An Extracellular Pathway for Dystroglycan Function in Acetylcholine Receptor Aggregation and Laminin Deposition in Skeletal Myotubes*

The dystroglycan (DG) complex is involved in agrin-induced acetylcholine receptor clustering downstream of muscle-specific kinase where it regulates the stability of acetylcholine receptor aggregates as well as assembly of the synaptic basement membrane. We have previously proposed that this entails coordinate extracellular and intracellular interactions of its two subunits, α- and β-DG. To assess the contribution of the extracellular and intracellular portions of DG, we have used adenoviruses to express full-length and deletion mutants of β-DG in myotubes derived from wild-type embryonic stem cells or from cells null for DG. We show that α-DG is properly glycosylated and targeted to the myotube surface in the absence of β-DG. Extracellular interactions of DG modulate the size and the microcluster density of agrin-induced acetylcholine receptor aggregates and are responsible for targeting laminin to these clusters. Thus, the association of α- and β-DG in skeletal muscle may coordinate independent roles in signaling. We discuss how DG may regulate synapses through extracellular signaling functions of its α subunit.

The dystroglycan (DG) complex is involved in agrin-induced acetylcholine receptor clustering downstream of muscle-specific kinase where it regulates the stability of acetylcholine receptor aggregates as well as assembly of the synaptic basement membrane. We have previously proposed that this entails coordinate extracellular and intracellular interactions of its two subunits, ␣and ␤-DG. To assess the contribution of the extracellular and intracellular portions of DG, we have used adenoviruses to express full-length and deletion mutants of ␤-DG in myotubes derived from wild-type embryonic stem cells or from cells null for DG. We show that ␣-DG is properly glycosylated and targeted to the myotube surface in the absence of ␤-DG. Extracellular interactions of DG modulate the size and the microcluster density of agrin-induced acetylcholine receptor aggregates and are responsible for targeting laminin to these clusters. Thus, the association of ␣and ␤-DG in skeletal muscle may coordinate independent roles in signaling. We discuss how DG may regulate synapses through extracellular signaling functions of its ␣ subunit.
Dystroglycan (DG) 2 is a member of the dystrophin-associated glycoprotein complex (DGC). It is encoded by a single gene (dag1) and is expressed initially as a propeptide that gets cleaved to generate two distinct subunits as follows: ␣-dystroglycan (␣-DG), a peripheral mucin-like protein, and ␤-dystroglycan (␤-DG), a transmembrane protein (1,2). In mature skeletal muscle, both subunits associate noncovalently and localize to the plasma membrane where they form the functional core of the DGC. Dystroglycan is crucial for muscle fiber integrity and survival and is widely thought to act as a structural element of the cell surface by linking the extracellular matrix (ECM) to the cytoskeleton to protect the cell against the stress of contractions (3,4). Other studies suggest that DG mediates intracellular signaling in noncontractile cells (5,6) and can regulate cell death (7,8), replication (9), and polarity (10,11).
In skeletal muscle, the absence of DG leads to muscular dystrophy (12,13). At neuromuscular junctions (NMJs), DG associates directly with rapsyn (14) and is colocalized with acetylcholine receptors (AChRs) (15) by their movement to nascent synapses (16). DG-deficient muscle fibers in vivo and in vitro show diffuse and unstable aggregates of AChRs (12,17,18), indicating a role for DG in the condensation and stabilization of AChR microclusters. In addition, laminin, perlecan, and acetylcholinesterase (AChE) are greatly reduced at the NMJs of DGdeficient muscle (12,18).
The NMJ represents a highly specialized compartment within the multinucleated myotube that extends from the nerve terminal through the intervening synaptic basement membrane and plasma membrane to the underlying cytoskeleton and subsynaptic nucleus (19). Formation of this synaptic compartment is associated with expression, highly restricted localization, and stabilization of molecules. The mechanisms involved are fundamental to our understanding of structural and functional mosaicism within the cells. For example, one of the hallmarks of mature NMJs is the presence of a high density of AChRs (10 4 receptors/ m 2 ) at the tops of folds in the postsynaptic membrane (20). These receptors are trapped in a region of the plasma membrane directly apposed to the nerve terminal, degrade very slowly (21), and are unable to diffuse within the plasma membrane. Initially, AChR aggregation is a muscle-intrinsic process that does not require innervation (22,23), but reorganization of the AChRs within the membrane is regulated by secretion from the motor nerve of agrin, a heparan-sulfate proteoglycan (24). Agrin acts via MuSK, a muscle-specific receptor tyrosine kinase, together with a myotube-associated specific coreceptor within the postsynaptic membrane (25). Phosphorylation of MuSK recruits rapsyn, which is closely associated with the major cytoplasmic loop of the AChRs (26 -31) and can self-associate to form microclusters of AChRs (27,32). Activation of MuSK leads to the phosphorylation of the ␤ subunit of the AChR (33), an important requirement for enhanced linkage of AChRs to the cytoskeleton (34). As a specialized domain of skeletal myofibers, the NMJ offers insights into DG functions in fundamental cellular processes. For example, DG functions in basement membrane assembly by anchoring AChE via interactions with perlecan (18,35,36).
