Phosphacan Short Isoform, a Novel Non-proteoglycan Variant of Phosphacan/Receptor Protein Tyrosine Phosphatase-β, Interacts with Neuronal Receptors and Promotes Neurite Outgrowth*

Phosphacan, one of the principal proteoglycans in the extracellular matrix of the central nervous system, is implicated in neuron-glia interactions associated with neuronal differentiation and myelination. We report here the identification of a novel truncated form of phosphacan, phosphacan short isoform (PSI), that corresponds to the N-terminal carbonic anhydrase- and fibronectin type III-like domains and half of the spacer region. The novel cDNA transcript was isolated by screening of a neonatal brain cDNA expression library using a polyclonal antibody raised against phosphacan. Expression of this transcript in vivo was confirmed by Northern blot hybridization. Analysis of brain protein extracts reveals the presence of a 90-kDa glycosylated protein in the phosphate-buffered saline-insoluble 100,000 × g fraction that reacts with antisera against both phosphacan and a recombinant PSI protein and that has the predicted N-terminal sequence. This protein is post-translationally modified with oligosaccharides, including the HNK-1 epitope, but, unlike phosphacan, it is not a proteoglycan. The expression of the PSI protein varies during central nervous system development in a fashion similar to that observed for phosphacan, being first detected around embryonic day 16 and then showing a dramatic increase in expression to plateau around the second week post-natal. Both the native and recombinant PSI protein can interact with the Ig cell adhesion molecules, F3/contactin and L1, and in neurite outgrowth assays, the PSI protein can promote outgrowth of cortical neurons when used as a coated substrate. Hence, the identification of this novel isoform of phosphacan/receptor protein tyrosine phosphatase-β provides a new component in cell-cell and cell-extracellular matrix signaling events in which these proteins have been implicated.

Differentiation and morphogenesis of neural tissues involve a diversity of interactions between neural cells and their environment (1). Although the organization of the extracellular matrix (ECM) 1 in the vertebrate central nervous system (CNS) is not well understood it is marked by the relative abundance of chondroitin sulfate proteoglycans (CSPGs) and hyaluronan (2,3). Previously we have characterized DSD-1-PG, one of the more abundant of the soluble CSPGs in the post-natal brain, showing this to be the mouse homolog of phosphacan and demonstrating that it can have opposing effects upon neurite outgrowth according to the neuronal lineage (4,5).
Although phosphacan occurs in the CNS as a large CSPG (Ͼ800 kDa) (4); it is in fact a secreted splice variant of an even larger transmembrane receptor protein tyrosine phosphatase (RPTP), RPTP-␤ (6), also known as PTP-(-zeta) (7). Hence phosphacan corresponds to the entire extracellular part of the long RPTP-␤ receptor (the relative structures of the proteins are illustrated in Fig. 1). These proteins are characterized by a carbonic anhydrase-like (CA) domain at their extracellular N terminus. A third splice variant, the short RPTP-␤ receptor form, contains the same intracellular phosphatase domains as the long receptor, but, unlike phosphacan and the long RPTP-␤, it is distinguished by the absence of an 850-amino acid highly glycosylated sequence, the IS region, which possesses many of the predicted glycosaminoglycan (GAG) attachment sites.
In vitro functional studies have provided evidence for such roles in the development and maintenance of the CNS, and there have been a number of reports of the effects of phosphacan on process outgrowth from neuronal cultures. These have shown that the proteoglycan can either promote or inhibit neurite outgrowth dependent upon the neuronal type tested and the conditions under which it is presented. For example, we have shown previously that phosphacan has outgrowthpromoting properties on mesencephalic and hippocampal neurons (4), whereas it is inhibitory for laminin-promoted outgrowth from neonatal dorsal root ganglion explants (5). * This work was supported in part by the CNRS, the German Research Council (DFG, Fa 159/11-1,2,3), the International Spinal Research Trust, and the Association pour la Recherche contre le Cancer. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) AJ428208.
§ Awarded a Poste rouge from the CNRS during part of this work. To whom correspondence should be addressed. Tel. Biochemical studies have demonstrated a number of potential binding partners for phosphacan, both in the ECM, such as the tenascins, TN-C (17,18) and TN-R (19), and on the cell surface, such as the immunoglobulin cell adhesion molecules (IgCAMs), F3/contactin (20), and NrCAM (21). In addition, a number of growth factors can bind to phosphacan, including bFGF (22) and HB-GAM (19). Interaction sites on phosphacan have been characterized at three levels (13). First, there are the long negatively charged CS GAG polymer chains, which can bind, for example, to TAG-1/Axonin-1 (23), and the cytokines, amphoterin (19), midkine (24), and HB-GAM/pleiotrophin (19,25). Second, on the core glycoprotein, there are other classes of glycosylation, represented by oligosaccharides carrying epitopes such as HNK-1 (sulfated glucuronic acid) and Lewis-X (sialyl-Lewis-X) (5), which have also been implicated in IgCAM interactions (17). Third, there is the primary protein sequence, which includes the CA, and a fibronectin type III-like (FNIII) domain, but the remaining 80% of the molecule bears no strong homology to other characterized protein data base structures.
Here we report the identification and characterization of a novel short isoform of phosphacan. This new isoform corresponds to the first third of phosphacan, and although it is glycosylated, it is not a proteoglycan. It is not as readily extracted from brain tissue as phosphacan, perhaps because of stronger interactions with the cell surface. A recombinant protein corresponding to the new isoform sequence can, like phosphacan, also promote neurite outgrowth from cortical neurons. Hence regulated expression of this protein may introduce an extra level of complexity to the proposed functional interactions of phosphacan and the RPTP-␤ receptor forms during CNS development and regeneration.

