The Endocytic Recycling Protein EHD2 Interacts with Myoferlin to Regulate Myoblast Fusion

Abstract

Skeletal muscle is a multinucleated syncytium that develops and is maintained by the fusion of myoblasts to the syncytium. Myoblast fusion involves the regulated coalescence of two apposed membranes. Myoferlin is a membrane-anchored, multiple C2 domain-containing protein that is highly expressed in fusing myoblasts and required for efficient myoblast fusion to myotubes. We found that myoferlin binds directly to the eps15 homology domain protein, EHD2. Members of the EHD family have been previously implicated in endocytosis as well as endocytic recycling, a process where membrane proteins internalized by endocytosis are returned to the plasma membrane. EHD2 binds directly to the second C2 domain of myoferlin, and EHD2 is reduced in myoferlin null myoblasts. In contrast to normal myoblasts, myoferlin null myoblasts accumulate labeled transferrin and have delayed recycling. Introduction of dominant negative EHD2 into myoblasts leads to the sequestration of myoferlin and inhibition of myoblast fusion. The interaction of myoferlin with EHD2 identifies molecular overlap between the endocytic recycling pathway and the machinery that regulates myoblast membrane fusion.

Muscle development and regeneration rely on the fusion of singly nucleated myoblasts to initiate the formation or augment the growth of multinucleated myotubes. Myoblast fusion involves multiple steps including cell recognition, adhesion, and membrane coalescence. Genetic analyses of Drosophila mutants that fail to form normal muscles during development have been instrumental in elucidating not only the critical steps of myoblast fusion but also proteins that mediate recognition, adhesion, and the cytoskeletal elements that mediate fusion (1). In mammalian cells additional proteins that regulate myoblast fusion have been characterized, as well as many proteins implicated in myoblast differentiation (2–4). In both vertebrate and invertebrate systems, ultrastructural analysis of fusing myoblasts reveal the presence of vesicles accumulating at the sites of membrane fusion (5, 6). Because vesicle fusion is a common feature of most cells, it is possible that aspects of the myoblast fusion machinery utilize pathways more commonly used by other cell types for intracellular trafficking. Based on their involvement in vesicular fusion, the ferlin proteins are candidates for mediating the membrane coalescence stage of myoblast fusion.

In Caenorhabditis elegans, mutations in the gene encoding the prototypical ferlin protein, fer-1, produce defective membrane fusions in developing spermatozoa that lead to fertility defects (7, 8). Myoferlin is a member of the ferlin family and shares homology with fer-1; both proteins each contain multiple C2 domains in the long cytoplasmic portion followed by a carboxyl-terminal transmembrane domain. C2 domains consist of ∼130 amino acids that adopt conformation independent of neighboring sequences. Crystallographic studies of C2 domains support a structure with multiple aligned β strands and a calcium binding motif formed at one end from the intervening loops (9–11). There are ∼100 different C2 domain-containing proteins, and many of these show membrane association. Most C2 domain-containing proteins have one or two domains where one may participate in binding phospholipids, whereas the second may participate in protein-protein interactions. The C2 domains in myoferlin are highly related to those in the synaptotagmins; synaptotagmins participate in the rapid exocytosis that occurs at nerve terminals. Synaptotagmins have two C2 domains, one of which binds phospholipids in response to calcium (12). Models for synaptotagmin function suggest that its C2 domains can insert directly within a lipid bilayer and participate in the fusion of two independent membranes (13).

Myoferlin and members of the ferlin family are unique because of their multi-C2 domain nature. The myoferlin amino-terminal C2 domain, C2A, directly binds negatively charged phospholipids in response to calcium (14, 15). Myoferlin is abundantly expressed in myoblasts that are preparing to undergo fusion. In the elongated “prefusion” myoblast, myoferlin is found within vesicular structures in the cytoplasm and concentrated near the plasma membrane. In fusing myoblasts, myoferlin is concentrated at the sites of fusion (14, 15). Myoferlin null myoblasts fuse poorly in culture leading to smaller myotubes, and in vivo, the loss of myoferlin leads to reduced muscle mass with the loss of large diameter myofibers (15).

The Eps15 homology domain (EH domain),3 is a conserved domain important for protein interactions during vesicular trafficking (16, 17). The mammalian genome encodes four carboxyl-terminal EH domain-containing proteins, EHD1–4 (18–20). These carboxyl-terminal EHD proteins are characterized by an amino-terminal nucleotide binding domain followed by a central coiled-coil domain (21). Functionally, these carboxyl-terminal EHD proteins have been linked to endosomal trafficking including the recycling of cell surface receptors back to the plasma membrane (19, 22, 23). C. elegans and Drosophila each contain a single EHD gene. In the worm, mutations in the presumed nucleotide binding domain of EHD, also known as RME-1, lead to delay of recycling membrane components after endocytosis (24). This role is conserved in mammals (25). Mice engineered to lack EHD1 similarly display reduced endocytic recycling (22). EHD-binding proteins have been described that regulate interactions with the cytoskeleton, the small GTPases, and several cell surface receptors that may be carried as cargo as they are returned to the plasma membrane (26–29).

