Experimental Cell Research 259, 378 –388 (2000) doi:10.1006/excr.2000.4987, available online at http://www.idealibrary.com on
Modulation of Extracellular Matrix Adhesiveness by Neurocan and Identification of Its Molecular Basis Ulrika Talts,* ,† Ursula Kuhn,* Gunnel Roos,‡ and Uwe Rauch‡ ,1 *Department of Protein Chemistry, Max Planck Institute for Biochemistry, 82152 Martinsried, Germany; †Department of Cell and Molecular Biology, Lund University, 221 00 Lund, Sweden; and ‡Department of Experimental Pathology, Lund University, 221 85 Lund, Sweden
Neurocan is one of the major chondroitin sulfate proteoglycans of perinatal rodent brain. HEK-293 cells producing neurocan recombinantly show changes in their behavior. The expression of full-length neurocan led to a detachment of the secreting cells and the formation of floating spheroids. This occurred in the continuous presence of 10% fetal bovine serum in the culture medium. Cells secreting fragments of neurocancontaining chondroitin sulfate chains and the C-terminal domain of the molecule showed a similar behavior, whereas cells expressing fragments of neurocan-containing chondroitin sulfate chains but lacking parts of the C-terminal domain did not show spheroid formation. Cells secreting the hyaluronan-binding N-terminal domain of neurocan showed an enhanced adhesiveness. When untransfected HEK-293 cells were plated on a surface conditioned by spheroid-forming cells, they also formed spheroids. This effect could be abolished by chondroitinase treatment of the conditioned surface. The observations indicate that the ability of the chondroitin sulfate proteoglycan neurocan to modulate the adhesive character of extracellular matrices is dependent on the structural integrity of the C-terminal domain of the core protein. © 2000 Academic Press
Key Words: cell adhesion; extracellular matrix; chondroitin sulfate; proteoglycan; neurocan.
INTRODUCTION
A characteristic feature of the extracellular matrix of the central nervous system is the lack of fibrillar collagenous structures and the abundance of chondroitin sulfate proteoglycans and hyaluronan [1]. Although in some cases it has been observed that the distribution of chondroitin sulfate coincides with axonal extensions [2, 3], and isolated chondroitin sulfate proteoglycans or 1 To whom correspondence and reprint requests should be addressed at Department of Experimental Pathology, Lund University Hospital, So¨lvegatan 25, 22185 Lund, Sweden. Fax: ⫹46 46 158202. E-mail:
[email protected].
0014-4827/00 $35.00 Copyright © 2000 by Academic Press All rights of reproduction in any form reserved.
their core proteins promote neurite outgrowth [4, 5], chondroitin sulfate proteoglycans have mainly been implicated in the generation of barriers, in vivo and in vitro, for migrating cells or extending axons [6]. The mechanism for these properties of chondroitin sulfate proteoglycans is not clear. No signal-transmitting receptor for chondroitin sulfate chains has been elucidated so far and not many molecules, among them interestingly microbial antigens [7], have been shown to bind specifically to chondroitin sulfate. Therefore, the repulsive activity of chondroitin sulfate might just be caused by sterical interference with adhesion systems based on integrins or other cell adhesion molecules. Experiments with lipid-derivatized glycosaminoglycans indicate that chondroitin sulfate chains have to be topologically immobilized to exert the repulsive activity [8]. In vivo such an immobilization would be a result of specific interactions of the core protein. With respect to the also observed permissive properties of chondroitin sulfate-containing matrices, it has been suggested that permissive or repulsive cues exist, which are functionally associated with chondroitin sulfate chains and require chondroitin sulfate either for their biological activity or for their localization [9]. Also in this case the core proteins of the proteoglycans and their interactions with other matrix or cell surface molecules would finally be responsible for the distribution of such cues. A major family of chondroitin sulfate proteoglycans, defined by the structure of the core proteins, is the lectican family, covering the four chondroitin sulfate proteoglycans aggrecan, versican, neurocan, and brevican [10]. These molecules share homologous N-terminal globular domains, which would be able to bind to hyaluronan. With the exception of a GPI-linked brevican splice variant [11] they also share homologous C-terminal globular domains, containing C-type lectin modules. The attachment sites for glycosaminoglycan chains are located in the central regions of the lectican proteoglycans. These regions are the most divergent parts of the molecules. Neurocans central region is covering 600 amino acids and contains in addition to
378
NEUROCAN MODULATES MATRIX ADHESIVENESS
two to three chondroitin sulfate chains numerous Olinked oligosaccharides [12, 13]. These substitutions are likely to give the central part of neurocan a mucinlike character, which would be in agreement with rotary shadowing electron micrographs, showing an extended central region connecting two globular structures [14]. The expression of neurocan appears to be mainly restricted to the central nervous system [13, 15], although it has recently also been detected elsewhere [16, 17]. In rodent brain neurocan is one of the major chondroitin sulfate proteoglycans expressed during perinatal development [18, 19]. The function of neurocan during this and other periods of development is elusive. The C-terminal domain of neurocan binds to tenascin-C [20]. Therefore, neurocan could participate in the organization of the extracellular space by linking hyaluronan to tenascin-C [21]. Aggregates of hyaluronan, neurocan, and tenascin-C could assume large supramolecular structures in brain, similar to hyaluronan–aggrecan aggregates observed in cartilage [22]. During postnatal development of the rat an increasing fraction of neurocan is proteolytically processed, and in adult rats the N- and C-terminal fragments occur almost entirely as separate molecules [12, 23, 24]. These fragments of neurocan can obviously no longer link