Local translational program at the muscle-tendon junction endows domain identity in muscle syncytia

42 How cells establish specialized subdomains is a fundamental question in cell and tissue biology. 43 Skeletal muscle fibers, among the largest cells in the body, are multinucleated and form distinct 44 regions such as the neuromuscular and myotendinous junctions (MTJ), the latter forming a critical 45 interface between muscle and tendon that transmits contractile force. While transcriptional 46 heterogeneity among myonuclei has been described, whether local translation contributes to domain 47 identity remains unknown, largely due to the lack of tools for domain-specific manipulation. Here, 48 we introduce MTJ-AAV, a viral system that enables selective genetic targeting of MTJ myonuclei. 49 This approach allowed MTJ -specific ribosome tagging and revealed extensive translational 50 regulation underlying MTJ biology and its remodeling during exercise. Interestingly, untranslated 51 regions of these transcripts were sufficient to control regionalized translation. Notably, the KLF -52 family transcription factors emerged as translationally upre gulated targets at the MTJ, where they 53 drive local gene expression. Our findings establish local translation as a key layer of subcellular 54 specialization and provide a versatile toolkit for dissecting spatial molecular regulation within 55 muscle syncytia. 56 57

58 To precisely coordinate diverse biochemical activities and biological functions, cells must establish 59 specialized subcellular domains. How these domains form, how they are maintained and remodeled 60 remain a fundamental question in cell biology. Skeletal muscle cells, or myofibers, provide a unique 61 model to study this process because of their exceptionally large cytoplasm, which contains hundreds 62 to thousands of nuclei . Different myofiber regions interact with distinct external cues, giving rise 63 to specialized functional domains. For example, myofiber central regions interact with motor 64 neurons, forming the neuromuscular junction (1, 2), while their ends attach to tendons, creating the 65 myotendinous junction (MTJ) (3, 4). The MTJ plays a critical role in dissipating the mechanical 66 forces generated during contraction and transmitting them to tendons. Consequently, the MTJ is 67 highly susceptible to injury, and its dysfunction leads to muscle rupture (5). Understanding how 68 these muscle domains are regulated not only advances our knowledge of myology and disease, but 69 also provides broader insights into how cells spatially and functionally organize their intracellular 70 space. 71 Previous studies using single -nucleus RNA sequencing (snRNA -Seq) have shown that muscle 72 nuclei at specialized domains adopt distinct gene expression programs, conceptually resembling 73 cellular differentiation in multicellular tissues (6-9). These studies highlight transcriptional control 74 as an important mechanism for domain formation. However, the contribution of post-transcriptional 75 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2026.01.01.697307doi: bioRxiv preprint Page 3 of 27 regulation remains largely unexplored. Translational regulation , the control of protein synthesis 76 from existing mRNA , has emerged as a central mechanism across diverse biological contexts, 77 including gametogenesis, neurodevelopment, cancer, and inflammatory signaling (10-14). In 78 muscle, translational profiling has thus far been limited to whole myofibers (15-17), leaving little 79 insight into how localized protein synthesis supports muscle domain biology . Interestingly, 80 although the MTJ undergoes structural remodeling during exercise and aging (18, 19), snRNA-Seq 81 studies have not detected major transcriptomic changes in MTJ myonuclei (7, 20). This discrepancy 82 suggests that post-transcriptional regulation, particularly translation, may play a key role in these 83 adaptations. 84 Conventional genetic approaches in the muscle field lack spatial resolution and target all myonuclei 85 indiscriminately. The distinct lack of genetic tools that can manipulate specific domains is the major 86 roadblock preventing the field from studying regionalized molecular events in myofibers. To 87 overcome this limitation, we reasoned that promoters of MTJ-specific genes identified from 88 snRNA-Seq datasets could be leveraged to drive gene expression specifically within this domain. 89 Here, we develop a versatile adeno-associated virus (AAV) system that enables expression of 90 genetic tools specifically at the MTJ. Using this platform, we profile the local translational 91 landscapes of the MTJ and uncover extensive translational regulation that becomes markedly 92 amplified during endurance exercise. We further demonstrate that untranslated regions (UTRs) of 93 MTJ-enriched transcripts are sufficient to confer localized translation and that KLF transcription 94 factors are translationally upregulated to drive local gene expression. Together, these findings 95 reveal a previously unrecognized layer of spatial regulation underlying this specialized muscle 96 domain, and provide a powerful toolkit for dissecting subcellular regulation in a syncytium. 97 98 99 100 101 102 103 104 105 106 107 108 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2026.01.01.697307doi: bioRxiv preprint Page 4 of 27

109 110 Development of MTJ-AAV 111 To selectively target the MTJ, we focused on Tigd4, a highly specific marker of MTJ nuclei (6). 112 We synthesized a fragment of the Tigd4 promoter, spanning from the transcription start site to -3.3 113 kb, and cloned it into an AAV plasmid to drive the expression of super -fold GFP (sfGFP) fused 114 with four SV40-derived nuclear localization signals (NLS). The 3.3 kb length was chosen to 115 accommodate AAV genome size constraints. The resulting pAAV -Tigd4-NLS-sfGFP construct 116 was packaged i nto the MyoAAV4a serotype (21) and intramuscularly injected into 10 -week-old 117 wild-type ( WT) mice (Fig. 1A). As a control, we performed parallel experiments using the 118 conventional CK8 promoter, which is active in all myonuclei. 119 As expected, CK8 promoter-driven AAV broadly labeled the entire muscle tissue with NLS-sfGFP 120 when tibialis anterior (TA) muscles were examined one -month post -injection. In contrast, the 121 Tigd4 promoter restricted reporter expression to the MTJ, identifiable by characteristic DAPI -122 stained structures penetrating into the muscle (Fig. 1B). Quantification of labeling efficiency using 123 GFP immunostaining combined with RNAscope for the MTJ markers Tigd4 and Col22a1 revealed 124 that approximately 80% of MTJ nuclei (defined by co-expression of Tigd4 and Col22a1) were GFP-125 positive (Fig. 1C and 1D). Conversely, only ~ 20% of GFP -positive nuclei were not MTJ nuclei, 126 indicating high labeling specificity. Importantly, all GFP-positive nuclei were myonuclei as they 127 also expressed Ttn, a pan-myonuclei marker (Fig. 1D), and located adjacent to the MTJ. 128 To confirm the molecular identity of the labeled nuclei via an orthogonal method, we isolated GFP-129 positive nuclei by FACS (Fig. 1E) and performed bulk RNA sequencing (Fig. 1 F). The analysis 130 showed that many representative MTJ markers, including Tigd4, Col22a1, Resf1, Lama2 and 131 Ankrd1, to be enriched in GFP-positive nuclei from the Tigd4 promoter-driven AAV compared to 132 those labeled by the CK8 promoter (Fig. 1F). From here on, we refer to our new AAV system that 133 uses Tigd4 promoter as ‘MTJ-AAV’. 