Address correspondence and reprint requests to Constanze I. Seidenbecher, AG Molecular Mechanisms of Plasticity, Department of Neurochemistry/Molecular Biology, Leibniz Institute for Neurobiology, 39118 Magdeburg, Germany. E-mail: email@example.com
Brevican is a neural-specific proteoglycan of the brain extracellular matrix, which is particularly abundant in the terminally differentiated CNS. It is expressed by neuronal and glial cells, and as a component of the perineuronal nets it decorates the surface of large neuronal somata and primary dendrites. One brevican isoform harbors a glycosylphosphatidylinositol anchor attachment site and, as shown by ethanolamine incorporation studies, is indeed glypiated in stably transfected HEK293 cells as well as in oligodendrocyte precursor Oli-neu cells. The major isoform is secreted into the extracellular space, although a significant amount appears to be tightly attached to the cell membrane, as it floats up in sucrose gradients. Flotation is sensitive to detergent treatment. Brevican is most prominent in the microsomal, light membrane and synaptosomal fractions of rat brain membrane preparations. The association with the particulate fraction is in part sensitive to chondroitinase ABC and phosphatidylinositol-specific phospholipase C treatment. Furthermore, brevican staining on the surface of hippocampal neurons in culture is diminished after hyaluronidase or chondroitinase ABC treatment. Taken together, this could provide a mechanism by which perineuronal nets are anchored on neuronal surfaces.
sodium dodecylsulfate–polyacrylamide gel electrophoresis
Brain tissue is characterized by a large variety of extracellular matrix (ECM) constituents comprising secreted glycoproteins and proteoglycans as well as cell surface-bound matrix receptors. One of the most prominent constituents of mature rodent brain ECM is brevican, a neural-specific chondroitin sulfate proteoglycan (CSPG), which belongs to the lectican family (reviewed in Gary et al. 1998; Yamaguchi 2000). In accordance with the other lecticans aggrecan, versican and neurocan, the brevican core protein primary structure suggests a dumb-bell-like shape consisting of two globular domains at the N- and C-termini and a central elongated region, which bears the chondroitin sulfate (CS) side chains. As a conditional proteoglycan, brevican occurs both as a CSPG and in a CS-free non-proteoglycan form. The presence of CS chains has been reported to be necessary for the neurite outgrowth-inhibiting action of the molecule (Yamada et al. 1997). The N-terminal brevican structure harbors an Ig-like domain and a so-called proteoglycan tandem repeat, which is capable of binding hyaluronic acid (HA). HA, also termed hyaluronan, is the central organizing polysaccharide of the brain ECM that forms huge protein–glycan aggregates. In the C-terminal part of brevican an epidermal growth factor (EGF)-like motif, a C-type lectin domain and a sushi-repeat (also termed complement regulator protein (CRP)-like motif) represent potential protein–protein interaction sites. Accordingly, the lectin domain has been shown to interact specifically with the extracellular glycoprotein tenascin R (Aspberg et al. 1997; Brückner et al. 2000) and with sulfated glycolipids at the cell surface (Miura et al. 1999; Miura et al. 2001). The central region contains matrix metalloprotease cleavage sites. Specific proteolytic cleavage leads to the separation of the N-terminal HA-binding region from the C-terminal cell surface- and protein-interacting part of the molecule, and may have implications for the integrity and biophysical features of brain ECM. Interestingly, brevican cleavage was reported to be of particular importance for the invasiveness of glioma tumors (Matthews et al. 2000).
The core protein occurs in at least two discrete isoforms, which arise from the same gene via alternative splicing (Seidenbecher et al. 1995; Rauch et al. 1997). The smaller isoform lacks the C-terminal globular interacting domains and instead harbors an attachment sequence for a glycosylphosphatidylinositol (GPI) anchor, which ties this isoform to the cell membrane (Seidenbecher et al. 1995). Biochemical experiments have demonstrated that brevican immunoreactivity is partially susceptible to phosphatidylinositol-specific phospholipase C (PI-PLC; EC 18.104.22.168). However, this is not definite proof for the presence of a GPI anchor, because an indirect influence of the release of other glypiated molecules on brevican cannot be excluded. In this paper, we provide direct evidence for brevican glypiation by use of metabolic labeling and incorporation protocols.
