Proteoglycans from the swarm rat chondrosarcoma. Structure of the aggregates extracted with associative and dissociative solvents as revealed by electron microscopy.

Proteoglycan aggregates were extracted from Swarm rat chondrosarcoma tissue in the native state and compared with proteoglycan aggregates isolated dissociatively with 4 M guanidine HCl. Purified aggregates were examined with a variety of electron microscopic techniques. In some cases they showed a structure of the central filament identical to that of the link-stabilized central filament observed in earlier experiments where the separated constituents were allowed to reconstitute (Mörgelin, M., Paulsson, M., Hardingham, T. E., Heinegård, D., and Engel, J. (1988) Biochem. J. 253, 175-185). The tight packing of proteoglycan monomers along the hyaluronate with a minimum distance of 12 nm between adjacent E1 strands also could thus be confirmed for never dissociated aggregates. The results therefore show that the organization of proteoglycan aggregates assembled in vitro from the participating molecules is representative for conditions in situ. An additional structural type of central filament was observed in the preparations. This contained long stretches of free hyaluronate interspaced by short stretches of central filament with condensed arrays of link protein-proteoglycan. Chemical cross-linking in combination with low shear electron microscopical techniques showed that this discontinuous central filament structure is not an artifact of specimen preparation. The addition of suprastoichiometric amounts of exogenous link protein did not affect the central filament structure with the low packing density. Densely and loosely packed types of central filament were isolated in varying relative amounts with different associative and dissociative solvents.

imum distance of 12 nm between adjacent El strands also could thus be confirmed for never dissociated aggregates. The results therefore show that the organization of proteoglycan aggregates assembled in vitro from the participating molecules is representative for conditions in situ.
An additional structural type of central filament was observed in the preparations. This contained long stretches of free hyaluronate interspaced by short stretches of central filament with condensed arrays of link protein-proteoglycan. Chemical cross-linking in combination with low shear electron microscopical techniques showed that this discontinuous central filament structure is not an artifact of specimen preparation. The addition of suprastoichiometric amounts of exogenous link protein did not affect the central filament structure with the low packing density.
Densely and loosely packed types of central filament were isolated in varying relative amounts with different associative and dissociative solvents.
Proteoglycans constitute the major non-collagenous extracellular matrix component of normal hyaline cartilage. They occur predominantly as a very large aggregating species with *This study was supported by grants from the Swiss National Science Foundation and the Swedish Medical Research Council. The work was greatly facilitated by a short-term European Molecular Biology Organization fellowship (to M. M.). The continuous support of the M. E. Muller Foundation has been essential for this investigation. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisernent" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The aggregating proteoglycans (aggrecan) bind specifically to a decasaccharide or larger segment of hyaluronate . Because this non-covalent interaction and the ensuing formation of large aggregates is fundamental for retaining these molecules in cartilage, much attention has been focused on the mechanism of aggregation. It involves a specialized portion of the core protein, the hyaluronate binding region (Heinegdrd and , which has been identified as the globular domain G1 at the NH2 terminus of the molecule (Doege et al., 1987;Paulsson et al., 1987;Morgelin et al., 1988). The interaction is further stabilized by the binding of link protein (Heinegdrd and Hardingham, 1979) with affinity for both the hyaluronate and hyaluronate binding region (Tengblad, 1981;Franz6n et al., 1981), leading to the formation of a very stable ternary complex (Bonnet et al., 1985). A large number of proteoglycans may bind to hyaluronate, and the resulting aggregated structure with its very high negative charge density is responsible for cartilage elasticity and resilience.
Chaotropic solvents such as 4 M guanidine HC1 effectively dissociate the components in the aggregates by abolishing non-covalent interactions. The dissociated components are readily extracted in high yields (Sajdera and Hascall, 1969). Dialysis into an associative solvent such as 0.5 M guanidine HCl allows a large portion of the molecules to reaggregate.
Taking advantage of this behavior, we have previously studied the assembly of intact monomers and fragments thereof with hyaluronate and link protein by electron microscopy (Morgelin et al., 1988) using glycerol spraying/rotary shadowing (Shotton et al., 1979;Tyler and Branton, 1980). Aggregates reconstituted in the absence of link protein exhibited a rather loose structure of the central filament consisting of G1 and hyaluronate, where individual GI domains were distinguished and apparently had a statistical distribution along the hyaluronate. Although gaps of variable size were present, closest center-to-center distances of 12 nm between adjacent G1 domains were measured. A more condensed continuous central filament structure, with a tight packing of monomers along hyaluronate, was observed in the presence of link protein. Individual domains of either G1 or link protein of Rat Chondrosarcoma Proteoglycans were not resolved. Long stretches of densely packed central filament regions alternating with short regions of free hyaluronate indicated a high degree of cooperativity in protein binding. Again the closest distances between neighboring monomers were 12 nm. The same continuous central filament structure was previously demonstrated when A l l preparations of bovine nasal cartilage were visualized by this technique (Wiedemann et al., 1984).
