Myomerger induces fusion of non-fusogenic cells and is required for myoblast fusion

Despite the importance of cell fusion for mammalian development and physiology, the factors critical for this process remain to be fully defined1. This lack of knowledge has severely limited our ability to reconstitute cell fusion, which is necessary to decipher the biochemical mechanisms driving plasma membrane merger. Myomaker (Tmem8c) is a muscle-specific protein required for myoblast fusion2,3. Expression of myomaker in fibroblasts drives their fusion with myoblasts, but not with other myomaker-fibroblasts, highlighting the requirement of additional myoblast-derived factors for fusion. Here, we demonstrate that Gm7325, named myomerger, induces the fusion of myomaker-expressing fibroblasts. Cell mixing experiments reveal that while myomaker renders cells fusion-competent, myomerger induces fusogenicity. Thus, myomaker and myomerger confer fusogenic activity to normally non-fusogenic cells. Myomerger is skeletal muscle-specific and only expressed during developmental and regenerative myogenesis. Disruption of myomerger in myoblast cell lines through Cas9-mutagenesis generated non-fusogenic myocytes. Genetic deletion of myomerger in mice results in a paucity of muscle fibers demonstrating a requirement for myomerger in normal muscle formation. Myomerger deficient myocytes exhibit an ability to differentiate and harbor organized sarcomeres, however remain mono-nucleated. These data identify myomerger as a fundamental myoblast fusion protein and establishes a system that begins to reconstitute mammalian cell fusion.

five genes: Tmem182, Gm7325,Cdh15,Tspan33,and Tm6sf1,however Cdh15 was omitted 25 from further analysis because it is not necessary for myoblast fusion or muscle formation 22 . We 26 retrovirally expressed each gene in myomaker + GFP + fibroblasts and assayed for fusion. 27 Appropriate expression in fibroblasts was verified through quantitative reverse transcription 28 polymerase chain reaction (qRT-PCR) analysis (Extended Data Fig. 1a). We observed mainly 29 mono-nucleated GFP + cells in all cultures except when Gm7325 was expressed where 30 widespread multi-nucleated cells were present (Fig. 1a). Based on the ability of Gm7325 to 31 induce fusion of myomaker + fibroblasts and the observations described below we named the 32 control. Flow cytometry of GFP + cells followed by genotyping through PCR analysis revealed 121 disruption of the myomerger locus (Extended Data Fig. 5b). Myomerger was not detectable in 122 myomerger KO C2C12 cells confirming efficient disruption of the locus (Fig. 3a). Control and 123 myomerger KO C2C12 cells were then analyzed for their ability to differentiate and form 124 myotubes. WT myoblasts differentiated, as indicated by myosin + cells, and fused to form multi-125 nucleated myotubes (Fig. 3b). In contrast, myomerger KO C2C12 cells exhibited the ability to 126 differentiate but lacked fusogenic activity to form myotubes (Fig. 3b). Indeed, quantification of 127 the differentiation index revealed no difference in the percentage of myosin + cells between WT 128 and myomerger KO cultures (Fig. 3c). Additionally, quantification of fusion demonstrated that 129 myomerger KO myosin + cells remain mono-nucleated while WT cells fuse (Fig. 3d). qRT-PCR 130 analysis for the myogenic genes Myogenin, Myh4,Ckm,and Tmem8c (myomaker) further 131 indicated that myomerger KO myoblasts activate the differentiation program (Fig. 3e). 132 Interestingly, myogenic transcripts were elevated in myomerger KO cells potentially suggesting 133 a feedback mechanism by which non-fusogenic cells attempt to further differentiate (Fig. 3e). 