Skeletal muscle is continually subjected to microinjuries that must be repaired to maintain structure and function. Fluorescent dye influx after laser injury of muscle fibers is a commonly used assay to study membrane repair. This approach reveals that initial resealing only takes a few seconds. However, by this method the process of membrane repair can only be studied in part and is therefore poorly understood. We investigated membrane repair by visualizing endogenous and GFP-tagged repair proteins after laser wounding. We demonstrate that membrane repair and remodeling after injury is not a quick event but requires more than 20 min. The endogenous repair protein dysferlin becomes visible at the injury site after 20 seconds but accumulates further for at least 30 min. Annexin A1 and F-actin are also enriched at the wounding area. We identified a new participant in the membrane repair process, the ATPase EHD2. We show, that EHD2, but not EHD1 or mutant EHD2, accumulates at the site of injury in human myotubes and at a peculiar structure that develops during membrane remodeling, the repair dome. In conclusion, we established an approach to visualize membrane repair that allows a new understanding of the spatial and temporal events involved.
Skeletal muscle microinjuries are frequent since normal muscle activity permanently exposes the sarcolemma to mechanical stress . Repair is crucially important to maintain muscle structure and function. The most widely accepted model of sarcolemmal repair describes the process as a Ca2+-dependent membrane patch formation by the fusion of intracellular vesicles [2-6]. Another model is called the lipid flow hypothesis. Here, lateral flow of lipids over the injured area closes the wound by fusing with the complementary membrane edge [7, 8]. The earliest identified protein involved in membrane repair is dysferlin, a 230 kD membrane protein with seven intracellular C2 domains and a short extracellular domain. Mutated dysferlin causes muscular dystrophy in young adults [9, 10] and isolated dysferlin-deficient mice muscle fibers show an impaired ability to repair membranes . Binding partners of dysferlin, such as the annexins A1 and A2  and the tripartite-motif (TRIM) family member protein mitsugumin 53 (MG53) are part of the sarcolemmal resealing complex . After activation through Ca2+ influx, annexin A1 binds to membrane phospholipids and serves as a promoter of membrane aggregation and fusion, whereas the actin-associated protein annexin A2 is involved in the regulation of enlargosome fusion with the plasma membrane [14, 15]. MG53 facilitates intracellular vesicular transport as well as membrane fusion, and loss of MG53 leads to repair defects . Polymerase I and transcript release factor (PTRF) acts as a MG53 membrane anchor during membrane resealing  and regulates caveolae membrane structure . The involvement of the muscle-specific isoform caveolin 3 in sarcolemmal repair is unclear. Although interacting with dysferlin and MG53 [18, 19] caveolin 3 may play a more indirect role in the repair process [9, 20].
We established a laser-wounding assay on primary human myotubes to study the repair of microinjuries. We identified an Eps15 homology (EH)-domain-containing protein EHD2 that has hitherto not been appreciated as important to the repair process.
Results and Discussion
An advanced laser wounding approach highlights the repair complex formation and slow accumulation of dysferlin at the resealing site
Several methods have been described to investigate muscle membrane repair on cultured myotubes or isolated mouse muscle fibers. The strategies vary from chemically induced membrane disruption  to mechanical wounding by stretching or scraping using glass beads, microneedles or scratchers [12, 22]. Laser-induced disruption of single mice muscle fibers is a well-established method to investigate the ability of intact mice skeletal muscle fibers to conduct membrane repair [11, 13, 16, 18, 22]. Time-dependent influx of fluorescent dye into the muscle fiber is the most commonly applied readout of these assays. This technique documents the sequelae of wounding but does not allow investigation of the repair proteins. Therefore, we attempted to morphologically dissect membrane repair after laser-induced sarcolemmal injury. We established an assay that focuses on visualizing the healing process and used this assay with primary human myotubes. To allow real-time confocal observation of proteins predicted to be involved in membrane repair, we combined the laser wounding technique with microinjection of GFP-encoding cDNA constructs. In myotubes expressing GFP-dysferlin laser irradiation not only lead to membrane wounding but also to a GFP-signal bleaching at the injury site (Figure 1A, white arrow, upper row). Twenty seconds after wounding dysferlin could be redetected at the membrane and slowly accumulated thereafter with an apical enrichment at the membrane injury site (Figure 1A, green arrow), but not in the entire bleached area (Figure 1A, white arrows, lower row). This specific relocalization of dysferlin points toward a directed transport of dysferlin to the membrane wound underlining its role in sarcolemmal repair. The simultaneous detection of extracellular fluorescence dye FM 4-64 showed no influx into the myotube demonstrating intact membrane repair (Figure 1B).
