A human-like mouse model of Duchenne muscular dystrophy to help develop genetic treatments

Although it was previously reported that the parental hDMD (hDMD/mdx) mouse generated by 't Hoen and collaborators ('t Hoen et al., 2008; Yavas et al., 2020) contains a tail-to-tail duplication event (Yavas et al., 2020), this duplication event has not yet been evaluated in the hDMDΔ52/mdx model developed by Pickar-Oliver et al. (2021). Towards this end, we designed digital droplet PCR (ddPCR) probes spanning from the entire DMD locus, including 2 kb of sequence upstream and downstream of the annotated DMD gene (Fig. 1A). ddPCR definitively detected two copies in heterozygous hDMD/mdx and hDMDΔ52/mdx mice, and four copies in homozygous hDMD/mdx and hDMDΔ52/mdx mice, suggesting a complete duplication (Fig. 1B).

To further characterize the genomic composition of the hDMD/mdx control line ('t Hoen et al., 2008) and the hDMDΔ52/mdx disease line (Pickar-Oliver et al., 2021), we performed long-read whole-genome sequencing (WGS) (Table S1). A copy number variation analysis confirmed that there is complete duplication of the entire DMD locus, including 2 kb of sequence upstream and downstream of the annotated DMD sequence in the hDMD/mdx parental line and that the duplication is maintained in the daughter hDMDΔ52/mdx line generated by Pickar-Oliver et al. (2021) (Figs S1 and S2).

Insertion of the DMD transgene was originally reported to mouse chromosome 5 ('t Hoen et al., 2008). To similarly determine the DMD insertion location, we aligned the WGS data to the Hygromycin cassette sequence and yeast genome (as a proxy for the YAC sequence used to generate the hDMD transgenic mouse) and further used those reads to map against the mouse genome. This analysis pointed to regions with mapped reads common to the transgenic lines but absent in the control mdx line, with the most likely position for the insertion overlapping with the 5qG2 band towards the end of chromosome 5. We pinpointed the insertion around coordinates chr5:133,745,208-133,829,563, which closely agrees with the recent work of Chey et el. (2024) (Table S2).

Next, the Δ52 deletion was characterized using structural variation analysis. A unique 309 bp deletion event was detected spanning human chrX:31327616-31327924, which overlaps exon 52, as annotated for the main isoform expressed in muscle (NCBI transcript ID, NM_004006.3; ENSMBL transcript ID, ENST00000357033.9). The observation of a unique deletion event suggests that the identical edit was made on both copies during the creation of this mouse (Fig. S3). To further understand the region in which the deletion occurred, we generated a de novo assembly from the WGS data. We recovered one single contig containing the full DMD region, including at least 2 kb upstream and downstream of the genic region, from each of the hDMD/mdx and hDMDΔ52/mdx lines and none from the control mdx line (Fig. S4). A search of the sequences corresponding to the 79 annotated exons for the main isoform expressed in muscle (i.e. Dp427m) revealed that exon 52 and flanking regions in the adjacent introns are present in the assembled loci for the parent hDMD/mdx line (Fig. S4A) and absent from the hDMDΔ52/mdx line, as expected (Fig. S4B,C). Follow-up Sanger sequencing revealed a 16-bp insertion in the intronic region between the guide RNA (gRNA) cut sites (Fig. 1C), consistent with a previous report (Robinson-Hamm, 2018). The same insertion was confirmed in the de novo assembled DMD loci for each of the hDMDΔ52/mdx samples (Fig. S3B,C). An additional search of the 16-bp insertion sequence directly in the reads that map to the DMD allele showed similar counts to the neighbor exons in the hDMDΔ52/mdx mice and zero count in the hDMD/mdx mice, suggesting that this same 16-bp insertion can be found on both copies of the hDMDΔ52 allele (Table S3).

