Agrin-induced Activation of Acetylcholine Receptor-bound Src Family Kinases Requires Rapsyn and Correlates with Acetylcholine Receptor Clustering*

During neuromuscular synaptogenesis, neurally released agrin induces aggregation and tyrosine phosphorylation of acetylcholine receptors (AChRs) by acting through both the receptor tyrosine kinase MuSK (muscle-specific kinase) and the AChR-associated protein, rapsyn. To elucidate this signaling mechanism, we examined tyrosine phosphorylation of AChR-associated proteins, particularly addressing whether agrin activates Src family kinases bound to the AChR. In C2 myotubes, agrin induced tyrosine phosphorylation of these kinases, of AChR-bound MuSK, and of the AChR β and δ subunits, as observed in phosphotyrosine immunoblotting experiments. Kinase assays revealed that the activity of AChR-associated Src kinases was increased by agrin, whereas phosphorylation of the total cellular kinase pool was unaffected. In both rapsyn-deficient myotubes and staurosporine-treated C2 myotubes, where AChRs are not clustered, agrin activated MuSK but did not cause either Src family or AChR phosphorylation. In S27 mutant myotubes, which fail to aggregate AChRs, no agrin-induced phosphorylation of AChR-bound Src kinases, MuSK, or AChRs was observed. These results demonstrate first that agrin leads to phosphorylation and activation of AChR-associated Src-related kinases, which requires rapsyn, occurs downstream of MuSK, and causes AChR phosphorylation. Second, this activation intimately correlates with AChR clustering, suggesting that these kinases may play a role in agrin-induced AChR aggregation by forming an AChR-bound signaling cascade.

A common feature of most synapses is the local accumulation of proteins regulating synaptic responses in the postsynaptic membrane. At the neuromuscular junction, a model system for synaptogenesis, aggregation of acetylcholine receptors (AChRs) 1 is an early sign of postsynaptic differentiation during development and occurs at sites of nerve-muscle contact (1,2).
Clusters of AChRs and further synaptic components are in-duced by neurally released agrin, an essential factor for synaptogenesis (3). Accordingly, mice deficient for agrin lack differentiated synapses and nerve-associated clusters of postsynaptic proteins (4). Neurons or recombinant agrin induce aggregation of AChRs and other muscle components when added to cultured myotubes (5,6), a process that mimics events at developing endplates, as it is blocked by antibodies specific for nerve-released agrin (7).
Little is known about the signaling pathway of agrin, although a muscle-specific kinase, MuSK, is part of the agrin receptor (8,9), and proteoglycans are thought to play an accessory role in presenting agrin to its receptor (10). Agrin causes tyrosine phosphorylation of MuSK, and this kinase is essential for postsynaptic specialization, as shown in MuSK-deficient mice, in which the phenotype is similar to agrinϪ/Ϫ animals (11). Furthermore, myotubes lacking MuSK fail to respond to agrin (9,11). Agrin also causes tyrosine phosphorylation of the ␤ subunit of the AChR, which may play a role in AChR clustering because, among other observations, the kinase inhibitors herbimycin and staurosporine block both AChR ␤ phosphorylation and AChR clustering (12,13). However, mutant AChRs, lacking cytoplasmic tyrosine residues in their ␤ subunits, can still be recruited into clusters upon expression in myotubes, which also express wild-type AChRs (14). In addition, it has been shown that AChR tyrosine phosphorylation is accompanied by a change in receptor desensitization by cholinergic ligands (15).
One protein involved in agrin signaling downstream of MuSK is rapsyn, a 43-kDa protein closely associated with AChRs (16,17). RapsynϪ/Ϫ mice lack aggregates of AChRs and other postsynaptic proteins, and their cultured myotubes fail to form agrin-induced AChR clusters, illustrating the importance of rapsyn for postsynaptic organization (18). Furthermore, rapsyn causes aggregation of AChRs upon expression in fibroblasts and is necessary for agrin-induced AChR ␤ phosphorylation in myotubes (19,20). MuSK, however, is still activated by agrin and localized at mutant endplates of rapsyndeficient animals, suggesting that MuSK forms a primary synaptic scaffold to which rapsyn recruits AChRs and further components of the postsynaptic apparatus (20,21).
The individual steps of this postsynaptic assembly, in particular, the agrin signaling events downstream of MuSK, still remain largely unknown. Myotubes contain preassembled AChR protein complexes in which, independently of agrin, AChRs are associated with several muscle proteins, including rapsyn, MuSK, and Src-related kinases (21). Agrin selectively increases, in a rapsyn-dependent way, the association between AChRs and MuSK, suggesting that agrin causes a link of these preassembled AChR protein complexes with the MuSK primary synaptic scaffold, thereby driving postsynaptic assembly.
Proteins that bind to MuSK or the AChR are likely to regu-late this mechanism. Activated MuSK contains several phosphorylated tyrosine residues in its cytoplasmic tail, which may act as docking sites for signal transducing molecules (22). One of these residues, a juxtamembrane tyrosine within an NPXY consensus, is indeed required for agrin-induced AChR phosphorylation and clustering, suggesting the involvement of a PTB (phosphotyrosine-binding) domain adaptor protein (23,24). Another critical step in agrin's signaling pathway are calcium fluxes, as the fast chelator, BAPTA-AM, inhibits agrin-induced AChR clustering (25). However, the nature of this calcium-dependent step is unknown. Moreover, several observations indicate the existence of a critical kinase downstream of MuSK. Thus, staurosporine does not affect agrin-induced activation of MuSK but inhibits AChR ␤ phosphorylation and AChR clustering, strongly suggesting that a staurosporine-sensitive intermediate kinase downstream of MuSK causes AChR phosphorylation and aggregation (26). Such a kinase(s) may be one or several members of the Src family, because Fyn and Fyk associate with AChRs in the Torpedo electric organ (27), whereas Src (c-Src) and Fyn are bound to AChRs in mammalian muscle (28). Furthermore, Src can phosphorylate AChR ␤ subunit fusion proteins (28), and Src members are regulated by rapsyn when expressed in fibroblasts, causing AChR phosphorylation in this heterologous system (29).
