Domain structure of the basement membrane heparan sulfate proteoglycan

We have used proteolytic digestions and immunological reactivity to map regional domains of the 400-kilodalton (kDa) core protein of the heparan sulfate containing basement membrane proteoglycan from the Englebreth-Holm-Swarm tumor. Digestion with V8 protease caused the rapid release of numerous large peptides ranging in size from 80 to 200 kDa and a 44-kDa peptide. The 44-kDa peptide (P44) was stable to further digestion, but the larger peptides were eventually degraded to a 46-kDa peptide (P46). Both the P44 and P46 fragments migrate slower in the presence of a reducing agent, indicating intrachain disulfide bonding, and do not have heparan sulfate side chains. Antisera to the P46 fragment, however, did not react with P44 fragment, and the amino acid compositions of P46 and P44 fragments were different. This suggests that these two fragments were unrelated. Trypsin digestion of the proteoglycan immediately released a 200-kDa peptide (P200) that also lacked heparan sulfate side chains. Digestion of the P200 fragment with V8 protease produced the P44 and P46 fragments in the same temporal sequence seen with V8 protease digestion of the proteoglycan. Antisera to the P200 fragment reacted strongly with the P44 and P46 fragments. These results show that the P44 and P46 domains are contained within the P200 domain. The rapid release of the P44 domain indicates that it is located at one end of the core protein. The large size of these proteolytic fragments suggests the core protein contains considerable conformational structure, and the absence of heparan sulfate on the P200 domain indicates that the side chains are asymmetrically located on the core.     

Reassembly of Golgi stacks from mitotic Golgi fragments in a cell-free system

Rat liver Golgi stacks were incubated with mitotic cytosol for 30 min at 37 degrees C to generate mitotic Golgi fragments comprising vesicles, tubules, and cisternal remnants. These were isolated and further incubated with rat liver cytosol for 60 min. The earliest intermediate observed by electron microscopy was a single, curved cisterna with tubular networks fused to the cisternal rims. Elongation of this cisterna was accompanied by stacking and further growth at the cisternal rims. Stacks also fused laterally so that the typical end product was a highly curved stack of 2-3 cisternae mostly enclosing an electron-lucent space. Reassembly occurred in the presence of nocodazole or cytochalasin B but not at 4 degrees C or in the absence of energy supplied in the form of ATP and GTP. Pretreatment of the mitotic fragments and cytosol with N-ethylmaleimide (NEM) also prevented reassembly. GTP gamma S and A1F prevented reassembly when added during fragmentation but not when added to the reassembly mixture. In fact, GTP gamma S stimulated reassembly such that all cisternae were stacked at the end of the incubation and comprised 40% of the total membrane. In contrast, microcystin inhibited stacking so that only single cisternae accumulated. Together these results provide a detailed picture of the reassembly process and open up the study of the architecture of the Golgi apparatus to a combined morphological and biochemical analysis.     

Terminal short arm domains of basement membrane laminin are critical for its self-assembly

Laminin self-assembles into large polymers by a cooperative two-step calcium-dependent mechanism (Yurchenco, P. D., E. C. Tsilibary, A. S. Charonis, and H. Furthmayr. 1985. J. Biol. Chem. 260:7636-7644). The domain specificity of this process was investigated using defined proteolytically generated fragments corresponding to the NH2-terminal globule and adjacent stem of the short arm of the B1 chain (E4), a complex of the two short arms of the A and B2 chains attached to the proximal stem of a third short arm (E1'), a similar complex lacking the globular domains (P1'), and the distal half of the long arm attached to the adjacent portion of the large globule (E8). Polymerization, followed by an increase of turbidity at 360 nm in neutral isotonic TBS containing CaCl2 at 35 degrees C, was quantitatively inhibited in a concentration-dependent manner with laminin fragments E4 and E1' but not with fragments E8 and P1'. Affinity retardation chromatography was used for further characterization of the binding of laminin domains. The migration of fragment E4, but not of fragments E8 and P1', was retarded in a temperature- and calcium-dependent fashion on a laminin affinity column but not on a similar BSA column. These data are evidence that laminin fragments E4 and E1' possess essential terminal binding domains for the self-aggregation of laminin, while fragments E8 and P1' do not. Furthermore, the individual domain-specific interactions that contribute to assembly are calcium dependent and of low affinity.     