MATERIALS AND METHODS
Antibodies-Monoclonal antibody (mAb) 473HD is a rat IgM antibody directed against the chondroitin sulfate DSD-1 epitope (4,26), 324 is a rat IgG against the cell adhesion molecule L1 (27), and 412 is a rat IgG recognizing both sulfated and non-sulfated epitopes of the HNK-1 carbohydrate (28). Mouse mAb 4 -121 against chick F11 cross-reacts with mouse F3/contactin (29) and was provided by F. Rathjen (Max Delbruck Center for Molecular Medicine, Berlin, Germany). 3F8 is a mouse mAb against rat phosphacan (9,30) that cross-reacts with mouse phosphacan (5), and it is available from the Developmental Studies Hybridoma Bank (Department of Biological Sciences, University of Iowa, Iowa City, IA 52242).
Antibody Screening of cDNA Expression Libraries-A mouse brain cDNA expression library was screened using the polyclonal antibody, KAF13, at 1 g/ml as described previously (5). The library used was BALB/c neonatal whole brain oligo(dT) and random-primed Uni-ZAP XR l (Stratagene). Screening 3 ϫ 10 6 recombinant phages, yielded four positive clones, corresponding to bases 1144 -1509 and 1032-3591 of phosphacan/long RPTP-␤ and 1186 -7850 of short RPTP-␤. The fourth cDNA clone contained the truncated ORF corresponding to a shorter form of phosphacan. This begins at base 1276 and ends with a termination codon at 1792. This is followed by a 3Ј-UTR of 1988 bases. A Marathon cDNA amplification kit (Clontech, Heidelberg, Germany) was used to generate P0 and P7 total mouse brain cDNA libraries from 1 g of poly(A) ϩ RNA. Using PCR amplification of this cDNA, bands from the ORF to 3Ј-UTR were cloned and sequenced (primers were as follows: 1 (sense), 5Ј-ATG CGA ATC CTG CAG AGC TTC CTC; 1700 (sense), 5Ј-GCC TCC TTA AAC AGT GGC T; 1891 (antisense), 5Ј-CTA TCC AGC TGA AGA GTC ATC GGC; 2120 (antisense), 5Ј-ACT TAG AAC TGG TGC GGA C).
Northern Blot Analysis-Poly(A) ϩ RNA was prepared from 1 g of mouse brain from different developmental stages using a Fast-Track 2.0 mRNA isolation kit (Invitrogen). It was separated on formaldehyde/1% agarose gels and then blotted by capillary action onto Hybond nylon membrane (Amersham Biosciences). Blots were pre-hybridized for 1 h and then hybridized overnight at 50°C for DNA probes and at 65°C for riboprobes. Prehybridization was in hybridization buffer without probe (7% SDS, 50% deionized formamide, 5ϫ SSC, 50 mM sodium phosphate, pH 7, 0.1% N-lauroyl sarcosine, 2% blocking reagent (Roche Applied Science)). Washing for DNA probes was in three changes of wash buffer (40 mM sodium phosphate, pH 7.2, 1% SDS, 1 mM EDTA) at 68°C for 40 min, and for riboprobes, at 65°C, 2 ϫ 5 min in 2ϫ SSC (where 1ϫ SSC ϭ 0.15 M NaCl, 15 mM sodium citrate, pH 7), and in 0.1ϫ SSC, 3 ϫ 15 min. Signals were revealed with Hyperfilm (Amersham Biosciences) and amplifying screens at Ϫ80°C. The DNA probe for the 5Ј-ORF of all RPTP-␤ isoforms (1-1785; see Fig. 1) was labeled with [ 32 P]dCTP by random priming. Because of homology in the 3Ј-UTR with the complementary 28 S rRNA sequence, it was not possible to employ a DNA probe without a strong cross-reaction with rRNA, because sequences from the sense strand hybridize to 28 S rRNA. Hence, it was necessary to employ an antisense riboprobe that only hybridized to the sense strand of the PSI transcript. This corresponded to bases 2653-3784 of the 3Ј-UTR of the PSI transcript and was synthesized with incorporation of radioactive ␣-35 S-UTP (Amersham Biosciences) using T7 RNA polymerase (MBI Fermentas GmbH, St Leon Rot, Germany) and the linearized pBluescript II plasmid (Stratagene) containing the PSI cDNA. A riboprobe against ␤-actin was used as an indicator of relative RNA concentrations (bases 739 -970).
Tissue Fractionation-P7 mouse brains were homogenized in different buffers to test the efficiency of different extraction conditions. Protease inhibitors were used throughout (1 mM phenylmethylsulfonyl fluoride, 1 M trypsin inhibitor, 0.1 M Aprotinin, 1 M pepstatin, 5 M leupeptin, 1 M antipain, 1 mM benzamidine, 1 mM EDTA). Initially, 20 brains were homogenized in PBS on ice with a mechanical dounce. Nuclei and insoluble debris were removed by centrifugation at 600 ϫ g for 20 min. Half of this supernatant was detergent-solubilized by addi-FIG. 1. Schematic comparison of the phosphacan short isoform protein with the known forms of phosphacan/RPTP-␤. All four isoforms share a common N-terminal sequence including the carbonic anhydrase (residues 38 -292) and fibronectin type III (312-401) domains. Although the other forms contain the S region between the FNIII domain and the splice site for the short RPTP-␤ (402-762), the PSI protein ends 196 residues after the FNIII domain. Phosphacan and the long RPTP-␤ receptor form in addition contain the IS region with the GAG attachment sites. Both receptor forms possess a single-pass transmembrane domain and two cytoplasmic protein tyrosine phosphatase domains. pAb KAF13 (anti-phosphacan) recognizes extracellular epitopes common to all four isoforms. The predicted molecular masses of the proteins based on their primary amino acid sequences are indicated. However, these proteins are subject to extensive post-translational glycosylation, e.g. phosphacan occurs as a CSPG with a molecular mass of Ͼ800 kDa (4). aa, amino acids. tion of 1% Nonidet P-40 and mixing at 4°C for 1 h. The remaining 600-g PBS supernatant was ultracentrifuged at 100,000 ϫ g for 1 h at 4°C to precipitate PBS-insoluble material including total cellular membranes. This PBS-insoluble pellet was subsequently resuspended in 1% Nonidet P-40/PBS. As an alternative to detergent-solubilization with Nonidet P-40, 10 brains were homogenized in 60 mM n-octyl-␤-D-glucopyranoside, 50 mM Tris, pH 8, 50 mM sodium acetate and then centrifuged at 100,000 ϫ g for 1 h at 4°C. Finally, for urea extraction, 10 brains were homogenized in 8 M urea, 10 mM sodium acetate, pH 6, and then centrifuged at 100,000 ϫ g for 1 h at 4°C. Protein determination used the DC assay kit (Bio-Rad).