In several signaling pathways, the functional roles of Src kinases are indicated by an increase in their kinase activities following extracellular stimulation (30). Similarly to muscle, Src family kinases associate with ionotropic neurotransmittor receptors in the central nervous system, e.g. Src with NMDA receptors and Lyn with AMPA receptors (31,32). The associated kinases regulate receptor channel properties and form receptor-bound signaling cascades, as shown for Lyn, which is activated by AMPA stimulation of cerebellar neurons, leading to activation of the mitogen-activated protein kinase pathway and possibly synaptic plasticity (31).
To begin to understand the role of the Src family in postsynaptic differentiation in muscle and in agrin-induced signaling events downstream of MuSK, we examined the effect of agrin on tyrosine phosphorylation of proteins associated with the AChR, in particular the Src family members. We demonstrate that agrin causes activation and phosphorylation of AChRbound Src class kinases, which requires rapsyn and correlates with AChR phosphorylation and aggregation. Src-related kinases thus cause AChR phosphorylation induced by agrin and may play a role in AChR clustering, suggesting that the Src family not only regulates ionotropic receptors channel properties but also contributes to the synaptic localization of these receptors.

EXPERIMENTAL PROCEDURES
Cell Culture-Cell culture reagents were purchased from Life Technologies, Inc. C2 (C2C12) and S27 mouse muscle cells were maintained at 37°C in 8% CO 2 and propagated in Dulbecco's modified Eagle's medium with 4.5 g/liter D-glucose containing 20% fetal bovine serum, 0.5% chick embryo extract, 2 mM glutamine, and penicillin/streptomycin. After reaching 90 -100% confluence, cells were shifted to fusion medium containing Dulbecco's modified Eagle's medium, 5% horse serum, and 2 mM glutamine and fed daily. C2 myotubes were used for experiments after 2 days in this medium, while S27 cells were used after 2-3 days, by which time myotubes similar to C2 had formed. We noticed that the morphology of the S27 myotubes varied between different experiments, in that myotubes were occasionally wider than in C2 (see Fig. 8, S27 cells treated with 50 nM agrin). Furthermore, the fusion of myoblasts to form myotubes was less efficient in the case of S27, such that most cultures had more unfused cells than C2. Nevertheless, these morphological variations did not affect the outcome of our experiments, because agrin-induced clustering of AChRs and phosphotyrosine-containing proteins, as well as phosphorylation of MuSK and AChR-bound Src family kinases, were consistently absent in S27 cells irrespective of their morphology. RapsynϪ/Ϫ and the corresponding wild-type myoblasts were grown at 33°C, 5% CO 2 in the same medium as C2 cells with an additional 4 units/ml ␥-interferon, as described previously (21). Confluent cultures were fed with C2 fusion medium every second day and used for experiments after 3-4 days.
Antibodies-To detect AChR-associated phosphotyrosine proteins by immunoblotting, we used a mixture of two commercially available mouse monoclonal antibodies, 4G10 (Upstate Biotechnology, Lake Placid, NY) and PY20 (Transduction Laboratories, Lexington, KY). The mouse monoclonal antiserum 88B, reactive with the AChR ␥ and ␦ subunits, was provided by Dr. S. C. Froehner (University of North Carolina, Chapel Hill). Rat monoclonal antibodies 124 against the AChR ␤ subunit were a gift from Dr. J. Lindstrom (University of Pennsylvania, Philadelphia). To detect Src-related kinases, we used src-CT (Santa Cruz Biotechnology, Santa Cruz, CA), a rabbit polyclonal antibody raised against the C terminus of Src (c-Src), which is highly conserved in several Src-related kinases. Src-CT reacts with Src, Fyn, Yes (28), and possibly further kinases of the Src family. Fyn was detected by a rabbit polyclonal Fyn-specific antiserum (Santa Cruz Biotechnology). The antiserum against MuSK was as described previously (26).
Isolation of AChR Protein Complexes and Immunoprecipitation-Myotubes were lysed in a mild extraction buffer containing 1% Nonidet P-40 and an abundance of protease and phosphatase inhibitors, and AChRs were precipitated with ␣-bungarotoxin (␣-BT) coupled to Sepharose beads (␣-BT-Sepharose beads) as described previously (21). Alternatively, AChRs were isolated using 200 nM soluble biotin-conjugated ␣-BT followed by streptavidin-coupled agarose beads (Molecular Probes, Eugene, OR), which gave identical results. As controls, an excess (10 M) of free ␣-BT was added, or unconjugated Sepharose was used. To immunoprecipitate Src members representing the total cellular Src family pool, myotube extracts were incubated with 1 g of src-CT antibodies followed by protein A-Sepharose. Cellular MuSK was isolated accordingly using polyclonal anti-MuSK antibodies and protein A-Sepharose.
For both methods, the quantitation of signals on films was performed by scanning with a computerized densitometer (Nikon Scantouch 210) and using the NIH Image J 1.04b software (National Institutes of Health, Bethesda, MD). Background signals, originating from control precipitations including excess of free ␣-BT, were subtracted from ␣-BT AChR precipitations. Values from untreated cells were set to 100%, and signals after agrin treatment were calculated accordingly. Statistical evaluation was performed by ANOVA followed by pairwise Bonferroni's t tests.
