Versican V2 Is a Major Extracellular Matrix Component of the Mature Bovine Brain

We have isolated and characterized the proteoglycan isoforms of versican from bovine brain extracts. Our approach included (i) cDNA cloning and sequencing of the entire open reading frame encoding the bovine versican splice variants; (ii) preparation of antibodies against bovine versican using recombinant core protein fragments and synthetic peptides; (iii) isolation of versican isoforms by ammonium sulfate precipitation followed by anion exchange and hyaluronan affinity chromatography; and (iv) characterization by SDS-polyacrylamide gel electrophoresis and Coomassie Blue staining or immunoblotting. Our results demonstrate that versican V2 is, together with brevican, a major component of the mature brain extracellular matrix. Versicans V0 and V1 are only present in relatively small amounts. Versican V2 migrates after chondroitinase ABC digestion with an apparent molecular mass of about 400 kDa, whereas it barely enters a 4-15% polyacrylamide gel without the enzyme treatment. The 400-kDa product is recognized by antibodies against the glycosaminoglycan-alpha domain and against synthetic NH2- and COOH-terminal peptides. Our preparations contain no major proteolytic products of versican, e.g. hyaluronectin or glial hyaluronate-binding protein. Having biochemical quantities of versican V2 available will allow us to test its putative modulatory role in neuronal cell adhesion and axonal growth.

The term hyalectans (or lecticans) defines a family of large hyaluronan-binding proteoglycans (1, 2) whose members currently include versican (3), aggrecan (4), neurocan (5), and brevican (6). Hyalectans share highly similar domain structures at either end of the core protein, whereas the chondroitin sulfate-carrying middle portions are clearly distinct. The homologous NH 2 -terminal regions consist of an Ig loop and a tandem repeat element that interacts with hyaluronan (7)(8)(9). The COOH-terminal globular structure includes epidermal growth factor-like repeats, a C-type lectin domain, and a sushi (or complement regulatory protein) element. Recombinantly expressed C-type lectin domains of hyalectans interact specifically with tenascin-R (10) and tenascin-C (11). The binding is mediated by a calcium-dependent protein-protein interaction.
Alternative splicing and transcription termination add greatly to the structural diversity of hyalectans. Brevican, for instance, exists either as a secreted or glycosylphosphatidylinositol-anchored molecule as a result of alternative transcription termination (12,13). Variable usage of exons encoding the epidermal growth factor-like elements and the sushi domain lead to several aggrecan isoforms (14,15), and alternative splicing of human (16) and mouse (17,18) versican transcripts may generate up to four versican core protein variants. The differently spliced exons encode the central glycosaminoglycancarrying GAG 1 -␣ and GAG-␤ domains (19).
All known hyalectans are temporally expressed in the central nervous system (2,20,21). Neurocan and aggrecan appear transiently during early developmental stages (22,23) and are replaced later by brevican and versican (24 -26), the major postnatal hyalectans in brain tissues. Neurocan and brevican are restricted to the central nervous system, whereas aggrecan is expressed predominantly in cartilage. The longest versican splice variants (V 0 and V 1 ) display an even wider distribution and are present in a number of mesenchymal (27) and epithelial (28) tissues and in cultures of endothelial cells (18).
In vitro experiments suggest that hyalectans play a modulatory role in cell-cell and cell-matrix interactions and hence may participate in the nervous system in the control of axon growth and guidance. Inhibitory activity on neuronal cell adhesion and neurite outgrowth has been described for neurocan (29) and for brevican (24). In addition, aggrecan from nasal cartilage is a potent inhibitor of neurite outgrowth (30). Because brain and cartilage aggrecan differ largely in the extent of post-translational modifications (31), this finding still needs to be confirmed with the more relevant brain-derived aggrecan. Evidence that hyalectans may also be involved in vivo in the control of axonal growth and guidance is provided by our recent observation of the transient versican V 0 /V 1 expression in tissues that act as barriers during the development of the peripheral nervous system (32).
Although versicans V 0 and V 1 are also present in the central nervous system (27), there are indications that a third splice variant, versican V 2 , is the major isoform in adult brain. The assumption is based on results from reverse transcriptase-PCR (16) and Northern blot experiments (18) as well as on comparative immunohistochemical studies (26). In contrast to the other versican splice variants, versican V 2 transcripts are only detectable in the central nervous system (16). On the protein level, a versican-like proteoglycan (25) and several small mo-lecular size hyaluronan-binding proteins (33,34) have been identified in brain extracts. Partial peptide sequences suggest that these components are derived from versican V 1 by a proteolytic process (35). No biochemical data are currently available to confirm the presence of the versican V 2 proteoglycan in the central nervous system.
