Muscle-specific CRISPR/Cas9 dystrophin gene editing ameliorates pathophysiology in a mouse model for Duchenne muscular dystrophy

Induction of dystrophin expression was tested following AAV6-mediated delivery of CRISPR/Cas9 components derived from either Streptococcus pyogenes (SpCas9)22 or Staphylococcus aureus (SaCas9)23 using dual- or single-vector approaches, respectively (Fig. 1a–e). Cas9 expression was restricted to skeletal and cardiac muscle by use of the muscle-specific CK8 regulatory cassette (RC)24 to reduce the risk of off-target events in non-muscle cells and to minimize elicitation of an immune response25,26. We tested several approaches to either excise exons 52 and 53 (Δ5253; strategy 1) or to directly target the mutation in exon 53 (53*; strategy 2). Due to the ∼5 kb packaging limit of AAV we designed dual AAV vectors to work in tandem: a nuclease vector expressing SpCas9 under control of the CK8 RC and a set of targeting vectors containing two single-guide RNA (sgRNA) expression cassettes unique to strategies 1 or 2 (Fig. 1a–e). A variant of strategy 1 relying on CK8-regulated expression of the smaller SaCas9 enabled use of a single vector (Fig. 1a).

The overall approaches used in strategy 1 (Δ5253) are potentially applicable to a majority of DMD patients with mutations affecting one or more exons whose removal via editing would allow production of a mRNA with an ORF. For this, we designed sgRNAs to direct Cas9-mediated DNA cleavage within the introns flanking exons 52–53 (Fig. 1a). Following DNA repair via non-homologous end joining (NHEJ) these would result in deletion of ∼45 kb of genomic DNA and 330 bp in the encoded mRNA. Successful deletion with strategy 1 will remove the nonsense mutation and lead to the expression of a dystrophin lacking 110 amino acids in a non-essential portion of the protein (Fig. 1b). Strategy 2 (53*) was developed to target small mutations directly, in this case in exon 53, using two distinct methods. These approaches could be applicable to patients with mutations in exons encoding essential domains of dystrophin, such as the dystroglycan-binding domain27. The first approach within strategy 2 relies on the introduction of a ‘mutation-corrected’ DNA template to allow for potential HDR following Cas9-mediated DNA cleavage, resulting in full-length endogenous dystrophin expression (Fig. 1c,d). In the absence of successful HDR, this approach could still enable dystrophin expression where NHEJ repair of the cleaved exon 53 leads to excision of the nonsense mutation while maintaining an ORF in the resultant mRNA (Fig. 1c,e).

Dystrophin gene targeting was initially evaluated in vitro using the T7 endonuclease 1 assay in mdx^4cv-derived primary dermal fibroblasts. The respective targeting efficiencies for sgRNA-i51 and sgRNA-i53 were 9 and 16%, while a combined targeting efficiency of 8% was observed for the 5′ and 3′ sgRNAs within exon 53 (which due to their close proximity were analysed together; Supplementary Fig. 1). For initial in vivo testing 10–12 week old male mdx^4cv mice were injected in the tibialis anterior (TA) muscles with 5 × 10¹⁰ vector genomes (v.g.) of the AAV6 CK8-nuclease plus targeting vectors and sacrificed at 4 weeks post-injection. In vivo targeting efficiency was estimated via deep sequencing across target regions within the dystrophin gene. For strategy 1 PCR amplification of the genomic DNA region spanning the intron 51–53 target sites revealed low levels of a unique Δ5253 deletion product whose sequence was verified following isolation and cloning (Supplementary Fig. 2). Due to the large size of the genomic deletion, quantification of NHEJ events resulting from the deletion of both exons 52 and 53 could not be determined via deep sequencing. However, deep sequencing of PCR amplicons generated across the individual target sites could be used to quantify the instances where on-target DNA cleavage did not result in the excision of the intervening 45 kb segment. Using this approach, gene editing efficiencies at introns 51 and 53, respectively, were 8.6% and 8.2% for the dual-vector (Sp) approach and 3.5% and 2.7% for the single vector (Sa) approach (Fig. 2a; Supplementary Fig. 2; Supplementary Table 1). Reverse transcription PCR (RT–PCR) analysis revealed a predominant shorter dystrophin transcript that lacked the sequences encoded on exons 52 and 53 as determined by sequencing of the excised unique band (Fig. 2b,c).

