to the resin and little was removed in subsequent washing steps using the binding buffer. The protein remained strongly bound in 0.5 M NaCl, but completely eluted at 1 M NaCl ( Fig. 2A, lanes 6 and 7). The type XV collagen-containing protein pool was concentrated by ultrafiltration and applied to a Sephacryl S-500 gel filtration column under denaturing and disulfide bond-reducing conditions. Dotblot screening of all fractions showed that the ␣1(XV) chain eluted in a broad peak with an apparent mass of 300 -400 kDa (data not shown).
The above concentrated pool, examined by Western blotting using the COOH-Ab, showed the expected 27-kDa collagenaseresistant fragment after digestion (Fig. 2B, lane 2). Now, for the first time, a strong immunoreactive signal was seen using the NH 2 -Ab (Fig. 2B, lanes 3 and 4), explaining that the lack of this signal in the unpurified extract was due to an unacceptably low concentration. In the untreated sample, a high molecular mass, very diffuse form was identified, whereas in the sample treated with collagenase, a smaller and equally diffuse ϳ215-kDa (at the midpoint) form was seen (Fig. 2B, lane 4). This collagenase-resistant species was presumed to have originated from the 530-residue amino-terminal domain (see Fig. 9).
Type XV Collagen in Placenta Is a Chondroitin Sulfate Proteoglycan Involving GAG Chain Attachment to a High Molecular Mass Amino-terminal Domain-The ability of urea and/or guanidine-HCl and detergent to enhance extraction, the binding to Q-Sepharose at high ionic strength (25), and the diffuse band on Western blots of both the intact protein and the putative amino-terminal domain suggested that type XV collagen may be a proteoglycan. To test this possibility, purified type XV collagen from placenta (see above) was incubated with either heparitinase or chondroitinase ABC, and the digests were analyzed by Western blotting using the COOH-Ab (Fig. 3, A  and B). The results showed that, although heparitinase digestion had no effect on mobility of the immunoreacted material (Fig. 3A, lane 2), chondroitinase digestion generated a 250/225-kDa doublet (Fig. 3B, lane 2). (Following collagenase digestion, the carboxyl-terminal 27-kDa fragment was not retained on the 5% gel illustrated in Fig. 3B, lane 3, although it could be seen in a 10% or 12% gel (see Fig. 1, lane 2 for example).) To verify the identity of the 250/225-kDa protein, and to try to define the amino-terminal collagenase-resistant fragment, a duplicate set of placenta samples was treated as above and Western blotted using the NH 2 -Ab (Fig. 3C). In the sample treated with chondroitinase, the same collagenase-sensitive 250/225-kDa doublet was recognized (Fig. 3C, lane 2). Moreover, following both chondroitinase and bacterial collagenase digestion, the type XV NH 2 -Ab recognized a discrete 135-kDa fragment (Fig. 3C, lane 3), which is derived from the 215-kDa diffuse band in samples treated with collagenase but not chondroitinase (Fig. 2B, lane 4). Taken together, the results showed that type XV is a chondroitin sulfate proteoglycan, that most, if not all, of the GAG chains are attached to the amino-terminal domain, and that the latter polypeptide migrates anomalously on SDS-polyacrylamide gels. The presence of just one size of amino fragment also indicated that the difference in mass between the 250/225-kDa doublet did not reside within this domain.
Direct Identification of the Purified Type XV Collagen-proteoglycan on Silver-stained Gels-To determine if the type XV  6 and 7). In lane 8 is a sample of the Q-Sepharose resin (following elution) that was mixed with 2ϫ gel loading buffer and boiled for 2 min. Note that essentially all the type XV collagen bound to the resin in the 0.05 M NaCl buffer and was eluted at 1.0 M NaCl. Panel B illustrates size fractionations-purified type XV collagen Western blotted using the COOH-Ab (lanes 1 and 2) or NH 2 -Ab (lanes 3 and 4). In the samples incubated without collagenase, the immunoreactive material remained in the 5% stacking gel of the 12% or 8% separating gel (lanes 1 and 3, respectively). In the samples incubated with collagenase, the COOH-Ab identified the 27-kDa fragment (lane 2) whereas the NH 2 -Ab identified a diffuse band with a mean mass of ϳ215 kDa (lane 4, bracket).  B and C, lanes 2 and 3), followed by bacterial collagenase where indicated (panels A, B, and C, lane 3) and as detailed under "Materials and Methods." Samples were then electrophoresed on a 5% SDS-polyacrylamide gel, electroblotted, and reacted with the COOH-Ab (panels A and B) or the NH 2 -Ab (panel C). Note that heparitinase digestion had no effect on the mobility of the immunoreactive diffuse bands (panel A, lane 2), whereas chondroitinase ABC digestion generated a 250/225-kDa doublet (panels B and C, lane 2). The same two type XV bands, identified using the COOH- Ab (panel B, lane 2), were seen using the NH 2 - Ab (panel C, lane 2). In the sample treated with chondroitinase followed by bacterial collagenase, a single band of 135 kDa, representing the type XV amino-terminal domain was now detected (panel C, lane 3). The 27-kDa collagenase-resistant carboxyl-terminal fragment (seen in Fig. 1, lane 2) was not retained on the 5% gel in panels A and B (lanes 3).
