Direct Interaction of CASK/LIN-2 and Syndecan Heparan Sulfate Proteoglycan and Their Overlapping Distribution in Neuronal Synapses

CASK (CASK-FYG) antibodies and syndecan-2 (Syn-2C) antibodies were raised in rabbits against the synthetic peptide CFYGDPPEELPDFSED, corresponding to amino acids 320–334 of CASK, and the synthetic peptide RKKDEGSYDLGERKPSC, corresponding to amino acids 182– 197 of rat syndecan-2, respectively. The terminal cysteine residues were added for coupling purposes. Both antibodies were affinity-purified using their respective immunogen peptide coupled to a Sulfolink column (Pierce Chemical Co., Rockford, IL). Rabbit anti-human CASK (hCASK) antiserum is described in Cohen et al. (cosubmitted manuscript). Anti-SAP97 and anti-PSD95 CSK antibodies were described in Kim et al. (1996). Anti-Myc mouse monoclonal antibody (9E10) was purchased from Santa Cruz Biotechnology (Santa Cruz, CA). MSE-2 antiserum, a gift from Dr. Merton Bernfield, was raised against the nonconserved extracellular region of syndecan-2, and has been described (Kim et al., 1994).

Rat syndecan-2 cDNA was generously provided by Dr. Graham Cowling (Manchester University, United Kingdom). The coding region of rat syndecan-2 was PCR-amplified and subcloned into the KpnI and EcoRI site of mammalian expression vector GW1-CMV. Two oligonucleotides, one containing an XbaI site, the myc epitope (EQKLISEEDL), and the sequence GACCTTGGAGAACGCAAACCG (corresponding to aa 190– 196 of syndecan-2), and the other containing an XbaI site and the sequence GTAGCTTTCTTCGTCTTTC TT (corresponding to aa 183–189 of rat syndecan-2), were applied in reverse PCR for construction of myc-tagged syndecan-2. The amplified product was digested with XbaI and then religated. The Kv1.4-syndecan-2 chimeric protein (Kv1.4-EFYA) was constructed by PCR using Kv1.4 cDNA as template and a 3′ end primer encoding the last four amino acids (aa) of syndecan-2 and aa 648– 651 of Kv1.4.

CASK cDNA was amplified by reverse transcription PCR from rat brain first-strand cDNA library, and was subcloned into pGW1-CMV. In addition, we obtained CASK cDNA as a gift from Dr. Thomas C. Südhof. For myc-tagged CASK construction, an AscI site was created between aa 599 and 560 of CASK by inverse PCR. A double-stranded oligonucleotide encoding the myc epitope was inserted into the AscI site.

COS-7 cells in 35-mm plates at 50–70% confluency were incubated with a 1-ml OPTI-MEM (GIBCO BRL, Gaithersburg, MD) mixture containing 1.6 μg DNA and 6 μl Lipofectamine (GIBCO-BRL) for 5 h followed by incubation in DMEM medium. 48 h later, cells were harvested and lysed in RIPA buffer. Immunoprecipitation from RIPA cell lysates has been described (Hsueh et al., 1997), except that Myc antibody 9E10-conjugated agarose (Santa Cruz Biotechnology) was used to precipitate myc-tagged proteins directly. Fractionation of brain membranes and PSDs was performed as described by Cho et al. (1992). Immunoblotting with enhanced chemiluminescence reagents was performed as described (Sheng et al., 1993). hCASK antiserum was used at a 1:2,000 dilution; affinity-purified primary antibodies were used at 2 μg/ml (Syn-2C antibody), 0.5 μg/ml (CASK-FYG antibody), 0.2 μg/ml (PSD-95 CSK antibody), 1 μg/ml (SAP97 antibody), or 0.2 μg/ml (anti-Myc antibody 9E10).

