Collapsin Response Mediator Proteins of Neonatal Rat Brain Interact with Chondroitin Sulfate*

Chondroitin sulfate proteoglycans are structurally and functionally important components of the extracellular matrix of the central nervous system. Their expression in the developing mammalian brain is precisely regulated, and cell culture experiments implicate these proteoglycans in the control of cell adhesion, neuron migration, neurite formation, neuronal polarization, and neuron survival. Here, we report that a monoclonal antibody against chondroitin sulfate-binding proteins from neonatal rat brain recognizes collapsin response mediator protein-4 (CRMP-4), which belongs to a family of proteins involved in collapsin/semaphorin 3A signaling. Soluble CRMPs from neonatal rat brain bound to chondroitin sulfate affinity columns, and CRMP-specific antisera co-precipitated chondroitin sulfate. Moreover, chondroitin sulfate and CRMP-4 were found to be localized immuno-histochemically in overlapping distributions in the marginal zone and the subplate of the cerebral cortex. CRMPs are released to culture supernatants of NTera-2 precursor cells and of neocortical neurons after cell death, and CRMP-4 is strongly expressed in the upper cortical plate of neonatal rat where cell death is abundant. Therefore, naturally occurring cell death is a plausible mechanism that targets CRMPs to the extracellular matrix at certain stages of development. In summary, our data indicate that CRMPs, in addition to their role as cytosolic signal transduction molecules, may subserve as yet unknown functions in the developing brain as ligands of the extracellular matrix.

The functions of chondroitin sulfate proteoglycans (CS-PGs) in neural tissue can be categorized into effects on cell adhesion, cell migration, neurite formation, neuron polarization, synaptic modulation/plasticity, and neuron survival (for reviews, see Ref. 6 and references therein and Ref. 11). These effects often critically depend on glycosaminoglycans or may even be attributed solely to glycosaminoglycan chains. Many of the listed functions, however, have been discovered using in vitro experiments that were designed in a way that the proteoglycan or glycosaminoglycan component became limiting in the assays. In vivo, on the other hand, there is marked structural redundancy of ECM components (1, 3) if one considers for example the existence of four different lecticans (see above). Thus, phenotypes of knockout animals lacking a single ECM protein are often rather mild (for example in knockouts for neurocan (12) and tenascin-C (13,14)). Inactivation of enzymes involved in CS biosynthesis, on the other hand, could lead to more severe phenotypes since CS is a component of multiple ECM proteins. In the chondroitin-6-sulfate transferase knockout mouse, however, no major CNS pathology was found (15). C-6-S may be replaced by C-4-S, and expression data suggest that C-4-S is probably more important in the developing brain. Nevertheless, the importance of CS in vivo is underscored impressively by the observation that after local treatment with chondroitinase ABC, which degrades C-4-S, C-6-S, and dermatan sulfate, regeneration of functional neurites in the adult spinal cord is enabled (16). Thus, CS-PGs are considered to contribute to the inhibition of regenerative responses in the adult mammalian nervous system. The current insight into the mechanisms of how CS acts on neurons is still rudimentary. Binding partners of CS-PGs at the plasma membrane include sulfatide and several (eventually glycosylphosphatidylinositol-anchored) cell adhesion molecules of the Ig family such as N-CAM, L1, TAG-1, and F3/ contactin (for a review, see Ref. 6). The signaling events exerted after binding of CS-PGs to these molecules are unknown. On the other hand, CS-PGs bind a variety of soluble ligands including growth factors like bovine fibroblast growth factor and oligomeric glycoproteins of the ECM-like tenascins (6). Interestingly, CS is a critical component of a molecular scaffold to which diffusible molecules are bound that convey inhibitory or promoting actions, e.g. on the adhesion of thalamic neurons and the formation of neurites (17). The identity of these diffusible molecules, however, is as yet unknown.
To elucidate the molecular basis of the neurotrophic actions of chondroitin sulfate, we previously fractionated protein extracts from neonatal rat brain on a chondroitin sulfate affinity column and used the eluted binding proteins to generate monoclonal antibodies (18). One of these antibodies, termed mAb-9, recognizes a 65-kDa protein with laminar expression in the neocortex, which parallels the expression of CS. The protein is present in both the fraction of soluble proteins and in the particulate fraction of neonatal rat brain. The aim of the present study was to identify this chondroitin sulfate-binding protein and to characterize its interaction with glycosaminoglycans. We show that mAb-9 recognizes a soluble protein that is present in the cytosol, termed collapsin response mediator protein-4 (CRMP-4). This protein and its relatives interact with chondroitin sulfate, and they are released from the cytosol of neurons to the extracellular space most probably after cell death. This may explain why CRMP-4 was found to be colocalized with chondroitin sulfate in the developing neocortex of rat brain in regions where naturally occurring cell death is prevalent.