In previous studies, we have shown that the glycosylation of ␣-DG is regulated by the nerve and that it functions in NMJ formation (37). Also, we proposed that ␣and ␤-DG function coordinately to assemble an extracellular and intracellular matrix of proteins (38), which is consistent with a recent report suggesting novel extracellular interactions for DG (39). In this study, we have taken advantage of the well defined role of DG at NMJs to determine whether the extracellular domains of DG function independently in AChR aggregation. We describe here that the functions of DG in AChR aggregation and basement membrane assembly in myotubes can be rescued by viral expression of full-length DG in DGϪ/Ϫ myotube cultures. Interestingly, DG constructs consisting of ␣-DG alone or of ␣-DG and the extracellular regions of ␤-DG can regulate aspects of AChR aggregation, namely the size of AChR aggregates, the distribution of microclusters within them, and the assembly of laminin at these aggregates. The implications of these observations for synapse formation and signaling via DG are discussed.

EXPERIMENTAL PROCEDURES
Generation of Dystroglycan cDNA Constructs-Dystroglycan cDNA constructs (Fig. 1A) were generated by PCR amplification using specific oligonucleotides for mouse dag1. PCR products were subcloned inframe into the peGFP-N1 expression vector (Clontech) to generate eGFP fusion transcripts. The CD4 ectodomain/transmembrane sequence expression plasmid was obtained from Dr. R. Dunn (McGill University). Constructs ␣-DG and ␣/⌬ecto␤DG were obtained from Dr. S. Winder (University of Sheffield, UK). All constructs were sequenced for validation (Genome Center, McGill University).
Generation of Recombinant Adenoviruses-Replication-defective adenoviruses were generated using the AdEasy system (Qbiogene) as described by the manufacturer. In brief, eGFP, CD4-eGFP, DG, ␣-DG, ⌬cytoDG, and ␣/⌬ecto␤DG cDNA fusion constructs were subcloned into the multiple cloning site of pShuttle-CMV. All vectors were linearized using PmeI, and cotransfected with pAdEasy-1 into the BJ5183 electrocompetent E. coli strain. Recombinants were screened by colony size and by restriction digests as described (40). Recombinants were linearized using PacI, purified, and transfected into HEK293 cells using Lipofectamine (Invitrogen). Adenovirus production was observed by plaque formation, and particles were isolated from cellular lysates and amplified through several rounds of infections in HEK293 cells. Titers were determined by TCID 50 infection test as described by the manufacturer (Qbiogene).
Quantification of AChR Aggregation-Soluble recombinant rat agrin was a generous gift from Dr. M. Ferns (University of California, Davis). To assay AChR clustering in ES cells, differentiated cultures (20 -25 days old) were treated with 0.5 nM agrin for 16 h. Cells were washed twice in Dulbecco's phosphate-buffered saline (DPBS) (Invitrogen) and incubated with 2 g/ml rhodamine-conjugated ␣-bungarotoxin (Molecular Probes) for 20 min at 25°C. Cells were washed twice with DPBS and fixed with 2% paraformaldehyde/PBS at 37°C for 20 min. Cultures were mounted on glass coverslips using ImmunoFloure mounting medium (ICN) and observed with a Zeiss Axioskop fluorescence microscope. Digital images observed with a ϫ63 Zeiss objective were captured with a QImaging Retiga 1300 10-bit digital camera under normalized exposures and quantifications of cluster number, size, and density were generated using Northern Eclipse software (Empix). For confocal images, cultures were observed using a Leica DM LFSA microscope equipped with an Ultraview confocal scanner (PerkinElmer Life Sciences), and images were captured and processed using Metamorph software (Universal Imaging). For cells infected with adenovirus, AChR aggregate size was determined by circling each aggregate with a trace from the free hand tool and by measuring total area as calibrated for the objective used (Fig. 2B). AChR aggregate density was determined as a percentage of total aggregate area (18,41), where the area of each aggregate that was labeled with ␣-bungarotoxin was highlighted using the threshold function (values of 18 -255 pixels), and the highlighted area was determined within the traced area (Fig. 2B). AChR aggregate numbers were determined by counting the number of aggregates present in each microscopic field observed. At least 30 nonoverlapping fields were visualized for each set of cultures, and data were collected from at least three separate experiments. For statistical analyses, we normalized the data obtained as described above for AChR aggregate size and density by a logarithmic transformation, and we then performed multiple comparisons using unbalanced one-way analysis of variance followed by pairwise comparisons (Tukey-Kramer and Fisher's tests) on the normalized data. All probability (p) values stated in the text are for analysis on normalized values. All analyses and percentile plots were generated using StatView 4.5 package (Abacus) or SAS (SPSS). All images and figures were prepared using Adobe Photoshop and Illustrator.