Immunohistochemistry was performed on Vibratome-cut (Ted Pella, Inc., Redding, CA) 50-μm floating brain sections from rats (∼6 wk old; Harlan Sprague Dawley Inc., Indianapolis, IN) perfused transcardiacally with 4% paraformaldehyde and permeabilized with either 0.1% Triton X-100 or 50% ethanol. Rat brain sections were incubated with affinity-purified primary antibodies at 2 μg/ml at room temperature overnight. For DAB staining, Vectastain Elite ABC kit (Vector Labs, Inc., Burlingame, CA) was used as described (Sheng et al., 1994). For immunofluorescent staining, Cy3- or FITC-conjugated secondary antibodies (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA) were used at the dilution of 1:1,000 and 1:200, respectively.

Double-label immunofluorescence with two antibodies raised from the same species (rabbit) was performed as described in Shindler and Roth (1996) with some modifications. The brain sections were incubated with a very low concentration (0.05 μg/ml) of Syn-2C antibodies in PBS with 0.5% blocking reagent (TSA direct kit; Dupont-NEN, Boston, MA) at room temperature overnight. After washing three times with PBST (1× PBS with 0.05% Tween-20), the sections were incubated with biotinylated anti-rabbit antibody (Jackson ImmunoResearch Laboratories, Inc.) at 1:1,000 at room temperature for 2 h, and were incubated with streptavidin-conjugated horseradish peroxidase (TSA direct kit; Dupont-NEN ) at 1:100 for a further 30–60 min. Finally, FITC-conjugated tyramide (TSA direct kit; Dupont-NEN) was applied at 1:50 in amplification diluent (TSA direct kit; Dupont-NEN) for 5–10 min. After washing three times with PBST, these rat brain sections were then processed for conventional immunofluorescent CASK staining using Cy-3–conjugated secondary antibodies as described above. At the low concentrations of Syn-2C antibodies used, no Syn-2C signal could be detected with Cy3-conjugated anti-rabbit secondary antibodies. Results were viewed with an Axioskop microscope (Carl Zeiss Inc., Thornwood, NY) or a confocal microscope (MRC-1000; Bio-Rad Laboratories, Hercules, CA), and digitized images were processed for publication using Adobe Photoshop.

For electron microscopy, pentobarbital-anesthetized (60 mg/kg) male Sprague-Dawley rats (250–350 g) were perfused intraaortically with mixed aldehyde fixatives (0.2% glutaraldehyde and 4% depolymerized paraformaldehyde for immunoperoxidase staining; 2% glutaraldehyde and 2% paraformaldehyde for immunogold) in phosphate buffer (PB), pH 7.4, after a brief flush with heparinized saline. The brain was removed and immersed in the same fixative for 2 h; 40-μm sections were cut on a Vibratome and collected in PB. For immunogold, sections were embedded directly into resin (see below). For preembedding immunoperoxidase staining, floating sections in glass vials were treated for 30 min in 1% sodium borohydride in PB, permeabilized for 30 min with 50% EtOH, and then incubated for 10 min in 3% hydrogen peroxide, blocked in 10% donkey serum, and incubated in primary antibody (diluted 1:100–1:1,l000) on a shaker overnight. Sections were then rinsed, incubated in secondary antibody (biotinylated, donkey anti–rabbit; Jackson ImmunoResearch Laboratories, Inc., 1:200) for 2 h, rinsed, incubated in ExtrAvidin (1:5,000; Sigma Chemical Co., St. Louis, MO) for 2 h, rinsed, and reacted with diaminobenzidine to visualize antibody-binding sites. Sections were then poststained 1 h in 1% platinum chloride in maleate buffer (MB) and infiltrated with resin.