tenascin-C to hyaluronan, but are still able to modulate homophilic interactions of neural cell adhesion molecules such as NCAM and L1/NgCAM [25, 26]. In the course of the production of recombinant neurocan and neurocan fragments for various interaction studies [14, 26], cells secreting certain recombinant chondroitin sulfate proteoglycans showed a significantly decreased adhesiveness to the culture dish surface. By visual inspection of the cells it was immediately obvious that the reduction in adhesiveness was due to the reorganization of the cells from monolayers into free-floating cell aggregates, apparently spheroids. Spheroids have been described as spherical masses composed of cells and extracellular matrices, which can be produced by culturing cells in dishes coated with nonadhesive substrates [27]. The visual impression could be verified and extended by cell dissociation assays. The results of the expression of neurocan molecules with various deletions indicate that it is not the mucin-like central part, but rather the globular Cterminal domain of neurocan and the integrity of this domain that is of critical importance for the ability of the molecule to alter the behavior of cells. MATERIALS AND METHODS Construction of expression vectors, transfections, and cell culture. The preparation of constructs 359H, L639, and D925 and the characterization of the expression products have been described previously [14]. These are neurocan fragments ending with amino acid histidine 359, starting with amino acid leucine 639, and starting
379
with amino acid aspartic acid 925, respectively. Fragment 550G represents a neurocan molecule ending with glycine 550 [13] and contains about 200 amino acids of the mucin-like domain. For the construction of NC⌬M, the N-terminal part of the rat neurocan sequence was fused after histidine 359 via the artificial sequence GSSGAPLAKLT with the C-terminal part of the rat neurocan sequence at aspartic acid residue 925 [13]. For the construction of fragments lacking the second EGF-repeat and the lectin like domain (D925⌬L and NC⌬L) glutamic acid 985 was fused via an artificial leucine replacing isoleucine 986 (thereby generating a SacI restriction site) with glycine 1152. The rationale for the deletion of the second EGF module together with the C-type lectin module was the presence of one of two cysteines, which are not conserved in the other three proteoglycans of the aggrecan family and which could establish an additional disulfide bond (Fig. 1). In both molecules missing the second EGF module, D925⌬L, and NC⌬L, the C-terminal cysteine, cysteine 1257, was mutated to an alanine. In fragments D925⌬C, D925⌬S, and D925⌬T the replacement of isoleucine 986 by leucine was maintained, and the nonconserved cysteine 993 in the second EGF repeat was replaced by alanine. In addition, in fragment D925⌬C the C-terminal cysteine, cysteine 1257 was mutated to an alanine, in fragment D925⌬S a stop codon was introduced after threonine 1153, and in fragment D925⌬T a stop codon was introduced after arginine 1216. Neurocan fragment-expressing clones of human embryonic kidney cells (HEK-293, American Type Culture Collection) were produced by stable transfection with constructs in expression vectors with a CMV promotor and either neomycin or puromycin resistance, identified by the analysis of the medium by SDS–polyacrylamide gel electrophoresis (PAGE) and were maintained as described [14] in DMEM/F12 medium (Gibco) supplemented with 10% fetal bovine serum (FBS) (Gibco). 3T3 fibroblasts and C17 cells (provided by Dr. Evan Y. Snyder, Harvard Medical School, Boston, MA) were maintained in the same medium. The cDNA for neurocan, NC⌬L, NC⌬M, D925⌬C, D925⌬S, and D925⌬T were inserted into an episomal expression vector containing a CMV promotor and a puromycin resistance gene [28] and transfected into human HEK-293 EBNA cells. Proteoglycan isolation. Proteoglycans were isolated from 800 l of medium containing 10% FBS, by adding 100 l of 10⫻ TBPI (TBPI, Tris-buffered protease inhibitors, containing 5 mM EDTA, 5 mM benzamidine, 5 mM N-ethylmaleinimide, buffered with 20 mM Tris/HCl, pH 8), 100 l of 1% Triton X-100, and 100 l DEAE– Sephacel. The resin was washed twice with TBPI containing 150 mM NaCl and 0.1% Triton and twice with TBPI containing 250 mM NaCl and 0.1% Triton X-100. The proteoglycans were eluted with TBPI containing 1 M NaCl and 0.1% Triton X-100 and precipitated with 1/5 vol of 55% trichloroacetic acid (TCA), and the precipitated material was washed with ice-cold acetone. Reextractions of the first supernatant with another 100 l of DEAE–Sephacel yielded no further material. D925, D925⌬L, D925⌬S, and D925⌬T samples for the Western blot and Coomassie blue staining were precipitated from 1 ml FBS-free conditioned medium containing 0.1% Triton with 1/5 vol of 55% TCA. Analytical methods. SDS–PAGE was performed on slab gels [29] and stained with Coomassie blue (Serva) according to standard protocols. Before running the gels digestion with protease-free chondroitinase ABC (Seikagaku) was carried out for 1 h at 37°C in 100 mM Tris/HCl, pH 8.0, 30 mM sodium acetate using 1 mU enzyme per proteoglycan preparation and 2 mU enzyme per 6-cm tissue culture dish. Western blots were performed by transfer of proteins separated by SDS–PAGE to supported nitrocellulose (Bio-Rad) in Tris/glycine buffer containing 10% methanol for 1 h at 100 V using the Bio-Rad mini gel system. The blots were blocked with 1% bovine serum albumin, incubated with the monoclonal antibody 1D1 [12], and developed with alkaline phosphatase-conjugated anti-mouse antibodies.
380
TALTS ET AL.
FIG. 1. The design of recombinant neurocan fragments. The protein modules, potential glycosylation sites, and specific cysteines are indicated. Fragments L639, D925, D925⌬L, D925⌬C, D925⌬S, D925⌬T, NC⌬L, and NC⌬M are aligned according to the first EGF module, which is present in all of those fragments. For neurocan (NC), NC⌬L, and NC⌬M an icon has been designed reflecting to the shape of neurocan observed by rotary shadowing electron microscopy [14]. Those icons will be used in some of the following figures to support the identification of these molecules.
