Brevican, a chondroitin sulfate proteoglycan of rat brain, occurs as secreted and cell surface glycosylphosphatidylinositol-anchored isoforms.

cDNA clones encoding proteins related to the aggrecan/versican family of proteoglycan core proteins have been isolated with antisera against rat brain synaptic junctions. Two sets of overlapping cDNAs have been characterized that differ in their 3′-terminal regions. Northern analyses with probes derived from unique regions of each set were found to hybridize with two brain-specific transcripts of 3.3 and 3.6 kilobases (kb). The 3.6-kb transcript encodes a polypeptide that exhibits 82% sequence identity with bovine brevican and is thought to be the rat ortholog of brevican. Interestingly, the polypeptide deduced from the open reading frame of the 3.3-kb transcript is truncated just carboxyl-terminal of the central domain of brevican and instead contains a putative glypiation signal. Antibodies raised against a bacterially expressed glutathione S-transferase-brevican fusion protein have been used to show that both soluble and membrane-bound brevican isoforms exist. Treatment of the crude membrane fraction and purified synaptic plasma membranes with phosphatidylinositol-specific phospholipase C revealed that isoforms of brevican are indeed glycosylphosphatidylinositol-anchored to the plasma membrane. Moreover, digestions with chondroitinase ABC have indicated that rat brevican, like its bovine ortholog, is a conditional chondroitin sulfate proteoglycan. Immunohistochemical studies have shown that brevican is widely distributed in the brain and is localized extracellularly. During postnatal development, amounts of both soluble and phosphatidylinositol-specific phospholipase C-sensitive isoforms increase, suggesting a role for brevican in the terminally differentiating and the adult nervous system.

Morphogenesis and differentiation as well as functional plasticity in the brain involve a variety of different interactions between neurons and their environment. The extracellular matrix (ECM), 1 a complex agglomerate of glycoproteins and pro-teoglycans located in the extracellular space, is prominently involved in these interactions. More than 20 different proteoglycan core proteins have been reported to occur in the developing and adult rat brain (1), most of them bearing either chondroitin sulfate or heparan sulfate as glycosaminglycan (GAG) moieties. A variety of different genes encodes brain proteoglycan core proteins that can occur as secreted molecules, such as CAT 301 (2) or the T1 antigen (3), as integral membrane proteins, e.g. NG2 (4) or N-syndecan (5), or as membrane-anchored forms, such as glypican (6,7) or cerebroglycan (8). The diversity of proteoglycan core proteins is further amplified by alternative processing at the post-transcriptional and post-translational levels (for review, see Ref. 9).
One of the most intensely studied families of proteoglycans is the aggrecan/versican family. Members of this family include the cartilage proteoglycan aggrecan (10 -12), versican, a molecule originally identified in fibroblasts (13) but also highly expressed in the central nervous system (14), as well as the brain-specific proteoglycans neurocan (15) and brevican (16). They share a number of structural features. The amino-terminal domains mediate the interaction with hyaluronic acid (HA) as well as with the link protein, a relatively small polypeptide involved in the aggregation of proteoglycans (17). The link protein consists basically of a HA binding domain. BEHAB, a protein deduced from its cDNA sequence, was suggested to constitute a brain-specific link protein (18,19).
The domain structure of the core proteins of aggrecan/versican family members is reflected at the genomic level. The genes for rat (20) and mouse aggrecan (21) as well as human versican (22) have a very similar exon/intron organization that resembles the arrangement of functional domains of the corresponding proteins.
All known members of the aggrecan/versican family are soluble chondroitin sulfate proteoglycans, whereas most of the membrane-spanning or glycosylphosphatidylinositol (GPI)-anchored proteoglycans carry heparan sulfate or keratan sulfate side chains (9). Here we report the identification and characterization of two sets of cDNAs encoding isoforms of a member of the aggrecan/versican family of proteoglycan core proteins expressed in the rat brain. One isoform constitutes most likely the rat homolog of bovine soluble brevican, whereas another group of isoforms represents the first example of GPI-anchored proteins of the aggrecan/versican family. GPI-anchored brevican isoforms are up-regulated late during postnatal development.