Next, the hDMDΔ52/mdx mice were assessed for histological outcomes across skeletal muscles and heart, which are tissues known to have clinical manifestations in patients with DMD (Deconinck and Dan, 2007). Because DMD primarily affects young boys, all animals assessed were male to eliminate any impact of sex on the outcomes. At all ages assessed, Masson's trichrome staining showed increased fibrosis in the diaphragm of hDMDΔ52/mdx mice compared with that in hDMD/mdx mice (Fig. 2A,B). However, minimal fibrosis was apparent in the heart at later timepoints (Fig. S5). Fiber diameter in the TA muscle of hDMDΔ52/mdx mice varied considerably, with hDMDΔ52/mdx mice showing a broader distribution of fiber diameters and reduced mean fiber diameter at older ages compared to hDMD/mdx controls at similar ages (16 weeks, 39.8 vs 42.7 µm; 20 weeks, 41.2 vs 41.6 µm; and 52 weeks, 34.1 vs 37.4 µm, respectively) (Fig. 2C,E). The same trends were observed in the gastrocnemius (GAS) and triceps brachii (TRI) muscles (Fig. S6A-C). Furthermore, indicative of the high turnover rate of dystrophic muscle fibers, hDMDΔ52/mdx mice had increased central nuclei at all ages in several muscles, including the TA (4 weeks, 53.1% vs 2.4%; 16 weeks, 82.5% vs 3.6%; 20 weeks, 82.4% vs 4.7%; 52 weeks, 63.4% vs 6.5% at 52 weeks) (Fig. 2C,D), GAS (4 weeks, 39.8% vs 1.3%; 16 weeks, 74.1% vs 2.2%; 20 weeks, 70.6% vs 6.6%; 52 weeks, 57.6% vs 12.9%) and TRI (4 weeks, 56.4% vs 3.3%; 16 weeks, 76.6% vs 3.3%; 20 weeks, 69.7% vs 6.8%; 52 weeks, 70.0% vs 14.1%) (Fig. S6A,D,E), compared to those in the hDMD/mdx control strain. This finding is consistent with numerous other mouse DMD models (Yucel et al., 2018), including reports of 50% central nuclei in other humanized mouse models of DMD (Veltrop et al., 2018) and up to 80% central nuclei in mdx mice (Coulton et al., 1988). The hDMDΔ52/mdx strain described here shows a range of central nuclei from 39.8% to 82.5%, depending on the age and tissue examined.

The phenotypic hallmark of DMD is muscle damage and weakness due to contraction (Deconinck and Dan, 2007). As shown in Fig. 3A, the specific force (normalized to cross-sectional area of the tissue) of the TA was reduced in hDMDΔ52/mdx mice compared to that in age-matched hDMD/mdx control animals at 4 weeks (22% reduction), 16 weeks (27% reduction), 20 weeks (33% reduction) and 52 weeks (41% reduction) of age. This is consistent with the reported reduction of 29% in TA specific force in mdx mice compared to that in wild-type controls (Sharp et al., 2011). Resistance to damage by eccentric contractions was also reduced in the TA of hDMDΔ52/mdx mice starting at 16 weeks of age (Fig. 3B,C), consistent with only an average 25% maintenance of maximal force observed in mdx mice (Sharp et al., 2011). hDMDΔ52/mdx mice also display cardiac rhythm defects as detected by ECG under anesthesia (Fig. 4). Many DMD mouse models have not recapitulated cardiac defects that are associated with human disease, whereas hDMDΔ52/mdx mice displayed an increased QRS interval (Fig. 4), which does mimic human disease (van Putten et al., 2020). Other aspects of cardiac rhythm were unaffected (Fig. S7). Mild deficits in grip strength and rotarod were observed at select timepoints (Fig. S8).

