Co-localization of 1,4-dihydropyridine receptor α₂/δ subunit and N-CAM during early myogenesis in vitro

After myoblast fusion into myotubes, the differentiation of muscle is characterized by the acquisition of contractile properties accompanied by a high degree of plasma membrane specialization. The transverse-(T-) tubule network, the depolarization of which induces a signal transmitted to the sar-coplasmic reticulum resulting in Ca²⁺release and contraction (see Rios et al., 1992), is such a specialized domain: formed by invagination of the plasma membrane, it is connected to the sarcoplasmic reticulum at the level of the triads (Franzini-Armstrong and Porter, 1964; Huxley, 1964). When innervation takes place, another highly specialized membrane domain is defined, the neuromuscular junction. Proteins such as acetylcholine receptor (Sytkowski et al., 1973; Fischbach and Cohen, 1973) and neural cell adhesion molecule (N-CAM) (Tassin et al., 1991) accumulate at the membrane surface and become restricted to the neuromuscular junction (Rieger, 1985; Covault and Sanes, 1986). Other proteins, such as 1,4-dihydropyridine (DHP)-sensitive calcium channels, first found at the surface membrane, migrate to the T-tubules at later stages of development (Romey et al., 1989). Thus, the plasma membrane represents a common site of expression for many proteins prior to their redistribution in different and highly specialized membrane regions.

The DHP receptor of muscle has been the subject of intensive research over the past ten years. Initially defined as a channel mediating a slow calcium current (Sanchez and Stefani, 1978; Almers et al., 1981), it was later found to act mainly as a voltage-sensor (Rios and Brum, 1987), mediating excitation-contraction coupling through a still poorly defined mechanism independent of inward calcium fluxes (see Rios et al., 1992). Purification of the muscle DHP receptor (see Hosey and Lazdunski, 1988; Campbell et al., 1988) has led to the identification of at least four subunits, α₁, α2/δ, β and g, which have all been cloned (Tanabe et al., 1987; Ellis et al., 1988; Ruth et al., 1989; Jay et al., 1990). The α₁subunit (M_r170,000-200,000) carries the binding sites for pharma-cological ligands such as the 1,4-dihydropyridines (Striessnig et al., 1986; Hosey et al., 1987; Sieber et al., 1987), as well as the calcium channel activity and the voltage sensor function (Tanabe et al., 1987, 1988; Bois et al., 1991; Beam et al., 1992). α2/δ is a heavily glycosylated polypeptide of M_r170,000, constituted of the α₂subunit per se (M_r145,000) linked to the d subunit (M_r17,000-25,000) by disulfide bonds (Barhanin et al., 1987; Jay et al., 1991). The β (M_r55,000) and g (M_r30,000) subunits seem to affect the level of expression and the electrophysiological properties of α₁from different tissues (see Catterall, 1991; Snutch and Reiner, 1992).

The case of α2/δ is particularly interesting. In muscle, it is involved in the modulation of DHP binding on α₁(Woscholski and Marme, 1992), but how α2/δ might regulate the channel and the voltage sensor functions of α₁is still unclear (Varadi et al., 1991; Singer et al., 1991, 1992). The fact that the association of α2/δ with α₁is relatively weak (Takahashi et al., 1987), and that α2/δ expression precedes that of α₁during development of muscle in vivo (Morton and Froehner, 1989) and in vitro (Varadi et al., 1989) has led some to believe that α2/δ may have other functions than regulating the DHP receptor during myogenesis (Morton and Froehner, 1989). In the nervous system, the neuronal α2/δ isoform (α_2b) seems necessary for an efficient cell surface expression of the N-type calcium channel (Brust et al., 1993), suggesting that α2/δ is involved in cellular mechanisms related to molecular targeting to the cell membrane.