MATERIALS AND METHODS
Unless otherwise stated, chemicals were from Serva (Heidelberg, Germany), Sigma, Roche Molecular Biochemicals, or Merck. ZERO Blunt vector for PCR cloning was from Invitrogen, and pQE-30 vector for bacterial expression of histidine-tagged proteins was from Qiagen (Hilden, Germany).
Protein Preparation-Whole brains from neonatal Wistar rats were shock-frozen in liquid nitrogen and homogenized in a Dounce homogenizer in 10 mM HEPES, pH 7.4, 2 mM MgCl 2 (HEPES buffer) containing 2 mM Pefabloc, 1 mM leupeptin, and 1 mM pepstatin. The homogenate was centrifuged at 10,000 ϫ g for 15 min, and the supernatant was subsequently centrifuged at 100,000 ϫ g for 1 h to obtain soluble protein. Forty milligrams of this protein material were loaded on a 1-ml Mono Q fast protein liquid chromatography column (Amersham Biosciences), washed with 10 column volumes of loading buffer (HEPES buffer), and eluted stepwise with 5 column volumes of HEPES buffer containing 100, 300, 500, and 2000 mM NaCl. One-milliliter fractions were collected and analyzed by SDS-PAGE, silver staining and Western blot with mAb-9. Western positive fractions were concentrated (Centri-con10, Millipore) and loaded on a 25-ml Superose-12 column (Amersham Biosciences). Again, 1-ml fractions were collected and analyzed by Western blotting. For mass spectrometry, proteins were prepared according to the method of Gevaert et al. (23). Briefly, the positive fractions were concentrated by ultrafiltration and separated by SDS-PAGE. After staining with zinc/imidazole the protein bands corresponding to the Western signal of the monoclonal antibody were excised. After destaining in 5% citric acid for 15 min and another 3 times for 15 min in deionized water and incubation in SDS-PAGE sample buffer containing 0.1% SDS, 10% glycerol, 50 mM dithiothreitol, 12 mM Tris/HCl, pH 6.8, and 0.1% bromphenol blue for 1 h, the gel was cut into small pieces (5 ϫ 5 mm) and loaded on top of a concentration gel (5% acrylamide, 0.26% bisacrylamide, 125 mM Tris/HCl, pH 6.8, and 0.1% SDS) inside of a Pasteur pipette (length, 145 mm). The pipette was transferred to an isoelectro-focusing unit (Bio-Rad), and the gel pieces were carefully overlayered with running buffer (50 mM Tris, 190 mM glycine, 0.1% SDS). Electrophoresis at 250 V was continued until the bromphenol blue approached the lower edge of the pipette. The gel was removed from the Pasteur pipette and stained with Coomassie Blue, and the sharp blue protein band in the lower part of the pipette was excised, destained, and stored at Ϫ20°C until further analysis. Aliquots of this material were analyzed by Western blotting to confirm hat the desired protein had been excised.
In-gel Digestion for Nanospray-ESI-MS-The destained gel piece was washed twice in digestion buffer (10 mM NH 4 HCO 3 ) for 15 min and twice for 15 min in digestion buffer/acetonitrile 1:1 (v/v). The gel piece was re-swollen with 2 l of protease solution (trypsin at 0.05 g/l in digestion buffer), and after 20 min another 10 l of digestion buffer was added. After digestion overnight at 37°C the supernatant was collected and dried down to about 0.5 l. For nanospray ESI-MS 2 l of 70% formic acid were added, and this solution was used in 0.5-l aliquots.
ESI-MS-ESI-MS was done as described (24) using a TSQ 7000 Triple quadrupole mass spectrometer (Finnigan MAT, San Jose, CA) equipped with the standard ESI source or an in-house constructed nanospray source. The ESI-voltage was between 0.6 and 1.1 kV for nanospray. Mass spectra were acquired with a scan speed of 1000 Da/s. For ESI-MS/MS analysis, argon at a pressure of 3 millitorr was used as collision gas. Data acquisition and evaluation were done on a DEC work station using the ICIS software, version 8.2.1. Peptide mass calculation was done with the BIOWORKS software, version 8.2.1.