Serum biomarkers for skeletal muscle and cardiac injury were assessed at various ages in both hDMD/mdx and hDMDΔ52/mdx animals through in-life serum collections prior to assessment of functional or physiological outcomes. Markers of skeletal muscle injury include skeletal troponin I (sTnI; also known as TNNI1/2), myosin light chain 3 (Myl3), fatty acid binding protein 3 (FABP3), muscle-type creatine kinase (CKM) and matrix metalloproteinase 9 (MMP-9). sTnI and Myl3 are myofibrillar proteins involved in the contraction of muscles (Schiaffino and Reggiani, 1996). The release of myofibrillar proteins into the circulation increases in response to muscle damage; thus, these biomarkers are expected to increase in response to muscle damage in models of muscular dystrophy (Burch et al., 2015; Pavis et al., 2021). As expected, serum levels of Myl3 and sTnI were elevated in hDMDΔ52/mdx animals at most ages (Fig. 5A,B). FABP3 and CKM are secreted by muscle fibers into the blood upon muscle damage (Burch et al., 2015), particularly in cases of muscular dystrophy. Levels of FABP3 were elevated in some hDMDΔ52/mdx mice at 4 and 16 weeks of age (Fig. S9A). Levels of CKM were significantly elevated in hDMDΔ52/mdx mice, compared to those in hDMD/mdx mice, at all ages (Fig. 5E), consistent with reports of elevated serum FABP3 and CKM levels in mdx mice (Glesby et al., 1988).

Cardiac biomarkers included cardiac troponin I (cTnI; also known as TNNI3), tissue inhibitor of metalloproteinase 1 (TIMP-1), FABP3 and MMP-9 (Niebroj-Dobosz et al., 2015). Serum levels of TIMP-1 were elevated in hDMDΔ52/mdx mice compared to those in hDMD/mdx controls at all ages (Fig. 5C). cTnI is a biomarker of coronary artery disease. The levels of cTnI were below fit-curve range in the hDMD/mdx animals but were elevated in numerous hDMDΔ52/mdx animals at later time points, as shown in Fig. 5D, indicating a notable difference between the two strains of mice. MMP-9 is secreted from both skeletal muscle and cardiac muscle into the extracellular matrix, where it exacerbates skeletal muscle and cardiac muscle phenotypes in DMD by contributing to increased inflammation and muscle damage and reduced skeletal and cardiac muscle function (Kherif et al., 1999; Li et al., 2009; Ogura et al., 2014). No differences in MMP-9 serum levels were detected between the two strains (Fig. S9B). However, it is possible that our hDMD/mdx mice retain higher levels of MMP-9 than wild-type animals owing to the mdx background of the strain; thus, differences between hDMDΔ52/mdx and hDMD/mdx mice could not be observed (Dahiya et al., 2011a,b; Kherif et al., 1999; Li et al., 2009).

To further support the use of this model for the development of human-targeting therapies, we next conducted a proof-of-concept study to assess whether hDMDΔ52/mdx deficits could be rescued through therapeutic intervention using a human-targeting CRISPR/Cas9-based gene-editing strategy to delete exons 45-55 and reframe the DMD gene. This is a different therapeutic strategy from that previously reported by Pickar-Oliver et al. (2021), which used a CRISPR/Cas9 homology-independent targeted integration method to integrate an exon 52 copy into the hDMDΔ52/mdx mouse. Here, we utilized a CRISPR/Cas9-based gene-editing strategy with two gRNAs targeting introns 44 and 55 to delete exons 45-55 and thereby produce an internally shortened, functional dystrophin protein (Gersbach et al., 2022↗). These two guides were packaged into AAVrh74 with a Staphylococcus aureus Cas9 coding sequence. Systemic administration resulted in appreciable levels of edited DMD transcripts in the heart and TA as measured by ddPCR (Fig. 6A,B). These edited transcripts led to 14.8% dystrophin protein restoration in the heart (Fig. 6C; Fig. S10B) and 2% dystrophin restoration in the TA (Fig. 6D; Fig. S10D) as quantified by Jess Automated Western Blot. Mice treated with gene-editing therapeutics also showed functional restoration of resistance to injury in the TA muscle at 12 weeks post-treatment, as shown by significant improvement in the retention of muscle force eccentric muscle contractions (Fig. 6E-G).

hDMDΔ52/mdx and hDMD/mdx transgenic mice were obtained from Duke University. The animals were generated as previously described (Pickar-Oliver et al., 2021). The animal studies herein were conducted with adherence to the guidelines for the care and use of laboratory animals of the National Institutes of Health (NIH). All the experiments with animals were approved by the Institutional Animal Care and Use Committee (IACUC) at Sarepta Therapeutics, Inc. hDMD/mdx and hDMDΔ52/mdx mice heterozygous for the respective human DMD allele were assessed at various ages for histological pathology via Masson's trichrome and Hematoxylin and Eosin (H&E) staining, cardiac and skeletal injury via serum biomarkers, as well as performance in grip strength and rotarod, in situ physiology of the TA, and ECG (Table S4).

