CRISPR/Cas Correction of Muscular Dystrophies

Abstract

Muscular dystrophies are a heterogeneous group of monogenic neuromuscular disorders which lead to progressive muscle loss and degeneration of the musculoskeletal system. The genetic causes of muscular dystrophies are well characterized, but no effective treatments have been developed so far. The discovery and application of the CRISPR/Cas system for genome editing offers a new path for disease treatment with the potential to permanently correct genetic mutations. The post-mitotic and multinucleated features of skeletal muscle provide an ideal target for CRISPR/Cas therapeutic genome editing because correction of a subpopulation of nuclei can provide benefit to the whole myofiber. In this review, we provide an overview of the CRISPR/Cas system and its derivatives in genome editing, proposing potential CRISPR/Cas-based therapies to correct diverse muscular dystrophies, and we discuss challenges for translating CRISPR/Cas genome editing to a viable therapy for permanent correction of muscular dystrophies.

1. Introduction

Skeletal muscle is one of the largest tissues in the human body and is essential for physical movement, metabolism and energy homeostasis. Despite the remarkable regenerative capacity of skeletal muscle, it is vulnerable to pathological disorders, including congenital myopathies, inflammatory myopathies, and muscular dystrophies [1–3]. Inherited muscle disorders such as congenital myopathies are caused by genetic mutations, which result in muscle weakness and hypotonia. Inflammatory myopathies comprise a heterogeneous group of muscle diseases, which can be further classified into polymyositis, dermatomyositis, necrotizing myopathy, and inclusion body myositis. Muscular dystrophies represent a group of more than 30 genetic diseases clinically described by progressive skeletal muscle weakness and muscle mass atrophy[4]. Genetic mutations causing muscular dystrophy generally affect sarcolemma structures while others alter nuclear positioning, membrane repair, genome transcription, and RNA splicing[3]. Heart dysfunction such as cardiomyopathy is also a common phenotype seen in muscular dystrophies.

To date, there are only two clinically approved therapies to control the symptoms of muscular dystrophies and slow disease progression: 1) steroid supplementation and 2) antisense oligonucleotide (AON)-based therapy. It is well established that steroid treatment reduces muscle inflammatory damage and promotes muscle membrane repair, alleviating the symptoms of muscular dystrophy [5]. However, long-term exposure to steroids is associated with severe side effects, including bone loss, high blood pressure and even muscle weakness and atrophy. (AON)-based therapy can induce skipping of mutant exons by targeting RNA transcripts. Although this approach offers correction to the mutation, the efficiency remains extremely low [6]. In addition to the aforementioned therapies, gene replacement therapy is a potential treatment for muscular dystrophies. It relies on using adeno-associated virus (AAV) as a vector to deliver a functional gene product and replace the endogenous mutant one. Several gene replacement therapy clinical trials are actively being performed in patients with Duchenne muscular dystrophy and limb-girdle muscular dystrophies [7]. Nevertheless, major challenges exist in gene replacement therapy, such as the limited packaging capacity of AAV, restricting the gene product size and necessitating truncation of the product which might compromise protein function. Moreover, potential long-term benefits of gene replacement therapy are unknown since the endogenous gene mutation still exists in the human genome.

With no apparent cure for most muscular dystrophies, there is an urgent need to develop new therapeutic approaches to treat these debilitating muscle diseases. Correcting genetic mutations at the genome level using programmable nuclease for site-specific DNA double-stranded breaks (DSBs) provides a powerful tool for precise genome editing [8]. Among different programmable nucleases developed so far, clustered regularly interspaced short palindromic repeats (CRISPR) technology offers simplicity and accuracy in genome editing, representing a promising therapy for permanent correction of genetic diseases [9]. This review provides a comprehensive overview of CRISPR/Cas genome editing technology and its application in correcting muscular dystrophies, as well as challenges and future perspectives in CRISPR/Cas-mediated therapeutic genome editing.

2. CRISPR/Cas GENOME EDITING

CRISPR/Cas Genome Editing Components

The CRISPR/Cas system consists of two components, the CRISPR array and CRISPR-associated (Cas) proteins. The first report of CRISPR arrays dates back to the 1980s but at that time the purpose was undefined. Years later, it was discovered that the CRISPR array has a viral origin and some Cas proteins are nucleases or helicases, leading to the hypothesis that the CRISPR/Cas system may function as an adaptive immune system in bacteria and archaea for defending against bacteriophages [10, 11]. The CRISPR/Cas system is categorized into two classes: the class 1 system requires multiple small Cas effectors for antiviral defense; in contrast, the class 2 system only utilizes a single large Cas effector for RNA-guided DNA cleavage [12]. Most CRISPR/Cas systems engineered for genome editing are derived from the class 2 CRISPR system, including Cas9 from Streptococcus pyogenes (SpCas9) and Staphylococcus aureus (SaCas9).