To close this gap, we have now determined the complete cDNA sequence of bovine versican, we have prepared several domain-specific antibodies via immunization with recombinant core protein fragments or synthetic peptides, and we have purified and characterized the different versican isoforms from bovine brain tissues.

EXPERIMENTAL PROCEDURES
Tissues-Bovine tissues were obtained from the local abattoir. The tissues were immediately placed on ice and processed within 1 h postmortem.
PCR Cloning and Sequencing-Total RNA was isolated from bovine forebrain using a RNeasy total RNA isolation kit (Qiagen, Hilden, Germany). 1 g of total RNA was reverse transcribed with a first-strand cDNA synthesis kit (Boehringer Mannheim, Germany) and random hexamer primers (total volume 20 l). Either 1 l of this reverse transcription reaction or 100 ng of bovine genomic DNA was used as the template in different PCRs. Primer sequences and cycling conditions are listed in Table I. Initial cDNA fragments were obtained by PCR with guessmer primers that were based on comparisons of chick (36), mouse (18), and human versican (3,16) cDNA sequences. The PCR products were either sequenced directly or cloned into a pGEM-T vector using an A/T cloning kit (Promega, Madison, WI). All sequences were determined in both directions on an Applied Biosystems 373 DNA Sequencer using dye terminator chemistry.
All antisera were purified by affinity chromatography. Antisera against GAG-␣ I , GAG-␣ II , and GAG-␤ were preabsorbed on bacterial control extract columns followed by purification on corresponding fusion protein columns (NHS-activated HiTrap matrix, Amersham Pharmacia Biotech). Antibodies against the NH 2 -and COOH-terminal ends of versican were affinity purified by binding to the corresponding peptides immobilized on UltraLink iodoacetyl columns (Pierce).
Isolation and Characterization of Versican from Bovine Brain-300 g of bovine forebrain was homogenized in a Waring blender using 4 volumes of extraction buffer containing 0.5 M NaCl, 50 mM Tris, pH 7.5, 25 mM EDTA, 0.5% Nonidet P-40, 1 mM 4-(2-aminoethyl)-benzenesulfonylfluoride hydrochloride (Pefabloc SC; Pentapharm, Basel, Switzerland), 1 g/ml pepstatin, and 1 g/ml leupeptin (buffer modified from Ref. 37). The homogenate was stirred for 5 h at 4°C and subsequently centrifuged at 100,000 ϫ g for 1 h. Proteins were differentially precipitated from the supernatant by the addition of solid ammonium sulfate. After an initial step at 30% saturation, versican-containing fractions were precipitated by increasing the ammonium sulfate concentration in the supernatant to 60%. All precipitates were allowed to form overnight at 4°C and were collected by centrifugation at 27,000 ϫ g for 45 min. The 60% precipitate was resuspended in buffer A (6 M urea, 50 mM Tris, 10 mM EDTA, pH 6) and batch absorbed to Q-Sepharose FF (Amersham Pharmacia Biotech) at 4°C overnight. After column packing, the resin was washed with buffer A containing 0.3 M NaCl, and elution was achieved by applying a NaCl gradient from 0.3 to 2 M. Versican-containing fractions, identified on a slot blot with anti-GAG-␣ II polyclonal antibodies, were pooled and dialyzed extensively against 0.5 M NaCl, 20 mM Tris, 10 mM EDTA, pH 8 (buffer B). Subsequently, hyaluronan-EAH-Sepharose, prepared by coupling hyaluronan (Sigma) to EAH-Sepharose (Amersham Pharmacia Biotech) through cross-linking with N-(3-dimethylaminopropyl)-NЈ-ethylcarbodiimide hydrochloride (Fluka, Buchs, Switzerland) (38), was added to the dialyzed fractions. The batch absorption was performed at 4°C for 20 h. After packing the resin into a column, a washing gradient from 0.5 to 2 M NaCl in buffer B was applied, and elution was achieved with 4 M guanidine-HCl, 20 mM Tris, 10 mM EDTA, pH 8. Versican-containing fractions (identified as described above) were pooled, concentrated in a Biomax-100 spin concentrator (Millipore, Bedford, MA), and dialyzed against 40 mM NaOAc, pH 8, 10 mM EDTA. 100-l aliquots were digested with 1 milliunit of carrier-free chondroitinase ABC (Seikagaku, Tokyo, Japan) in 40 mM NaOAc, 10 mM EDTA, 1 mM 4-(2-aminoethyl)-benzenesulfonylfluoride hydrochloride, pH 8 at 37°C for 7 h or overnight. Electrophoresis and Immunoblotting-Isolated fractions were analyzed by SDS-PAGE on 4 -15% gels (Bio-Rad) and Coomassie Blue stained with a colloidal blue staining kit (Novex, San Diego, CA).