Deglycosylation . The homogenate was centrifuged at 20,000 ϫ g for 15 min at 4°C and then ammonium sulfate was added to the supernatant to a final concentration of 1.5 M (35% saturation). Precipitated proteins were removed by centrifugation at 20,000 ϫ g for 1 h at 4°C followed by filtration through 0.45-m filter units. This supernatant was loaded onto a methyl HIC (hydrophobic interaction chromatography) column (Bio-Rad) in a Biologic Duo-Flow gradient chromatography system (Bio-Rad), equilibrated with 1.5 M AmSO 4 / PBS. The column was washed until the baseline was attained and then eluted with a linear gradient from 1.5 to 0 M AmSO 4 /PBS. The 90-kDa protein eluted at around 1 M AmSO 4 . The eluate was diluted by addition of 2ϫ binding buffer (20 mM Tris, pH 7.4, 0.5 M NaCl, 1 mM MnCl 2 , 1 mM CaCl 2 ) and then loaded onto a concanavalin A-Sepharose affinity chromatography column (Amersham Biosciences). The column was washed with 20 volumes of binding buffer before being eluted with elution buffer (0.5 M methyl-␣-D-mannopyranoside, 20 mM Tris, pH 7.4, 0.5 M NaCl). Sample buffer was added to this eluate, which was then separated on 10% SDS-PAGE. The proteins were transferred by electroblotting onto ProBlott PVDF membrane (Applied Biosystems). Bands were revealed by Ponceau Red coloration, and the 90-kDa band was confirmed by alignment with an adjacent strip, which was revealed using immunodetection with pAbs KAF13 and anti-PSI. The N-terminal protein sequence of the 90-kDa KAF13-positive band was determined by automatic Edman degradation on an Applied Biosystems 473A microsequencer.
Developmental Expression Blot-Whole brains from mice at different developmental stages were homogenized on ice in PBS with protease inhibitors. The homogenate was centrifuged at 600 ϫ g for 20 min at 4°C and then the supernatant was further centrifuged at 100,000 ϫ g for 1 h at 4°C. The pellet was resuspended in PBS/20 mM n-octyl-␤-Dglucopyranoside with protease inhibitors.
Preparation of Recombinant PSI Protein-The recombinant PSI protein was generated by expression in Escherichia coli of a plasmid vector containing the PSI cDNA sequence fused in-frame to GST. The PSI cDNA was amplified by PCR using primers 5Ј-ATG CGG ATC CTG CAG AGC TTC C (which introduces a BamHI restriction site just after the initiation codon by changing the "A" at the sixth base into a "G") and 5Ј-CTA TCC AGC TGA AGA GTC ATC GGC (which includes the termination code in the PSI transcript). The PCR cDNA product was purified from an agarose gel and digested with BamHI before being ligated into the plasmid vector, pGEX-5X-3 (Amersham Biosciences), which had been pre-digested with the restriction endonucleases, BamHI and SmaI. Competent E. coli cells (TOP10FЈ; Invitrogen) were transformed, and clones were selected. The correct insertion of the PSI cDNA sequence into the vector was verified by DNA sequencing of the entire construct. For recombinant protein expression, the plasmid-bearing bacterial clone was grown in LB broth supplemented with 50 g/ml ampicillin at 37°C to an A 595 nm of 0.8, when protein expression was induced by addition of isopropyl-1-thio-␤-D-galactopyranoside to a final concentration of 0.05 mM, and the culture was transferred to 24°C for a further 3 h. Bacteria were pelleted and then lysed in PBS/1% Nonidet P-40 with 50 g/ml lysozyme. The PSI-GST recombinant protein was purified by affinity chromatography using a HiTrap glutathione column (Amersham Biosciences). The GST control protein was produced in the same way using bacteria transformed with the same plasmid vector but without insert.
GST Pull-down-Purified GST and GST-PSI protein was incubated with glutathione-agarose beads (2 g/10 l pre-swollen beads; Amersham Biosciences) in PBS/1 mM dithiothreitol/0.5% Nonidet P-40 for 1 h at room temperature with mixing and then washed three times with PBS. 20 l of beads were incubated with the 1 mg (2 g/l) P7 PBSinsoluble 100,000 ϫ g brain fraction in PBS, 1 mM dithiothreitol, 0.5% Nonidet P-40, 0.5% n-octyl-␤-D-glucopyranoside, 1 mM EDTA, plus protease inhibitors at room temperature with mixing for 1 h, and then placed on ice for 30 min before washing with 3 ϫ 1 ml of ice-cold PBS. The beads were finally boiled in SDS-PAGE sample buffer, and the proteins were separated on 10% SDS-PAGE gels before transfer onto PVDF membrane.
Phosphacan Purification-Phosphacan was purified from detergentfree physiological saline-buffered brain lysates from post-natal day 7 to 14 mice as described previously (4) using a combination of affinity chromatography with the mAb, 473HD, bound to Sepharose resin and anion-exchange chromatography (4). It was quantitated using the protein assay (Bio-Rad) and by the determination of uronic acid equivalents (33).