134 Infection of myomerger KO C2C12 cells with either myomerger-S or myomerger-L rescued the 135 fusion defect demonstrating that the phenotype in these cells is specifically due to the loss of 136 myomerger (Extended Data Fig. 5c). Western blot analysis from these lysates shows re-137 expression of myomerger in KO cells (Extended Data Fig. 5d). As a potential mechanism for the 138 lack of fusion in myomerger KO myocytes, we examined expression and localization of 139 myomaker. On day 2 of differentiation, myomerger KO cells exhibited normal expression and 140 localization of myomaker (Extended Data Fig. 6a). Moreover, we did not detect widespread co-141 localization between myomaker and myomerger suggesting that myomerger does not directly 142 regulate myomaker distribution (Extended Data Fig. 6b). These data reveal that myomerger is 143 necessary for myoblast fusion in vitro.
To examine the function of myomerger in vivo, we disrupted exon 3 using the same 145 CRISPR/Cas9 strategy described for C2C12 myoblasts. Injection of Cas9 and myomerger 146 gRNAs into blastocysts resulted in lethality of 9 of the 10 F 0 pups, suggesting that the high 147 efficiency of Cas9 lead to homozygous deletion of myomerger. The one remaining pup was 148 heterozygous for myomerger and was mated to generate Gm7325 -/mice (Extended Data Fig.  149 7a). Sequencing the mutant PCR product from the heterozygous founder revealed the presence 150 of the same mutation as was achieved in C2C12 cells. We failed to observe any Gm7325 -/mice 151 upon genotyping two litters at P7 suggesting that, in conjunction with the lethality detected 152 through initial generation of F 0 pups, myomerger is essential for life. Indeed, E17.5 Gm7325 -/-153 embryos exhibited minimal skeletal muscle upon gross examination (Fig. 4a). Specifically, bones 154 of the limbs and rib cage were noticeable due to a scarcity of surrounding muscle as observed 155 in WT embryos. Myomerger KO mice also display a hunched appearance with elongated snouts, 156 hallmark characteristics in embryos with improper muscle formation (Fig. 4a). Detection of 157 myomerger by western blot of WT and Gm7325 -/tongues showed elimination of myomerger 158 protein in KO samples (Extended Data Fig. 7b). E15.5 forelimb sections show that myomerger 159 KO myoblasts express myogenin indicating that specification and differentiation are activated 160 despite loss of myomerger (Fig. 4b). Moreover, histological analysis of multiple muscle groups 161 at E15.5 revealed the presence of myosin + muscle cells and sarcomeric structures in myomerger 162 KO mice, (Fig. 4c and Extended Data Fig. 7c). While multi-nucleated myofibers were present in 163 WT mice, these structures were not readily detected in myomerger KO mice indicating that 164 genetic loss of myomerger renders myocytes non-fusogenic ( Fig. 4c and Extended Data Fig. 7c). 165 Analysis of forelimbs from E17.5 WT and myomerger KO embryos confirm that myomerger KO 166 myoblasts are unable to properly fuse, although we did detect myocytes with two nuclei 167 (Extended Data Fig. 7d-f). These results, together with our in vitro analysis, reveals that myomerger is required for muscle formation during mammalian development through regulation 169 of myoblast fusion. 170 In summary, we report the discovery of an additional muscle-specific factor required for 171 myoblast fusion and developmental myogenesis. While myomaker and myomerger are both 172 necessary for muscle formation, E17.5 myomerger KO embryos exhibit more myocytes 173 compared to embryos lacking myomaker suggesting that these two key myoblast fusion proteins 174 may have distinct functions. Given that myomaker and myomerger are both essential for 175 myoblast fusion, an understanding of their precise function will aid in the delineation of 176 mammalian plasma membrane fusion mechanisms. The fibroblast cell fusion system developed 177 here, through expression of myomaker and myomerger, provides a unique platform to decipher 178 fusion mechanisms. Moreover, the induction of cell fusion by two factors represents an avenue 179 for gene delivery to potentially any tissue.