WGS was performed using high-throughput long-read sequencing of the hDMDΔ52/mdx line. For a comprehensive characterization, we included two samples of the line of interest (i.e. one sample from the hDMDΔ52/mdx heterozygous and one sample from hDMDΔ52/mdx homozygous for the respective DMD allele). To perform comparisons, we included a sample of the hDMD/mdx parental line as heterozygous for the respective DMD allele, and a sample of the C57BL/10-mdx (The Jackson Laboratory strain #001801↗) as a representative of the background, in which all the lines are maintained (Table S1).

High-molecular mass genomic DNA was obtained using a Genomic-tip 100/G kit (Qiagen, 10243). Briefly, 80 mg of frozen brain and muscle tissue from each of the mice was processed. High-molecular mass DNA samples were submitted to Azenta Life Sciences to perform PacBio library preparation and continuous long-read sequencing. The PacBio library was prepared using a SMRTbell prep kit as per the manufacturer's protocol. SMRTbell libraries were then sequenced on the PacBio Sequel II platform with v3.0 chemistry. After quality control of PacBio raw reads [LongQC v1.2.1 (Fukasawa et al., 2020)] (Table S1), PacBio raw reads were used (1) to contrast against a reference genome that included the full genome sequence of the Mus musculus strain C57BL/GJ (version GRCm39) plus appended genomic sequence corresponding to a fragment of the primary Homo sapiens genome (version GRCh38.p14) spanning the region chrX:31117222-33341388(−), which includes the annotated hDMD gene ±2 kb of the flanking region (NCBI gene ID, 1756; ENSMBL gene ID, ENSG00000198947); and (2) to obtain a reference-free reconstruction of the DMD gene by building independent de novo genome assemblies from each sample using hifiasm [v0.16 (Cheng et al., 2022)] and further comparing the reconstructed sequence against the already-described DMD gene extracted from the human reference genome to determine which contigs contained the assembled sequence.

The full set of PacBio reads was used for the alignment against the reference genome using minimap2 with option '-ax map-hifi' [v2.28 (Li, 2021)], and the resulting alignment file was used as input for calling of genomic variation. Because variants of different sizes represent different challenges, we used (1) Sniffles2 [v2.6.0 (Smolka et al., 2024)] for the characterization of medium (gain and loss of ≥50 bp) variation and (2) CNVkit [0.9.12 (Olshen et al., 2011; Talevich et al., 2016)] for the characterization of large structural (gain and loss of ≥1 kb to larger genomic regions) variation. Sniffles2 was used in the presence of the reference genome to allow for the output of deletions and run in two steps as indicated for multi-sample variant calling (i.e. including the individual variant calling and next pooling of the first pass of variants for the combined and final calling) to obtain a variant calling file. CNVkit was executed in two passes. First, we built a flat reference for the DMD transgene and a reference from the sample C57BL/10-mdx for the mouse genome. Second, we used the '-method wgs' against the concatenate references to allow for read counts to be normalized by correcting for GC content and repetitive sequences, which resulted in an estimated depth from which log2 and copy ratios could be further calculated. The circular binary segmentation algorithm was employed to infer copy number segments. The threshold values for copy number 0 were selected as the minimum log2 from the regions in the sample used as representative of the genetic background, and the threshold value for 1 was selected as the log2 value for percentile 75 of the regions in the same sample. Values for copy gains were identical to the defaults for diploid samples as specified by CNVkit.

To determine the completeness of the DMD assembled loci, the exon sequences for the MANE transcript (NCBI RefSeq, NM_004006.3; ENSMBL transcript ID, ENST00000357033) were used to search the assembled contigs using Seqkit and the command 'locate -i -m 1 -p' [v2.3.0 (Shen et al., 2016)] to allow a maximum of one mismatch of difference between the sequences. To detect the raw PacBio reads containing the insertion generated during the generation of the disease line, we used Seqkit and the command 'locate -i -p' [v2.3.0 (Shen et al., 2016)] for no mismatches allowed.

