Proteoglycan Lt from Chicken Embryo Sternum Identified as Type IX Collagen*

Proteoglycan Lt (PG-Lt), isolated from 17-day-old chicken embryo sterna, appeared to differ from its counterpart from tibia and femur (Noro, A., Kimata, K., Oike, Y., Shinomura, T., Maeda, N., Yano, S., Takahashi, N., and Suzuki, S. (1983) J. Biol. Chern. 258, 9323-9331). The intact -disulfide-bonded molecule of approximately 300 kDa was separable into three chains of 115,84, and 68 kDa on reduction, the molecular masses being relative to those of collagen standards on sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). This is in contrast to tibial cartilage PG-Lt, from which there was no ob- served release of a 68-kDa chain (100 kDa relative to globular protein standards) after reduction. The 115- kDa chain of sternum PG-Lt consists of a core 68-kDa polypeptide to which the chondroitin sulfate chains are attached. The ratio of 4-sulfated to 6-sulfated disaccharides released after either chondroitinase ABC or AC digestion is 3:l. Identity of PG-Lt with type IX collagen was indicated by their similar elution profiles on DEAE-Trisacryl and by the presence in both pro-teins of co-migrating 84- and 68-kDa bands on SDS- PAGE. This identity was confirmed by immunoblotting PG-Lt after

Proteoglycan Lt (PG-Lt), isolated from 17-day-old chicken embryo sterna, appeared to differ from its counterpart from tibia and femur (Noro, A., Kimata, K., Oike, Y., Shinomura, T., Maeda, N., Yano, S., Takahashi, N.,  J. Biol. Chern. 258,[9323][9324][9325][9326][9327][9328][9329][9330][9331]. The intact -disulfide-bonded molecule of approximately 300 kDa was separable into three chains of 115,84, and 68 kDa on reduction, the molecular masses being relative to those of collagen standards on sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). This is in contrast to tibial cartilage PG-Lt, from which there was no observed release of a 68-kDa chain (100 kDa relative to globular protein standards) after reduction. The 115-kDa chain of sternum PG-Lt consists of a core 68-kDa polypeptide to which the chondroitin sulfate chains are attached. The ratio of 4-sulfated to 6-sulfated disaccharides released after either chondroitinase ABC or AC digestion is 3:l. Identity of PG-Lt with type IX collagen was indicated by their similar elution profiles on DEAE-Trisacryl and by the presence in both proteins of co-migrating 84-and 68-kDa bands on SDS-PAGE. This identity was confirmed by immunoblotting PG-Lt after SDS-PAGE, with affinity-purified polyclonal antibodies specific for a triple helical domain (HMW) of type IX collagen. The nonreduced high molecular mass material and all three bands of the reduced PG-Lt were immunoreactive, giving immunostaining patterns similar to autoradiographs from the ['4C]glycine-labeled protein.
Bruckner from the Swiss National Science Foundation and through a generous donation to L. Vaughan from B. Weiss. These results were presented in preliminary form to the IXth Meeting of the Federation of European Connective Tissue Societies (FECTS) held on the 17-20 July, 1984 in Budapest. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$ To whom correspondence should be addressed.
The second family has as yet only one clear member. This is PG-Lt from chicken embryo cartilage with an estimated 2 chondroitin (dermatan) sulfate chains (32). PG-Lt has three disulfide-linked polypeptide chains of 120 and 100 kDa, comprising 70-80% of the molecular mass of the intact molecule of 420 kDa. This is in contrast to PG-Lb in which the polypeptide represents only 20% of the total molecular mass.
A unique feature of PG-Lt was that both the 120-and 100-kDa polypeptides were digestible with bacterial collagenase. The collagenous nature of the protein was further substantiated by the finding of hydroxyproline and hydroxylysine on amino acid analysis.