The highly heterogeneous cell surface glycoprotein N-CAM is clearly implicated in the formation and differentiation of the neuromuscular system, as well as in morphogenetic events in other tissues (Edelman, 1986, 1987). Like N-cadherin, which is involved in Ca²⁺-dependent adhesion events (Knudsen et al., 1990b), N-CAM is implicated in the Ca²⁺-independent aggregation of myoblasts and the rate of myotube formation (Dickson et al., 1990; Knudsen et al., 1990a; Mège et al., 1992), and probably in the regulation of gene expression at later stages of myogenesis (see Knudsen, 1990). It has been proposed that homophilic N-CAM–N-CAM binding, rather than representing an adhesive interaction stricto sensu, could also influence the function of other proteins (Rutishauser et al., 1988). In neurons, for example, N-CAM is involved in neurite outgrowth via an interaction with G proteins, which in turn modulate L-type calcium channels associated with the DHP receptor (Doherty et al., 1990, 1991).

Pilot experiments dealing with the immunolocalization of α2/δ on rabbit muscle cells during myogenesis in vitro have shown a membrane distribution strongly resembling that of N-CAM as described by Tassin et al. (1991). These results prompted us determine whether these two proteins would be co-distributed in muscle cell culture. We have used a polyclonal antibody recognizing all N-CAM forms in vivo and in vitro (Rieger et al., 1985; Tassin et al., 1991), and the monoclonal antibody (mAb) 3007, which was previously shown to immunoprecipitate the DHP receptor (Vandaele et al., 1987). In addition to demonstrating unambigously that the epitope recognized by mAb 3007 is located on α2/δ, the present data reveal that the α2/δ subunit and N-CAM are to a large extent co-localized during early myogenesis.

T-tubule membranes from adult rabbit skeletal muscle were prepared according to Galizzi et al. (1984). Solubilization of T-tubules in digitonin, preparation of wheat-germ agglutinin (WGA) extract, iodination by the lactoperoxidase method, and immunoprecipitation with mAb 3007 were performed as described (Vandaele et al., 1987), except for the use of additional proteolysis inhibitors in all buffers: 1 mM iodoacetamide, 100 μM paramethyl-sulfonylfluoride, 1 μM pepstatin A, 10 μg/ml leupeptin, 1 μg/ml aprotinin and 20 μg/ml soybean trypsin inhibitor (Sigma). The immunoprecipitated material was analyzed by SDS-PAGE in the presence of β-mercaptoethanol on 4% to 14% linear gradient gels as described (Laemmli, 1970). Dried gels were exposed for 1-3 days to XAR 5 Kodak films.

Primary cultures of skeletal myogenic cells were obtained from muscles from the backs of 3- to 8-day-old rabbits. Muscles were aseptically dissected out, minced and rinsed in a Ca2+/Mg2+-free solution containing 8 mM NaH₂PO₄, 22.6 mM NaHCO₃, 116 mM NaCl, 5.3 mM KCl, 5.5 mM glucose and 10 mg/ml Phenol Red, pH 7.4, prewarmed at 37°C. The tissue was dissociated by 4 successive exposures of 15 minutes each to a 0.1% (w/v) trypsin solution (Seromed) made in the same buffer under gentle agitation at 37°C. Cells were filtered through a sterile 80 μm nylon gauze and centrifuged for 10 minutes at 800 g. The resulting pellet was resuspended in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 20% fetal calf serum (Gibco) and antibiotics (200 units/ml penicillin, 50 μg/ml streptomycin). A differential adhesion procedure (Yaffe, 1968) was then used to remove most fibroblasts from the culture, and to obtain myoblast-enriched primary cultures. The cells were first plated for 1 hour at a density of 2×10⁶cells per 75 cm²flask in a humidified 37°C incubator equilibrated at 5% CO₂. The fast-adhering cells, mainly fibroblasts, did not acquire the elongated shape characteristic of myoblasts, did not align and did not form myotubes after reaching confluence. The cells remaining in the supernatant were mostly myoblasts and were then plated at a density of 0.5×10⁶cells per 100 mm gelatin-coated dishes. When the cultures reached confluency, 4 to 7 days later, growth medium was replaced by differentiation medium (DMEM supplemented with 5% fetal calf serum), and this day was considered as T₀. Under these conditions, myoblasts differentiated rapidly into multinucleated myotubes. Differentiation medium was changed every 4 days.