File format manipulation, filtering and indexing of alignments was performed using Samtools [v1.2.0 (Danecek et al., 2021)] and Bedtools [v2.27.1 (Quinlan and Hall, 2010)]. File format manipulation of read files, such as conversion between fastq and fasta formats, was performed with Seqkit [v2.3.0 (Shen et al., 2016)]. Assembled contigs containing the DMD gene were visualized using SnapGene↗ (v 6.2.1). PacBio reads alignments to the reference genome were visualized using IGV [v2.11.3 (Robinson et al., 2017)]. Copy number variation results were visualized using custom scripts in python (v3.12).

Genomic DNA was isolated from C57BL/10SCSn-DMDmdx/J mice, as well as from heterozygous and homozygous hDMD/mdx and hDMDΔ52/mdx mice, using a NucleoMag^® Tissue kit (Macherey-Nagel, 744300). Copy number analysis by ddPCR was performed using the QXOne Droplet Digital PCR system (Bio-Rad). Assays targeting different regions of the DMD transgene were multiplexed with the Tfrc copy number assay (Thermo Fisher Scientific) using ddPCR Supermix for Probes (No dUTP) and SpeI restriction enzyme (2 U/reaction). The assay targeting exon 47 was used in singleplex reactions, with accompanying singleplex reactions for Tfrc detection on the same plate. Samples were normalized to 1 ng/µl in molecular water with 3 ng DNA analyzed per 20 µl reaction. Data were analyzed using QXOne Software, setting thresholds between positive and negative droplets. Copy numbers were determined by normalizing the DMD target assay concentration to the Tfrc concentration, with Tfrc representing two copies per genome. Samples were analyzed in duplicate. Sequences for primers and probes were as follows (forward/reverse/probe, respectively): intron 2, 5′-TTTAGCTGTGCAAATCTGGAAATA-3′/5′-TGCAGCTATAAACTCTGAGGACT-3′/56-FAM/AGGAGTAAG/ZEN/CAGCAGAAGATATGGC/3IABkFQ/; intron 20, 5′-TGTAAGACGTAGATACCCGGA-3′/5′-CCTCATGCTTACGCAAACCG-3′/56-FAM/CAATTA GCG/ZEN/CTGTGTAACTACGCCAAAT/3IABkFQ/; exon 77, 5′-AGCACAGGGTTAGAGGAGGT-3′/5′-ACTGCGTGTTGGCTTCCATA-3′/56-FAM/AAGCGAGTG/ZEN/GCCTGATCCCA/3IABkFQ/; exon 79 3′ UTR, 5′-TTGTTTTGCATCCTTTTGGCGTG-3′/5′-CAGTTCTCAAATGAGCAGTGTGT-3′/56-FAM/AAGACTTCC/ZEN/TCTACCACCACACCA/3IABkFQ/; promoter region, 5′-CCCGCCTTCTCTCTCAAGTT-3′/5′-TGTCCACTGTGCTATTCTGGTT-3′/56-FAM/CCAGCATGG/ZEN/CAAGCTCTGTGA/3IABkFQ/; and exon 47, 5′-AGCTCAAGCAGACAAATCTCCA-3′/5′-CCAAAGCAAACGGTCAGGTT-3′/5HEX/TCCCCGACC/ZEN/AATGAAGCACC/3IABkFQ/. The thermal cycling conditions followed the Bio-Rad recommended Supermix for Probes (No dUTP) cycling protocol, with an annealing temperature of 60°C.

Genomic DNA was isolated from hDMDΔ52/mdx heterozygous mice using the NucleoMag^® Tissue kit. PCR reactions were performed using NEBNext^® Ultra II Q5 Master Mix (New England Biolabs) and the following primers, according to the manufacturer's specifications: forward, 5′-CTGGAGGGAGGCAAAAAGGT-3′; reverse, 5′-AAGATGGCCCAGGAAGAAGC-3′. PCR products were purified using a NucleoSpin^® Gel and PCR Clean-up kit (Macherey-Nagel, 740609), and samples were submitted to GENEWIZ (Azenta Life Sciences) for Sanger sequencing.