Embedding was performed according to modifications of an osmium-free protocol (Phend et al., 1995). In brief, sections were incubated over ice in 0.1% filipin/0.1% digitonin (Sigma Chemical Co.) in MB for 20 min, for 40 min with 1% tannic acid in MB, for 20 min in 0.1% CaCl₂, for 40 min in 1% uranyl acetate, and then for 20 min in 0.5% IrBr₄ (Pfaltz and Bauer, Inc., Waterbury, CT). Sections were dehydrated through graded alcohols and propylene oxide, and then infiltrated with Epon-Spurr resin (Electron Microscopy Sciences, Fort Washington, PA). Wafers prepared with ACLAR plastic (Pelco; Ted Pella, Inc.) were polymerized at 60°C for 24– 36 h. Small chips of the polymerized wafers were glued to plastic blocks; and thin sections cut with a diamond knife were collected on copper mesh grids, poststained with uranyl acetate and lead citrate, and examined with a CX200 microscope (JEOL USA, Inc., Peabody, MA). For immunogold staining, sections were collected on nickel mesh grids and processed according to methods described in Phend et al. (1992). Primary antibody concentrations were 1:100–1:400; the secondary was donkey anti-rabbit serum, conjugated to 18-nm gold particles (Jackson ImmunoResearch Laboratories, Inc.). After immunoprocessing, grids were briefly immersed in 1% glutaraldehyde, rinsed, and poststained as described above.

Two-hybrid screening was done using the L40 yeast strain harboring reporter genes HIS3 and LacZ under the control of upstream LexA-binding sites, as described previously (Kim et al., 1995). The PDZ domain of hCASK (which is identical at the aa level with rat CASK) was subcloned into pBHA (LexA fusion vector), and was used to screen rat and human brain cDNA libraries constructed in pGAD10 (GAL4 activation domain vector; CLONTECH Laboratories, Inc., Palo Alto, CA). COOH-terminal deletion mutants of syndecan-2 were created by PCR and subcloned into pGAD10 to generate GAL-4 activation domain fusions. The interactions of CASK and syndecan were tested in yeast two-hybrid assays by using HIS3 and LacZ as reporter genes (Kim et al., 1995).

To study CASK expression in rat brain at the protein level, we raised antibodies (termed CASK-FYG) against a peptide sequence located between the CaMK and PDZ domains of CASK. The specificity of these anti-CASK antibodies was examined by Western blotting of transfected heterologous cells and brain. CASK-FYG antibodies recognized a dominant band of ∼110 kD in extracts of COS-7 cells transfected with CASK cDNA, but not with vector alone (Fig. 1,A). In rat brain, CASK was detected as an ∼110-kD protein that is mainly associated with membranes, though a significant fraction was also present in soluble fractions (Fig. 1,B). By immunoblotting, CASK is broadly expressed in several brain regions, including neocortex, cerebellum, hippocampus, and brain stem. The same ∼110-kD band in brain and in CASK-transfected COS cells was recognized by an independent antiserum raised against an hCASK fusion protein (Fig. 1 B and data not shown).

The membrane-associated CASK further purified into the postsynaptic density (PSD) fraction, in which it was resistant to extraction by Triton X-100 (PSD I and PSD II) and by Triton X-100 and sarkosyl (PSD III; Fig. 1,C). The degree of CASK enrichment in the PSD fractions, however, was weaker than that seen with PSD-95, a standard marker for PSDs (Fig. 1 C). Nevertheless, enrichment of CASK in purified PSDs showed specificity, since SAP97, an axonal and presynaptic protein (Müller et al., 1995), was depleted in PSDs. These biochemical fractionation studies suggest that CASK is present in the PSD, but that it is not as selectively concentrated in synapses as is PSD-95. This interpretation is supported by immunocytochemical studies (see below).

Affinity-purified CASK-FYG antibodies and anti-hCASK antibodies were used to localize CASK in sections from rat brain. As with Western blotting, these two independent antibodies gave specific staining patterns that were strikingly similar to each other (data not shown), providing strong support that they reveal the true distribution of CASK protein. CASK protein is expressed widely in the brain, and is distributed in a somatodendritic pattern, predominantly in neurons (Fig. 2). In cerebral cortex, pyramidal cells showed strong immunoreactivity in cell bodies and in apical dendrites; nuclei appeared unstained. The apical dendritic staining extended from the proximal trunk to the terminal branches, and was most prominent for layer 5 pyramidal neurons (Fig. 2, A and B). Layer 2/3 pyramidal neurons were also stained, though at lower intensity, and layer 4 cells were relatively spared of labeling. The punctate nature of CASK staining seen on pyramidal cell dendrites is consistent with a synaptic localization (Fig. 2 B), though there was also substantial intracellular immunoreactivity in somata and proximal dendrites.