To understand better the components that direct myoblast fusion, we evaluated myoferlin interacting proteins in myoblasts undergoing fusion. Using immunoprecipitation and mass spectrometry, we identified EHD2 as a myoferlin-associated protein. Myoferlin contains the amino acid sequence asparagine-proline-phenylalanine (NPF), a known EHD binding motif, within its second C2 domain, and it is this region of myoferlin that binds directly to EHD2. EHD2 protein expression was significantly reduced in myoferlin null myoblasts. Moreover, myoferlin null myoblasts display delayed endocytic recycling accompanied by intracellular aggregation of labeled transferrin. Expression of mutant EHD2 protein in myoblasts leads to sequestration of myoferlin and an inhibition of myoblast fusion. This work demonstrates that impaired endocytic recycling is associated with defective myoblast fusion.

MATERIALS AND METHODS

Culture of C2C12 Cells and Primary Myoblasts—C2C12 cells were obtained from ATCC (catalog #CRL-1772). Cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and penicillin/streptomycin in 7% CO₂. Primary myoblasts were isolated from neonatal wild type and myoferlin null pups and grown and maintained as described (15). All tissue culture media and sera used were from Invitrogen.

Immunoprecipitation—C2C12 cells or primary myoblasts were grown to confluency (5 days) in growth media. Cells were lysed in coimmunoprecipitation buffer (150 mm NaCl, 50 mm Tris-HCl, pH 7.4, 0.15% CHAPS with Complete Mini Protease Inhibitor mixture, Roche Applied Science). Cellular debris was removed, and the protein concentration of the supernatant was determined using Bio-Rad protein assay. Two hundred μl of protein G-Sepharose beads (Amersham Biosciences) was blocked for 1 h with 20 mg/ml bovine serum albumin in coimmunoprecipitation buffer, preincubated for 30 min with anti-myoferlin antibody, MYOF3 (15), then incubated overnight with 200 μg of protein. Beads were washed 4 times in coimmunoprecipitation buffer and boiled 10 min in loading buffer (50 mm Tris, pH 6.8, 100 mm dithiothreitol, 2% SDS, 0.1% bromphenol blue, 10% glycerol), and the supernatant was loaded on a 4–20% acrylamide gel. Gels were either silver-stained (Bio-Rad Silver Stain Plus) or stained with Coomassie or transferred to polyvinylidene difluoride Immobilon-P membrane (Millipore).

Identification of Interacting Proteins—The band of interest was excised from a Coomassie-stained acrylamide gel. Five bands of the same molecular weight were combined and subjected to MALDI-TOF mass spectrometry at the Proteomics Core Facility at University of Chicago. Using the Mascot search engine, full-length murine proteins corresponding to the peptides generated by the mass spectrometry were identified in the non-repetitive NCBI data base, and statistical significance of the matches was determined.

Immunoblotting—Proteins were transferred to polyvinylidene difluoride membrane, stained with Ponceau as a loading control, and immunoblotted with polyclonal anti-myoferlin (MYOF3, 1:2000) (15), polyclonal goat anti-EHD2 (1:10–25,000, Abcam), or anti-Xpress (1:5000, Invitrogen). Secondary antibodies, goat anti-rabbit, goat anti-mouse, and donkey anti-goat conjugated to horseradish peroxidase (Jackson ImmunoResearch) were used at a dilution of 1:5000. Blocking and antibody incubations were done in 5% milk in 1× Tris-buffered saline with 0.1% Tween 20 for all antibodies except anti-EHD2 and donkey anti-goat conjugated to horseradish peroxidase. These were incubated in 1× Tris-buffered saline containing 4% donkey serum and 0.1% Tween 20. ECL-Plus chemiluminescence (Amersham Biosciences) and Kodak Biomax MS film or a Amersham Biosciences PhosphorImager was used for detection.