RESULTS
cDNA Cloning of Bovine Versican-The starting points for the cloning of the entire bovine versican cDNA sequence were short cDNA stretches amplified from total bovine brain RNA with reverse transcriptase-PCR and guessmer primers. These guessmer primers were derived from versican cDNA sequences that were highly conserved among the human, mouse, and chicken homologs and lay close to the alternative splice junctions (Fig. 1, primer pairs a*/b* and c*/d*). In this way we obtained bovine cDNA fragments encoding parts of the hyaluronan binding region and the COOH-terminal domains. Based on these sequences, we generated primers (e/f) for a long reverse transcriptase-PCR, which yielded a 3.2-kb amplification product extending over the entire GAG-␣-encoding portion of versican V 2 and a short PCR fragment, which originated from the versican V 3 transcript. No amplification product derived from the V 1 splice variant was obtained in this reaction (expected size 5.5 kb). We therefore amplified two small versican V 1 cDNA fragments that crossed over the splice junctions into the GAG-␤-encoding sequence. For each of these two reverse transcriptase-PCRs we used one perfect match and one guessmer primer (primer pairs: e/k* and l*/f, respectively). Because the GAG-␤ domain is encoded by a single exon in the human and the mouse genome, we then tried to amplify the remaining bovine GAG-␤ sequence in a long PCR using genomic DNA as template. The 5.1-kb fragment we obtained corresponded to the expected size. Finally, the portions including the translation initiation or translation termination sequences were each amplified with a perfect match and a guessmer primer (g* or j*), coined from species-conserved 5Ј-and 3Ј-untranslated regions, respectively.
Using this strategy, we PCR cloned overlapping cDNA fragments that covered the entire coding region of bovine versican corresponding to a 3,361-amino acid core protein in bovine versican V 0 , 2,380 amino acids in V 1 , 1,623 amino acids in V 2 , and 642 amino acids in V 3 (all exclusive of the 20-amino acid secretory signal sequence). The cDNA and deduced amino acid sequences of the four splice variants of bovine versican are not shown in this paper but are available from the GenBank data base.
Expression of Versican in Adult Bovine Brain-From our PCR results we could conclude that mRNAs of all versican splice variants (V 0 -V 3 ) are present in bovine brain tissue. However, no quantitative assertion could be made based on these experiments. We therefore prepared Northern blots with probes that covered portions encoding the NH 2 -terminal region, the GAG-␣ domain, or the COOH-terminal portion of versican (Fig. 2). All three probes detected a single band about 6.5 kb long corresponding to the expected size of the V 2 splice variant. Because no signals were detected in the range predicted for the mRNAs of the V 0 , V 1 , and V 3 isoforms, we concluded that these transcripts remained below the detection limit of the Northern blot.
Identification of Versican Isoforms in Protein Extracts from Bovine Brain-Based on the cDNA sequence, we prepared polyclonal antibodies against different parts of the versican core protein (Fig. 3). Some of the antigens chosen for immunization are present in all four versican isoforms (NH 2 -and COOHterminal peptides), and others are only included in the splice variants V 0 and V 2 (fragments GAG-␣ I and GAG-␣ II ) or in V 0 and V 1 (fragment GAG-␤), respectively. In immunoblotting experiments with bovine brain extracts, the polyclonal antibodies specific for the GAG-␣ domain detected a predominant core protein band of approximately 400 kDa after chondroitinase ABC digestion (Fig. 4). This band was also weakly visible before the enzyme treatment, probably reflecting unprocessed versican core protein from the intracellular pool. An almost identical picture was obtained with antibodies directed against the NH 2 -terminal versican peptide, whereas antibodies specific for the GAG-␤ domain did not recognize the 400-kDa component. In contrast, GAG-␤-specific antibodies detected two larger core proteins, which migrated well above the 400-kDa marker protein. Because these two bands were barely visible on blots developed with the polyclonal antibodies against the NH 2terminal peptide we concluded that these two core proteins must be present in minute amounts and are only detectable as a result of the high immunoreactivity of the anti-GAG-␤ antibodies. Finally, a minor band migrating at around 80 kDa was recognized by the NH 2 -terminal peptide antibodies. The mobility of this band was unaffected by chondroitinase ABC treatment.
