Interaction of a brain extracellular matrix protein with hyaluronic acid.

A glial hyaluronate-binding protein (GHAP) was isolated from bovine spinal cord and partially characterized. Bovine GHAP consisted of three immunologically related polypeptides with molecular masses of 76, 64, and 54 kDa and isoelectric points of 4.1, 4.2, and 4.4, respectively. Peptide mapping and partial amino acid sequencing showed that all three polypeptides derive from the same protein. The protein was localized immunohistochemically with rabbit antisera in the white matter surrounding the myelinated axons. Sugar analyses indicated that the three polypeptides are glycosylated and the sugar residues account for at least 30% of their weight. After enzymatic deglycosylation, the apparent molecular mass of the bovine GHAP was reduced to 43 kDa. The biochemical properties of bovine GHAP were compared to those of human GHAP. Initial peptide mapping indicated similarities between bovine and human GHAP. Partial amino acid sequencing of bovine GHAP showed a striking identity (up to 90%) with human GHAP and with the hyaluronate binding domain of the large human fibroblast proteoglycan, versican. Bovine and human GHAP were demonstrated to bind specifically to hyaluronic acid (HA) with one protein molecule binding to an average 17 disaccharide repeating units. The binding of bovine and human GHAP was inhibited by oligosaccharides of HA and specifically by the octamer. Salt concentrations of up to 1 M NaCl had very little effect on the binding of the GHAP to HA. The GHAP-HA interaction was pH dependent. Dissociation only took place at low pH (less than 3.5). Analysis of several polypeptides derived from GHAP by limited proteolysis allowed us to conclude that one of the tandem repeated sequences is sufficient for HA binding and that the aminoterminal domain (which contains an immunoglobulin-like fold) is not involved in the GHAP-HA-binding event.     

Isolation and partial characterization of a glial hyaluronate-binding protein

A glial hyaluronate-binding protein (GHAP) with an isoelectric point of 4.3-4.4 was isolated from human brain white matter. The 60-kDa glycoprotein appeared to be quite resistant to proteolysis, and comparison with GHAP from a viable glioma removed at surgery showed that the protein isolated from autopsy material was not a degradation product resulting from postmortem autolysis. The protein was localized immunohistochemically with mouse monoclonal and rabbit polyclonal antibodies in cerebral white matter. Only small amounts could be found in the gray matter. After enzymatic deglycosylation, an immunoreactive 47-kDa polypeptide was obtained. Two amino acid sequences of GHAP showed a striking similarity (up to 89%) with a highly conserved region of cartilage proteins (bovine nasal cartilage proteoglycan and rat and chicken link protein). However, the amino acid composition and other amino acid sequences suggested that there are also differences between brain-specific GHAP and cartilage proteins.     

Hyaluronate binding properties of versican.

We have previously cloned a large chondroitin sulfate proteoglycan (versican) from human fibroblasts. The primary sequence shows that the N terminus contains sequence homology with known hyaluronate-binding molecule, suggesting that versican can bind hyaluronate. To test this hypothesis we have reconstructed a full-length versican cDNA and a versican cDNA fragment encoding the N terminus and have transfected Chinese hamster ovary cells and mouse 3T3 fibroblasts, respectively, with these constructs. The transfected Chinese hamster ovary cells make a proteoglycan shown to be versican by enzymatic and immunologic analysis. No corresponding proteoglycan was seen in the control cells. Using hyaluronate affinity chromatography, we show that recombinant versican specifically binds hyaluronate and does not bind to heparin or chondroitin sulfate. The transfected fibroblasts make a 78-kDa truncated form of versican that also binds hyaluronate and does not bind the related polysaccharides, showing that the hyaluronate binding activity resides at the N terminus of versican. The binding of versican to hyaluronate is substrate-concentration dependent and time dependent and can be competed with unlabeled versican. The dissociation constant for versican binding to hyaluronate was determined to be 4 x 10(-9) M.     