Despite several studies on brevican interaction partners the function of the proteoglycan still remains enigmatic. Available data suggest that brevican plays a role as a bifunctional regulator in neurite outgrowth during development and after injury (summarized in Yamaguchi 2000) as well as in the invasiveness of glioma tumors (Nutt et al. 2001). Electrophysiological studies with brevican knockout mice revealed a phenotype with clearly reduced synaptic plasticity (Brakebusch et al. 2002). These facts indicate that brevican might be involved in the cross-talk between the inside and outside of the neuron. In this context it is interesting that brevican was shown to be a structural constituent of perineuronal nets (Hagihara et al. 1999; Brückner et al. 2000). These net-like specializations of the brain ECM are discussed in the context of synapse stabilization and insulation, growth factor supply and many other features of the ECM in the mature CNS. However, it is not completely understood how brevican is integrated into these networks and how perineuronal nets are anchored at neuronal surfaces. Therefore, we analyze the extent and mode of brevican association with particulate subcellular brain fractions and with cultured neurons to elucidate how the proteoglycan might be anchored in the ECM meshwork and at neural surfaces.
Materials and methods
Radiochemicals, enzymes and antibodies
[3H]ethanolamine hydrochloride (999 GBq/mmol, 37 MBq/mL) was purchased from Amersham Life Science (Amersham, UK). PI-PLC was obtained either from Oxford Glycosystems (Abingdon, UK) (Bacillus thuringiensis PI-PLC), or from Sigma–Aldrich (Poole, UK) or Boehringer Mannheim (Mannheim, Germany) (B. cereus PI-PLC). Chondroitinase ABC (chondroitin ABC lyase from Proteus vulgaris; EC 22.214.171.124) was supplied by Sigma–Aldrich or Boehringer Mannheim and hyaluronidase (from Streptomyces hyaluronlyticus; EC 126.96.36.199) by Calbiochem–Novabiochem (San Diego, CA, USA). Rabbit anti-brevican polyclonal antiserum has been described previously (Seidenbecher et al. 1995), monoclonal mouse antiSAP90/postsynaptic density (PSD) 95 antibody was purchased from Transduction Laboratories (BD Biosciences, San Diego, CA, USA) and the monoclonal mouse anti-neuroplastin antibody (anti-gp65) was a gift from Dr P. W. Beesley (Hill et al. 1988). Fluorescence-labeled anti-rabbit Alexa-488 was obtained from Molecular Probes (Eugene, OR, USA).
Isolation of subcellular protein fractions and western blotting
Brain tissue from adult rats (total forebrain) was homogenized in 5 mm Hepes buffer, pH 7.4, containing 0.32 m sucrose and a protease inhibitor cocktail (Complete; Boehringer Mannheim). Subcellular membrane protein fractions were obtained essentially as described previously (tom Dieck et al. 1998) and digested with chondroitinase ABC at 0.1 U per 60 µg protein in 40 mm Tris-HCl, 40 mm sodium acetate, pH 8.0, for 90 min at 37°C. In the partitioning experiment 600 µg rat brain homogenate was adjusted to 40 mm Tris-HCl, 40 mm sodium acetate, pH 8.0, centrifuged at 14 000 g for 20 min, and both the soluble and the resuspended particulate fractions were incubated with chondroitinase ABC. Separation of proteins by sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS–PAGE) on 5–20% gels under fully reducing conditions and transfer on to nitrocellulose was performed according to standard protocols. Western blots were immunodeveloped by overnight incubation with primary antiserum and processed employing the enhanced chemiluminescence detection system (Amersham Corp.). Quantification of relative amounts of immunoreactivity was performed with a Bio-Rad GS800 scanner (Bio-Rad Laboratories, Hercules, CA, USA) and QuantityOne software (Bio-Rad).