MATERIALS AND METHODS
Isolation and Sequencing of cDNA Clones-The cDNA clone 37c2 was isolated from a gt11 expression library constructed from rat brain poly(A) ϩ RNA screened with polyclonal antisera against synaptic junctional protein fractions from the rat brain (23,24). Radiolabeled 37c2 cDNA was used to screen a rat brain gt10 cDNA library (Clontech) for overlapping clones. Hybridization was carried out for 4 h at 65°C in Rapid-hyb buffer (Amersham Corp.). The cDNA inserts were subcloned into pBluescript vectors (Stratagene) and sequenced from both strands using the fluorescent dye dideoxy termination method in combination with an automated DNA sequencer (Applied Biosystems). Sequences were handled using the GCG sequence analysis package (Genetics Computer Group, Inc.).
Northern Analysis-Total RNA was prepared from liver, heart, muscle, C6 glioma cell line, cerebellum, cortex, hippocampus, and remaining brain regions (i.e. brain stem, striatum, thalamus, hypothalamus) according to Ref. 25. RNA was electrophoretically separated in formamide/formaldehyde-agarose gels (1%) and transfered onto Hybond-N nylon membranes (Amersham Corp.) following standard protocols (26). Filters were hybridized with ␣-32 P-labeled probes using the Rapid-hyb system (Amersham Corp.) and washed following the manufacturer's instructions. Hybridization signals were detected using Kodak X-Omat x-ray films or visualized by a Fujix BAS1000 Bio Imager.
Construction of Bacterial Fusion Proteins and Production of Antisera-The cDNA insert of subclone 37c2/12 spanning nucleotides 763-2076 of the rat brevican cDNA (compare Fig. 1) was cloned into the bacterial expression vectors pEX34c (27) and pGEX-3X (Pharmacia Biotech Inc.) to produce chimeric proteins fused to MS2 polymerase with an apparent molecular mass of 80 kDa and to glutathione Stransferase with an apparent molecular mass of 92 kDa, respectively. The glutathione S-transferase-37c2/12 fusion protein was affinity-purified on glutathione-Sepharose as described by the manufacturer (Pharmacia) and used to produce polyclonal antisera. Three different antisera, two raised in mice and one produced in a rabbit, were used in this study. Antisera were tested for specificity on MS2 polymerase-37c2/12 fusion proteins. Purified IgG fractions were obtained by protein A-Sepharose (Pharmacia) chromatography.
Isolation of Subcellular Protein Fractions and Western Blotting-Tissue samples were isolated from rat liver, heart, and brain using 10 mM HEPES buffer containing 0.64 M sucrose. Protease inhibitors to a final concentration of 0.5 mM EDTA, 0.5 mM EGTA, 1 g/ml leupeptin, 2 g/ml aprotinin, 1 mM benzamidine, 0.3 mM 4-(2-aminoethyl)-benzenesulfonylfluoride, HCl were added to prevent proteolytic degradation. Tissue fractionation was carried out essentially as described previously (28). Briefly, after homogenization cell debris and nuclei were removed by a 1000 ϫ g centrifugation. The supernatant was spun for 20 min at 12,000 ϫ g resulting in supernatant S2 and pellet P2. The soluble protein fraction is the 100,000 ϫ g supernatant of S2. Triton X-100 extracts of crude membranes were obtained as described in Ref. 3. Isolation of synaptic membranes was performed as described previously (28).
Separation of proteins by SDS-polyacrylamide gel electrophoresis on 5-20% gels under fully reducing conditions and transfer onto nitrocellulose were performed as described previously (24). Western blots were immunodeveloped by overnight incubation with primary antibody and processed employing the ECL detection system (Amersham Corp.). Chondroitinase ABC and PI-PLC-Treatment-Soluble protein and detergent-soluble and membrane fractions were incubated with chondroitinase ABC (Boehringer Mannheim) at 0.03 units/30 g of protein in 40 mM Tris-HCl, pH 8.0, 40 mM sodium acetate, 0.01% bovine serum albumin, for 2 h at 37°C.
Phospholipase treatment was carried out using 0.6 units of PI-PLC (Oxford GlycoSystems) and 0.03 units of chondroitinase ABC/30 g of membrane proteins as described by Ref. 29.