Peptide Mass Fingerprint Data Base Search-Peptide masses obtained from in-gel digestion were used for searching the SwissProt.r34 data base with MS-FIT (falcon.ludwig.ucl.ac.uk/msfit.htm). The standard parameters were: species Rattus norvegicus, molecular mass 40 -100 kDa, tryptic digest with a maximum of 1 missed cleavage site. Peptide masses were assumed to be monoisotopic, and cysteine was assumed to be not modified. The allowed mass error was set at 0.1%.
Glycosaminoglycan Affinity Chromatography-Chondroitin sulfate was coupled to EAH-Sepharose (Amersham Biosciences) as described (18). As control columns, heparin Hitrap, SP Hitrap, and CM Hitrap 1-ml columns (Amersham Biosciences) were used. Approximately 10 mg of soluble neonatal rat brain proteins obtained after ultracentrifugation of postnuclear supernatants at 100,000 ϫ g for 1 h were filtered (0.45 m) and chromatographed on 1-ml analytical columns using an Ä kta Explorer equipment (Amersham Biosciences). Runs on CS columns and control columns were carried out in parallel, taking advantage of the column-scouting routine of the Unicorn 3.1 software using sequential step elution with PBS containing 300 mM NaCl, 750 mM NaCl, 2 M NaCl, and 4 M guanidinium hydrochloride (GuaHCl) as described (18). To remove the salt from the eluent fractions and to concentrate the proteins, they were precipitated with acetone (Ϫ20°C) and washed twice with 80% ethanol (4°C) before Western blot analysis with the peptide-specific antibodies against CRMPs.
Immunoprecipitation-1-ml aliquots of soluble neonatal rat brain proteins prepared as described above were incubated with 1 l of the different peptide-specific CRMP antisera for 1 h at 4°C and with 10 l of protein A-agarose (Calbiochem) for an additional hour at 4°C with continuous rocking. The precipitates were collected by centrifugation, washed 3 times in phosphate-buffered saline, dissolved in 100 l of Laemmli buffer, and analyzed by Western blotting with the biotinylated CRMP-Fam antibody or CS56 (Sigma) against CS.
Immunohistochemistry-Before brain dissection, animals were perfused with standard mammalian Ringer's solution, pH 7.4, followed by 3.7% formaldehyde. Brains were post-fixed for 16 h and washed extensively in tap water. After dehydration in a series of increasing ethanol concentrations, brain tissue was embedded in paraffin, and 10-m sections were cut on a microtome (Leica HM 355 S). Sections were mounted on Histobond slides, dried for 2 days at 37°C, and used for immunohistochemistry. First, sections were deparaffinized and hydrated by decreasing concentrations of ethanol in H 2 O. Afterward, sections were incubated in boiling 2ϫ SSC (1ϫ SSC ϭ 0.15 M NaCl and 0.015 M sodium citrate) for 20 min. After equilibrating in PBS, sections were treated with 1% H 2 O 2 in PBS for 10 min to remove endogenous peroxidase activity. After permeabilization with 0.5% Triton X-100 in PBS for 10 min, sections were washed with PBS and incubated in a 4% bovine serum albumin, PBS solution for at least 30 min. First, antibodies to CRMP-4 were used in a dilution of 1:1000, and for CS56 (Sigma), in a dilution of 1:200 (all in 4% bovine serum albumin, PBS). Sections were incubated at 4°C overnight. After washing, sections were treated with the secondary biotinylated antibody diluted 1:200 in 4% bovine serum albumin, PBS, washed again, and incubated in a streptavidinperoxidase complex (ABC-kit, Vector). After a 1-h incubation, sections were washed intensively and stained with diaminobenzidine (0.05% in Tris-buffered saline). Counterstaining was done with hemalum (Mayers hemalum, 1:6 dilution in H 2 O, Merck) for 5 min followed by several washes in H 2 O and a final wash in tap water. Pictures were taken with a digital camera (Polaroid DMC Ie, Cambridge, UK) connected to a Zeiss Axioskop 2 (Jena).