134 We next tested whether we could express other genetic tools at the MTJ as it will broaden the utility 135 of MTJ -AAV. As such, we asked whether MTJ-restricted recombination could be achieved by 136 delivering Cre recombinase. MTJ -AAV carrying Cre -NLS was intramuscularly injected into 137 Rosa26-LSL-H2B-GFP reporter mice (Fig. 2A), resulting in GFP labeling around the MTJ (Fig. 138 2B). Quantification using MTJ markers showed a similar labeling efficiency of ~75% (Fig. 2C). 139 However, off -target labeling was higher in this model, with ~40% of GFP -positive nuclei not 140 corresponding to MTJ nuclei (Fig. 2C). Similar to our previous approach, all labelled nuclei were 141 Ttn-positive, confirming their myonuclear identity. 142 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2026.01.01.697307doi: bioRxiv preprint Page 5 of 27 To investigate why there was higher off -target labelling with Cre compared with NLS -sfGFP, we 143 first tested whether Cre protein diffuses to neighboring nuclei. Cre/GFP co-immunohistochemistry 144 revealed that only ~56% of GFP -positive nuclei exhibited a detectable Cre signal (Fig. S1A and 145 S1B). Given the ~40% off -target labeling, this result suggests that Cre protein remains largely 146 restricted to MTJ nuclei with minimal diffusion, although we cannot rule out the possibility that 147 undetectable amounts of Cre protein contribute to leaky recombination. Next, we tested whether 148 H2B-GFP protein is more diffusive than the NLS-sfGFP protein used earlier by performing a side-149 by-side comparison . Expressing H2B -GFP via MTJ -AAV resulted in a similar ~80% labeling 150 efficiency but a higher off-target labeling rate of ~40% compared to ~20% with NLS-sfGFP (Fig. 151 S1C). These findings indicate that diffusion properties of expressed factors can vary and require 152 careful assessment before downstream experiments. 153 Finally, w e tested whether MTJ -AAV could be delivered systemically (Fig. 2D) . Intra-orbital 154 injection of MTJ -AAV carrying Cre -NLS into the H2B -GFP reporter successfully labeled MTJ 155 nuclei across multiple muscles, including the gastrocnemius and diaphragm (Fig. 2E). 156 157 Genetic labeling of MTJ ribosomes and translational profiling 158 Although protein synthesis is central to cellular identity and function, regionalized translation at the 159 MTJ has not been explored, largely due to the absence of suitable tools. We hypothesized that 160 combining MTJ-AAV with genetic ribosome tagging would enable selective profiling of ribosome-161 associated transcripts. 162 We therefore intramuscularly injected MTJ-AAV Cre-NLS or CK8-Cre-NLS into LSL-HA-Rpl22 163 (RiboTag) mice (22), in which Cre-mediated recombination introduces an HA tag into the 164 endogenous Rpl22 ribosomal subunit (Fig. 3A). HA immunohistochemistry confirmed the tight 165 expression of HA-Rpl22 at the MTJ (Fig. 3B). We also performed the same experiment in Ribo -166 Trap mice ( Rosa26-LSL-GFP-L10a) (23) and observed similar results. However, GFP -L10a 167 displayed broader expression than HA -Rpl22 (data not shown), likely due to its overexpression 168 from the Rosa26 locus. For this reason, we focused on the RiboTag model for subsequent analysis. 169 To isolate ribosome -associated transcripts, TA tissue lysates were subjected to anti -HA 170 immunoprecipitation. Bioanalyzer profiles confirmed robust RNA recovery from Cre -delivered 171 RiboTag muscles, with no detectable RNA in saline -injected controls (Fig. 3C). RNA sequencing 172 was then performed to map the MTJ translational landscape (n = 4 for whole muscle, n = 3 for 173 MTJ). Notably, MTJ markers Tigd4 and Col22a1 were among the most enriched riboso me-174 associated transcripts, validating the specificity of our dataset (Fig. 3D and 3E). 175 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2026.01.01.697307doi: bioRxiv preprint Page 6 of 27 Strikingly, ~80% of MTJ ribosome -enriched transcripts (17 0 of 209; fold -change > 2, p < 0.05) 176 have not been identified as transcriptionally enriched in published snRNA-Seq datasets or in our 177 bulk RNA-Seq from GFP -positive MTJ nuclei in Figure 1F (Fig. S2A). This indicates that they 178 correspond to transcripts that are translationally controlled at the MTJ (Supplementary Table 1). 179 Consistent with MTJ function, Gene Ontology analysis of these genes showed enrichment for those 180 related to extracellular matrix biology, cytoskeletal organization, and plasma membrane structur e 181 (Fig. 3E). Intriguingly, angiogenesis-related genes were also enriched, though the significance of 182 this remains unclear. Manual inspection identified additional targets of diverse function, including 183 the E3 ligase s Klhl42, Rnf152 and Rnf43 and transcription factors such as Klf2 and Klf5. 184 Conversely, translationally repressed genes were mainly linked to mitochondria and sarcoplasmic 185 reticulum biology (Fig. S2B and S2C), suggesting distinct organelle composition or density at the 186 MTJ. 187 188 Rewiring of translational program upon sustained exercise 189 Endurance exercise remodels the MTJ by increasing the depth and branching of its interdigitated 190 structures (24). To examine whether this event involves translational control, we subjected whole-191 muscle and MTJ ribosome-tagged mice to an endurance training regimen, followed by Ribo some 192 immunoprecipitation and sequencing (Fig. 3A). 193 We first identified a set of genes that were translationally induced throughout the muscle in response 194 to exercise, mainly being inflammatory mediators ( Nfkbia, Nfkb2, Stat4 , Socs3 ) (Fig. S 2D). 195 Notably, many of these were also induced in the M TJ translatome upon exercise, indicating a 196 genuine muscle-wide response (Fig. S2D and S2E). Indeed, pathway analysis of genes upregulated 197 in both CK8 and MTJ datasets revealed strong enrichment for cytokine-related signaling pathways 198 (Fig. S2F). 199 We next co mpared MTJ ribosome -enriched transcripts to CK8 -labeled ribosomes in exercised 200 muscles (Fig. 3G). The number of MTJ -enriched genes increased substantially under exercise 201 (1,054 versus 209 in sedentary conditions), with significant overlap between the two s ets (Fig. 202 S2G). Importantly, even shared targets exhibited markedly stronger enrichment after training (Fig. 203 3H). For instance, Cpne2 increased from a 1.3-fold enrichment (log₂ scale) at rest to 2.4-fold after 204 training, while Klf5 rose from 1.1- to 2.9-fold (Supplementary Table 1). Together, these findings 205 demonstrate that endurance exercise profoundly rewires the translational landscape at the MTJ, 206 amplifying a pre-existing local translational program to meet the heightened mechanical demands 207 of sustained activity. 