Immunohistochemistry-For immunohistochemical detection of brevican, 30-day-old rats were deeply anaesthetized with sodium pentobarbital (Nembutal; 40 mg/kg) and transcardially perfused with a brief saline rinse followed by 0.25-1 liter of 4% paraformaldehyde, 0.1% glutaraldehyde in sodium phosphate buffer, pH 7.2. After removal of the brain, horizontal vibratome sections (50 m) were cut and immersed in 50% methanol, 1% H 2 O 2 , and 5% bovine serum albumin in phosphate-buffered saline. Freely floating slices were then incubated with a 1:20 dilution of an affinity-purified fraction of rabbit anti-brevican antiserum for 2 days at room temperature. The biotin/streptavidin/ peroxidase system (Dianova) with ␣-chloronaphthol as a substrate was used to detect brevican immunoreactivity.
Expression of GPI-anchored Brevican in HEK293 Cells-The open reading frame for the mature polypeptide encoded by the 3.3-kb brevican transcript was fused to the BM40 signal peptide, which has been proven to act very efficiently in the production of extracellular proteins and protein fragments (30). The construct was introduced into the pRC/CMV vector (Invitrogen) and lipofected into 293 human embryonic kidney cells. Stably transfected cell lines were selected under G418, and resistant colonies were propagated. 2

RESULTS
Cloning and Identification of Brevican cDNAs-From a collection of cDNA clones isolated by expression screening with antisera against a rat brain synaptic protein preparation (23,24) one cDNA, clone 37c2, displayed distinct sequence similarity to cDNAs encoding proteins of the aggrecan/versican family of proteoglycan core proteins (10,11,13,15,16). The clone was used to isolate a set of 10 overlapping cDNAs. Two of these covered a long open reading frame and were analyzed in detail (Figs. 1 and 2A). As deduced from the nucleotide sequence the encoded protein has 883 amino acid residues and a calculated M r of 96,056. The translation initiation site was assigned to nucleotides 110 -112 encoding the first methionine downstream of an in-frame stop codon (nucleotides 2-4). As revealed 2 I. Arnold-Ammer, manuscript in preparation.

FIG. 1. Physical map and overlapping cDNA clones encoding brevican isoforms.
The structural features of brevican isoforms (boxed) are indicated with respect to their mRNAs. Numbering of amino acids is as in Fig. 2. Abbreviations are as follows: Ig, immunoglobulin; PTR, proteoglycan tandem repeat; CRP, complement regulatory protein. The extension of four fully analyzed cDNAs is shown. Clone 37c2/12 was used to produce MS2 polymerase and glutathione S-transferase fusion proteins of brevican. The sequence deviation of clone 37c2/10 from other brevican cDNAs is indicated by a boldface line. The extension of PCR probes derived from the unique regions of the 3.6-kb mRNA (probe 1) and the 3.3-kb mRNA (probe 2) for the detection of these transcripts on Northern blots (compare Fig. 3) is indicated. by sequence comparison with other members of the aggrecan/ versican protein family, the first 22 amino acids represent a potentially cleavable signal peptide. Thus, the putative mature protein consists of 861 amino acids and has a calculated M r of 93,871.
Various functional domains can be identified in the protein encoded by the 37c2 cDNAs based on sequence homology to known proteins (Fig. 1). These include an Ig-like fold (amino acids 12-135) and two proteoglycan tandem repeats (amino acids 136 -236 and 237-337). All three elements are supposed to act as HA binding region (for review see Ref. 14). Amino acid residues 338 -601 represent a region unique to the 37c2 protein. This central domain is relatively rich in glutamic acid residues, which account for 16.5% of all amino acid residues in this region. The carboxyl-terminal portion harbors a single EGF-like repeat (amino acids 602-635), a lectin-like domain (amino acids 636 -764), and a complement regulatory proteinlike domain (amino acids 765-829). Two potential N-glycosylation sites are located at amino acid positions Asn-107 and Asn-314 ( Fig. 2A). In total, 13 Ser-Gly or Gly-Ser dipeptides, potentially serving as GAG attachment sites, occur in the deduced amino acid sequence. Five of these dipeptides have at least one acidic amino acid in the vicinity of either side, which is thought to be essential for GAG attachment (13). Four of these five sites are located in the Glu-rich central domain where GAG attachment to members of the aggrecan/versican family of proteoglycans is thought to occur.