Isolation of GAG-containing Isoforms of Versican from Brain Extracts-Versicans were solubilized efficiently from brain tissues with a detergent-containing high salt buffer, confirming previous experiments by Yamagata and co-workers (37). Subsequent extraction of the insoluble fraction with 4 M guanidine-HCl yielded less than 5% additional material recognized by GAG-␣-and GAG-␤-specific antibodies (data not shown). More than 90% of the soluble immunoreactive material could be precipitated with ammonium sulfate at saturation levels between 30 and 60% (data not shown). Further purification of the versican proteoglycan isoforms could be achieved by anion exchange chromatography on Q-Sepharose (Fig. 5), in which the versican-containing fractions eluted in a broad peak between 0.7 and 1.2 M sodium chloride. The subsequent affinity chromatography step on a hyaluronan column demonstrated the high stability of the versican-hyaluronan interaction, which could be disrupted only under relatively harsh conditions (4 M guanidine-HCl).
To monitor the purification process, aliquots of versicancontaining fractions were analyzed with SDS-PAGE (Fig. 6). The 400-kDa core protein component observed by immunoblotting the crude brain extract became clearly visible in the Coomassie Blue-stained gels after anion exchange chromatography and was highly enriched after affinity purification on the hyaluronan column. Without chondroitinase ABC treatment, this component barely entered the gel, indicative of its proteoglycan character. At least three other GAG-carrying components were present after the Q-Sepharose chromatography step. Their core proteins either migrated on SDS-PAGE as a double band around 150 kDa or as single band slightly above the 84-kDa protein marker. None of these smaller size bands displayed FIG. 3. Localization of the fragments and peptides used for raising polyclonal antibodies. The splice variant-specific fragments ␣ I and ␣ II (both found in V 0 and V 2 ), and ␤ (present in V 0 and V 1 ) were overexpressed as histidine-tagged fusion proteins in a bacterial expression system. The two peptides N and C contain the 15 most NH 2 -and COOH-terminal amino acids, respectively, which are present in all four splice forms. HABR, hyaluronan binding region. E/L/S, epidermal growth factor-like domains/C-type lectin domains/sushi elements.

FIG. 4. Immunoblot analysis of bovine brain extract.
A protein extract from bovine brain was separated on a 4 -15% SDS-Phastgel under reducing conditions before (Ϫ) and after (ϩ) chondroitinase (Ch'ase) ABC digestion. Blots were incubated with polyclonal antibodies against the NH 2 -terminal peptide (N) and against the fusion proteins GAG-␣ (␣ II ) and GAG-␤.
FIG. 5. Purification of versican isoforms with anion exchange and hyaluronan affinity chromatography. Anion exchange chromatography was performed under dissociative conditions using a Q-Sepharose FF column (upper). Fractions that were positive on a slot blot using the anti-GAG-␣ II antibody were pooled (solid bar) and subjected to affinity chromatography on a hyaluronan column. During washing only every second fraction was monitored (white gaps indicate fractions that were not tested). Anti-GAG-␣ II -positive fractions were pooled (solid bar) and processed for SDS-PAGE analysis.
immunoreactivity with anti-versican antibodies (data not shown), and they mostly disappeared after hyaluronan affinity chromatography. The final proteoglycan preparation usually contained 50 -100 g of protein/100 g of brain tissue (wet weight).
Characterization of the Isolated Versican Core Proteins-The proteoglycan fraction obtained by hyaluronan affinity chromatography was characterized further by immunoblot analysis using the entire panel of domain-specific polyclonal antibodies (Fig. 7). Compared with the immunoblotting analysis of the crude brain extracts (Fig. 4), no major changes in the relative amounts of the large molecular mass core proteins were observed. Only the quantitatively minor 80-kDa component of the crude brain extract had been lost during our purification procedure. Again, the two large core proteins migrating around 500 kDa were recognized by the GAG-␤-specific polyclonal antibodies and displayed very weak immunoreactivity with antibodies against the NH 2 -and the COOH-terminal ends of versican. Because the upper band was also apparent on both GAG-␣-specific immunoblots, we could conclude that these two components corresponded to the intact V 0 and V 1 core proteins of bovine versican, respectively. Both of these large versican splice variants had formerly evaded detection in the less sensitive Coomassie Blue staining (Fig. 6). In contrast to these quantitatively minor splice variants, the predominant 400-kDa core protein band displayed no immunoreactivity with antibodies against the GAG-␤ domain. It was, however, recognized by both anti-GAG-␣ antibodies and by antibodies directed against the NH 2 -and COOH-terminal peptides. We therefore could conclude that the predominant band in our brain extract preparation corresponded to the complete versican V 2 core protein.