Cell dissociation assay. Cells were plated in six-well plates (Falcon) at 2 ⫻ 10 5 cells/well in 1.5 ml DMEM/F12 medium (Gibco) supplemented with 10% FBS (Gibco) and puromycin (0.5 g/ml) until confluency. Cells were washed twice in DMEM/F12 medium and grown in serum-free DMEM/F12 medium overnight. Cells were washed in phosphate-buffered saline and then treated with 0.8 ml of a nonenzymatic EDTA-containing cell dissociation solution (Sigma) for the times specified on a shaker at 37°C in 5% CO 2. Detached cells were aspirated and cells that remained attached were trypsinized and counted.
RESULTS
Neurocan and various fragments of this chondroitin sulfate proteoglycan had been recombinantly expressed in the human embryonal kidney (HEK) cell line 293 [14, 26]. Clones producing high levels of secreted recombinant proteins or proteoglycans were identified by SDS–PAGE analysis of serum-free conditioned medium and used for further production. It was noticed that cells expressing full-length neurocan and neurocan fragment D925 (Fig. 1) showed detachment and spheroid formation. Fragment D925 comprises the last 25 amino acids of the central region of neurocan, which contain the last potential glycosaminoglycan attachment site, and the entire C-terminal globular domain, which is homologous to the C-terminal domains
of the other three proteoglycans of the lectican family [14]. The obviously increased tendency of cells expressing either fragment D925 or the entire neurocan molecule to detach was examined in cell dissociation assays. Included in these assays were also cells expressing neurocan fragments L639, 550G, and D925⌬L (Fig. 1), which all showed no obvious behavioral differences from untransfected cells. D925⌬L represents a mutant form of D925 without the second EGF module and the C-type lectin module (Fig. 1). Both molecules, D925 and D925⌬L, were secreted by the cells as chondroitin sulfate proteoglycans (Fig. 2). Also included in the dissociation assay were cells expressing neurocan fragment 359H (Fig. 1) which appeared to have an increased adhesiveness and during routine cell cultivation procedures often required a second trypsin treatment to be released from the culture plate surface. The dissociation assays confirmed the observations made by visual inspection and showed that cells expressing neurocan and cells expressing fragment D925 dissociated more rapidly from the culture plate surface than untransfected cells (Fig. 3A). However, it must be noted that due to different expression levels of the recombinant molecules the effect on the cell dissocia-
NEUROCAN MODULATES MATRIX ADHESIVENESS
FIG. 2. Expression of neurocan fragments D925 and D925⌬L. Serum-free conditioned medium of 293 cells secreting D925⌬L (A) and D925 (B) was precipitated with TCA. The precipitated proteins were treated with (⫹) chondroitinase ABC (Ch’ase) or not (⫺), were electrophoresed under nonreducing conditions on a 10% SDS–polyacrylamide gel, blotted on nitrocellulose, and detected with the monoclonal anti neurocan antibody 1D1. Core proteins with apparent M r values of 42 and 45 kDa (D925) and 31 kDa (D925⌬L) could be revealed.
tion rate of each molecule can only be compared with the dissociation rate of untransfected cells. With respect to those cells, also a faster dissociation of cells expressing fragment L639 could be revealed (Fig. 3A). Cells expressing fragment D925⌬L and 550G showed a dissociation rate quite similar to untransfected cells (Fig. 3A). Cells expressing fragment 359H, the hyaluronan-binding fragment of neurocan, showed no dissociation after 22 min, a time after which all other cells were completely detached. Taken together, the results of these assays indicated a contribution of the second EGF/C-type lectin part to the ability of secreted neurocan fragments to reduce cell adhesion, which might be critical. They also suggested that the mucin-like region of neurocan plays no significant role in changing cellular behavior. To clarify these issues, several new neurocan deletion mutants were designed. Since mucins have strongly been implicated in preventing intercellular and matrix adhesion [30], potential effects of the mucin-like region were further elucidated with the fragment NC⌬M (Fig. 1). In this construct the entire central region except for the last 25 amino acids, but not the N-terminal hyaluronan-binding domain was deleted. With fragment NC⌬L (Fig. 1) it was possible to analyze the contribution of the second EGF/C-type lectin part by comparing the effects of neurocan (NC) and NC⌬L, molecules more similar in size and biochemical character than the fragments D925 and D925⌬L. To elucidate a potential contribution of other structures of the C-termi-
381
nal globule, fragment D925⌬C, with two nonconserved cysteines mutated into alanines, and two C-terminally truncated fragments, D925⌬S and D925⌬T, were designed (Fig. 1). To induce more uniform expression levels, independent of the chromosomal insertion of the expression constructs, a perisomal expression vector and HEK-293 EBNA cells were used. HEK-293 EBNA cells expressing NC showed the same behavior as HEK-293 cells expressing NC. One or 2 days after splitting they detached from the culture plate and formed spheroids (Fig. 4A). HEK-293 EBNA cells expressing NC⌬M also formed spheroids (Fig. 4C). They appeared to have an even stronger tendency to detach and form spheroids than cells expressing NC. HEK-293 EBNA cells expressing NC⌬L stayed attached to the culture plate and showed no formation of floating spheroids (Fig. 4B). Secreted proteoglycans present in the conditioned medium in the presence of 10% FBS were isolated from the culture medium of cells showing the described behavior. After chondroitinase digestion they were analyzed by SDS–PAGE and Coomassie blue staining (Fig. 5A). This analysis revealed that all three recombinant molecules were present in the conditioned medium as chondroitin sulfate proteoglycans. For NC⌬M a core protein with an apparent M r of 80 kDa could be revealed, whereas the core proteins of NC and NC⌬L, which were present in similar amounts, had apparent M r values of more than 200 kDa, reflecting a more than proportional contribution of the mucin-like region to the apparent M r’s of the core proteins. Cells expressing neurocan fragment D925⌬C, where two cysteines not conserved in the other three lecticans were mutated, showed detachment and spheroid formation (results not shown). No formation of spheroids could be observed from cells expressing the C-terminally truncated molecules D925⌬S and D925⌬T (results not shown), although they were secreted into the medium in amounts easily detectable by Coomassie blue staining (Fig. 6). However, a cell detachment assay showed a decreased adhesiveness of cells secreting those fragments (Fig. 3B). In this assay, cells expressing D925⌬S or D925⌬T were less adhesive than cells expressing NC⌬L, which came close to the behavior of untransfected cells. Cells secreting NC⌬M could not be investigated in this type of assay, since these cells did not stay attached when the medium was changed to the dissociation medium. In conclusion the above results indicate a major contribution of the second EGF/C-type lectin part and minor contributions of the parts C-terminal of the lectin module to the observed ability of the C-terminal neurocan domain to induce spheroid formation of the expressing cell. A putative disulfide bridge between two cysteines not conserved in the lectican family seems not to be relevant. When cells expressing NC, NC⌬L, and NC⌬M were
382
TALTS ET AL.