Cell Culture-NTera-2 precursor cells (Stratagene) were grown in Dulbecco's modified Eagle's medium/F-12 supplemented with 10% fetal calf serum (FCS), L-glutamine, and penicillin/streptomycin. Cell death was induced by feeding the cells with medium without FCS after repeated washings with serum-free medium. Conditioned medium was harvested after 3 days and centrifuged at 1000 ϫ g for 15 min to remove floating cells and debris. For Western blot analysis, conditioned medium was subjected to Q-Sepharose (Hitrap 1-ml column, Amersham Biosciences) chromatography to capture CRMPs using elution with a linear gradient from 150 mM to 1 M NaCl in phosphate buffer.
Primary neocortical neurons were prepared as described (27) and plated onto poly-D-lysine-coated 10-cm cell culture Petri dishes (Falcon) or 6-well cell culture plates (Sarstedt). Cultures were incubated at 37°C in humidified 10% CO 2 , 90% air for 16 -24 h and analyzed by phase contrast microscopy. To visualize the morphology of nuclei, cultures were fixed for 10 min in 4% paraformaldehyde, washed with PBS, and stained briefly with 4,6-diamidine-2-phenylindole (100 g/ml in PBS). Pictures were taken with a digital camera (Axiovision, Zeiss) connected to a Zeiss Axiovert 100M, and images were processed with the Axiovision software (Zeiss, Göttingen, Germany).
Analytical Procedures-Protein determination was performed with the Bradford assay or the detergent-compatible protein assay (both purchased from Bio-Rad) using bovine serum albumin as the standard. Lactate dehydrogenase activity was determined as described (28).

Purification of CRMP-4 from Neonatal Rat Brain-
The monoclonal antibody mAb-9 used in this study recognizes a 65-kDa protein that is abundant in the soluble fraction of neonatal rat brain (18). From this material the 65-kDa protein was captured on a Mono Q anion exchange column (Fig. 1). When steps of increasing ionic strength were applied to the column, about 30% of the cross-reacting protein was eluted with 200 mM sodium chloride, and the remaining 70% was released at 500 mM sodium chloride according to Western blot analysis (Fig. 1C). Interestingly, the electrophoretic mobility of the cross-reacting protein bands from the eluate fractions was slower in comparison to the starting material (Fig. 1C). Because the 200 mM NaCl eluate contained a lower amount of contaminating proteins than the 500 mM eluate, according to silver-stained SDS-gels (Fig. 1B), fraction 7 from the Mono Q column was subsequently fractionated using size exclusion chromatography on a Superose 12 column (Fig. 2). From this column the cross-reacting protein eluted after 11-12 ml, corresponding to a molecular mass of ϳ200 kDa (Fig. 2, A and C). Finally, purification to homogeneity was achieved by preparative SDS-PAGE (data not shown). In-gel digestion with trypsin yielded 22 peptides, which were analyzed by ESI-MS. In the SwissProt.r34 rat sequence data base (Table I) 10/22 peptide masses fitted to a rat sequence homologous to "Turned on after division" protein of 64 kDa (TOAD-64 (29), also called collapsin response mediator protein-2 (CRMP-2 (30)). Extension of the search to the entire mammalian data base (nrdb, EMBL Heidelberg), however, showed a more close match of 15/22 peptides to mouse CRMP-4 (mUlip, Unc33-like phosphoprotein (31)). Furthermore, MS/MS analyses of three different peptides from the spectrum confirmed that these were derived from the rat homologue of CRMP-4. Because only a truncated cDNA sequence for rat CRMP-4 was available in the data base, a chimeric sequence was assembled from the available truncated rat CRMP-4 and mouse CRMP-4/mUlip. To this chimeric sequence 19 of the 22 obtained masses could be matched exactly (data not shown). Thus, the mass spectrometric analyses indicate that mAb-9 recognizes collapsin response mediator protein-4 from rat brain.
Western Blot Analysis of Recombinant CRMPs with mAb-9-To confirm that mAb-9 recognizes CRMP-4 and to determine the specificity of the antibody, the coding regions of rat CRMP-1-4 sequences were amplified by PCR, sequenced, and expressed as Nterminally His-tagged fusion proteins in E. coli. All four CRMPs were found in inclusion bodies. When equal amounts of the four recombinant proteins were analyzed by Western blotting, all recombinant proteins were detected with an anti-polyhistidine antibody (Fig. 3B). Recombinant CRMP-4 clearly reacted with mAb-9, whereas CRMP-2 was not detected even after prolonged exposure of radiography films (Fig 3A). Moreover, CRMP-1 and -3 crossreacted weakly with mAb-9 (Fig. 3A). In summary, mass spectrometric experiments and the immunological analysis of recombinant CRMPs consistently demonstrated that mAb-9 recognizes an epitope that is present on CRMP-4.