In vitro labeling of rabbit muscle cells was carried out at T₀. Some control experiments were carried out one day after plating on fibrob-last-enriched culture (pre-plating step). Cell cultures were incubated for 30 minutes at 37°C with DMEM without methionine (Seromed, Paris) and without fetal calf serum, then for 4 hours in the same medium containing 200 μCi/ml [³⁵S]methionine. The experiment was then continued at 4°C, and all buffers except DMEM contained the following protease inhibitors: 1 mM iodoacetamide, 100 μM phenylmethylsulfonyl fluoride, 1 μM pepstatin A, 1 mM leupeptin, 1 μg/ml aprotinin. Cell cultures were washed twice in cold normal DMEM, and were scraped from the plate in a buffer containing 0.32 M sucrose, 20 mM HEPES/NaOH, pH 7.4. Cells were harvested by centrifugation (15 minutes, 2,000 g), resuspended in 140 mM NaCl, 20 mM HEPES/NaCl, pH 7.4 (buffer A), and homogenized with a Dounce homogenizer. After a 15 minute centrifugation at 10,000 g, the pellet was resuspended and incubated for 30 minutes in a buffer containing 500 mM KCl, 20 mM Hepes, pH 7.4. After another 15 minute centrifugation at 10,000 g, the pellet was resuspended in 0.2 ml of buffer A, and solubilized with the same amount of buffer A containing 2% CHAPS ([3-[(3-cholamidopropyl)-dimethylammonio]-1-propanesul-fonate, Sigma). After incubation for 1 hour, the material was centrifuged for 30 minutes at 100,000 g. The supernatant was incubated overnight under agitation with 20 μl of a 2% (w/v) suspension of Protein A-Sepharose Cl-4B (Pharmacia) preincubated with either 2 μl of ascites fluid 3007 or 20 μl of control ascites fluid 104F (equivalent to ∼15 μg IgG). The supernatant was then removed and the beads were washed six times with 1 ml of buffer containing 1% CHAPS, 1% Triton X-100, 140 mM NaCl, 20 mM HEPES/NaOH, pH 7.4, and three times with a buffer containing 0.1% CHAPS, 0.1% Triton X-100, 140 mM NaCl, 20 mM HEPES/NaOH, pH 7.4. Bound proteins were eluted from the beads with 0.1 M glycine, pH 2.8, and the pH was equilibrated to 7.0 using 1 M Tris-HCl, pH 7.0. The samples were denatured in Laemmli buffer containing 2.5% β-mercaptoethanol (reducing condition) or 40 mM iodoacetamide (non-reducing condition), and analyzed by SDS-PAGE. After fixation and staining in Coomassie Blue, 4% to 14% linear gradient gels were soaked in Enhancer (Amersham) for 30 minutes, dried, and exposed for 12-15 days to XAR 5 Kodak films.

Cell cultures were washed twice with DMEM without serum at 37°C, then fixed in 2% paraformaldehyde in phosphate buffered saline (PBS), pH 7.4, for 30 minutes. After two washes in PBS, they were incubated for 10 minutes in 0.1 M glycine in PBS, and then washed with PBS. All antibodies were diluted in 0.1% gelatin in PBS, and 4% goat serum was added to primary antibody dilutions. All incubation steps were carried out directly in the culture dish. They lasted 2 hours each, except with mAb 3007 (overnight) and avidin-rhodamine (30 minutes), and were each followed by six washes of 5 minutes. Fixed cells were first incubated at 4°C with mAb 3007 (ascites, 1/50), then with biotinylated goat anti-mouse IgGs (1/100; Vector). They were then incubated with polyclonal rabbit anti-N-CAM IgGs (140 μg/ml; Rieger et al., 1985), and with fluorescein-labeled (FITC) goat anti-rabbit IgGs (1/100, Biosys, Paris) to reveal N-CAM immunore-activity. mAb 3007 immunoreactivity was then revealed with rhodamine₆₀₀-labeled (XRITC) avidin D (1/100, Vector). The emission peak of XRITC (600 nm) is better separated than that of regular rhodamine (580 nm) in double-labeling experiment with FITC (525 nm). After six additional washes, the dishes were mounted in Mowiol 4-88 mounting medium containing 5 mM paraphenylenedi-amine. Control experiments included labeling with FITC-labeled goat anti-rabbit IgGs without first antibody, or after mAb 3007, and biotinylated goat anti-mouse IgGs followed by XRITC-labeled avidin without first antibody.