Cell Culture-Neuronal cell cultures were established from embryonic day 17 (E17) mouse brains. The cerebral hemispheres were dissected in PBS, and the meninges were removed. Dissociation was achieved by addition of 0.25% trypsin at 37°C for 10 min, followed by passage through a sieve with 48-m pores. The resulting cell suspension was plated on coverslips at a density of 7,000 cells/cm 2 in minimal Eagle's medium (Invitrogen) supplemented with the N2 mixture, namely 5 g/ml insulin, 20 nM progesterone, 100 m putrescine, 30 nM selenite (34), 1 mM pyruvate, 0.1% (w/v) ovalbumin, and 100 g/ml transferrin (Sigma-Aldrich). The cultures were kept in a humidified atmosphere with 5% CO 2 at 37°C.
Neurite Outgrowth Assays-E17 mouse cortical neurons were plated on supports coated with different substrates; glass coverslips were treated with 15 g/ml poly-L-lysine in 0.1 M borate buffer, pH 8.2, for 1 h at 37°C. The coverslips were then washed with water and dried. Purified phosphacan (50 g/ml uronic acid equivalents) or recombinant proteins (50 g/ml) were coated for 4 h at 37°C. After coating, the coverslips were washed three times with PBS. After 24 h of culture, neurons were fixed with 4% paraformaldehyde for 15 min. After permeabilization with 0.1% Triton X-100 for 3 min and a blocking step with 3% bovine serum albumin in PBS, cells were stained with a mAb directed against the neuronal marker, ␤3-tubulin (mouse; Sigma; 2-h incubation in 3% bovine serum albumin/PBS) and then revealed with a cy3-conjugated antibody (goat anti-mouse-cy3; Jackson Immunoresearch Laboratories; 30-min incubation in 3% bovine serum albumin/ PBS). The coverslips were then mounted on slides in Mowiol. Neurons were observed using a DMRB microscope (Leica), and pictures were taken with an axiocam camera (Zeiss). A first parameter considered was the fraction of process-bearing cells (with at least a process longer than one neuronal cell body) from at least 100 neurons per experiment, chosen at random. Subsequently the neurite lengths of the longest neurite on each process-bearing neuron was measured, and morphometric analysis was performed using the ImageTool soft-ware (University of Texas Health Science Center, San Antonio, TX). Three independent experiments were performed, and the data were analyzed with the Kyplot software (Kyence Inc.) using one-way analysis of variance statistics (␣ ϭ 0.05), followed by a Tukey test.

Antibody Screening of Brain cDNA Expression Libraries Yields a Novel Splice Variant of Phosphacan/RPTP-␤-
The cDNA corresponding to phosphacan was cloned previously (5) from a mouse cDNA expression library using KAF13, a pAb raised against the entire proteoglycan purified from post-natal mouse brain. In addition to clones corresponding to phosphacan and the long and short receptor forms of RPTP-␤, the antibody also recognizes a clone expressing a novel truncated form of the extracellular domain.
In this cDNA clone, the ORF is identical to the other three isoforms up to base ϩ1787 but then terminates at ϩ1792 with a stop codon (Fig. 2). It encodes a protein of 597 residues. The ORF is followed by a 2-kb 3Ј-UTR that has homology with rRNA sequences on the complementary strand and finishes with a putative polyadenylation signal. This 3Ј-UTR is different from those for phosphacan and for the two RPTP-␤ receptor transcripts. The 5Ј-end of the clone was confirmed by reverse transcriptase PCR analysis employing primers from the 5Ј-end and from the 3Ј-UTR to generate the complete ORF region.
By comparison with the three known splice variants of RPTP-␤, the truncated protein, which we designate PSI, represents 37% of phosphacan and the extracellular part of the long RPTP-␤ receptor form and 78% of the extracellular portion of the short RPTP-␤ receptor (Fig. 1). The PSI protein shares the same N-terminal region, with the signal peptide (first 23 residues) followed by the CA and FNIII domains present in the other forms plus an additional 188 residues of the S region. This latter region lies between the FNIII domain and the splice site for the short receptor form, which occurs at residue 762.
The expression of the PSI transcript was investigated by Northern blot analysis of poly(A) ϩ RNA from E15 and P0 mouse brains, detected with probes against the ORF and 3Ј-UTR regions (Fig. 3). In addition to previously characterized bands at around 6.5, 8, and 9.5 kb, corresponding, respectively, to the short RPTP-␤ receptor, phosphacan, and the long RPTP-␤ receptor transcript (35), a band is detected with the ORF probe at around 4 kb, which as predicted is also hybridized by the probe against the 3Ј-UTR of the PSI transcript.
In Vivo Expression of a Protein Corresponding to the Phosphacan Short Isoform Transcript-Previously, we have worked with phosphacan, the soluble secreted proteoglycan isoform of RPTP-␤. This protein can be readily extracted from brain tissue using detergent-free physiological buffers (4). It was originally characterized by the presence of a chondroitin sulfate epitope, DSD-1, which is recognized by the mAb, 473HD (hence the original name, DSD-1-PG). As is shown in Fig. 4a), most of the 473HD-bearing proteoglycan is present in the PBS-soluble 100,000 ϫ g supernatant (molecular mass Ͼ 500 kDa), although there is also material in the PBS-insoluble 100,000 ϫ g pellet fraction, which contains all of the cell membranes. This protein band corresponds to the transmembrane, long RPTP-␤ receptor form, which contains the whole of the phosphacan sequence in its extracellular half and is also a CSPG. Analysis of these brain extracts with an antibody against TN-C (Fig. 4c), another major component of the neural ECM, reveals a similar profile, the majority of the secreted glycoprotein being extracted with PBS.