protein could be directly visualized after chondroitinase treatment, an aliquot of the pooled fractions eluting from the Sephacryl S-500 column was electrophoresed on a 5% polyacrylamide gel and stained with alcian blue and neutral silver. This highly sensitive method has a detection limit of 0.04 -1 ng of proteoglycan (23). As illustrated in the sample incubated without chondroitinase (Fig. 4, lane 1), almost all of the staining was seen as a widely diffuse area in the upper portion of the separating gel. However, in the chondroitinase-digested sample, the majority of this stained material disappeared; instead, two new bands of the expected size, 250 and 225 kDa, were clearly evident (Fig. 4, lane 2). Their identity as type XV collagen was verified by Western blotting of a parallel set of samples in adjacent lanes (Fig. 4, lanes 3 and 4). This profile also revealed the equivalent ratio of the 250-to the 225-kDa band as was seen in the stained sample. The estimated yield of type XV collagen through the purification steps (ϳ35 ng/60 g of tissue) showed that this is an extremely low-abundance protein.
Type XV Collagen Is a Chondroitin Sulfate Proteoglycan in All Five Human Tissues Examined-We had previously determined by immunohistochemistry that type XV collagen is present in BMZs of the 10 human tissues examined (7,9). To biochemically characterize type XV collagen found in different tissues, protein extracts were prepared from umbilical cord, skeletal muscle, lung, and colon. As shown in Fig. 5A, in the absence of chondroitinase, type XV was consistently seen as a diffuse, high molecular mass entity. In the presence of chondroitinase (Fig. 5B), this material resolved into 250-and 225-kDa bands of varying intensity, except for umbilical cord, where only the 250-kDa band was seen (Fig. 5B, lane 2). Thus type XV collagen exists as a chondroitin sulfate proteoglycan in all five tissues analyzed.
The Type XV 250/225-kDa Forms Differ at Their Carboxyl Terminus-Emanating from the chondroitinase digestion of the tissue extracts was the question of the multiple core protein forms. Since among the five tissues examined, only umbilical cord (from four independent sources) showed a single 250-kDa form, we decided to compare the collagenase-resistant products generated from this tissue to those generated from placenta, a tissue containing both the 250-and 225-kDa forms. Having seen no heterogeneity at the amino terminus (Fig. 3C, lane 3), we focused on the carboxyl-terminal domain for structural differences. Speculating that only a small portion of the carboxyl terminus may still be attached to the 225-kDa form and not readily retained on Immobilon after collagenase digestion and electrophoretic transfer, we changed the blotting buffer and the membrane to those recommended for enhanced binding of low molecular mass proteins (see "Materials and Methods"). Placenta and umbilical cord extracts taken from fractions enriched for type XV collagen were incubated without or with collagenase, electrophoresed in an 18% polyacrylamide-Tricine gel, and Western-blotted using Immobilon-P SQ . As seen in Fig. 6 (lane  2), an 8-kDa band (in addition to the 27-kDa band) was present only in the placenta sample treated with collagenase; no such fragment was released by digestion of the umbilical cord protein (Fig. 6, lane 4). Thus, in some tissues, the 225-kDa form is apparently generated from the 250-kDa form by cleavage within the carboxyl-terminal domain. The 8-kDa fragment should include 12 collagenous residues (extrapolated from the recognition site for bacterial collagenase) and therefore about 60 residues of the carboxyl-terminal domain (see Fig. 9).