Tyrosine Kinase Inhibitors-To examine the effects of kinase inhibitors on tyrosine phosphorylation of proteins bound to the AChR, C2 myotubes were incubated for 5 to 6 h with 1-20 nM staurosporine (Sigma), 1 M herbimycin A (Life Technologies, Inc.), or carrier controls. After stimulation with agrin, cells were extracted and analyzed by ␣-BT AChR precipitation followed by phosphotyrosine immunoblotting. Quantitation of signals was carried out as described previously (26).
Immunoblotting-To detect proteins in AChR, src-CT, or MuSK precipitations, proteins were eluted into SDS sample buffer, separated by SDS-PAGE, and transferred to nitrocellulose. Membranes were probed with the appropriate antibodies, and immunoreactive bands were visualized using horseradish peroxidase-conjugated secondary antibodies and enhanced chemiluminescence (ECL, Amersham Pharmacia Biotech). For reprobing, blots were stripped by incubating them for 20 min in 200 mM glycine, 0.1% Tween 20, pH 2.5.
Immunocytochemical Staining and Fluorescence Microscopy-To visualize the distribution of AChRs and phosphotyrosine proteins in C2 and S27 myotubes, cells were grown and fused on chamber slides (Nunc, Life Technologies, Inc.) and treated for 15 h with agrin concentrations ranging from 0.5 to 100 nM. Cells were incubated with 100 nM tetramethylrhodamine-conjugated ␣-bungarotoxin (Molecular Probes) in fusion medium for 1 h at 37°C followed by fixation in methanol for 7 min at Ϫ20°C and permeabilization (5 min in phosphate-buffered saline containing 1% Triton X-100 and 1 mM sodium orthovanadate). Phosphotyrosine proteins were detected by PY20 antibodies (Transduction Laboratories) and fluorescein isothiocyanate-conjugated goat-antimouse secondary antisera (Jackson ImmunoResearch Laboratories, La Roche, Switzerland) in the presence of 1 mM sodium orthovanadate. Cells were mounted in glycerol containing p-phenylenediamine and examined with a microscope (Axioskop 2, Carl Zeiss) equipped with the appropriate fluorescence filters. Images were processed with a computer-supported cooled 3CCD camera (Hamamatsu, Hamamatsu-City, Japan).

Agrin Induces Tyrosine Phosphorylation of Several Proteins
Associated with the AChR-In our initial experiments, we analyzed agrin-induced tyrosine phosphorylation of proteins associated with the AChR by preparing extracts from agrintreated and untreated C2 myotubes; AChRs were precipitated with ␣-bungarotoxin (␣-BT) covalently coupled to Sepharose beads, or using biotinylated ␣-BT in combination with streptavidin-agarose, followed by phosphotyrosine immunoblotting. To assess nonspecific protein binding to the beads, free ␣-BT was added prior to the precipitation, or unconjugated Sepharose was used. Using this procedure allows the specific isolation of AChRs in association with several postsynaptic proteins (21).
We stimulated C2 myotubes for increasing times with either recombinant neural agrin (agrin 4,8) or the predominant agrin isoform of muscle (agrin 0,0), which is much less effective in aggregating AChRs (6,33). ␣-BT precipitation and phosphotyrosine immunoblotting revealed a consistent pattern of AChRassociated phosphotyrosine bands induced by neural but not muscle agrin (Fig. 1). In addition to AChR-bound MuSK and the AChR ␤ subunit (26), we detected a band at about 70 kDa. This protein appears to represent the AChR ␦ subunit, based on the cross-reactivity with ␦-specific antibodies in stripping and reprobing experiments (Fig. 1). In addition, an AChR-associated protein of about 60 kDa, i.e. the molecular mass of Src family kinases, consistently became tyrosine-phosphorylated in response to neural agrin. Its increase in phosphorylation occurred in parallel with that of AChR-bound MuSK and the AChR ␤ and ␦ subunits, as it was first detected after a 15 min treatment with 0.1 nM neural agrin and reached a peak after 40 min (Fig. 1). Dephosphorylation of this 60-kDa band strictly paralleled that of the AChR subunits, as all three proteins were still phosphorylated to an intermediate degree after a 15-h treatment with neural agrin. MuSK, however, was no longer phosphorylated after this time. In these assays, agrin-induced tyrosine phosphorylation events were observed only for proteins associated with the AChR and not when total cellular lysates were analyzed without precipitation ( Fig. 1, L lanes). This selectivity indicates local, AChR-bound signaling, which does not involve changes in phosphorylation on the level of a whole myotube.
To determine whether this AChR-associated 60-kDa phosphotyrosine protein represents one or several members of the Src family, immunoblots were stripped and reprobed with pan-Src antibodies (src-CT) that recognize Src, Fyn, and Yes in muscle cells (28). These assays revealed that the protein indeed comigrates precisely with AChR-bound Src members and that the protein amounts of these associated kinases are not affected by agrin ( Fig. 2A). We have previously reported that two members of the Src family, Src and Fyn, are associated with AChRs and that they migrate with identical mobility in gel electrophoresis and immunoblotting experiments (28). Further reprobing studies using Fyn-specific antibodies confirmed that the observed 60-kDa phosphotyrosine band indeed comigrates with AChR-bound Fyn (Fig. 2B). The amounts of AChR-bound Src and Fyn are unchanged by agrin (26), a finding similar to the results using src-CT antibodies ( Fig. 2A). Therefore, both kinases are associated constitutively with the AChR independently of agrin, and either one or both of them may become phosphorylated strictly in response to agrin.
