In this study we used two Panx1 null mouse strains: the Panx1KO/B6 (Genetech Inc.) and the CMV-Cre/Panx1 strain. The Panx1KO/B6 (Qu et al., 2011) was backcrossed to the C57BL/6 background (see Supplementary Methods) for 11 generations. The Panx1 null strain CMV-Cre/Panx1^fl/fl carries a floxed Panx1 allele that was activated with CMV-Cre to create a global knockout. This strain also carries a “passenger” Casp11 mutation (Dvoriantchikova et al., 2012). Caspase11 is induced in macrophages in response to injury or exposure to bacterial metabolites but is not detectable in homeostatic conditions in tissues such as muscle (Kang et al., 2000; Py et al., 2014). The Casp11^–/– [B6.129S4(D2)-Casp4^tm1Yuan/J] mouse strain in the C57BL/6 background (Jackson Labs) was used as a control for any effects that might be mediated by loss of Casp11 in the CMV-Cre/Panx1 strain. Wildtype C57BL/6 mice were used as an additional control.

Mice were housed under standard conditions of temperature and humidity, with a 12-h light/dark cycle and free access to food and water. All experiments were performed in compliance with the Guidelines for the Care and Use of Laboratory Animals published by the United States National Institutes of Health (NIH Publication No. 85-23, revised 1996) and were pre-approved by the Scripps Research Institute Animal Care and Use Committee.

Primary myoblasts were isolated from limb muscles of 3-week-old mice. Satellite cells were isolated using gentleMACS^TM Octo Dissociator and either the MACS Microbid satellite cell isolation kit (Miltenyl Biotech, CA, United States) or Fluorescence Activated cell sorting (FACS) as described previously (Sacco et al., 2010; Meech et al., 2011). Cells were grown on collagen-coated plates or chamber slides in growth medium (GM: 1:1 Ham’s F10/DMEM, supplemented with 20% FBS, and 2.5 ng/ml of basic FGF; Rando and Blau, 1994; Makarenkova et al., 2009). To induce myoblast differentiation, the GM was replaced with a differentiation medium (DM: DMEM supplemented with 2% horse serum).

Tibialis anterior (TA) muscles of anesthetized 3-month-old Panx1-null and wild-type mice were injected with cardiotoxin (CTX, 100 ng) to induce muscle injury, or with saline (control) as described previously (Sacco et al., 2010; Meech et al., 2011). Muscles were dissected at 2–10 days post-injury, fixed with buffered 4% paraformaldehyde (PFA). Paraffin sections prepared from the muscle midsection were stained with Hematoxylin and Eosin (H&E), and Masson’s Trichrome to detect collagen deposition, and imaged using a Leica scanner.

For gene expression analysis, RNA was extracted using Trizol or Zymo Direct-Zol RNA Kit (Zymo Research Corporation, Irvine, CA, United States) either from dissected TA tissues or from myoblast cultures. The amount of total RNA was estimated using a Nanodrop ND-1000 spectrophotometer and RNA purity and integrity (RIN) was assessed using a Bioanalyzer-2100 device (Agilent Technologies, Inc., Santa Clara, CA, United States). RNAs with RIN between 9 and 10 were used in all experiments. RNA was reverse transcribed to cDNA using the RT² First Strand Kit (#330404, Qiagen, Valencia, CA, United States). The Mouse ECM & Adhesion Molecules RT² Profiler PCR Array (PAMM-013Z; Qiagen) and RT² SYBR Green ROX qPCR Mastermix (#330522; Qiagen, Carlsbad, CA, United States) were used to measure expression levels of 84 individual genes involved in cell–cell and cell–matrix interactions. All other gene-specific primers were obtained from Qiagen. Quantitative real-time PCR was performed using an ABI 7300 Real-Time PCR System (Applied Biosystems, Life Technologies, Carlsbad, CA, United States). Statistically significant differences in threshold cycle (Ct) for each reaction were determined using the RT² Profiler PCR Array Data Analysis software v3.5 (Qiagen). Gene expression differences were filtered for >1.5-fold change and significance P < 0.05. Genes with highly variable Ct values were excluded from the final analysis. Reference genes for normalization of real-time PCR data were b-actin (ACTB), β-2 microglobulin (B2M), and glyceraldehyde-3-phosphate dehydrogenase (GAPDH).