Immunocytochemistry-Cultures were stained with ␣-bungarotoxin as described above and then fixed in 2% paraformaldehyde/DPBS at 37°C for 20 min. If applicable, cells were permeabilized in 0.5% Triton X-100 in DPBS for 10 min at 25°C. Cells were blocked with 10% donkey or horse serum at 25°C for 1 h. This was followed by incubation with either monoclonal antibody IIH6 to ␣-DG (1:50) (Upstate Biotechnologies, Inc.), rabbit anti-sera to laminin (homemade, 1:100), anti-GFP antibody (Living Colors; 1:50) (Clontech), or rabbit anti-sera to ␤-DG (homemade, 1:100) at 25°C for 1 h. Cells were washed extensively with DPBS and incubated with rhodamine-conjugated goat anti-mouse IgM (1:100), aminomethylcoumarin acetate-conjugated donkey anti-rabbit (1:100), or rhodamine-conjugated donkey anti-rabbit IgG (1:100) secondary antibodies (Jackson ImmunoResearch), at 25°C for 1 h. Following a second series of washing, cells were mounted on glass coverslips and analyzed as described above. For laminin-AChR overlap observations, we scored positive only the aggregates that discretely overlapped with laminin immunoreactivity. We normalized to the total number of clusters observed in each case. A minimum of 20 nonoverlapping fields was scored in three experiments for each set of cultures, and differences were analyzed by analysis of variance followed by Fisher's test.

RESULTS
A simple model for DG function predicts that extracellular interactions mediated by ␣-DG would in turn signal intracellularly via ␤-DG. This model is complicated by the fact that most ligands for ␣-DG bind to an extended carbohydrate side chain (for review see Ref. 42), and transmembrane signaling would require the unprecedented translation of this interaction into conformational changes in both ␣and ␤-DG. Considering that interactions of the ␣ subunit are extracellular and of the ␤ subunit are intracellular, we suggest to determine whether ␣and ␤-DG might have independent functions in organizing cell surface proteins by generating a series of DG cDNA constructs bearing deletions of ␤-DG (Fig. 1). We fused these cDNA constructs at their C terminus to the enhanced green fluorescent protein (eGFP; see "Experimental Procedures"). In this study, we focused on constructs that reveal distinct functions of ␣-DG versus full-length DG (Fig. 1). Construct DG encodes full-length ␣and ␤-DG subunits; construct ⌬cytoDG lacks all cytoplasmic domains of ␤-DG; construct ␣-DG encodes the ␣ subunit of DG; construct ⌬ecto␤DG lacks the extracellular domains of ␤-DG leaving its transmembrane region following the C-terminal end of ␣-DG. All constructs were transiently expressed by infecting muscle cells derived from DGϪ/Ϫ (dag1) ES cells (12,18) using serotype Ad5 replicationdefective adenoviruses under the CMV promoter (see "Experimental Procedures"). DGϪ/Ϫ cells form diffuse aggregates of AChRs in response to agrin ( Fig. 2A), and the area occupied by the AChR microclusters as well as their density were quantified by computer-assisted image analysis ( Fig. 2B) (18).