In parallel to the enrichment of dysferlin a repair ‘dome’ developed at the injury site (Figure 1C). The repair dome increased in size over a period of 60 min. Dysferlin-deficient myotubes did not form this repair dome (Figure S1C,D, Supporting Information). Enlarging the wounding area did not alter the formation process or the relative size of the dome (Figure S1E,F). Previous work published on sarcolemmal repair in skeletal muscle is restricted to findings observed within a period up to 10 min after wounding [11-13, 17, 18, 20-23]. Here, we highlight membrane healing as a dynamic process lasting beyond this previously investigated time frame. The repair process includes ‘dome formation’ as well as continuous dysferlin accumulation.
To investigate whether the dome formation is limited to wound healing after laser induced damage, we performed mechanical wounding by interventional atomic force microscopy (AFM). Here, the sarcolemma is injured with the tip of the cantilever. By this second approach, we demonstrate that the repair dome is an integral part of the healing process independent of the wounding method (Figure 1D).
Ca2+-dependent membrane patch formation is the most widely accepted mechanism of membrane repair. We attempted to visualize the patch formation process at the injury site by AFM. This technique revealed a raising and highly corrugated surface at the site of membrane injury in comparison to the adjacent unwounded areas (Figure 1E). This suggests an intracellular directed transport of internal membrane compartments toward the injury site and would therefore be consistent with the membrane patch formation hypothesis. Our image is limited to the periphery of the wounded area because the large height gradient and the tip size and tilt of the cantilever prevented imaging of the entire repair dome.
Annexin A1 is involved in skeletal muscle membrane repair, whereas caveolin 3 is not
Other important components, such as annexin A1 and F-actin participate in membrane repair [15, 24]. Therefore we verified the presence of these proteins in wounded human myotubes and analyzed annexin A1 and F-actin localization 20 min after laser-mediated membrane wounding. Annexin A1 accumulated robustly at the wounded area (Figure 2A,B and D). The area of membrane healing could easily be identified even at low magnification (Figure 2A). F-actin was highlighted using phalloidin-atto 565 and was also considerably enriched within the same region (Figure 2C,D). We expressed GFP-dysferlin in a human myotube and localized annexin A1 after laser irradiation. Both proteins were present at the repair dome at the same time (Figure 2E). These results confirm the physiological relevance of our assay. The membrane repair process is Ca2+-dependent. Accordingly, we did not observe a repair dome formation without extracellular Ca2+, while fluorescence dye FM 1-43 continuously enters the myotube (Video S1, Supporting Information), and annexin A1 did not accumulate at the injury site (Figure S1A,B).
Caveolin 3 binds to dysferlin and co-localizes with dysferlin during intracellular trafficking [9, 10, 18]. Caveolin 3 also co-localizes with PTRF. Together, these two proteins are believed to construct and secure the shape of caveolae [17, 25]. PTRF has also been identified as a dysferlin binding partner  and PTRF co-localizes with MG53 in resealing events . However, neither CAV3 nor PTRF mutations give rise to a disease that resembles dysferlinopathy [27, 28]. We therefore wondered whether or not we could identify caveolin 3 during the membrane repair process in human primary myotubes following membrane injury. Different antibodies recognizing various epitopes of caveolin 3 were used to identify the protein. Furthermore, z-stack images covering a depth of more than 6 µm were taken to recognize potential regions of caveolin enrichment (Figures 3 and S2). We found that caveolin 3 was never present in the area of patch formation and membrane remodeling following sarcolemmal laser injury. We conclude from these findings that co-localization of caveolin 3 and dysferlin might be significant in dysferlin functions relating to intracellular trafficking but not in membrane repair.