Since it was possible that proteolysis during the extraction process could have created the 225-kDa form, we conducted six independent extractions of placenta tissue to determine whether there was a decrease in the ratio of the 250/225-kDa bands. Type XV was extracted for 0.5, 3, and 5 h in the absence or presence of protease inhibitors (our standard conditions are 30 min with inhibitors present; see "Materials and Methods") and analyzed by SDS-PAGE and Western blotting. The profiles in the six lanes were indistinguishable (data not shown); there was no decrease in the intensity of the 250-kDa form, nor were there any new lower molecular mass bands, further suggesting that the cleavage in the type XV chain occurred in situ.
Glycosaminoglycan Chain Modification of Type XV Collagen in Cultured Human Cells-Type XV collagen examined to this point represented the in vivo form of the protein deposited in tissues. An important complement to this source would be an in vitro model system, which would allow us to address specific aspects of type XV biosynthesis. A viable candidate was the human rhabdomyosarcoma cell line, CCL136. In the course of previous studies, we had established that these rapidly growing cells expressed several newer nonfibrillar collagen proteins/ RNAs, including type XV (7,20), 3 in addition to matrix molecules previously characterized (26 -29).
An experiment was designed to analyze the relative distribution of type XV in the CCL136 cell layer and medium and to determine if the collagen in these fractions was posttranslationally modified by the addition of GAG chains. Cells were plated at high density such that within 24 h the cultures would tolerate both addition of vitamin C and reduction in serum concentration to 0.1% (for later concentration by ultrafiltration). Following 2 days' incubation under these conditions, the media and cell layer fractions were collected. Western blotting using the COOH-Ab was performed with samples incubated in the absence or presence of chondroitinase. The results showed that the cell layer contained a paucity of type XV collagen compared with the amount in the medium (Fig. 7). A single band of ϳ225-kDa was found in the cell layer fraction, and the intensity and mobility of this band were the same regardless of chondroitinase treatment (Fig. 7, lanes 1 and 2). (Double the amount of cell layer sample was applied to the gel in order to visualize the bands.) In contrast, the strong type XV collagen signal in the medium was found exclusively in the form of a 3 J. C. Myers, unpublished data. FIG. 4. Identification of the purified type XV collagen-proteoglycan by combined alcian blue and silver staining. An aliquot of the peak fractions eluted from the S-500 column was electrophoresed in a 5% SDS-polyacrylamide gel before (lanes 1 and 3) and after chondroitinase ABC digestion (lanes 2 and 4). Part of the gel (lanes 1 and 2) was stained with alcian blue and neutral silver (see "Materials and Methods" and Ref. 23), and the adjacent part (lanes 3 and 4) was Western-blotted using the COOH-Ab (the same results were seen using the NH 2 -Ab; data not shown). The amount of the sample used for Western blotting was 30% of the amount loaded in the lanes that were stained. Note also that the ratio of the two type XV bands, 250 and 225 kDa, visualized by silver staining, was very similar to that seen in the Western blots.
250-kDa band seen only after chondroitinase digestion (Fig. 7,  lane 4). Therefore, all of the type XV collagen in the cell layer was present as the unmodified core protein, and all of the type XV collagen in the medium was present as the chondroitin sulfate proteoglycan.
Type XV Chains Are Disulfide-linked Only through the Cysteines in the Collagenous Region-Another important question to be resolved was whether the cysteines in the type XV chain participate in interchain disulfide bonds. Two cysteines are present in the amino terminus, two are in interruptions within the collagenous region, and four are found in the carboxyl terminus (see Fig. 9). An aliquot of the CCL136 medium was electrophoresed in the absence of reductant, following treatment with either chondroitinase ABC, or chondroitinase plus collagenase, and Western blotted using both COOH-and NH 2 -Abs. As seen in Fig. 8 (A and B), without DTT in the sample buffer (lanes 1) type XV migrated near the top of the gel, whereas in the presence of DTT (lanes 3) the 250-kDa ␣1(XV) chain was seen. In contrast, electrophoresis of the collagenaseresistant, 27-kDa carboxyl-terminal fragment was unaffected by DTT (Fig. 8A, lanes 2 and 4; Ref. 9). Likewise, migration of the chondroitinase-treated, collagenase-resistant 135-kDa amino-terminal fragment was the same without or with reduction (Fig. 8B, lanes 2 and 4), demonstrating that just the two cysteines located within the collagenous region ( Fig. 9) could participate in interchain disulfide bonds. (The same result was found using placenta tissue, and the chondroitinase-treated protein extract in the absence and presence of DTT is shown in Fig. 8B (lanes 5 and 6, respectively).) DISCUSSION Type XV and Other Collagen-proteoglycan Molecules-Of the 19 collagen types so far identified, 5 are known to be proteoglycans in at least some species or tissues: types IX (30,31), XII (32,33), XIV (32), XVIII (12), and now type XV. Types IX, XII, and XIV belong to the FACIT subgroup (fibril-associated collagens with interrupted triple helices) due to their common structural features and molecular interactions with fibrillar collagens in dense connective tissues (reviewed in Refs. 34 and 35). In some tissues, especially cartilage, a significant portion (but not all) of the FACIT proteins are substituted with chondroitin/ dermatan sulfate chains, making them "part-time" proteoglycans (30,32,33,36).