Expression of DG Constructs in Myotubes Derived from DGϪ/Ϫ ES Cells-Infection of agrin-treated myotubes with DG-expressing adenoviruses resulted in a diffuse intracellular accumulation of the eGFPtagged constructs (Fig. 3, A-L), a consequence of the constitutive nature of the promoter. To assess the expression of these constructs at the cell surface in infected myotubes, we immunolabeled cultures using three antibodies as follows: monoclonal antibody (mAb) IIH6 against glycosylated ␣-DG on nonpermeabilized cultures (43) to visualize cell surface expression; a rabbit antiserum to the last 15 amino acids of ␤-DG (38) on permeabilized cultures; and an anti-serum to eGFP. DGϩ/ϩ cells infected with a control eGFP-expressing adenovirus (Fig. 3, A and B) showed regions of aggregated surface labeling of ␣-DG (Fig. 3AЈ) and of ␤-DG (Fig. 3BЈ) in a distribution that is identical to uninfected DGϩ/ϩ cells (data not shown). As expected, DGϪ/Ϫ cells infected with eGFP ( Fig. 3, C and D) expressed neither ␣-DG (Fig. 3CЈ) nor ␤-DG (Fig. 3DЈ). Expression of full-length DG in DGϪ/Ϫ cells (Fig. 3, E and F) showed aggregates of ␣-DG (Fig. 3EЈ) and ␤-DG (Fig. 3FЈ) similar to that in DGϩ/ϩ cells (Fig. 3, AЈ and BЈ). Our immunofluorescence analysis indicated approximately equivalent levels of expression of DG constructs in DGϪ/Ϫ cells when compared with endogenous levels in DGϩ/ϩ cells. Expression of ⌬cytoDG (Fig. 3, G and H) resulted in surface expression of this construct which encompasses both ␣-DG (Fig. 3G') and the extracellular and transmembrane regions of ␤-DG as detected by anti-eGFP (Fig. 3HЈ). In this instance, labeling with mAb IIH6, which recognizes a functional carbohydrate epitope on ␣-DG, indicated that proper glycosylation of ␣-DG does not require ␤-DG cytoplasmic domains. Expression of ␣-DG lacking any anchor to the cell surface via ␤-DG (Fig. 3, I and J) also resulted in its surface deposition on myotubes as detected by both IIH6 and anti-GFP antibodies in nonpermeabilized cells (Fig. 3, IЈ and JЈ). The distribution of this ␣-DG construct on DGϪ/Ϫ cells was similar to the endogenous ␣-DG expression observed on DGϩ/ϩ cells (Fig. 3AЈ). ␣-DG was not readily detected on the surface of infected non-muscle cells (data not shown), suggesting that ␣-DG, when expressed alone, preferably bound to muscle cells. The heterogeneity of ES-derived cell cultures and their low abundance in myotubes prevented us from assessing glycosylation or expression of ␣-DG by Western blot analysis. Nonetheless, in another study, we determined that expression of ␣-DG alone in adenovirus-infected COS cells resulted in its secretion into the culture medium (data not shown), and it could be identified by Western blotting using mAb IIH6 (data not shown). Finally, expression of construct ␣/⌬ecto␤DG (Fig. 3, K and L) showed surface labeling of ␣-DG (Fig. 3KЈ) and ␤-DG (Fig. 3LЈ) as observed for endogenous or ectopic full-length DG (Fig. 3, EЈ and FЈ). Taken together, these observations show that DG cDNA constructs could be expressed in myotubes derived from DGϪ/Ϫ ES cells using adenoviruses and that both ␣and ␤-DG were normally expressed at the cell surface. Proper surface targeting and glycosylation of ␣-DG could occur in the absence of ␤-DG expression. Moreover, the association of  Expression of a membrane-bound eGFP (CD4eGFP) in DGϪ/Ϫ myotubes does not alter the distribution of AChR microclusters (right panel). Pictures show a single aggregate for each myotube. B, methods for measuring size and density of AChR aggregates on myotubes. Using Northern Eclipse software (version 6.0, Empix), captured images of bungarotoxin-labeled AChR aggregates were inverted and calibrated for the ϫ63 Zeiss objective used; using the freehand tool, aggregates were encircled, and the area was measured from the "measure" command to evaluate the total aggregate area (upper right panel). The same aggregate was processed using the threshold tool to highlight the AChR microclusters labeled within each aggregate (lower left panel). The area from the highlighted AChRs was measured. To express the relative distribution of AChRs within the aggregate (density), we used the ratio between the measured value for area of AChR microclusters within aggregates to the measured value for total aggregate size (lower right panel). Bar, 15 m. a form of ␣-DG lacking any anchorage to ␤-DG still was able to interact with cell surface proteins either in the ECM or in the plasma membrane possibly affecting interactions and functions of the DG complex.