Bioinformatic Analysis 191
Microarray data from the GEO DataSet GSE34907 28 was interrogated using GEO2R analysis to 192 identify 1826 genes displaying an increase greater than 1 log fold-change in  fibroblasts. In parallel, a transcriptional profile of 10T 1/2 fibroblasts transduced with empty virus 194 was generated using RNA-seq analysis (paired-end library layout using Illumina sequencing 195 platform) and a list of all genes with RPKM values below 1.5 compiled using Strand NGS 196 software (Ver. 2.6; Build: Mouse mm10 (UCSC) using Ensembl transcript annotations). These 197 two gene lists were then compared to generate a final tally comprised of 531 genes that were 198 both upregulated in MyoD-expressing fibroblasts and had low or no detectable expression in 199 10T 1/2 fibroblasts. Finally, the top 100 genes were interrogated for genes that contain 200 transmembrane domains and not previously studied for their role during myoblast fusion. 201 202

Animals 203
We used a dual sgRNA targeting strategy to create Gm7325 -/mice. We selected the sgRNAs 204 according to the on-and off-target scores from the web tool CRISPOR 29 . The selected gRNAs 205 were 5'-GCAGCGATCGAAGCACCATC-3' and 5'-GAGGCCTCTCCAGAATCCGG-3' that target 206 exon 3 of Gm7325. The sgRNAs were in vitro synthesized using the MEGAshortscript T7 kit 207 (ThermoFisher) and purified by the MEGAclear Kit (ThermoFisher), as previously described 30 . 208 sgRNAs (50 ng/ul of each) were mixed with 100 ng/ul Cas9 protein (ThermoFisher) and 209 incubated at 37°C for 15 min to form a ribonucleoprotein complex. We then injected the mix into 210 the cytoplasm of one-cell-stage embryos of the C57BL/6 genetic background using a piezo-211 driven microinjection technique as described previously 30 . Injected embryos were immediately 212 transferred into the oviducal ampulla of pseudopregnant CD-1 females. Live born pups were 213 genotyped by PCR with the following primers: F: 5′-GAAGGGAGGACTCCACACCC-3' and R: 214 5'-CGCCTGGACTAACCGGCTCC-3'. The edited allele was further confirmed by Sanger 215 sequencing. One heterozygous founder was obtained and mated with WT C57Bl6 mice to 216 eventually generate KO mice. Mdx 4cv mice were purchased from Jackson Laboratory (#002378). 217 Muscle overload of the plantaris muscle was achieved through bilateral synergistic ablation of 218 soleus and gastrocnemius muscles as described by others 31 . Briefly, the soleus and 219 gastrocnemius muscles were exposed by making an incision on the posterior-lateral aspect of 220 the lower limb. The distal and proximal tendons of the soleus, lateral and medial gastrocnemius 221 were subsequently cut and carefully excised. All animal procedures were approved by Cincinnati 222 Children's Hospital Medical Center's Institutional Animal Care and Use Committee. 223

CRISPR-Mediated Genome Editing in C2C12 Cells 225
Freshly plated low passage C2C12 cells were transfected with 4μg of a modified pX458 plasmid 226 (Addgene #48138, gift from Yueh-Chiang Hu), which contained a high fidelity Cas9, an optimized 227 sgRNA scaffold, and an IRES-GFP cassette. The same gRNAs used to generate KO animals 228 were used for C2C12 cells. 16 μL of Lipofectamine 2000 was used for this transfection. 5 x 10 5 229 C2C12 cells were transfected in a 60 mm culture dish. Forty-eight hours after transfection GFP + 230 cells were sorted into 96 well plates using FACS. These cells were maintained in DMEM 231 containing 20% FBS with antibiotics at subconfluent densities. The cell lines were genotyped by 232 amplifying a 420 bp region surrounding the site of Cas9 activity using the primers used to 233 genotype Gm7325 -/animals. 234 235

Cloning and viral infection 236
We initially cloned a region of the Gm7325 locus, containing all genomic information for 237 expression of myomerger-short and myomerger-long, from C57Bl6 mouse genomic DNA using 238 the following primers: F: 5'-AGTGATGCTGAATCCACCGCA-3' and R: 5'-239 CCAATAACAACACACTGTCCT-3'. We cloned myomerger-short and long coding sequences 240 from cDNA of differentiating C2C12 cells using the following primers: myomerger-long F: 5'-241 ATGCCAGAAGAAAGCTGCACTG-3', myomerger-short F: 5'-242 ATGCCCGTTCCATTGCTCCCGA-3', and a common myomerger R: 5'-TCA CTT CTG GGG 243 GCC CAA TCT C-3'. Myomerger cDNA and genomic DNA was cloned into the retroviral vector 244 pBabe-X 2 using EcoRI. Myomaker and GFP retroviral plasmids have been described previously 2 . 245 NLS-TdTomato was subcloned from pQC-NLS-TdTomato (Addgene #37347) into the retroviral 246 vector pMX (Cell Biolabs). Plasmids containing cDNA for Tmem182, Tspan33, and Tm6sf1 from 247 the Mammalian Gene Collection were purchased from Open Biosystems and subcloned into 248 pBabe-X. Ten micrograms of retroviral plasmid DNA were transfected with FuGENE 6 (Roche) 249 into Platinum E cells (Cell Biolabs), which were plated 24 hours before transfection on a 10 cm 250 culture dish at a density of 3-4x10 6 cells per dish. Forty-eight hours after transfection, viral media 251 were collected, filtered through a 0.45 µm cellulose syringe filter and mixed with polybrene 252 (Sigma) at a final concentration of 6 µg/ml. Target cells were plated on 10 cm culture dishes at a density of 4x10 5 cells per dish 16-18 hours before infection. Eighteen hours after infection,