In the hippocampal formation, CASK protein was present in dentate granule cells and in CA1 and CA3 pyramidal cells in a somatodendritic distribution (Fig. 2, C–F). In addition, however, staining was noted in the mossy fiber tract (Fig. 2,C) where axons from the dentate gyrus granule cells run adjacent to the cell bodies of CA3 pyramidal neurons. CASK immunoreactivity was also prominent in neurons of the thalamus, especially in somata and proximal dendrites (Fig. 2,G). In the cerebellum, the cell bodies and dendrites of Purkinje cells were intensely labeled, and the granule cell layer showed moderate staining (Fig. 2 H). In both forebrain and hindbrain, the neuropil showed signficant CASK staining, whereas white matter tracts were relatively spared.

Higher resolution immunofluorescence confocal microscopy with CASK-FYG antibodies revealed punctate staining on top of a relatively diffuse staining in intracellular compartments and in major dendrites (Fig. 3). Punctate labeling was most prominent along dendritic processes, suggesting a possible synaptic localization of CASK. This subcellular distribution pattern is exemplified in pyramidal neurons of cortex and hippocampus (Fig. 3, A–D). In cerebellum, CASK antibodies also revealed punctate staining in somata and along major dendrites of Purkinje cells (Fig. 3, E and F).

Ultrastructural localization of CASK was obtained by postembedding immunogold EM. The highest density of CASK labeling was found at asymmetric (presumed excitatory) synapses (Fig. 4, A, B, E, and F), but was also seen at symmetric (presumed inhibitory) synapses (Fig. 4, B, C, and D). Quantitative analysis of immunogold distribution showed that the peak of CASK labeling was closely associated with the synaptic membranes, and was roughly symmetrically disposed across the synaptic cleft (see Fig. 8, left), implying the presence of CASK in both presynaptic and postsynaptic compartments close to the membrane. In addition to its association with synaptic membranes and adjacent cytoplasm, CASK was also present at a lower density at nonsynaptic membranes (Fig. 4, B, C, and F; see Fig. 8, right). Cytoplasmic CASK labeling was observed in somata and in dendritic shafts, where it appeared to be often associated with polyribosomes (Fig. 4,B). The presence of CASK in intracellular compartments is also reflected in the particle density plot (Fig. 8, left), which shows a broad major peak with smaller peaks of immunogold density some distance (∼100 nm) from the synaptic membranes. These ultrastructural findings correlate well with the localization of CASK at the LM level.

In summary, CASK is distributed in a somatodendritic pattern in neurons, and is enriched but not exclusively localized in synapses. CASK is also present at nonsynaptic plasma membranes and in intracellular compartments. These immunocytochemical findings are consistent with the biochemical fractionation profile of CASK, which shows incomplete membrane association (Fig. 1,B) and only a moderate enrichment in PSD fractions (Fig. 1 C).

The COOH terminal tail of neurexin has been shown to interact with CASK (Hata et al., 1996), possibly via CASK's PDZ domain. Since PDZ domains in general can bind to multiple partners with compatible COOH-terminal sequences (Kornau et al., 1997; Sheng, 1996; Sheng and Kim, 1996; Songyang et al., 1997) as well as to heterologous PDZ domains (Brenman et al., 1996), we sought to identify other potential ligands for CASK. Using the single PDZ domain of CASK as bait in a yeast two-hybrid screen, three interacting cDNA clones were isolated from brain cDNA libraries (Table I). The most strongly interacting cDNA encoded the COOH-terminal region of a known protein, the transmembrane HSPG syndecan-2 (which terminates in the sequence -EFYA). The remaining two cDNAs encoded the COOH-terminal regions of two proteins of unknown function and with no homology to other sequences in the database, one terminating in -EFQV, and the other ending in -YFDV. Conservation of the COOH-terminal sequence of the interacting clones suggests that this region is involved in binding to the PDZ domain of CASK. This result was confirmed by deletion of the last three residues of syndecan-2, which abolished its interaction with CASK's PDZ in the two-hybrid system (Table I). In the course of this analysis, we also confirmed that the COOH terminus of neurexin can bind to the CASK PDZ domain.