In Vitro Binding Experiments—Human myoferlin C2 domains were ligated into pGEX4T-1 as previously described (14). Mutant versions of myoferlin C2B were generated by site-directed mutagenesis using Pfu polymerase and the following sets of primers: AAG AGA GGA AAC AGC CCT TTG TTT GAT GAG TTG TT and AAC AAC TCA AAC AAA GGG CTG TTT CCT CTC TCT T (NPF to dysferlin C2B sequence SPL) and AAG AGA GGA AAC TGC CCT TTT TTT GAT GAG TTG TT and AAC AAC TCA AAA AAA GGG CTG TTT CCT CTC TCT T (NPF to fer1L4 C2B sequence CPF) as described (14). Cultures of BL21 cells (Invitrogen) expressing glutathione S-transferase-myoferlin C2 domain fusion proteins were induced for 1 h with isopropyl 1-thio-β-d-galactopyranoside at a final concentration of 85 μg/μl. Cells were resuspended in 10 mm Tris, pH 7.5, and added to SDS loading buffer. After boiling, 10 μl of lysed cells was loaded on 10% acrylamide gel. For each experiment two gels were used; one was transferred to polyvinylidene difluoride membrane used for gel overlay, and the other was stained with GelCode (Bio-Rad).

A clone corresponding to the full-length murine cDNA of EHD2 (NCBI accession number AK046566) was obtained from RIKEN (clone ID B430101D12). The cDNA was amplified with the primers EHD2 (forward, ATG TTC AGC TGG CTG AAG AGG GG; reverse, TCA TTC AGC AGA GCC CTT CTG TC) and cloned into pCR T7/NT-TOPO (Invitrogen). ³⁵S-Labeled in vitro transcribed/translated EHD2 was generated using [³⁵S]methionine with the Promega TnT Coupled Reticulocyte Lysate system. The product was separated on a G-25 Sephadex column (Roche Applied Science). The membrane was blocked for 2 h at 4 °C in blocking buffer (0.1% gelatin, 5% bovine serum albumin, 0.1% Tween in 1× PBS, pH .5) then incubated overnight at 4 °C in overlay buffer (150 mm NaCl, 20 mm HEPES, 2 mm MgCl₂, 5% bovine serum albumin, pH 7.5) containing the radioactive EHD2 as described (30). The membrane was washed 4 × 20 min at room temperature in overlay buffer and exposed to film and/or imaged on a PhosphorImager.

Immunostaining and Microscopy—Primary myoblasts were grown on gelatin-coated glass coverslips, fixed in 4% paraformaldehyde for 10 min on ice, rinsed with PBS, then fixed 2 min on methanol on ice. C2C12 cells were fixed in 4% paraformaldehyde 10 min on ice followed by 10 min in 0.3% Triton-X100 in 1× PBS on ice. Polyclonal MYOF2 serum (31) was used at 1:200, goat polyclonal anti-EHD2 (Abcam, ab23935) was used at 1:500, and monoclonal anti-Xpress (Invitrogen) was used at 1:500. Donkey anti-rabbit conjugated to Alexa488, donkey anti-goat conjugated to Alexa 594, goat anti-rabbit conjugated to Alexa 594, and goat anti-mouse conjugated to Alexa488 were obtained from Molecular Probes and used at 1:2000. Blocking and antibody incubations were done in 1× PBS, 4% donkey serum, 0.1% Triton X-100 with donkey anti-goat and donkey anti-rabbit secondary antibodies. With goat anti-mouse and goat anti-rabbit secondary antibody incubations, blocking and antibody incubations were done in 1× PBS, 5% fetal bovine serum, 0.1% Triton X-100. Coverslips were mounted using Pro-Long Gold with 4′,6-diamidino-2-phenylindole. Images were captured using either a Zeiss Axiophot microscope and Axiovision software (Carl Zeiss) or a Leica SP2 scanning laser confocal microscope and LCS Leica Confocal Software.

Transferrin Internalization Assay—Myoblasts were grown on sodium hydroxide-treated glass coverslips. Cells were incubated in Dulbecco's modified Eagle's medium, 0.1% bovine serum albumin, 20 mm HEPES, pH 7.2. Cells were incubated with Alexa488-conjugated transferrin (25 μg/ml, Molecular Probes) for 60 min followed by chase with 50 mm deferoxamine, 250 μg/ml holotransferrin in 20% fetal bovine serum, 20 mm HEPES, pH 7.2. Cells were fixed in 4% paraformaldehyde and mounted for visualization.

Myoblast cultures analyzed by flow cytometry were prepared as above but were trypsinized then fixed in 4% paraformaldehyde. Cells were washed and resuspended in PBS for analysis in a FACS Canto (BD Biosciences). After gating to remove dead cells, aggregated cells, and debris, a minimum of 20,000 events was scored per time point per culture. Cultures were derived from three independent animals per genotype. Fluorescent gates were set to include 97% of wild type cells at time 0 after transferrin removal. The percentage of fluorescent cells was determined on subsequent time points by the number of cells remaining in this gate. Median fluorescence values were determined using a histogram of Alexa488 fluorescence. FACS data were analyzed using FlowJo software (Tree Star, Inc.).