For strategy 2, the combined gene editing efficiency for both target sites within exon 53 was 2.3%, as determined by deep sequencing (Fig. 2d; Supplementary Fig. 3; Supplementary Table 1). Encouragingly, successful HDR was detected in 0.18% of total genomes (Fig. 2d; Supplementary Fig. 4; Supplementary Tables 1 and 2). While this efficiency was low (∼8% of the edited genomes resulted from HDR), the data show that myogenic cells within dystrophic muscles are at least modestly amenable to HDR-mediated dystrophin correction following CRISPR/Cas9 targeting. Analysis of dystrophin transcripts isolated from four treated samples revealed a unique shorter RT–PCR product that, following sequencing of individual cloned RT–PCR products, was shown to correspond to a complete deletion of exon 53 (Supplementary Fig. 3). This unanticipated exclusion of exon 53 from the mRNA likely resulted from larger indel mutations disrupting splicing enhancer signals located within this exon28. Successful editing within the main exon 53 RT–PCR product was detected via both T7 endonuclease 1 digestion and Sanger sequencing of individual clones (Supplementary Fig. 3). Deep sequencing of RT–PCR amplicons spanning exons 52 and 53 revealed an overall editing efficiency of 9.2% at the transcript level with 0.8% of total transcripts corresponding to successful HDR events (Fig. 2d; Supplementary Fig. 3 and 4; Supplementary Tables 1 and 3), thus indicating successful Dmd gene editing and HDR within exon 53. Analysis of the sequence reads revealed several types of editing events. For example, 44% (genomic DNA) and 36% (mRNA) of the edited sequences carried insertions, deletions or substitutions that did not shift the reading frame (Fig. 2e). However, only 3% (genomic DNA) and 16% (mRNA) of all edited sequences were in-frame deletions that also removed the mdx^4cv stop codon. Since ∼8% of all edited genomes and ∼9% of all edited transcripts resulted from HDR (Fig. 2d,e), a total of ∼11% (genomic) and ∼25% (transcript) of the strategy 2 editing events were able to express dystrophin (Fig. 2e, Supplementary Fig. 4; Supplementary Tables 1–3). Overall, on-target editing frequency was significantly higher than for predicted off-target sites sharing the most sequence similarity to the sgRNAs used in strategies 1 and 2 (Supplementary Table 4).

Establishment of a functional ORF led to significant induction of dystrophin expression in treated TAs as detected by immunostaining of muscle cryosections (Fig. 3a; Supplementary Fig. 5) and by western blotting of whole muscle lysates (Fig. 3b). CRISPR/Cas9-mediated gene correction resulted in full- to near-full-length dystrophin protein expression levels of 0.8–18.6% (dual vector, n=4) or 1.5–22.9% (single vector, n=4) for strategy 1 and 1.8–8.4% (53*, dual vector, n=4) for strategy 2, as compared with wild-type (WT) dystrophin levels (Fig. 3c). In addition to the detection of full- to near-full-length dystrophin, western analysis also revealed a range of shorter dystrophin isoforms (110–160 kD) of unclear therapeutic impact that were more frequent in strategy 2-treated muscles, possibly due to aberrant splicing.