EHD2 is a novel member of the membrane repair complex
We were intrigued by the appearance of tubular projections that extended radially from the repair dome structures after injury. These projections were best seen by live cell microscopy (Video S1) rather than fixed sections. The appearance of these projections brought to mind the dynamin superfamily of GTPases that are known to actively remodel membranes. Eps15 homology (EH)-domain-containing proteins (EHDs) comprise a class of highly conserved eukaryotic ATPases implicated in clathrin-independent endocytosis, and recycling from endosomes. The EHD protein family consists of a group of four known proteins that function in intracellular trafficking and that possess a dynamin-like structural fold. EHD1 and EHD2 are expressed in skeletal muscle, whereas EHD3 and EHD4 are predominantly present in heart and brain [29-31]. EHD1 is a key regulator of endocytic trafficking [32-34]. EHD2 is involved in nucleotide-dependent membrane remodeling . In skeletal muscle, both EHD1 and EHD2 are critical for myoblast fusion [36, 37]. We hypothesized that dynamin or a dynamin-like EHD may be involved at sites of sarcolemmal repair. EHD proteins are membrane benders and membrane binders in the presence of GTP or ATP. EHD2 dimers insert themselves into membranes using a highly curved interface and bend the membrane toward the dimer while hydrolyzing ATP. Purified EHD2 deforms liposomes into tubules in vitro and coats these tubules with ring-like oligomers . An important role for EHD2 in membrane fission and shaping has been predicted by mathematical modeling .
We identified endogenous EHD2 at the site of membrane repair after injury. F-actin and EHD2 co-localized at the site of the resealing area as detected immunohistochemically (Figure 4A–C). When EGFP-EHD2 was overexpressed in human myotubes it could be simultaneously detected with annexin A1 at the injury site (Figure 4D,E). Moreover, the tubular structures at the repair dome were clearly highlighted by EGFP-EHD2 15 min after wounding (Figure 5A). An EHD2 variant with a mutation in the membrane binding site, EGFP-EHD2 (K328D), was not recruited to the site of membrane repair (Figure 5B). The radial projections from the repair dome that drew our attention to the possible presence of EHD2 can be appreciated (Figure 5A,C and D). EH domains are protein interaction modules that recognize Asn-Pro-Phe (NPF) motifs in their biological ligands to mediate critical events during endocytosis and signal transduction. NPF motifs are present on actin-binding proteins such as the EH domain binding protein 1 (EHBP-1) . Such binding could connect the plasma membrane to the cytoskeleton. It is interesting that EHD2 recruited to the repair area appears fuzzy in staining, whereas the rest of EHD2 stains in a punctate manner. This would be consistent with vesicular storage of EHD2 but diffuse distribution in repair areas
EHD2 has been described to reside at caveolae structures controlling the stability of these organelles . We also find that in uninjured human muscle cells EHD2 co-localized with the muscle-specific marker of caveolae, caveolin 3 (Figure S3). In contrast, at the injury site, this co-localization is absent and a functional interaction during the membrane repair process can be excluded.
EGFP-EHD2 could be detected at the injury site approximately 40 seconds after wounding and like dysferlin continuously accumulated in the repair dome thereafter (Figures 1A and 6B). Besides EHD2, EHD1 is highly expressed in skeletal muscle. However, we did not detect recruitment of GFP-EHD1 to sites of injury (Figure 6A,B) despite the development of a repair dome (Figure 6C,D). This suggests that EHD2 may be more specifically involved in the wound healing process.
We adapted the laser-assisted wounding assay to primary human myotubes. We focused our attention on visualizing the area of membrane repair and patch formation for a period up to an hour after injury. Expression of GFP-labeled candidate proteins in combination with laser irradiation enabled us to confirm or exclude their involvement in membrane repair and track these proteins over time. Live cell imaging of the repair dome formation drew our attention to a protein that we could identify as a new member of the repair complex, EHD2.