Collagen types XV and XVIII constitute a very different subclass by virtue of their unique domain homology and widespread BMZ distribution (7)(8)(9)(10)(11)(12)37). Unlike the FACIT group, these two proteoglycans are distinguished by the type of GAG  1 and 2) and umbilical cord (lanes 3 and 4) following a 1 M NaCl extraction and 5 M NaCl precipitation were incubated in the absence (lanes 1 and 3) or presence (lanes 2 and 4) of bacterial collagenase, electrophoresed in an 18% polyacrylamide-Tricine gel, Western-blotted using Immobilon-P SQ membrane (see "Materials and Methods"), and reacted with the COOH-Ab. The 27-kDa fragment (lanes 2 and 4) represents the complete carboxyl terminus. The 8-kDa fragment was only generated from digestion of the placenta (and crude colon extract; data not shown) sample which contains both the 250-and 225-kDa forms in contrast to umbilical cord (and rhabdomyosarcoma medium proteins; data not shown), which contains only the 250-kDa form. In lane 3, the undigested umbilical cord sample is not detected in transfer from the high percent polyacrylamide gel and is only seen as a relatively low signal even in the 5% gel shown in Fig. 5A, lane 2. The figure shown is a 5-min exposure. Even in a 30-min (much darker) exposure, there was still no evidence of the 8-kDa fragment in the umbilical cord sample treated with collagenase (lane 4). chain attached; type XVIII collagen contains heparan sulfate (12), whereas data presented here established that type XV collagen contains chondroitin/dermatan sulfate. Although the organization of these collagens in the BMZ (or basement membrane) is not yet defined, it is important to note that, in contrast to the FACIT subgroup, types XV and XVIII collagen appear to be "full-time" proteoglycans.
Type XV Collagen Glycosaminoglycan Sites-The fact that type XV collagen is a chondroitin/dermatan sulfate proteoglycan is consistent with its enhanced extractability from tissues using urea or guanidine solutions, its elution from Q-Sepharose at high ionic strength (Fig. 2), and its diffuse pattern on SDS-PAGE prior to chondroitinase ABC digestion (Fig. 3). The observation that the collagenase-resistant amino terminal fragment (which is likely also to include the first interruption in the collagenous region) is chondroitinase-sensitive showed that it contains most, if not all, of the GAG chains. The type XV amino-terminal domain per se (Fig. 9) has sequence similarities to aggrecan, the major cartilage proteoglycan (14), and has a total of eight candidate serines in the D/E-X 1-2 -S-G/A consensus configuration, where X 1-2 is one or two hydrophobic and/or small, neutral residues (38). Four such additional sites are situated in two of the largest interruptions in the collagenous region ( Fig. 9 and Ref. 3). Although the available data do not permit an accurate determination of how many GAG chains are present, molecular mass estimates before and after chon-droitinase digestion suggested that they contribute Ն200 kDa to the ␣1(XV) chain (Fig. 3, B and C). If one assumes that the average chondroitin/dermatan sulfate chain in matrix proteoglycans is 10 -50 kDa (for review, see Refs. 39 and 40), then each collagen chain could be decorated with 4 -20 GAG chains. It is thus conceivable that all consensus sites are occupied.

Identification of the Type XV Amino-terminal Domain Exhibiting an Unusually High Molecular Weight by SDS-PAGE-
Following chondroitinase ABC and bacterial collagenase digestions, the NH 2 -Ab identified a 135-kDa protein (Fig. 3C), a size initially difficult to reconcile with the primary structure. We had expected a ϳ65-kDa fragment representing the 530-residue amino terminus plus 65 residues of the collagenous domain, which precede the presumptive bacterial collagenase cleavage site (Fig. 9). However, the 135-kDa mass determined by SDS-PAGE was in fact consistent with the 250/225-kDa mass found for the intact ␣1(XV) core protein. The remaining 90 -115 kDa could be accommodated by the 27-kDa carboxyl terminus plus the 85 kDa expected for the collagenous region, which is generically known to display a mass by SDS-polyacrylamide gel electrophoresis about 50% greater than that estimated by amino acid content (7,41).