Interactions of cartilage proteoglycans with hyaluronate. Inhibition of the interaction by modified oligomers of hyaluronate.

Oligomers of hyaluronic acid were prepared by digestion of hyaluronic acid from rooster combs with testicular hyaluronidase (hyaluronate 4-glycanohydrolase, EC 3.2.1.35), leech head hyaluronidase (hyaluronate 3-glycanohydrolase, EC 3.2.1.36), and with fungal hyaluronidase (hyaluronate lyase from Streptomyces hyalurolyticus). The oligomers were fractionated by gel permeation, using Sephadex G-50. Oligomers isolated after incubation of the hyaluronic acid with the testicular hyaluronidase were further modified. To prepare oligomers with N-acetylglucosamine at both ends, terminal nonreducing glucuronic acid residues were removed with beta-glucuronidase. Reducing terminal N-acetylglucosamine residues were removed by reaction under mildly alkaline conditions. The reducing terminal N-acetylglucosamine residues were also reduced with sodium borohydride to form N-acetylglucosaminitol. The potentials of the various oligosaccharides to bind to the proteoglycan from bovine nasal septum cartilage were estimated by determining their effectiveness as inhibitors of the proteoglycan-hyaluronate interaction. The present study shows that, to bind maximally to the proteoglycan, the hyaluronate oligosaccharide must be at least 10 sugar residues in length and be terminated at the nonreducing and reducing ends with a glucuronate residue and an N-acetylglucosamine residue, respectively. Sugar residues extended beyond this basic decasaccharide, do not interact with the hyaluronate binding site on the proteoglycan.     

The ATPase core of a clathrin uncoating protein

Chymotryptic digestion of bovine brain uncoating ATPase produced a 60-kDa fragment that was subsequently proteolyzed to 44 kDa. Loss of clathrin cage uncoating activity paralleled the conversion of the intact 70-kDa enzyme to the 60-kDa fragment, while clathrin binding activity was lost as the 60-kDa fragment was degraded to 44 kDa. This 44-kDa fragment has been purified to homogeneity and characterized as a clathrin-independent ATPase. The 44-kDa ATPase domain has been localized within the intact enzyme by the use of amino-terminal specific antibodies. This localization relates to the conserved nature of the 70-kDa heat shock protein family, of which bovine brain uncoating ATPase is a constitutively expressed member.     

Hyaluronate binding and CD44 expression in human glioblastoma cells and astrocytes.

CD44 is an integral membrane glycoprotein of approximately 90 kDa which has been implicated in the binding of hyaluronate to the cell surface. The expression of CD44 in astrocytes was investigated by means of indirect immunofluorescence on cultured cells. The vast majority of these cells were found to express CD44. Western blot analysis of these cells revealed a highly polydisperse species having an M(r) corresponding to 74-86 kDa. In order to visualize hyaluronate-binding cells, living cultures were probed with fluorescein-conjugated hyaluronate (FI-HA). Some astrocytes were able to bind FI-HA, provided that they were first treated with hyaluronidase. Streptomyces hyaluronidase, which is hyaluronate-specific, was effective in exposing the hyaluronate-binding capacity of these cells. This leads one to conclude that hyaluronate is bound to the surface of these cells and that it masks their capacity to bind hyaluronate. Provided that they were first treated with hyaluronidase, the U-87 MG (glioblastoma-astrocytoma), U-373 MG (glioblastoma), and Hs 683 (glioma) cell lines were also able to bind FI-HA. The U-138 MG (glioblastoma) cell line was unable to bind FI-HA, with or without prior hyaluronidase treatment. A quantitative assay was developed with the use of [3H]hyaluronate ([3H]HA). This revealed the binding to be highly specific, inasmuch as the addition of unlabeled hyaluronate, but not other glycosaminoglycans, was effective in inhibiting the binding of the [3H]HA. An anti-CD44 monoclonal antibody, 50B4, was able to inhibit the binding of the [3H]HA to the U-373 MG cell line. In this cell line, then, CD44 functions as a hyaluronate receptor and one may infer that this is also the case in some astrocytes.     