KAF13, the anti-phosphacan pAb employed for the expression cloning of the PSI cDNA, was used to analyze Western blots of these brain extracts for the corresponding protein (Fig.  4b). Phosphacan and the long RPTP-␤ are visualized as large proteoglycan species (Ͼ500 kDa) in the PBS-soluble and insoluble fractions, respectively. Protein species around 250 and 200 kDa are present in all fractions, although these are considerably larger than the size of the predicted PSI protein (67 kDa), and because they are present in the PBS-soluble fraction, they are unlikely to correspond to the short RPTP-␤ transmembrane receptor form. Instead it has been suggested that they represent less glycosylated forms of the phosphacan core protein (36). Based on its amino acid sequence, the PSI protein should be secreted into the extracellular environment, but there was no evidence of a smaller protein species Ͻ180 kDa recognized by pAb KAF13 in this fraction. However, a smaller KAF13positive protein band around 90 kDa was detected in brain tissue extracted with detergent and urea-containing buffers (Fig. 4b, open arrow). This 90-kDa protein could be extracted with the detergents, Nonidet P-40 or octylglucoside, but was absent from the PBS-soluble fraction, being precipitated by centrifugation at 100,000 ϫ g. It was also possible to extract this 90-kDa protein with urea, suggesting that it was associated with the PBS-insoluble fraction, which includes the cell membranes, through non-lipid molecular interactions. Sequence analysis using several computer logarithms (e.g. Prosite (37) and Glycosylphosphatidylinositol Modification Site Prediction (38)) did not indicate the presence of any known lipid binding domains in the PSI protein. In addition, F3/contactin, a receptor protein that is associated with the cell membrane through lipid interactions (glycosylphosphatidylinositol anchor) was also found in the detergent extracts and the PBSinsoluble 100,000 ϫ g fraction, but, unlike the 90-kDa protein, it was not found in the urea extract (Fig. 4d).
The 90-kDa PSI Protein Is Glycosylated, but It Is Not a Proteoglycan-To further test whether the 90-kDa protein corresponds to the PSI cDNA, the presence of glycosylation was investigated using glycosidases (Fig. 5). To determine the presence of GAG chains, protein fractions were treated with GAG lyases. Digestion of the PBS-insoluble 100,000 ϫ g fraction with chondroitinase ABC (ChABC) removes the CS GAGs from the CSPGs. The "core protein" of the long RPTP-␤ receptor form, recognized by pAb KAF13 (indicated by a bar in Fig. 5), now migrates as a smaller, although still heavily glycosylated, species around 300 -400 kDa. The mAb, 3F8, recognizes an epitope on phosphacan and the long RPTP-␤ receptor (but not on a keratin sulfate variant of phosphacan, called phosphacan-KS (6,9,30) or on the short RPTP-␤ receptor (39)) and detects both the undigested long RPTP-␤ receptor at the top of the separating gel and the core protein of long RPTP-␤ following digestion of the CS GAGs with ChABC (presumably the 3F8 epitope is to some extent masked by the CS GAGs, because the signal is much stronger for the core protein), but it does not detect any other protein in this fraction. Hence, the PSI protein does not seem to bear the 3F8 epitope. Keratanase digestion (to remove keratin sulfate GAG chains) does not appear to result in any significant deglycosylation of the long RPTP-␤ receptor. Meanwhile the 90-kDa protein (arrow) detected with KAF13 shows no shift on either ChABC or keratanase treatment, and hence, as predicted for PSI, it does not appear to be a proteoglycan. However, the presence of other carbohydrate modifications on PSI is probable, because the N-terminal region of phosphacan is glycosylated (17), and there is a difference between the migration of the 90-kDa protein on SDS-PAGE and the calculated size from the primary PSI sequence (around 67 kDa). To test for the presence of N-linked carbohydrates, the membrane fraction was treated with peptide N-glycosidase F. As expected, this results in a significant reduction in the size of the 90-kDa protein detected by KAF13 by around 15-20 kDa, indicating that the 90-kDa band is indeed a glycoprotein (Fig.  FIG. 2. Nucleotide sequence of phosphacan short isoform cDNA. The 3.8-kb sequence is shown together with the deduced 597-amino acid sequence of the 1.8-kb open reading frame; * indicates the stop codon, and the underlined sequence indicates the potential polyadenylation site. The schema below illustrates the fragments used to compile the sequence; Ͼ and Ͻ are the PCR primers, and the dotted lines represent the sequences used to make the probes used in the Northern analysis. The accession number for this sequence in the EMBL data base is AJ428208. 5). This indicates that without sugars, the 90-kDa protein migrates at around 70 kDa, which would be the expected size of the PSI protein sequence. A weaker band around 190 kDa is also shifted down by N-glycosidase F treatment and may correspond to the short RPTP-␤ receptor form, which is not a proteoglycan and does not carry the 3F8 epitope but does contain the entire PSI sequence in its extracellular part.
To further confirm the identity of the 90-kDa protein, a new polyclonal antisera was raised against a bacterially expressed recombinant PSI protein. This anti-PSI sera also clearly recognized the 90-kDa protein detected by the anti-phosphacan sera, KAF13 (Fig. 6a). In addition, it appears that the HNK-1 carbohydrate epitope is also present on the 90-kDa protein. This sugar has been found on a range of extracellular proteins in the CNS, including CAMs, TN-C and TN-R, and we have previously shown it to be present on phosphacan (5). As further confirmation of the recognition of the 90-kDa HNK-1 positive protein by the anti-phosphacan and anti-PSI sera, immunoprecipitation experiments were performed. These showed that both the anti-phosphacan and anti-PSI sera recognized the 90-kDa protein precipitated from brain extracts by the anti-HNK-1 mAb and that the 90-kDa band precipitated by both anti-PSI and anti-phosphacan carried the HNK-1 epitope (Fig.  6b). Hence, the 90-kDa band precipitated by HNK-1 is recognized by both anti-phosphacan and anti-PSI, whereas HNK-1 recognizes the 90-kDa protein precipitated by both of these polyclonal sera.
Purification and Protein Analysis Confirms the Identity of the 90-kDa Band as Phosphacan Short Isoform-To confirm the identity of the 90-kDa protein band, it was further enriched from brain detergent extracts, and protein sequence was obtained. Given the presence of the other splice variants of phosphacan/RPTP-␤, which share common epitopes with PSI, it was necessary to follow the purification steps by Western blot analysis of fractions using both pAb KAF13 and anti-PSI. Following a series of pilot assays, a relative enrichment of the 90-kDa protein was obtained using a combination of hydrophobic interaction and concanavalin A lectin-affinity chromatography.