Among the members of the aggrecan/versican family, bovine brevican (16) is most closely related to the protein encoded by the 37c2 cDNAs. The overall sequence identity of the two proteins is 82% with the highest degree of identity located in the HA binding domain (90%) and the carboxyl-terminal homology region comprising the EGF-repeat, the lectin-like domain, and the complement regulatory protein-like domain (84%). The central domain is least conserved, both in length (274 amino acids for the 37c2 protein, 298 amino acids for bovine brevican) and sequence identity (72%), although in both proteins this region is relatively rich in glutamic acid residues. The positions of nine out of 13 Ser-Gly or Gly-Ser dipeptides, the two potential N-glycosylation sites, and all cysteine residues that determine the structure of functional domains are conserved between the 37c2 protein and brevican ( Fig. 2A). Thus, the two proteins are likely to be species homologs, and we refer to the new protein as rat brevican.
The nucleotide sequence of one of the isolated cDNAs, clone 37c2/10, differed from that of the other cDNAs starting from nucleotide 1982 and contains 553 additional nucleotides (Figs. 1 and 2B). A nested reverse transcriptase PCR assay confirmed that a corresponding transcript exists in total brain RNA and the cDNA was not artificially recombined during library construction (data not shown). As the nucleotide sequence around the point of sequence divergence (CAG/GTAATT, Fig. 2B) perfectly resembles a splice donor site (31), we wondered whether the clone represents simply an unspliced transcript. To test this hypothesis, Northern blot hybridization was performed using specific PCR amplification products derived from the unique region of each set of cDNAs. As shown in Fig. 3 (lanes 9  and 10), the hybridizing transcript (3.3 kb) is smaller than the major brevican transcript of 3.6 kb, indicating that the RNA represented by the 37c2/10 cDNA must be an alternatively processed mRNA species.
The common protein sequence encoded by both transcripts FIG. 2. Nucleotide sequence and deduced amino acid sequence of brevican isoforms. A, sequence of the cDNAs encoding the secreted rat brevican isoform. The putative signal peptide is underlined; potential N-glycosylation sites are indicated by an open triangle; Ser-Gly and Gly-Ser dipeptide sequences representing potential chondroitin sulfate attachment sites are double underlined with dashed lines; cysteine residues conserved between rat and bovine brevican (16) are indicated by dots. B, sequence of cDNA clone 37c2/10 spanning the unique 3Ј region that encodes the glypiation signal of GPI-anchored brevican. The beginning of the specific sequence at nucleotide 1982 is indicated. The potential splice donor site is dotted underlined. The Ser residue serving as potential site of GPI anchor addition is marked by a filled triangle. The nucleotide sequence of A is available from the GenBank/EMBL/ DDBJ data bases under the accession number X79881. The sequence of B has been submitted. ends exactly after the central domain (Fig. 1). The specific portion of the 3.3-kb transcript extends the reading frame by only 21 additional amino acids beyond the central domain (Fig.  2B). The extreme carboxyl terminus of the deduced protein is highly hydrophobic and resembles the signal for the attachment of GPI anchors (32). Indeed, digests with the enzyme PI-PLC confirmed that GPI-anchored isoforms of brevican exist (see below). Comparison of the GPI anchor signal with the sequences signaling the addition of GPI-anchors to other proteins suggest processing and GPI addition to brevican at Ser-600 ( Fig. 2B; Ref. 32). Thus, the putative mature GPI-anchored isoform consists of a protein moiety of 600 amino acids with a calculated M r of 64,405.
Tissue Distribution of Brevican Transcripts and Protein Isoforms-Northern blot analysis revealed a strongly hybridizing band of 3.6 kb and a weaker band of 3.3 kb that are detectable in various brain regions but not in heart, muscle, liver, kidney, testes, lung, thymus, and spleen ( Fig. 3 and data not shown). Brevican transcripts are also neither found in C6 glioma cells (Fig. 3, lane 8) nor in PC12 phaeochromocytoma cells (not shown). In all brain regions examined, 3.6-and 3.3-kb transcripts coexist.
Antisera were produced against a bacterially expressed glutathione S-transferase-brevican fusion protein comprising amino acids 194 -631 of rat brevican, which should recognize polypeptides encoded by both classes of transcripts (compare Fig. 1). Consistent with the Northern data, these antibodies detect brevican immunoreactivity on Western blots of brain protein preparations but not in protein extracts of liver nor of heart tissue (Fig. 4).