The protein components employed in these reconstitution experiments had presumably been denatured during isolation and purification. The significance of the results depends on the proper refolding of the molecules into a native conformation and their self-assembly into supramolecular structures representative of in situ organization. Therefore, in the present study, we compared reassociated to native aggregates. The latter can be extracted in large amounts from the Swarm rat chondrosarcoma (Choi et al., 1971). Proteoglycan monomers from rat chondrosarcoma tissue have a molecular structure similar to those obtained from cartilage of other species, except that they do not contain any keratan sulfate chains (Choi et al., 1971), and the proline-rich keratan sulfate attachment polypeptide sequence is lacking (Doege et al., 1987;Antonsson et al., 1989). Whereas proteoglycans of hyaline cartilage resist quantitative extraction with associative solvents, those from the tumor tissue can be easily isolated in non-denaturing buffers, i.e. with 0.5 M guanidine HC1,0.05 M sodium acetate, pH 5.8. Under these conditions the noncovalent interactions of the central filament components are not dissociated (Oegema et al., 1975;Faltz et al., 1979a). In the present study the detailed structure of native aggregates was studied using different electron microscopical replica techniques and turned out to be indistinguishable to that of dissociatively prepared (Al) or reconstituted link-stabilized aggregates. We could also demonstrate that two forms of central filament structure in native aggregates can be differentially extracted from the chondrosarcoma by use of a variety of associative solvents. A preliminary report on some aspects of this work has been presented elsewhere (Morgelin et al., 1990).

MATERIALS AND METHODS
Maintenance of Swarm Rat Chondrosarcoma-Frozen tumor cells were kindly provided by A. Blattler, Ciba-Geigy AG, Basel, Switzerland. The tumor was maintained in male Sprague-Dawley rats, weighing 250-350 g, by subcutaneous injection of 0.5-1 ml of tumor tissue minced in 0.15 M NaC1, 0.05 M Tris-HC1, pH 7.4 (TBS, ahout 200 mg/ml suspension). At 4-5-week intervals, the rats were killed and tumor nodules, 20-30 g (wet weight) per animal, were harvested and dissected free of surrounding fascia. The cleaned tumor tissue was rapidly frozen on dry ice and stored at -20 "C.
Extraction Procedures-Extractions were carried out at 4 "C and, unless indicated otherwise, in the presence of the following protease inhibitors: 0.05 M EDTA, 0.1 M 6-aminohexanoic acid, 5 mM Nethylmaleimide, 5 mM benzamidine hydrochloride, and 0.5 mM phenylmethanesulfonyl fluoride. The latter two inhibitors were added to the solvents just prior to the extraction. Tumor tissue was sliced with a razor blade and briefly (10 s) homogenized in 5 ml/g wet weight of prechilled 0.5 M guanidine HC1 (GdnHCl)? 0.05 M sodium acetate, pH 5.8, using a Polytron homogenizer (Kinematica, Littau, Switzerland) at half-maximal speed, and subsequently extracted for 4 h with stirring. Alternatively, briefly homogenized tumor tissue was ex-' The nomenclature of Heinegird (1977) is used throughout. Additionally, the term a-A1 is used to indicate associately extracted A1 fractions.
tracted dissociatively for 4 h with 5 ml/g wet weight of prechilled 4 M GdnHC1,0.05 M sodium acetate, pH 5.8, in the presence of protease inhibitors.
In sequential extraction experiments, dissected tumor nodules were homogenized in 5 volumes of prechilled TBS, containing 5 mM benzamidine hydrochloride, 5 mM N-ethylmaleimide, and 0.5 mM phenylmethylsulfonyl fluoride. The resulting tissue dispersion was extracted for 1 h with stirring. Residues were collected by centrifugation for 20 min at 8,000 rpm (4 "C, Sorvall, GSA rotor, 10,000 X ga.). This extraction cycle was repeated twice, and the three supernatants were immediately frozen and stored at -20 "C. The tissue residues were then extracted three times as described above in the same buffer containing in addition 10 mM EDTA (TBS-EDTA). Pellets were extracted for 4 h with 5 volumes of 0.5 M GdnHCl, 0.05 M sodium acetate, pH 5.8, and finally with 4 M GdnHCl, 0.05 M sodium acetate, pH 5.8. These extracts were dialyzed into associative conditions, i.e. against 10 volumes 0.05 M sodium acetate, pH 5.8. Aliquots (1 ml) of the first extract obtained with each solvent were chromatographed on Sepharose CL-2B as described below.