Myoblast cultures were fixed with 2% PFA in PBS pH 7.5 for 20 min, permeabilized in Tris–buffered Saline with 0.1% Tween 20 (TBST), and blocked with 5% goat serum. Cells were immunostained with primary antibodies overnight at 4°C, followed by staining with appropriate secondary antibodies (antibodies are detailed in Supplementary Methods) and nuclear staining with DAPI. Images were taken using Zeiss LSM 710 laser scanning confocal microscope (LSCM).

Primary antibodies used: affinity purified rabbit Panx1 antibody CT-395 (Px-34; Penuela et al., 2007) was kindly provided by Dr. D. W. Laird (University of Western Ontario, Canada), rabbit polyclonal anti-human Panx1 antibodies (Chemicon, Inc.); mouse monoclonal anti-Myogenin (clone F5D, BD Bioscience Pharmingen, San Diego, CA, United States), mouse monoclonal anti-MyoD (clone MoAb5.8A, BD PharMingen, San Diego, CA, United States), pan myosin heavy chain (MyHC; A4.1025 Developmental Studies Hybridoma Bank, Iowa City, IA, United States), and mouse monoclonal anti-sarcomere myosin (clone MF20, Developmental Studies Hybridoma Bank, Iowa City, IA, United States), antibodies.

Total protein was prepared from primary myoblast cultures using Nonidet P-40 lysis buffer and sonicated. Equal aliquots of protein were resolved by SDS-PAGE using the 4% to 12% Bis-Tris gels (NuPAGE; Invitrogen), transferred to polyvinylidene difluoride membrane, and probed with polyclonal antibodies to Panx1 (Chemicon) and to β-actin, β-tubulin, or GAPDH. Bands were visualized by chemiluminescence using X-ray film exposure and quantified by densitometry. Each experiment was performed in duplicate.

Single muscle fibers were isolated from the extensor digitiorum longus (EDL) muscle of 12-week old male C57Bl6 as described previously (Otto et al., 2008, 2011). Briefly, the EDL was dissected with both tendons intact and myofibers liberated by digestion with (0.2%) type I collagenase (Sigma) in Dulbecco modified eagle medium (DMEM) at 37°C under 5% CO₂. Using tapered glass pipettes, single fibers were plated in single fiber culture medium (SFCM, DMEM supplemented with 10% horse serum, 0.5% chick embryo extract, and 1% penicillin/streptomycin).

Following 24 h of culture, myofibers were monitored using a phase contrast time-lapse microscope system. Myofibers grown in chamber sides were imaged in an environmental chamber maintained at 37°C and supplemented with 5% CO₂. Time-lapse video was taken at a rate of 1 frame every 15 min over 24 h for satellite cells using a 10X objective, a previously well-studied time course for satellite cell migration following single fiber isolation (Otto et al., 2008, 2011). All video analysis was carried out using freeware package ImageJ (version 1.49 m). Satellite cells were individually manually tracked using the MTrackJ plugin on ImageJ.

For electron microscopy, single myofibers were fixed following 48 h in standard culture with 4% PFA for 15 min, and then processed.

Fixed myofibers were dehydrated through 30, 50, 70, 80, 90, and 100% ethanol solution series (15 min for each step) and transferred to a critical point drier (Balzers CPD 030, using liquid carbon dioxide) thereafter. Dried myofibers were then carefully transferred to scanning electron microscopy (SEM) chucks using micro-forceps under a light microscope. Myofibers were then gold-coated using an Edwards S150B sputter-coater. Coated myofibers were then visualized under an FEI 600F SEM using the accompanying analysis software for image collection.