The aberrant size observed for the type XV amino terminus is likely due to its highly acidic nature (a pI of 4.01 compared with 6.85 for the rest of the protein). A similar electrophoretic anomaly was reported for type VII collagen NC-2-derived recombinant fragments, which have a pI of 4.3 and migrate 1.7-1.8 times larger than their predicted mass would indicate (42). This finding is in accord with previous characterization of several acidic histone-binding proteins and neurofilament proteins (43,44). Their high glutamic acid composition was deemed responsible for overestimation of the molecular mass by SDS-PAGE, a feature directly shown for the latter group by sedimentation equilibrium centrifugation (43).
Type XV Interchain Disulfide Bond Formation Occurs Exclusively via the Two Cysteines Contained within the Collagenous Domain-Our results have shown that type XV chains are linked by interchain disulfide bonds, and this manner of chain association is limited to the two cysteines within the collagenous region. These cysteines are separated by 231 residues; one begins a 31-amino acid interruption, and the second is near the center of a 34-amino acid interruption ( Fig. 9 and Ref. 3). The extended distance between these two cysteines in a type XV homotrimer would require that this portion of the chain (encompassing approximately one-third noncollagenous sequences) must loop out in order to bring the involved residues within acceptable proximity for bond formation. Therefore, it may be particularly noteworthy that the ␣1(XVIII) chain lacks these two cysteines (5,45). Instead, the only cysteines within the confines of the ␣1(XVIII) collagenous region (in the human chain (Ref. 45) but absent in the mouse chain (Ref. 5)) are eight residues apart in the first interruption, and included in a short segment that may undergo alternative splicing (45).
In contrast to the collagenous region, the distribution of cysteines in the types XV and XVIII noncollagenous terminal domains can be directly compared (11,13,14). Each collagen contains four homologous cysteines in their respective carboxyl terminus, and the two cysteines in the type XV amino terminus correspond to those in the short (S) form of the type XVIII amino-terminal variants (11,13,14). It has not yet been determined whether the type XVIII collagen chains are covalently bound via any of their cysteines, and, therefore, the manner of disulfide bond formation (and implied secondary/tertiary structure alluded to above) may prove to be quite different from type XV collagen.
Newly Synthesized Type XV Collagen in Human Cultured The cell layer and medium fractions from CCL136 cells were processed as described under "Materials and Methods." The medium (containing 0.1% fetal calf serum) was concentrated to ϳ1/30 of the original volume. A proportional amount of the cell layer homogenate (v/v) was equivalent to ϳ21 g of protein. Double that amount (ϳ42 g) was electrophoresed on the gel (lanes 1 and 2) in order to detect the type XV signal; therefore, the relative signal shown in the figure reflects a 2-fold overestimation of the amount of protein retained in the cell layer. Samples were Westernblotted following incubation without (lanes 1 and 3) or with (lanes 2 and 4) chondroitinase and reacted with the COOH-Ab. The molecular mass of type XV collagen in the cell layer was ϳ225 kDa regardless of chondroitinase digestion (lanes 1 and 2), and the intensity of this protein was identical in both lanes. The type XV collagen in the medium was ϳ250 kDa, approximately the same size as the upper band in the placenta doublet in Fig. 4. No signal was detected in this exposure of the medium fraction before chondroitinase treatment (lane 3), but it was visible in longer exposures. Both the 225-and 250-kDa proteins were completely digested with bacterial collagenase (data not shown). The difference in mass between these two forms of type XV may be due to "stubs" of GAG remaining after chondroitinase digestion since the amino-terminal collagenase-resistant band in the medium sample is 135 kDa, identical to the tissue form (Fig. 3C), but larger than the 114-kDa band obtained from the cell layer fraction (data not shown).