Preparation of a-A1 Fractions-0.5 M GdnHCl extracts were cleared by centrifugation in capped polycarbonate tubes (13,000 rpm, 27,000 X g , , , 45 min, Sorvall GSA rotor). The density was adjusted to give 1.63 g/ml by the addition of 1 g of solid CsCl/g of extract. The solutions were then centrifuged in Beckman Quick Seal polyallomer tubes at 34,000 rpm (100,000 X gav), 15 "C for 48-72 h in a Kontron TFT 50.38 rotor. The gradients were separated into four fractions after freezing the centrifuge tubes in liquid nitrogen. Fractions were analyzed for protein and proteoglycan contents by measuring the optical density at 280 nm, dot blots on nitrocellulose developed with toluidine blue, SDS-polyacrylamide gel electrophoresis, and electron microscopy. Material in the bottom one-fourth was collected and dialyzed twice against 40 volumes of 0.5 M sodium acetate, pH 7.0. This fraction is subsequently referred to as a-A1 (associative extract, A1 fraction of associative gradient).
Preparation of Link Protein-A 4 M GdnHCl extract was clarified by centrifugation (8,000 rpm, 10,800 X g,,, 20 min, Sorvall GS3 rotor) and dialyzed overnight at 4 "C into associative conditions. The dialyzed extract was then centrifuged in cesium chloride density gradients as described above. The A1 fraction was recovered and mixed with an equal volume of 8 M GdnHC1, 0.1 M sodium acetate, pH 5.8, and centrifuged under dissociative conditions (starting density 1.53 mg/ml). The resulting A1D4 fraction was purified further by chromatography on Sepharose CL-GB, followed by chromatography and rechromatography on Sephacryl S-200. Columns were eluted with 4 M GdnHCl, 0.05 M sodium acetate, pH 5.8.
Digestion with Chondroitinase ABC-Proteoglycan aggregate samples from an a-A1 preparation were either dialyzed extensively against 0.1 M sodium acetate, 0.1 M Tris, pH 7.3, or mixed with the appropriate volume of concentrated buffer. They were usually digested with 1 unit of chondroitinase ABC/100 mg proteoglycan for 4-6 h at 37 "C in the presence of ovomucoid (10 pg/ml) as a protease inhibitor. Proteoglycan concentrations were 5-10 mg/ml. Digested samples were subsequently purified by gel filtration on Sepharose CL-2B as described below.
Column Chromatography-Gels for column chromatography were obtained from Pharmacia Fine Chemicals, Uppsala, Sweden. For large scale preparative purposes a Sepharose CL-2B column (37 X 5 cm) was equilibrated with 0.5 M sodium acetate, pH 7.0. Samples of extracts, a-A1 or chondroitinase ABC treated a-A1 (40 ml, about 5% of bed volume), corresponding to 100-200 mg of proteoglycan, were loaded, and the column was eluted with the same buffer. Flow rates were 40 ml/h, and fractions of 10 ml were collected. Additional preparative columns were packed with either Sepharose CL-GB (100 X 2.5 cm) or Sephacryl S-200 (67 X 2.5 cm). Link protein (20 mg), dissolved in 10 m14 M GdnHCl, 0.05 M sodium acetate, pH 5.8, was loaded onto the column and eluted with the same buffer. Flow rates were 9 ml/h, and fractions of 4.5 ml were collected.
A semipreparative Sepharose CL-2B column (32 X 1 cm) was equilibrated with 0.5 M sodium acetate, pH 7.0, at a flow rate of 4 ml/h. Samples (1 ml, about 3% of bed volume) were applied and eluted with the same buffer. Fractions of 1 ml were collected. Column effluents were analyzed by SDS-polyacrylamide gel electrophoresis and carbazole assay as described below.
Analytical Procedures-Proteoglycan and hyaluronate samples from column effluents were analyzed for hexuronic content by the procedure of Bitter and Muir (1962). For the determination of hexuronic acid contents in crude extracts, samples were hydrolyzed with alkali prior to the carbazole assay (RodBn et al., 1972).
Protein was determined with the bicinchoninic acid Pierce protein assay (Pierce Chemical Co.). Column effluents from Sepharose CL-2B gel filtration of proteoglycan preparations were assayed for protein contents by a simplified microtiter plate bicinchoninic acid assay.
SDS-polyacrylamide slab gel electrophoresis was performed according to the method of Laemmli (1970). Usually aliquots of extracts or proteoglycan samples were concentrated by ethanol precipitation. Approximately 25-50 pg of protein were applied to 3-15% polyacrylamide gradient gels. When desired, samples were reduced with 2% flmercaptoethanol in the sample buffer. Proteins were detected by staining with Coomassie Brilliant Blue R.