Mathematical models of particle motion are able to provide us with characterizations, which can be compared with data in order to predict how cells are moving (Woolley et al., 2013, 2014). Although we have shown that the shape of the cell and its blebs are critical to its migration properties (Woolley et al., 2015a,b), here, we simply characterize the motion characteristics in terms of the migration data. Specifically, we model the cells as random walkers on a straight, cylindrical fiber. Previously (Collins-Hooper et al., 2012), we have shown that, under this assumption, we can separate the cellular motion into two components; (i) a purely angular motion of the cell, around the muscle fiber and (ii) a purely longitudinal motion, along the axis of the fiber. We primarily focus of this longitudinal motion as it is the most crucial mode of movement.

Due to the probabilistic nature of the problem, we cannot specify exactly where the cell will be at all times in the future, however, we can predict the probability distribution of the cell locations. We define this probability distribution to be p(x,t), which represents the probability of finding a cell at a position x (measured along the fiber) at a given time t. The distribution satisfies the standard diffusion equation in one dimension (Belmonte-Beitia et al., 2013),

∂ ⁡ p ∂ ⁡ t = D ⁢ ∂ 2 ⁡ p ∂ ⁡ x 2 , p ⁢ ( x , 0 ) = 0 , p ⁢ ( ∞ , t ) = 0 .

This equation simply encapsulates the idea that the cells are moving randomly, all of the cell’s positions are normalized to start at zero and the fiber has no boundaries. The D parameter is a positive constant that measures that rate at which the cells spread out from the normalized origin. From this equation we derive that the mean squared displacement of the cells is proportional to time,

⟨ x 2 ⟩ = 2 ⁢ D ⁢ t .

Thus, taking the trajectory data we normalize the start position of all trajectories to zero and rotate all trajectories such that the fiber lies along the x-axis. The mean square distance moved by the cell along the fiber is calculated, and a straight line is fitted to this derived statistic. A goodness of fit R² value is also computed to express how well the straight line fits the data. The closer the R² value is to one, the closer the movement of the cells is to a memoryless, random diffusion.

To determine the fusion index, cultures were stained with the MF20 antibody and the numbers of nuclei were counted in MyHC+ cells. Fusion index was calculated as the number of nuclei in multinucleated MyHC+ cells (≥2 nuclei) as a proportion of total nuclei. Three independent experiments were performed for each condition; approximately 500 nuclei were counted per condition.

The erythrocyte exclusion assay was performed as previously described (Hattori et al., 2011; Stupka et al., 2013). Briefly, Panx1 and WT primary myoblasts were plated into two-well chamber slides and cultured for 24–48 h. The culture medium was replaced by freshly prepared erythrocyte suspension (1 × 10⁷ erythrocytes/ml) in serum-free DMEM. After erythrocytes had settled to the bottom, images (20–30 per experimental group) were taken with an inverted microscope. In some assays, Streptomyces hyalurolyticus hyaluronidase (H1136, Sigma) was added prior to erythrocyte addition.

Statistical significance was assessed using unpaired two-tailed Student’s t-test and results were considered significant at P < 0.05, or P < 0.01. Results were expressed as mean ± s.d.m (standard deviation of the mean).

Panx-1 was previously shown to be expressed in skeletal muscle (Langlois et al., 2014); however, its role during muscle differentiation and regeneration remains largely unknown. We examined the temporal pattern of PANX1 protein expression during myoblast differentiation using immunocytochemical, quantitative Western blot and gene expression analyses of primary cultures. Panx1 was expressed in isolated satellite cells (Figure 1A) and was increased rapidly at both mRNA and protein levels after the induction of myoblast differentiation (Figures 1A–C).

Muscle development and regeneration in the absence of Panx1 have not been previously studied. Basic morphometric analysis of muscle groups (TA, gastrocnemius, and soleus) in WT vs. Panx1^–/– mice showed that neither postnatal nor adult Panx1^–/– mice have any significant change in the size of limb muscles (data not shown). These results suggest that loss of Panx1 does not significantly affect muscle development and/or postnatal muscle growth.