Expression of Wild Type and G65R EHD2—The full-length EHD2 cDNA was amplified with the primers EHD2 (forward, GGT ACC CAT GTT CAG CTG GCT GAA GAA GGG C; reverse, GCG GCC GCT CAT TCA GCA GAG CCT TCT GTC GT) and ligated into the KpnI and NotI sites of pcDNA3.1 His B (Invitrogen), which contains an amino-terminal Xpress epitope tag and a cytomegalovirus promoter. The G65R mutation was generated by site-directed mutagenesis using the primers EHD2 G65R (forward, GTG CTG GTG GCC CGC CAG TAT AGC ACC GGC; reverse, G65R GCC GGT GCT ATA CTG GCG GGC CAC CAG CAC) and Pfu polymerase as described (14). C2C12 cells were plated at 50,000 cells per well on glass coverslips in 6-well plates. The following day each well was transfected with 5 mg of DNA using Lipofectamine Plus and Opti-MEM (Invitrogen). Twenty four hours after transfection, cells were differentiated or fixed, stained, and imaged as described above.

RESULTS

Identification of EHD2 as a Myoferlin Interacting Protein—Because myoferlin is abundantly expressed in prefusion myoblasts, C2C12 cells were grown to confluence in 10% fetal bovine serum to enrich for myoblasts in the early phases of fusion. As shown in Fig. 1, the polyclonal anti-myoferlin antibody, MYOF3, immunoprecipitated myoferlin as well as a protein of ∼65 kDa (arrow in Fig. 1B). After determining that this band appeared reproducibly, the results of repeated experiments were combined and analyzed by MALDI-TOF mass spectrometry. The Mascot search engine (32) was used to identify full-length proteins in the NCBI data base corresponding to the peptides generated by mass spectrometry analysis and to determine which matches were statistically significant. A probability-based Mowse score was determined, and a significance score was set at p < 0.5 for this analysis. From this approach, fourteen full-length proteins contained a sufficiently high match to peptides identified by mass spectrometry. Of these 14, 13 were likely contaminants, as they encoded keratin-related intermediate filament proteins (8 of 14), α globin, the β-2 chain of hemoglobin, complement C3, and two RNA-binding proteins. Two peptides in this analysis corresponded to EHD2 (Eps15 homology domain 2), and the position of these sequences along the EHD2 schematic is indicated in Fig. 1A. We selected this protein for further analysis because EHD2 is a protein of the expected size that localizes in the cytoplasm and the membrane similar to myoferlin.

FIGURE 1.

Co-immunoprecipitation and mass spectrometry identify EHD2 as a myoferlin interacting protein. A, schematic representation of myoferlin and EHD2. Myoferlin is a large protein (230 kDa, 2060 amino acids) containing a carboxyl-terminal transmembrane domain, and six C2 domains, referred to as C2A–F. EHD2 contains a carboxyl-terminal EH domain as well as an amino-terminal ATPase sequence. The amino acids listed above the EHD2 schematic are those identified by mass spectrometry of the 65-kDa band shown in (B). Amino acid residue numbers are indicated along the top. B, an anti-myoferlin (αmyof) antibody (ab) was used to immunoprecipitate (IP) myoblasts lysates and consistently showed a 65-kDa protein (arrow). C, an immunoblot of myoblast cell lysates using anti-EHD2 antibody confirmed that EHD2 migrates slightly higher than its predicted mass of 61 kDa, equal to the 65-kDa band of interest. D, immunoprecipitation with the anti-myoferlin antibody from myoblasts demonstrated that EHD2 associates with myoferlin.

Eps15 homology (EH) domains are found in more than 50 eukaryotic proteins, most of which are involved in regulating membrane traffic and events such as receptor internalization, vesicle transport, and actin polymerization (33). EHD2 is a member of a subclass of EH domain-containing proteins differentiated by the presence of a carboxyl-terminal EH domain. EHD2 is predicted to be 61 kDa, very close to the observed size of the protein co-immunoprecipitated by the anti-myoferlin antibody. We immunoblotted cell extracts from the myoblast C2C12 line demonstrating that EHD2 consistently migrated slower than the 62-kDa marker at ∼65 kDa (arrow in Fig. 1C). In addition, immunoblotting with an anti-EHD2 antibody confirmed that EHD2 was immunoprecipitated along with myoferlin in the presence of anti-myoferlin antibodies (Fig. 1D). The antibody to EHD2 was unable to immunoprecipitate EHD2 and, thus, could not be used to determine whether EHD2 antibodies could immunoprecipitate myoferlin (data not shown).