Addition of Link Protein to Proteoglycan Aggregates-Link protein from chondrosarcoma, dissolved at a concentration of 1 mg/ml in 4 M GdnHCl, 0.05 M sodium acetate, pH 5.8, was added in increasing amounts (1.5, 3, 6, and 15% of proteoglycan mass, respectively) to aliquots of a purified a-A1 preparation in 0.5 M sodium acetate, 0.5% CHAPS (Fluka, Buchs, Switzerland), pH 7.0. The samples were allowed to react for 48 h at 4 "C during dialysis against 0.2 M ammonium hydrogen carbonate, pH 7.9. In parallel control experiments link-free aggregates were reconstituted by mixing proteoglycan monomers from bovine nasal cartilage or rat chondrosarcoma with 0.1% (w/w) high molecular weight hyaluronate (Healon, Pharmacia, Uppsala, Sweden) in 0.5 M sodium acetate, pH 7.0. The samples were allowed to react overnight at 4 "C, and then equimolar amounts of link protein were added. The CHAPS concentration was adjusted to 0.5% by adding appropriate amounts of a concentrated CHAPS solution. The samples were then dialyzed for 48 h at 4 "C against 0.2 M ammonium hydrogen carbonate, pH 7.9. Reconstituted and link protein-treated aggregates were diluted to final concentrations of 50-100 pg/ml with the same buffer and visualized by rotary shadowing electron microscopy.
Chemical Protein Cross-linking-Intact or chondroitinase-digested proteoglycan a-A1 preparations were purified by Sepharose CL-2B chromatography and dialyzed overnight at 4 "C against 0.15 M NaC1, 10 mM HEPES, pH 7.0. Samples (0.5-1 mg/ml) were incubated after addition of 2% (v/v) dithiobis(succinimidy1)propionate (DSP, Pierce) in dimethyl sulfoxide (10 mg/ml) for 1 h at room temperature. Alternative samples (0.5-1 mg/ml) were fixed in solution by the addition of 1% (v/v) glutaraldehyde and incubation for 1 h at room temperature. The reagents were subsequently removed by dialysis overnight at 4 "C against 0.2 M ammonium hydrogen carbonate, pH 7.9, and the samples were either subjected to SDS-polyacrylamide gel electrophoresis or prepared for electron microscopy as described below.
Electron Microscopy-Extracted or reconstituted aggregates were dialyzed overnight at 4 "C against 0.2 M ammonium hydrogen carbonate, pH 7.9, in a dialysis apparatus designed for small volumes (5-50 pl, Biowerk, Basel, Switzerland). Proteoglycan samples were then diluted with the same solvent to obtain a final concentration of 50-100 pg/ml. They were used for electron microscopy after either glycerol spraying (Shotton et al., 1979;Tyler and Branton, 1980), mica sandwich squeezing (Mould et al., 1985), or mica centrifugation (Nave et al., 1989). Specimens were subsequently dried at high vacuum for 1-2 h and rotary shadowed at a 9" angle with platinum/carbon, followed by carbon coating at 90'. Replicas were floated onto distilled water and picked up with 400 mesh copper grids. Electron micrographs were taken on a Zeiss 109 transmission electron microscope operated at 50 or 80 kV accelerating voltage. Magnifications were calibrated by photography of a calibration grid (Bakers, Liechtenstein) under the same electron-optical conditions. Measurements were performed on a screen after x10 enlargement of the negatives.

Isolation of Proteoglycan Aggregates from the Swarm Rat
Chondrosarcoma-Proteoglycan aggregates were isolated from the tumor tissue under mildest possible conditions, i.e. by extraction with 0.5 M GdnHCl containing protease inhibitors according to previously published protocols (Oegema et al., 1975;Faltz et al., 1979a). Extracts were purified by size exclusion chromatography on Sepharose CL-2B either directly or after associative CsCl density gradient centrifugation. Proteoglycan aggregates were recovered in high yields from void volume peaks of column effluents (results not shown).
In some cases the tumor tissue was sequentially extracted with different associative solvents, and finally with 4 M GdnHCl, in the presence of protease inhibitors. For comparison, in some experiments the tumor was directly extracted dissociatively with 4 M GdnHCl containing protease inhibitors, followed by dialysis into associative conditions. A schematic presentation of the different extraction procedures is given in Fig. 1. Aliquots of the clarified extracts were directly chromatographed o n a Sepharose CL-ZB column. All extracts gave essentially the same profile, with a sharp aggregate peak in the void volume, minimal amounts of material in the included portion, and a large peak due to small proteins and protease inhibitors in the total volume (results not shown). Material from the different extracts, excluded from the column, was identified by electron microscopy as proteoglycan aggregates (see below). Interestingly, large amounts of aggregates were solubilized from the tissue already in physiological buffer, i.e. TBS, without use of chaotropic or chelating additives ( Table I).