AChR-associated Kinases of the Src Family Are Activated by Neural Agrin-To confirm that the 60-kDa AChR-associated phosphoprotein indeed represents one or several members of the Src family bound to the AChR, we isolated large amounts of  (4,8) or muscle (0,0) agrin. AChRs were isolated from cellular lysates using ␣-BT-Sepharose beads followed by SDS-PAGE and phosphotyrosine immunoblotting (Tox-P/PY-blot). Middle panel, the blot was stripped and reprobed with MuSK-specific antibodies (Tox-P/MuSK-reprobe), whereas a second reprobing using AChR ␥and ␦-specific antisera (monoclonal antibodies 88b) is shown at the bottom (Tox-P/␥/␦-reprobe). Controls for AChR precipitation included the addition of 10 M free ␣-BT (ϩT) and the use of unconjugated Sepharose (C). The solid arrowheads indicate proteins associated with AChRs isolated by ␣-BT precipitation. The open arrowhead designates the AChR-bound protein(s) of ϳ 60 kDa in which tyrosine phosphorylation is induced by neural agrin. L indicates a fraction (0.3%) of the initial lysate analyzed without precipitation.
AChR protein complexes from agrin-treated and untreated C2 myotubes using ␣-BT-Sepharose. Complexes were dissociated by an incubation at 80°C in a buffer containing deoxycholate and SDS. The proteins released were then reprecipitated with phosphotyrosine antibody beads, immunoblotted, and probed with src-CT antibodies. This assay specifically monitors the overall levels of tyrosine phosphorylation of Src class kinases bound to the AChR. We indeed found a significant increase in the phosphotyrosine content of these kinases in response to neural agrin (Fig. 3A). As Src members undergo autophosphorylation when activated, we additionally used kinase assays to examine the effect of neural agrin on phosphate incorporation into and thus activation of AChR-associated Src kinases. To this end, ␣-BT-precipitated AChR complexes, isolated from agrin-treated or untreated cells, were subjected first to in vitro phosphorylation, using [␥-32 P]ATP at 25°C. The complexes were then dissociated at 80°C and released Src members immunoprecipitated with src-CT antibodies followed by gel electrophoresis and autoradiography. In these assays, neural agrin significantly increased the phosphate incorporation into Srcrelated kinases by ϳ3-fold, indicative of their activation (Fig.  3B). Together, these results show that one or several AChRassociated kinases of the Src family are activated by neural agrin and that the AChR-bound phosphotyrosine band at 60 kDa, as seen on immunoblots, indeed originates, at least in part, from Src members.
Interestingly, when Src members were isolated from agrintreated myotubes by a direct immunoprecipitation with src-CT antibodies, their phosphotyrosine content was unaffected in comparison with untreated cells (Fig. 3C). Thus, agrin specifically activates those Src-related kinases that are associated with AChRs without affecting the total cellular kinase pool.

Agrin-induced Phosphorylation of AChR-bound Src Members
Correlates with AChR Phosphorylation-We asked how activation of AChR-associated Src members correlates with phosphorylation of the AChR subunits and MuSK. For this purpose, C2 myotubes were treated for various times with a wide range of agrin concentrations (20 pM-5 nM). A 5-min incubation with 0.5 nM neural agrin was sufficient to induce significant phosphorylation of AChR-bound Src family kinases and the receptor ␤ and ␦ subunits, as well as of AChR-associated MuSK (Fig. 4). Maximal phosphorylation of these proteins, particularly at lower agrin concentrations, required a 40-min incubation with neural agrin. At 20 pM agrin, a slight decrease in the extent of Src family phosphorylation (relative to maximal stimulation at higher concentrations) was paralleled by a decreased phosphotyrosine signal of the AChR ␤ and ␦ subunits. Muscle agrin had no effect on any phosphorylation. Thus, phosphorylation of AChR-bound Src-related kinases intimately correlates with AChR ␤ and ␦ phosphorylation, raising the possibility that AChR phosphorylation occurs in direct response to Src member activation. The quantitation (right) shows the mean Ϯ S.D. from six experiments evaluated by densitometric scanning as described in A. C, Src-related kinases were isolated from agrin-treated cells using src-CT antibodies and protein A-Sepharose precipitation followed by phosphotyrosine immunoblotting. As controls, either the src-CT antibody (ϪAb) or the cellular lysate (ϪL) was omitted. *, differs significantly from control, p Ͻ 0.05 (by ANOVA followed by Bonferroni's t test).

Staurosporine Selectively Blocks Agrin-induced Phosphorylation of AChR-associated Src Family Kinases and the
AChR-To analyze the role of Src family activation in agrininduced AChR phosphorylation and clustering, we used the kinase inhibitors herbimycin and staurosporine. Both of these inhibitors block AChR ␤ subunit phosphorylation and AChR aggregation (12,13), whereas staurosporine does not affect agrin-induced activation of MuSK (26). In this present study, we observed that staurosporine inhibits agrin-induced phosphorylation of AChR-bound Src family kinases and of AChR ␤ and ␦ subunits (Fig. 5A). Using increasing concentrations of staurosporine, Src member phosphorylation was gradually reduced strictly in parallel with AChR phosphorylation (Fig. 5B). Most dramatically, at 20 nM of staurosporine, agrin did not induce significant phosphorylation of AChR-associated Src members and AChR ␤ and ␦ subunits, whereas phosphorylation of MuSK bound to the AChR occurred normally. Thus, MuSK activation can be uncoupled from phosphorylation of Src kinases and the AChR, suggesting that the AChR ␤ and ␦ subunits are direct substrates for AChR-bound Src members but not for MuSK.
Herbimycin, in contrast, abolished phosphorylation of all proteins seen in ␣-BT-precipitates, including MuSK, Src kinases, and the AChR (Fig. 5A). Together, these results demonstrate that herbimycin and staurosporine exert their inhibitory effect on AChR phosphorylation and clustering by interfering with distinct steps in the agrin signaling pathway. Whereas herbimycin inhibits MuSK, staurosporine blocks autophosphorylation of AChR-bound Src kinases, suggesting that Src activation is downstream of MuSK and involved in AChR clustering.