Expression of Dystroglycan in DGϪ/Ϫ Myotubes Restores the Size of AChR Aggregates-Myotubes derived from both DGϩ/ϩ and DGϪ/Ϫ ES cells express AChRs and respond to soluble agrin by forming aggregates of receptors on their surfaces (17,18), indicating that DG is downstream of the agrin/MuSK signaling pathway (44). However, in the absence of DG, these agrin-induced aggregates are an abnormally large, diffuse, and unstable collection of microclusters of AChRs ( Fig. 2A) (18). Furthermore, these DG-deficient aggregates lack laminin, merosin, perlecan, and AChE (18). To ascertain whether ␣and ␤-DG act only as a complex in response to MuSK activation, we infected DGϪ/Ϫ myotubes with adenoviruses expressing either eGFP or DG constructs (Fig.  1), and 5-7-day post-infection, treated the cultures with a saturating concentration (0.5 nM) of soluble agrin to induce AChR aggregation at the surface of myotubes (45), and labeled AChRs using rhodamineconjugated ␣-bungarotoxin. As controls for the assays, we used soluble cytoplasmic eGFP or submembrane eGFP (CD4-eGFP). This latter construct allowed us to control for any effect due to membrane expression of eGFP on AChR aggregation. Expression of either control constructs in DGϩ/ϩ or DGϪ/Ϫ cells did not affect the aggregation of AChRs after agrin treatment ( Figs. 2A and 4).
To assess the relative contribution of extracellular and intracellular interactions of DG to AChR aggregation, DGϪ/Ϫ cells were infected with adenoviruses expressing either ⌬cytoDG, ␣-DG, or ␣/⌬ecto␤DG (Figs. 3 and 4) and treated with soluble agrin. Expression of ⌬cytoDG, which includes full-length ␣-DG and the extracellular domains of ␤-DG to which it binds, in DGϪ/Ϫ cells (Fig. 4) resulted in a highly significant   but partial reduction in aggregate size (mean, 96.9 Ϯ 6.3 m 2 ) when compared with DGϪ/Ϫ:eGFP (p Ͻ 0.0005; see Fig. 5). Expression of ␣-DG also resulted in reduction of aggregate size (mean, 77.6 Ϯ 4.2 m 2 ) when compared with DGϪ/Ϫ:eGFP (p Ͻ 0.0001; see Fig. 5). Interestingly, expression of ␣/⌬ecto␤DG resulted in full recovery of AChR aggregate size (mean size, 47.8 Ϯ 2.9 m 2 ) when compared with DGϪ/Ϫ:eGFP (p Ͻ 0.0001) and similar to the expression of endogenous or ectopic full-length DG. We did not observe any increase in numbers of AChR aggregates per myotube after expression of the different DG constructs in DGϪ/Ϫ cells (data not shown). These observations suggest that interactions mediated via extracellular regions of DG can function to regulate AChR aggregate size. Moreover, expression of ␣-DG and of its ␤-DG-binding sites can function as well as the noncovalently associated complex of ␣and ␤-DG.
Dystroglycan Regulates the Distribution of AChR Microclusters within Aggregates-In cultured myotubes, agrin in solution (46,47) or released by the nerve (48) stimulates the formation of microclusters of AChRs, which subsequently condense into a mature postsynaptic plaque. Myotubes deficient for DG fail to form such plaques and as a result the area of an aggregate occupied by AChRs is greater and the microclusters remain distinct puncta ( Fig. 2A). Because full-length or extracellular DG regulate aggregate size (see above), we asked whether extracellular interactions of DG were also capable of controlling AChR packing density of microclusters within agrin-induced AChR aggregates. In this case, we defined packing density as the relative area of an aggregate covered by AChR microclusters labeled with fluorescent ␣-bungarotoxin (Fig. 2B) (18,41). When quantified in myotubes derived from DGϩ/ϩ cells treated with agrin, this packing density yielded a frequency histogram with a normal distribution and a mean of ϳ70%. Uninfected or eGFP-expressing DGϩ/ϩ cells showed the same mean (66.7 Ϯ 2.0%; see Fig. 6), but uninfected or eGFP-expressing DGϪ/Ϫ cells had a significantly lower relative area occupied by microclusters (mean, 45.1 Ϯ 2.8%; p Ͻ 0.0001). Expressing full-length DG in DGϪ/Ϫ cells clearly increased the relative area occupied by microclusters with a mean of 56.7 Ϯ 1.9% when compared with DGϪ/Ϫ:eGFP (p ϭ 0.022), but it remained somewhat lower than in DGϩ/ϩ:eGFP cells (p ϭ 0.0474; see Fig. 6). Thus, expression