The process of sarcolemmal repair might be divided into at least two parts: The early phase takes place immediately after injury and secures a very rapid resealing process to guarantee a primary cellular barrier. Consecutively, remodeling of the primary patch reconstitutes the membrane ad integrum. Visualizing the repair process itself for up to an hour led us to conclude, that this subdivision is highly artificial and that both processes are closely linked to each other. Both tracked proteins, dysferlin and EHD2, are quickly recruited to the injury site (seconds) but do still accumulate during the remodeling process. The same applies to the dome development. A primary intracellular patch formation relies on Ca2+ influx and probably on subcortical actin contraction, whereas the repair dome still increases in size while the membrane itself is restructured.
Our assay is not a high-throughput test and myotubes differ from intact myofibers. Further, laser induced membrane injury differs from physiological wounding elicited by exercise. Keeping these limitations in mind, we conclude that our approach is suitable and sufficient to study the molecular events of membrane repair, protein function and possibly the effects of therapeutic interventions in human myotubes obtained from patients suffering from a variety of different muscle disorders.
Materials and Methods
Cells and cell culture
Primary human myoblasts were isolated by protease digestion from fresh muscle biopsies. The biopsies were obtained for diagnostic purposes after informed consent was obtained. Muscle specimens were placed in 30 mm HEPES, 130 mm NaCl, 3 mm KCl, 10 mm d-Glucose, 3.2 μm Phenol red (pH 7.6) and subsequently digested in phosphate buffered saline (PBS) containing 254 U/mL Collagenase type CLS II (Biochrom AG, Germany), 2.5 U/mL Dispase II (Roche), 0.03% Trypsin and 125 μm EDTA for 45 min at 37°C. Outgrowing cells were expanded in Skeletal Muscle Cell Growth Medium (Promocell) containing the Supplement Mix for Skeletal Muscle Growth Medium, 10% fetal bovine serum (Lonza), 3 mm glutamine (GlutaMAX) and gentamicin (40 µg/mL) (Gibco). Myoblasts were purified using anti-CD56 coated magnetic beats (Miltenyi Biotech) according to the manufacturer's instructions. Purity of the myoblast preparation was confirmed by antibody staining (anti-human desmin, clone 33, Dako). Differentiation of myoblasts into myotubes was induced in a 3- to 5-day culture in Opti-MEM (Invitrogen).
Plasmids and microinjection
N-terminal EGFP-tagged mouse EHD2 and the K328D-mutant were previously described . The cDNA from EHD1 (a kind gift of M. Plomann, University of Cologne) was cloned into the pEGFP-C3 vector for overexpression in human myotubes. The pcDNA4/TO vector (Invitrogen) containing N-terminal GFP-tagged full-length human dysferlin was a gift of Steven Laval (Newcastle Upon Tyne, UK). For DNA-microinjections primary human myotubes were grown on culture dishes with a cover glass bottom (FluoroDish, World Precision Instruments Inc.). Microinjection was performed using the FemtoJet system combined with the InjectMan NI 2 (Eppendorf). Plasmid DNA was diluted in ddH2O to a final concentration of 250 µg/mL and centrifuged at 4°C for 45 min. The supernatant was loaded immediately onto the capillaries (Femtotip II, Eppendorf). Microinjections into the cytoplasm of myotubes were done with an injection pressure of 80 to 120 hPa. Twenty-four hours after microinjection protein expression was controlled by EGFP- or GFP-epifluorescence.
For laser wounding experiments primary human myoblasts were plated onto μ-Slide 8 well (ibidi, Germany) and fused to myotubes as described before. Plasmid DNA transfected myotubes were irradiated in glass bottom culture dishes (FluoroDish, WPI) 24 h after DNA microinjection. Briefly before laser wounding medium was switched to Tyrode solution (140 mm NaCl, 5 mm KCl, 2 mm MgCl2 and 10 mm HEPES, pH 7.2) containing 2.5 μm FM 1-43 or 2.5 μm FM 4-64 (Molecular Probes) and 2.5 mm CaCl2. For a Ca2+-free experiment, CaCl2 was not added to the Tyrode solution. Myotubes were wounded by irradiating a 2.5 × 2.5 µm boundary area of the plasma membrane at 100% maximum power (10 mW diode laser, 488 nm laser line) for 38 seconds (FM 1-43) or 76 seconds (FM 4-64) using a Zeiss LSM 700 confocal microscope with a LCI Plan-NEOFLUAR 63×/1.3 glycerin immersion objective (Zeiss). For comparison to experiments performed by others we also increased the wounding area to 5 × 5 µm in some experiments as indicated in the text. After injury 15 images were captured every 20 seconds. Digital images were processed using the Zeiss LSM ZEN software 2010.