Grip strength assessments were captured using DFE2 grip force meters (Columbus Instruments). Grip strength assessments were performed at 4, 16, 20, 24, 32 and 52 weeks of age. At each timepoint, forelimbs were assessed for visible sores or injuries prior to testing, and only those passing visual examination were tested. To obtain a grip force (g-force), rodents were allowed to grasp the pull bar with both forepaws before being pulled away from and parallel to the pull bar. This process was repeated until five grip forces for each rodent were obtained over a maximum interval of 2 min. All five grip forces were averaged to get the forelimb grip strength, which was then normalized to body weight (g).

Rotarod assessments were captured using a RotaMex-5 acquisition system (Columbus Instruments). Rotarod assessments were performed at 4, 16, 20, 24, 32 and 52 weeks of age. At each timepoint, forelimbs were assessed for visible sores or injuries, and only those passing visual examination were tested. Rodents were assessed four to six times per day, with a minimum 15-min break between each assessment, for 2 consecutive days. Rodents were placed on a rod moving at 4 rpm (one per lane). Once all rodents were placed, the assessment was started, and the rod speed increased by 1 rpm every 8 s until all rodents fell or 40 rpm was reached. Latency to fall was recorded in seconds, and the three highest runs between the 2 days were averaged together to obtain the average latency to fall.

In situ physiology assessments were performed as terminal procedures and captured using an Aurora Scientific muscle physiology acquisition system. Rodents were anesthetized with ketamine and xylazine. Once a sufficient plane of anesthesia was obtained, the TA tendon was exposed, and 4-0 silk sutures were used to secure the tendon to the lever arm of the force transducer. The knee and foot were secured to the platform under a heat lamp. Electrodes were then inserted into the sciatic nerve. With the stimulator set to 10 mN, an instant stimulation was performed to confirm proper electrode placement. After a 5-min rest, warm up protocols were run, and optimal baseline tension was determined. Force frequencies were run, followed by a 5-min wait prior to eccentric contractions being performed. Eccentric contractions performed a 10% muscle length pull to induce injury. Data were collected from both left and right TA muscles. Muscle length and weight were collected during necropsy and used to calculate the cross-sectional area, which was used for force normalization.

Subsurface ECG evaluations were performed at 16, 20, 24 and 52 weeks of age. In vivo surface ECGs were captured using a PowerLab acquisition system (ADInstruments). Animals were anesthetized using 1-2% isoflurane for maintenance with oxygen (1-2 l/min). Electrodes were placed sub-dermally in the corresponding limbs parallel to the limb, running from distal to proximal. The chest echo lead was placed parallel to the sternum in the direction of head to toe for the V1 position. For the V3 position, the needle was placed medially to the left midclavicular line. The V5 was placed at the anterior axillary line, caudal to the V3. A 1-min tracing was recorded in leads I and II. A 15- to 20-s tracing was recorded in leads III, aVR, aVL, aVF, V1, V3 and V5. Lead II data were analyzed using LabChart Software (ADInstruments). QTc was calculated using the Mitchell method (Mitchell et al., 1998).

At necropsy, muscles (TA, GAS, TRI, diaphragm and heart) were mounted onto wooden blocks with 7% Gum Tragacanth and fresh frozen in 2-methylbutane cooled in liquid nitrogen and stored at −80°C. Tissues were cryosectioned (10-µm sections) and stained with H&E. Centrally nucleated fibers and fiber diameters were quantified using Aperio ImageScope software (Leica Biosystems). Images were first annotated to select regions of interest and remove artifacts. The percentage of myofibers with central nuclei and fiber diameter were determined by analyzing 10-μm thick cryosections of H&E-stained TA, GAS and TRI muscles, as previously described (Rodino-Klapac et al., 2007). Analysis included four fields of 20× magnification per animal per muscle, and a minimum of 250 fibers (n=5-10).