208 209 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2026.01.01.697307doi: bioRxiv preprint Page 7 of 27 Role of untranslated regions in conferring translational specificity 210 Next, we asked how certain transcripts are selected for local translation . Untranslated regions 211 (UTRs) often contain regulatory elements that control where and how efficiently mRNAs are 212 translated (25, 26 ). To test whether such elements mediate local translation at the MTJ, we 213 generated translation reporters based on MTJ-enriched genes. We selected Arhgap27 and Cpne2 as 214 models because they were not previously identified as transcriptional markers of MTJ myonuclei 215 but showed enrichment in our RiboTag-Seq profiling. Moreover, their UTRs are sufficiently long 216 (~1 kb) to harbor potential regulatory motifs yet compact enough to remain compatible with AAV 217 packaging constraints. Arhgap27 encodes a poorly characterized Rho GTPase –activating protein 218 (27), whereas Cpne2 encodes a similarly underexplored calcium -dependent phospholipid-binding 219 protein that links membranes to the cy toskeleton (28). Both genes therefore represent plausible 220 effectors of MTJ function, where precise coordination between cytoskeletal and membrane 221 dynamics is essential. 222 We constructed AAVs in which the pan-myonuclear CK8 promoter drives an NLS-sfGFP reporter 223 either alone or fused to the 5′ and 3′ UTRs of Arhgap27 or Cpne2 (Fig. 4A). In the absence of 224 UTRs, the reporter protein was uniformly expressed throughout the myofiber, whereas inclusion of 225 the UTRs resulted in markedly enriched expression at the fiber tips (Fig. 4B). RNAscope analysis 226 confirmed that GFP transcripts remained uniformly distributed even in the presence of the UTRs 227 (Fig. 4C), indicating that localization occurs at the translational level rather than through 228 transcriptional control or mRNA transport. Therefore, the UTRs of Cpne2 and Arhgap27 are 229 sufficient to confer MTJ-specific translation. 230 231 Translational control of KLF contributes to MTJ specific gene expression 232 While previous snRNA-Seq studies identified hundreds of MTJ -enriched transcripts, they did not 233 uncover any transcription factors that are strongly and uniquely enriched at the MTJ. This contrasts 234 with the neuromuscular junction, where key transcription factors such as Etv4 and Etv5 are 235 specifically expressed (6-9). However, motif analysis of a published snATAC-Seq dataset predicted 236 a strong enric hment of KLF -binding motifs in MTJ -specific open chromatin regions (8). We 237 verified this with an independently acquired snATAC -Seq dataset (full data to be published 238 elsewhere): 8,601 MTJ-specific chromatin accessible regions were identified that majorly l ocated 239 in promoters (~50% of all peaks) (Fig. 5A). Motif analysis of the accessible regions predicted KLF 240 factors as the top candidates (Fig. 5B). These analyses raise the question of how KLF factors might 241 selectively function at the MTJ. 242 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2026.01.01.697307doi: bioRxiv preprint Page 8 of 27 In this light, we were intrigued to find that KLF factors, Klf2 and Klf5, to be enriched in our 243 RiboTag-Seq datasets of the MTJ (Fig. 3E and 3G). We first confirmed the enrichment of Klf2 244 protein at the MTJ. Klf2 immunohistochemistry in tissue sections showed higher Klf2 signal in the 245 nuclei located at the MTJ (Fig. S3A). We also surgically fractionated MTJ and non-MTJ areas using 246 Soleus muscle. We chose this muscle type because of its flat -shaped MTJ anatomy, making the 247 fractionation more suitable. Western blotting for Co l22a1 confirmed the validity of this strategy, 248 where Klf2 protein was also found to be more abundant at the MTJ fraction (Fig. S3B). 249 To probe KLF function while circumventing redundancy issue among KLF family members, we 250 employed a dominant-negative (DN) approach. We fused the DNA-binding domain of Klf2 to the 251 transcriptional repressor domain of Drosophila Engrailed (Fig. 4C), a strategy previously validated 252 in other biological contexts (29). For proper comparison, we intramuscularly injected a control virus 253 (Tigd4 promoter expressing NLS-sfGFP) on one hindlimb, and KLF2-DN virus on the contralateral 254 muscle (Fig. 5C). Targeted expression of Klf2 -DN at the MTJ suppressed MTJ gene expression, 255 particularly for genes containing KLF motifs identified in snATAC-Seq datasets like Col22a1 and 256 Tigd4 (Fig. 5D and S3C). A reduction in Col22a1 expression was also observed at the protein level 257 by immunohistochemistry, displaying reduced signals in KLF-DN transduced muscles (Fig. S3D). 258 Therefore, KLF transcription factors are translationally controlled at the MTJ, where they drive a 259 downstream transcriptional program required for MTJ specialization (Fig. 5E). 260 261

262 Our study establishes a framework for investigating regulatory mechanisms that allow the 263 formation of s pecialized domains in multinucleated muscle fibers. By developing a viral system 264 with MTJ specificity, we achieved domain-restricted genetic manipulations and ribosome tagging. 265 These approaches uncovered extensive translational regulation at the myotendino us junction . A 266 broad range of protein are controlled by this mechanism, encompassing extracellular matrix and 267 cytoskeletal regulators as well as transcription factors. Previous studies largely attributed the 268 domain formation to a transcriptional specialization of local myonuclei . The results extend our 269 understanding of how muscle domains are formed and maintained, and expand the growing 270 appreciation that translational control represents a central regulatory layer in cell and tissue biology. 271 Mechanistically, our data indicate that untranslated regions of MTJ-enriched genes play a key role 272 in conferring domain-restricted translation. Moreover, we show that the MTJ translational program 273 is intensified by endurance exercise, indicating that this mechanism is dyn amically regulated 274 according to physiological demand. The precise cis -regulatory motifs and RNA -binding proteins 275 involved remain to be identified. Future studies should determine how specific transcripts are 276 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2026.01.01.697307doi: bioRxiv preprint Page 9 of 27 selected for local translation (for example, through UTR motifs, RNA secondary structures, 277 ribosome heterogeneity, or other mechanisms) and how these processes integrate with tissue-level 278 signals at the MTJ such as tendon-derived or mechanical cues (Fig. 5E). 279 We also found that KLF transcription factors are translationally upregulated at the MTJ and 280 contribute to local gene expression. This explains why KLF motifs are highly enriched in accessible 281 chromatin regions of MTJ nuclei , although Klf transcript levels at the MTJ ar e similar to those 282 observed in bulk myonuclei . However, KLF factors are not strictly MTJ -exclusive; they are also 283 detectable in non -MTJ nuclei, albeit at lower levels. This suggests that KLF expression alone is 284 insufficient to convey MTJ identity and must act in concert with additional, yet unidentified 285 regulators to establish the full MTJ transcriptome. 286 A key innovation of our work is the flexibility of the MTJ -AAV strategy. Beyond transcriptional 287 and translational profiling, it can be adapted to deliver diverse genetic tools to investigate the MTJ. 288 Moreover, the same strategy could be extended to other muscle domains, such as the neuromuscular 289 junction, and to other syncytial systems. For instance, distinct nuclear subtypes have been described 290 in placental syncytiotrophoblasts by snRNA -Seq (30), suggesting that similar approaches could 291 illuminate nuclear subtype -specific regulation in these contexts as well. Nonetheless, such 292 applications must be pursued with caution, as biomolecular diffusion can confound specificity, as 293 demonstrated by our comparisons of NLS -sfGFP versus H2B -GFP and Rpl22 -HA versus GFP -294 L10a. With careful validation, however, MTJ-AAV and related strategies promise to be come 295 powerful tools for dissecting the spatial logic of gene regulation in syncytial cells. 296 297