In the Absence of Rapsyn, AChR-bound Src Members Are Not Activated by Agrin-To further correlate Src family autophosphorylation with AChR clustering, we analyzed the effect of neural agrin on tyrosine phosphorylation of AChR-bound proteins in rapsyn-deficient myotubes. These cells were derived from rapsyn knockout mice and do not aggregate AChRs in response to agrin (21). Agrin induces normal phosphorylation of total cellular as well as AChR-associated MuSK in these cells, whereas AChR ␤ subunit phosphorylation is strongly reduced, suggesting that rapsyn is required for a step downstream of MuSK that leads to AChR phosphorylation and clustering (20,21). Therefore, we examined phosphorylation of AChR-bound Src members by agrin in rapsyn Ϫ/Ϫ myotubes. In control wildtype cells, agrin induced phosphorylation of AChR-bound Src family kinases and the AChR ␤ and ␦ subunits as in C2 myotubes (Fig. 6). In contrast, in the rapsyn-deficient myotubes, these phosphorylations were strongly reduced in both agrintreated and untreated cells. Control experiments confirmed that rapsyn-deficient and wild-type cells express identical levels of MuSK and Src members and that the amount of ␣-BTisolated AChR is indistinguishable ( Fig. 6; Ref. 21). These results concur with those on staurosporine in C2 myotubes, demonstrating that rapsyn is required for agrin-induced phosphorylation of AChR-bound Src family kinases and that this activation occurs downstream of MuSK. The data further suggest that AChR ␤ and ␦ subunits are direct substrates for these Src class kinases.
In S27 Myotubes, Agrin Fails to Cause Phosphorylation of AChR-bound Src Family Kinases-As a final correlation between agrin-induced Src member activation and AChR clustering, we analyzed Src class phosphorylation in S27 myotubes, genetic variants of C2 defective in heparan and chondroitin FIG. 4. Correlation between agrin-induced phosphorylation of AChR-bound Src kinases and AChR subunits. C2 myotubes were treated with neural (4,8) or muscle (0,0) agrin as indicated, and AChR protein complexes were precipitated with ␣-BT-Sepharose followed by phosphotyrosine immunoblotting (Tox-P/PY-blot). The bottom panel is a short exposure (Short exp.) of the blot to better visualize the dose dependence and time course of AChR ␤ phosphorylation.

FIG. 5. Effects of tyrosine kinase inhibitors on agrin-induced phosphorylation of AChR-bound MuSK, Src family kinases, and
the AChR ␤ and ␦ subunits. A, C2 myotubes were treated for 5 h with 1 M herbimycin (H), the indicated concentrations of staurosporine (S), or a Me 2 SO carrier control (c) and were then stimulated for 40 min with 1 nM neural agrin. Lysates were processed by ␣-BT-Sepharose precipitation and subjected to phosphotyrosine immunoblotting. The lower panel shows a shorter exposure (Short exp.) of the blot to better illustrate tyrosine phosphorylation of the AChR ␤ subunit. B, four experiments as shown in A were quantitated by densitometric scanning. Relative intensities of phosphorylation of AChR-bound MuSK, Src members, and AChR subunits from cells not treated with agrin were set to 0%. Agrin-induced signals of cells incubated with carrier and lacking staurosporine were set to 100% (control), and intensities of signals from staurosporine-treated cells were calculated accordingly. The data are shown as the mean Ϯ S.D. sulfate proteoglycan biosynthesis (34,35). In these cells, even high concentrations of agrin up to 100 nM do not induce AChR clustering and ␤ phosphorylation (10,12).
To determine which step in the agrin signaling pathway is affected in S27 cells, we examined agrin-induced phosphorylation and clustering events in these cells. In C2, 0.5 nM neural agrin was sufficient to induce abundant co-clustering of AChRs and phosphotyrosine proteins, as observed by double-label immunofluorescence microscopy using rhodamine-conjugated ␣-BT and phosphotyrosine antibodies (Fig. 8). The same agrin concentration also caused phosphorylation of AChR-bound MuSK, Src family kinases and the AChR ␤ and ␦ subunits in C2 cells (Fig. 7A, see also Fig. 4). In contrast, in S27 myotubes, 50 and 100 nM agrin did not induce comparable phosphorylation of any of these proteins, even though Src family kinases were associated normally with the AChR and AChRs were present in normal amounts (Fig. 7A). Similarly, 100 nM neural agrin failed to induce the clustering of AChRs and phosphotyrosine-containing proteins in these cells (Fig. 8). Some aggregates of phosphotyrosine-containing proteins were occasionally observed in agrin-treated S27 cells, but these were much smaller and less intense than the typical clusters in C2 (Fig. 8).
Based on the absence of agrin-activated AChR-associated MuSK in S27, we examined phosphorylation of the total cellular pool of MuSK by neural agrin in these cells, using MuSK immunoprecipitation followed by phosphotyrosine immunoblotting. 100 nM neural agrin did not cause prominent MuSK activation in S27, whereas in C2 cells, 0.5 nM caused a strong phosphorylation (Fig. 7B). We noticed that the majority of MuSK isolated from S27 cells migrated with a lower molecular weight than in C2, consistent with defective glycosylation of MuSK in the mutant cells. Our phosphotyrosine blotting assays, however, sometimes revealed weak agrin-induced phosphorylation of a MuSK population with the same apparent molecular weight as in C2 (Fig. 7B). It is therefore possible that MuSK exists in several forms in S27 cells, because of differential glycosylation, and that some of these subpopulations react slightly differently to treatment with high doses of agrin. Together, these experiments show that even high concentrations of agrin fail to cause clustering of AChRs and phosphorylation of AChR-bound Src members in S27. Although we cannot exclude the possibility that further, unknown genetic defects in these cells prevent AChR clustering, our results are consistent with the assumption that, similar to herbimycin-treated C2 cells, MuSK is not activated by agrin in S27 cells, which causes defects in the activation of AChR-bound Src family kinases, in AChR phosphorylation, and in clustering of AChRs and phosphotyrosine-containing proteins. DISCUSSION In this study, we demonstrate that agrin activates Src family kinases that are associated with the AChR. The activation occurs downstream of MuSK, requires rapsyn, and correlates closely with agrin-induced AChR clustering. These data indicate that Src family kinases are part of a local AChR-bound signaling cascade and that they may play a role in the assembly of AChR protein complexes into a postsynaptic apparatus.