Characterization of Brevican Isoforms-Brain tissue of 30day-old rats was differentially fractionated to study the subcellular distribution of brevican isoforms. The vast majority of brevican immunoreactivity is present in the soluble fraction (100,000 ϫ g supernatant). On immunoblots, two major bands, one of 140 kDa and a broad band between 70 and 80 kDa, as well as diffuse material distributed over a wide molecular weight range, are recognized by brevican antibodies (Fig. 4,  lane 1). After treatment of the soluble fraction with chondroitinase ABC, the diffuse material disappears while the 140 and 70 -80 kDa bands are sharpened (Fig. 4, lane 2). These results are consistent with the data for bovine brevican where a 145-kDa core protein and an 80-kDa processing product have been described (16). In addition, two minor protein bands of about 110 and 60 kDa are detected by the antisera, the identity of which is yet unclear.
Additional brevican immunoreactivity is associated with the crude membrane fraction of brain extracts (Figs. 4 and 5). Detergent extracts of this fraction contain a diffuse high molecular weight smear and immunoreactive bands of 140, 125, and about 80 kDa (Fig. 4, lane 5). Again, the high molecular weight material is chondroitinase-sensitive (Fig. 5A, lanes 2  and 3).
Chondroitinases ABC and AC (not shown) completely eliminate the high molecular weight material, whereas incubation with heparitinase III does mot affect these high molecular weight forms, neither in the soluble fraction nor in membranes (data not shown), indicating that brevican bears principally chondroitin sulfate side chains.
The crude membrane fraction was digested with PI-PLC to prove the existence of GPI-anchored isoforms. As shown in Fig.  5A (lanes 4 and 5), most of the immunoreactivity is released from the membranes by this treatment. Digestion of the crude membrane fraction with GPI-specific phospholipase C produces similar results (not shown). Incubation of the membranes under identical conditions but without enzyme does not release brevican immunoreactivity into the supernatant. Thus, we conclude that GPI-anchored isoforms of brevican do occur. These isoforms are a 140-kDa protein that is similar in size as the major soluble form, a 125-kDa protein that does not occur in the soluble protein fraction, and an 80-kDa isoform that comigrates with the upper half of the soluble 70 -80-kDa material.
In order to clarify which GPI-anchored brevican isoforms are derived from the 3.3-kb transcript, human embryonic kidney cells 293 were transfected with the 3.3-kb cDNA. On Western blots of isolated crude membranes from untransfected 293 cells, no brevican immunoreactivity is detectable (Fig. 6, lane  1). Stable transfectants express high amounts of brevican. After chondroitinase treatment, three distinct isoforms with apparent molecular masses of 90, 125, and 140 kDa are observed (Fig. 6, lane 2). Crude membrane fractions of these transfectants were treated with PI-PLC to test whether GPI anchoring occurs. Whereas the 140-kDa isoform is not affected by the enzyme, the 90-and 125-kDa isoforms are at least in part released into the supernatant (Fig. 6, lanes 3 and 4). This demonstrates that correct glypiation of the 90-and the 125-kDa isoform can occur in 293 cells.
ined the rat brain synaptic membrane fraction for the presence of brevican immunoreactivity. As shown in Fig. 5B, PI-PLCreleasable brevican isoforms are present in this fraction. In addition to the 140-, 125-, and 80-kDa isoforms, a 150-kDa isoform is highly enriched in this membrane preparation. Whether the 150-kDa protein is a synapse-specific brevican isoform or co-purifies with synaptic membranes for other reasons, e.g. because of its biochemical characteristics, remains to be solved.
Expression of Brevican during Postnatal Development-The expression of many proteoglycans is subjected to massive changes during development (1). To study the developmental regulation of brevican isoforms, we have performed an immunoblot analysis of chondroitinase-digested soluble and PI-PLCreleased protein extracts from rat brains collected at postnatal days P1 through P64. The soluble 140-and 70 -80-kDa isoforms are present at birth, and their amounts steadily increase from days P4 to P64 (Fig. 7A). The complexity of GAG attachment also increases during this developmental period (not shown). The GPI-anchored isoforms are barely detectable around birth. A strong increase of these isoforms occurs after day P8 (Fig. 7B). In conclusion, brevican expression is generally up-regulated during terminal differentiation of the rat brain.