Immunostaining of muscle cross-sections revealed that an average of 41% (Δ5253) and 45% (53*) of myofibers expressed dystrophin (Fig. 3d). Of note, dystrophin-positive myofibers in treated TAs were significantly larger than myofibers of untreated mdx^4cv controls and than dystrophin-negative fibres within treated muscles (Fig. 3e,g; Supplementary Fig. 6), constituting an average of 54% (Δ5253) and 61% (53*) of the myogenic cross-sectional area with a maximum observed positive area of 68% (Δ5253) and 71% (53*). Dystrophin-positive myofibers within treated muscles also displayed a significant reduction in central nucleation (Fig. 3h).

Induction of dystrophin expression also allowed for sarcolemmal localization of neuronal nitric oxide synthase (nNOS), an important component of the dystrophin-glycoprotein complex that modulates muscle performance (Fig. 4a)11. To assess whether CRISPR/Cas9-mediated induction of dystrophin expression would translate into functional improvements we performed in situ measurements of muscle force generation at 18 weeks post-transduction of 2-week-old male mdx^4cv mice. Encouragingly, the observed dystrophin levels in muscles treated using strategy 1 were maintained at this later time point, resulting in significant increases in specific force generating capacity and protection from contraction-induced injury (Fig. 4b,c). Conversely, muscles treated according to strategy 2 only displayed a slight but non-significant increase in specific force development, likely due to the lower levels of dystrophin production.

On the basis of the higher dystrophin-correction efficiency observed for strategy 1, we proceeded to test this approach following systemic delivery of the AAV nuclease and targeting vectors using a range of doses between 1–10 × 10¹² v.g. per mouse. Both single- and dual-vector approaches yielded widespread dystrophin expression in the heart, with up to 34% of cardiac myofibers expressing dystrophin at 4 weeks post-transduction (Fig. 5). While both high- and low-vector doses were able to generate dystrophin expression in the heart (Fig. 5b–d), only the high dose was able to generate widespread, albeit variable, dystrophin expression in all muscle tissues analysed (ranging from <10% dystrophin-positive fibres in the quadriceps and EDL muscles to >50% in soleus muscles; Fig. 5e–h). Furthermore, higher cardiac dystrophin expression levels were also obtained with increasing vector dose (Fig. 5i).

Plasmids containing regulatory cassettes for expression of Cas9 or gRNAs flanked by AAV serotype 2 inverted terminal repeats (ITRs) were generated using standard cloning techniques. The spCas9 nuclease expression cassette was generated by PCR cloning of NLS-SpCas9-NLS from LentiCRISPRv1 (ref. 39), and insertion into pAAV (Stratagene) containing the ubiquitous elongation factor-1 alpha short promoter (EFS)39 (for in vitro studies in fibroblasts) or the muscle-specific creatine kinase 8 (CK8) regulatory cassette (RC)24,26 (for in vivo studies). (Sp)sgRNA target sequences were selected using the online software ZiFiT Targeter ( http://zifit.partners.org/ZiFiT/↗) and inserted into pLentiCRISPRv1 following BsmB1 restriction enzyme digestion. Two targeting constructs to work in conjunction with SpCas9 were generated by PCR cloning of the U6-(Sp)sgRNA expression cassette from pLentiCRISPRv1 followed by insertion into pAAV plasmids on either side of a CMV-mCherry expression cassette and a HDR template spanning positions X84575274 to X84576081 of the murine DMD gene cloned from C57BL/6 genomic DNA. The corresponding protospacer adjacent motif (PAM) sites at positions X84575612 (G-A) and X84575639 (G-A) within the HDR template were mutated using PCR-mediated mutagenesis while preserving the encoded amino acids (silent mutations) to eliminate or reduce targeting of the DNA repair template by Cas9. The modified HDR sequence, guide RNA sequences as well as primer sequences for cloning and PCR amplification of genomic DNA and complementary DNA (cDNA) are provided in Supplementary Tables 5–6. The SaCas9 single vector expression cassette was generated by replacing the CMV immediate early enhancer and promoter and the bovine growth hormone poly-adenylation (pA) signal in plasmid #61591 (Addgene)23 with the CK8 RC and a rabbit beta-globin pA signal, followed by PCR cloning and insertion of a second U6-(Sa)sgRNA expression cassette sequential to the first. (Sa)sgRNA target sequences were manually selected to target the same locations as the (Sp)sgRNAs and inserted into the U6-(Sa)sgRNA expression cassette via Bsa1 restriction enzyme digestion before inserting the second U6-(Sa)sgRNA cassette into the final construct. Nuclease and targeting pAAV plasmids were co-transfected with the pDG6 packaging plasmid into subcultured HEK293 cells (American Type Culture Collection) using calcium phosphate-mediated transfection to generate AAV6 vectors that were harvested, purified via heparin-affinity chromatography and concentrated using sucrose gradient centrifugation40. Resulting titres were determined by Southern analyses using probes specific to the poly-adenylation signal or CMV promoter for nuclease and targeting vectors, respectively.