The flotation assay was performed essentially as described by Sanmarti-Vila et al. (2000). Briefly, 600 µg crude rat brain membranes (postnuclear 12 000 g pellet P2) were adjusted to 55% sucrose, homogenized and after a short spin (5 min at 5000 g) placed underneath a continous 25–52.5% sucrose gradient using a 20-gauge needle. Centrifugation was performed in a SW55 rotor (Beckman L7 Centrifuge; Beckman Coulter, Munich, Germany) for 16 h at 100 000 g and 4°C. Afterwards, 14 fractions (250 µL each) were collected from the gradient, incubated with 25 mU chondroitinase ABC in 40 mm Tris-HCl, 40 mm sodium acetate, pH 8.0, for 90 min at 37°C. After addition of SDS-loading buffer the fractions were finally treated with iodacetamide to inactivate free sulfhydryl groups.
Cell culture and metabolic labeling
Human embryonic kidney (HEK) 293 cells stably expressing GPI-anchored brevican were described previously (Seidenbecher et al. 1995). The oligodendrocyte progenitor cell line Oli-neu was established by Jung et al. (1995). Protocols for metabolic labeling of cultured cells were adapted from Koch et al. (1997). Basically, cells grown in six-well plates were incubated with 3.7 MBq [3H]ethanolamine per ml medium for 18 h, harvested and washed once with Tris-buffered saline (TBS). The cell pellet was homogenized in 50 mm Tris-HCl, 60 mm octylglucoside, pH 7.2, including Complete protease inhibitor, kept for 30 min on ice, spun at 14 000 g for 30 min and the resulting supernatant was subjected to immunoprecipitation with polyclonal rabbit anti-brevican antibodies or the corresponding preimmune serum bound to GammaBind proteinG sepharose (Pharmacia, Amersham Biosciences, Freiburg, Germany). Immunoprecipitates were washed with TBS including 0.02% sodium azide, separated by SDS–PAGE, blotted on to nitrocellulose and exposed to 3H-sensitive PhosphoImager plates (Roytest, Berlin, Germany) for approximately 6 days. Afterwards, the blots were immunodeveloped with anti-brevican antiserum.
Primary hippocampal cultures
Hippocampal cultures were prepared and grown on coverslips according to Goslin et al. (1998). Enzymatic treatment was performed on living cells after 21 days in vitro for 60 min at 37°C with 50 turbidity reducing units per mL hyaluronidase in 100 mm phosphate buffer, 150 mm sodium acetate, pH 6.0, or 0.5 U/mL chondroitinase ABC in 40 mm Tris-HCl, 40 mm sodium acetate, pH 8.0, each in the presence of a protease inhibitor cocktail. Control experiments were performed in the same buffer without enzyme. After washing in phosphate-buffered saline, cultures were fixed in 4% ice-cold paraformaldehyde for 15 min, washed in 25 mm glycine and blocked with 10% normal goat serum, 1% bovine serum albumin with 0.1% Triton X-100. Triton was added to allow double staining with anti-microtubule associated protein 2 (MAP2) antibodies to evaluate physical integrity of the neurons after glycosidase digestion. Cultures were stained with affinity-purified rabbit anti-brevican antiserum (1 : 50) and anti-rabbit Alexa-488 (1 : 800), and examined under a Zeiss Axioplan 2 imaging microscope (Göttingen, Germany). Images were taken with a Visitron Systems camera (Puchheim, Germany).
Brevican partitions into soluble and insoluble subfractions
The majority of brevican immunoreactivity is soluble and can be extracted from brain homogenates under physiological buffer conditions. As shown in Fig. 1a approximately 25% of the immunoreactive bands are found in the particulate fraction after centrifugation at 14 000 g. This holds true both for the intact molecules migrating at 145 kDa and for the 80-kDa proteolysis products. The 14 000 g supernatant still contains membranes and after consecutive centrifugation (100 000 g) a considerable amount of brevican is found in the microsomal fraction (Fig. 1b). After further fractionation of the 14 000 g particulate fraction P2 in a sucrose step gradient brevican is found to be preferentially associated with the myelin fraction, the light membranes and synaptosomes. The relative enrichment towards the synaptosomal protein fraction was confirmed by immunodeveloping the same blot with a monoclonal antibody against SAP90/PSD95, a truly synapse-associated protein (Fig. 1b).