Extraction of the chondrosarcoma either for 4 h with 0.5 M GdnHCl, or in sequence with TBS, TBS-EDTA, and 0.5 M GdnHCl, solubilized about 70% of t h e total hexuronic acid. Further extraction of the residues with 4 M GdnHCl yielded an additional 20% of the total hexuronic acid. Upon extraction with 4 M GdnHC1, yields of about 86% of the total hexuronic acid were obtained (Table I). These results are in general agreement with published data (Faltz et al., 1979a).
Electron Microscopy of Native Aggregates-To find a specimen preparation technique that exposed aggregates t o a of Rat Chondrosarcoma Proteoglycans minimum of shear forces and thereby gave optimal structure preservation, samples of purified a-A1 fractions from 4 h 0.5 M GdnHCl extracts were visualized in the electron microscope by different preparation techniques. These were glycerol spraying/rotary shadowing (Shotton et al., 1979;Tyler and Branton, 1980), mica sandwich squeezing (Mould et al., 1985), and mica centrifugation (Nave et al., 1989).
Long aggregates with varying amounts of monomers bound t o hyaluronate strands a t different packing densities were observed with each method. The central filament of a more densely packed aggregate appeared identical to that of aggregates isolated under dissociative conditions from a number of tissues and then reconstituted (Figs. 2 and 3). When aggregate samples were prepared for electron microscopy by glycerol spraying, only about 5% of the monomers were in aggregates, the major fraction of the molecules remaining unbound (Fig.  3a). There was large variation in the ratio of bound to unbound particles between individual droplets, and especially  the smaller ones were often completely devoid of aggregate structures. Sometimes hyaluronate strands with only some proteoglycan monomers bound in a very oriented fashion were visible (Fig. 3a, inset), suggesting partial removal of proteoglycan monomers due to shear during spraying or upon impact on the mica surface.
In contrast, aggregate preparations after mica sandwich squeezing or mica centrifugation exhibited large fields densely covered with aggregates (Fig. 3, b and c ) . About 90% of the monomers were bound, and strands of unsubstituted hyalu-of Rat Chondrosarcoma Proteoglycans ronate were not observed. A higher degree of preservation was also demonstrated by the much larger aggregate length observed by the two latter methods (results not shown). It seems as if the mica squeezing or centrifugation techniques exert less shearing stress on the samples, yielding less destruction of long and thin structures compared to the glycerol spraying technique. This has previously been shown for other filamentous systems (Mould et ol., 1985;Nave et al., 1989). On the other hand glycerol spraying/rotary shadowing consistently gave the lowest background staining of 2-3 nm platinum crystallites, whereas clusters of varying sizes were often present in specimens prepared with the other two techniques. These artifactual structures are probably due to traces of salt and do sometimes cover large areas of the grids. Furthermore, glycerol spraying/rotary shadowing usually gave the highest resolution of the central filament and the glycosaminoglycan chains. In contrast, mica squeezing and centrifugation frequently showed the protein cores of individual monomers along their whole lengths, sometimes with particularly high contrast due to collapsed side chains. Hence, the three techniques were applied in parallel for optimal information.
Examination of the associatively prepared aggregates with a high packing density of monomers along the hyaluronate showed a structure of the central filament identical to that of link-stabilized aggregates seen in our previous reconstitution experiments (Morgelin et al., 1988(Morgelin et al., , 1989. Continuous stretches of central filament were seen as compact proteincovered strands extending over long distances, occasionally interspaced by free hyaluronate strands (Fig. 2). Individual G1 domains could not be resolved, but tight packing of proteoglycan monomers along the hyaluronate, with closest center-to-center distances of 12 nm, was demonstrated by measuring the distances between adjacent El strands.
The length of the extended domain E2 was 282 k 21 nm and 210 f 24 nm for those intact and chondroitinase digested monomers that contained G3, respectively. This is somewhat shorter than the corresponding lengths of the E2 domain of the bovine nasal cartilage proteoglycan (405 f 37 nm, intact and 263 & 27 nm, chondroitinase digested, Morgelin et al., 1989). The observed difference correlates with the finding that a proline-rich peptide (about 100 amino acids long), corresponding to the keratan sulfate attachment region, is present in the sequence of mature bovine cartilage core protein, whereas it is not found in the chondrosarcoma proteoglycan sequence (Doege et al., 1987;Antonsson et al., 1989).