The relationship of PG-Lt to the known collagens, in particular to the minority collagens of cartilage, remained to be established. We, therefore, isolated PG-Lt from chicken embryo sternum using similar procedures to those of Nor0 et al. (32) and report here that the PG-Lt so isolated is immunologically and structurally identical to type IX collagen. were from Sigma. Unsaturated disaccharide standards were prepared as previously described (33). Bacterial collagenase (grade CLSPA) was purchased from Wortbington. Nitrocellulose filters were provided by Schleiffer (Feldbach, Switzerland). The horseradish peroxidaselabeled goat anti-rabbit antibody was from Nordic Immunological Laboratories (Tillburg, Netherlands) and goat anti-rabbit immunoglobulin G was from Gibco (New York). DEAE-Trisacryl was from LKB (Bromma, Sweden).

Methods
Incubation, Extraction, and Separation of the Proteoglycans-The sternal cartilages from 17-day-old chicken embryos were dissected free of perichondria. They were rinsed with, and placed into, 30 ml of Dulbecco's modified essential medium supplemented with 0.25 mM 3-aminopropionitrile fumarate and 0.25 mM ascorbic acid. Sterna from 120 embryos were used in each experiment. Incubation at 37 "C was commenced after addition of either ['*C]glycine (2 pCi/ml), Naz[35S]sulfate (10 pCi/ml), or [2-3H]mannose (30 pCi/ml) and continued for 20-24 h with gentle shaking. The medium was then decanted off and the cartilages immersed in cold (0 "C) extraction solution of 4 M guanidinium HC1 buffered with 50 mM Tris-HC1 to pH 8.0 and containing a cocktail of proteinase inhibitors (1 mM phenylmethylsulfonyl fluoride, 100 mM 6-aminocaproic acid, 10 mM N-ethylmaleimide, and 10 mM EDTA). The sterna were disrupted by a brief homogenization and then extracted with slow stirring at 0 "C for 16-24 h. The resultant suspension was clarified by centrifugation at 27,000 X g for 30 min at 0 "C. The macromolecular components were precipitated with 3 volumes of cold ethanol containing 1.3% (w/ v) potassium acetate. They were redissolved in a minimal volume of extraction solution prior to rate zonal centrifugation under dissociative conditions.
Ion-Exchange Chromatography-PG-Lt was purified by DEAE-Trisacryl column chromatography following the procedures of Nor0 et al. (32). Briefly, 0.2% (w/v) Triton X-100 was added to the top fraction from the CsCl isopycnic gradient which was then dialyzed against the starting buffer of 7 M urea, 50 mM Tris-HC1, pH 7.4, and 0.2% (w/v) Triton X-100. After dialysis the sample was applied to a 1.5 X 12-cm column of DEAE-Trisacryl. The column was first washed with 50 ml of the starting buffer, material bound to the resin was eluted with a linear gradient of 0-0.6 M NaCl in the same buffer, and 3-ml fractions collected. All procedures were carried out at 4 "C.
Chain Separation-Purified PG-Lt labeled with either [14C]glycine or [35S]sulfate was precipitated with 3 volumes of cold ethanol containing 1.3% (w/v) potassium acetate in 1.5-ml plastic centrifuge tubes. After centrifugation at 4 "C, the supernatant was discarded and the precipitate redissolved in 100 pl of 50 mM Tris-HC1, pH 7.0, 8 M urea, 2% (w/v) Triton X-100, and 5% (v/v) 2-mercaptoethanol. The samples were reduced at 60 "C for 1 h, with intermittent mixing. The samples were then diluted with an equal volume of buffer A (50 mM citrate HCl buffer, pH 2.5 containing 8 M urea, 0.2% Triton X-100 (w/v), and 1% (v/v) 2-mercaptoethanol) and applied to a 4.5 X 50-mm column of DEAE-Trisacryl equilibrated with the same buffer. Nonbound material was eluted with 3 ml of buffer A, and 300-p1 fractions were collected. A linear gradient of 0-0.6 M NaCl in buffer A desorbed the majority of the bound sample. Increasing the NaCl concentration to 1.0 M failed to elute further radiolabeled material.