Activation of Src Family Kinases by Agrin-Several of our FIG. 6. In rapsyn-deficient myotubes, neural agrin does not cause phosphorylation of AChR-bound Src members and AChR ␤ and ␦ subunits. Rapsyn Ϫ/Ϫ and wild-type (W.T.) myotubes were treated with 1 nM neural (4,8) or muscle agrin (0,0) as indicated and analyzed by ␣-BT-Sepharose precipitation followed by phosphotyrosine immunoblotting. A shorter exposure of the film revealed strong agrininduced AChR ␤ phosphorylation in the wild type and no signal at all in mutant cells (not shown). Src family kinases (src's) were identified by stripping and reprobing the blot with src-CT antibodies (lower panel). The amounts of ␣-BT-isolated AChRs were equal between the two cell types (not shown).

FIG. 7. Agrin fails to activate AChR-bound Src family kinases in S27 myotubes.
A, S27 and C2 myotubes were treated with neural (4,8) or muscle agrin (0,0) as indicated. AChRs were isolated with biotinylated ␣-BT followed by streptavidin-agarose precipitation and phosphotyrosine immunoblotting. As controls, an excess (10 M) of free toxin was added (ϩT). To ensure equal amounts of ␣-BT-precipitated AChRs and AChR-bound Src family kinases, parallel cells were subjected to ␣-BT precipitation and AChR ␤ subunit or src-CT immunoblotting (middle and bottom). For comparison, fractions (L) of the initial lysate were loaded. B, cells treated with agrin as indicated were lysed and processed by MuSK immunoprecipitation and phosphotyrosine immunoblotting (top). As a control, the MuSK antibody was omitted (ϪAb). To compare the amounts of MuSK between both cell types, fractions of cellular lysates (L) were analyzed by MuSK immunoblotting (bottom). The majority of MuSK (solid arrowheads) migrates with a lower molecular weight in S27 than in C2 cells. A fragment of the MuSK antibody that comigrates with the majority of MuSK in S27 cells is denoted by the asterisk; the open arrowhead indicates phosphorylation of a minor population of MuSK in S27 that migrates as in C2.
observations indicate that AChR-associated kinases of the Src family are activated by agrin. In agrin-treated myotubes, the phosphotyrosine content of these kinases is increased, as shown by both kinase reprobing and phosphotyrosine reprecipitation experiments. Furthermore, in kinase assays in vitro, incorporation of phosphate into the kinases is augmented by agrin. Such in vitro phosphorylation assays are widely used to analyze the activation of Src class kinases in response to various stimuli, e.g. growth factors, and an increase in phosphate incorporation indicates kinase autophosphorylation and thus activation (30). We can exclude the possibility that AChRbound Src family kinases are phosphorylated by MuSK, because staurosporine blocks Src phosphorylation without affecting MuSK autophosphorylation (see also below). Thus, our kinase assays most likely reflect Src family autophosphorylation. In our experiments, we used a high affinity Src family antibody against a conserved epitope of Src, Fyn, and Yes. As both Src and Fyn have been shown to be associated with AChRs (28), we therefore assume that either one or both of these kinases are activated by agrin. The affinities of specific anti-Src or anti-Fyn antibodies in immunoprecipitations were too low to show an activation of the respective individual kinase. 2 Kinases of the Src family are regulated in complex ways, both by tyrosine phosphorylation and by intramolecular protein interactions mediated by their SH2 and SH3 domains (30,36). Activation can be achieved by binding of proline-rich sequences of other proteins to the SH3 domain, by interaction of external phosphotyrosine peptides with the SH2 domain, or by C-terminal dephosphorylation, all of which lead to the active open Src conformation (36,37). In all cases, activation requires autophosphorylation of a tyrosine in the activation loop of the kinase domain.
Based on these mechanisms, our observed Src family activation does not appear to be mainly mediated by C-terminal dephosphorylation but rather by interacting proteins that bind to the Src family SH2 and/or SH3 domains, thereby leading to the increase in overall phosphotyrosine content. A good candidate for such an interacting protein is MuSK, a portion of which is associated with the AChR and becomes phosphorylated in response to agrin (26). Interestingly, the cytoplasmic domain of MuSK contains two proline-rich regions, one of which corresponds to the consensus requirements for Src SH3 binding, as it contains the typical PXXP motif preceded by a leucine (38,39). These MuSK peptide motifs and/or the autophosphorylated cytoplasmic tyrosine residues of activated MuSK (22) could thus bind to Src class kinases causing kinase opening, autophosphorylation, and activation. Such a mechanism would work most efficiently and rapidly if the proteins involved (AChRs, MuSK, and Src family kinases) were to form a preassembled complex. Indeed, we have previously demonstrated that these proteins interact in myotubes, before the addition of agrin (21), and data in this study indicate that Src family activation occurs rapidly, as early as 5 min after addition of agrin, and is highly correlated with the activation of MuSK.