298 299 Animals 300 Wild-type C57BL/6N mice were purchased from Charles River. Rosa26-LSL-H2B-GFP mice were 301 described previously (6). RiboTag mice were obtained from Jackson Laboratory (#029977). Mice 302 were housed under standard conditions: constant ambient temperature (23 °C), humidity (56%), and 303 a 12 -hour light/dark cycle (lights on at 6:00 am, off at 6:00 pm). Animals were euthanized by 304 gradual CO₂ inhalation (up to 100%) over a 3-minute period. All procedures were approved by the 305 institutional animal ethics committee of IGBMC (Comite d'Ethique). Animal health and welfare 306 were continuously monitored by trained staff and veterinarians to minimize suffering. 307 308 309 310 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2026.01.01.697307doi: bioRxiv preprint Page 10 of 27 AAV construction and generation 311 3.3kb of Tigd4 promoter was synthesized by GenScript and cloned into pAAV plasmid using 312 pAAV-MCS2 (Addgene) as the backbone. Recombinant adeno -associated virus (rAAV) were 313 generated by a triple transfection of HEK2 93T/17 cell line using Polyethylenimine (PEI) 314 transfection reagent and the 3 following plasmids: the expression plasmids (pAAV), the 315 pMyoAAV4A encoding Rep and Cap genes, and the pHelper (Agilent) encoding the adenovirus 316 helper functions. 48 h after transfection, rAAV vectors were harvested from cell lysate and treated 317 with Benzonase (Merck) at 100U/mL. They were further purified by gradient ultra -centrifugation 318 with Iodixanol (OptiprepTM density gradient medium) followed by dialysis and concentration 319 against Dulbecco’s Phosphate Buffered Saline (DPBS) using centrifugal filters (Amicon Ultra -15 320 Centrifugal Filter Devices 100K, Millipore). Viral titres were quantified by Real -Time PCR using 321 the LightCycler480 SYBR Green I Master (Roche) and primers targeting respective insert 322 sequences ( e.g., sfGFP). Titers are expressed as genome copies per milliliter (GC/mL). The 323 pMyoAAV4A plasmid was self -constructed by IGBMC’s molecular biology platform (31), 324 benchmarking the published construct (21). 325 326 AAV injection 327 Mice were anesthetized via intraperitoneal injection of a ketamine/xylazine mixture composed of 328 10% ketamine (100 mg/mL), 5% xylazine (20 mg/mL), and 85% sterile NaCl, administered at a 329 total dose of 100 µL per 10 g of body weight. Anesthesia depth was verified by the absence of pedal 330 reflex before proceeding with the injections. 331 For intramuscular (IM) injection, the skin over the target muscle was disinfected with 70% ethanol 332 under anesthesia. MyoAAV4a vectors carrying different constructs were injected intramu scularly 333 into the tibialis anterior muscle using a 30 -gauge needle. Each construct was administered at its 334 respective final viral genome (vg) dose per muscle: 5 × 10¹⁰ - 10¹¹ vg, depending on the viral 335 concentration and experimental design. The total injec tion volume per muscle was adjusted with 336 sterile NaCl to ensure a constant final volume of 20 µL across all conditions. Contralateral muscles 337 received vehicle or a control vector where applicable. Muscles were collected at 1 -month post-338 injection for molecular and histological analyses. 339 For systemic administration, anesthetized mice were placed in a prone position, and MyoAAV4a 340 particles were delivered via retro -orbital sinus injection using a 30 -gauge insulin syringe. A total 341 volume of 100 µL corresponding to 1.5 x 1013 vg per mouse was injected slowly over approximately 342 15 seconds to minimize reflux. Animals were monitored until full recovery and returned to their 343 cages. Tissue collection was performed at 2 months post-injection. 344 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2026.01.01.697307doi: bioRxiv preprint Page 11 of 27 Ribosome immunoprecipitation and sequencing (RiboTag-Seq) 345 10 weeks old homozygous RiboTag mice were intramuscularly injected with CK8 - or MTJ-AAV 346 carrying Cre recombinase. TA muscles were harvested one month later, snap-frozen, and stored at 347 –80°C until processing. 348 Frozen tissues were thawed on ice and transferred to homogenization tubes containing silica beads 349 (Precellys P000918 -LYSK1-A). For the MTJ -AAV condition, lysates were pooled from four 350 injected muscles from different mice. For the CK8 -AAV condition, lysates were pooled from two 351 injected muscles and two un-injected WT muscles to normalize total lysate concentration between 352 conditions, as this might affect HA-based immunoprecipitation efficiency and purity. 353 Each sample was homogenized in 1 ml lysis buffer containing 50 mM Tris-HCl (pH 7.5), 100 mM 354 KCl, 12 mM MgCl₂, 1% NP -40, 1 mM DTT, 200 U/ml RNAsin (Promega), 1 mg/ml heparin 355 (Sigma), 100 µg/ml cycloheximide (Sigma), and protease inhibitor cocktail (Roche). DTT, 356 RNAsin, heparin, cycloheximide and protease inhibitor were added freshly before use. Tissues were 357 minced with sterile scissors and incubated on ice for 20 minutes before homogenization using the 358 Precellys Evolution system (5,000 rpm, 25 seconds, twice, with 10-minute intervals on ice). Lysates 359 were clarified by centrifu gation at 10,000g for 20 minutes at 4°C, and the supernatants were 360 transferred to fresh tubes. 361 Protein concentration was determined via Bradford assay (Bio-Rad). Approximately 3 mg of lysate 362 was incubated with 5 µl HA antibody (Covance, clone 12CA5) for 4 hours at 4°C with gentle 363 rotation. Meanwhile, magnetic Protein A/G beads (Pierce) were equilibrated in lysis buffer. Forty 364 microliters of beads were added to each sample, followed by overnight incubation at 4°C with 365 rotation. 366 Beads were washed three times with high-salt buffer (same as lysis buffer, but with 300 mM KCl 367 and 0.5 mM DTT), each wash involving 5 minutes of rotation. After the final wash, beads were 368 resuspended in 800 µl high-salt buffer and transferred to a new tube. RNA was extracted by adding 369 350 µl of RLT buffer (Qiagen RNeasy kit) and following the manufacturer’s protocol, including 370 DNase treatment. Purified RNA was stored at –80°C until further use. 371 Library preparation was performed at the GenomEast platform at the Institute of Genetics and 372 Molecular and Cellular Biology using Illumina Stranded Total RNA Prep Ligation with Ribo-Zero 373 Plus (Reference Guide - PN 1000000124514). Total RNA-Seq libraries were generated from 50 ng 374 of total RNA using Illumina Stranded Total RNA Prep, Ligation with Ri bo-Zero Plus kit and IDT 375 for Illumina RNA UD Indexes, Ligation (Illumina, San Diego, USA), according to manufacturer’s 376 instructions. DNA library were amplified using 14 cycles of PCR. Surplus PCR primers were 377 further removed by two successive purifications using SPRIselect beads (Beckman -Coulter, 378 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2026.01.01.697307doi: bioRxiv preprint Page 12 of 27 Villepinte, France). The final libraries were checked for quality and quantified using Bioanalyzer 379 2100 system (Agilent technologies, Les Ulis, France). Libraries were sequenced on an Illumina 380 NextSeq 2000 sequence r as paired -end 50 base reads. Image analysis and base calling were 381 performed using RTA version 2.7.7 and BCL Convert version 3.8.4. 