Cells Is Glycosylated and Efficiently Secreted-
The studies shown here on newly synthesized type XV collagen provide the first information on its biosynthesis in mammalian cultured cells. In the CCL136 line, most (Ͼ90%) of the protein was found in the medium and little was retained in the cell layer fraction. This finding argues against type XV being a pericellular collagen. Essentially the same dominant media distribution was seen for type III (the major collagen produced by the rhabdomyosarcoma cells; Ref. 27) when the blots, initially reacted with a type XV antibody, were stripped and reprobed (data not shown). However, most importantly, the results demonstrated that the type XV collagen secreted was the same size as that found in vivo and contained the same post-translational modifications as the protein extracted from fresh tissues. All of the type XV protein in the medium was disulfide-bonded and sensitive to chondroitinase ABC digestion, whereas the small amount associated with the cell layer was an unmodified, nondisulfide-linked form of the core protein (Figs. 7 and 8, and data not shown). The tissue culture data further inferred that the cell layer-associated material was in fact intracellular since GAG chain elongation occurs in the Golgi apparatus (for review, see Ref. 46).
In other laboratories, a 200-kDa protein was produced in insect cells using a recombinant baculovirus construct of the human type XV chain sequences (10). This molecule was not secreted and was found instead in the detergent-insoluble fraction of these cells. The lack of modification is not surprising since published reports state that "it is unlikely that insect cells have the post-translational machinery required for effective processing of a proteoglycan" (47).
Differential Cleavage of Type XV Collagen Chains at the Carboxyl Terminus-Our data support the hypothesis that some 250-kDa type XV chains are cleaved in situ within the carboxyl-terminal noncollagenous domain to produce a 225-kDa form (Fig. 9). This event occurs in placenta, colon, lung, and skeletal muscle but not in umbilical cord or the rhabdomyosarcoma cells; 250-and 225-kDa forms are present in the former, but only the 250-kDa form is present in the latter (Figs. 5B and 8). Accordingly, an 8-kDa carboxyl fragment, in addition to the 27-kDa carboxyl terminus, was generated upon collagenase digestion of placenta (and colon) type XV collagen, FIG. 9. Schematic diagram showing major structural features of the mature human type XV collagen chain, the epitopes for the NH 2 -and COOH-Abs, and the location of the multiple polypeptide forms and bacterial collagenase cleavage fragments. The three domains of type XV collagen are drawn proportional to their size. The amino terminus (diagonal stripes) contains 530 residues, the discontinuous collagenous region (white and gray areas) encompasses 577 residues, and the carboxyl terminus (dotted pattern) includes 256 residues (3,13,14). White areas in the collagenous domain indicate the presence of continuous G-X-Y triplets; intervening gray areas and dashed lines in the collagenous subdomains lines show the location of large and small interruptions, respectively. Potential Olinked GAG attachment sequences (D/E-X 1-2 -S-G/A: eight in the amino terminus and four in the interruptions) are designated by vertical ball and stick symbols. C ϭ cysteines: two in the amino-terminal domain, two in the collagenous region, and four in the carboxyl-terminal domain. The two cysteines in the collagenous region that are predicted to be involved in interchain disulfide bonds are tagged with an asterisk. The heavy black lines, drawn above the type XV map, show the location of amino-and carboxyl-terminal sequences used to prepare the synthetic peptide and recombinant protein antigens for production of the NH 2 -and COOH-Abs, respectively. The predicted positions of the 250and 225-kDa forms, and the collagenase-resistant fragments consisting of the 135-kDa amino-terminal domain and the 27-and 8-kDa carboxyl fragments are designated. The internal boundaries of the 135-, 27-, and 8-kDa fragments coincide with the positions of the nearest recognition sites for bacterial collagenase (G-P-Y2G-P) (48). Note that molecular mass values of the 250-, 225-, and 135-kDa forms have been derived only from polyacrylamide gel electrophoresis of chondroitinase-treated samples.
but not from digestion of umbilical cord protein or rhabdomyosarcoma medium proteins ( Fig. 6 and data not shown).
The results presented here, therefore, raise the intriguing question of whether a putative ϳ19-kDa released fragment (resulting in the 225-kDa type XV form, Fig. 9) corresponds to those fragments isolated by John et al. (19) from 10,000 liters of human hemofiltrate. In their search for endostatin (type XVIII carboxyl)-related fragments, these authors identified two type XV peptides beginning at residues 66 and 81, respectively, of the 256-residue carboxyl domain. It is important to note that the proposed cleavage site in the type XV chain required to generate these fragments closely corresponds to where we predicted the 225-kDa form to terminate, approximately at residue 60 of the carboxyl-terminal domain (Fig. 9). The concordance of these studies provides a novel focus for future experiments aimed at characterizing the processing of a select population of type XV chains, and determining possible biological relevance of the cleaved fragments.