Heart and diaphragm tissue sections (10-μm sections) were cryosectioned and allowed to air dry or were stored at −40°C until staining was performed. Frozen slides were allowed to warm to ambient temperature, then fixed in 10% neutral buffered formalin for 10 min. After incubation, slides were rinsed in deionized water up to four times, followed by fixation in 95% ethanol for 10 min. Slides were then air dried for at least 15 min and placed in Bouins fixative for ∼30 min. Slides were then washed in tap water until excess stain was no longer visible, followed by a deionized water rinse. Slides were placed in Weigert's Iron Hematoxylin Solution for 5 min. After washing in running water for 5 min, slides were rinsed in distilled water before placing in Biebrich Scarlet-Acid Fuchsin for 2 min, washed in distilled water, and incubated in phosphotungstic-phosphomolybdic acid for 15 min. Slides were then placed in Aniline Blue for 5 min, washed with distilled water, and incubated in 1% aqueous acetic acid for 5 min. Slides were quickly dehydrated in two changes of 95% ethanol, followed by three changes of 100% ethanol, then transferred to xylene (Thermo Fisher Scientific, X3P-1GAL) and cleared in three changes. Glass coverslips were mounted using Cytoseal 60 medium (Electron Microscopy Sciences).

Full slide scans were prepared using an Aperio VERSA 200 Imaging System (Leica Biosystems) and analyzed for percentage collagen quantification. Analysis was performed using Aperio ImageScope software through eSlide Manager (v12.4.3.5008; Leica Biosystems). Images were first annotated to select regions of interest and to remove artifacts, such as tissue folds, off-target saining and debris. The positivity of blue chromogen (ratio of blue pixel count to red pixel count) was reported as a percentage to denote the collagen content in the annotated areas. The following equation was used to generate collagen percentage:

% collagen=blue pixel count/(red pixel count+blue pixel count)×100,

where the blue pixel count corresponded to collagen area, and the red pixel count corresponded to non-collagen muscle area.

In-life whole blood was collected from the facial mandibular vein under isoflurane anesthesia, with blood volume not exceeding 10% of rodent total blood volume. Terminal serum was collected via cardiac puncture following anesthetization by intraperitoneal injection with ketamine/xylazine cocktail. Blood was collected in serum separator tubes, and coagulation was allowed for at least 30 min. Samples were centrifuged at a speed of at least 1780 g for 10 min. Serum was collected, snap frozen and stored at ≤−70°C before serum biomarker quantification.

Whole blood was collected, and serum was processed at 4, 8, 12, 16, 20, 24, 32 and 52 weeks of age prior to functional outcome assessments for the quantification of CKM (MM and MB isoforms), MMP-9 and TIMP-1 serum biomarkers by enzyme-linked immunosorbent assay (ELISA). Assays for MMP-9 and TIMP-1 were performed by BioAgilytix Labs (Durham, NC, USA) using the R&D Systems Quantikine^® ELISA Mouse Immunoassay kit for the respective biomarker and following the manufacturer's instructions. Absorbance was read using a microplate reader set to 450 nm against 570 nm. Assays for CKM (MM and MB isoforms) were performed by BioAgilytix Labs using the MSD Muscle Injury Panel (rat) 2 kit (Meso Scale Discovery).

A Muscle Injury Panel 3 (mouse) electro chemiluminescent immunoassay (Meso Scale Discovery) was used to measure skeletal and cardiac muscle injury by quantitative analysis of endogenous cTnI, FABP3, Myl3 and sTnI serum levels at 4, 8, 12, 16, 20, 24, 32 and 52 weeks of age. A standard curve for each biomarker was prepared, and all reported values were within the detection range of the standard curve. Assay plates were read using an MSD Sector Imager 120 MM plate reader (Meso Scale Discovery), and signal and mean concentration of biomarkers were determined using the Discovery Workbench (v4.0.13). Analyses were performed using Prism Software version 8.0 for Windows (GraphPad Software).