382 383 Endurance exercise 384 RiboTag mice were intramuscularly injected with CK8 -Cre-NLS or Tigd4 -Cre-NLS AAV. One 385 month post-injection, the mice underwent an endurance exercise regimen using an animal treadmill 386 (LE8710, Panlab, Harvard Apparatus, Spain). To ensure proper acclimatization, the mice were 387 introduced to the treadmill environment over a period of three days prio r to the start of the 388 experimental protocol. During this familiarization phase, the treadmill operated at a low speed of 389 10–12 cm/s for 15 minutes each day. The exercise training protocol began at 12 cm/s for 15 minutes, 390 increased to 20 cm/s for 50 minutes, and finished with a 5-minute cooldown at 10 cm/s. All exercise 391 sessions were consistently conducted at the same time of day for 4 weeks (5 days consecutively and 392 2 days rest). 393 394 FACS isolation and low-input RNA-Seq 395 TA muscles injected with CK8 -NLS-sfGFP or Tigd4 -NLS-sfGFP AAVs were collected, snap 396 frozen, and stored at –80°C until further use. The TA muscles were minced and incubated in 300µl 397 of cold hypotonic buffer (250mM sucrose, 10mM KCl, mM MgCl2, 10mM Tris-HCl pH 8, 25mM 398 HEPES pH 8, 0.3% Triton x-100, 0.2mM PMSF, 0.1 mM DTT, and 0.2U/µL RNase inhibitor) for 399 5 mins. Samples were then transferred to 2ml ‘Tissue homogenizing CKMix’ (Bertin Technologies) 400 with an additional 700µl hypotonic buffer. Following a further 15mins incubation, the samples were 401 then homogenized with the ‘Precellys 24 tissue homogenizer’ (Bertin Technologies) for 25s at 402 5,000rpm. The homogenized samples were then passed through a 100µm filter (Sysmex) followed 403 by a 20µm filter (Sysmex) before t he nuclei were pelleted by centrifugation at 400g for 10 min at 404 4 °C. The pellets were then washed and resuspended in a washing buffer (2% BSA in PBS + RNase 405 inhibitor 0.2U/µL). The centrifugation and wash steps were then repeated before the homogenized 406 samples were passed through a ‘5ml polystyrene round -bottom tube with cell -strainer cap’ 407 (Corning). DAPI was then added to the resuspended nuclei at a final concentration of 200mM. The 408 isolated nuclei were sorted with FACs Aria Fusion 2022, with BD FACSDiva and FlowJo (v10) 409 software, to sort out the DAPI and GFP+ nuclei. Approximately 1,000 nuclei were collected from 410 each sample. The sorted nuclei were immediately collected in 11.5µL of CDS sorting solution, 411 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2026.01.01.697307doi: bioRxiv preprint Page 13 of 27 SMART-Seq® mRNA LP (with UMIs) recipe (Takara), bri efly spun down, then flash -frozen on 412 dry ice. The frozen isolated nuclei were then stored at -80°C until further use. 413 Library preparation was performed at the GenomEast platform at the Institute of Genetics and 414 Molecular and Cellular Biology using Takara Bio USA, Inc., SMART-Seq mRNA User Manual - 415 PN 062223 + Illumina Nextera XT DNA Library Prep Kit (Reference Guide - PN 15031942). Full 416 length cDNA were generated from 100 to 2000 nuclei using SMART -Seq mRNA (Takara Bio 417 Europe, Saint Germain en Laye, France) according to manufacturer’s instructions with 11 cycles of 418 PCR for cDNA amplification by Seq -Amp DNA polymerase. The entire volume of each pre -419 amplified cDNA was then used as input for Tn5 transposon tagmentation followed by 12 cycles of 420 library amplifica tion using Nextera XT DNA Library Preparation Kit and IDT for Illumina 421 DNA/RNA UD Indexes, Tagmentation (Illumina, San Diego, USA). Following purification with 422 SPRIselect beads (Beckman -Coulter, Villepinte, France), the size and concentration of libraries 423 were assessed by capillary electrophoreris (Bioanalyzer 2100 system, Agilent technologies, Les 424 Ulis, France). Full length cDNA was generated from from 100 to 2000 nuclei using SMART -Seq 425 mRNA Kit (Takara Bio Europe, Saint Germain en Laye, France) according to manufacturer’s 426 instructions with 11 cycles of PCR amplification by Seq -Amp polymerase. Totality of pre -427 amplified cDNA were then used as input for Tn5 transposon tagmentation followed by 12 cycles 428 of library amplification using Nextera XT DNA Library Preparation Kit and IDT for Illumina 429 DNA/RNA UD Indexes (Illumina, San Diego, USA). Following purification with SPRIselect beads 430 (Beckman-Coulter, Villepinte, France), the size and concentration of libraries were assessed using 431 Bioanalyzer 2100 system (Agilent technologies, Les Ulis, France). Libraries were sequenced on an 432 Illumina NextSeq 2000 sequencer as paired -end 50 base reads. Image analysis and base calling 433 were performed using RTA version 2.7.7 and BCL Convert version 3.8.4. 434 435 Bioinformatic analyses of RNA-Seq datasets 436 All sequencing was performed on an Illumina NextSeq 2000 platform in a 2x50bp paired -end 437 configuration. For RiboTag-Seq, sequencing data was processed processed using PiGx -RNA-seq 438 (30277498) pipeline. In short, the data was mapped onto the GRCm39/mm11 version of the mouse 439 transcriptome (downloaded from the ENSEMBL database 29155950) using SALMON (28263959). 440 The quantified data was processed using tximport (26925227), and the differential expression 441 analysis was done using DESeq2 (25516281). Genes with less than 5 reads in all biological 442 replicates of one condition were filtered out before the analysis. Two groups of differentially 443 expressed genes were defined - a relaxed set containing genes with an absolute log2 fold change of 444 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2026.01.01.697307doi: bioRxiv preprint Page 14 of 27 0.5, and a stringent set containing games with an absolute log2 fold change of 1. The fold change 445 was deemed significant if the adjusted p-value was less than 0.05 (Benjamini-Hochberg corrected). 446 For bulk RNA -Seq, r eads were preprocessed with cutadapt 4.2 to remove adaptor sequences, 447 poly(A) tails, and low-quality bases. Reads shorter than 40 bp were discarded. The remaining reads 448 were mapped to Mus musculus rRNA sequences using bowtie2 2.3.5 (32) and reads aligning to 449 these sequences were removed. The filtered reads were then aligned to the Mus musculus reference 450 genome (GRCm39 assembly) using STAR 2.7.10b (33). Gene-level quantification was performed 451 with HTSeq-count 1.99.2 (34) in “union” mode, using Ensembl 111 annotations. Differential gene 452 expression analysis was carried out with the DESeq2 1.34.0 (35) R/Bioconductor package using 453 default parameters. P -values were adjusted for multiple testing with the Benjamini -Hochberg 454 method. 455 Gene ontology and pathway reactome analyses were performed using 456 ‘https://maayanlab.cloud/Enrichr/’ website. 