Interestingly, a different aggregate type with a more loosely packed central filament structure was found in the same extracts (Fig. 4b). The prominent feature thereof is a similarly tight packing of monomers in densely staining clusters, but these are rather short and do usually not exceed 10 adjacent proteoglycan monomers. In addition to the clusters, single monomers with a binding region globe enlarged in diameter, presumably representing one G1 domain complexed with one or more link proteins, were often seen to be bound to the hyaluronate. Both types of central filament structure were observed with all electron microscopic preparation techniques, and it is therefore unlikely that the loosely packed central filaments are artifacts caused by the shearing forces of the glycerol spraying method. It should be noted, however, that occasionally both central filament types were found to somewhat overlap by transitional forms, or to be present within one single aggregate (Fig. 4c). Additional details on the molecular structure of the aggregate types are shown in Table 11. An average a-A1 preparation contained 60% of densely packed central filament structure, with mean dis-  (a and b ) , mica sandwich squeezing (c), and mica centrifugation ( d ) of non-dissociatively extracted proteoglycan aggregates from the Swarm rat chondrosarcoma. A representative densely packed aggregate is shown in a, aggregates with loosely packed central filaments can he seen in h, and an aggregate with regions of different central filament packing density is presented in c. In d, a very long aggregate with an almost completely uninterrupted central filament structure is shown.

TABLE I1
Structural details of proteoglycan agxregates extracted from the Swarm rat chondrosarcoma with 0.5 M GdnHCl The fraction of particles containing G3 was the same for particles present in densely and loosely packed central filament stretches. d denotes average distances between adjacent monomers along the hyaluronate. Monodisperse and heterodisperse distributions of d were determined for densely and loosely packed central filament t-ypes, resoectivelv.  (Fig. 5 b ) . The loosely packed central filament stretches accounted for 40% of the total central filament length. They showed broad distributions of distances between neighboring monomers (27 f 21 nm, Fig. 5d). Similar distances between adjacent monomers were measured for whole aggregates with overall predominant dense or loose central filament packing density (Fig. 5 , a and c). In both central filament types closest center-to-center distances between monomers were 12 nm. The COOH-terminal globular domain G3 was present in about 50% of the particles both in the densely and loosely packed types.
In some experiments 0.5 M CdnHCl extracts were directly applied to a preparative Sepharose CL-2B column in order to determine whether exposure to high salt concentrations and prolonged high centrifugal force during cesium chloride density gradient centrifugation might influence the central filament structure. In all cases, however, both densely and loosely packed central filaments were seen in the electron microscope, showing the same structural features as summarized for a-A1 preparations in Table 11. When the lengths of central filament patches or interspersing free hyaluronate strands were measured we obtained similar results as for aggregates from a-A1 preparations (results not shown). Therefore, it is unlikely that the pool of molecules with loosely packed central filaments is created during purification.
Chemical Cross-linkin~--xperiments were specifically designed to examine whether the interrupted structure of loosely packed central filaments was due to electron microscopic preparation artifacts. Therefore, proteoglycan aggregates were fixed by chemical cross-linking in solution either with glutaraldehyde or DSP. Subsequent, SDS-polyacrylamide gcl electrophoresis under reducing and non-reducing conditions showed that both reagents were able to covalently cross-link proteoglycan monomers and link protein in the central filament (Fig. 6) and thus presumably stabilized its structure. When aliquots of native a-A1 samples were electrophoresed under reducing or non-reducing conditions a link protein band at about M , 40,000 appeared. The link protein band was neither present in preparations after fixation with glutaraldehyde nor in non-reduced DSP-treated samples. I t was, however, visible in the lane of reduced DSP-treat.ed samples, included as a control for successful cross-linking by the thiol cleavable DSP spacers. A similar electrophoretic behavior was seen for fixed chondroitinase-digested aggregates from the a- A 1 preparation, where the core protein was seen as a faint smear with a mobility at about 400 kDa in addition to the link protein band (Fig. 6).
When cross-linked aggregates were visualized in the electron microscope, again densely and lonselv packed central filament structures were seen at the stme r d o s HS for rlntreated aggregate preparations (Fig. 7). Furthermore, when aliquots of a-A1 samples were digested with chondroitinase ABC and subsequently purified by Sepharose CL-2B chromatography, free stretches of hyaluronate had been digested and long aggregates were no longer visible, but inst.ead short, continuous aggregates were observed by electron microscopy. The length distributions of their central filarnerlts were the same as the sum of those from the cent.ral filament patches of undigested aggregates with densely and loosely packed regions (Fig. 8). These observations suggest that both central filament types are present in situ.