FIG. 3. Cell dissociation assay. The percentage of detached cells was determined at the indicated time points (1, 2, 3, 4, 5, 8, 10, 12, 14, 16, 18, 20, and 22 min) after application of the cell dissociation solution as described under Materials and Methods. No dissociation of HEK-293 cells secreting fragment 359H could be observed within this 22-min period. (A) Fragments expressed in HEK-293 cells. (B) Fragments expressed in HEK-293 EBNA cells. The bars represent mean ⫾ SD.
seeded on tissue culture plates conditioned by untransfected cells, which had been removed by trypsin/EDTA treatment, all three cell lines were able to form monolayers (Fig. 7). An analysis of the proteoglycans iso-
lated from the conditioned medium showed that this effect was not due to significant changes in the secretion of recombinant proteoglycans (Fig. 5B). Cells transfected with NC and NC⌬M could be trypsinized
NEUROCAN MODULATES MATRIX ADHESIVENESS
383
FIG. 4. HEK-293 EBNA cells secreting recombinant proteoglycans. HEK-293 EBNA cells secreting neurocan (A), NC⌬L (B), and NC⌬M (C) after two days in culture. (D) The typical appearance of native HEK-293 EBNA cells.
and reseeded on the same plate several times without significant alteration of the adhesiveness of the matrix. This observation made it actually possible to generate single-cell suspensions of these cells by trypsin/EDTA treatment. When untransfected cells were seeded on plates conditioned with NC-, NC⌬L-, and NC⌬M-secreting cells, which had been removed by trypsin/EDTA treatment, they formed normal monolayers on plates conditioned by NC⌬L-secreting cells (results not shown), but they had difficulties in attaching and formed spheroids when seeded on plates conditioned with NC- and NC⌬M-secreting cells (Figs. 8A and 8C). Also in these experiments with untransfected cells a more antiadhesive character of surfaces conditioned by cells secreting NC⌬M compared with surfaces conditioned by cells secreting NC was apparent. The antiadhesive character of surfaces of tissue culture plates conditioned either with NC-secreting cells or with NC⌬M-secreting cells could be eliminated by treatment of the surfaces of the dishes with chondroitinase ABC (Figs. 8B and 8D). Also other cells seeded on culture plates conditioned by NC⌬M-secreting HEK-293 EBNA cells showed a different behavior from that on control surfaces of culture plates conditioned by themselves or by untransfected HEK-293 EBNA cells. 3T3 cells could be observed as floating spheroids 1 day after transfer on NC⌬M-conditioned surfaces, whereas all cells were attached to the control surfaces at that time. C17 cells, a neural precursor cell line [31], attached more efficiently to NC⌬M-conditioned surfaces and formed some attached, but no floating, aggregates. With both
cell lines a reduction of the surface area occupied by the cells on NC⌬M-conditioned surfaces compared to control surfaces was obvious (results not shown). DISCUSSION
In the course of the production of recombinant neurocan and neurocan fragments we observed a considerable change in the behavior of these cells. The cells detach from the culture plate surface and form freefloating spheroids, induced by the matrix deposited to the surface of the tissue culture plate. A literature search revealed that a similar effect has been observed with chondroitin sulfate-containing proteoglycan fractions from rat liver reticulum fibers, which induced the formation of multicellular spheroids of adult rat hepatocytes with a high albumin-producing ability [32–34]. The formation of liver cell spheroids was, however, completely inhibited in FBS-containing medium [34], whereas neurocan-producing HEK-293 cells formed spheroids in the continuous presence of 10% FBS. The apparent uniqueness of this observation encouraged a determination of the molecular basis for this change in cellular behavior. Another impetus was that quite different changes in behavior have been reported from cells which were transfected with the proteoglycan appican, a splice variant of the Alzheimer precursor protein [35]. These changes are in marked contrast to the morphological changes observed in HEK-293 cells expressing neurocan. Appican-transfected C6 cells became much more adhesive and formed monolayers similar to untransfected HEK-293 cells. Untransfected C6 cells, in turn, behaved more like neuro-
384
TALTS ET AL.