457 458 Analyses of snATAC-Seq dataset and motif enrichment 459 Single-nucleus ATAC-seq (snATAC-seq) libraries were processed using Signac v1.10.0 and Seurat 460 v4.3.0 in R. Raw peak files were imported and converted to GRanges ob jects for genomic 461 coordinate standardization. Cell -type-specific open chromatin peaks were identified for 462 myotendinous junction (MTJ) nucleus through Seurat object integration with parallel 463 transcriptomic data (GEX), utilizing clustering and differential a ccessibility analysis to isolate 464 peaks enriched in MTJ -specific clusters. From the total accessible chromatin landscape, 465 myotendinous junction-specific peaks were extracted by filtering for regions showing significant 466 accessibility exclusively in MTJ nucle us relative to other muscle nucleus populations. Peak 467 coordinates were standardized using StringToGRanges coordinate conversion (mm10 genome 468 build). Selected peaks were then used as input for downstream transcription factor binding site 469 discovery via motif scanning with JASPAR 2020 vertebrate PWMs. 470 Transcription factor binding site (TFBS) enrichment analysis was performed on MTJ-specific open 471 chromatin regions using a computational motif -scanning approach. MTJ -specific accessible 472 chromatin peaks were identi fied from ATAC -seq data and used as input for systematic motif 473 discovery. Vertebrate transcription factor position weight matrices (PWMs) were obtained from the 474 JASPAR 2020 database using TFBSTools (v1.38.0). All vertebrate PWMs with non -redundant 475 versions were retrieved, encompassing 746 motif models representing diverse transcription factor 476 families. Genomic sequences corresponding to each peak region were extracted from the mouse 477

genome (mm10) using BSgenome.Mmusculus.UCSC.mm10 (v1.4.3). Motif scanning 478 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2026.01.01.697307doi: bioRxiv preprint Page 15 of 27 was performed with motifmatchr (v1.22.0) using the matchMotifs() function, which implements a 479 log-odds scoring algorithm based on PWM models. For each peak -motif pair, a match score was 480 computed by sliding the PWM across both DNA strands. Binding sites exceeding the default log-481 odds threshold (typically p < 10⁻⁴) were classified as putative TFBS occurrences. 482

were integrated into a Seurat object (v4.3.0) using the Signac framework (v1.10.0) for 483 chromatin accessibility data. A binary motif occurrence matrix was constructed with dimensions n 484 peaks × m motifs, where each entry indicates presence (1) or absence (0) of a motif within a given 485 peak. This matrix was stored as a ChromatinAssay object within the Seurat framework, enabling 486 downstream integration with single-cell chromatin accessibility profiles when applicable. 487 Global motif enrichment was calculated as the proportion of peaks containing each motif: 488 Enrichment (%) = (Number of peaks with motif / Total peaks) × 100 489 Motifs were ranked by fr equency of occurrence across all analyzed peaks compared to random 490 occurences. The top-ranking motifs representing the most enriched transcription factor binding sites 491 in MTJ -specific open chromatin were identified for downstream validation and functional 492 interpretation. 493 MTJ-specific accessible chromatin peaks were functionally annotated using ChIPseeker v1.36.0 494 with the mouse genome annotation database (TxDb.Mmusculus.UCSC.mm10.knownGene and 495 org.Mm.eg.db). Peak coordinates were mapped to their nearest geno mic features, including 496 promoters (±3 kb from transcription start sites), gene bodies, introns, and intergenic regions, using 497 annotatePeak(). The distribution of peaks across genomic features was visualized with annotation 498 pie charts to assess regulatory l andscape composition. Gene symbols associated with each peak 499 were extracted, and genes with multiple associated peaks (≥150 peaks per gene) were identified as 500 high-confidence MTJ regulatory targets. This threshold -based filtering enriched for genes with 501 extensive regulatory architecture characteristic of master regulators and structural genes defining 502 myotendinous junction identity. 503 All analyses were performed in R (v4.3.1). The complete analysis pipeline, including motif 504 occurrence matrices and enrichment statistics, was saved for reproducibility and downstream 505 chromVAR-based activity inference. 506 507 Histology and imaging 508 For histological analysis, samples were rapidly embedded in Cryomatrix (Epredia) in cryomolds 509 then snap-frozen in isopentane pre-cooled in liquid nitrogen for approximately 15 seconds until the 510 cryomatrix solidified completely. Samples embedded in cryomatrix were equilibrated to cryostat 511 temperature (–20 °C) prior to sectioning. Transverse cryosections were cut at 10 μm thickness using 512 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2026.01.01.697307doi: bioRxiv preprint Page 16 of 27 a Leica CM3050S cryostat (Leica Microsystems, Germany) and collected on Superfrost Plus 513 adhesion microscope slides (Epredia). Sections were then stored at –20 °C for 30 mins then stored 514 at -80 °C until further use. 515 For RNAscope, the tissue sections were fixed with 4% PFA in PBS for 15 minutes at 4  °C. After 516 two times washing with PBS, the sections were serially dehydrated through increasing ethanol 517 (50%, 70%, 100%) 5 minutes each. Subsequently, RNAscope was performed according to 518 manufacturer’s guideline using the RNAscope Multiplex Fluorescent Reagent Kit v2 (Bio-Techne 519 323100). Proteinase IV was used for our procedure. The following RNAscope probes were used 520 for our study: Col22a1 (590911-C2), Tigd4 (598761), Ttn (483031), GFP (409011) and Engrailed 521 (newly designed for this study). After all RNAscope procedures, samples were counterstained with 522 DAPI and mounted using Prolong Gold Antifade (Thermofisher scientific). When combined with 523 antibody staining, tissue sections were processed to blocking buffer (see below) after the last wash 524 and proceeded to regular immunohistochemistry. 525 For immunofluorescence staining, transverse cryosections were fixed in 4% paraformaldehyde 526 (PFA) for 10 minutes at room temperature (RT), followed by three washes in PBT (PBS containing 527 0.1% Tween-20) for 5 minutes each. Permeabilization was performed in PBX (PBS containing 528 0.5% Triton X -100) for 6 minutes, after which sections were rinsed in PBS for 5 minutes. Non-529 specific binding was blocked by incubating the sections in blocking buffer (PB S containing 5% 530 BSA, 3% horse serum and 0.1% Triton X-100) for 1 hour at RT. Sections were then incubated with 531 primary antibody diluted in blocking buffer overnight at 4 °C in a humidified chamber. The 532 following day, slides were washed three times in PBT f or 10 minutes each, then incubated with 533 fluorophore-conjugated secondary antibody diluted in blocking buffer for 1 hour at RT. DAPI (1 534 μg/mL) was included in the secondary incubation step for nuclear counterstaining. Finally, sections 535 were washed three tim es in PBT for 10 minutes each in the dark, mounted with ProLong Gold 536 Antifade, and stored at 4 °C until imaging. The following antibodies were used in this study: GFP 537 (Aves Labs; 1:500), Klf2 (Cell Signaling, 1:500), Dystrophin (Abcam, 1:250), and Col22a1 (Gift 538 from Manuel Koch, 1:1000). 539 Images were acquired on a Leica confocal at the IGBMC imaging core facility. 20× oil-immersion 540