Immunofluorescent double labeling was examined with a Leitz epifluorescence photomicroscope, equipped for fluorescein and rhodamine fluorescence detection and for phase-contrast optics. Leitz Ploemepak^®filters were used to ensure a maximal separation of rhodamine and fluorescein emitted light. Each fluorophore was detected only with its respective filter. Proportions of co-localized versus non-co-localized N-CAM aggregates on myotubes were determined on photographs of two independent primary cultures at different times after T₀, and were expressed as % of total number of N-CAM aggregates. At least 100 aggregates were counted in each experiment.

The SDS-PAGE analysis of the proteins immunoprecipitated by mAb 3007 and control mAb 104F (Vandaele et al., 1987) is shown in Fig. 1A. In some experiments, the immunoprecipitates were washed in the presence of Triton X-100. All lanes were run in the presence of β-mercapthoethanol. Lane 1 corresponds to the whole starting material: iodinated digitonin-solubilized T-tubule WGA extract. Lane 2 shows that, in the absence of Triton X-100, three main polypeptides with M_rof 210,000, 175,000 and 145,000 are greatly enriched in the mAb 3007 immunoprecipitate. Other smaller polypeptides are co-precipitated, but in small amounts compared to the starting material. In the presence of Triton X-100 (lane 3), the immunoprecipitate consists of the M_r145,000 polypeptide, accompanied by two smaller polypeptides (M_rof 20,000 to 25,000), already present in lane 2. Control mAb 104F does not precipitate any of these polypeptides (lane 4).

The specificity of mAb 3007 was further examined in vitro on rabbit muscle cell cultures, which were biosynthetically labeled with [³⁵S]methionine at T₀. Fig. 1Bpresents the SDS-PAGE analysis of this material, in reducing (lanes 1-3) and non-reducing conditions (lanes 4-6). Lanes 1 and 4 show the starting material: [³⁵S]methionine-labeled solubilized cell membranes run in the presence of β-mercaptoethanol and iodoacetamide, respectively. As demonstrated in lanes 2 and 5, mAb 3007 precipitated only one polypeptide of M_r125,000, the size of which reflected the ongoing biosynthesis (see Discussion) and was not changed in reducing conditions. Control mAb 104F did not precipitate these or other polypeptides (lanes 3, 6).

The immunocytochemical localization of α2/δ and N-CAM was investigated in primary cultures of rabbit muscle cells at different stages of differentiation, up to T₁₂. Control experiments showed no specific reactivity of the secondary antibodies (Fig. 2A,C), and an absence of FITC-anti-rabbit IgG cross-reactivity with mAb 3007 (Fig. 2B). As presented in matching panels of Figs 2, 3and 4, both α2/δ and N-CAM immunola-beling were mostly detected as small spots and aggregates thereof. On myotubes, slight variations of the fine focus revealed that spots and aggregates were localized to the membrane. In the literature, such types of arrangement have sometimes been designated as clusters (Flucher et al., 1991) or spots (Tassin et al., 1991), for the present spots, and as clusters (e.g. see Conolly, 1984; Tassin et al., 1991) or aggregates (Olek et al., 1986), for the present aggregates. Some diffuse N-CAM staining was also observed on contracting myotubes (Figs 3, 4).

One day after plating, neither proliferating myoblasts nor fibroblasts (pre-plating step of the culture) showed any labeling higher than background for both α2/δ and N-CAM. In contrast, one or two days before T₀(fusion time), both labelings were detected in an identical, irregular distribution in elongated myoblasts typically aligned before fusion (Fig. 2G,Hand corresponding phase-contrast, I). Small spots, some of them overlying the unlabeled nuclear profile, were likely to be located at the membrane, but some perinuclear patchy immunoreactivity for both molecules was likely to be in the cytoplasm. Other cells, with a fibroblast-like shape, were not labeled.