Isolation and characterization of developmentally regulated chondroitin sulfate and chondroitin/keratan sulfate proteoglycans of brain identified with monoclonal antibodies

A panel of monoclonal antibodies prepared to the chondroitin sulfate proteoglycans of rat brain was used for their immunocytochemical localization and isolation of individual proteoglycan species by immunoaffinity chromatography. One of these proteoglycans (designated 1D1) consists of a major component with an average molecular size of 300 kDa in 7-day brain, containing a 245-kDa core glycoprotein and an average of three 22-kDa chondroitin sulfate chains. A 1D1 proteoglycan of approximately 180 kDa with a 150-kDa core glycoprotein is also present at 7 days, and by 2-3 weeks postnatal this becomes the major species, containing a single 32-kDa chondroitin 4-sulfate chain. The concentration of 1D1 decreases during development, from 20% of the total chondroitin sulfate proteoglycan protein (0.1 mg/g brain) at 7 days postnatal to 6% in adult brain. A 45-kDa protein which is recognized by the 8A4 monoclonal antibody to rat chondrosarcoma link protein copurifies with the 1D1 proteoglycan, which aggregates to a significant extent with hyaluronic acid. A chondroitin/keratan sulfate proteoglycan (designated 3H1) with a size of approximately 500 kDa was isolated from rat brain using monoclonal antibodies to the keratan sulfate chains. The core glycoprotein obtained after treatment of the 3H1 proteoglycan with chondroitinase ABC and endo-beta-galactosidase decreases in size from approximately 360 kDa at 7 days to approximately 280 kDa in adult brain. In 7-day brain, the proteoglycan contains three to five 25-kDa chondroitin 4-sulfate chains and three to six 8.4-kDa keratan sulfate chains, whereas the adult brain proteoglycan contains two to four chondroitin 4-sulfate chains and eight to nine keratan sulfate chains, with an average size of 10 kDa. The concentration of 3H1 increases during development from 3% of the total soluble proteoglycan protein at 7 days to 11% in adult brain, and there is a developmental decrease in the branching and/or sulfation of the keratan sulfate chains. A third monoclonal antibody (3F8) was used to isolate a approximately 500-kDa chondroitin sulfate proteoglycan comprising a 400-kDa core glycoprotein and an average of four 28-kDa chondroitin sulfate chains. In the 1D1 and 3F8 proteoglycans of 7-day brain, 20 and 33%, respectively, of the chondroitin sulfate is 6-sulfated, whereas chondroitin 4-sulfate accounts for greater than 96% of the glycosaminoglycan chains in the adult brain proteoglycans.(ABSTRACT TRUNCATED AT 400 WORDS)     

Link protein interactions with hyaluronate and proteoglycans. Characterization of two distinct domains in bovine cartilage link proteins

Hyaluronic acid-binding region and trypsin-link protein were prepared from bovine nasal cartilage proteoglycan complex after trypsin digestion. Binary complexes were reformed between trypsin-link protein and hyaluronic acid-binding region or hyaluronate. Upon trypsin treatment of these complexes, two fragments deriving from trypsin-link protein were characterized. One of them, of 20 kDa, corresponds in fact to a 140-amino acid long fragment and bears the glycosylated site of trypsin-link protein; it appears to be involved in proteoglycan/link protein interaction. The other, of 22 kDa, corresponds to the 200 C-terminal amino acids of trypsin-link protein; it appears to be involved in the binding of link protein with hyaluronic acid. A structural model of bovine trypsin-like protein depicting two distinct domains involved in hyaluronate and proteoglycan subunit interactions is proposed.     

Purification and characterization of adult diarrhea rotavirus: identification of viral structural proteins.