expression of any recombinant proteoglycan in HEK293 cells should have similar effects and, moreover, the antiadhesive effect of the substrate should have been enhanced and not reduced by chondroitinase treatment. The matrix with the antiadhesive properties is formed in the presence of recombinant neurocan, all molecules endogenously secreted by HEK-293 cells, and all molecules present in fetal bovine serum, and the antiadhesive character of the matrices was preserved after treatment of the culture dish surfaces with trypsin/EDTA. These observations indicated that HEK-293 cells are a particularly robust, but at the same time sensitive model for the analysis of antiadhesive properties of matrix-incorporated proteoglycans. Also the time course of the dissociation of untransfected HEK-293 or HEK-293 EBNA cells with a detachment of half of the cells after 15 min made these cells an ideal model system for the analysis of the repulsive properties of matrices modified by recombinant molecules. The observed ability of C17 cells to attach more efficiently to NC⌬M-conditioned matrices than HEK-293 cells or 3T3 cells indicates differences between those cells in their expression of adhesion molecules. It might reflect the observation that certain axons are able to extend through chondroitin sulfaterich matrices which are apparently avoided by other axons, for example, in the thalamocortical system [3]. The observation that NC- and NC⌬M-expressing HEK-293 cells were unable to convert an adhesive substrate into an antiadhesive substrate suggests that once a matrix has been assembled, either with or withFIG. 5. Proteoglycan preparations from media conditioned by HEK-293 EBNA cells. (A) Proteoglycans isolated by ion-exchange chromatography from media-conditioned by cells secreting neurocan (A), NC⌬L (B), and NC⌬M (C) and from untransfected cells (D). Isolated proteoglycans were (⫹) or were not (⫺) treated with chondroitinase ABC (Ch’ase). The proteoglycan and core protein preparations were electrophoresed under nonreducing conditions on a 6% SDS–polyacrylamide gel. (B) Proteoglycans isolated by ion-exchange chromatography from media-conditioned by cells cultured on tissue culture plates conditioned by untransfected cells (⫹) or unconditioned plates (⫺), secreting neurocan (A), NC⌬L (B), or NC⌬M (C). All proteoglycan preparations were treated with chondroitinase and electrophoresed under nonreducing conditions on a 6% SDS–polyacrylamide gel.
can-transfected HEK-293 cells. Also in this case the effects were dependent on the matrix assembled by the transfected cells. Since recombinantly expressed proteoglycan core proteins can be expected to compete with endogenously produced core proteins for the same activated monosaccharide pool, a reduced chondroitin sulfate substitution of endogenous appican or other proteoglycans produced by HEK-293 cells might be considered as a potential cause for the unusual behavior of neurocan-expressing HEK-293 cells. However, this seems very unlikely because in such a case an
FIG. 6. Expression of neurocan fragments D925⌬T and D925⌬S. Serum-free conditioned medium of 293 cells secreting D925⌬T (A) and D925⌬S (B) was precipitated with TCA. The precipitated proteins were treated with (⫹) chondroitinase ABC or not (⫺) and were electrophoresed under nonreducing conditions on a 8% (A) or 10% (B) SDS–polyacrylamide gel. Core proteins with apparent M r values of 45 kDa (D925⌬T) and 40 kDa (D925⌬S) could be revealed. S, molecular weight standards.
NEUROCAN MODULATES MATRIX ADHESIVENESS
385
FIG. 7. HEK-293 EBNA cells secreting recombinant proteoglycans cultured on different surfaces. HEK-293 EBNA cells secreting neurocan (A, D), NC⌬L (B, E), and NC⌬M (C, F) cultured on unconditioned tissue culture plates (A–C) and on plates conditioned by untransfected HEK-293 EBNA cells (D–F).
out antiadhesive molecules, it is difficult to significantly extend or remodel this matrix. It also indicates a firm integration of neurocan or neurocan fragments with antiadhesive activity into this matrix. The observed differences in the antiadhesive effect of matrices assembled in presence of proteoglycans which contained the complete C-terminal domain and of proteoglycans missing parts of this domain indicate the importance of the structural integrity of the C-terminal domain of the core protein for a firm integration. The thereby immobilized chondroitin sulfate chains could, in turn, be responsible for the repulsive effect. Fragments which were not modified with chondroitin sulfate chains were also observed in the samples not treated with chondroitinase (Figs. 2 and 6). In the above-described model these unmodified fragments might compete with glycosaminoglycosylated fragments for binding sites to other matrix proteins. However, fragment D925-secreting cells showed the formation of spheroids even in the presence of considerable amounts of unmodified fragments, whereas fragment D925⌬L-secreting cells showed no reduction in adhe-
siveness, although almost all fragments secreted by those cells were modified (Fig. 2). It should be mentioned that our results cannot completely exclude the opposite scenario that the chondroitin sulfate chain mediates matrix integration and the protein part induces repulsion. In either case it appears to be necessary to combine those two components to generate molecules with antiadhesive activity. In neurocan a functional chondroitin sulfate attachment site is separated from the first EGF module only by a few amino acids, whereas in versican the closest consensus sequence for attachment of chondroitin sulfate chains is more than 100 amino acids away from the C-terminal domain [36]. This could make appropriately composed neurocan fragments, combining a chondroitin sulfate chain and the C-terminal domain, much less susceptible for a proteolytic separation than equivalent fragments of versican, which is, at least on the mRNA level, endogenously expressed by HEK-293 cells (C. Retzler and U. Rauch, unpublished observation). HEK-293 cells expressing the secreted brevican variant never showed spheroid formation, but these brevi-
386
TALTS ET AL.
FIG. 8. Untransfected HEK-293 EBNA cells cultured on different surfaces. Untransfected HEK-293 EBNA cells cultured on tissue culture plates, which had been conditioned by neurocan-secreting cells (A), by NC⌬M-secreting cells (B), by neurocan-secreting cells which have afterward been treated with chondroitinase ABC (C) and on plates conditioned by NC⌬M-secreting cells which have afterward been treated with chondroitinase ABC (D).