Characterization of Peptide-specific Antibodies against CRMPs-To distinguish different members of the CRMP family, we produced peptide-specific polyclonal antisera. Peptides were designed from regions of the CRMP amino acid sequence that displayed marked divergence between the CRMPs. Furthermore, a sequence was selected that is conserved among the CRMP family members to generate a pan-specific antibody. After immunization of rabbits, five different antibodies were obtained termed anti-CRMP-1 to -4 and anti-CRMP-Fam. Western blots of the bacterially expressed recombinant CRMPs demonstrated binding of anti-CRMP-Fam to all four recombinant CRMPs (Fig. 3C) and mono-specific binding of the antibodies anti-CRMP-1, -2, and -4 to the corresponding recombinant proteins (Fig. 4, A, B, and D). Anti-CRMP-3 strongly bound to recombinant CRMP-3 and showed weak cross-reactions with CRMP-1 and -2 (Fig. 4C).
Interaction of CRMPs with Glycosaminoglycans-The mAb-9 was raised against chondroitin sulfate-binding proteins that were solubilized with CHAPS from the particulate fraction of neonatal rat brain. This raises the possibility that CRMPs could be cytosolic associates of membrane-associated protein complexes that contain receptors for ECM proteoglycans. In this case, soluble CRMPs might not necessarily bind to chondroitin sulfate. Thus, we examined if CRMPs from the soluble fraction of neonatal rat brain bind to chondroitin sulfate-Sepharose columns. First, eluates from a chondroitin-6-sulfate-Sepharose column were screened for the presence of CRMPs using the peptide-specific antibodies. All four CRMPs bound to the column (Fig. 5). Although substantial amounts of the bound CRMPs were released on washing of the column at moderate stringency (Fig 5, lane 1 of each panel), limited quantities of each CRMP remained on the column even after harsh washing with 2 M NaCl and were only eluted by the chaotropic salt GuaHCl (Fig. 5, lane 4 of each panel). Almost identical elution profiles of CRMPs were observed with a chondroitin-4-sulfate column (data not shown). As the control we performed in parallel chromatographies on a heparin-Sepharose column under exactly the same experimental conditions. Although all four CRMPs bound quantitatively to the heparin column, they were eluted completely by washing the column with 300 mM NaCl (Fig. 6, lane 1 of each panel), none of the CRMPs was eluted with GuaHCl, as seen for both chondroitin sulfate columns (Fig. 6, lane 4 of each panel). Taken together, these chromatography profiles indicate a weak charge-mediated interaction of CRMPs with heparin but tight binding of sub- To rule out the possibility that soluble CRMPs were retained on chondroitin sulfate columns because they interact with putative contaminants of these glycosaminoglycans from their biological sources (i.e. shark cartilage for C-6-S and bovine trachea for C-4-S), we carried out immuno-co-precipitation experiments as an independent approach. Anti-CRMP4 and anti-CRMP-Fam precipitated detectable amounts of CRMPs (Fig.   7A). Western blot analysis with the monoclonal antibody CS56 against chondroitin sulfate demonstrated co-precipitation of characteristic smeary proteoglycan bands after precipitation of soluble brain-derived proteins with anti-CRMP4 and the anti-CRMP-Fam but not after incubation with a preimmune serum (Fig. 7B). Thus, immuno-coprecipitation data confirmed that soluble CRMPs bind to chondroitin sulfate.