Insoluble brevican is in part associated with membranes
The particulate fraction P2 is a mixture of integral membrane proteins, membrane-attached molecules, and components of dense and insoluble proteinaceous meshworks inside and outside the cells such as cytomatrix and ECM. To investigate the nature of the membrane association of insoluble fractions of brevican we employed flotation assays, in which membrane-associated proteins float up when placed under a continuous sucrose gradient, while membrane-free proteins and their aggregates remain at the bottom. Interestingly, a significant amount of brevican found in the P2 fraction indeed floats up in the gradient (Fig. 2a). Comparison with the synaptic transmembrane glycoprotein neuroplastin shows that both molecules peak in the same medium-density fraction, indicating a true association with membranes. However, a considerable portion of membrane-free brevican remains in the bottom fractions. Preincubation with chondroitinase ABC does not influence the flotation behavior of brevican (data not shown). However, after membrane solubilization by detergent pretreatment of the P2 fraction with 1% Triton X-100 before gradient loading brevican immunoreactivity is lost from the floating fractions (Fig. 2b).
Metabolic ethanolamine labeling confirms the existence of a GPI anchor
As we have shown previously (Seidenbecher et al. 1995) brevican occurs in at least two discrete isoforms encoded by alternative transcripts from the brevican gene (Rauch et al. 1997): a major secreted isoform and a minor variant which is smaller and potentially membrane-anchored via a GPI moiety. A putative GPI anchor attachment site was identified in the open reading frame of the corresponding cDNA and successful release of the isoform employing PI-PLC is indicative of glypiation. To confirm the presence of a GPI anchor we performed incorporation studies with [3H]ethanolamine. To this end we used in a first attempt HEK293 cells stably transfected with a cDNA encoding GPI-brevican (Seidenbecher et al. 1995) and untransfected HEK cells as a control. This cDNA gives rise to a 90-kDa primary translation product and two mature protein bands at 125 and 145 kDa (Fig. 3a), possibly owing to post-translational modifications. Cells were grown in the presence of [3H]ethanolamine, harvested and brevican was immunoprecipitated with a polyclonal brevican antibody. After SDS–PAGE and western blotting the blot was exposed to a Phosphoimager plate and finally immunodeveloped (Fig. 3a). Strong incorporation into the 125 kDa band and faint labeling of the band at approximately 145 kDa can be visualized in the immunoprecipitate from transfected cells. In a second series of experiments we used the oligodendrocyte precursor cell line, Oli-neu, which is known to express this brevican isoform endogenously (Seidenbecher et al. 1998). Again, metabolic labeling and consecutive immunoprecipitation yield a radioactive band which corresponds to the 145-kDa band seen in the transfected cells (Fig. 3b) confirming conclusively the existence of a GPI anchor structure. Currently, it is unclear why Oli-neu cells synthesize only the 145-kDa glypiated isoform.
Glycosidases can solubilize brevican from particulate fractions
When brain crude membrane preparations are incubated with chondroitinase ABC a large portion of brevican immunoreactivity, in particular 145 -kDa and 80-kDa isoforms, are released from the particulate fraction into the supernatant (Fig. 4a). Co-incubation with PI-PLC in addition liberates the smaller 125-kDa brevican isoform. Both enzymatic treatments are sensitive to Zn2+ ions, a known inhibitor of these enzymes. Of note, a small amount of brevican can be set free during incubation at 37°C without enzyme addition, potentially as a result of the presence of endogenous lysosomal chondroitinases in the crude membrane fraction. To obtain further insight into the mechanisms that trap brevican in the particulate fraction we performed sequential release experiments with different agents (Fig. 4b). The largest fraction is solubilized in the first incubation with chondroitinase ABC at pH 8.0, the optimal pH for CS degradation. Afterwards, PI-PLC treatment leads to liberation of the GPI-anchored subfraction. Consecutive chondroitinase ABC incubation at pH 6.8, the optimal pH for HA degradation, releases only a small additional amount of brevican. The immunoreactivity remaining in the particulate fraction is not sensitive to high-salt treatment (1 m NaCl) and only moderately sensitive to detergents (1% Triton X-100) indicating a tight association with insoluble protein complexes.