Adult diarrhea rotavirus (ADRV) is a newly identified strain of noncultivable human group B rotavirus that has been epidemic in the People's Republic of China since 1982. We have used sodium dodecyl sulfate-polyacrylamide gel electrophoresis and Western (immuno-) blot analysis to examine the viral proteins present in the outer and inner capsids of ADRV and compared these with the proteins of a group A rotavirus, SA11. EDTA treatment of double-shelled virions removed the outer capsid and resulted in the loss of three polypeptides of 64, 61, and 41, kilodaltons (kDa). Endo-beta-N-acetylglucosaminidase H digestion of double-shelled virions identified the 41-kDa polypeptide as a glycoprotein. CaCl2 treatment of single-shelled particles removed the inner capsid and resulted in the loss of one polypeptide with a molecular mass of 47 kDa. The remaining core particle had two major structural proteins of 136 and 113 kDa. All of the proteins visualized on sodium dodecyl sulfate-polyacrylamide gel electrophoresis were antigenic by Western blot analysis when probed with convalescent-phase human and animal antisera. A 47-kDa polypeptide was most abundant and was strongly immunoreactive with human sera, animal sera raised against ADRV and against other group B animal rotaviruses (infectious diarrhea of infant rat virus, bovine and porcine group B rotavirus, and bovine enteric syncytial virus) and a monoclonal antibody prepared against infectious diarrhea of infant rat virus. This 47-kDa inner capsid polypeptide contains a common group B antigen and is similar to the VP6 of the group A rotaviruses. Human convalescent-phase sera also responded to a 41-kDa polypeptide of the outer capsid that seems similar to the VP7 of group A rotavirus. Other polypeptides have been given tentative designations on the basis of similarities to the control preparation of SA11, including a 136-kDa polypeptide designated VP1, a 113-kDa polypeptide designated VP2, 64- and 61-kDa polypeptides designated VP5 and VP5a, and several proteins in the 110- to 72-kDa range that may be VP3, VP4, or related proteins. The lack of cross-reactivity on Western blots between antisera to group A versus group B rotaviruses confirmed that these viruses are antigenically quite distinct.     

Mapping of ATP-dependent trypsin-sensitive sites on the beta chain of outer-arm dynein from sea urchin sperm flagella.

The beta chain of sea urchin outer-arm dynein showed a peculiar tryptic digestion pattern in the presence of ATP (or ADP) plus Vi. Examination of the molecular mass of the products formed by photocleavage of tryptic fragments indicated that the trypsin-sensitive sites on the 165-kDa ATP-binding polypeptide in the presence of ATP and Vi are located 15 kDa apart from its amino-terminus, 2 kDa apart from its carboxy-terminus, and near the middle portion between the adenine- and gamma-Pi-binding sites. On the other hand, the carboxy-terminal region of the beta chain, the 135-kDa polypeptide, was cleaved into a 96-kDa polypeptide by tryptic digestion in the presence of ATP and Vi. Peptide mapping of 135-kDa, 96-kDa, and carboxy-terminally truncated polypeptides of the 135-kDa polypeptide revealed that the 96-kDa region is located at the amino-terminal portion of the 135-kDa region. These results indicate that the changes of trypsin susceptibility of dynien beta chain caused by binding ADP and Vi occur not in local region but over an extensive region on the beta chain.     

Formation of hyaluronan- and versican-rich pericellular matrix by prostate cancer cells promotes cell motility