Functional studies of the domains of talin

The protein talin has two domains of approximately 200 and 47 kD, which can be cleaved apart by a variety of proteases. To examine the function of these two structural domains of talin, we have digested purified talin with a calcium-dependent protease and separated the resulting fragments chromatographically. Both fragments were radioiodinated and used to probe Western blots of whole fibroblasts and chicken gizzard extracts. The large talin fragment bound to vinculin and metavinculin. The small fragment did not demonstrate any binding in this assay. The fragments were labeled fluorescently and microinjected into fibroblasts in tissue culture. The large talin fragment incorporated quickly into focal adhesions where it remained stable for at least 14 h. The small fragment associated with focal adhesions of fibroblasts but was also distributed diffusely in the cytoplasm and the nucleus. These experiments suggest that talin has at least two sites that contribute to its localization in focal adhesions. Intact talin microinjected into Madin-Darby bovine kidney epithelial cells localized to the focal adhesions but was excluded from the zonulae adherentes, despite the localization of vinculin to both of these sites. In contrast, the large talin fragment, when microinjected into these epithelial cells, incorporated into both focal adhesions and zonulae adherentes. The difference in localization between the large talin fragment and intact talin seems to be due to the removal of the small domain. This difference in localization suggests that talin binding sites in zonulae adherentes have limited accessibility.     

Characterization of porcine plasma fibronectin and its fragmentation by porcine liver cathepsin B

Fibronectin was isolated from porcine plasma by affinity chromatography with gelatin-linked Sepharose 4B. Porcine fibronectin had a chemical composition similar to those of human and other fibronectins and reacted with antiserum raised against human fibronectin. It showed hemagglutination activity with trypsin-treated rabbit erythrocytes, though the activity was far less than that of human fibronectin. Porcine plasma fibronectin consisted of two subunit chains of about 230,000-daltons linked by disulfide bonds(s). Limited proteolysis of this protein with porcine liver cathepsin B yielded five major fragments which were investigated by affinity chromatography with gelatin- and heparin-linked Sepharose 4B. One fragment (Mr = 50,000) was bound to gelatin but not to heparin, while the remaining four were bound to heparin but not to gelatin, suggesting that plasma fibronectin takes a discrete domain structure with respect to interaction with these two macromolecules. The three larger heparin-binding fragments, Mr = 175,000, 150,000, and 130,000 were eluted with different concentrations of a mixture of NaCl and urea from the heparin-column, suggesting that they have different interactions with heparin, the 130,000-dalton fragment being the one with the strongest interaction. After reduction with 2-mercaptoethanol, the 175,000-dalton fragment was converted to the 150,000-dalton region fragment, which, together with the unchanged 150,000-dalton fragment, appeared to be equivalent in amount to the 130,000-dalton fragment. This finding suggests that the 150,000- and 130,000-dalton fragments may have originated from different subunit chains. Since the 175,000-dalton fragment was not produced by cathepsin B digestion of fibronectin which had been treated with plasmin, it was concluded that the 175,000-dalton fragment contained interchain disulfide bond(s) which had linked the native subunit chains. These results suggest that porcine plasma fibronectin has non-identical subunit chains composed of domains which differ in interaction with heparin and in susceptibility to cathepsin B.     

Proteolysis of the bifunctional methionine-repressible aspartokinase II-homoserine dehydrogenase II of Escherichia coli K12. Production of an active homoserine dehydrogenase fragment.

The dimeric bifunctional enzyme aspartokinase II-homoserine dehydrogenase II (Mr = 2 X 88,000) of Escherichia coli K12 can be cleaved into two nonoverlapping fragments by limited proteolysis with subtilisin. These two fragments can be separated under nondenaturing conditions as dimeric species, which indicates that each fragment has retained some of the association areas involved in the conformation of the native protein. The smaller fragment (Mr = 2 X 24,000) is devoid of aspartokinase and homoserine dehydrogenase activity. The larger fragment (Mr = 2 X 37,000) is endowed with full homoserine dehydrogenase activity. These results show that the polypeptide chains of the native enzyme are organized in two different domains, that both domains participate in building up the native dimeric structure, and that one of these domains only is responsible for homoserine dehydrogenase activity. A model of aspartokinase II-homoserine dehydrogenase II is proposed, which accounts for the present results.     