Differences between Proteoglycan Aggregates Sequentially Extracted by Different Solvents-Rat chondrosarcoma tissue was consecutively extracted with TBS, TBS-EDTA, 0.5 M GdnHCl, and 4 M GdnHCl. Proteoglycan aggregates from each fraction were recovered by Sepharose CL-ZB chromatography and compared by electron microscopy (Fig. 9). From each extract, aggregates with continuous and interrupted central filament structures were identified in the relative amounts given in Table I. The different extracts also varied in their content of G3-containing monomers. Representative aggregate structures are shown in Fig. 9, a-d proteoglycans from the TBS-EDTA extract exhibited a much higher proportion of densely packed central filament structure (60%) and an unusually large fraction of GB-bearing particles (85%). In contrast in the 4 M GdnHCl extracts, only loosely packed aggregates with a rather degraded appearance were observed. It should be noted, however, that these are reconstituted and were not observed in the presence of 4 M GdnHCl. Only few intact monomers were bound to the hyaluronate strands, and quite often whole aggregates consisting of only globular structures, presumably the G1 fragment, attached to hyaluronate were found. Aggregates solubilized in TBS had a similar but less degraded appearance. In each case closest center-to-center distances between adjacent monomers of 12 nm were observed within central filament patches. In all densely packed central filament stretches which were examined, average monomer distances of about 12 nm were found. In contrast, the loosely packed central filaments from all extracts exhibited rather broad distributions of intermolecular distances of 27 21 nm. In control experiments chondrosarcoma tissue was directly extracted dissociatively with 4 M GdnHCl, and the clarified extracts were dialyzed into associative conditions and chromatographed over Sepharose CL-PB. About 70% of the proteoglycan monomers were present in aggregates and appeared in the void volume (results not shown). Electron microscopical inspection of aggregates recovered from void volume fractions showed only loosely packed aggregates (Fig. 9e). Average distances of 40 f 2 nm between neighboring monomers along the hyaluronate were measured for such aggregates. Similar results were obtained when a-A1 samples were dialyzed into 4 M GdnHCl and back into associative conditions (results not shown). These data indicate an overall molar excess of hyaluronate in the tissue, resulting in low aggregate packing densities after dissociation and reassociation of central filament patches.
Treatment of Aggregates with Link Protein-A possible explanation for the presence of a loosely packed central filament structure would be that link protein is present only in substoichiometric amounts in the tissue, since in the absence of link protein proteoglycans bind to hyaluronate in a statistical way (Morgelin et al. 1988). To test this hypothesis, link protein was purified from an A1D4 preparation and added in varying amounts to purified a-A1 aggregate samples. In parallel control experiments it was added to link-free complexes reconstituted in vitro from hyaluronate and proteoglycan monomers. After dialysis into 0.2 M ammonium hydrogen carbonate such aggregates were visualized in the electron microscope. After addition of equimolar amounts of link protein, the reconstituted, previously link-free aggregates exhibited a link-stabilized central filament structure (Fig. 10, c and  d). In aggregates from the a-A1 preparation, however, even a large excess of added link protein did not affect the relative amounts of densely and loosely packed central filament forms (Fig. 10, a and b). The gaps between adjacent short central filament pieces were not closed. These results provide evidence that substoichiometric amounts of link protein are not responsible for the noncontinuous structure of loosely packed central filaments.

DISCUSSION
Proteoglycan aggregates can be extracted in high yields from the Swarm rat chondrosarcoma with associative solvents which do not disrupt their native supramolecular structure. It was shown in earlier work that an effective associative solvent is 0.5 M GdnHC1, 0.05 M sodium acetate, pH 5.8, which when used together with protease inhibitors yields the highest pro-d . portion of aggregates in an a-A1 preparation with least indication of proteolytic degradation of the proteoglycans (Oegema et al., 1975;Faltz et al., 1979a). In addition, as shown here, intact proteoglycan aggregates can be solubilized in physiological saline, i.e. TBS or TBS containing EDTA. The a-A1 preparation, or comparable preparations obtained in consecutive extraction steps with associative buffers, represented about 70% of the total tissue hexuronic acid, and an additional 20% were solubilized in 4 M GdnHCl. Dissociative extraction of the tumor with 4 M GdnHCl yielded 86% of the total hexuronic acid present. These values are in accordance with previously published results (Faltz et al., 1979a).
Electron microscopical examination of aggregates recovered from different associative extracts did in all cases reveal a structure of the densely packed central filament type identical to that of link-stabilized aggregates seen in earlier reconstitution experiments performed with dissociatively extracted proteoglycans from cartilage (Morgelin et al., 1988) or other tissue sources (Morgelin et al., 1989). These data, however, were obtained by the recombination of molecules which had been exposed to 4 M GdnHCl and high cesium chloride concentrations and centrifugal force during isolation and purification. It cannot be taken for granted that the structure of such reconstituted aggregates is in all respects representative of native aggregates. Our present results, however, confirm that this organization is representative for in situ conditions and that associatively extracted aggregates will be a suitable material for further detailed examinations of the central filament structure as well as of the molecular arrangement of hyaluronate binding region and link protein within the central filament.