Primary dermal fibroblasts were isolated from 3-week-old male mdx^4cv mice41. Electroporation of ∼600,000 cells per strategy were performed in Invitrogen R-buffer containing 4 μg of both nuclease (EFS-SpCas9)- and targeting (Δ5253/53*) plasmid expression constructs using a Neon transfection system (Invitrogen) with three 10 ms pulses of 1,650 volts. Cells were subsequently seeded on 0.1% gelatin-coated culture vessels and maintained for 12 days in Dulbecco’s modified Eagle medium supplemented with Penicillin-Streptomycin, Sodium pyruvate, L-Glutamine and 15% fetal bovine serum (Thermo Fisher Scientific) before harvest and DNA isolation (DNeasy, Qiagen).

All animal experiments were approved by the Institutional Animal Care and Use Committee of the University of Washington. Intramuscular delivery of 2.5–5 × 10¹⁰ v.g. of each vector (nuclease and targeting) was performed via longitudinal injection into tibialis anterior (TA) muscles of 2–12-week-old male C57BL/6-mdx^4cv (mdx^4cv) mice. For strategy 1, systemic delivery of 1 × 10¹² v.g. (low dose) to 1 × 10¹³ v.g. (high dose) was performed via retro-orbital injection into 11 week-old male mdx^4cv mice (n=3). Both dual- and single-vector approaches were evaluated at the low dose of 1 × 10¹² v.g. of each vector, while the dual-vector approach was also evaluated at a high dose of 1 × 10¹³ v.g. of the nuclease vector and 4 × 10¹² v.g. of the targeting vector. The mdx^4cv mouse model of DMD harbours a nonsense C to T mutation in exon 53 leading to a loss of dystrophin expression20. These mice exhibit ∼10-fold lower frequencies of revertant dystrophin expressing muscle fibres than the original mdx^scsn mouse strain, which provides much greater assurance that dystrophin-corrected fibres resulted from gene targeting rather than spontaneous reversion.

Muscles were collected and analysed at 4 weeks post-transduction and compared with age-matched male non-injected mdx^4cv and WT mice, except for mice undergoing physiological measurements which were analysed at 18 weeks post-transduction. Medial portions of TA muscles were embedded in Optimal Cutting Temperature (O.C.T.) compound (VWR International) and fresh frozen in liquid nitrogen cooled isopentane for immunofluorescence analysis. The remaining portions of TA muscles were snap frozen in liquid nitrogen and ground to a powder under liquid nitrogen in a mortar kept on dry ice for subsequent extraction of DNA, RNA and protein.