To determine whether Panx1 may play a role in regeneration, we first examined the dynamics of pannexin expression during muscle regeneration in vivo. The TA muscle of WT mice was injured with notexin and changes in the levels of Panx 1–3 transcripts post-injury were assessed by qRT PCR. The Panx1 transcript levels were higher than those of the Panx2 and Panx3 genes at all stages (data not shown). On day 2 after notexin-injury (muscle inflammation/early regeneration phase), there was a 7.6 ± 1.9-fold increase in the amount of Panx1 transcript in injured vs. saline-injected control TA muscle (Figure 1D), while the amount of Panx2 transcript remained unchanged (Figure 1E). On day 5 after injury (late differentiation phase), Panx1 levels decreased relative to the day 2 peak level and was 2.4 ± 0.7 (Figure 1D). In contrast, Panx2 levels at this time point increased 11 ± 1.2-fold (Figure 1E). These findings suggest that these two pannexins may play different roles in the succession of injury-induced muscle repair phases. As expected (Langlois et al., 2014), the Panx3 mRNA level during muscle regeneration was very low and in some samples undetectable (data not shown).

We next investigated the role of Panx1 in muscle regeneration using two different Panx1 null models: CMV-Cre/Panx1^fl/fl mice for which Casp11^–/– mice served as the relevant control, and Panx1KO/B6 mice with WT mice as the control. TA muscles were injured with notexin and regeneration studied 10 days later. All mice showed regeneration as evidenced by appearance of new myofibers with centrally located nuclei (Figure 1F). However, the newly regenerated myofibers in both Panx1^–/– models were significantly smaller in diameter when compared to those of the control strains (Figure 1G). In addition, we found that regenerated TA muscles of Panx1^–/– mice were depleted for large myofibers and enriched in smaller myofibers, as compared to wild-type TA muscles (Figure 1H).

The reduction in myofibre size in regenerating Panx1^–/– muscle suggested a possible myoblast fusion defect. We thus compared differentiation of Panx1^–/– (both Panx1-null strains) and control (Casp11^–/– and/or WT) myoblasts in culture. Control myoblasts cultured in DM formed numerous multinucleated myotubes (Figures 2A,B and Supplementary Figure 1A) and by 72 h the cultures were almost free of unfused myoblasts (Figures 2C,D and Supplementary Figure 1A). In contrast, Panx1^–/– myoblasts developed very few multinucleated cells over 72 h in DM (Figures 2E–H and Supplementary Figure 1B). Analysis of Casp11^–/– myoblast differentiation showed that they differentiated normally and did not differ morphologically from WT myoblasts. Both differentiating control and Panx1^–/– cultures expressed MyHC protein (Figures 2A–H).

The process of myoblast differentiation is tightly controlled by myogenic regulatory factors (MRFs). Cultured myoblasts normally express MyoD in GM and initiate expression of myogenin as early as 1–3 h after induction of differentiation (Makarenkova et al., 2009). There was no difference in the number of MyoD expressing cells in Panx1^–/– relative to control myoblasts at 3 and 24 h-hours post induction of differentiation (Figure 2I). Moreover, the proportion of myogenin (MYOG)-positive nuclei in Panx1^–/– and control myoblast cultures 3 and 24 h after induction of differentiation was not significantly different (Figure 2J). The time to induction of MyHC and percentage of MyHC expressing cells were also similar in both cultures (Figures 2A–H). However, relative to control (WT and Casp11^–/–) myoblasts, Panx1^–/– myoblasts formed significantly fewer multi-nucleated (3 or more nuclei per cell) MyHC positive myotubes (Figure 2K and Supplementary Figure 1C). Interestingly, unfused myoblasts in Panx1^–/– cultures also expressed MyHC despite the impaired myotube formation (Figures 2G,H). In summary, our results suggest that activation of the myogenic differentiation program is not affected in Panx1^–/– myoblasts. This finding corroborates previous reports that myoblast fusion is not required for the expression of MyHC (Devlin and Emerson, 1978, 1979).