can molecules were not notably substituted with chondroitin sulfate chains [37]. In attempts to express the C-terminal domains of brevican and versican, the brevican fragment, but not the vesican fragment, was secreted in amounts which were detectable by Coomassie blue staining of SDS–PAGE gels of proteins precipitated from conditioned serum-free medium (C. Retzler and U. Rauch, unpublished observations). The lack of the observation of high expression levels of the Cterminal versican domain could be due to its growth factor like activity which might compromise the survival of the cells expressing higher amounts of this fragment [38]. Unfortunately, a fusion proteoglycan containing the last 25 amino acids of the central neurocan region and the C-terminal brevican domain was secreted in very low quantities only detectable by Western blot (U. Rauch, unpublished observation). Thus, it was not possible to distinguish whether the ability of neurocan to modulate the character of the extracellular matrix of HEK-293 cells depends on the short distance of the chondroitin sulfate attachment site to the C-terminal domain or on specific interactions of this domain with the matrix deposited by HEK293 cells. However, it appears more likely that the HEK-293-cell-derived matrix is not an equally appropriate matrix for all of the C-terminal domains of the lectican proteoglycans. It has been shown that the Ctype lectin modules within this domain have considerably different affinities for the glycoproteins tenascin-R and fibulin-1 [39, 40]. In line with those results, plasmon resonance-based interaction studies, where
tenascin-C bound with high affinity to the C-terminal neurocan domain [20], showed no binding of tenascin-C to immobilized brevican (W. Goehring and U. Rauch, unpublished observation). Tenascin-C is a hexameric glycoprotein which can be found in serum [41] and can be incorporated into fibronectin matrices [42]. Since it colocalizes with neurocan in brain [43], it would be able to contribute to an integration of neurocan into extracellular matrices, in vitro and in vivo. Neurocan binds to tenascin-C in a calcium-dependent fashion [20]. Therefore, this interaction could be important during assembly of the matrix, but would have to be supported by other interactions during trypsin/EDTA treatments of the matrix. The HEK-293 cells expressing fragment 359H, the N-terminal fragment of neurocan, showed a strongly enhanced binding to the culture dish surface. This hyaluronan-binding domain [14] without attached glycosaminoglycan chains could be expected to have quite similar biological properties as the other equivalent lectican domains, since all are likely to share the same ligand, hyaluronan. However, in contrast to our observation with HEK-293 cells, the homologous domain of aggrecan reduced the adhesion of chondrocytes and fibroblasts to the matrix deposited by those cells [44]. Further, it has been reported that the N-terminal domain of brevican, expressed in 9L glioma cells, increased strongly the invasiveness of those cells into normal brain tissue [45]. A common mechanism linking all of these observations is difficult to imagine. A possible activity of all hyaluro-
NEUROCAN MODULATES MATRIX ADHESIVENESS
nan-binding domains could be an interference with the formation of space-filling hyaluronan–proteoglycan aggregates, by competing with hyaluronan-binding chondroitin sulfate-substituted proteoglycans. A study about the localization of proteoglycans and their ligands in the pericellular matrix of cultured fibroblasts showed a similar distribution of the molecules PG-M/versican, hyaluronan, CD44, and tenascin-C, and all these molecules were notably excluded from focal contacts [46]. These molecules could create space-filling aggregates in the subcellular space, and an inhibition of the formation of strongly hydrated voluminous hyaluronan–proteoglycan aggregates in the subcellular space could decrease its volume and make the focal contacts less accessible for proteases and dissociative agents. This concept might explain the increased adhesiveness of the 293 cells expressing the N-terminal neurocan domain, but not the decreased adhesiveness of chondrocytes and fibroblasts expressing the N-terminal aggrecan domain. However, extending this concept to the extracellular brain matrix, it could also help to understand the increased invasiveness of gliomas. Also in this case an inhibition of the formation of voluminous hyaluronan–proteoglycan aggregates, potential barriers for migrating cells, by small hyaluronan-binding molecules might be involved. It must be noted that physiologically it is unlikely that a polysaccharide-free N-terminal domain of neurocan is generated, since two potential glycosaminoglycan attachment sites are in close proximity to this domain [13], whereas polysaccharide-free hyaluronanbinding fragments are established processing products of aggrecan [47], versican [48], and brevican [45]. Thus, the physiological role of neurocan might rather reside in its ability to perturb the attachment of certain cells and axons to particular matrices. Neurocan could thereby either promote cell aggregation and axonal fasciculation and participate in the compartmental organization of the developing nervous system or contribute to its plasticity, since it also can be observed in zones permissive for axonal extension [3]. The authors thank Gerlinde Kulbe for excellent technical assistance, Dieter Zimmermann and Ray Boot-Handford for helpful comments on the manuscript, and Rupert Timpl, Reinhard Fa¨ssler, and Peter Ekblom for support. Funding was provided by the German research council (DFG, Ra 544/4-1 to UR), the Swedish Natural ¨ sterlund Stiftelse, Carl Science Research Council (NFR), the Alfred O Tesdorpfs Stiftelse, and Alice och Knut Wallenbergs stiftelse.
3.
Bicknese, A. R., Sheppard, A. M., O’Leary, D. D., and Pearlman, A. L. (1994). Thalamocortical axons extend along a chondroitin sulfate proteoglycan-enriched pathway coincident with the neocortical subplate and distinct from the efferent path. J. Neurosci. 14, 3500 –3510.
4.
Faissner, A., Clement, A., Lochter, A., Streit, A., Mandl, C., and Schachner, M. (1994). Isolation of a neural chondroitin sulfate proteoglycan with neurite outgrowth promoting properties. J. Cell Biol. 126, 783–799.
5.
Iijima, N., Oohira, A., Mori, T., Kitabakate, K., and Kohsaka, S. (1991). Core protein of chondroitin sulfate proteoglycan promotes neurite outgrowth from cultured neocortical neurons. J. Neurochem. 56, 706 –708.
6.
Margolis, R. K., and Margolis, R. U. (1993). Nervous tissue proteoglycans. Experentia 49, 429 – 445.
7.
Buffet, P. A., Gamain, B., Scheidig, C., Baruch, D., Smith, J. D., Hernandez-Rivas, R., Pouvelle, B., Oishi, S., Fujii, N., Fusai, T., Parzy, D., Miller, L. H., Gysin, J., and Scherf, A. (1999). Plasmoidum falciparum domain mediating adhesion to chondroitin sulfate A: A receptor for human placental infection. Proc. Natl. Acad. Sci. USA 96, 12743–12748.
8.