TABLE I MS-Fit search results of the peptide masses obtained after digestion of the unknown protein, recognized by mAb-9 with trypsin
Shown are search results for 22 different peptides obtained by tryptic digest of the antigen of mAb-9 after screening the SwissProt.r34 data base using the MS-Fit program. Parameter settings were as detailed under "Materials and Methods." Ten peptides matched the rat Turned On After Division 64 kDa protein (TOAD-64, 62,277.9 Da). "Data submitted," measured peptide masses according to MALDI-TOF; "MϩH matched," mass of matched peptides from the database assuming mono-protonation; ⌬, relative mass differences between observed and predicted peptides; "Start" and "End" refer to the position of the peptides in the TOAD-64 sequence; "Peptide sequence," the amino acid sequences are given in single-letter code (residues in parentheses refer to the amino acids immediately preceding or following the peptides, respectively); "Modifications," oxidation of methionine was the only allowed protein modification (IMet-ox). The 12 unmatched masses are 732. 8 ping Regions of the Cerebral Cortex-To determine sites in brain tissue where an interaction between CRMPs and chondroitin sulfate may take place, we examined the immunohistochemical distribution of CRMP-4 and chondroitin sulfate in the cerebral cortex of neonatal rat brain. CRMP-4-positive cells were present in the upper part of the cortical plate. Interestingly, within these cells nuclei were strongly stained (Fig. 8, A  and C). Moreover, there was a fine reticular CRMP-4 staining without obvious relationship to cellular structures in the mar-ginal zone and a diffuse labeling of the subplate and prospective white matter (Fig. 8, A and C). Chondroitin sulfate, on the other hand, was expressed as a reticular meshwork in the marginal zone (Fig. 8, B and D). Furthermore, the subplate was diffusely stained. In summary, CRMP-4 and chondroitin sulfate were partly expressed in the same regions of the neocortex.
In agreement with published data on the naturally occurring cell death in the developing rodent brain (32-34), we found numerous pyknotic nuclei in the upper cortical plate, immediately beneath the marginal zone corresponding to layer II of the mature neocortex (Fig. 8D, arrowheads).

Release of CRMPs to the Extracellular Space-The biochemical interaction of CRMPs with chondroitin sulfate in vitro obviously raises the question of under what conditions this segregation might break down and in what compartment the interaction could become relevant in vivo.
The almost congruent reticular staining patterns of CRMP-4 and CS in the marginal zone of the cerebral cortex suggested that CRMPs may be released to the extracellular space, e.g. after the programmed death of neurons. To substantiate that CRMPs are released to the extracellular space, we screened cell culture supernatants of neural cells for the presence of CRMPs. First, in serum-free conditioned media of NTera-2 precursor cell cultures we detected CRMP immunoreactivity after capture on a Q-Sepharose column (Fig. 9A). The fact that no CRMP-like molecules were detected in fetal calf serum (Fig. 9B) rules out the possibility that residual traces of FCS are the source of CRMP-like immunoreactivity in these supernatants. Because serum-starved cultures contained many dead cells on microscopic examination and considerable activity of the cytosolic marker enzyme lactate dehydrogenase (LDH) (Fig. 9C), we assume that CRMPlike molecules possibly were released from the cytosol of NTera-2 precursor cells that underwent cell death. Our data do not exclude the possibility, however, that active mechanisms of release may exist. Second, primary cultures of neocortical neurons were studied for the release of CRMPs. These cultures had been characterized previously and contained Ͼ95% microtubule-associated protein 2-positive neurons (11,18,27). When these neuron cultures were grown in the presence of HEK293 cell-conditioned medium to provide trophic support, no lactate dehydrogenase activity was detected in the supernatants of these cultures. However, we observed that ϳ50% of cells underwent cell death with compacted rounded cell morphology (Fig. 10, A and B) and pyknosis of the nuclei (Fig. 10, C, D, and E) independent on the initial plating density. Supernatants of these cultures contained CRMPs and their amount paralleled the extent of cell death (Fig. 10F). Almost all of the released CRMP was CRMP-4 ( Fig. 10G) since we were not able to detect significant amounts of the other CRMPs in the supernatant even after prolonged film exposure (data not shown). Taken together, these experiments indicated that NTera-2 precursor cells and neocortical neurons release CRMPs to the extracellular compartment and that naturally occurring cell death may be a possible release mechanism. DISCUSSION Previously, we generated the monoclonal antibody mAb-9 against chondroitin sulfate-binding proteins from neonatal rat FIG. 6. Heparin affinity chromatography of soluble CRMPs from rat brain. Soluble proteins (10 mg) from neonatal rat brain were loaded on a heparin-Sepharose column, washed with 300 mM NaCl (1), and eluted with 750 mM NaCl (2), 2 M NaCl (3), or 4 M GuaHCl (4). All fractions were tested in Western blot with the different mono-specific CRMP antisera as indicated on the right. L, load; F, flow-through. Bars on the left indicate the position of apparent molecular weight (MW) marker bands (in kDa). brain. This antibody recognizes a 65-kDa protein with laminar expression in the cerebral cortex (18). In the present study, we identified this protein as collapsin response mediator protein-4 (CRMP-4/Ulip (31,35)) based on mass spectrometric analysis of the purified protein and Western blot analysis of recombinant CRMP-4. Furthermore, we obtained evidence that CRMP1, -2, -3, and -4 interact with chondroitin sulfate proteoglycans. The collapsin response mediator proteins form a family of five homologues, the first of which (formerly called CRMP-62, now termed CRMP-2) was identified by expression cloning as a signal transduction molecule, mediating the growth cone collapse activity of semaphorin 3A/collapsin on peripheral sensory neurons (30). The rat orthologue (called TOAD-64) was identified as a marker of differentiating post-mitotic neurons re-expressed after nerve lesions in the adult animal (29). CRMP-4 (mUlip) (31) was cloned in mouse as a phosphoprotein crossreacting with an anti-stathmin antibody and was later identified in rat and human by homology screening (35,36). Recently, CRAM/CRMP-5 was cloned as a protein that interacts with CRMP-3 (37), with a glycine transporter (38), and that crossreacts with an anti-ZAP-70 antibody (39). CRMPs share sequence similarities with dihydropyrimidinase and with the gene product of the unc-33 gene of Caenorhabditis elegans, which is involved in axonal pathfinding (30).