EHD2 and Myoferlin Interact Directly—The Eps15 EH domain is composed of ∼100 amino acids, and EH domains are 50–60% conserved among proteins (34). NMR spectroscopy has shown that EH domains contain two EF-hands linked by an anti-parallel sheet (18). The EH domain recognizes and binds to the motif asparagine-proline-phenylalanine (NPF) (35). This sequence was found in both human and murine myoferlin at amino acids 238–240. When engaged, the NPF motif is predicted to project into a conserved hydrophobic pocket in the EH domain where a conserved tryptophan interacts with the asparagine (36, 37). The C2B domain of myoferlin shows significant homology to the C2A domain of synaptotagmin I (Fig. 2A). Using the known structure of synaptotagmin, we modeled the myoferlin C2B domain. With this configuration, the NPF motif is located at the beginning of the fourth β strand, protruding from the surface of the domain and accessible for protein-protein interactions (Fig. 2B).

FIGURE 2.

The NPF motif in myoferlin C2B is required for interaction with EHD2. A, the C2B domain of myoferlin was aligned with the C2A domain of synaptotagmin I (PDB code 1RSY, amino acids 43–153). The NPF motif and the corresponding amino acids in synaptotagmin (blue) are indicated by the arrow. Conserved residues are highlighted in gray, and calcium-coordinating residues are in red. B, the crystal structure of synaptotagmin I C2A is shown as a ribbon backbone. The side chains of NPF are shown (blue) as are those of the calcium coordinating aspartic acid residues (red). The NPF sequence is found in the fourth β strand, projecting out from the surface of the C2 domain where it is accessible for protein-protein interactions. C, a blot overlay was used to demonstrate a direct interaction between EHD2 and myoferlin. All six myoferlin C2 domains fused to glutathione S-transferase are expressed at equal levels (Coomassie-stained gel, top). An equivalent gel was transferred to polyvinylidene difluoride membrane and incubated with ³⁵S-labeled EHD2 (bottom) showing that only myoferlin C2B bound EHD2. The remaining C2 domains lack an NPF domain and also lack binding, indicating specificity. D, mutation of the NPF motif decreases the binding of EHD2 to myoferlin C2B by 4-fold. Gel overlay experiments were done with glutathione S-transferase-myoferlin C2B domains containing NPF motifs substituted with the sequence from dysferlin C2B (SPL-Mut1) or another member of the ferlin family, fer1l4 (CPF-Mut2). In these experiments the center proline residue was left intact to avoid substantially misfolding the C2 domain. WT, wild type.

Because C2 domains have been previously shown to fold independent of their surrounding sequences (38), we assessed each C2 domain of myoferlin individually. We demonstrated a specific and direct interaction between the NPF amino acids in myoferlin C2B and EHD2 (Fig. 2C). EHD2 bound directly to myoferlin C2B but did not bind detectably to the other C2 domains in myoferlin including C2A, C2C, C2D, C2E, or C2F. Consistent with this lack of binding, none of these domains possesses the NPF motif. Two different mutants of the NPF motif were generated to document the specificity of the EHD2-myoferlin interaction. In both mutants we left intact the central proline residue that forms the core of the NPF motif to avoid disrupting the entire C2 domain. The first mutant tested (SPL) represents the residues found in the related protein dysferlin, and the second mutant (CPF) represents the sequences found in fer1l4, another ferlin sequence. This experiment was performed in triplicate and quantified to show that EHD2 binding to both mutants was reduced by 4-fold (Fig. 2D). This binding was not affected by the presence of calcium, and this is consistent with the lack of calcium binding of C2B (data not shown).

EHD2 Is Reduced in Myoferlin Null Myoblasts—We found that EHD2 was expressed in myoblasts isolated from muscle. As shown in Fig. 3, EHD2 was expressed in undifferentiated myoblasts enriched near the plasma membrane in punctate structures. When myoblasts were co-stained with the anti-myoferlin antibody MYOF2 and anti-EHD2, overlap in the merged images was seen (Fig. 3A). Confocal imaging (Fig. 3A) showed myoferlin and EHD2 immunostaining at sites enriched along the plasma membrane, consistent with colocalization and a potential role in vesicle trafficking. Myoferlin null mice were previously characterized and have a phenotype of reduced muscle mass and smaller diameter myofibers (15). When cultured, myoferlin null myoblasts do not fuse efficiently and generate smaller myotubes (15). Myoferlin null myoblasts showed a marked reduction in EHD2 compared with wild type control myoblasts (Fig. 3B) and was confirmed by immunoblotting undifferentiated primary myoblasts from control and myoferlin null mice (Fig. 3C).

FIGURE 3.