The intact versican V 2 carries several chondroitin sulfate side chains preventing it from entering 4 -15% Phastgels without chondroitinase ABC digestion. Although the size of intact versican V 2 cannot be assessed from the electrophoretic mobility, it seems likely that the molecular mass exceeds 600 kDa.

DISCUSSION
In our earlier immunohistochemical study of human brain tissues (26) we observed a more abundant and widely spread staining with GAG-␣-specific antibodies than with antibodies recognizing the GAG-␤ domain. Yet these antibodies did not allow a distinction between the versican isoforms V 0 and V 2 , or V 0 and V 1 , respectively. We therefore could only assume on the basis of the differential staining patterns that the V 2 isoform might be the predominant versican splice variant in mature brain tissues. To prove this hypothesis, we initiated experiments isolating hyalectans from brain extracts. Because au-topsy samples of human brain displayed various degrees of autolysis, we used bovine brain as the tissue source. Unfortunately, our previously prepared polyclonal antibodies against human (16,28) and chick versican (32) showed only weak or no cross-reactivity with the bovine homolog. In consequence, a more elaborate approach including cDNA cloning and preparation of recombinant core protein fragments had to be used to prepare highly reactive antibodies specific for bovine versican.
By comparing the cDNA-deduced primary structures of different species homologs of versican (Table II), we observed a very high amino acid sequence conservation in the NH 2 -terminal and particularly in the COOH-terminal portion of the core protein, whereas a significant drop in sequence similarity was noted in the chondroitin sulfate-carrying GAG-␣ and GAG-␤ domains. This is not surprising because both highly conserved globular ends of the core protein are involved in protein-mediated interactions with other extracellular ligands. In vitro, the NH 2 -terminal region of versican binds to hyaluronan (8) and possibly also to link protein (39). The recombinantly expressed C-type lectin domain, which is localized in the COOH-terminal globule of intact versican, interacts with the fibronectin type III repeats of tenascin-R (10). In the chondroitin sulfate-carrying middle portion of the versican core proteins, we observed high sequence conservation in the vicinity of serine-glycine pairs, which are likely to be recognized by ␤-D-xylosyltransferase (40). Furthermore, the highly acidic amino acid sequence present in a potential cystine loop at the NH 2 -terminal end of the human GAG-␤ domain could also be identified in bovine versican. The function of this core protein stretch is still unknown.
All versican isoforms are expressed in bovine brain. However, the majority of the versican splice variants, namely the mRNAs encoding versican V 0 , V 1 , or V 3 , are only picked up by the very sensitive reverse transcriptase-PCR analysis. It seems therefore likely that these splice forms are only present at very low levels in mature brain tissues. The sole versican mRNA detected in the Northern blot experiment corresponds to the V 2 isoform. Human versican splice variants appear usually as double bands on the Northern blot (16) because of multiple polyadenylation signals in the human versican gene (19). That the bovine versican V 2 transcript appears as a single band may indicate either that the bovine gene only contains one polyadenylation signal or that they are localized only a few bases apart. From our reverse transcriptase-PCR and Northern blot experiments we did not obtain evidence for the occurrence of an additional "plus" domain that is present in chick versican (41), nor was there any indication for the existence of a membraneassociated versican splice variant similar to the glycosylphosphatidylinositol-anchored form of brevican (12).