Identification of disulfide-bridged substructures within human von Willebrand factor

In the course of identifying substructural domains within the homooligomeric protein von Willebrand factor [270 kilodaltons (kDa) per polypeptide chain], seven large fragments of 8-90 kDa have been generated by limited proteolysis. A monomeric fragment that binds coagulation factor VIIIc is identified as residues 1-272. A fragment that binds platelet glycoprotein Ib is identified as a homodimer containing two pairs of identical chains, i.e., residues 273-511 and 674-728. Disulfide bonds have been identified by several methods, including direct observation of the phenylthiohydantoin of cystine during Edman degradation of isolated peptides. Among half-cystine residues in the amino-terminal 1365-residue region, 52 have been paired. They place structural constraints on folding possibilities within three structural domains. Additional clusters of disulfide bonds are evident. It has been shown that at least 35 disulfides must form intrachain bridges, specifically the cystines among residues 1-272 and 906-1492. Intersubunit disulfide bonds are partially localized in an interior region (residues 283-695) and a carboxyl-terminal region (residues 1908-2050). Each of these regions appears to be linked to a corresponding region of a neighboring subunit in the network of interconnected chains. The difficulties of pairing all 169 half-cystines (per chain) and of distinguishing intrachain from interchain disulfides are evaluated.     

The crucial role of trimerization domains in collagen folding

Collagens contain large numbers of Gly-Xaa-Yaa peptide repeats that form the characteristic triple helix, where the individual chains fold into a polyproline II helix and three of these helices form a right-handed triple helix. For the proper folding of the triple helix collagens contain trimerization domains. These domains ensure a single starting point for triple helix formation and are also responsible for the chain selection in heterotrimeric collagens. Trimerization domains are non-collagenous domains of very different structures. The size of trimerization domains varies from 35 residues in type IX collagen to around 250 residues for the fibrillar collagens. These domains are not only crucial for biological functions, but they are also attractive tools for generating recombinant collagen fragments of interest as well as for general use in protein engineering and biomaterial design. Here we review the current knowledge of the structure and function of these trimerization domains.     

Crystal structure of a soluble form of the human T cell coreceptor CD8 at 2.6 A resolution.

A secreted fragment of the extracellular portion of human CD8 alpha has been expressed in CHO cells, and a deglycosylated and proteolyzed form of this fragment has been crystallized. We report here the crystal structure of this fragment as refined at 2.6 A resolution. The structure was solved by molecular replacement using a superposition of ten variable domains from immunoglobulin light chains as the search model. Only the N-terminal 114 amino acids of CD8 alpha are visible in the electron density maps. The domain formed by these residues possesses a fold typical of immunoglobulin variable domains and associates to form Fv-like homodimers.     

Chemical studies on the isolated collagen-like and globular fragment of complement component C1q. Comparative studies on bovine and human C1q

Both the collagen-like and the globular fragments of a subcomponent C1q of the first component of bovine and human complement were highly purified by enzymic digestion followed by gel filtration. Analyses by polyacrylamide gel electrophoresis showed that the former was composed of covalently linked peptide chains with an average molecular weight of 14 000, and that the latter was composed of three non-covalently linked peptide chains each having a molecular weight of approximately 15 000. Great similarities between amino acid compositions of the globular fragments and some similarities between those of the collagen-like fragments were found. Moreover, great similarities of amino acid compositions were found among three non-covalently linked chains of each globular fragment as well as between the corresponding chains of both globular fragments. These results suggested that both the collagen-like and the globular domains on the C1q molecule remained highly conserved in its evolution.     

Correlation between the exposure of aromatic chromophores at the surface of the Fc domains of immunoglobulin G and their ability to bind complement.

The recognition that certain biological effector functions associated with the Fc region of human IgG are mediated exclusively by either the Cgamma2 or Cgamma3 domains prompted a study of some of the physical properties of the isolated domains in an attempt to correlate these with functional differentiation. The degree of aromatic chromophore exposure of intact Fc and fragments corresponding to the Cgamma2 and Cgamma3 domains were determined by solvent perturbation difference spectroscopy using 20% ethylene glycol. For the monomeric Cgamma2 fragment one of the two tryptophans and all four of the tyrosines were exposed to solvent. In the pFc' fragment, which represented a dimer of two intact Cgamma3 domains, an average of 0.4 of the two tryptophans of 3.3 of the five tyrosines per chain were exposed. These data were consistent with the suggested involvement of tryptophan in complement fixation since Cgamma2 binds C1q but pFc' does not. Several fragments derived from the Cgamma3 region had previously been shown to have differing environments for their aromatic side chains from circular dichroism studies. These fragments have now been shown to exhibit different degrees of chromophore exposure to solvent. Removal of the carboxy-termimal heptapeptide from the intact, Cgamma3 domain resulted in a fragment not only showing a greater exposure of aromatic residues but also having the ability to bind Clq. Our data suggest that the structural requirements for C1Q binding may be quite commonplace within Fc, but tertiary folding limits their expression except in Cgamma2 in the native molecule. The solvent perturbation observed with Fc was somewhat lower than would have been expected from the results with the isolated domains, suggesting that interdomain interactions may result in burial of aromatic residues.     