To confirm that full-length CASK and syndecan-2 proteins can associate with each other, we performed coimmunoprecipitation assays in heterologous cells. From COS cells cotransfected with CASK and myc-tagged syndecan-2, myc antibodies were able to precipitate CASK in addition to myc-tagged syndecan-2 (Fig. 5 A). CASK expressed by itself, however, was not immunoprecipitated by myc antibodies. These data indicate that full-length syndecan-2 and full-length CASK can form a complex in a mammalian cell environment.

The results of yeast two-hybrid assays (Table I) indicated that the COOH terminus of syndecan-2 is necessary for binding to the CASK PDZ domain. To test whether the COOH-terminal residues are sufficient for this interaction, we constructed a Kv1.4-EFYA in which the last four amino acids of the K⁺ channel Kv1.4 (-ETDV) are replaced by the corresponding syndecan-2 COOH terminus (-EFYA). In heterologous cells, this Kv1.4-EFYA chimeric protein could be coprecipitated with myc-tagged CASK by CASK-FYG antibodies (Fig. 5,B). In contrast, wild-type Kv1.4 was not coprecipitated with CASK (Fig. 5 B), although it could be efficiently coimmunoprecipitated with PSD-95 (Kim et al., 1996). These results indicate that the COOH-terminal four amino acids of syndecan-2 (EFYA) are sufficient to confer CASK PDZ-binding on a heterologous membrane protein. Since all four known members of the syndecan family terminate with the identical sequence EFYA (Bernfield et al., 1992; Carey, 1997), this finding implies that all four syndecans can bind to the PDZ domain of CASK.

To investigate the putative interaction of syndecan HSPGs and CASK in vivo, we used Syn-2C to localize syndecan-2 in rat brain. At high resolution, the Syn-2C immunostaining pattern in all regions of the brain was notably punctate and suggestive of synaptic labeling (Fig. 6). The specific concentration of Syn-2C immunoreactivity in synapses was revealed by its colocalization with the synaptic vesicle marker synaptophysin in double-labeling studies (Fig. 6). Examples of punctate colocalization of Syn-2C and synaptophysin staining are shown from the stratum lucidum of the hippocampus (Fig. 6,A), the neuropil of the cerebral cortex (Fig. 6,B), the glomeruli of the granule cell layer of the cerebellum (Fig. 6,C), and the glomeruli of the olfactory bulb (Fig. 6,D). In all areas of the brain, cell bodies and major dendrites were relatively spared of Syn-2C labeling (Fig. 6).

To confirm the specificity of Syn-2C staining, we performed immunocytochemistry with a second independent antiserum raised against the extracellular region of syndecan-2. This antibody (MSE-2; Kim et al., 1994) gave a synaptic pattern of staining similar to that of Syn-2C, including the glomeruli of the cerebellar granular layer (Fig. 6 E). The similar staining pattern with two independent antibodies to syndecan-2 provides strong evidence that syndecan-2 is specifically localized in neuronal synapses.