EHD2 is reduced in myoferlin null myoblasts. A, in myoblasts cultured from neonatal mice, EHD2 was found throughout the cytoplasm with increased concentrations at the membrane (red staining). Myoferlin showed a similar pattern using the MYOF2 antibody (green). Colocalization was visualized as yellow in the merged image; 4′,6-diamidino-2-phenylindole (blue) identifies nuclei. Higher magnification confocal images (boxed areas) are shown on the right. Myoferlin and EHD2 colocalized at sites along the membrane. B, wild type (WT) and myoferlin null (MKO) primary myoblasts stained with MYOF2 anti-myoferlin (green), anti EHD2 (red), and 4′,6-diamidino-2-phenylindole (blue). Images were acquired identically for control and myoferlin null myoblasts and show that EHD2 is reduced in myoferlin null myoblasts. C, immunoblotting of wild type and myoferlin null myoblasts showed reduction of EHD2 in myoferlin null myoblasts. Actin is shown as a loading control.

Delayed Endocytic Recycling in Myoferlin Null Myoblasts—Transferrin is a small molecule that is rapidly internalized into cells via the transferrin receptor. Once internalized, transferrin is recycled to the plasma membrane where it can be released to bind iron. Fluorescently labeled transferrin is used commonly to image and measure endocytic recycling. Myoferlin null and wild type control myoblasts were isolated and incubated with fluorophore-labeled transferrin. Myoferlin null myoblasts, compared with wild type, displayed internalized transferrin that was aggregated in perinuclear structures (Fig. 4). We quantified the percentage of cells that displayed intracellular aggregates after incubation with labeled transferrin and found that 80% of myoferlin null myoblasts had this appearance, whereas less than 30% of wild type control cells did. These data were gathered from 100 myoblasts per culture and from myoblast cultures isolated from independent animals (n = 3, myoferlin null, and n = 2 wild type control, for a total of 300 and 200 myoblasts, respectively).

FIGURE 4.

Intracellular aggregation of fluorescently labeled transferrin in myoferlin null myoblasts. Wild type (WT) and myoferlin null (MKO) myoblasts were labeled for 1 h with Alexa488-transferrin and imaged. Myoferlin null cells showed marked accumulation of fluorescent transferrin in perinuclear aggregates, whereas fewer than 30% of wild type cells showed similar aggregation. A typical example of aggregates of transferrin is shown for MKO myoblasts. The accumulation of transferrin in myoferlin null cells reflects delayed recycling to the plasma membrane.

We next evaluated the rate of endocytic recycling using labeled transferrin in pulse-chase experiments on myoferlin null and wild type myoblasts. In wild type myoblasts, internalized, labeled transferrin was recycled between 20 and 30 min of chase with unlabeled transferrin (Fig. 5A, top row). In contrast, in myoferlin null cells fluorophore remained visible even after 40 and 50 min of incubation of unlabeled transferrin (Fig. 5A, bottom row). To quantify these findings, similarly cultured and treated myoblasts were subjected to FACS analysis. At base line, the median fluorescence intensity was five times greater in myoferlin null myoblasts than wild type, consistent with an accumulation of labeled transferrin (Fig. 5B). FACS sorting of cells at each time point showed that labeled transferrin was undetectable in 50% of wild type myoblasts after 20 min, whereas myoferlin null cells required 60 min to reduce labeled transferrin by 50% (Fig. 5C). These data demonstrate delayed endocytic recycling of transferrin in myoferlin null myoblasts and are remarkably similar to what has been noted in C. elegans RME-1 mutants and mice lacking EHD1 (22, 24).

FIGURE 5.

Myoferlin null cells display delayed recycling of fluorescently labeled transferrin. A, wild type and myoferlin null myoblasts were cultured and then incubated with Alexa488-labeled transferrin for 60 min followed by chase with unlabeled transferrin for 0, 10, 20, 30, 40, or 50 min (indicated as 0, 10, 20, 30, 40, 50 on A). Most internalized labeled transferrin was absent by 20–30 min in wild type cells. In contrast, transferrin fluorescence was visible until 40 or 50 min in myoferlin null cells. B, after incubation with Alexa488-labeled transferrin, the median fluorescence intensity of myoferlin null cells was five times greater than that for wild type (WT). C, myoblasts were collected at time points after incubation with unlabeled transferrin and subjected to FACS. Myoferlin null myoblasts cleared labeled transferrin less effectively than wild type. Fifty percent of labeled transferrin was cleared after 20 min in wild type myoblasts compared with 60 min for myoferlin null myoblasts.