Five- to seven-week-old hDMD/mdx and hDMDΔ52/mdx mice were injected via the tail vein with rh74.AAV solution at 2.5×10¹⁴ vector genomes/kg or saline control solution. The single adeno-associated virus (AAV) encoded SaCas9, sgRNAs ATATAGTAATGAAATTATTGGCAC and CATTTGTATAGAGAGGAAATGT (Gersbach et al., 2022↗). Animals were euthanized at 12 weeks post-administration by intraperitoneal administration of ketamine/xylazine cocktail, and tissues were fresh frozen.

Tissues were homogenized by mechanical disruption using a Bullet Blender Tissue Homogenizer (Next Advance) in protein lysis buffer (125 mM Tris-HCl, 4 M urea, 4% SDS) supplemented with cOmplete™, Mini Protease Inhibitor Cocktail (Sigma-Aldrich). Dystrophin protein expression was analyzed on the Jess Automated Western Blot system (Bio-Techne). Protein concentrations were quantitated using a NanoDrop OneC Spectrophotometer (Thermo Fisher Scientific). A protein concentration of 0.112 mg/ml was prepared into 1× fluorescent master mix for each sample following the manufacturer's protocol using the 66-440 kDa Separation Module. Detection of dystrophin was performed using the rabbit monoclonal antibody (Abcam, ab154168) at 1:600 dilution in milk-free antibody diluent, and detection of alpha-actinin-4 loading control was performed using mouse monoclonal antibody (Abcam, ab254074) at 1:50 dilution in milk-free antibody diluent. Secondary antibodies used were donkey anti-mouse NIR (ProteinSimple, DM-009) and goat anti-rabbit HRP (ProteinSimple, DM-001) following the manufacturer's protocol. Samples were run on the 66-440 kDa fluorescence separation plates using the 8×25 capillary cartridges. Plate layouts included the replex module and were loaded according to the manufacturer's specifications.

To calculate the percentage dystrophin restoration, the dystrophin area under the curve (AUC) was normalized to the AUC obtained for actinin-4. The average actinin-4-normalized dystrophin value from hDMD/mdx mice (N=5) was used to determine 100% dystrophin expression. For each treated animal sample, the actinin-4-normalized edited dystrophin expression was divided by the average healthy actinin-4-normalized dystrophin expression and multiplied by 100 to generate the percentage dystrophin restored.

RNA was isolated from tissue samples using a NucleoMag^® RNA Kit (Macherey-Nagel) on a KingFisher Apex instrument. cDNA was generated from 70-100 ng RNA using a SuperScript VILO cDNA Synthesis Kit (Thermo Fisher Scientific) following the manufacturer's protocol with a 2-h 42°C incubation step. Transcript editing was analyzed by ddPCR using 2× ddPCR Supermix for probes (no dUTP) (Bio-Rad) and custom PrimeTime assays to quantify edited and unedited cDNA transcripts in the sample. For Δ45-55 transcript-specific edits, the custom PrimeTime Assay was designed with a HEX-labeled probe against the Δ45-55 edited cDNA sequence (forward primer, 5′-CTGAGAATTGGGAACATGC-3′; reverse primer, 5′-CATCGGAACCTTCCAGGG; probe, ACAAATGGTATCTTAAGGACCTCCAAGGTG-3′). The final primer concentration was 0.25 μM per reaction, and the final probe concentration was 0.9 μM per reaction. The reference assay used for the Δ45-55 edit-specific master mix was TaqMan Assay Hs02562862_s1 (Thermo Fisher Scientific) targeting exon 55 of the DMD transcript and labeled with a VIC-MGB probe. The exon 55 assay was diluted to a 1× concentration per reaction. All samples and controls were run in duplicate on a QXOne analyzer (Bio-Rad). Automated positive thresholds were generated for all samples for analysis as possible. Manual thresholds were drawn as needed.

Statistical analysis was performed using GraphPad Prism 10.1.2 software. Data were expressed as mean±s.d. (error bars), with individual points representing individual animals. One-way or two-way ANOVA was performed for analysis of normally distributed data (histology quantification, TA physiology, ECG, protein quantification). Kruskal–Wallis test with Dunn's multiple comparisons was performed for nonparametric data (muscle fiber diameter). Mann–Whitney test was used for nonparametric data (serum biomarker quantification). No animals assessed were excluded from analysis for any outcomes reported.