The protein was finally separated on SDS-PAGE and transferred to sequencing grade PVDF membrane. Based on alignment with the KAF13-positive band, the 90-kDa protein corresponding to the candidate isoform was excised and subject to Edman degradation, which yielded an N-terminal sequence, XXRQQRKLVEEI. This corresponds to the predicted start of all of the RPTP-␤ splice variants following cleavage of the 23-amino acid signal peptide (6). Thus, in conclusion, we have found a protein that has the biochemical properties predicted for the protein product of the PSI transcript.
Developmental Expression of the Phosphacan Short Isoform Protein-A developmental profile of PSI protein expression in mouse brain indicates that it is already present by E16 and that, like phosphacan (5,40), its expression levels rise steadily to plateau in the first and second weeks post-natal, before decreasing a little in the adult (Fig. 7). This peak of expression correlates with the phase of myelination in the developing CNS, a process in which phosphacan/RPTP-␤ has been impli- FIG. 3. Northern blot analysis. Blot shows whole brain poly(A) ϩ RNA from E15 and P0. This was probed with a 5Ј-ORF probe covering the region common to the four isoform transcripts (left panel) and an antisense riboprobe for the 3Ј-UTR of the PSI transcript (right panel). In addition to the bands for phosphacan and the two RPTP-␤ receptor forms, there is an additional band around 4 kb that is hybridized by both the 5Ј-ORF and the 3Ј-UTR probes. A ␤-actin probe was employed as a marker of relative RNA concentration (bottom).
cated (12,41,42). There are several other bands seen on the blot at higher molecular masses (around 180 kDa and Ͼ400 kDa). These could correspond to the receptor forms of RPTP-␤ that may also show developmental regulation, both with respect to their relative levels of expression and also in terms of the degree of glycosylation present on the core proteins (5,40).
Binding Partners for Phosphacan Short Isoform-The asso-ciation of the secreted PSI protein with the PBS-insoluble 100,000 ϫ g brain fraction suggests that it is complexed with material in this fraction. The 100,000 ϫ g fraction contains all of the cellular membranes, and hence it is possible that PSI is retained in this fraction because of interactions with membrane-bound receptor proteins. A number of neuronal receptor proteins have already been identified for phosphacan, including the IgCAMs, NCAM, NrCAM, L1/NgCAM, (19,21) F3/ contactin (20), and TAG-1/Axonin-1 (23). In addition, studies with recombinant protein constructs corresponding to different parts of the phosphacan protein sequence have shown that the CA domain can bind to F3/contactin (20) and can also be immunoprecipitated as a complex with a third transmembrane protein, Caspr (42), whereas the S region up to residue 630 binds to NCAM, NrCAM, and L1/Ng-CAM (21). Because the PSI protein sequence contains the entire CA and FNIII domains and 188 of the 223 amino acids (84%) of the S region used in the mapping studies, a similar interaction profile might be expected. Immunoprecipitation studies of brain membrane extracts suggest that the PSI protein can indeed be co-precipitated with L1 and F3/contactin (Fig. 8a). Further evidence for such interactions came from binding experiments with a recombinant PSI protein. The PSI protein was expressed as a GST recombinant construct in E. coli and attached to glutathione-agarose beads prior to incubation with mouse brain extracts. After washing the beads, the proteins that had been "pulled down" were assessed on Western blots. These showed that, compared with the GST control protein, there was an interaction of the GST-PSI protein with both F3/contactin and L1 (Fig. 8b). Hence, these results confirm the previous reports of interactions between the N-terminal domains of phosphacan/ RPTP-␤ and the IgCAMs, F3/contactin (20) and L1 (21), and support the possibility that, in addition to phosphacan and the extracellular domains of the receptor forms of RPTP-␤, the PSI protein might also interact in vivo with such cell membrane receptor molecules. In addition, these results indicate that the interactions between the PSI protein and these cell adhesion molecules can occur in the absence of glycosylation. PSI Recombinant Protein and Phosphacan Promote Neurite Outgrowth from Cortical Neurons-One of the most studied functional effects of phosphacan has been its influence upon process outgrowth from neurons, a process that it may influence as an ECM component in differentiating cortical layers and along axon tracts, as well as in CNS lesions. It has been FIG. 5. Deglycosylation studies of brain protein extracts. The PBS-insoluble, 100,000 ϫ g pellet fraction from Fig. 4 was digested with chondroitinase ABC, N-glycosidase F, or keratanase, prior to separation on SDS-PAGE and blotting onto PVDF membrane. The 90-kDa band detected by pAb KAF13 (arrow) is shifted down by N-glycosidase F treatment. A band around 190 kDa (circle) corresponds to the short RPTP-␤ receptor and is also shifted down by N-glycosidase F treatment. The core protein of the long RPTP-␤ receptor at around 350 -400 kDa (bar) is obtained following ChABC digestion. In the right-hand panel, the mAb 3F8 against an epitope on phosphacan/long RPTP-␤ recognizes the ChABC-digested core protein of long RPTP-␤ (bar) but neither the 90-kDa band nor the short RPTP-␤ receptor.
FIG. 6. The 90-kDa protein is recognized by antibodies against the PSI protein sequence and the HNK-1 carbohydrate epitope. a, Western blot of concanavalin A affinity-purified protein fraction revealed with the pAbs KAF13 and anti-PSI and the mAb 412 against the HNK-1 epitope. The 90-kDa protein (arrow) reacts with all three antibodies and yields an N-terminal sequence corresponding to the PSI transcript. b, immunoprecipitation of the 90-kDa protein (arrow) by anti-PSI, KAF13, and 412 (anti-HNK-1). Shown is a Western blot showing proteins precipitated from the PBS-insoluble 100,000 ϫ g fraction using the indicated antibodies (IP), and revealed using the antibodies indicated below the figure (Blot). Controls were performed without addition of antibody. shown previously (43) that phosphacan can promote outgrowth from cortical neurons and that the N-terminal domains of phosphacan can also promote outgrowth, some of these effects being mediated via the IgCAMs, F3/contactin and NrCAM (20,21). Based upon these studies, it seems likely that PSI can also affect neurite outgrowth. To investigate this, the effects of a recombinant PSI protein on outgrowth from embryonic (E17) mouse cortical neurons were compared with those obtained with the purified phosphacan proteoglycan.