Src family activation was observed only for that fraction of kinases associated with AChRs but not on the level of the total cellular kinase pool. In contrast, the activation of Src members in other signaling systems, such as mitogenic responses to growth factors or adhesion on fibronectin, involves and is detectable on the whole cellular kinase pool (40,41). This illustrates an important difference between Src family signaling in cell proliferation or adhesion and in postsynaptic assembly. The latter process requires correct protein interactions, presumably driven by local signaling events, and would be hard to reconcile with signaling occurring throughout myotubes as a whole. The need for local signaling, combining kinase activation with protein interactions, is illustrated by MuSK, which, in addition to signaling, also mediates protein interactions involving its extracellular domain, and where, unlike in the case of growth factor receptors, the activated kinase domain is not sufficient to reproduce all aspects of signaling (20,42).
Src Family Activation and AChR Phosphorylation-Our AChR precipitation procedure combined with phosphotyrosine immunoblotting also revealed agrin-induced phosphorylation of AChR ␦ subunits, which has been described previously in the chick (43). Our present data are the first report of this modification in mammalian myotubes, and they indicate that AChR ␦ phosphorylation caused by agrin highly correlates with AChR clustering. Thus, neural but not muscle agrin induces both phosphorylation and aggregation. Furthermore, herbimycin and staurosporine inhibit both agrin-induced AChR clustering and ␦ phosphorylation, and phosphorylation of ␦ is not observed in rapsynϪ/Ϫ and S27 myotubes in which AChRs are not clustered by agrin. As it reaches a peak after 40 min of agrin treatment, AChR ␦ phosphorylation may play a role early in agrin-induced clustering and compensate for phosphorylation of AChR ␤ subunits, allowing agrin-induced clustering of mutated AChRs lacking ␤ cytoplasmic tyrosine residues (14). Because phosphorylated AChR ␦ subunits bind to the adaptor molecule Grb2 in the Torpedo electric organ (44), one role of ␦ phosphorylation in the agrin signaling pathway in myotubes may be to recruit Grb2 to AChRs, thereby linking further signaling pathways to the AChR. Colledge and Froehner (44) have speculated that such an interaction may have to be of low affinity, which fits our observed low relative extent of ␦ phosphorylation. In fact, one intriguing possibility is that agrininduced ␤ phosphorylation, which is of high relative stoichiometry, regulates the channel properties of the AChR such that clustered receptors have accurate desensitization kinetics, whereas agrin-induced ␦ phosphorylation regulates the association of AChRs with signaling pathways, thereby facilitating postsynaptic assembly.
Which tyrosine kinase phosphorylates the AChR in response to agrin? Several of our experiments suggest that both the AChR ␤ and ␦ subunits are direct substrates for agrin-activated Src members bound to the AChR but not for MuSK. First, 2 C. Fuhrer, unpublished observations. FIG. 8. Agrin induces co-clustering of AChRs and phosphotyrosine-containing proteins in C2 but not in S27 myotubes. C2 and S27 myotubes were treated for 15 h with the indicated concentrations of neural agrin. Cells were double-labeled for AChRs with rhodamine-conjugated ␣-BT (left panels) and for phosphotyrosine proteins using anti-phosphotyrosine antibodies followed by fluorescein isothiocyanate-coupled secondary antisera (right panels). Whereas 0.5 nM of agrin induces pronounced co-clustering of AChRs and phosphotyrosinecontaining proteins in C2, no comparable protein aggregates are observed in S27 myotubes, not even at 100 nM of agrin. Scale bar, 20 m. phosphorylation of Src members and of AChRs are strictly correlated over a wide range of agrin concentrations and incubation times. Second, staurosporine inhibits Src family activation, which intimately correlates with reduced AChR ␤ and ␦ phosphorylation, whereas MuSK is unaffected. Third, staurosporine is known to inhibit both tyrosine kinase autophosphorylation and phosphorylation of their direct substrates in parallel, as shown in the case of the insulin receptor and its substrate IRS-1 and TrkA and its substrate phospholipase C-␥1 (45,46). Finally, the differential effect of herbimycin, which also inhibits MuSK, versus staurosporine suggests a hierarchy of phosphorylation, where MuSK activation is upstream of the activation of AChR-bound Src kinases, which phosphorylate the AChR ␤ and ␦ subunits.
Interestingly, as in staurosporine-treated C2 myotubes, agrin-induced Src member activation and AChR phosphorylation are also inhibited in rapsyn-deficient cells, showing that rapsyn is necessary for Src family activation by agrin. In fibroblasts, rapsyn interacts with and activates both MuSK and Src members, causing AChR ␤ and ␦ phosphorylation (29,47). These previous findings and our present studies suggest a complex of the AChR in association with rapsyn, MuSK, and Src family kinases, in which rapsyn holds AChR-bound MuSK and Src members in a conformation that allows activation of Src kinases by MuSK. The latter scenario could occur via an interaction of MuSK sequences with the Src family SH2 or SH3 domains as mentioned earlier. Src members thus appear to phosphorylate the AChR in response to agrin, a process that occurs downstream of MuSK and requires rapsyn.
Agrin-induced Src Family Activation and Postsynaptic Differentiation-Studies in several signaling systems show that an activation of Src family kinases by extracellular stimuli is an indication of a functional role of these kinases in the respective pathways (30). In our case, the activation of AChR-bound Src family kinases by agrin correlates closely with agrin-induced AChR clustering as judged by several criteria, including the differential effects of neural and muscle agrin, the inhibition by staurosporine and herbimycin, and the absence of agrin-induced Src member activation and AChR clustering in rapsyn-deficient myotubes. Furthermore, also in S27 myotubes treated with high concentrations of agrin, AChR clustering and phosphorylation of AChR-bound Src members are not observed (Table I). As staurosporine blocks Src family but not MuSK autophosphorylation, AChR-associated Src class kinases indeed appear to be the previously proposed kinase(s) downstream of MuSK that are important for AChR clustering (26). Furthermore, by inference, Src family activation shows the same dose dependence on neural agrin as AChR clustering, because it correlates strictly with AChR ␤ phosphorylation, which, in turn, highly correlates with AChR aggregation (12). Together, these striking correlations suggest that AChR-asso-ciated Src family kinases play a role in agrin-induced postsynaptic differentiation, to which AChR phosphorylation itself may contribute.