At T₀, and with increasing frequency therafter, fusing cells, typically constituted of two aligned mononucleated cells connected by a relatively thin bridge, were readily identified by phase-contrast observation. These cells were labeled on their cell body with both α2/δ and N-CAM (Fig. 3A,B; phase-contrast not shown). In addition, bright aggregates of N-CAM spots only were now visible at their junction (Fig. 3B). No α2/δ was found in this particular location (Fig. 3A). Such labeling was observed at the junction of mononucleated cells fusing with other myoblasts or with already formed myotubes at T₆(Fig. 3C,D). Fusing cells were less frequent at T₁₀, and rarely found at T₁₂.

Around T₆, many myotubes with a regular elongated shape (Fig. 3C-F), but without spontaneous contractile activity, were visible in the cultures. As soon as myotubes were formed, aggregates of co-localized α2/δ and N-CAM were found at their surface (Fig. 3C,D). When observed at higher magnification, these aggregates were constituted of small, bright spots of varying dimensions (Fig. 3E,F). Few such spots were visible outside of aggregates and the co-localization of α2/δ and N-CAM was apparent at the level of individual spots. A slight diffuse N-CAM staining was also detected (Fig. 3D,F).

As the culture matured, myotubes widened out and formed a branching network with some local enlargements. By T₁₂, most myotubes showed this morphology (Fig. 4A-D). Spontaneous contractile activity was apparent on most myotubes. The size of the immunofluorescent aggregates was generally smaller, and numerous spots were now dispersed at the myotube surface. The diffuse N-CAM immunoreactivity was now prominent at the surface of myotubes (Fig. 4D,F). Some nuclear profiles were sometimes visible, indicating that some diffuse N-CAM immunoreactivity was also located in the cytoplasm (Fig. 4F). A few smaller myotubes as described at T₆were still present in the T₁₂cultures (Fig. 4A,B).

At T₁₂, the biggest α2/δ spots still showed co-localized N-CAM (Fig. 4C,D). Some α2/δ spots, smaller and more homogeneous in size, began to show regular alignments on contracting myotubes (Fig. 4E,G,H). Corresponding N-CAM arrangements were occasionally visible (Fig. 4C-F).

At every time interval examined, all α2/δ aggregates co-localized N-CAM, but a small proportion of N-CAM aggregates were not labeled for α2/δ (see example at T₁₂in Fig. 4A-D). The proportion of these aggregates of N-CAM alone was always evaluated between 2 and 3% of total.

mAb 3007 has been previously selected for its ability to precipitate [³H]PN200-110-labeled DHP receptors from solubilized T-tubule membrane from rabbit muscle (Vandaele et al., 1987). T-tubule membrane preparations routinely contained a high concentration of [³H]PN200-110 binding sites (70-90 pmol/mg protein) and were well suited for characterizing the antibodies. As demonstrated in our previous work (Vandaele et al., 1987) and in the results presented here, at least α₁and α2/δ were co-precipitated by mAb 3007 from iodinated digitonin-solubilized T-tubules. However, it remained unclear whether the epitope recognized by mAb 3007 was located on α₁or α2/δ, because mAb 3007 was ineffective in Western blotting experiment (Vandaele et al., 1987). In the present study, DHP receptor subunits were therefore dissociated by treating the immunoprecipitate with Triton X-100 (Takahashi et al., 1987). Under these conditions, α2/δ was the only subunit found in the precipitate, as demonstrated by its typical cleavage into the α₂(M_r145,000) and d components (M_r22-25,000) under reducing conditions of SDS-PAGE (Barhanin et al., 1987; De Jongh et al., 1990).