Sugiura, N., Sakurai, K., Hori, Y., Karasawa, K., Suzuki, S., and Kimata, K. (1993). Preparation of lipid-derivatized glycosaminoglycans to probe a regulatory function of the carbohydrate moieties of proteoglycans in cell–matrix interactions. J. Biol. Chem. 268, 15779 –15787.
9.
Emerling, D. E., and Lander, A. D. (1996). Inhibitors and promoters of thalamic neuron adhesion and outgrowth in embryonic neocortex: Functional association with chondroitin sulfate. Neuron 17, 1089 –1100.
10.
Iozzo, R. V., and Murdoch, A. D. (1996). Proteoglycans of the extracellular environment: Clues from the gene and protein side offer novel perspectives in molecular diversity and function. FASEB J. 10, 598 – 614.
11.
Seidenbecher, C. I., Richter, K., Rauch, U., Fa¨ssler, R., Garner, C. C., and Gundelfinger, E. D. (1995). Brevican, a chondroitin sulfate proteoglycan of rat brain, occurs as secreted and cell surface glycosylphosphatidylinositol-anchored isoforms. J. Biol. Chem. 270, 27206 –27212.
12.
Rauch, U., Gao, P., Janetzko, A., Flaccus, A., Hilgenberg, L., Tekotte, H., Margolis, R. K., and Margolis, R. U. (1991). Isolation and characterisation of developmentally regulated chondroitin sulfate and chondroitin/keratan sulfate proteoglycans of brain identified with monoclonal antibodies. J. Biol. Chem. 266, 14785– 14801.
13.
Rauch, U., Karthikeyan, L., Maurel, P., Margolis, R. U., and Margolis, R. K. (1992). Cloning and primary structure of neurocan, a developmentally regulated, aggregating chondroitin sulfate proteoglycan of brain. J. Biol. Chem. 267, 19536 –19547.
14.
Retzler, C., Wiedemann, H., Kulbe, G., and Rauch, U. (1996). Structural and electron microscopic analysis of neurocan and recombinant neurocan fragments. J. Biol. Chem. 271, 17107–17113.
15.
Prange, C. K., Pennacchio, L. A., Lieuallen, K., Fan, W., and Lennon, G. G. (1998). Characterisation of the human neurocan gene, CSPG3. Gene 221, 199 –205.
16.
Lilien, J., Balsamo, J., Hoffman, S., and Eisenberg, C. (1997). Beta-Catenin is a target for extracellular signals controlling cadherin function: The neurocan-GalNAcPTase connection. Curr. Top. Dev. Biol. 35, 161–189.
17.
Oleszewski, M., Beer, S., Katich, S., Geiger, C., Zeller, T., Rauch, U., and Altevogt, P. (1999). Integrin and neurocan binding to L1 involves distinct Ig domains. J. Biol. Chem. 274, 24602–24610.
REFERENCES 1.
Margolis, R. U., Margolis, R. K., Chang, L. B., and Petri, C. (1975). Glycosaminoglycans of brain during development. Biochemistry 14, 85– 88.
2.
Sheppard, A. M., Hamilton, S. K., and Pearlman, A. L. (1991). Changes in the distribution of extracellular matrix components accompany early morphogenetic events of mammalian cortical development. J. Neurosci. 11, 3928 –3942.
387
388 18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29. 30.
31.
32.
33.
TALTS ET AL. Margolis, R. K., Rauch, U., Maurel, P., and Margolis, R. U. (1996). Neurocan and phosphacan: Two major nervous tissuespecific chondroitin sulfate proteoglycans. Perspect. Dev. Neurobiol. 3, 273–290. Milev, P., Maurel, P., Chiba, A., Mevissen, M., Popp, S., Yamaguchi, Y., Margolis, R. K., and Margolis, R. U. (1998). Differential regulation of hyaluronan binding proteoglycans in developing brain: Aggrecan, versican, neurocan, and brevican. Biochem. Biophys. Res. Commun. 247, 207–212. Rauch, U., Clement, A., Retzler, C., Fro¨hlich, L., Fa¨ssler, R., Go¨hring, W., and Faissner, A. (1997). Mapping of a defined neurocan binding site to distinct domains of tenascin-C. J. Biol. Chem. 272, 26905–26912. Rauch, U. (1997). Modeling an extracellular environment for axonal pathfinding and fasciculation in the central nervous system. Cell Tissue Res. 290, 349 –356. Mo¨rgelin, M., Heinegard, D., Engel, J., and Paulson, M. (1994). The cartilage proteoglycan aggregate: Assembly through combined protein– carbohydrate and protein–protein interactions. Biophys. Chem. 50, 113–128. Oohira, A., Matsui, F., Watanabe, E., Kushima, Y., and Maeda, N. (1994). Developmentally regulated expression of a brain specific species of chondroitin sulfate proteoglycan, neurocan, identified with a monoclonal antibody 1G2 in rat cerebrum. Neuroscience 60, 145–157. Matsui, F., Nishizuka, M., Yasuda, Y., Aono, S., Watanabe, E., and Oohira, A. (1998). Occurrence of a N-terminal proteolytic fragment of neurocan, not a C-terminal half, in a perineuronal net in the adult rat cerebrum. Brain Res. 790, 45–51. Grumet, M., Flaccus, A., and Margolis, R. U. (1993). Functional characterization of chondroitin sulfate proteoglycans of brain: Interaction with neurons and neural cell adhesion molecules. J. Cell Biol. 120, 815– 824. Retzler, C., Go¨hring, W., and Rauch, U. (1996). Analysis of neurocan structures interacting with the neural cell adhesion molecule N-CAM. J. Biol. Chem. 271, 27304 –27310. Takezawa, T., Mori, Y., Yonaha, T., and Yoshizato, K. (1993). Characterization of morphology and cellular metabolism during the spheroid formation by fibroblasts. Exp. Cell Res. 208, 430 – 441. Kohfeldt, E., Go¨hring, W., Mayer, U., Zweckstetter, M., Holak, T. A., Chu, M.-L., and Timpl, R. (1996). Conversion of the kunitz type module of collagen IV into a highly active trypsin inhibitor by site-directed mutagenesis. Eur. J. Biochem. 238, 333–340. Laemmli, U. K. (1970). Cleavage of structural proteins during the assembly of the head of bacteriophage T4. Nature 227, 680 – 685. Kemperman, H., Wijnands, Y., Wesseling, J., Niessen, C. M., Sonnenberg, A., and Roos, E. (1994). The mucin epiglycanin on TA3/Ha carcinoma cells prevents alpha 6 beta 4-mediated adhesion to laminin and kalinin and E-cadherin mediated cell– cell interaction. J. Cell Biol. 127, 2071–2080. Snyder, E. Y., Deitcher, D. L., Walsh, C., Arnold-Aldea, S., Hartwieg, E. A., and Cepko, C. L. (1992). Multipotent neural cell lines can engraft and participate in development of mouse cerebellum. Cell 68, 33–51. Shinji, T., Koide, N., and Tsuji, T. (1988). Glycosaminoglycans partially substitute for proteoglycans in spheroid formation of adult rat hepatocytes in primary culture. Cell Struct. Funct. 13, 179 –188. Koide, N., Shinji, T., Tanabe, T., Asano, K., Kawaguchi, M., Sakaguchi, K., Koide, Y., Mori, M., and Tsuji, T. (1989). Continued high albumin production by multicellular spheroids of
Received May 23, 2000
adult rat hepatocytes formed in the presence of liver-derived proteoglycans. Biochem. Biophys. Res. Commun. 161, 385–391. 34.