Dominant Negative EHD2 Impairs Myoblast Fusion—Similar to the other three EHD proteins, the amino terminus of EHD2 contains a predicted P-loop nucleotide triphosphatase (Fig. 1A). It has been shown that EHD1 binds and hydrolyzes ATP but not GTP (21, 23). A series of loss of function and dominant negative mutations in the orthologous gene RME-1 were characterized based on their role in endocytic recycling in C. elegans. The RME-1 point mutation, G81R, is predicted to disrupt nucleotide binding since it falls within the nucleotide binding domain. This mutant exhibited dominant negative behavior in C. elegans resulting in altered subcellular localization of RME-1 and impairment of endocytosis (24, 25). The orthologous mutation, G65R, has been studied in mammalian EHD1 where it similarly was noted to prevent the proper formation of the endocytic recycling complex (39). To determine EHD2 function in myoblast fusion, epitope-tagged wild type EHD2 or dominant negative G65R-EHD2 was introduced into C2C12 myoblasts. We found that G65R-EHD2, unlike wild type EHD2, impaired myoblast fusion and sequestered myoferlin into intracellular aggregates (Fig. 6). In these experiments, wild type EHD2 and G65R-EHD2 were expressed in the same percent of cells (23.2 and 22.2%, respectively) and at comparable levels (Fig. 6A). After 6 days in differentiation media, 23.5% of cells expressing wild type EHD2 were multinucleate myotubes. In contrast, only 6.7% of the cells expressing G65R-EHD2 were multinucleate myotubes (n = 3 experiments), indicating a block in myoblast fusion. High magnification images (Fig. 6C, left) of transfected myoblasts at 24 h demonstrated that dominant negative G65R-EHD2 formed aggregates. These aggregates not only contained EHD2 but also myoferlin, indicating that G65R-EHD2 was capable of sequestering myoferlin (Fig. 6C, right). Transfection of wild type EHD2 did not disturb myoblast fusion. Consistent with normal myoblast fusion, wild type EHD2 and myoferlin were colocalized at sites on the membrane and throughout the cytosol in a pattern consistent with previous reports describing a role for EHD2 in coupling endocytosis to actin cytoskeletal dynamics (28).

FIGURE 6.

Expression of G65R-EHD2 inhibits myoblast fusion and sequesters myoferlin. C2C12 myoblasts were transfected with Xpress-tagged wild type or G65R-EHD2. Twenty-four hours later cultures were fixed and stained or immunoblotted (A and C) or differentiated into myotubes (B). A, wild type and G65R constructs transfected at the same efficiency (23 and 22.2%, respectively) and at comparable levels. Myosin is shown as a loading control. B, after 6 days in low serum differentiation media, very few G65R-EHD2-expressing cells were multinucleated myotubes (6.5%), whereas many wild type EHD2-expressing cells were multinucleated (23.5%). The lower panels contain higher magnification images of representative transfected cells. Scale bars are 20 μm. C, higher magnification images of cells 24 h after transfection indicate that myoferlin colocalized with wild type EHD2 at the membrane and distributed throughout the cytoplasm. However, mutant G65R-EHD2 mislocalized to intracellular aggregates (green) that sequestered myoferlin (red and merge in yellow) and reduced myoblast fusion.

DISCUSSION

Topologically, the process of myoblast fusion is not well understood but has been observed that transient vesicular structures appear at the sites of membrane fusion (6). Our data suggest that these vesicles may derive from the endocytic recycling compartment. These studies and others conducted in cultured mammalian myoblasts are hampered by an inability to coordinate all myoblasts undergoing fusion. Within a given culture, cells are at many different stages of myoblast fusion including dividing myoblasts, prefusion elongated cells, fusing cells forming small myotubes, and myotubes that are being augmented by additional myoblast fusion.

During development, Drosophila myogenesis occurs discretely, and so this model has been used to identify a series of ultrastructural features and mutations that alter these structures. Electron dense puncta are visible at tightly apposed membranes just before fusion. After this, vesicles align on both sides of the fusing membranes (5). Dissolution of apposed membranes is associated with adjacent electron-dense material and, eventually, a single lipid bilayer. Analysis of mutants that contribute to impaired myogenesis in Drosophila has led to the identification of a number of gene products that regulate these distinct stages of myoblast fusion. At least three different genes have been identified that contribute the intracellular, cytoskeletal aspects of membrane fusion. These include Myoblast City (mbc), a DOCK180/CED-5-like protein (40), Loner, an ARF6-binding protein (41), and solitary (sltr), a WASP interacting protein (42, 43). Importantly, in the sltr mutants, actin accumulation was perturbed at the sites of fusion. Intriguingly, vesicles are noted at the sites of cytoskeletal rearrangement. This observation along with the identification of these other genes implicates aspects of the cytoskeletal reorganization that must accompany the fusion of myoblasts to myoblasts. In mammalian muscle, myoblast to myotube fusion also occurs readily and requires even more significant cytoskeletal remodeling on the myotube aspect of fusion.