The present studies led to the detection of apparently different central filament structures in aggregates. The more densely packed aggregates had the same central filament structure as known from reconstitution experiments, with the proteoglycan monomers being tightly packed in a continuous protein shell extending over long distances. The other novel structural type exhibited rather loosely packed central filament stretches with only short patches of clustered or even single monomers bound to the hyaluronate. Several experiments were designed in order to determine whether the loosely packed central filaments were derived from the densely packed ones by partial destruction of the continuous central filament structure during isolation and purification steps or electron microscopic specimen preparation. For this purpose aggregates were visualized in the electron microscope with low shear methods after conventional or mild extraction and purification procedures, and after chemical cross-linking in solution. In fact we could demonstrate that normal glycerol spraying/rotary shadowing destroyed a large portion of the aggregates as compared to mica sandwich squeezing or mica centrifugation. Nevertheless, in all cases densely and loosely packed aggregates with similar central filament structures became apparent in the electron microscope, independently of the purification and specimen preparation procedures used.
Interestingly, different fractions of densely and loosely packed central filament structure were solublized in various extraction buffers as shown in Table I. EDTA-containing physiological saline solubilized aggregates with the highest proportion of continuous central filament structure and with the largest fraction of G3 containing monomers. Taken together with earlier work demonstrating a sequence homology (Doege et al., 1986(Doege et al., , 1987Sai et al., 1986;Oldberg et al., 1987) and possibly a functional similarity (Halberg et al., 1988) of the COOH-terminal globular domain G3 to a Ca2+-dependent group of animal lectins, this result might indicate that intact proteoglycans are able to interact in a Ca2+-dependent way with a still unknown ligand in the matrix. It is likely, in the light of the large proportion of G3 retained, that this EDTAsoluble pool represents recently synthesized aggregates that have not yet been subject to proteolytic degradation.
When chondrosarcoma tissue residues, after consecutive extraction with different associative solvents (TBS, TBS-EDTA, 0.5 M GdnHCl), were treated further with 4 M GdnHCl, only aggregates with a rather degraded appearance were recovered from extracts after reconstitution. Relatively large amounts of hyaluronate binding region fragment bound to hyaluronate were observed. Faltz et al. (1979a) have examined proteoglycan aggregates from 4 M GdnHCl chondrosarcoma tissue extracts. They found ( a ) large amounts of hyaluronate binding region fragments, seen in SDS-polyacrylamide gel electrophoresis of A1D4 preparations, compared with material from a-AlD4 samples, and ( b ) size distributions of proteoglycan monomers after Kleinschmidt spreading of A1 preparations which were skewed toward shorter lengths than those spread from a-A1 samples. These and our results may indicate proteolytic degradation during extraction steps, or otherwise the proteoglycans extractable only by 4 M GdnHCl may represent the oldest and, during normal tissue turnover, most degraded aggregate fraction, which may be strongly bound in the tissue.
Rat chondrosarcoma proteoglycan monomers, digested with chondroitinase ABC, exhibited lengths of the extended domain E2 which were somewhat shorter (210 +-24 nm) than those measured from bovine nasal cartilage (263 f 27 nm). The differences were even more pronounced in whole monomers, namely 282 f 21 nm uersus 405 f 37 nm for particles from chondrosarcoma and nasal cartilage, respectively. These findings probably indicate a lower degree of glycosylation of the rat proteoglycan, leading to less stretching of its core protein by charge repulsion than observed for the nasal cartilage proteoglycan. On the other hand the differences in the core lengths obtained after chondroitinase digestion agree well with the fact that the extended domain E2 from the rat proteoglycan lacks a proline-rich stretch consisting of 23 consecutive repeats of 6 amino acids, as demonstrated by cDNA sequencing (Antonsson et al., 1989). Similar results have been reported by Doege et al. (1990), when rat and human proteoglycan core protein cDNA sequences were compared. These authors showed a comparably longer sequence in man caused by insertions in the E2 domain which are not present in the rat sequence. Some of the experiments indicate that substoichiometric amounts of link protein are not the cause for the observed loosely packed central filament structures. When link protein was added to a-A1 aggregate preparations even large molar excesses of link protein did not close the gaps between adjacent short central filament patches. In contrast, control experiments clearly showed that link-free aggregates, in which individual monomers were bound statistically to the hyaluronate, exhibited the well-known continuous densely packed central filament structure after addition of link protein. In previous studies strong evidence was provided that equimolar amounts of hyaluronate binding region and link protein are present in the Swarm rat chondrosarcoma (Oegema et al., 1977;Faltz et al., 1979b). We assume that this is also the case in our preparations, and in future work we will examine in more detail the molar ratios of the different components in the central filament.