Previous studies have demonstrated that high levels of hyaluronan (HA) and the chondroitin sulfate proteoglycan, versican in the peritumoral stroma are associated with metastatic spread of clinical prostate cancer. In vitro integration of HA and versican into a pericellular sheath is a prerequisite for proliferation and migration of vascular smooth muscle cells. In this study, a particle exclusion assay was used to determine whether human prostate cancer cell lines are capable of assembling a pericellular sheath following treatment with versican-containing medium and whether formation of a pericellular sheath modulated cell motility. PC3 and DU145, but not LNCaP cells formed prominent polarized pericellular sheaths following treatment with prostate fibroblast-conditioned medium. The capacity to assemble a pericellular sheath correlated with the ability to express membranous HA receptor, CD44. HA and versican histochemical staining were observed surrounding PC3 and DU145 cells following treatment with prostatic fibroblast-conditioned medium. The dependence on HA for integrity of the pericellular sheath was demonstrated by its removal following treatment with hyaluronidase. Purified versican or conditioned medium from Chinese hamster ovary K1 cells overexpressing versican V1, but not conditioned medium from parental cells, promoted pericellular sheath formation and motility of PC3 cells. Using time lapse microscopy, motile PC3 cells treated with versican but not non-motile cells exhibited a polar pericellular sheath. Polar pericellular sheath was particularly evident at the trailing edge but was excluded from the leading edge of PC3 cells. These studies indicate that prostate cancer cells recruit stromal components to remodel their pericellular environment and promote their motility.     

A comparison of isometallothionein synthesis in rat liver after partial hepatectomy and parenteral zinc injection.

Proteoglycan aggregates free of non-aggregating proteoglycan have been prepared from the annuli fibrosi and nuclei pulposi of intervertebral discs of three human lumbar spines by extraction with 4M-guanidinium chloride, associative density gradient centrifugation, and chromatography on Sepharose CL-2B. The aggregate (A1-2B.V0) was subjected to dissociative density-gradient ultracentrifugation. Three proteins of Mr 38 900, 44 200 and 50 100 found in the fraction of low buoyant density (A1-2B.V0-D4) reacted with antibodies to link protein from newborn human articular cartilage. After reduction with mercaptoethanol, two proteins of Mr 43 000 and two of Mr 20 000 and 14 000 were seen. The A1-2B.V0-D4 fraction, labelled with 125I, coeluted with both hyaluronate and a hyaluronate oligosaccharide (HA14) on a Sepharose CL-2B column. HA10 and HA14 reduced the viscosity of A1 fractions; HA4, HA6 and HA8 did not. HA14 decreased the viscosity of disc proteoglycans less than it did that of bovine cartilage proteoglycans. Thus, although a link protein was present in human intervertebral disc, it stabilized proteoglycan aggregates less well than did the link protein from bovine nasal cartilage.     

Distribution in cesium chloride gradients of proteoglycans of chick embryo brain and characterization of a large aggregating proteoglycan

Proteoglycans were extracted from 14-day chick embryo brains, which had been labelled in vitro with [35S]sulfate or 3H-labelled amino acids. 4.0 M guanidinium chloride (containing proteinase inhibitors) extracted 94% of the 35S-labelled glycoconjugates. Following cesium chloride equilibrium centrifugation, the proteoglycans in each fraction were characterized by chromatography on Sepharose CL-2B. The most dense fraction (D1), which contained no detectable non-proteoglycan proteins, contained a large, aggregating chondroitin sulfate proteoglycan in addition to small chondroitin sulfate and heparan sulfate proteoglycans. The less dense fractions (D2-D6) contained both small chondroitin sulfate and heparan sulfate proteoglycans. Removal of hyaluronate from the D1 sample by digestion with Streptomyces hyaluronidase in the presence of proteinase inhibitors showed that aggregation of the large chondroitin sulfate proteoglycan is hyaluronate-dependent. Aggregation was restored by re-addition of hyaluronate. Reduction and alkylation, which blocked aggregation of a cartilage A1 proteoglycan, did not interfere with aggregation of the large brain proteoglycan.     