The EHD proteins have been implicated in endocytic recycling (24, 39, 44). We found that EHD2 was reduced markedly in myoferlin null myoblasts, and myoferlin null myoblasts had delayed recycling of transferrin in pulse-chase experiments. This delayed loss of labeled transferrin was also associated with abnormal cytoplasmic aggregates. We infer from this observation that the endocytosis of transferrin may occur normally, but that recycling is delayed, potentially due to failure to efficiently exit the endocytic recycling compartment. Abnormal transferrin recycling reflects one receptor, the transferrin receptor. However, we expect that other receptors required for normal cell fusion and myogenesis are likely adversely affected by the loss of both myoferlin and EHD2.

In myoblasts, the interaction between EHD2 and myoferlin may indirectly regulate disassembly or re-organization of the cytoskeleton that accompanies myoblast and myotube fusion. It has previously been observed that EHD2 and its binding partner EHD2-binding protein 1 (EHBP1), localize to adjacent structures in regions of endocytosis and actin dynamics such as ruffled membranes and only colocalize in regions where the adjacent structures overlap (28). It is possible that myoferlin and EHBP1 compete for binding on EHD2 (Fig. 7).

FIGURE 7.

Model for the role of myoferlin and EHD2 in vesicle cycling and myoblast fusion. Many cell surface, membrane-bound receptors undergo recycling after endocytosis. These receptors and their bound ligand are internalized, shuttled to the endocytic recycling compartment, and then are recycled back to the plasma membrane where they can function in another round of ligand binding. EHD2 and myoferlin are implicated in vesicle cycling. It is hypothesized that improper endocytic recycling of receptors involved in propagating myoblast fusion signals could cause myoblast fusion defects due to an interruption in the receptor signaling cascade.

The introduction of the G65R EHD2 point mutant sequestered myoferlin to internalized structures and inhibited myoferlin. This result implicates the nucleotide binding state of EHD2 in myoblast fusion. EHD proteins 1–4 contain a central coiled-coil region that mediates oligomerization (39, 45). Most recently, the structure of EHD2 was determined where the importance of the nucleotide hydrolysis state was established for its interaction with membrane components (23). Point mutations in the nucleotide binding domain were found to localize distinctly with membrane structures, suggesting that dissociation from internal membranes may be a key step regulated by nucleotide hydrolysis. The orthologous EHD protein in C. elegans RME-1 carrying a G81R mutation, which is equivalent to G65R, is not able to form homodimers with either wild type or mutant RME-1 (21). Assuming the analogous mutant in EHD2 is similarly impaired in oligomerization, our data suggest that myoferlin is sequestered by mutant EHD2 monomers leading to inhibition of myoblast fusion. Inhibition of myoblast fusion by mutant EHD2 was more marked than what is observed from loss of myoferlin alone. This suggests that EHD2 may sequester not only myoferlin but also potentially other proteins that are important for myoblast fusion, thus leading to a more severe block of myoblast fusion.

Both the EHD proteins and the ferlins are encoded by multigene families. The four known carboxyl-terminal EHD proteins are highly related to each other, so it is possible that other EHD proteins participate in myoblasts fusion or in the process of membrane repair in myofibers. Only recently have isoform-specific antibodies been reported for each of the EHD proteins, and these antibodies show that EHD1 and EHD2 are most highly expressed in skeletal muscle, whereas little to no EHD3 and EHD4 is present (46). However, it should be noted that these studies were conducted on skeletal muscle, and the expression pattern of EHD proteins in myoblasts at different stages of differentiation may differ from mature muscle. Similarly, there is significant diversity among the ferlin family proteins since there are six full-length ferlin proteins predicted by the mammalian electronic databases. These include dysferlin, otoferlin, myoferlin, fer1l4, fer15, and fer1l6. Dysferlin protein expression increases markedly in mature myotubes, whereas myoferlin expression increases in myoblasts undergoing fusion. The expression patterns of otoferlin, fer1l4, fer1l5, and fer1l6 have not been demonstrated in myoblasts and myotubes, although unpublished data suggest that fer1l4, fer1l5, and fer1l6 are each found in C2C12 myoblasts.4 The complexity of the EHD and ferlin family may correlate with unique stages of myoblast differentiation and fusion as myoblast fusion is a highly regulated process.

Our data suggest that vesicles seen in myoblast fusion may represent a specialized aspect of the endocytic recycling compartment. The interaction of myoferlin with EHD2 may facilitate membrane fusion at sites of contact between cells where cytoskeletal rearrangements are needed. Both myoferlin and EHD2 are widely expressed, so the role of this interaction may be important for endocytic recycling in other cell types and systems beyond myoblast fusion.