To form myofibers, differentiating myoblasts must migrate and establish stable cell-cell contacts (Asakura et al., 2001). Real-time in vivo imaging has revealed rapid migration of myoblasts to the site of myofiber damage in vivo (Ishido and Kasuga, 2011, 2012; Webster et al., 2016). Several research groups, including ours, reported that rapid movement of myoblasts requires formation of dynamic membrane blebs (Charras and Paluch, 2008; Makarenkova et al., 2009; Otto et al., 2011), as they utilize an amoeboid-based rather than lamellipodia-mediated propulsion mechanism (Otto et al., 2011). Morphometric analysis showed that cultured primary Panx1^–/– myoblasts had fewer blebs than control myoblasts (Figure 3A and Supplementary Figure 1, compare A and B). Analysis of Panx1^–/– myoblasts showed that they were significantly less migratory and stayed clustered, in contrast to Control (WT and Casp11^–/–) myoblasts that readily spread throughout the dish (Figures 3C,E). Moreover, while control myoblasts had wide, rounded surface blebs, the blebs on Panx1^–/– myoblasts were thinner. The average length-to-width ratio of Panx1^–/– myoblast blebs was 2-fold less than that of controls (1.23 ± 0.42 for CMV-CrePanx1^fl/fl vs. 0.79 ± 0.18 for Casp11^–/– and 1.45 ± 0.54 for Panx1KO/B6 vs. for WT myoblasts, p > 0.001), (Figures 3B,D,F).

We next tested whether pharmacological blockade of Panx1 channels in culture could phenocopy Panx1 genetic ablation. Carbexonolone (CBX) inhibits ATP release through pannexin channels (Hayoz et al., 2012; Xia et al., 2012) and has high specificity for Panx1 channels at <100 μM (Bao et al., 2012). Undifferentiated wildtype myoblasts were exposed to CBX (10, 25, and 50 μM) for 24 h prior to induction of differentiation and then transferred into DM also containing CBX. Myoblast treatment with CBX for 24 h produced significant changes in the morphology and number of surface blebs. With increasing CBX concentration, the number of surface blebs per cell decreased significantly after 10 μM CBX applicationammount of protein; at 25–50 μM CBX myoblasts retained only a few small protrusions (Figures 3G–K).

The morphological changes induced by CBX negatively correlated with adhesion of treated myoblasts to the collagen substrate, causing cells to detach in plate-washing assays more readily (Figure 3L and Supplementary Methods). Thus, 50 μM CBX treatment caused a 5-fold increase in myoblast detachment relative to control (Figure 3L). In contrast, treatment with 10–25 μM of CBX did not induce significant myoblast detachment and we used these drug concentrations in all further experiments. After 24 h exposure to 10 and 25 μM CBX, myoblasts were induced to differentiate in DM also containing CBX, and expression of MYOG and MyHC was assessed by immunocytochemistry (Figures 3M–Q). Similar to Panx1^–/– myoblasts, the number of MYOG expressing cells remained unchanged in CBX treated myoblasts (Figures 3M–O), implying that the core transcriptional differentiation program remained unaltered. The induction of MyHC protein expression also appeared unaffected in control and CBX treated cultures (Figures 3P,Q and Supplementary Figures 2A,B). However, unlike control myoblasts (Figures 3N,P), myoblasts treated with CBX failed to elongate and showed almost no fusion (Figures 3O,Q). The similarities in differentiation deficiencies of CBX-treated and Panx1^–/– cells, suggest that Panx1 and ATP mediated signaling controls myoblast migration and fusion.

Next, we examined whether signaling downstream of Panx1, particularly release of intracellular ATP, is involved in myoblast migration and fusion. We treated myoblasts with apyrase, an enzyme that catalyzes the breakdown of ATP into AMP and inorganic phosphate. Apyrase treatment did not alter the timing of MYOG expression or the number of MYOG-expressing cells (Figures 4A,B, white arrows, C), but dramatically inhibited myotube formation (Figures 4D–L). After 24 h in DM, apyrase treated myoblasts elongated but did not form myotubes (Figure 4, compare E, F with G, H) and after 48 h the percentage of myotubes with 3 or more nuclei was much lower than in control cultures (Figure 4D, compare I, J with K, L). This effect phenocopied Panx1^–/– and CBX-treated myoblasts (see Figures 2, 3). Thus, both ablation of Panx1 and pharmacological blockade of ATP lead to altered cell shape, migration, and fusion properties, suggesting that Panx1 mediates these events through ATP signaling.