To verify the synaptic localization of syndecan-2 at ultrastructural resolution, we performed EM immunocytochemistry. Type 1 (asymmetric) synapses were specifically labeled by Syn-2C antibodies as revealed by preembedding immunoperoxidase staining (Fig. 7,A). Electron-dense reaction product was prominent at the PSD as well as in the presynaptic axon terminal. Labeling was observed close to the synaptic membrane, and in many cases could also be seen some distance from the synapse, particularly on the presynaptic side. Postembedding immunogold staining confirmed a heavy concentration of Syn-2C immunoreactivity, specifically at asymmetric synapses. By far the highest density labeling was closely associated with the synaptic membranes, but in addition a lower density of particles was detected over the presynaptic axon terminal some distance from the synapse (Fig. 7, B–D, F, and G; see quantitation in Fig. 8). Outside of asymmetric synapses, Syn-2C labeling was present sparsely at nonsynaptic sites of membrane–membrane apposition, and only occasionally within the cytoplasm of dendritic shafts (Fig. 7, D and E). In short, these LM and EM studies indicate that syndecan-2 is highly concentrated at asymmetric (presumably excitatory) synapses of rat brain, apparently on both pre- and postsynaptic sides of the synapse.

Based on subcellular localization of the individual proteins, we reasoned that CASK and syndecan-2 distribution might overlap in brain synapses. This prediction was confirmed by colocalization of these two proteins in rat brain by double-label confocal microscopy (Fig. 9). An example of syndecan-2 colocalization with CASK is in the mossy fiber tract of the hippocampus (Fig. 9,A). Partial colocalization (i.e., overlap) was also seen in a punctate pattern along the dendrites of hippocampal pyramidal neurons (Fig. 9, B and C). The partial overlap occurs because Syn-2C immunoreactivity is specifically localized to synaptic puncta, whereas CASK shows additional extensive staining in somata and major dendrites. Distinctive subcellular distributions of CASK and syndecan with overlapping localization in neuronal synapses is also shown graphically by quantitative comparison of their immunogold distributions (Fig. 8).

We tried to demonstrate biochemical association of CASK and syndecan-2 in brain, but failed to obtain convincing coimmunoprecipitation of these two proteins from detergent extracts of synaptosomal membranes. Several reasons may account for this inability: first, membrane-bound CASK is relatively insoluble and requires harsh detergents for extraction that may lead to disruption of the complex; second, our syndecan-2 antibodies were relatively inefficient at immunoblotting (possibly due to the very heterogeneous size of syndecan-2 protein resulting from glycosaminoglycan side chains); third, CASK almost certainly binds to multiple proteins via its PDZ domain (e.g., other syndecan family members and neurexin), and hence only a fraction of CASK is likely to be complexed with syndecan-2 in the brain. In any case, the degree and significance of the CASK–syndecan interaction in vivo still needs to be confirmed by further biochemical and genetic experiments.

Previous biochemical studies have shown that the CASK protein is enriched in the synaptic plasma membrane fraction of brain (Hata et al., 1996). Our subcellular fractionation extends this observation and reveals that CASK is also enriched in the PSD fraction where it is at least partly resistant to detergent extraction. The synaptic localization of CASK, revealed for the first time by immunocytochemistry, is consistent with its interaction with syndecan-2 (an HSPG shown here to be concentrated in synapses) and with neurexins (which are believed to be at least partly synaptic in distribution; Hata et al., 1996). However, CASK is not exclusively localized in synapses; it is also present at nonsynaptic plasma membranes and in intracellular compartments. CASK may therefore be present at a variety of membrane domains in neurons, and its compartmentalization between membrane and cytoplasm may be dynamically regulated. Components of other cell junctional complexes have been shown to exist in both membrane-associated and cytoplasmic compartments, one example being β-catenin, which is both an adherens junction–associated and a cytoplasmic protein (reviewed in Miller and Moon, 1996).

The interacting clones captured by our two-hybrid screen suggest that the CASK PDZ domain has a binding selectivity for the COOH-terminal tetrapeptide consensus sequence -E-F-X-V/A (Table I). Neurexins, a family of transmembrane proteins previously shown to bind CASK (Hata et al., 1996), terminate in -EYYV, which conforms to this consensus. We show here that the neurexin COOH terminus also binds to CASK via the PDZ domain (Table I). The -E-F/Y-X-V/A consensus sequence, reached here on the basis of yeast two-hybrid assays, is consistent with that derived from peptide library screening using the CASK PDZ domain (Songyang et al., 1997). Both approaches reveal that the CASK PDZ domain selects for a valine or alanine at the COOH terminus (the 0 position) and an aromatic side chain residue (tyrosine or phenylalanine) at the -2 position. Thus, the CASK PDZ domain belongs to the class II PDZ domains, in contrast to the class I PDZ domains of the PSD-95 family of MAGUKs, which prefer a serine or threonine at the -2 position (Songyang et al., 1997). These recognition specificities are in keeping with the fact that in C. elegans LIN-2 does not interact directly with LET-23/EGFR, which terminates with the sequence -ETCL.