of full-length DG significantly but not completely rescued the normal distribution of microclusters within aggregates on myotubes derived from DGϪ/Ϫ ES cells to produce a density of AChRs approximately that in DGϩ/ϩ cells. Interestingly, expression of ⌬cytoDG in DGϪ/Ϫ cells also significantly increased the area occupied by microclusters (mean, 60.2 Ϯ 1.5%) when compared with DGϪ/Ϫ:eGFP (p ϭ 0.005; see Fig. 6), suggesting that the extracellular domains of DG are responsible for most if not all of the activity of full-length DG in regulating AChR density. We observed a similar significant increase when either ␣-DG (mean, 60.8 Ϯ 2.2%; p Ͻ 0.0001) or ␣/⌬ecto␤DG was expressed (mean, 65.7 Ϯ 2.0%; p Ͻ 0.0001). When the populations are ranked by percentiles (Fig. 6B), the distribution plot for DGϪ/Ϫ:⌬cytoDG cells looked similar to that for DGϩ/ϩ:eGFP or DGϪ/Ϫ:DG cells but was distinct from control DGϪ/Ϫ:eGFP cells. In contrast, the plots for DGϪ/Ϫ:␣-DG and DGϪ/Ϫ:␣/⌬ecto␤DG cells showed a greater variation across the percentiles. Together, these data suggest that DG regulates the distribution of microclusters of AChRs within aggregates induced by agrin, and extracellular interactions via ␣-DG function to achieve this regulation.
Localization of Dystroglycan to AChR Aggregates-Dystroglycan interacts with rapsyn (14) and localizes precisely to nascent aggregates of agrin-induced AChRs on skeletal myofibers in vivo (16). However, in C2C12 myotubes treated with agrin, DG accumulates in most but not all AChR aggregates (8,43). Furthermore, in the same myotubes treated with laminin-1 that can also stimulate AChR aggregation via DG, DG is not colocalized with AChR aggregates (49). Because DG regulates AChR aggregation, we asked whether minimal constructs of DG are localized to these aggregates in DGϩ/ϩ and DGϪ/Ϫ myotubes. To assess this, we immunolabeled DG using mAb IIH6 (see Fig. 3) to detect an overlap between surface ␣-DG and AChR aggregates (Fig. 7). In DGϩ/ϩ cells infected with eGFP, DG was expressed along the myofiber surface and accumulated with several but not all AChR aggregates (Fig.  7, DGϩ/ϩ:eGFP, arrows). As expected, in DGϪ/Ϫ cells infected with eGFP, there was no DG detected (Fig. 7, DGϪ/Ϫ: eGFP, arrowheads). Only full-length DG expressed in DGϪ/Ϫ cells localized to most AChR aggregates (Fig. 7, DGϪ/Ϫ:DG, arrows) but not all (arrowheads). Expression of ⌬cytoDG, ␣DG, or ␣/⌬ecto␤DG in DGϪ/Ϫ cells was detected along the myotube surface (Fig. 3) but was seldom found colocalized with AChR aggregates (data not shown). That full-length DG is found more frequently at AChR aggregation may reflect targeting to clusters via interactions with rapsyn. Constructs lacking these domains are not similarly targeted but nevertheless can function in regulating aggregate size and density. Possibly this is mediated by participation of DG in ECM assembly, which at sites of AChR aggregation, even distant ones, can affect AChR aggregation by interactions with MuSK (50) or some proteins essential to the primary AChR scaffold (discussed below).

DISCUSSION
We have demonstrated previously that ␣and ␤-DG are recruited to developing NMJs (16) where they regulate the size and the stability of postsynaptic densities of AChRs and the assembly of a synaptic basement membrane (12,18). We hypothesized that DG functions via an extracellular as well as an intracellular matrix of proteins (38); consistent with this, a recent report suggests novel extracellular interactions of ␣-DG in AChR aggregation (39). Here we show that ␣-DG is glycosylated and targeted to the cell surface in absence of ␤-DG (Fig. 3). We also show that the DG complex can function extracellularly through ␣-DG  to regulate the size of AChR aggregates (Figs. 4 and 5) and the density of AChR microclusters within them (Figs. 4 and 6). Moreover, DG constructs lacking either the extracellular or cytoplasmic domains of ␤-DG can mediate laminin assembly at agrin-induced AChR aggregates (Fig.  8). We conclude that the ␣ and ␤ subunits of the DG complex, which have been widely viewed as two interacting and inter-dependent subunits in skeletal muscle, can function independently (discussed below).