Brevican localized at the surface of cultured neurons is glycosidase sensitive
To verify the biochemical behavior of brevican in a physiologically relevant system we used hippocampal primary cultures, which express brevican in significant amounts. Cells cultured for 3 weeks develop mature properties and nearly all neurons are decorated with brevican immunoreactivity on their surfaces (Figs 5a and c). After chondroitinase ABC treatment of the living culture, staining is clearly reduced but not completely eliminated (Fig. 5b). When hyaluronidase is applied instead, the reduction of surface labeling is even more pronounced (Fig. 5d). Consistent with the biochemical data described above these findings show that significant amounts of brevican can be released by glycosidase treatment.
The structural organization of the brevican core protein makes the molecule an ideal candidate to interconnect the extracellular HA-based proteinaceous meshwork and the surface of neural cells. In the present paper we addressed the question of how brevican is attached to cellular membranes and ECM. It is well documented that ECM components are more tightly cross-linked and less soluble in the mature brain than at early postnatal stages (Hockfield et al. 1990; Rauch 1997; Yamaguchi 2000). Indeed, we found approximately 25% of brevican to be insoluble under physiological buffer conditions. In the embryonic brain, however, brevican is almost completely contained in the soluble fraction (Miura et al. 2001). Moreover, in contrast to its closest relative neurocan (for review see Rauch et al. 2001) the total amount of brevican immunoreactivity is strongly up-regulated during postnatal development and reaches highest levels after terminal differentiation of the rodent brain (Milev et al. 1998).
Particulate brevican immunoreactivity is associated with several subcellular fractions: microsomes, myelin, light membranes and synaptosomes. The microsomal fraction mostly comprises intracellular endoplasmic reticulum-derived membrane fragments and therefore might possibly represent newly synthesized brevican. The myelin fraction contains low-density lipids that are derived from axonal myelin sheaths. The presence of brevican in this fraction has already been reported (Seidenbecher et al. 1998; Niederöst et al. 1999) and it was suggested to contribute to the non-permissive properties of CNS myelin (Niederöst et al. 1999). The light membrane fraction is a less well defined membrane subpopulation, which contains a mixture of endoplasmic reticulum, Golgi and plasma membranes as estimated by electron microscopy (Cohen et al. 1977; Hu et al. 1998). This fraction contains extrasynaptic marker molecules of neuronal membranes, such as BDNF receptor gp145trkB and gp95trkB or the 5-hydroxytryptamine receptor 2A (Hu et al. 1998). Finally, the synaptosomes represent closed spherical presynaptic endings filled with cytosolic components and vesicles that are still in contact with the adjacent postsynaptic membrane and the underlying PSD. The integrity of these structures might be preserved via cell–cell and cell–matrix adhesive molecules. Our finding that significant amounts of brevican are present in this subfraction agrees with immunohistochemical evidence of a perisynaptic localization (Seidenbecher et al. 1997; Hagihara et al. 1999) and with the fact that we originally cloned rat brevican from a screen for novel synapse-associated proteins (Seidenbecher et al. 1995; Langnaese et al. 1996).
In principle, the presence of brevican in particulate fractions can result from both membrane association and the existence of brevican-containing insoluble lattices. In our flotation assays we could discriminate between these two subpopulations; approximately 60% of the particulate brevican immunoreactivity was truly membrane associated. The observation that chondroitinase ABC treatment before loading the gradient did not influence the flotation behavior suggests that CS chains are not directly involved in brevican linkage to the cell membrane. Plasmalemmal binding was observed in in vitro assays previously (Yamada et al. 1997) and is thought to be mediated via a direct interaction of the C-type lectin domain with sulfated membrane glycolipids, such as sulfatides and HNK-1-reactive sulfoglucuronylglycolipids (Miura et al. 1999; Miura et al. 2001), which in turn could provide a cell-substrate recognition system.