Immunohistochemical Localization of Brevican-We have used three different antisera against the glutathione S-transferase-37c2/12 fusion protein to study the distribution of brevican in the rat brain at postnatal day P30. The antisera cannot discriminate between the various brevican isoforms. Brevican immunoreactivity is found in all parts of the brain. A typical extracellular staining surrounding cell bodies is observed, for instance around hippocampal pyramidal cells (Fig. 8, A and B) or cerebellar granule cells and Purkinje cells (Fig. 8C). Primary dendrites of hippocampal pyramidal cells and Purkinje cells are also stained at their surfaces. In neuropil areas, the distribution of brevican immunoreactivity is diffuse. Staining appears not to be restricted to neurons. DISCUSSION This study describes the cloning of the complete coding sequence for two isoforms of a rat member of the aggrecan/ versican family of proteoglycan core proteins. One isoform has the features of ECM core proteins that are secreted into the extracellular space, the other one is likely to be attached to the cell membrane via a GPI anchor. Biochemical data demonstrate that both soluble and GPI-anchored isoforms of this proteoglycan exist. Immunohistochemical studies with antisera that recognize both types of isoforms revealed an extracellular localization throughout the brain, suggesting that the protein is indeed a component of the ECM.
The secreted isoform of the rat protein characterized here has 82% sequence identity with bovine brevican (16). Both proteins are of similar size. Therefore, we assume that the two proteins are orthologs. This assumption is supported by protein data; soluble bovine brevican occurs in two isoforms, a 145-kDa mature form and an 80-kDa amino-terminally truncated form (16). Antisera against rat brevican fusion proteins also recognize proteins of similar sizes in the soluble fraction of rat brain homogenates. Moreover, bovine brevican as well as the rat protein are brain-specific. Both proteins act as conditional proteoglycans, i.e. they can occur as proteoglycan and as free proteins.
The central domains of the two proteins have substantially diverged. The sequence identity is only 72% as compared with Another putative member of the aggrecan/versican family is BEHAB (18). BEHAB is more than 99% identical with the amino-terminal 363 amino acids of pre-brevican, indicating that both are most likely derived from the same gene and that BEHAB is a carboxyl-terminally truncated form of brevican. However, the existence of BEHAB as an actual protein has not been demonstrated yet. The distribution of BEHAB transcripts in the rat brain as revealed by in situ hybridization (18,19) is in favorable accord with the distribution of brevican immunoreactivity observed in this study.
Differential RNA processing and the alternate use of exons have been described as mechanisms to increase the molecular diversity of ECM components (e.g. Refs. [33][34][35]. Our data show that this is also the case for brevican, where secreted and GPI-anchored isoforms are synthesized from alternatively processed transcripts of the same gene. A potential splice donor site located on the 3.3-kb brevican transcript at the point of divergence from the 3.6-kb transcript suggests that the glypiation signal may be encoded by a DNA segment, which is removed as an intron from transcripts for the secreted brevican isoform. Indeed, the organization of the mouse brevican gene implies that transcripts for the GPI-anchored isoform are synthesized by "read through" into a 2-kb intron located between the exons for the carboxyl terminus of the central domain and the EGF-like domain. 3 Besides alternative processing of transcripts, the heterogeneity among brevican isoforms may reflect differential modification of primary translation products. After chondroitinase ABC digestion of the soluble protein fraction, the two major bands of 140 and 70 -80 kDa are likely to represent unprocessed and amino-terminally processed brevican, respectively (16). In the crude membrane fraction, three major PI-PLCsensitive protein bands of 140, 125, and 80 kDa are detectable. Human embryonic kidney 293 cells stably transfected with the cDNA of the 3.3-kb transcript express 140-, 125-, and 90-kDa core proteins with the latter two being sensitive to treatment with PI-PLC. Thus, 293 cells are capable of correctly inserting brevican isoforms into the plasma membrane via a GPI-anchor. The 90-kDa isoform synthesized by transfected cells coincides with the size of the in vitro translation product obtained from in vitro transcripts of the 3.3-kb cDNA, 3 suggesting that it represents the primary glypiated translation product. Glypiation is thought to be an early event in the cascade of posttranslational modifications that occurs in the endoplasmic reticulum (32). In the rat brain, only minor amounts of the 90-kDa material are detectable. The 125-kDa isoforms found in brain membranes and in transfected cells are both released from membranes by PI-PLC. They may be identical products. In contrast, the 140-kDa isoforms differ in their PI-PLC sensitivity; the brain membrane isoform is PI-PLC-releasable, and the isoform expressed in 293 cells is not. The reason for this difference is unclear. In the brain, an additional 80-kDa isoform occurs, which is released from membranes by PI-PLC. This may be a proteolytic degradation product of the larger forms, as the relative amount of the 80-kDa material varies from preparation to preparation. Taken together, these data show that multiple isoforms are derived from the 3.3-kb transcript, although the modifications leading to the differences in migration in SDS-polyacrylamide gel electrophoresis are yet unknown.