Extracellular ATP is known to transmit signals by activating purinergic ionotropic P2X or metabotropic P2Y receptors (Wilkinson et al., 1993; Meyer et al., 1999; Ryten et al., 2004; Martinello et al., 2011); involvement of P2X receptors in C2C12 myoblast differentiation was documented previously (Araya et al., 2004). To test whether purinergic signaling via P2X7 is involved in myoblast migration and differentiation, we blocked this receptor in WT primary myoblasts with P2X7 receptor antagonist A740003. A740003 treatment inhibited myotube formation relative to control (Figures 4M,N), while again MYOG expression was unaffected (Figure 4O). These data suggest an important role for P2X7-mediated signaling in myotube formation.

We next examined the effect of inhibiting the P2X7 channel on cell blebbing and migration using single myofiber cultures where individual satellite cells migrate along the myofibers. Treatment of fiber associated-satellite cells with A740003 significantly reduced bleb formation and migration speed (Figures 5A–E). Lipid signaling mediated by LPA was shown to be a major determinant in bleb formation in a variety of cell types (Hagmann et al., 1999; Jia et al., 2006). Moreover, P2X7 receptors activate a signaling pathway that induces Phospholipase D (PLD) to convert phosphatidylcholine to phosphatidic acid (PA) which in turn is transformed into lysophosphatidic acid (LPA) by Phospholipase A₂ (Panupinthu et al., 2007). The hydrolytic reaction mediated by PLD requires water and is inhibited by primary but not tertiary alcohols. To assess whether P2X7 may control blebbing through (L)PA production, we repressed PLD activity by addition of 1-butanol to satellite cells on their native substrate and found that it resulted in a significant decrease in the number of blebs displayed on the cell surface, as well as a reduction in migration speed (Figures 5F–J). In a control experiment, an identical concentration of tertiary butanol had no effect on either parameter (Figures 5G–J). These data indicate that P2X7-facilitated blebbing and migration is reliant on PLD activity to form PA. The action of lipids to induce signaling cascades is heavily dependent on the conversion of PA into LPA by phospholipase A₂, which in turn activates the LPA receptor, a G-coupled protein that controls a diverse activity including cell migration. We examined whether LPA activity was important in blebbing and satellite cell migration using VPC-32182, a selective LPA1 and LPA3 receptor antagonist. Treatment with VPC-32182 dramatically reduced bleb formation and rate of satellite cell migration (Figures 5K–O), showing a role for LPA activity in these processes. LPA receptors can activate Rho, which ultimately leads to myosin light chain phosphorylation hence controlling actin/myosin-based contraction which may drive bleb dynamics. We found that inhibition of myosin light chain kinase with a ML7 inhibitor reduced satellite cell blebbing and migration speed (Figures 5P–T).

To show a more direct link between Panx1/P2X7-mediated ATP signaling and the activity of the LPA receptor in controlling blebbing and migration, we inhibited Panx1 channels with CBX or P2X7 receptors with A740003, and simultaneously introduced LPA. The inhibitory effect of both CBX and A740003 on bleb formation and migration was rescued by LPA (Figures 5U–Y for A740003 and Z-Ac for CBX).

Finally, we used mathematical modeling to examine whether interfering with the Panx1-mediated lipid based signaling pathway in myoblasts influenced not only migration speed but also directionality. We have previously demonstrated that satellite cells from young mice move in a random, directionless manner on the myofiber that contrasts the behavior of cells from old mice, which show directionality (Otto et al., 2011). Plotting the mean square displacement over time showed that although A740003, Butanol, VPC-32183 and ML7 treatments all decreased migration speed of young myoblasts on single myofiber, they still moved in a random manner indicated by a straight-line plot (Figures 5E,J,O,T). Furthermore, rescue of the migration deficit induced by A740003 or CBX with LPA did not alter the migration mode (Figures 5Z, Aa–Ad).