Because the globular domains of hyalectans are highly similar, the distinct sequences of the GAG-carrying middle portions are often used to generate specific antibodies (8,16,28). Unfortunately, these antibodies recognize only some of the splice variants, but none is able to detect versican V 3 . We therefore looked for highly variable sequence portions at either end of the versican core protein and prepared anti-peptide antibodies. These antibodies recognize all four versican isoforms and allow us to test whether the core proteins are intact. Because a number of reports have described the isolation of small versican fragments named glial hyaluronate-binding protein (34) or hyaluronectin (33), the issue of proteolytic degradation has gained particular interest. Aware of this fact, we used for the extraction a detergent containing high salt buffer at neutral pH (37) supplemented with a mixture of protease inhibitors including the water-stable serine protease inhibitor 4-(2-aminoethyl)-benzenesulfonylfluoride hydrochloride. As a result, we detected with our set of antibodies mainly the intact core proteins of versican V 2 , V 1 , and V 0 on immunoblots of bovine brain extracts. The only lower molecular mass component we observed with antibodies against the NH 2 -terminal end of versican migrated in the range of 80 kDa. This is significantly larger than the 60 kDa found for glial hyaluronatebinding protein, which is possibly derived from versican V 1 by proteolytic cleavage with metalloproteinases (35). In fact, it seems more likely that our 80-kDa component corresponds to the V 3 isoform of versican because its size is in good agreement with the calculated molecular mass of about 73 kDa for this splice variant. Unfortunately, from all of our polyclonal antibodies against bovine versican the antibodies directed against the COOH-terminal core peptide displayed by far the lowest immunoreactivity, only weakly recognizing the most abundant versican isoforms in the crude brain extracts. Hence, further experiments are needed to prove finally that the 80-kDa polypeptide is indeed versican V 3 .
The most abundant splice variant of mature brain tissues is versican V 2 , as demonstrated by Northern and immunoblotting experiments as well as by analyzing the isolated versican proteoglycan isoforms in SDS-PAGE followed by Coomassie Blue staining. Versican V 2 is expressed exclusively in the central nervous system (16), whereas most other tissues synthesize the larger V 0 and V 1 splice variants. On SDS-PAGE, the core proteins of versican V 2 , V 1 , and V 0 migrate after chondroitinase ABC digestion with a significantly higher relative size than expected from the calculated molecular masses of 180, 260, and 368 kDa. This discrepancy may result in part from a high proportion of N-and O-linked oligosaccharides covalently attached to the core protein and to poor SDS binding because of the highly acidic primary structures (calculated pI values around 4).
At present we do not know how the 400-kDa versican V 2 core protein relates to versican isolated earlier from human brain tissue (25). This proteoglycan has a significantly smaller core protein of 345 kDa and seems to carry only a few short chondroitin sulfate side chains with a total size of 20 kDa. However, because one of the peptide sequences deduced from the core protein of this proteoglycan matches a sequence within the GAG-␤ domain, it seems likely that this proteoglycan is derived from the V 1 splice variant through a proteolytic process. Other only partly characterized proteoglycans with core proteins in the size range of versican V 2 are the CAT-301 antigen (42) and the DSD-1-proteoglycan (43). Whether these proteoglycans are identical or only similar to versican V 2 will be the subject of further experiments.
Apart from versican V 2 mature bovine brain tissues express another major proteoglycan that belongs to the hyalectan family. Brevican (6), the smallest currently known proteoglycan with hyaluronan binding capacity, exists in two alternatively spliced versions (12), one secreted and one membrane-associated isoform. In our proteoglycan preparation we identified core proteins in the range of about 150 and 85 kDa, matching the reported sizes of full-length brevican and its COOH-terminal proteolytic product. After anion exchange chromatography similar amounts of versican V 2 and brevican were present in Coomassie Blue-stained gels. In contrast, only trace amounts of brevican could be detected after hyaluronan affinity chromatography, which included stringent wash steps with high salt buffers. This observation may indicate that brevican binds to hyaluronan with a significantly lower affinity than versican.
The functional role that versican V 2 and brevican play in mature brain tissues is currently unknown. Hockfield et al. (44) hypothesized that changes in the extracellular matrix composition during brain maturation lead to a stabilization of the synaptic structures. This could in part be achieved by the strengthening of adhesive mechanisms and by inhibiting further neurite growth. Although the loss of synaptic plasticity will certainly not only depend on changes in hyalectan expression, it seems conceivable that matrices rich in versican V 2 and/or brevican act as inhibitors of cell interactions and therefore reduce the formation of new synapses in the mature central nervous system. Such an inhibitory activity of brevican has been demonstrated recently by Yamada et al. (24) in a neurite outgrowth assay, whereas functional studies with versican V 2 are just beginning. Now having biochemical quantities of versican V 2 available will allow us to explore potential roles of this brain-specific splice variant in in vitro and in vivo model systems.