mAb3007 was demonstrated to be well suited for examining α2/δ immunoreactivity on muscle tissue. In adult rabbit muscle sections, a dotted pattern corresponding to the triadic structure of the T-tubule network was revealed by immunofluorescence (Toutant et al., 1990), in keeping with results on the localization of the α₁and α2/δ DHP receptor subunits obtained with other antibodies (Jorgensen et al., 1989; Flucher et al., 1990; Yuan et al., 1991). Before proceeding to the immunocytochemical studies on rabbit muscle cell cultures, we verified that mAb 3007 precipitated a unique polypeptide from solubilized rabbit muscle cells (T₀) biosynthetically labeled with [³⁵S]methionine. The M_r(125,000) of this polypeptide corresponded to the size of the amino acid sequence of the cloned α2/δ from muscle (Ellis et al., 1988), and to the size of the enzymatically deglycosylated polypeptide core of purified α2/δ (Barhanin et al., 1987; Jay et al., 1991). Indeed, the α₂and d components of α2/δ are encoded by a single gene. Post-translational modifications of this gene product include peptide cleavage and glycosylation, which increase the M_rof α2/δ up to 170,000 (Jay et al., 1991). As already observed for other ion channels biosynthetically labeled for a short period of time (Thornhill and Levinson, 1992), our result suggests that glycosylation did not have time to occur during the 4 hour period of our experiments. Interestingly, in this preparation, the size of the α2/δ molecule was not affected by the presence or absence of reducing agents, indicating that the α2/δ propeptide (Jay et al., 1991) had not yet been cleaved into its constituent α₂and d. Alternatively, the processed model of α2/δ inferred from purification studies in vivo might be an artifact of membrane isolation (Jay et al., 1991).

Starting from freshly prepared iodinated digitonin-solubilized T-tubules, at least two large polypeptides (M_r175,000 and M_r210,000) were co-precipitated by mAb 3007 along with α2/δ in the absence of Triton X-100. The M_r175,000 polypeptide has already been identified as the DHP receptor subunit, now referred to as α_1-175, which carries the high affinity DHP binding site (Vandaele et al., 1987; see Hosey and Lazdunski, 1988). The M_r210,000 polypeptide (further referred to as α_1-210) has also been identified as an independent α₁subunit differing from α_1-175by 174 to 188 additional amino acid residues at the C terminus of the molecule (De Jongh et al., 1989, 1991; Lai et al., 1990). The association of α2/δ and α_1-175has been abundantly documented (e.g. see Hosey and Lazdunski, 1988; Catterall et al., 1988). The fact that both α₁subunits were co-precipitated with α2/δ indicates that α_1-210is also associated to α2/δ.

In the present study, the respective amounts of α_1-175and α_1-210polypeptide immunoprecipitated by mAb 3007 appeared to be roughly equivalent. Since both molecules have the same number of tyrosine residues (Tanabe et al., 1987; De Jongh et al., 1991), they were likely to be iodinated in similar proportion. The equivalent amount of both molecules was at variance with earlier reports of α_1-175/α_1-210ratios close to 10/1 to 20/1 in rabbit adult muscle (De Jongh et al., 1989, 1991). However, it is clear from other studies that the α_1-210/α_1-175ratio may vary from one T-tubule preparation to another (Rotman et al., 1992).

The present experiments did not reveal the β (M_r55,000) and the g (M_r30,000) subunits of the DHP receptor molecular complex (Morton and Froehner, 1987; Leung et al., 1988). This was perhaps due to poor iodination of these subunits, which contain a smaller number of tyrosine residues than α₁and α2/δ (Tanabe et al., 1987; Ruth et al., 1989; Jay et al., 1990); it could also reflect the fact that purification procedures lead to variable recovery of these subunits (see Hosey and Lazdunski, 1988).

The evolution during myogenesis in vitro of α2/δ immunoreactivity consisting of patches and aggregates of small spots is reported here for the first time. The general aspect and the distribution of these accumulations underwent considerable changes with myotube maturation. Whether α2/δ exists independently of a molecular complex with the DHP receptor, or is always associated with other molecules, notably the DHP receptor α₁subunit, remains to be determined. The precocious appearance of α2/δ on myoblasts together with the fact that synthesis of α2/δ transcripts synthesis precedes muscle α₁mRNA expression (Varadi et al., 1989; Morton and Froehner, 1989) argues in favor of the first possibility, and thus for a role for α2/δ other than regulating the DHP receptor α₁subunit properties. However, as observed in the present study, the co-precipitation of α2/δ along with different α₁isoforms expressed in adult muscle (Brawley and Hosey, 1992; De Jongh et al., 1989, 1991; Lai et al., 1990; Rotman et al., 1992) suggest that it might also be associated with development-specific α₁isoforms during myogenesis (Malouf et al., 1992; Chaudhari and Beam, 1993). Further work using different α₁isoform-specific antibodies is needed to determine which of these might eventually be associated with the α2/δ subunit during myogenesis.