Characterization of the precursor form of type VI collagen

Well characterized monospecific antisera against pepsin-extracted bovine type VI collagen were used to identify and characterize the intact form of type VI collagen. In immunoblotting experiments the antisera reacted with the pepsin-resistant fragments of the alpha 1(VI) and alpha 3(VI) chains, but not with the fragment of the alpha 2(VI) chain. Extracts obtained from uterus and aorta with 6 M guanidine HCl contained two immunoreactive polypeptides of Mr = 190,000 and 180,000 based on globular protein standards. Cleavage of extracts with pepsin generated the previously characterized pepsin-resistant fragments of alpha 1(VI) and alpha 3(VI), indicating that the higher molecular weight polypeptides represent the intact parent chains, alpha 1(VI) and alpha 3(VI). Digestion of extracts with bacterial collagenase released an Mr = 100,000 noncollagenous fragment from the alpha 1(VI) chain. Thus, intact type VI collagen in tissues contains a relatively short triple helical domain and at least one very large globular domain which is sensitive to pepsin but resistant to collagenase digestion. Immunoblotting revealed a polypeptide of Mr = 240,000, which we suggest represents the pro-alpha 1(VI) chain, in the culture medium of bovine fibroblasts. Bands intermediate in molecular weight between 240,000 and 190,000 were identified in cell layers. These findings establish type VI collagen as a protein with very large nontriple helical domains, a property that undoubtedly plays an important role in its function.     

Helix-stabilized Fv (hsFv) antibody fragments: substituting the constant domains of a Fab fragment for a heterodimeric coiled-coil domain.

Antibody Fv fragments would in principle be useful for a variety of biotechnological applications because of their small size and the possibility to produce them in relatively large amounts in recombinant form; however, their limited stability is a drawback. To solve this problem, both domains are usually fused via a peptide linker to form a single-chain Fv (scFv) fragment, but in some cases this leads to a dimerization. We present an alternative format for stabilizing antibody Fv fragments. The C(H)1 and C(L) domain of the Fab fragment were replaced with a heterodimeric coiled coil (WinZip-A2B1), which had previously been selected using a protein-fragment complementation assay in Escherichia coli. This new antibody format was termed helix-stabilized Fv fragment (hsFv), and was compared to the corresponding Fv, Fab and single-chain Fv format. Bacterial growth and expression of the hsFv was significantly improved compared to the Fab fragment. The hsFv fragment formed a heterodimer of heavy and light chain with the expected molecular mass, also under conditions where the scFv fragment was predominantly dimeric. The hsFv fragment was significantly more stable than the Fv fragment, and nearly as stable as the scFv fragment under the conditions used (80 nM protein concentration). Thus, the format of a helix-stabilized Fv (hsFv) fragment can be a useful alternative to existing recombinant antibody formats, especially in cases where poor expression of Fab fragments or multimerization of scFv fragments is a problem.     

Interaction between complementary polymerization sites in the structural D and E domains of human fibrin.