Zmmunoblotting-Proteins were electrotransferred from SDS-PAGE slab gels to nitrocellulose filters (36). The transfer was achieved with 100 mA for 16-24 h. To prevent nonspecific binding of antibodies, the filters were incubated for 1 h with a blocking buffer of 10 mM Tris-HC1, pH 7.4, containing 5% (v/v) fetal calf serum, 1% (w/v) bovine serum albumin, 100 mM MgC12, and 0.5% (w/v) Tween 20 (37). Subsequently the filters were immersed in a minimal volume of this buffer containing the antibodies against type IX collagen at an appropriate dilution and incubated with gentle shaking overnight. The filters were then washed three times for 10 min each with blocking buffer. A peroxidase-labeled second antibody was diluted 1:lOOO in the blocking buffer and incubated with the filter for 3 h at room temperature. The excess second antibody was removed by three washes of 10 mM Tris-HC1, pH 7.4,150 mM NaCl. The substrate for the peroxidase was 3-chlornaphthol.
Preparation of Antibodies-Affinity-purified antibodies to pepsinreleased HMW domain of type IX collagen from chicken sternum (4) were prepared and characterized as described previously (17).
Enzyme Digestions-Incubation of PG-Lt with chondroitinase ABC or AC was performed in the presence of proteinase inhibitors (38). Digestion of PG-Lt (5 X lo3 cpm as '*C or %S, 2 X 10' cpm as 3H) was carried out at 35 "C for 5 h with 10 pg of purified bacterial collagenase, 100 pl of 100 mM Tris-HC1, pH 8.0, buffer containing 0.4 mM CaC12, 0.25 mM N-ethylmaleimide and 2 mg/ml bovine serum albumin.
Paper Chromatography-Descending paper chromatography of the products of the chondroitinase ABC or AC digestions was performed as described in Ref. 39.

RESULTS
The extracted proteoglycans of 17-day chicken embryo cartilages were separated under denaturing conditions into the PG-H and PG-L populations (31) by rate zonal centrifugation (Fig. la). PG-L was further separated into PG-Lb (highest density fractions) and PG-Lt (lowest density fractions) by CsCl isopycnic density gradient centrifugation (Fig.  lb). The profiles obtained with [35S]sulfate-and with ["C] glycine-labeled material are shown. Profiles similar to those shown for the ['4C]glycine-labeled preparations were obtained when [2-3H]mannose was used as a tracer (results not shown).
The pooled lowest density fractions from the CsCl centrifugation were then applied to a DEAE-Trisacryl column. Elution was with a linear NaCl gradient under dissociating conditions. The [3SS]sulfate-labeled PG-Lt from sternum eluted as a major broad peak centered at 0.35 M NaCl (Fig. 2b). A corresponding peak was obtained for each of the other radiolabels employed (Fig. 2, a and c), albeit with considerable differences in the remainder of the chromatograms. In each case the fractions under the 0.35 M NaCl peak were pooled and rerun on a similar column. This produced a single peak, well resolved from minor contaminants (not shown). The fractions from this peak were then pooled and concentrated by ultrafiltration.
Aliquots from the pooled peak fractions were then subjected to SDS-PAGE on 5-20% (w/v) gradient-acrylamide gels followed by fluorographic detection. The [14C]glycine, [35S]su1fate, and [3H]mannose tracers all labeled a protein which ran unreduced as a broad but single band migrating with an apparent molecular mass of approximately 300 kDa (Fig. 3,  b-d). Upon reduction the ['4C]glycine-labeled material yielded three bands (Fig. 3e). One ran as a diffuse band centered at 115 kDa and two as sharper bands at 84 and 68 kDa, relative to collagen standards. In contrast, [35S]sulfate labeled only the diffuse band at 115 kDa (Fig. 3f). Both the 84-and the 68-kDa bands could be labeled with [3H]mannose, but labeling of the 115-kDa band was inconclusive (Fig. 3g). Occasionally a high molecular mass doublet could be observed in nonreduced samples. After reduction of these samples one diffuse and one sharp high molecular mass band of 210 and 168 kDa, respectively, could be discerned. These presumably represented dimers joined through lysine-derived cross-links. The appearance of these bands was minimized by the inclusion of 3-aminopropionitrile fumarate in the organ culture medium.