Processing and Targeting of Dystroglycan in Muscle-Surface expression and glycosylation of DG are crucial for proper function of the DGC in skeletal muscle and in the brain (60,61). Aberrant glycosylation of ␣-DG, due to mutations in glycosylating intermediates, or reduced surface expression of ␣-DG, due to lack of dystrophin, leads to the development of muscular dystrophies (60 -63). Previous studies from Campbell and co-workers (64) have identified a site for the interaction of Large, an important step in the glycosylation pathway of DG within its ␣ subunit. Our results extend this by showing that not only is ␤-DG unnecessary for this glycosylation but ␣-DG can be targeted to the cell surface in the absence of ␤-DG (Fig. 3). Indeed, expression of constructs encoding ␣-DG and ⌬cytoDG are detected by mAb IIH6 (Fig. 3), which recognizes an O-linked carbohydrate side chain on ␣-DG at the myotube surface that interacts with laminin G domain-containing ligands (65,66). This indicates that the DG cytoplasmic domain is not necessary for the functional glycosylation of the ␣ subunit. In fact, this glycosylation may be important in targeting ␣-DG to the cell surface, which could be facilitated by the association with basement membrane components, most likely perlecan, agrin, or laminin.
␣-DG contains at its C-terminal end a region highly homologous to a sperm protein, enterokinase, agrin (SEA) module (67) that is found in membrane-tethered mucin proteins (68). In human MUC-1, the SEA module has been shown to undergo autoproteolysis. Point mutations of a conserved serine (Ser-53) residue within the SEA module abrogate this cleavage activity (69). Similarly, in vertebrates, cleavage of DG into ␣and ␤-DG depends upon a serine (Ser-655) found in this putative SEA module region (67). However, inhibition of cleavage of DG subunits in point mutants does not prevent the cell surface localization of DG in heterologous cells. These observations raise the question of why vertebrate DG evolved into two subunits. One possibility is to generate distinct functions for the two subunits cleaved from the DG precursor peptide. Noncovalent association between ␣and ␤-DG may allow the two subunits to dissociate and possibly interact independently with other partners in the cell or in the ECM. Consistent with this, there are several reports of disjunction in ␣and ␤-DG localization in non-muscle tissue (70 -72). 3 ␣-Dystroglycan in AChR Aggregation-Here we have utilized AChR aggregation as a relatively simple and well characterized cellular event that requires DG (12,17,18,44,73) to study its function in muscle where it is essential for cell integrity in vivo (12) as well as synapse formation. We show that expression of full-length DG or of DG lacking the ␤ subunit in DGϪ/Ϫ myotubes alters AChR aggregate morphology. Both the size of AChR aggregates and the density of the microclusters within aggregates are the same as those in DGϩ/ϩ cells following the expression of DG in DGϪ/Ϫ myotubes (Figs. 4 -6). Expression of DG constructs lacking the ␤ subunit or its cytoplasmic regions was also sufficient to recover the wild-type morphology of DG-null AChR aggregates (Figs. 4 -6). These observations suggest that ␣-DG can act independently from ␤-DG in regulating AChR aggregation. The initial interactions of ␣-DG must be extracellular, distinguishing it from classical transmembrane signaling. By extension, these data indicate that DG, and more broadly the DGC, may function extracellularly in formation of NMJs. Recent studies have reported that inactive muscle agrin can be potentiated to stimulate AChR aggregation when complexed with laminin into an ECM (57,74). The association of ␣-DG with laminin, agrin, and perlecan may similarly facilitate the assembly of endogenous ligands to stimulate AChR aggregation. More generally, extracellular interactions appear critical for NMJ formation. For example, in Caenorhabditis elegans, LEV-10, a type I transmembrane protein containing Clr/Cls, Uegf, and bone morphogenic protein and low density lipoprotein a domains (75), and CAM-1 (Canal-associated neurons abnormal migration protein 1), a retinoic acid-related orphan receptor tyrosine kinase resembling MuSK (76), are necessary for the aggregation of AChRs at nascent NMJs in a process involving only extracellular interactions. This extracellular signaling may serve to coordinate multiple transmembrane pathways during the assembly of the extremely complex structure of the postsynaptic membrane.