The studies above define a signaling pathway activated by Panx1 that regulates bleb formation and migration of satellite cells. Here we determined whether the same signaling components also regulate myoblast fusion. Freshly isolated satellite cells were cultured at high density permitting cell-cell contact and then maintained in GM or DM for 72 h under a range of drug treatment conditions. In the control conditions, cells in GM remained largely mono-nucleated with few cells expressing MyHC (Figure 6A), while cells in DM formed myotubes with an average of 4.5 nuclei and robust MyHC expression (Figure 6B). Addition of Panx1 channel inhibitor CBX or P2X7 channel inhibitor A740003, to high-density cultures in DM inhibited fusion despite the expression of MyHC (Figures 6C–F). Likewise, fusion was attenuated by addition of butanol, LPA receptor inhibitor VPC32183, or ROCK inhibitor Y-27632 (Figures 6G–L). Finally, inhibition of fusion by blockade of either Panx1 or P2X7 receptor could be reversed by LPA addition (Figures 6M–P). Together these results show Panx1 signaling facilitates myoblast fusion through the P2X7 receptor, downstream lipid intermediaries and the ROCK pathway that activates blebbing.

Pannexins mediate dynamic adhesive contacts and thus regulate diverse functions including cell alignment, adhesion, migration and fusion (Gorbe et al., 2007; Bao et al., 2012). To test whether loss of Panx1 affects expression of key genes involved in these processes, we isolated Panx1^–/– and WT myoblasts from 4-month-old mice by FACS and profiled the expression of 84 genes important for cell-cell and cell-matrix interactions. We found nine genes that were significantly altered (>1.5-fold) in Panx1^–/– vs. WT myoblasts (Table 1). Among the genes significantly up-regulated in Panx1^–/– vs. WT myoblasts were Cadherin-2 (Cdh2), Integrin-β4 (Itgb4), Collagen VI (Col6a1), and sparc/osteonectin (SPOCK1). Among the genes down-regulated in Panx1^–/– cultures, we identified transforming growth factor, beta-induced gene (TGFBI/BIGH3; 3.04-fold change, p = 0.008). TGFBI activation has been previously shown to promote myofibril bundling and ECM deposition (Kim and Ingham, 2009). We also found significant downregulation of Elastin Microfibril Interface-Located Protein 1 (Emilin1) that was recently shown to be responsible for the formation of the elastic fibers and anchoring muscle cells to them (Gundry et al., 2009). Multiple Matrix Metalloproteinases (MMPs) were also down-regulated, particularly MMP12 (5.6-fold; p = 0.02), and the disintegrin metallopeptidase with thrombospondin type 5 motif gene (Adamts5, 1.79-fold; p = 0.01; Table 1).

ADAMTS versicanases such as ADAMTS5 have been shown to mediate proteolysis of a hyaluronan and versican-rich matrix leading to pericellular matrix clearance enabling migration, cell-cell interaction, and fusion (D’hondt et al., 2010; Stupka et al., 2013). Moreover, a recent report shows that Adamts5 gene knockdown impairs myoblast fusion in vitro (Stupka et al., 2013). Hypothesizing that this gene may play a role in the migration- and fusion-defective phenotypes of Panx1^–/– myoblasts, we tested whether impaired Panx1 signaling disrupts pericellular matrix accumulation using an erythrocyte exclusion assay. Consistent with our hypothesis, we observed significantly more pericellular matrix in Panx1^–/– myoblast cultures than in WT (Supplementary Figure 3, compare A and B, D). Treatment of Panx1^–/– myoblasts with hyaluronidase blocked erythrocyte exclusion (Supplementary Figures 3C,D) indicating that the pericellular matrix is hyaluronan-based (and thus versican containing). These data suggest that reduced levels of ADAMTS5 and/or other extracellular proteinases in Panx1^–/– myoblasts may interfere with fusion by attenuating extracellular protein remodeling.