Treatment of the [I4C]glycine or [3H]mannose-labeled material with either chondroitinase ABC or AC sharpened the high molecular mass nonreduced band of PG-Lt on SDS-PAGE and decreased its apparent molecular mass by approximately 50 kDa (Fig. 4, c and e). Upon reduction of digested material containing either of the above labels, the diffuse 115-kDa band was no longer visible, but a corresponding increase in the intensity of the 68-kDa band was noted (Fig. 4, g and   i). 14C-or 3H-Labeled material could not be detected in the supernatant after ethanol precipitation of the digest nor was there material running as a low molecular mass band(s) on SDS-PAGE in 5-20% acrylamide gels. These results indicated that the change in apparent mass of the high molecular mass band was not due to proteolytic degradation. Conversely, more than 98% of the radioactivity could be recovered in the ethanol supernatant after chondroitinase ABC or AC treatment of

b,
CsCl isopycnic density gradient centrifugation to fractionate PG-L into PG-Lb and PG-Lt. The lowest density fractions corresponding to PG-Lt were pooled as shown (-). and c, [2-3H]mannose-labeled PG-Lt. Radioactivity is denoted by . " -. and the NaCl gradient by --.
[36S]sulfate-labeled material, with a corresponding disappearance of bands on SDS-PAGE (Fig. 4, d and h).
To more rigorously address the question of whether the 115-kDa diffuse band converts to a 68-kDa band upon chondroitinase ABC treatment, it was first necessary to separate the polypeptide chains. This was achieved by reduction of the protein under denaturing conditions, followed by ion-exchange chromatography at pH 2.5 on DEAE-Trisacryl. Absorbed material was eluted in denaturing conditions with a linear gradient of NaCI. Submitted to this procedure, ["C] glycine-labeled PG-Lt chromatographed as two peaks (Fig.  5a). The first peak eluted with the wash buffer and comprised 38% of the label. The second peak was desorbed by 0.3-0.4 M NaCl and contained 53% of the applied radioactivity. When PG-Lt labeled with [36S]sulfate was treated similarly, the 36S label eluted exclusively in one peak with 0.35 M NaCl (Fig.  5 b ) . Electrophoresis of ["Clglycine-labeled material from peak 1 (Fig. 5a) showed that it consisted almost entirely of the 68-kDa chain (Fig. 6d), with minimal contamination by the 84-kDa chain. In contrast, peak 2 contained mostly the heterogeneous 115-kDa chain (Fig. 6e), which co-migrated with material from the corresponding peak of the [%]sulfatelabeled chromatogram (Fig. 6c).
Aliquots from the various peaks were digested with chondroitinase ABC. All the [%]sulfate label was released, causing the diffuse 115-kDa band to disappear on SDS-PAGE (not shown). Similar treatment caused the [14C]glycine label to reappear in a band of 68-kDa mobility (Fig. Sf). The relative proportion and mobility of the contaminating 84-kDa band (lanes e and f ) was unchanged by chondroitinase ABC treatment. These results provided further evidence that the polypeptide carrying the chondroitin sulfate chain(s) has an apparent molecular mass of 68 kDa and migrates with the same electrophoretic mobility as the nonsulfated 68-kDa chain (Fig.  6, d and f ) . PG tinase ABC or chondroitinase AC, and the products were analyzed by paper chromatography (Table I). The ratio of 4to 6-sulfated disaccharides released by either enzyme was identical. Chondroitinase AC effected complete digestion to disaccharides, with none of the oligosaccharide fragments observed which would otherwise be indicative of dermatan sulfate. This shows that the PG-Lt isolated from chicken embryo sternum does not contain dermatan sulfate as was found in PG-Lt isolated from tibia and femur (32). Treatment of the PG-Lt preparation with bacterial collagenase resulted in the disappearance of the high molecular mass band in the unreduced and all three bands (115,84, and 68 kDa) in reduced samples run on SDS-PAGE (results not shown). The [14C]glycine and [3H]mannose labels ran at the buffer front. Based on the electrophoretic mobilities of CNBr fragments from type I collagen, this result indicates that any noncollagenous domains of PG-Lt must be around 10 kDa or smaller.