The lack of α2/δ immunoreactivity in proliferating myoblasts, and its appearance at a time when these cells acquire the capacity to fuse into myotubes, was consistent with previous studies indicating that α2/δ mRNA expression begins before fusion in vitro (Varadi et al., 1989). The absence of α2/δ immunoreactivity on fibroblasts did not exclude the possible existence of a low density of L-type channels that would include α2/δ, as detected with electrophysiological techniques on 3T3 fibroblasts (Chen et al., 1988).

Whether the formation of aggregates on young myotubes was due to a redistribution of already synthetized α2/δ in the membrane of the fused myoblasts, and/or to a new targeting of molecules synthetized de novo remains to be determined. As no diffuse staining on myotubes anteceded the formation of aggregates, these could represent the site of incorporation of vesicles containing newly synthesized molecules. Afterwards, the distribution of α2/δ continued to change: the irregular-sized spots constituting the aggregates on young, non-contracting myotubes were replaced by smaller, dispersed, regular-sized spots, which eventually aligned on contracting myotubes. Such alignments of α2/δ spots were reminiscent of the spot-like α2/δ immunoreactivity correlated to the T-tubule network as described by Flucher et al. (1991)on well-contracting mouse myotubes in vitro (three-week-old cultures).

Different models have been proposed to account for the formation of T-tubules (see Flucher, 1992). According to the ‘pull-in’ model, T-tubules derive from inpocketing of the plasma membrane (‘caveolae’) by a mechanism similar to endocytosis, without pinching off of coated vesicles (Ishikawa, 1968). According to the ‘add-on’ model, T-tubules form by addition of new membrane by a mechanism similar to exocytosis, involving coated vesicles directly derived from the Golgi complex (Schiaffino et al., 1977). More recent data have suggested a combination of these two mechanisms (Chan et al., 1990). The expression and the redistribution of α2/δ spots on the plasma membrane, followed by their eventual alignment on contracting myotubes, therefore suggest that α2/δ participates in the early morphogenesis of the T-tubules according to a pull-in mechanism.

The detection of patchy N-CAM immunoreactivity on myoblasts was consistent with previous reports (Covault et al., 1986; Moore et al., 1987; Tassin et al., 1991; Bloch, 1992). A localization of N-CAM, independent of α2/δ, was also transiently observed at the junction between fusing cells, whether myoblasts or mononucleated cells and myotubes. Such N-CAM staining was never seen at early stages of the cultures that are characterized by numerous myoblast divisions, and appeared most prominent at the peak of cell fusion. The fact that no α2/δ was found in this location while both markers were still present on the soma of fusing cells suggests that α2/δ does not play any direct role in the fusion process and/or that N-CAM molecules found at this particular location are synthetized de novo. Indeed, an increase in N-CAM synthesis and the expression of different N-CAM isoforms have already been reported to coincide with fusion (Covault et al., 1986; Moore et al., 1987; Tassin et al., 1991). The distribution of N-CAM on myotubes was in general agreement with the description of Tassin et al. (1991), notably the uniform staining and the presence of small spots scattered on the membrane at T₁₂. The diffuse intracellular N-CAM immunoreactivity observed at later stages was reminiscent of that observed in denervated adult muscle in vivo (Rieger et al., 1985; Sanes et al., 1986). At all stages of differentiation examined, a small percentage of N-CAM aggregates were not co-localized with α2/δ (between 2 and 3%). This result could hardly be explained on a purely technical basis, as the intensity of labeling was similar for co-localized and single N-CAM aggregates. These single N-CAM aggregates might correspond to those that remain visible on fully mature myotubes (Tassin et al., 1991), since N-CAM aggregates co-localizing α2/δ seem to be disrupted as the alignments of α2/δ appear.