A further mechanism of membrane anchoring for extracellular proteins is glypiation. Our metabolic labeling studies confirmed the primary structure-based hypothesis that the smaller brevican isoform which lacks the C-terminal globular domain indeed harbors a GPI moiety. This is an outstanding feature within the lectican family and it might enable the protein to serve as a tenascin R-independent ECM receptor. Furthermore, GPI-brevican might mediate cellular signaling events as several glypiated cell surface molecules are known to be signal transducers (Kasahara and Sanai 2000).
Our release experiments show that enzymatic treatment not only results in digestion and removal of the extracellular carbohydrate components but also affects the anchoring of core proteins themselves. The target sites of the enzymes used in this study, together with all known interactions of brevican with cell surfaces, are depicted schematically in Fig. 6.
Like brevican, other brain ECM molecules have also been reported previously to be extractable by glycosidase digestion: hyaluronidase treatment drastically reduces brain- specific hyaluronan-binding protein Bral1 immunoreactivity in cerebellar sections (Oohashi et al. 2002), solubilizes neurocan from hippocampal slices (Förster et al. 2001) and releases versican from pre-oligodendrocytes (Asher et al. 2002) indicative of HA-mediated cell surface retention of these molecules. Furthermore, glial hyaluronate binding protein is also released from brain homogenate fractions by enzymatic treatment with chondroitinase ABC (Asher et al. 1991). These data point to a central role not only for HA but also for CS in the network formation and aggregation of the mature neural ECM. As our primary culture experiments demonstrate, this also holds true for isolated neuronal cells grown in vitro, which lack the higher-order structural hierarchy of brain tissue. Therefore, cultured primary hippocampal neurons might represent a suitable test system with which further to investigate the formation of perineuronal ECM.
The modulation of cell–matrix interactions is assumed to be associated with neuronal plasticity in general and synaptic plasticity in particular. In functional terms several attempts have been made to influence brain properties via application of matrix-degrading enzymes like hyaluronidase or chondroitinase. Strikingly, in a recent report Bradbury et al. (2002) provide evidence that in vivo chondroitinase ABC treatment of injured spinal cord clearly promotes structural and functional recovery of the affected neuronal connections. After chondroitinase injection into rat brain, unexpectedly long-lasting changes in molecular ECM composition, including fragmented staining of core-protein components, were investigated by Brückner et al. (1998). Furthermore, in an elegant developmental study Förster et al. (2001) describe hyaluronidase-sensitive cues responsible for lamina-specific adhesion and innervation in entorhino–hippocampal co-cultures. Finally, it was reported that treatment of hippocampal slices with chondroitinase ABC leads to a 50% reduction in long-term potentiation although short-term synaptic plasticity was unaffected (Bukalo et al. 2001). At present, it is unclear if these treatments directly affect brevican anchoring on neuronal surfaces.
Recently, it became evident that brevican might play a role in synaptic plasticity as a complete lack of brevican in knockout mice leads to a clear reduction in the maintenance of long-term potentiation (Brakebusch et al. 2002) possibly due to a mis-structured perisynaptic ECM. Interestingly, this phenotype can be mimicked in rat hippocampal slices treated with anti-brevican antibodies. Therefore, it will be of vital importance to identify binding partners of membrane-associated brevican isoforms to elucidate the molecular mechanisms underlying this phenomenon.
This work was supported by grants from the Deutsche Forschungsgemeinschaft (DFG) to CIS (Se 952/2-1, 2-2), the Land Sachsen-Anhalt (LSA 3004 A/0088H), the LSA and European Social Fonds to KHS, the Fritz-Thyssen-Stiftung and the Fonds der Chemischen Industrie. We are grateful to Dr P. Beesley for the generous gift of anti-neuroplastin antibody, to Dr J. Trotter for kindly providing Oli-neu cells and for helpful suggestions on the metabolic labeling protocol, and to M. Kreutz for supplying primary hippocampal cultures. The authors acknowledge the expert technical assistance of Mrs M. Marunde.