To test whether PG-Lt and type IX collagen were indeed the same protein, samples were run on SDS-PAGE and then electrophoretically transferred to nitrocellulose plates. The bands were detected using affinity-purified antibodies, spe-icken Embryo Sternum 4761 cific for one of the triple helical portions (HMW) of type IX collagen (17). The immunostaining pattern (Fig. 7) was similar to that of the ['4C]glycine-labeled protein (refer to Figs.  3 and 4). An immunoreactive high molecular mass band was observed in nonreduced samples (Fig. 7a). Treatment with chondroitinase ABC increased the electrophoretic mobility of the antibody-stained band in a comparable way to that observed with the ['4C]glycine-labeled material (Fig. 7b). Following reduction and electrophoresis the 115-, 84-, and 68-kDA bands were all immunoreactive to the type IX antisera (Fig. 712). The immunostaining over the diffuse 115-kDa band was cleared by prior treatment with chondroitinase ABC. This resulted in a concomitant increase in antibody staining over the 68-kDa band (Fig. 7 4 .

DISCUSSION
Despite the differences between our preparation of PG-Lt and that of Nor0 et al. (32), the bulk of the evidence favors them being the same molecule. The centrifugation profiles of proteoglycans extracted from sternum (Fig. 1, this paper) and those extracted from tibia and femur (32) are similar. This enables the comparable PG-H, PG-Lb, and PG-Lt fractions to be isolated. In both cases PG-Lt is the major [35S]sulfatelabeled macromolecule with a buoyant density of less than 1.34 g/ml on CsCl density gradient centrifugation. This is evident from the DEAE-column chromatography of this lowest buoyant density fraction, where all but a small proportion of the incorporated [35S]sulfate label elutes in the PG-Lt peak.
On SDS-PAGE a disulfide-bonded molecule of approximately 300 kDa, reducible to a diffuse [35S]sulfate-labeled chain of 115 kDa is observed. This is accompanied by a narrow nonsulfated protein band at 84 kDa, a pattern similar to that previously described by Nor0 et al. (32) for tibial PG-Lt. The molecular masses of 420 kDa for the intact molecule, 190 kDa for the [35S]sulfate-labeled band, and 120 kDa for a nonsulfated [3H]serine-containing band were calculated by these authors relative to globular protein standards. In our hands, calibration against globular protein standards gave comparable molecular masses of 380, 160, and 110 kDa, respectively. In the interests of simplicity and because of the domina-lt collagenous nature of PG-Lt, these chains will be subsequently referred to in terms of the values estimated here relative to collagen standards.
A discrepancy between our results and those of Nor0 et al. (32) is that no 68-kDa band (100 kDa relative to globular protein standards) can be seen in fluorographs published by, nor is it noted by, these authors. This may be due to the labeling intensity, which is very light in the figure shown. It may also be that the content of serine is high in the 84-kDa chain relative to the 68-kDa chain such that the latter is not readily detected when [3H]serine is used as tracer. However in both cases the mobility of the 115-kDa band changes to a band of 68 kDa after chondroitinase ABC digestion. Finally, the unique collagenous nature of the intact proteoglycan as well as that of the individual polypeptide chains is evident from both preparations with the disappearance of all bands on 5-20% SDS-PAGE following digestion with bacterial collagenase.