The PSI protein was expressed as a GST recombinant construct and purified using glutathione affinity chromatography. As can be seen in Table I and Fig. 9, there was a significant promotion of neurite outgrowth by the PSI-GST protein relative to both the poly-L-lysine control and the GST protein alone, although the promotion on the phosphacan substrate was more robust. It is possible that the native phosphacan can exert stronger effects than the PSI construct because of the presence of additional promontory signals that may occur either in the much larger phosphacan protein sequence or in the extensive additional glycosylation, including the GAG chains, present on the proteoglycan. Nevertheless, it appears that the PSI protein sequence is capable of promoting neurite outgrowth from cortical neurons and that this effect can be obtained without eukaryotic post-translational modifications. DISCUSSION A New Isoform of Phosphacan: Phosphacan Short Isoform-In this paper, we describe evidence for a new isoform of phosphacan/RPTP-␤. The novel PSI transcript was cloned from a brain cDNA expression library using anti-phosphacan sera, and its expression in vivo was confirmed by Northern blot hybridization. Using the same antisera, we have searched brain extracts for the protein corresponding to the translation product of the PSI transcript and have identified a 90-kDa glycoprotein that fulfils the expected biochemical criteria. The deglycosylated protein has the predicted size and N-terminal peptide sequence. It is also recognized by antisera raised against the recombinant PSI protein and bears the HNK-1 sugar epitope, which is also found on phosphacan. Hence, it appears that the 90-kDa protein does indeed correspond to the product of the PSI transcript.
However, ECM proteins can be subject to proteolysis, and an alternative explanation for the 90-kDa protein we have characterized might be that it is a proteolytic breakdown product of phosphacan/RPTP-␤. There are several reasons, besides the existence of the PSI transcript, why this is unlikely to be the case. For a start, an analysis of the phosphacan protein sequence for potential cleavage sites by known proteases (Pep-tideCutter program (44)) indicates that it could be digested into relatively small peptides, especially in the N-terminal region. In addition, the only reported in vivo demonstration of proteolytic cleavage of phosphacan is for the tissue plasminogen activator and plasmin (45), where it was shown that I 125radiolabelled phosphacan could be digested by pure plasmin in vitro, but no digestion product was observed in the range of 40 to 150 kDa. Moreover, a clear demonstration of in vivo proteolysis of phosphacan compared with transgenic mice null for either tissue plasminogen activator or plasminogen was only observed in the hippocampus following sclerotic changes induced by kainite injections (45). Hence, there does not seem to be a constitutive large scale cleavage of phosphacan under non-pathological conditions, unlike the situation that is readily observed for another prominent CNS-specific CSPG, neurocan. Neurocan has been shown to be subject to a specific processing generating large N-and C-terminal fragments (46). However, this processing is developmentally regulated and can be readily observed both in vivo and in vitro, even occurring in astrocytic cell cultures that have been extensively supplemented with protease inhibitors (47).
In addition, although it has been suggested previously (48) that a 180-kDa protein species that reacts with anti-phosphacan sera may be a proteolytic product, this was because it was not recognized by the mAb 3F8. We observe similar bands migrating around 200 -250 kDa (Fig. 4b), and another explanation might be that the 180-kDa band corresponds to a form of mostly unglycosylated phosphacan (36) in which the 3F8 epitope is either hidden by a small oligosaccharide, or, indeed, it has even been suggested that 3F8 may itself be a sugar epitope (40). In addition, this 180-kDa band was found in cell-conditioned media (48), unlike the 90-kDa band that we have shown to be in the PBS-insoluble fraction. Hence, on balance, it seems unlikely that the 90-kDa protein that we have characterized results from a proteolytic cleavage event.
Molecular Interactions of PSI and Phosphacan-The identification of a novel isoform of phosphacan/RPTP-␤ provides a new component in the cell-cell and cell-ECM signaling events in which these proteins have been implicated. Compared with ECM structures in other tissues, the ECM of the CNS appears to have a relatively "loose" supramolecular organization. As such, many ECM components can be readily extracted from the CNS using physiological buffers, whereas the use of chaotropic reagents such as urea and guanidium hydrochloride is often necessary to dissociate ECM components from structures in other tissues (1). Phosphacan is a large CSPG that can be  quantitatively recovered from brain tissue using detergent-free physiological buffers (4). By contrast, the smaller PSI protein remains associated with the PBS-insoluble material, which includes the total cell membrane fraction. The PSI protein can be solubilized from this fraction using either detergent or urea, suggesting that, although it is strongly associated with this fraction, it is not directly associated with lipids. In addition, there is no evidence of any lipid binding domains in the PSI sequence. Consequently, it appears likely that the PSI protein is bound with PBS-insoluble proteins in this fraction. Based on previous reports of binding partners for phosphacan, several membrane-bound members of the IgCAM superfamily might account for such interactions. Thus, the transmembrane Ig-CAMs, L1/NgCAM, NrCAM, and NCAM-180, and the glycosylphosphatidylinositol-anchored IgCAMs, F3/contactin and TAG-1/Axonin, have all been shown previously to bind to phosphacan (20,21,23,49). The PSI protein comprises the CA and FNIII domains and half of the S region, and as such it could also be involved in the interactions that have already been shown for these parts of phosphacan. Hence, it has been shown that F3/contactin binds to the CA domain (20), whereas NCAM, L1/NgCAM, and NrCAM have all been shown to interact with the S region (21). Indeed of the neuronal IgCAMs found to interact with phosphacan, only TAG-1/Axonin-1, which binds to phosphacan via the CS GAG chains (23), interacts principally with parts of the molecule that are absent from the PSI protein.