The brevican fusion construct used for antibody production contains, in addition to the relatively unique central domain, a part of the HA binding region, and the EGF-like repeat, which are reasonably conserved among a variety of ECM proteins and cell surface receptors interacting with HA. Therefore, we cannot completely rule out that some of the immunoreactive bands, e.g. the weakly immunoreactive bands observed in both soluble and membrane fractions, represent other HA-binding proteins cross-reacting with the brevican antisera. Although, based on the sizes of recognized proteins, a cross-reaction with known members of the aggrecan/versican family (the identity of the HA-binding region ranges between 57 and 64% (16)) can be excluded.
Perhaps the most important finding of this study is the identification of isoforms of brevican that are anchored to the membrane via glypiation. This is the first example of GPI anchoring in the aggrecan/versican family. All members of this family of proteoglycan core proteins thus far reported to be expressed in the central nervous system, i.e. neurocan, versican, CAT-301, the S103L PG, and brevican, are secreted into the extracellular space and appear as soluble proteins upon cell fractionation (9). They are all chondroitin sulfate-bearing proteins. On the other hand, the two brain proteoglycan core proteins known to be GPI-anchored, glypican (6, 7) and cerebroglycan (8), also termed M12 and M13 (1), are heparan sulfate proteoglycans.
What could be the biological significance of a glypiated member of the aggrecan/versican family of brain proteoglycans? This mode of anchoring proteins to the cell membrane results in an increased lateral diffusibility along the membrane (36). It enables the protein to be released from the membrane by extracellular phospholipases in a controlled manner. Glypiated proteins are discussed as markers for physiologically important membrane microdomains (37). In polarized cells, such as neurons, a precise targeting to specific plasma membrane domains caused by the GPI signal has been observed (38). In addition, a possible interaction with signal transducing tyrosine kinases was proposed based on immunoprecipitation assays (for review, see Ref. 35). In essence, GPI anchorage makes proteins good candidates to mediate dynamic remodeling of neuronal membranes (39) receiving signals from the ECM and transducing them along and/or through the membrane. This postulated gain of function for the GPI-anchored as compared with secreted brevican isoforms is accompanied by a loss of possible other functions realized by the specific carboxyl-terminal domains of the extracellular brevican isoforms. This portion of the protein comprising the EGF-repeat, the lectin-like domain, and the complement regulatory protein-like domain is conserved among the other members of the aggrecan/versican family. Possible biological functions of this region may include the recognition of carbohydrate structures and the interaction with other proteins. As a working hypothesis, we can assume that the extracellular brevican isoform is an integral ECM component interacting both with HA via its amino terminus and with other ECM components via its carboxyl terminus, whereas the membrane-linked isoform could function as a chondroitin sulfate-bearing cell surface receptor targeting interaction with HA to specific membrane subdomains.
One such subdomain could be the synapse. Interestingly, the developmental appearance of the glypiated brevican isoforms parallels synaptogenesis in the rat central nervous system. Depending on the brain area, synapse formation starts around birth, culminates between postnatal days 15 and 25, and is basically completed after day P30 (40 -43). A possible role of particular brevican isoforms in synaptic development and function is also implicated by the fact that (i) the original cDNA clone was isolated using antisera against a rat brain synaptic protein preparation and (ii) a glypiated brevican isoform is enriched in the synaptic membrane fraction. A definitive proof for a synaptic localization of glypiated brevican will require specific antibodies for this particular isoform of brevican.