Different phosphorylation states of CRMPs exist (31,40) and may account for slightly different electrophoretic mobilities of the protein bands recognized by mAb-9 (Fig. 1C), since in the absence of phosphatase inhibitors like orthovanadate, the protein may undergo dephosphorylation, which may markedly influence the apparent molecular weight as determined by SDS-PAGE (41). Moreover, differences in phosphorylation may explain why CRMP immunoreactivity eluted from the Mono Q . F, Western blot (WB) analysis with anti CRMP-Fam of cell culture supernatants from low density culture (low), high density culture (high), and HEK-293 cells (HEK) as indicated above the lanes. The positions of apparent molecular weight standards are shown on the right. G, Western blot analysis with anti-CRMP-4 of high density culture supernatant (high) and soluble proteins from neonatal rat brain (brain) as indicated above the lanes. The positions of apparent molecular weight standards are shown on the right (in kDa). column in two distinct peaks (Fig. 1C). On the other hand, taking into account that mAb-9 weakly cross-reacts with CRMP-1 and CRMP-3 (Fig. 3A), the second peak could also represent a different CRMP. Interestingly, on size exclusion chromatography CRMP-4 migrated with a velocity corresponding to a molecular size of ϳ200 kDa (Fig. 2, A and C). This obvious discrepancy to the apparent M r of 65 kDa, as determined by SDS-PAGE (Figs. 1C and 2C), could reflect the tendency of CRMPs to form heterotetramers under native conditions (42).
In addition to their heterotetramerization, CRMPs interact with several other proteins. In sensory neurons, CRMP-2 is phosphorylated on Thr-555 by Rho kinase upon stimulation of growth cone collapse by lysophosphatidic acid (43). Although this lysophosphatidic acid-dependent signaling pathway does not depend on semaphorin 3A, activity of phospholipase D2 that is inhibited by CRMP-2 was found to be regulated by semaphorin 3A in PC12 cells (44). Furthermore, semaphorin 3A enhances tyrosine phosphorylation of CRMP-2 and CRMP-5 via Fes/Fps tyrosine kinase (45). Recently, it was shown that CRMP-2 binds to tubulin heterodimers and promotes microtubule assembly (46). Other interactions of CRMPs probably exist since CRMP-2 copurifies with dichlorophenol-indophenol oxidoreductase, aldolase C, and glyceraldehyde-3-phosphate dehydrogenase from adult bovine brain, suggesting complex formation of these proteins (47). All interactions of CRMPs mentioned so far involve proteins that are exposed to the cytosol. In this study, however, we show that CRMPs interact with chondroitin sulfates. Chondroitin sulfates represent an entirely novel category of interaction partners for CRMPs, since they are carbohydrates, and they are assumed to reside mainly in the extracellular space. Similar to heparin, chondroitin sulfates are polyanionic glycosaminoglycans, but they carry a lower density of negative charges.