Here we demonstrate that both the native and recombinant PSI protein can interact with L1/NgCAM and F3/contactin, and thus it seems likely that both the PSI protein and phosphacan could interact with the same receptor proteins at sites that are common to both isoforms. Therefore, it is possible that where the PSI protein and phosphacan are coexpressed they may compete for common receptors.
The retention of PSI in the PBS-insoluble fraction of postnatal brain under conditions in which phosphacan is readily extracted suggests that in fact the PSI protein may be more tightly bound to such membrane receptors than phosphacan. This weaker association of phosphacan with the membrane fraction could also be because of additional binding interactions with PBS-soluble factors via other sites present on the phosphacan molecule. Hence binding interactions have been described between the CS GAG chains on phosphacan and growth factors such as HB-GAM/pleiotrophin (19,25), midkine (24), and amphoterin (19), and perhaps such molecular binding interactions could modify binding interactions at other sites on the proteoglycan. In addition, the high negative charge and steric bulk of the CS GAG chains may in themselves represent factors that lower the relative affinity of phosphacan for the membrane compared with the PSI protein.
Phosphacan Short Isoform and Neurite Outgrowth Promotion-Previously we have shown that the phosphacan proteoglycan can have opposing effects upon neurite outgrowth dependent upon the neuronal lineage, which may be mediated either via the CS GAG chains (4) or the core glycoprotein (5). Similarly, promotion was found from E16 cortical neurons on a phosphacan substrate that was neutral in the same study for outgrowth from E16 thalamic neurons (43), whereas phosphacan was inhibitory for outgrowth from E17 cerebellar neurons (25) and P6-P8 retinal ganglion cells (50).
In this study, we have found that a recombinant PSI protein can promote neurite outgrowth from E17 cortical neurons. Previous studies have implicated both the GAG chains and the glycosylated core protein (following either enzymatic digestion or as a result of eukaryotic expression of recombinant constructs), but evidence is presented here that, even in the absence of any glycosylation or other eukaryotic post-translational modification, there are interaction sites on the PSI protein that can promote neurite outgrowth.
Studies with recombinant Fc fusion proteins containing the different domains of phosphacan have demonstrated previously (20) that the CA domain can support neuronal adhesion and neurite outgrowth by binding to F3/contactin on the surface of neurons (20), and NrCAM has been shown to promote outgrowth via interactions with the S region. Both of these effects could be blocked by specific antibodies directed against F3/ contactin and NrCAM (20,21). Our co-immunoprecipitation and GST pull-down studies indicate that both the native and recombinant PSI protein can, as predicted, interact with L1/ NgCAM and F3/contactin, and hence it is possible that the outgrowth promotion observed is because of binding interactions between neural IgCAM receptors and the PSI protein.
Phosphacan is subject to extensive glycosylation, with up to 16 predicted N-glycosylation and 41 potential O-glycosylation sites. By comparison, PSI, which lacks the large 850-amino acid IS region and the C-terminal half of the S region, contains only eight of the predicted N-glycosylation and three of the potential O-glycosylation sites, and it is not substituted with GAG chains. Consequently, the range of possible interactions of the PSI protein is reduced compared with phosphacan, and this might explain the weaker outgrowth promotion observed in our study. Nevertheless, it seems that even without carbohydrate modifications, the PSI protein sequence can still promote process outgrowth from cortical neurons.
Phosphacan Short Isoform and Phosphacan as Regulators of the Tyrosine Phosphatase Activity of the RPTP-␤ Receptor Forms-It has been suggested that phosphacan/RPTP-␤ isoforms are involved in a complex in vivo neuron-glia cross-talk via interactions with other receptor molecules and ECM proteins (13). The expression of the PSI protein could also intervene in such processes by competing selectively for ligands. A similar relationship has already been proposed for interactions between the secreted phosphacan proteoglycan and the transmembrane receptor forms of RPTP-␤. Hence extracellular sites common to the different isoforms could compete for similar extracellular ligands and thereby modulate the nature of the cytoplasmic tyrosine phosphatase signal from the RPTP-␤ receptor forms (51). Alternatively, they could interact in either cis or trans with other cell surface receptor molecules, affecting the respective signaling responses of these molecules (21). Thus, our identification of a novel truncated isoform of phosphacan representing 78% of the shorter extracellular region of the short RPTP-␤ receptor protein could be a complementary element to the competitive ligand model that has already been proposed for the larger extracellular domains of phosphacan and the long RPTP-␤ receptor. In addition, although both phosphacan and the long RPTP-␤ receptor are modified by the addition of CS GAG chains, the PSI protein and the short RPTP-␤ receptor are not proteoglycans. Hence, the expression of PSI suggests the possibility that both of the RPTP-␤ receptor forms have their own respective complementary secreted forms and that these could serve as competitive regulators of their ligand interactions and consequently of their intracellular enzymatic activity.
A possible functional interaction between the different isoforms of phosphacan/RPTP-␤ might occur during myelination, because it has been shown that the CA domain, common to all four isoforms, can interact with F3/contactin in a neuronal signaling complex that has been localized in vivo at paranodal junctions between axons and paranodal loops of myelinating glia (42). It has also been shown recently (52) that the recombinant CA domain can block the localization of this complex to the paranodes, suggesting that the complex may be targeted via ECM interactions with phosphacan/RPTP-␤ expressed by myelinating glia. Additionally a significant role for phosphacan/RPTP-␤ in myelination has been suggested by studies on mice deficient for the RPTP-␤ gene that indicate that, although there is no gross abnormality in the overall brain architecture, there is a fragility in the myelin sheaths of CNS neurons (12), and there is impaired recovery from demyelinating lesions in these mice (41).