Agrin-activated Src class kinases could act at different steps in the pathway leading from muscle innervation to formation of a fully differentiated postsynaptic apparatus. We have previously suggested that preassembled protein complexes containing AChRs in association with postsynaptic elements may be the building blocks for agrin-induced formation of the postsynaptic membrane (21) and that this mechanism involves a rapsyndependent link of such complexes to the MuSK-based primary synaptic scaffold (20). Preassembled AChR complexes contain rapsyn, Src members, and MuSK, and signaling by these kinases may render the complexes competent for such a link by the phosphorylation of parts of the complexes and/or further proteins. This mechanism could involve tyrosine phosphorylation of rapsyn, as rapsyn is a substrate for Src family kinases when coexpressed in fibroblasts (29). However, in preliminary experiments we did not observe tyrosine phosphorylation of AChR-bound rapsyn due to agrin, 3 although small amounts of phosphorylation may have remained undetected.
Alternatively, as AChR-associated Src class kinases, in contrast to MuSK, are still activated after prolongated incubation with agrin in vitro, they may play a role in growth and stabilization of postsynaptic protein clusters. Interestingly, besides rapsyn, several cytoskeletal proteins concentrated at the neuromuscular junction or in Torpedo synaptic membranes, including talin, paxillin, and ␣-fodrin (1,48), are known substrates for Src class kinases (30). Talin and paxillin are also found at focal adhesion sites, which have similarities to the postsynaptic apparatus. In both cases, large protein complexes mediating cell-cell contact are assembled by tyrosine kinase signaling (49). Moreover, integrins, which modulate agrin signaling, are found at both structures (50). Together, these similarities suggest that AChR-associated Src family kinases may contribute to postsynaptic assembly in a way analogous to their proposed role in focal adhesion site assembly, by phosphorylating additional proteins, including cytoskeletal elements, and thereby stabilizing growing AChR aggregates.
Finally, maturation of the postsynaptic apparatus involves a series of cytoskeletal rearrangements, which ultimately lead to formation of postsynaptic folds, including segregation of several postsynaptic proteins between crests and troughs (51). At the mammalian neuromuscular junction, as well as in Torpedo electrocytes, tyrosine phosphorylation in the postsynaptic membrane increases during this late stage of synaptic maturation (52,53). This activity could originate from AChR-bound tyrosine kinases, e.g. Src members, and may cause correct 3 M. Moransard and C. Fuhrer, unpublished observations. TABLE I Activation of AChR-bound Src family kinases by agrin correlates with AChR clustering For analysis of phosphorylation, C2, rapsynϪ/Ϫ, and S27 myotubes were incubated for 40 min with neural (4,8) or muscle (0,0) agrin at concentrations of up to 100 nM. In the case of staurosporine and herbimycin, inhibitors were added to cells 5-6 h prior to agrin treatment. AChRs were isolated from cell extracts by ␣-BT precipitation, and phosphorylation of receptor-associated proteins (MuSK, Src-related kinases) and AChR subunits was assayed by phosphotyrosine immunoblotting. For AChR clustering, myotubes were treated for 15 h with up to 100 nM agrin followed by staining with rhodamine-conjugated ␣-BT and analysis by fluorescence microscopy (see also Refs. 12 and 21 colocalization of proteins like utrophin or syntrophin isoforms with AChRs at the crests of the folds, whereas other proteins, such as sodium channels or dystrophin, are enriched in the troughs (51,54). Indeed, some of these proteins as well as dystrobrevin isoforms, which may also be differentially distributed, are tyrosine-phosphorylated in Torpedo or have tyrosine phosphorylation consensus sites (51,55). Thus, Src familymediated phosphorylation of such proteins may determine their precise postsynaptic localization with respect to AChRs and contribute to later aspects of postsynaptic maturation.
In the central nervous system, the Src family is thought to have various functions, including regulation of glutamate receptors. At central synapses, NMDA receptors are regulated by tyrosine phosphorylation and associated with Src (c-Src) in a manner similar to muscle AChRs (56). In CA1 pyramidal cells of rat hippocampus, Src activation induces long-term potentiation, apparently by potentiation of bound NMDA receptors (32). AMPA receptors of the cerebellum, on the other hand, interact with the Src family kinase Lyn, which is activated following receptor stimulation (31). Glutamate receptors thus associate with Src members, which, upon activation, contribute to synaptic plasticity in various ways. Compared with these studies, our findings are the first indication that activation of Src class kinases may also regulate the distribution and postsynaptic localization of ionotropic neurotransmittor receptors.
In summary, the present data, together with our previous studies (21), suggest a model for AChR clustering in which agrin activates AChR-bound MuSK, which requires rapsyn to activate receptor-associated Src members, which phosphorylate AChR ␤ and ␦ subunits and possibly further proteins in the vicinity. This AChR-bound signaling cascade causes preassembled AChR protein complexes containing rapsyn, tyrosine kinases, and postsynaptic proteins to be linked to a primary synaptic scaffold formed by MuSK (20), thereby providing a stepwise assembly mechanism of the postsynaptic apparatus. In addition, Src kinases may contribute to the stabilization and further differentiation of the postsynaptic apparatus, by phosphorylating cytoskeletal proteins, thereby modifying associations between the postsynaptic membrane and the cytoskeleton. Under several different experimental conditions, therefore, Src phosphorylation and activation are correlated with AChR clustering, and the precise role of Src family kinases in postsynaptic assembly remains to be investigated.