The mechanism of action of DG in AChR aggregation is dependent upon its levels of expression. The absence (12,18) and excessive levels (77, 78) 4 of full-length DG or ␤-DG alone in skeletal myotubes lead to aberrant postsynaptic densities of AChRs. In our study, extracellular portions of DG (␣-DG and ⌬cytoDG) seem to largely dictate AChR aggregate morphology (Figs. 5 and 6). The size of these aggregates on ␣-DGor ⌬cytoDG-expressing DGϪ/Ϫ cells, although significantly smaller than control DGϪ/Ϫ cells, are larger than in DGϪ/Ϫ cells expressing full-length DG or ␣/⌬ecto␤DG. The microcluster distribution within these aggregates is the same in these former cell populations. Thus, the total number of microclusters within aggregates on myotubes expressing ␣-DG and ⌬cytoDG is apparently greater than on those expressing full-length DG or ␣/⌬ecto␤DG. This effect of DG expression could be explained by an overaccumulation of AChR microclusters within aggregates resulting from an altered metabolic turnover (decrease) of AChRs mediated by the extracellular portions of DG. Related to this, Xu and Salpeter (79) have reported that AChR turnover in vivo is increased in mouse muscle lacking dystrophin.
Dystroglycan in Extracellular Signaling-DG is necessary for the formation of some (58) but not all (7) basement membranes. For example, DG is necessary in development for the maintenance of the extra-embryonic Reichert's membrane (80). Deletion of DG from the brain leads to aberrant basement membranes at astroglial end feet on blood vessels and meninges (60). On the other hand, DG-deficient skeletal muscle fibers assemble an ultrastructurally normal basement membrane containing laminin, collagen, fibronectin, and perlecan (12). In our myotube cultures, DG is necessary for the synaptic localization of laminin (Fig. 8) (18). The constructs ␣-DG and ⌬cytoDG clearly target laminin to AChR aggregates (Fig. 8). This suggests that the ␣-DG subunit may function as an ECM protein to assemble a basement membrane involved in the regulation of AChR aggregation (Figs. 4 -6) (74) and in the accumulation of laminin at AChR aggregates (Fig. 8).
Extracellular interactions via ␣-DG appear to be essential in DG function at NMJs because the disruption of ␣-DG interactions with laminin by blocking antibodies (43), with laminin-2 by deletions (57), or by hypoglycosylation of ␣-DG, as observed in myd mice (81), leads to aberrant NMJs with disrupted AChR aggregates at the end plate. Also, the idea that the extracellular domains of DG appear sufficient to target and maintain laminin to AChR aggregates (Fig. 8) and to contribute to basement membrane assembly is consistent with observations that laminin binding to sulfatides (6) can seed basement membrane assembly via DG and utrophin and emphasizes the contribution of extracellular interac-tions of the DGC to this process. In our study, full-length DG and ␣/⌬ecto␤DG contain the cytoplasmic regions of ␤-DG responsible for interaction with Grb2 (growth factor receptor-bound protein 2), rapsyn, caveolin 3, and dystrophin/utrophin (15,(82)(83)(84). These regions appear to functionally complement those of the extracellular portion of the DG complex by regulating AChR aggregate morphology (Figs. 4 -6) and basement membrane assembly (Fig. 8).
Our results raise the question of how extracellular domains of DG function in skeletal muscle in AChR aggregation in ways often attributed largely to direct or indirect interactions of AChRs with cytoskeletal proteins. Interactions of matrix proteins have the potential to introduce a level of extracellular protein-protein interactions that eventually signals into the cell interior via several mechanisms. These may include functions of glycosylated ␣-DG in signaling cross-talk with integrins; laminin bound to DG could bind via distinct sites to ␣6 or ␣7 integrins (7,57) in muscle to signal intracellularly. Other mechanisms could involve perlecan, a ligand for ␣-DG and a well known coreceptor for FGF2 (85,86); neuregulin-1, which regulates AChR synthesis and is a growth factor/ECM protein that is bound to heparan sulfate proteoglycans (87); or AChE, which binds to ␣-DG via perlecan (35,36) and interacts directly with MuSK to alter its distribution (50). Similar extracellular interactions may occur in central nervous system synapses where DG is expressed in neurons of the cerebral cortex and hippocampus (88) and has been implicated in long term potentiation (61) and in synaptic transmission in the retina (89). Conceivably, DG and the DGC may regulate the aggregation of neurotransmitter receptors by virtue of their stabilization in the plasma membrane in a fashion similar to that at NMJs. Further studies to elucidate the mechanisms of DG function in cell survival and synapse formation may contribute to our understanding of the muscle wasting and mental retardation associated with muscular dystrophies.