was used. Fluorescent signals were sequentially detected using appropriate laser lines and 541 emission filters for DAPI, Opal 520, Opal 570 and Opal 690 (Opal chemicals were from Akoya 542 Biosciences) to avoid channel bleed -through. Laser power, gain, and offset were optimized and 543 maintained constant across samples. Imaging was performed using the tile scan mode to capture 544 large tissue areas, with a frame size of 520 × 520 pixels and processed using imageJ. 545 546 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2026.01.01.697307doi: bioRxiv preprint Page 17 of 27 Western blotting 547 TA tissues were lysed as described in the ‘Ribosome immunoprecipitation and sequencing’ section. 548 Lysates were denatured by adding Laemmli buffer and boiling for 10 minutes. Denatured samples 549 were separated by SDS -PAGE (Bio -Rad) and transferred into nitrocellulose membrane 550 (Amarsham). Transferred membranes were blocked for one hour with 5% skim milk in TBS plus 551 0.1% Tween-20 (TBST) at RT. Afterwards, primary antibodies were incubated in 5% BSA in TBST 552 supplemented with 0.1% sodium azide overnight in cold room with gentle rocking. After three 553 times of washing with TBST (10 minutes each), membranes were incubated with secondary 554 antibodies (anti-mouse or anti -rabbit HRP; Cell Signaling) diluted in skim milk (1:5000) for one 555 hour in RT. After three times washing with TBST, membranes were developed using 556 chemiluminescence (ECL substrate; Pierce) and Amarsham ImageQuant 800. For Streptavidin-557 HRP, antibody was incubated for one hour in skim milk after blocking. 558 The following antibodies were used in this study: Col22a1 (Abcam; 1:500), β-actin (Cell Signaling; 559 1:1000), Klf2 (Cell Signaling, 1:500), HA (Covance, 1:2000), and Streptavidin -HRP (Sigma, 560 1:5000). 561 562 Statistical analysis 563 Graph generation and statistical analyses were performed using GraphPad Prism as described in 564 each figure legend. 565 566

567 1. L. A. Tintignac, H. R. Brenner, M. A. Ruegg, Mechanisms Regulating Neuromuscular 568 Junction Development and Function and Causes of Muscle Wasting. Physiol Rev 95, 809-569 852 (2015). 570 2. L. Li, W. C. Xiong, L. Mei, Neuromuscular Junction Formation, Aging, and Disorders. 571 Annu Rev Physiol 80, 159-188 (2018). 572 3. B. Charvet, F. Ruggiero, D. Le Guellec, The development of the myotendinous junction. 573 A review. Muscles Ligaments Tendons J 2, 53-63 (2012). 574 4. A. L. Mackey, The Myotendinous Junction-Form and Function. Cold Spring Harb 575 Perspect Biol, (2024). 576 5. U. Mayer et al., Absence of integrin alpha 7 causes a novel form of muscular dystrophy. 577 Nat Genet 17, 318-323 (1997). 578 6. M. Kim et al., Single-nucleus transcriptomics reveals functional compartmentalization in 579 syncytial skeletal muscle cells. Nat Commun 11, 6375 (2020). 580 7. M. J. Petrany et al., Single-nucleus RNA-seq identifies transcriptional heterogeneity in 581 multinucleated skeletal myofibers. Nat Commun 11, 6374 (2020). 582 8. M. Dos Santos et al., Single-nucleus RNA-seq and FISH identify coordinated 583 transcriptional activity in mammalian myofibers. Nat Commun 11, 5102 (2020). 584 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2026.01.01.697307doi: bioRxiv preprint Page 18 of 27 9. F. Chemello et al., Degenerative and regenerative pathways underlying Duchenne 585 muscular dystrophy revealed by single-nucleus RNA sequencing. Proc Natl Acad Sci U S 586 A 117, 29691-29701 (2020). 587 10. H. Li et al., A male germ-cell-specific ribosome controls male fertility. Nature 612, 725-588 731 (2022). 589 11. D. C. Rodrigues et al., Shifts in Ribosome Engagement Impact Key Gene Sets in 590 Neurodevelopment and Ubiquitination in Rett Syndrome. Cell Rep 30, 4179-4196 e4111 591 (2020). 592 12. M. L. Truitt, D. Ruggero, New frontiers in translational control of the cancer genome. Nat 593 Rev Cancer 16, 288-304 (2016). 594 13. N. R. Genuth, M. Barna, The Discovery of Ribosome Heterogeneity and Its Implications 595 for Gene Regulation and Organismal Life. Mol Cell 71, 364-374 (2018). 596 14. A. Dopler et al., P-stalk ribosomes act as master regulators of cytokine-mediated 597 processes. Cell. 598 15. K. Ramachandran et al., Transcriptional programming of translation by BCL6 controls 599 skeletal muscle proteostasis. Nat Metab 6, 304-322 (2024). 600 16. R. Tharakan, C. Ubaida-Mohien, Y. Piao, M. Gorospe, L. Ferrucci, Ribosome profiling 601 analysis of human skeletal muscle identifies reduced translation of mitochondrial proteins 602 with age. RNA Biol 18, 1555-1559 (2021). 603 17. M. Cui et al., The rRNA epitranscriptome and myonuclear SNORD landscape in skeletal 604 muscle fibers contributes to ribosome heterogeneity and is altered by a hypertrophic 605 stimulus. Am J Physiol Cell Physiol 327, C516-C524 (2024). 606 18. D. Curzi et al., Effect of Different Exercise Intensities on the Myotendinous Junction 607 Plasticity. PLoS One 11, e0158059 (2016). 608 19. K. B. Nielsen, N. N. Lal, P. W. Sheard, Age-related remodelling of the myotendinous 609 junction in the mouse soleus muscle. Exp Gerontol 104, 52-59 (2018). 610 20. Y. Wen et al., Myonuclear transcriptional dynamics in response to exercise following 611 satellite cell depletion. iScience 24, 102838 (2021). 612 21. M. Tabebordbar et al., Directed evolution of a family of AAV capsid variants enabling 613 potent muscle-directed gene delivery across species. Cell 184, 4919-+ (2021). 614 22. E. Sanz et al., Cell-type-specific isolation of ribosome-associated mRNA from complex 615 tissues. P Natl Acad Sci USA 106, 13939-13944 (2009). 616 23. P. Zhou et al., Interrogating translational efficiency and lineage-specific transcriptomes 617 using ribosome affinity purification. Proc Natl Acad Sci U S A 110, 15395-15400 (2013). 618 24. D. Curzi et al., Effect of Different Exercise Intensities on the Myotendinous Junction 619 Plasticity. Plos One 11, (2016). 620 25. S. Kuersten, E. B. Goodwin, The power of the 3' UTR: translational control and 621 development. Nat Rev Genet 4, 626-637 (2003). 622 26. K. Leppek, R. Das, M. Barna, Functional 5' UTR mRNA structures in eukaryotic 623 translation regulation and how to find them. Nat Rev Mol Cell Biol 19, 158-174 (2018). 624 27. Y. Katoh, M. Katoh, Identification and characterization of ARHGAP27 gene in silico. Int 625 J Mol Med 14, 943-947 (2004). 626 28. J. L. Tomsig, C. E. Creutz, Copines:: a ubiquitous family of Ca2+-627 dependent phospholipid-binding proteins. Cellular and Molecular Life Sciences 59, 1467-628 1477 (2002). 629 29. T. C. Branon et al., Efficient proximity labeling in living cells and organisms with 630 TurboID. Nat Biotechnol 36, 880-887 (2018). 631 30. M. Wang et al., Single-nucleus multi-omic profiling of human placental 632 syncytiotrophoblasts identifies cellular trajectories during pregnancy. Nat Genet 56, 294-633 305 (2024). 634 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2026.01.01.697307doi: bioRxiv preprint Page 19 of 27 31. J. Ji, E. Lefebvre, J. Laporte, Comparative in vivo characterization of newly discovered 635 myotropic adeno-associated vectors. Skelet Muscle 14, 9 (2024). 