Plasmic degradation products of human fibrin, fragments DD, D, and E, bind to fibrin. It has been inferred from this observation that the binding occurs by attraction of complementary sites located in the NH2- and COOH-terminal domains of the fibrin molecule. The interaction between fragments D1 and E1 has been investigated in this work since it represents the first step in the process of fibrin clot formation. Fragment D1, that was initially as active as fragment DD, lost most of its anticoagulant activity after purification by cation-exchange chromatography. The lability of fragment D1 function explained the previous unsuccessful attempts to form a complex between fragments D1 and E1. The loss of fragment D1 anticoagulant activity was not associated with the cleavage of the gamma 63-85 chain segment, since fragments D1A and D1 identically inhibited the fibrin monomer polymerization rate. In order to demonstrate the formation of a complex between fragments D1 and E1, three lines of experiments were advanced. First, the anticoagulant activity of fragment D1 was neutralized by fragment E1 in a dose-dependent manner, demonstrating that the association between these fragments involved polymerization sites. Second, two products, D1.E1 and D1.E1.D1, were stabilized in a reaction with bifunctional cross-linking reagents, proving the formation of D.E complexes in aqueous solution. Third, immobilized fragment D1 bound fragments E1 and E2, but not fragment E3, showing that fragments E1 and E2 attached via a polymerization site to the complementary one in fragment D1, since this association was disrupted by fibrin polymerization inhibitory peptide GPRP. These results provided direct evidence for specific binding between the structural D and E domains of fibrin mediated through complementary polymerization sites. Thus, the initial formation of fibrin clot fibers appears to be driven by specific association of these sites.     

Characterization of upFc, a fragment of human immunoglobulin G1 produced by pepsin in urea.

The digestion of human IgG1/K myeloma proteins with pepsin in the presence of 8 M-urea produces fragments that differ from those produced by aqueous peptic digestion, and from other characteristic immunoglobulin fragments. Fb'2, the larger urea/pepsin fragment, was previously shown to consist of the constant regions of the light chains, and the CH1 domains and hinge regions of the heavy chains. The smaller fragment, upFc, has now been characterized. After reduction, three peptides were released from fragment upFc. Amino acid sequencing, N- and C-terminal determinations and amino acid compositions have enabled these peptides to be identified as residues Ile-253 to Leu-306, residues Thr-307 to Asp-376 and residues Thr-411 to Gly-446 of the heavy chain. Fragment upFc therefore contains the entire Fc region, beginning at residue Ile-253, except for a 34-residue section from within the CH3-domain disulphide loop. Peptic digestion of IgG1/K proteins in 8M-urea therefore provides a method for isolating from gamma1 heavy chains five homogeneous peptides in good yield, which account for almost the entire constant region. Characterization of fragments Fb'2 and upFc has shown that the action of pepsin in urea is entirely different from that of aqueous pepsin. Two gamma1 heavy chains have been shown to differ in sequence at three positions from the sequence reported for protein Eu.     

Evidence for two extracellular domains in the human interleukin-2 receptor: localization of IL-2 binding.

The human interleukin-2 (IL-2) receptor was quantitatively cleaved into two large disulfide-bonded fragments by either trypsin or endoproteinase lys-C (endo lys-C). The smaller fragment contains both N-linked oligosaccharides found in the intact receptor and is derived from the amino terminus of the molecule. The larger proteolytic fragment was metabolically labeled with 32PO4 and represents the carboxy terminus. The predicted cleavage sites of both enzymes lie in the region of the molecule encoded by exon 3. This pattern of limited proteolysis provides biochemical evidence that the extracellular region of the receptor is organized into two domains. This supports a structural model of the receptor in which the regions of internal homology encoded by exons 2 and 4 form independent disulfide-bonded domains connected by a hydrophilic segment. To determine the role of these domains in IL-2 binding, [125I]IL-2 was chemically cross-linked to the proteolytically cleaved receptor on the cell surface. The 125I-labeled complex obtained displayed N-linked oligosaccharides and had an Mr consistent with one molecule of IL-2 cross-linked to the smaller proteolytic fragment of the receptor. Thus, the amino-terminal domain of the IL-2 receptor appears to form an integral part of the IL-2 binding site.     

Chymotryptic digestion of Tetrahymena ciliary dynein. II. Pathway of the degradation of 22S dynein heavy chains

As shown in the preceding paper (Toyoshima, Y. Y., 1987, J. Cell Biol., 105:887-895) three-headed Tetrahymena 22S dynein consists of three heavy chains (HCs) and is decomposed into two-headed (H) and one-headed (L) fragments by chymotryptic digestion. To accurately determine the presence of multiple ATPases and ultimately the location of various domains, it is necessary to determine the identity of each HC fragment relative to the original HCs in 22S dynein. The degradation pathway of each HC was determined by peptide mapping and immunoblotting. The three HCs (A alpha, A beta, and A gamma) were immunologically different; although SDS-urea gel electrophoresis showed that A gamma HC was apparently resistant to the digestion, actually three distinct HCs contributed to the same band alternately. H fragment was derived from A beta and A gamma HCs, whereas L fragment originated from A alpha HC. Since both fragments were associated with ATPase activity, these results directly demonstrate the presence of multiple ATPase sites in Tetrahymena 22S dynein.     