A Glycoprotein Expressed by Human Fibrous Astrocytes is a Hyaluronate-Binding Protein and a Member of the CD44 Family

We have isolated and characterized an antigen from normal human brain called p80, so called because it migrated with an Mr of 80 kDa on SDS PAGE. The Mr of 80 kDa consists of a protein of about 55-60 kDa and carbohydrate (20-25 kDa). The carbohydrate is almost entirely of the N-linked type, although a small amount of O-linked carbohydrate was detected. Cross-reactivity with monoclonal antibodies A3D8 and A1G3 showed that p80 could therefore be considered an isoform of the CD44 adhesion molecules. In addition, specific binding to hyaluronate which was not competed for by proteoglycan demonstrated that it involved different sites than the proteoglycan binding sites. We also observed that fucoidan and dextran sulphate increased the binding by 200-250% while chondroitin sulphate C also increased the binding but to a lesser extent. Heparin, heparan sulphate and chondroitin sulphates A and B did not have such an effect. The binding of p80 to hyaluronate was pH dependent with a maximum at pH 6.4. We concluded that p80 was an astrocyte specific adhesion molecule.     

Studies on the hyaluronate binding properties of newly synthesized proteoglycans purified from articular chondrocyte cultures

Primary cultures of rabbit articular chondrocytes have been maintained for 10 days and labeled with [35S]sulfate, [3H]leucine, and [35S]cysteine in pulse-chase protocols to study the structure and hyaluronate binding properties of newly synthesized proteoglycan monomers. Radiolabeled monomers were purified from medium and cell-layer fractions by dissociative CsCl gradient centrifugation with bovine carrier monomer, and analyzed for hyaluronate binding affinity on Sepharose CL-2B in 0.5 M Na acetate, 0.1% Triton X-100, pH 6.8. Detergent was necessary to prevent self-association of newly synthesized monomers during chromatography. Monomers secreted during a 30-min pulse labeling with [35S]sulfate had a low affinity relative to carrier. Those molecules released into the medium during the first 12 h of chase (about 40% of the total) remained in the low affinity form whereas those retained by the cell layer rapidly acquired high affinity. In cultures where more than 90% of the preformed cell-layer proteoglycan was removed by hyaluronidase digestion before radiolabeling the newly synthesized low affinity monomers also rapidly acquired high affinity if retained in the cell layer. Cultures labeled with amino acid precursors were used to establish the purity of monomer preparations and to isolate core proteins for study. Leucine- or cysteine-labeled core proteins derived from either low or high affinity monomer preparations migrated as a single major species on sodium dodecyl sulfate-polyacrylamide gel electrophoresis with electrophoretic mobility very similar to that of core protein derived from extracted proteoglycan monomer. Purified low affinity monomers were converted to the high affinity form by treatment at pH 8.6; however, this change was prevented by guanidinium-HCl at concentrations above 0.8 M. Conversion to high affinity was also achieved by incubation of monomers in aggregate with hyaluronic acid (HA) at pH 6.8 followed by dissociative reisolation of monomer. At both pH 6.8 and 8.6 the conversion process was slow, requiring up to 48 h for the maximum increase in affinity. It is suggested that the slow increase in HA binding affinity seen during extracellular processing of proteoglycans in cartilage and chondrocyte cultures is the result of an irreversible structural change in the HA binding domain following the binding of monomer to hyaluronate. The available evidence suggests that this change involves the formation or rearrangement of disulfide bonds.     

Isolation and characterization of a membrane glycoprotein from human brain with sequence similarities to cell adhesion proteins from chicken and mouse