The PG-Lt from tibia was classified as a proteodermatan sulfate (32). We were unable to find any evidence of dermatan sulfate in our preparations of PG-Lt from sternum, which possibly reflects differences between the two tissue sources. Certainly intratissue variation can be expected as the ratio of 4-to 6-sulfated chondroitin sulfate disaccharides in the PG-L fraction has been shown to decrease progressively from the epiphyseal through to the diaphyseal region of developing were pooled as shown (-) and analyzed by SDS-PAGE (Fig. 6).  GlcUA-GalNAc(6-SO4) 24 25 tibia from 12-day chicken embryos (40). Intertissue differences have also been observed. Dermatan sulfate, which is abundant in fibrous cartilage (41), was undetectable in hyaline cartilage (42). Interestingly, the ratio of 4-to 6-sulfated disaccharides of 3:l is the same for PG-Lt from either source and differs from the ratio of 1:3:1 found for the major aggregating proteoglycan (40).
The incorporation of [2-3H]mannose into PG-Lt, presumably as a constituent of asparagine-linked oligosaccharide chains, was noted previously (32). Here we confirm those observations and show that the 68-kDa chains incorporate the majority of the label. The structure of these oligosaccharides, their number, and sites of attachment must await further studies.
Interestingly, type IX collagen had also been found to migrate as a disulfide-bonded molecule with an apparent molecular mass of 300 kDa (17,19). In addition, polypeptide chains of 84 and 69 kDa could be released upon reduction. It had an unusually strong anionic nature for a collagen, as evidenced by its relatively high affinity for DEAE-cellulose (19). Furthermore, on closer examination of the published SDS-PAGE patterns of type IX collagen (17, 19) a diffuse band of about 115 kDa could be discerned in preparations of purified material, in addition to the 84-and 69-kDa bands. The presence of this diffuse band was not addressed in these publications.
The first direct evidence that these two molecules are identical is provided by the immunological identity between PG-Lt and type IX collagen demonstrated here. This conclusion is based not only on the immunoreactivity of the disulfide-linked high molecular mass band of PG-Lt but also on all three of its constituent polypeptide chains. This latter result also demonstrates that all three of these chains, including the glycosaminoglycan-containing 115-kDa chain, must participate in the triple helical HMW domain of type IX collagen. This conclusion relies on the specificity of the antibodies, which were raised against the purified HMW domain of type IX collagen (17). They were affinity purified and shown to be specific for HMW with no cross-reactivity with the la-, 2a-, and 3a-chain collagen, type I1 collagen, or indeed even the LMW domain of type IX collagen (17). The immunoreactivity of the antibodies was not affected by chondroitinase ABC treatment of PG-Lt.
The role(s) of this unique proteoglycan-collagen hybrid is at present unknown. PG-Lt has been shown to have a different tissue distribution to that of PG-H, PG-Lt being located in both the perichondrial tissue and the cartilage matrix proper, while PG-H is restricted to the latter (32). This result has a bearing on the debate over the distribution of type IX collagen. Although some groups favor a predominantly pericellular location (43,441, in recent studies from this laboratory at the level of the electron microscope we have found type IX collagen in both peri-and extracellular spaces. In addition, type IX collagen appears to be associated mainly with fibers of type I1 collagen.* Our results thus correspond well with those of Nor0 et al. (32), who were, in addition, able to immunostain fibrillar material on the cell surface and in the intercellular spaces in chondrocyte cultures with antibodies The primary role of PC-Lt/type IX collagen may well concern the regulation of collagen fibrils and their interactions, not only with other collagen fibrils, but'also with other proteoglycan species with which they must inevitably be in intimate contact. The unique proteoglycan/collagen hybrid nature of PG-Lt/type IX collagen may well provide the key to such a "link" role.
to PG-Lt.