TA cross-sections (10 μm) were co-stained with antibodies raised against alpha 2-laminin (Sigma, rat monoclonal, 1:200) and the C-terminal domain of dystrophin (a kind gift from Dr. Stanley Froehner at the University of Washington Department of Physiology and Biophysics, rabbit polyclonal, 1:500). Serial sections were stained with antibodies raised against neuronal nitric oxide synthase (Invitrogen, rabbit polyclonal, 1:200). Slides were mounted using ProLong Gold with DAPI (Thermo Fisher Scientific) and imaged via Leica SPV confocal microscope at the University of Washington Biology Imaging Facility ( http://depts.washington.edu/if/↗). Confocal micrographs covering the entirety of injected TA muscle sections were acquired and montaged using the Fiji toolset (ImageJ) and Photoshop (Adobe). Quantification of dystrophin-positive myofibers and dystrophin-positive muscle cross-sectional area was performed via semi-automated tracing and measurement of 1,250 to 3,500 individual myofibers per TA using Adobe Photoshop (n=5 TAs per treatment group). Automated quantification of central nucleation was performed using software developed in-house by Rainer Ng (CHAMP) running on the Matlab platform.

DNA and RNA were isolated using Trizol reagent (Invitrogen) according to the manufacturer’s recommendations. Approximately 500 bp amplicons across the targeted regions of genomic DNA were generated by PCR using Phusion proof-reading polymerase (New England Biolabs, NEB) and analysed for targeting efficiency using T7 endonuclease 1 (NEB), next generation sequencing (BGI International or in-house) or Sanger sequencing (Simpleseq, Eurofins MWG Operon) of subclones of PCR amplicons (Zero Blunt TOPO, Invitrogen). The T7 endonuclease assay was performed by denaturing and re-annealing the amplified PCR product followed by treatment with T7 endonuclease 1 to cleave indel-derived heteroduplex PCR products. Analysis of dystrophin-targeted transcripts by RT–PCR of the target regions was performed on cDNA made using Superscript III first-strand synthesis supermix (Invitrogen). Specific indel mutations or deletions in the dystrophin transcript were identified by Sanger sequencing of individual subclones of RT–PCR fragments. Muscle proteins were extracted in radioimmunoprecipitation analysis buffer (RIPA) supplemented with 5 mM EDTA and 3% protease inhibitor cocktail (Sigma, Cat# P8340), for 1 hour on ice with gentle agitation every 15 min. Total protein concentration was determined using Pierce BCA assay kit (Thermo Fisher). Muscle lysates from WT (10 and 1 μg), untreated mdx^4cv (30 μg) and treated mdx^4cv (30 μg) mice were denatured at 99 degrees Celsius for 10 min, quenched on ice and separated via gel electrophoresis after loading onto Bolt 4–12% Bis-Tris polyacrylamide gels (Invitrogen). Protein transfer to 0.45 μm PVDF membranes was performed overnight at constant 34 volts at 4 °Celsius in Towbin’s buffer containing 20% methanol. Blots were blocked for 1 hour at room temperature in 5% non-fat dry milk before overnight incubation with antibodies raised against the C-terminal domain of dystrophin (Froehner Lab, Rabbit polyclonal, 1:10,000), anti-SpCas9 (Millipore, mouse monoclonal, 1:2,000), anti-HA (Roche, Rat monoclonal-HRP conjugated, 1:2,000) for detection of HA-tagged saCas9 and Gapdh (Sigma, Rabbit polyclonal, 1:100,000). Horseradish-peroxidase conjugated secondary antibody staining (1:50,000) was performed for 1 h at room temperature before signal development using Clarity Western ECL substrate (BioRad) and visualization using a Chemidoc MP imaging system (BioRad). Gel- and blot- band densitometry measurements were performed on unsaturated images using ImageJ software (National Institutes of Health).