We have shown that a tetrapeptide sequence (-EFYA, corresponding to the COOH-terminus of syndecan-2) is sufficient for specific interaction with the PDZ domain of CASK. By extrapolation, any protein that ends in the -E-F/Y-X-V/A consensus sequence may be capable of binding to CASK. Thus, the CASK PDZ domain may have multiple binding partners in vivo, just as the PDZs of PSD-95 can bind to several proteins with the COOH-terminal -E-S/T-X-V motif, including Shaker K⁺ channels, NMDA receptors, and calcium pumps (Kim et al., 1998; Kim et al., 1995; Kornau et al., 1995; Niethammer et al., 1996). Consistent with the idea of multiple partners for CASK is that syndecan-2 distribution overlaps only partly with the distribution of CASK in neurons. Syndecan-2 is specifically localized in synaptic junctions, while CASK is more broadly distributed in synaptic and nonsynaptic sites and in intracellular compartments. Other potential ligands for CASK include the other members of the syndecan family, all four of which end in the identical COOH-terminal sequence of -EFYA (that we only isolated syndecan-2 in our two-hybrid screen may simply reflect the stochastic nature of the screening procedure). It is possible that other syndecans (particularly syndecan-3, which is relatively highly expressed in neurons) bind to CASK in nonsynaptic regions of the neuronal plasma membrane. In addition to the syndecans, of course, the neurexin intracellular COOH-terminal tail has specific affinity for the CASK PDZ (this study and Hata et al., 1996), and may be interacting with CASK in the brain. Neurexins are proposed to be at least in part present in synaptic membranes, based on their activity as latrotoxin receptors (Ushkaryov et al., 1992). The structure of neurexins, however, is extremely heterogeneous (Ullrich et al., 1995; Ushkaryov et al., 1992), and their subcellular distribution remains to be characterized in detail.

The syndecan family, consisting of four known members, constitutes a major form of heparan sulfate (HS) on the cell surface (reviewed in Bernfield et al., 1992; Carey, 1997; Couchman and Woods, 1996; David, 1993). Among the syndecans, conservation of sequence occurs only in their short intracellular COOH-terminal tails and in their transmembrane domains. Their striking sequence similarity had implied to many observers an important common function for intracellular COOH-terminal tails of syndecans. Here we show that one likely common function is to mediate binding to CASK, though the physiological significance of this interaction is still unknown.

Expression of syndecan HSPGs is regulated during development in a cell type–specific manner, and this regulation is believed to be important for dynamic cell–matrix interactions and for tissue morphogenesis (reviewed in Bernfield et al., 1992; Carey, 1997; Couchman and Woods, 1996; David, 1993). Transmembrane HSPGs such as syndecans may be involved in cell–matrix adhesion by binding to components of the extracellular matrix (ECM) including laminin, fibronectin, and collagen. By virtue of their interaction with CASK through their COOH-terminal tails, syndecans could provide a molecular bridge between the ECM and intracellular proteins. If CASK in turn interacts with cytoskeletal components such as protein 4.1 (Cohen et al., cosubmitted manuscript), then a syndecan–CASK connection could link ECM to the intracellular cytoskeleton. In such a way, a syndecan–CASK interaction may contribute to adhesion of synaptic junctions, perhaps in conjunction with the known cadherin-based adhesion complex found at the periphery of synaptic active zones (reviewed in Serafini, 1997).