Atomic force microscopy
For AFM imaging myotubes were wounded in the glass bottom culture dishes and fixed in 3.7% formaldehyde-PBS 20 min after laser irradiation. We used a Nanowizard II atomic force microscope (JPK Instruments) mounted on the Zeiss LSM 700. Due to sample's tall features we chose a cantilever with a very high tip (MikroMasch CSC38/B). This uncoated silicon cantilever has a nominal spring constant of 0.03 N/m, a resonant frequency of 10 kHz and a conical shaped tip with a 10 nm radius. We applied the thermal noise method to calibrate the cantilever . All AFM measurements were performed in contact mode under liquid (PBS) with an applied force setpoint of 0.5 nN. Using a Petri dish heater (JPK Instruments) we kept the temperature constant at 27°C. The AFM data were processed with JPK Data Processing 4.0.8 (JPK Instruments) and Gwyddion SPM software 2.2.4 (http://gwyddion.net). Very few isolated streaks were removed by line interpolation. Before extracting the height profiles we applied a Gaussian smoothing of three pixels to the height images.
For AFM wounding we used an uncoated silicon cantilever (MikroMasch CSC37/C) with a nominal spring constant of 0.35 N/m, a resonant frequency of 28 kHz, and a conical shaped tip with a 10 nm radius. Again, the thermal noise method was used to calibrate the cantilever. Corresponding to our laser wounding assay, we wounded the myotubes peripherally along a straight path of 2.5 µm applying a constant force of 30 nN and a speed of 1 µm/s. Starting with the wounding procedure we captured images every 6.7 seconds over a period of 5 min using a Leica DMI 6000 microscope with a HCX PL APO 63×/1.3 glycerin immersion objective. Digital images were processed using the LAS imaging software (Leica Microsystems GmbH).
Immunofluorescence and confocal microscopy
After laser wounding experiments, myotubes were washed with PBS and fixed in a 3.7% formaldehyde-PBS solution for 10 min at RT, rinsed in PBS and permeabilized with 0.2% Triton X-100 in PBS for 5 min. After washing repeatedly with PBS, blocking was performed with 1% BSA (Carl Roth) in PBS for 45 min. Primary antibody incubation was done overnight in PBS with 1% BSA at 4°C. After three washes with PBS secondary antibody incubation was performed for 1 h at RT in PBS. After two washes in PBS myotubes were mounted in Aqua PolyMount (Polysciences). Primary antibodies were used at following concentrations: 10 µg/mL polyclonal antibody to annexin A1 (abcam, ab88865), 0.5 µg/mL monoclonal antibody to caveolin 3 (santa cruz, sc-5310), 0.5 µg/mL polyclonal antibody to caveolin 3, (abcam, ab2912), 1 µg/mL polyclonal antibody to EHD2 (a kind gift of Richard Lundmark) . Secondary antibodies (anti rabbit Alexa Fluor 488, anti mouse Alexa Fluor 568, Invitrogen) were used at a dilution of 1:2000. F-actin staining was done with phalloidin-atto 565 (Sigma-Aldrich) in PBS after antibody incubation. Nuclei were stained using Hoechst 33342 (Invitrogen). All images were collected using a Zeiss LSM 700 confocal microscope equipped with a LCI Plan-NEOFLUAR 63×/1.3 glycerin immersion lens and a LD Plan-NEOFLUAR 40×/0.6 lens. Digital images were processed using Zeiss LSM ZEN software 2010. Images were assembled using CorelDRAW X5 software.
This study was supported by the German Research Society (DFG) (KFO192; DFG SP 1152/8-2). We thank the Jain Foundation for support of the AFM experiments. V. S. received a stipend of the Clinical Education Program of the Charité Medical Faculty and the Max Delbrück Center for Molecular Medicine, Berlin. We thank Stephanie Meyer for excellent technical assistance. There is no conflict of interests.