Interestingly, the elution patterns of CRMPs from heparin and chondroitin sulfate columns differed markedly. Although soluble CRMPs bound completely to a heparin column (Fig. 6) and were eluted quantitatively at moderate ionic strength (300 mM NaCl), indicating a charge-mediated interaction of low affinity, only about 50% of each CRMP bound to chondroitin sulfate columns (Fig. 5). This incomplete binding to CS columns could reflect competition of the abundant endogenous brain-derived CS proteoglycans with the immobilized glycosaminoglycan, as suggested by our immuno-coprecipitation experiments (Fig. 7B). Importantly, the CRMPs were not completely recovered from the CS columns even after stringent washing with buffer containing 2 M NaCl, as evidenced by the presence of CRMPs in eluates obtained with buffer containing the chaotropic salt guanidinium hydrochloride (Fig. 5, lane 4 in each panel). Thus, certain CRMP forms obviously engage in high affinity interactions with chondroitin sulfates that were not observed on heparin and which cannot be attributed solely to ion exchange effects.
Chondroitin sulfates are abundant in the extracellular matrix of the developing brain (1), whereas CRMPs as proteins without secretory leader peptides are assumed to exist well separated in the cytosol (30). The immunohistochemical fine reticular or diffuse staining patterns of CS in the marginal zone and subplate of the cerebral cortex (Fig. 8, B and D) are in agreement with published data (9,48,11) and presumably represent extracellular localizations of the glycosaminoglycan. No cytoplasmic CS was detected in the present study, and in the literature, evidence for cytoplasmic proteoglycans in the CNS is limited to the adult stage (49). On the other hand, CRMP-4 was found (i) in cells of the cortical plate mainly in nuclei, (ii) in the marginal zone with a reticular staining pat-tern very similar to the CS-staining, and (iii) in the subplate and prospective white matter (Fig. 8, A and C). The nuclear localization of CRMP-4 is consistent with reports on the targeting of a GFP-CRMP-1 fusion protein to nuclei in lung cancer cells (50) and of CRMP-2-positive nuclear inclusions after overexpression in Neuro2A cells (51). The similar reticular staining pattern of CS (Fig. 8, B and D) and CRMP-4 (Fig. 8, A and C) in the marginal zone, however, suggest that CRMP-4 may be present in extracellular compartment of the cerebral cortex. Taking into account that some proteins lacking a secretory leader peptide are targeted extracellularly (e.g. bovine fibroblast growth factor in MG-63 cells (52)), release of CRMPs to the extracellular space may occur under particular circumstances. Obviously, it is difficult to prove rigorously that a protein is located in the extracellular space based on immunohistochemical analysis of brain tissue. An important argument for the existence of extracellular CRMPs is the presence of these proteins in cell culture supernatants of NTera-2 precursor cells (Fig. 9A) and of neocortical neurons (Fig. 10, F and G). CRMPs were detected in the extracellular compartment, concomitant with cell death, as judged from the release of the cytosolic marker lactate dehydrogenase in the NTera-2 precursor cultures (Fig. 9C) and the presence of condensed, pyknotic chromatin ( Fig. 10, C, D, and E) in ϳ50% of the primary neocortical neurons, indicating naturally occurring (programmed) cell death. Naturally occurring cell death is a widely distributed phenomenon during the development of the central nervous system (33) and is found in the perinatal rat neocortex mainly in the future layers II/III (32,34). In these layers we found particularly strong cellular CRMP-4 staining (Fig. 8, A and C) and numerous pyknotic nuclei (Fig. 8D, arrowheads). Moreover, the non-cellular reticular CRMP-4 immunoreactivity in the marginal zone toward the pial surface and in the cortical plate is consistent with the formation of diffusion gradients of released CRMP-4 away from the zone of neuron death. Binding to CS proteoglycans may help to shape and stabilize gradients of diffusible CRMPs (53). On the other hand, CS could participate in the control of CRMP release from dying neurons since it inhibits the death of neocortical neurons in vitro (11). Thus, CS could be an important regulator of the release and distribution of extracellular CRMPs in the cerebral cortex. On the other hand, collapsin/semaphorin 3A could regulate the release of CRMPs from dying cells since it has been shown to promote the apoptosis of certain neuron classes (54).
The function of extracellular CRMPs could relate to the establishment of contacts with afferents, since the spatiotemporal pattern of naturally occurring cell death in the neocortex correlates with the arrival and settlement of cortical afferents at the different cortical levels (32). Interestingly, Emerling and Lander (17) obtained evidence that CS-bound soluble cues dramatically influence the growth of thalamic neurites within the cerebral cortex. Thus, it is tempting to speculate that CRMPs may belong to these CS-bound cues. However, screening for functional effects of purified CRMP applied in cell culture paradigms and careful analysis of CRMP transgenes will help to clarify the as yet unknown physiological roles of extracellular CRMPs in the future.