636 32. B. Langmead, S. L. Salzberg, Fast gapped-read alignment with Bowtie 2. Nat Methods 9, 637 357-359 (2012). 638 33. A. Dobin et al., STAR: ultrafast universal RNA-seq aligner. Bioinformatics 29, 15-21 639 (2013). 640 34. S. Anders, P. T. Pyl, W. Huber, HTSeq--a Python framework to work with high-641 throughput sequencing data. Bioinformatics 31, 166-169 (2015). 642 35. M. I. Love, W. Huber, S. Anders, Moderated estimation of fold change and dispersion for 643 RNA-seq data with DESeq2. Genome Biol 15, 550 (2014). 644 645 Acknowledgments 646 647 Funding: We thank grant supports to M.K. from the 648 European Research Council ERC-StG 101039531 (M.K) 649 ANR-22-CE13-0023-03 MYODOM (M.K) 650 ANR t-ERC StG 2021 (M.K) 651 LABEX INTR (M.K) 652 University of Strasbourg IDEX Attractivitae (M.K) 653 INSERM ATIP-Avenir (M.K) 654 AFM-Telethon Trampoline 24287 and n°22AA003-00 (M.K) 655 Grand Est PhD fellowship (N.E.K) 656 Fondation pour la Recherche Médicale postdoctoral fellowship (S.M) 657 AFM n°28842 (P.M) 658 ANR-21-CE14-0042-01 MOTOMYO (P.M) 659 660 Author contributions: 661 Conceptualization: M.K 662 Main Experiments: J.N and N.E.K 663 Other Experiments: C.S, S.M, and L.Y 664 Bioinformatic analyses: V.F and E.J 665 Supervision: P.M and A.A 666 Methodology (AAV generation): E.L 667 Writing: M.K 668 669 Competing interests: 670 All other authors declare they have no competing interests. 671 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2026.01.01.697307doi: bioRxiv preprint Page 20 of 27 672 Data and materials availability: 673 All data and materials used for this study are available from the corresponding author upon 674 reasonable request. All NGS datasets have been deposited to EBI (RNA -Seq: E-MTAB-16305; 675 RiboTag-Seq under sedentary: E -MTAB-16304; RiboTag -Seq after exercise: E -MTAB-16306). 676 The codes used in this study are available in Github server 677 (https://github.com/BIMSBbioinfo/MTJ_Figures). All data are available in the main text or the 678 supplementary materials. 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2026.01.01.697307doi: bioRxiv preprint Page 21 of 27 Figures and Tables 718 719 720 721 722 Figure 1. Development of MTJ-AAV 723 (A) Schematic of AAV constructs driven by CK8 or Tigd4 promoters and the expected expression 724 pattern of the 4×NLS-sfGFP reporter. 725 (B) Tile-scan images of tibialis anterior (TA) muscles injected with the AAVs shown in (A). The 726 right panel shows a magnified view of the boxed region. 727 (C) Combined GFP immunostaining and RNAscope detection of Col22a1 and Tigd4 transcripts. 728 Arrows indicate nuclei co-expressing all three markers. 729 (D) Quantification of labeling efficiency and specificity from experiments in (C) (n = 5). 730 (E) Representative FACS plots of CK8- and Tigd4-NLS-sfGFP injected TA muscles. GFP+ nuclei 731 were isolated and subjected to bulk RNA-Seq. 732 (F) Heatmap showing genes enriched or depleted in GFP + nuclei labeled by the Tigd4 promoter 733 relative to the CK8 promoter. Selected MTJ marker genes are indicated. 734 Scale bars, 100 µm. 735 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2026.01.01.697307doi: bioRxiv preprint Page 22 of 27 736 737 738 Figure 2. MTJ-AAV enables Cre-mediated recombination specifically at the MTJ 739 (A) Schematic of the AAV construct in which t he Tigd4 promoter drives expression of Cre -NLS. 740 When injected into Rosa26-LSL-H2B-GFP reporter mice, Cre activity induces H2B-GFP labeling 741 of MTJ nuclei. 742 (B) Tile-scan images of TA muscles injected with MTJ -AAV-Cre-NLS. Reporter mice injected 743 with an empt y MyoAAV4a vector (containing only the Tigd4 promoter without Cre) showed no 744 H2B-GFP expression. 745 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2026.01.01.697307doi: bioRxiv preprint Page 23 of 27 (C) Quantification of labeling efficiency and specificity from experiments in (B) (n = 4). 746 (D) Schematic of systemic delivery of MTJ -AAV-Cre-NLS into Rosa26-LSL-H2B-GFP reporter 747 mice via intra-orbital injection. 748 (E) Representative images showing MTJ -specific myonuclear labeling in the gastrocnemius and 749 diaphragm muscles. Similar results were obtained in three independent animals. 750 Scale bars, 100 µm. 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2026.01.01.697307doi: bioRxiv preprint Page 24 of 27 796 797 798 Figure 3. MTJ-AAV reveals a specialized translatome of MTJ 799 (A) Experimental workflow for MTJ -targeted ribosome labeling and translatome profiling in 800 sedentary and exercised muscles. 801 (B) HA immunohistochemistry confirming ribosome labeling along the MTJ. The right panel shows 802 a magnified view of the boxed region in the middle image. Scale bar, 100 µm. 803 (C) Bioanalyzer profile of RNA recovered after HA immunoprecipitation. 804 (D) Heatmap showing representative MTJ -ribosome-enriched transcripts under sedentary 805 conditions. Known transcriptional MTJ markers are indicated in red. 806 (E) Volcano plot comparing CK8- and MTJ-ribosome associated transcripts in sedentary muscles. 807 (F) Gene Ontology analysis of translationally enriched MTJ targets. 808 (G) Volcano plot comparing CK8- and MTJ-ribosome associated transcripts following endurance 809 exercise. 810 (H) Exercise amplifies the degree of translational enrichment for MTJ -specific targets relative to 811 sedentary conditions. 812 813 814 815 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2026.01.01.697307doi: bioRxiv preprint Page 25 of 27 816 817 818 819 Figure 4. UTRs of translational target transcripts enable localized protein synthesis at the 820 MTJ 821 (A) Schematic representation of MTJ translational reporter constructs. 822 (B) Single extensor digitorum longus (EDL) myofibers were isolated one month after intramuscular 823 injection of the AAVs shown in (A) and analyzed by epifluorescence microscopy. 824 (C) EDL myofibers from (B) were fixed and subjected to RNAscope detection of GFP mRNA to 825 assess transcript distribution. 826 Scale bars, 100 µm. 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2026.01.01.697307doi: bioRxiv preprint Page 26 of 27 856 857 858 859 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2026.01.01.697307doi: bioRxiv preprint Page 27 of 27 Figure 5. KLF transcription factors contribute to the MTJ transcriptome 860 (A) Classification of MTJ -specific ATAC -seq peaks according to their genomic locations and 861 sequence elements. 862 (B) De novo motif analysis of MTJ-specific ATAC-seq peaks identifying KLF family transcription 863 factors as top candidates. 864 (C) Schematic of KLF inhibition strategy using a dominant -negative (DN) construct expressed at 865 the MTJ. 866 (D) Expression of KLF2 -DN at the MTJ suppresses MTJ marker genes Tigd4 and Col22a1, as 867 determined by RNAscope. KLF2 -DN expression was detected via RNAscope targeting the 868 Engrailed (En) sequence. The result was reproduced in three independent animals. 869 (E) Model summarizing MTJ -specific translational regulation and its impact on local 870 transcriptional control through KLF factors. Local signals, such as mechanical or tendon -derived 871 cues, activate this translational program to reinforce domain specialization. 872 Scale bars, 100 µm. 873 874 875 876 .CC-BY 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted January 2, 2026. ; https://doi.org/10.64898/2026.01.01.697307doi: bioRxiv preprint