Subcellular fractionation 304
C2C12 cells were harvested on day 2 of differentiation in ice cold hypotonic buffer (10 mM Tris-305 HCl pH 8, 2 mM EDTA) and lysed using a dounce homogenizer. Lysates were then centrifuged 306 at 800 x g for 5 mintues at 4°C to separate nuclei and cell debris. That supernatant was then 307 centrifuged at 5000 x g for 10 minutes to pellet mitochondria and ER. ER and heavy vesicles 308 were further pelleted through centrifugation at 17,000 x g for 10 minutes. Finally, plasma 309 membrane, light vesicles, and organelles were pelleted at 100,000 x g for 20 minutes and the 310 supernatant from this spin was collected as the cytosolic fraction. All pellets were resuspended We used antigen-specific affinity purified products at a concentration of 4.3 µg/mL for 325 immunostaining. Esgp (myomerger) antibody was used at a concentration of 1 µg/mL. Anti-326 mouse myosin (my32, MA5-11748, ThermoFisher Scientific) antibody was used at 1:100. 327 Hoechst 33342 solution (ThermoFisher Scientific) was used to stain nuclei. Cells were imaged 328 using Nikon A1R+ confocal on a FN1 microscope (35 mm dishes) or Nikon A1R confocal on 329 Eclipse T1 inverted microscope (Ibidi slides). In Fig. 1c, the number of myosin + myotubes (myosin structures with 3 or more nuclei) and GFP + 348 myosin + myotubes were manually counted. To quantify fusion between myomaker + myomerger + 349 GFP + fibroblasts with either myomaker + NLS-Tom + or myomerger + NLS-Tom + fibroblasts 350 (Extended Data Fig. 3), we calculated the percentage of GFP + NLS-Tom + syncytial cells. The 351 differentiation index (Fig. 3c) was calculated as the percentage of nuclei in myosin + cells, and 352 the fusion index (Fig. 3d)  showing a screen for muscle genes that could activate fusion of GFP + myomaker + fibroblasts. Representative images of GFP + cells and nuclei after expression of the indicated genes. Arrows depict cells with multiple nuclei. b, Illustration of cell mixing approach to show fusion between the populations of fibroblasts. Co-localization of GFP and NLS-TdTomato (NLS-Tom) in the nucleus represents fusion. Representative images demonstrate fusion of myomaker + myomerger + fibroblasts but not empty-infected myomaker + fibroblasts. Arrows indicate fusion between GFP + and NLS-Tom + fibroblasts. The percentage of nuclei in syncytia after expression of empty or myomerger (n=3). c, Heterologous fusion experiment between C2C12 myoblasts and GFP + fibroblasts infected with either empty, myomaker, or myomerger. Representative immunofluorescent images to visualize co-localization of myosin and GFP (arrows), indicating fusion. Quantification of the percentage of GFP + myosin + cells (n=3). Data are presented as mean ± SEM. *P<0.05 compared to empty. # P<0.05 between myomaker and myomerger. Scale bars , 50 μm (a, b), 100 μm (c).   The short transcript is highly conserved in multiple species, including human, but not present in zebrafish. The upstream exon that produces the longer transcript is not highly conserved. Note that this annotation displays the gene on the reverse strand. c, The short (S) or long (L) myomerger transcripts were expressed in myomaker + 10T ½ fibroblasts and both induced fusion (n=3). d, Myomerger also induced fusion of myomaker + NIH/3T3 fibroblasts (n=3) and myomaker + mesenchymal stromal cells (n=3). Arrows indicate fusion. Scale bars, 50 μm. Diagram showing the cell mixing approach to assess fusion between the populations of fibroblasts. Co-localization of GFP and NLS-TdTomato (NLS-Tom) in the nucleus represents fusion (arrows). Representative images demonstrate fusion of myomaker + myomerger + GFP + fibroblasts with myomaker + NLS-Tom + fibroblasts but not myomerger + NLS-Tom + fibroblasts. The percent of GFP + NLS-Tom + syncytia (n=3). Data are presented as mean ± SEM. *P<0.05 compared to myomerger + NLS-Tom + fibroblasts. Scale bars, 50 μm. . c, Sequence alignment of both mouse myomerger protein products with multiple mammalian orthologs using Clustal Omega. A potential hydrophobic region is highlighted in gray. d, Immunoblotting from C2C12 cells infected with either empty, myomerger-short (S), myomerger-long (L) on day 2 of differentiation. Myomerger migrates as a single band around 12 kDa when endogenously produced (empty). Over-expression of myomerger-S leads to an increase in the endogenous band and a lower band is also detected suggesting that myomerger transcripts may be subjected to intricate mRNA processing or post-translational modifications. e, Graphic showing the regions of myomerger-S and myomerger-L as predicted by SignalP and Phobius.