The presence of uniform and clustered forms of N-CAM immunoreactivity on the membrane might reflect the heterogeneity of this molecule. Different N-CAM isoforms are produced by alternative splicing (see Cunningham et al., 1987), and are attached to the cell membrane either via a hydrophobic transmembrane peptide domain of two possible lengths, or via a glycosylphosphatidyl-inositol lipid-like anchor (He et al., 1986). Previous studies have suggested that patchy or spot-like immunoreactivity of N-CAM on myoblasts or myotubes corresponds to transmembrane domains bearing N-CAM isoforms, while more uniform N-CAM imunoreactivity might be associated with the glycosylphosphatidyl-inositol-anchored form (Tassin et al., 1991).

On myoblasts, the co-localization of intracellular α2/δ and N-CAM immunoreactivities suggests that they are transported through a similar pathway. The co-localization observed on the membrane would therefore result from a co-targeting rather than a reorganization of the two molecules in the membrane.

On myotubes, the predominant N-CAM immunoreactivity co-localized with α2/δ was essentially organized into spots and aggregates, as if α2/δ was solely associated with the membrane-anchored forms of N-CAM and not with the uniformly distributed glycosylphosphatidyl-inositol-anchored form. Antibodies specific to the various N-CAM isoforms, when they exist, will have to be used to answer this question correctly. Interestingly, it has been suggested that the spot-associated N-CAM forms interact with microtubules (Tassin et al., 1991), opening up the possibility of direct or indirect interactions between α2/δ and the microtubule network, at least during early myogenesis. The subsequent dispersion of α2/δ and N-CAM aggregates might therefore be consistent with the fact that lateral movement of certain molecules in the membrane requires microtubules (Connolly, 1984).

Irrespective of the state of myotube development examined, the overall aspect of α2/δ–N-CAM aggregates differed from that of acetylcholine receptors (AChR) clusters (Stya and Axelrod, 1983). This disparity, together with the fact that AchR does not co-localize either with N-CAM (Tassin et al., 1991), or with α2/δ (F. Rieger and S. Vandaele, unpublished results) argues against a non-specific trapping of N-CAM and α2/δ in the membrane.

It was of interest that from myoblast fusion to the acquisition of contractile properties by myotubes, α2/δ and N-CAM spots were mostly co-localized, until their alignment on myotubes, after which this type of N-CAM immunoreactivity seemed to disappear. Small amounts of N-CAM immunoreactivity have been detected by immunoelectron microscopy in T-tubules of adult rat muscle in vivo (Covault and Sanes, 1986), but such low amounts might be under the threshold of detection with our techniques.

The extensive co-localization of α2/δ and transmembrane N-CAM during early myogenesis is particularly intriguing, because in adult muscle most N-CAM immunoreactivity is found at the neuromuscular junction (Rieger et al., 1985), while α2/δ is located at the level of the triads in the T-tubule network (Flucher et al., 1991; F. Rieger, M. Pinçon-Raymond and S. Vandaele, unpublished results). The spatial proximity of N-CAM and α2/δ suggests a direct or indirect interaction between these two molecules, restricted to muscle development. The correlation of the changes of their distribution with myotube maturation suggests that such an interaction could regulate morphogenetic events such as the formation of T-tubules, the transient establishment of peripheral couplings between the sarcoplasmic reticulum and the sarcolemma in vivo (Kelly, 1971), and the elaboration and/or stabilization of the neuromuscular junction (Rieger, 1985). Indeed, a voltagedependent DHP-sensitive Ca²⁺channel appears to be involved in the development and maintenance of end-plates (Rotzier et al., 1992). Mechanisms by which N-CAM might be associated with α2/δ or with Ca²⁺channels in developing muscle could be akin to the second messenger system by which homotypic interactions between N-CAM molecules promote NGF-independent neurite elongation (Doherty et al., 1990, 1991).

We are particularly grateful to Dr L. Descarries for his encouragement and his careful review of the manuscript, and to Pr M. Lazdunski for his support during preliminary experiments. The anti-N-CAM antibodies were a generous gift from Dr G. Edelman. We also thank Gaston Lambert and Denis Dicaire for excellent artwork. This research was supported by grants to F.R. and M.L. from the Association Française des Myopathes and by institutional grants from CNRS and INSERM.