Isolation and characterization of fragments of ATP-dependent protease Lon from Escherichia coli obtained by limited proteolysis

Conditions of limited proteolysis of the protease Lon from Escherichia coli that provided the formation of fragments approximately corresponding to the enzyme domains were found for studying the domain functioning. A method of isolation of the domains was developed, and their functional characteristics were compared. The isolated proteolytic domain (LonP fragment) of the enzyme was shown to exhibit both peptidase and proteolytic activities; however, it cleaved large protein substrates at a significantly lower rate than the full-size protease Lon. On the other hand, the LonAP fragment, containing both the ATPase and the proteolytic domains, retained almost all of the enzymatic properties of the full-size protein. Both LonP and LonAP predominantly form dimers unlike the native protease Lon functioning as a tetramer. These results suggest that the N-terminal domain of protease Lon plays a considerable role in the process of the enzyme oligomerization.     

Isolation and characterization of a heparin-binding domain from the amino terminus of platelet thrombospondin

Calcium-replete thrombospondin has been purified from outdated platelets using heparin-Sepharose affinity chromatography, gelatin-Sepharose to remove fibronectin, and gel filtration to eliminate low-molecular-weight heparin-binding proteins. Edman degradation of six different preparations revealed the amino-terminal sequence of thrombospondin (TSP) to be Asn-Arg-Ile-Pro-Glu-Ser-Gly-Gly-Asp-Asn-Ser-Val-Phe-. This sequence was obtained in initial yields as high as 85%, indicating that no blocked chains are present. Cleavage of calcium-replete TSP with thermolysin or plasmin results in the production of relatively stable fragments. Chromatography of these digests on heparin-Sepharose followed by elution with 0.6 M NaCl affords purification of an Mr 25,000 fragment from the thermolysin digest and an Mr 35,000 fragment from the plasmin digest. The binding of these fragments to heparin-Sepharose does not require divalent metal ions. Neither fragment is disulfide-bonded to other fragments present in the digests. The heparin-binding domains from both digests have similar amino acid compositions and their tryptic peptide maps on high performance liquid chromatography are identical with the exception of one peptide unique to each fragment. Automated Edman degradation in a vapor-phase sequenator of the thermolytic heparin-binding domain electroeluted from sodium dodecyl sulfate-gels indicates that the heparin-binding domain resides at the amino terminus of the Mr 180,000 TSP peptide chain.     

Distribution of keratan sulfate in cartilage proteoglycans.

After chondroitinase digestion of bovine nasal and tracheal cartilage proteoglycans, subsequent treatment with trypsin or trypsin followed by chymotrypsin yielded two major types of polypeptide-glycosaminoglycan fragments which could be separated by Sepharose 6B chromatography. One fragment, located close to the hyaluronic acid-binding region of the protein core, had a high relative keratan sulfate content. This fragment contained about 60% of the total keratan sulfate, but less than 10% of the total chondroitin sulfate present in the original proteoglycan preparation. The weight average molecular weight of the keratan sulfate-enriched fragment was 122,000, as determined by sedimentation equilibrium centrifugation. The chemical and physical data indicate that this fragment contains an average of 10 to 15 keratan sulfate chains, if the average molecular weight of individual chains is assumed to be about 8,000, and about 5 chondroitin sulfate chains attached to a peptide of about 20,000 daltons. The other population of fragments was derived from the other end of the proteoglycan molecule, the chondroitin sulfate-enriched region, and contained mainly chondroitin sulfate chains. About 90% of the total chondroitin sulfate, but only 20 to 30% of the total keratan sulfate was recovered in these fragments. On the average, approximately 5 chondroitin sulfate chains and 1 keratan sulfate chain could be linked to the same peptide. Another 10 to 20% of the total keratan sulfate, originally found in or near the hyaluronic acid-binding region, was not separated from the chondroitin sulfate-enriched fragments. Hydroxylamine could be used to liberate a large molecular size, chondroitin sulfate-enriched fragment (Kav 0.54 on Sepharose 2B) from the proteoglycan aggregates. The remainder of the protein core, containing the keratan sulfate-enriched region, was bound to hyaluronic acid with the link proteins and recovered in the void volume on the Sepharose 2B column.