Approximately 200–250 bp PCR products were generated across target-, and the top predicted off-target sites for each sgRNA using Platinum Taq High-Fidelity polymerase (Invitrogen) or Phusion High-Fidelity Polymerase (NEB). Potential off-target sites were identified using ZiFiT Targeter software for SpCas9. CRISPR Rgen tools Cas-OFFinder software ( http://www.rgenome.net/cas-offinder/↗) was used to identify potential off-target sites for SaCas9, using a mismatch number of ≤3, DNA bulge size ≤1 and RNA bulge size ≤1. For Strategy B, genomic deep sequencing was performed on a ∼230 bp nested PCR product generated from an initial ∼500 bp product amplified across exon spanning both target sites. To eliminate false detection of the HDR template DNA present in AAV vectors, the primer pair used to generate the 500 bp product was designed with one primer annealing outside of the region complimentary to the HDR template. The resulting PCR product was isolated following gel electrophoresis (GeneJET gel extraction kit, Thermo Fisher Scientific) before performing nested PCR followed by a second gel extraction. For each site analysed, amplicons from 4–5 mice were pooled and subjected to standard Illumina library preparation (A-tailing, adaptor ligation and amplification using NEBNext library preparation kit (NEB)), and QC’d using a BioAnalyzer before paired end (PE150) sequencing on an Illumina MiSeq system (Illumina Inc., San Diego, CA). Libraries were barcoded for multiplexed sequencing and subsequent reads were parsed and QC’d using custom scripts (Trim galore software ( http://www.bioinformatics.babraham.ac.uk/projects/trim_galore/↗), phred33 score ≥ 30) and standard Illumina tools. On-target paired end (PE150) sequencing of DNA amplicons generated from muscles treated according to strategy 2 (53*) was performed by submitting the samples to BGI International (BGI Americas, Cambridge, MA). Uniquely mapping read pairs were used for downstream analysis using the CRISPResso software pipeline42. For CRISPResso analyses: 25 bp at each end of the amplicon were excluded from quantification, the window size around each cleavage site used to quantify NHEJ events was set to 5 bp and sequence homology for an HDR occurrence was set to 98%. Following CRISPResso analysis, manual analysis and quantification was performed by searching for defined sequences in the quality-filtered and adapter-trimmed deep sequencing FASTQ files to provide further information on specific genotypes generated by strategy 2. For DNA reads, search sequences were chosen to span the region containing both target sites and the site of the C–T mutation. For RNA reads, search sequences were defined to span a region starting from within exon 52 (>45 kb away from the target region) extending past the prototypical cut site at the 3′ end of the target region.

Eighteen weeks post-transduction, treated mdx^4cv mice together with age-matched controls were anaesthetized with 2,2,2-tribromoethanol (Sigma) and assayed in situ for force generation43. Briefly, a 4–0 silk suture was tied around the distal TA tendon and to a lever attached to a force transducer. After determination of optimal muscle fibre length (L₀) the maximum isometric tetanic force was measured during electrical stimulation using Dynamic Muscle Control v5.420 software (Aurora Scientific). Muscle cross-sectional area (CSA) was calculated by dividing muscle mass (mg) by fibre length (mm) and 1.06 mg mm⁻³ (density of mammalian skeletal muscle). Specific force values were obtained by normalizing maximum isometric tetanic force production to CSA. Protection against contraction-induced injury was evaluated by measuring force production during progressive lengthening contractions beyond optimal fibre length44.

Data values are represented as mean±s.e.m. and were analysed in Excel (Microsoft) and Prism6 (GraphPad). Measurements were analysed for statistical significance using one-way analysis of variance (ANOVA) multiple comparison tests with Turkey’s post hoc tests unless otherwise stated. Statistical significance was set to P<0.05.

Sequence data supporting the findings of this study have been deposited in the sequence read archive (SRA) with the BioProject accession code PRJNA358248 ( https://www.ncbi.nlm.nih.gov/bioproject/PRJNA358248↗).

The remaining data are available within the article and itsand from the corresponding author upon reasonable request. Full scans for western blots are available in. Supplementary Information files Supplementary Fig. 7

Sequence data supporting the findings of this study have been deposited in the sequence read archive (SRA) with the BioProject accession code PRJNA358248 ( https://www.ncbi.nlm.nih.gov/bioproject/PRJNA358248↗).

The remaining data are available within the article and itsand from the corresponding author upon reasonable request. Full scans for western blots are available in. Supplementary Information files Supplementary Fig. 7