Koide, N., Sakaguchi, K., Koide, Y., Asano, K., Kawaguchi, M., Matsuchima, H., Takenami, T., Shinji, T., Mori, M., and Tsuji, T. (1990). Formation of multicellular spheroids composed of adult rat hepatocytes in dishes with positively charged surfaces and under other nonadherent environments. Exp. Cell Res. 186, 227–235.
35.
Wu, A., Pangalos, M. N., Efthimiopoulos, S., Shioi, J., and Robakis, N. K. (1997). Appican expression induces morphological changes in C6 glioma cells and promotes adhesion of neural cells to the extracellular matrix. J. Neurosci. 17, 4987– 4993.
36.
Zimmermann, D. R., and Ruoslahti, E. (1989). Multiple domains of the large fibroblast proteoglycan, versican. EMBO J. 8, 2975–2981.
37.
Thon, N., Haas, C. A., Rauch, U., Merten, T., Faessler, R., Frotscher, M., and Deller, T. (2000). The chondroitin sulfate proteoglycan brevican is upregulated by astrocytes after entorhinal cortex lesion in adult rats. Eur. J. Neurosci., in press.
38.
Zhang, Y., Cao, L., Yang, B. L., and Yang, B. B. (1998). The G3 domain of versican enhances cell proliferation via epidermial growth factor-like motifs. J. Biol. Chem. 273, 21342–21351.
39.
Aspberg, A., Miura, R., Bourdoulous, S., Shimonaka, M., Heinegard, D., Schachner, M., Ruoslahti, E., and Yamaguchi, Y. (1997). The C-type lectin domains of lecticans, a family of aggregating chondroitin sulfate proteoglycans, bind tenascin-R by protein-protein interactions independent of carbohydrate moiety. Proc. Natl. Acad. Sci. USA 94, 10116 –10121.
40.
Aspberg, A., Adam, S., Kostka, G., Timpl, R., and Heinegard, D. (1999). Fibulin-1 is a ligand for the C-type lectin domains of aggrecan and versican. J. Biol. Chem. 274, 20444 –20449.
41.
Schenk, S., Muser, J., Vollmer, G., and Chiquet-Ehrismann, R. (1995). Tenascin-C in serum: A questionable tumor marker. Int. J. Cancer 61, 443– 449.
42.
Chung, C. Y., and Erickson, H. P. (1997). Glycosaminoglycans modulate fibronectin matrix assembly and are essential for matrix incorporation of tenascin-C. J. Cell Sci. 110, 1413–1419.
43.
Grumet, M., Milev, P., Sakurai, T., Karthikeyan, L., Bourdon, M., Margolis, R. K., and Margolis, R. U. (1994). Interactions with tenascin and differential effects on cell adhesion of neurocan and phosphacan, two major chondroitin sulfate proteoglycans of nervous tissue. J. Biol. Chem. 269, 12142–12146.
44.
Cao, L., and Yang, B. B. (1999). Chondrocyte apoptosis induced by aggrecan G1 domain as a result of decreased cell adhesion. Exp. Cell Res. 246, 527–537.
45.
Zhang, H., Kelly, G., Zerillo, C., Jaworski, D. M., and Hockfield, S. (1998). Expression of a cleaved brain specific extracellular matrix protein mediates glioma cell invasion in vivo. J. Neurosci. 18, 2370 –2376.
46.
Yamagata, M., Saga, S., Kato, M., Bernfield, M., and Kimata, K. (1993). Selective distribution of proteoglycans and their ligands in pericellular matrix of cultured fibroblasts: Implications for their roles in cell–substratum adhesion. J. Cell Sci. 106, 55– 65.
47.
Sandy, J. D., Neame, P. J., Boynton, R. E., and Flannery, C. R. (1991). Catabolism of aggrecan in cartilage explants: Identification of a major cleavage site within the interglobular domain. J. Biol. Chem. 266, 8683– 8685.
48.
Perides, G., Asher, R. A., Lark, M. W., Lane, W. S., Robinson, R. A., and Bignami, A. (1995). Glial hyaluronate-binding protein: A product of metalloproteinase digestion of versican? Biochem. J. 312, 377–384.