We previously described the production of monoclonal antibodies against a preparation of membrane glycoproteins from human brain [Berglund et al. (1987) J. Neurochem. 48, 809-815]. One of the glycoproteins, recognized by monoclonal antibody CF3, was specifically expressed in the brain. We now report the isolation and characterization of this glycoprotein, called glycoprotein 135 (Gp135). Gp135 was purified by means of lentil lectin affinity chromatography and immunoaffinity chromatography, using monoclonal antibody CF3, from a crude membrane extract of human brain cortex. Gp135 was shown to consist of a glycosylated single polypeptide chain with an apparent molecular mass of 135 kDa. The size of the polypeptide moiety was estimated to 115 kDa following N-glycanase digestion. The glycoprotein is anchored in the membrane by a glycosylphosphatidylinositol tail, as shown by phospholipase C digestion and liposome incorporation experiments. Amino acid sequence analysis of the amino terminal, and of an internal peptide obtained by V8 protease digestion of the glycoprotein, revealed a strong similarity to three previously described glycoproteins from chicken (contactin and F11) and mouse (F3) brains. These glycoproteins belong to the immunoglobulin superfamily and are implicated in cell adhesion phenomena in the developing brain. Gp135 may be the human counterpart to one or several of these glycoproteins.     

Inhibition of Hyaluronan Synthesis Reduces Versican and Fibronectin Levels in Trabecular Meshwork Cells

Hyaluronan (HA) is a major component of the extracellular matrix (ECM) and is synthesized by three HA synthases (HAS). Similarities between the HAS2 knockout mouse and the hdf mutant mouse, which has a mutation in the versican gene, suggest that HA and versican expression may be linked. In this study, the relationship between HA synthesis and levels of versican, fibronectin and several other ECM components in trabecular meshwork cells from the anterior segment of the eye was investigated. HA synthesis was inhibited using 4-methylumbelliferone (4MU), or reduced by RNAi silencing of each individual HAS gene. Quantitative RT-PCR and immunoblotting demonstrated a reduction in mRNA and protein levels of versican and fibronectin. Hyaluronidase treatment also reduced versican and fibronectin levels. These effects could not be reversed by addition of excess glucose or glucosamine or exogenous HA to the culture medium. CD44, tenascin C and fibrillin-1 mRNA levels were reduced by 4MU treatment, but SPARC and CSPG6 mRNA levels were unaffected. Immunostaining of trabecular meshwork tissue after exposure to 4MU showed an altered localization pattern of HA-binding protein, versican and fibronectin. Reduction of versican by RNAi silencing did not affect HA concentration as assessed by ELISA. Together, these data imply that HA concentration affects synthesis of certain ECM components. Since precise regulation of the trabecular meshwork ECM composition and organization is required to maintain the aqueous humor outflow resistance and intraocular pressure homeostasis in the eye, coordinated coupling of HA levels and several of its ECM binding partners should facilitate this process.     

Immunohistochemical evaluation of versican,in relation to chondroitin sulphate,in canine mammary tumours

The expression of increased amounts of versican, a chondroitin sulphate proteoglycan, in neoplastic tissues may play a role in promoting tumour cell proliferation and migration. This study investigated the immunolocalization of versican in normal and neoplastic canine mammary tissues, using antibodies 12C5 and 2B1, against different epitopes of the protein core of versican. Antibody CS56, recognising chondroitin sulphate (CS), was used to investigate the relation between versican and CS, which accumulates in canine mammary tumours. We found enhanced versican expression in both benign and malignant tumours, appearing in three main patterns: in periductal tissues, probably in association with basement membranes of ducts; in peripheral invasive areas of malignant tumours; and in spindle cell proliferations and myxoid areas of complex and mixed tumours. The 12C5 and 2B1 immunoreactivities co-localised in all types of tumours, and could be improved by chondroitinase digestion. The only exception was the abundant extracellular matrix (ECM) of spindle cell proliferations, particularly in myxoid areas of complex and mixed tumours, which displayed intense and diffuse 12C5 immunoreactivity and patchy or absent 2B1 and CS56 immunoreactivities; versican immunoreactivity could not be enhanced by chondroitinase digestion. The results indicate that versican is one of the extracellular matrix components characteristic of canine mammary tumours. It appears likely that in complex and mixed tumours versican exists in at least two forms, one of them lacking the CS attachment domain and the 2B1 epitope. Furthermore, the enhanced versican expression in the invasive areas of malignant tumours indicates the involvement of this proteoglycan in tumour cell invasion.