In addition to membrane proteins, MAGUK proteins can also interact with intracellular signaling molecules (such as nNOS, APC). By analogy with the integrin family of ECM receptors that nucleate an intracellular complex of proteins at the site of ECM contact (Clark and Brugge, 1995), the syndecan–CASK interaction might form the axis of a cytoskeletal/signaling complex at the plasma membrane. Since all syndecans terminate with the -EFYA sequence sufficient for binding to CASK, such a model may apply to other members of the family in addition to the synaptically localized syndecan-2.

Recently, interest has focused on cell surface HSPGs such as the syndecans because they bind to, and are essential for the biological actions of, certain secreted polypeptide factors (reviewed in Carey, 1997; Schlessinger et al., 1995). In the best-studied example of FGF signaling, cell surface HSPGs are believed to be important for appropriate presentation of FGF to its high-affinity tyrosine kinase receptor, or for increasing the effective concentration of FGF at the plasma membrane. In this way, HSPGs at the plasma membrane can be regarded as coreceptors for heparin-binding factors, though of lower affinity and lower specificity than the specific receptor tyrosine kinases (reviewed in Carey, 1997; Schlessinger et al., 1995).

Bearing in mind that the CASK homolog LIN-2 is required for LET23/EGFR signaling in C. elegans, the question is raised: could the interaction between syndecan and CASK/LIN-2 play a role in growth factor signaling? We hypothesize that one function of CASK/LIN-2 is to bind and cluster syndecan HSPGs in the vicinity of the growth factor receptor. The HS side chains of syndecan can bind to many polypeptide growth/differentiation factors; this binding could enhance the action of the growth factor via a concentration effect or via appropriate presentation to its receptor tyrosine kinase. In the context of this model, it is noteworthy that the C. elegans syndecan protein also terminates in -EFYA, and is thus predicted to bind to LIN-2's PDZ domain. This prediction raises the possibility that loss-of-function of LIN-2 in C. elegans could inhibit LET23/EGF receptor signaling as a result of mislocalization of syndecan HSPGs. In vertebrates, HSPGs have been implicated in the action of neuregulin/ARIA, a heparin-binding secreted factor that acts on EGFR-like receptor tyrosine kinases (reviewed in Fischbach and Rosen, 1997). More intriguing is that neuregulin/ARIA is concentrated in the synaptic cleft of neuronal synapses (Sandrock et al., 1995), where it would be in a good position to interact with syndecan-2 HSPG. We speculate that synaptically localized syndecan HSPG may be involved in binding to certain growth factors (like neuregulins) in synapses, and in potentiating their action on synaptic receptor tyrosine kinases.

Previously, HSPGs have been localized in the vertebrate NMJ (Sanes, 1989; Sanes et al., 1986), but these HSPGs are either not identified or are thought to be associated with the ECM/basal lamina of the NMJ (e.g., agrin and perlecan). In this paper we find that brain synapses also contain a specific HSPG, thereby revealing a parallel with the NMJ; however, this HSPG is a transmembrane protein of the syndecan family rather than a predicted ECM protein. The localization of a cell surface HSPG in neuronal synapses sheds some new light on the molecular organization of synaptic junctions in the brain, and should provide a basis for further investigation into the molecules of the synaptic space.

The authors thank Dr. Merton Bernfield for the gift of MSE-2 syndecan-2 antibody, Drs. Graham Cowling and Thomas Südhof for providing syndecan-2 and CASK cDNA plasmids, Dr. James Anderson for helpful discussion and providing and hCASK plasmid and antibody, and Elaine Aidonidis for help with the manuscript.

aa

amino acid

ECM

extracellular matrix

EGFR

EGF receptor

HCASK

rabbit anti-human CASK

HSPG

heparan sulfate proteoglycan

Kv1.4-EFYA

Kv1.4-syndecan-2 chimeric protein

MAGUK

membrane associated guanylate kinase

MB

maleate buffer

PB

phosphate buffer

PSD

postsynaptic density

SYN-2C

syndecan-2 antibody