CRISPR/Cas9-mediated DNA cleavage requires two components, an RNA-guided Cas9 endonuclease and a single guide RNA (sgRNA), which is a synthetic fusion of CRISPR RNA (crRNA) and transactivating CRISPR RNA (tracrRNA) [13] (Fig. 1A). DNA cleavage is induced by a Cas9-sgRNA ribonucleoprotein complex when the target DNA sequences match with the protospacer region of the sgRNA and when a protospacer-adjacent motif (PAM) is available. The PAM sequence for different Cas9 orthologs is distinct. For example, SpCas9 recognizes 5’-NGG-3’ or 5’-NAG-3’ PAMs, whereas SaCas9 prefers a 5’-NNGRRT-3’ PAM [13, 14]. The fundamental difference between CRISPR/Cas9 and other nucleases such as ZFNs and TALENs is that CRISPR/Cas9-mediated DNA cleavage is programmed by an sgRNA, whereas zinc-finger nucleases (ZFNs) and transcription activator-like effector nucleases (TALENs) require specifically engineered DNA binding domains for target binding.

Fig. 1. CRISPR/Cas9 and its derivatives used for genome editing.

(A) Schematic illustration of the CRISPR/Cas9 (clustered regularly interspaced short palindromic repeats-CRISPR associated protein 9) system. In the CRISPR/Cas9 system, target recognition is mediated by base pairing of DNA with the single guide RNA (sgRNA). (B) Schematic illustration of the CRISPR/Cas9 nickase system. Cas contains HNH and RuvC nuclease domains for DNA cleavage. The D10A Cas9 nickase has a mutation in the RuvC nuclease domain and is only active for target strand cleavage by the HNH nuclease domain (indicated by the arrowhead). The HNH nuclease domain is named for the histidine (H) and asparagine (N) amino acid residues. (C) Schematic illustration of the CRISPR/Cas9 Base Editing system. CRISPR/Cas9 nickase (nCas9) or deactivated CRISPR/Cas9 (dCas9) can be fused with an engineered cytidine deaminase (CBE) (orange) to induce C•G to T•A transitions or an adenine deaminase (ABE) (green) for A•T to G•C transitions. (D) Schematic illustration of the CRISPR/Cas9 Prime Editing system. CRISPR/Cas9 nickase (nCas9) fused with an engineered reverse transcriptase (magenta) can be used to introduce a variety of DNA modifications at the target site.

Engineered CRISPR/Cas Systems

Although wild-type (WT) Cas9 has been widely used for genome editing, many limitations still exist, including restricted PAM preference, high off-target activity and limited editing capacity. To address these issues, many engineered Cas9 effectors have been developed, including Cas9 with expanded PAM sequences, high-fidelity Cas9, Cas9 base editors, prime editors, and Cas9 activators and repressors.

Through directed evolution, a protein engineering method that copies the process of natural selection to drive proteins toward a specific goal, SpCas9 has been engineered to exhibit altered PAM profiles or relaxed PAM preferences, including 5’-NGA-3’, 5’-NGC-3’, 5’-NG-3’ and the near-PAMless 5’-NRN-3’ PAM [15–17]. These engineered SpCas9 variants significantly expand the number of potential gene editing sites within the genome. However, CRISPR/Cas9-mediated genome editing can result in off-target effects, which hinders its application in precision medicine. Several engineered high-fidelity Cas9 variants, such as eSpCas9, SpCas9-HF1 and HypaCas9 were developed by weakening the interactions between the non-target DNA strand and the Cas9 nuclease domain or by disrupting the interaction between Cas9 and the phosphate backbone of the target DNA strand [18–20]. These high-fidelity SpCas9 variants demonstrate significantly lower off-target activity without compromising on-target cutting efficiency.

SpCas9-mediated DNA cleavage is induced by two separate nuclease domains. Mutation of one or both of these nuclease domains partially or completely abolishes Cas9-mediated DNA cleavage activity, and therefore generates a nickase (nCas9) or a dead SpCas9 (dCas9) [13] (Fig. 1B). nCas9 and dCas9 can be fused with additional effectors for other genome editing purposes. For example, dCas9 fused with a transcriptional activator or repressor can be directed to enhancer or promoter regions to regulate gene expression. Similarly, dCas9 fused with an epigenetic modifier can be used for epigenome regulation, including changes in DNA methylation patterns and histone code modifications.

Base Editing is another powerful type of genome editing technology which was developed for single nucleotide conversions of the genome. nCas9 or dCas9 can be fused with an engineered adenine deaminase for A•T to G•C transitions or with a cytidine deaminase to induce C•G to T•A transitions [21] (Fig. 1C). Importantly, Base Editing does not induce DSBs and its editing outcome is more predictable than non-homologous end joining (NHEJ)-based DNA repair, and therefore represents an effective tool for correcting mutations with single nucleotide polymorphisms (SNPs).

Although Base Editing can precisely induce nucleotide transitions without introducing DSBs, additional modifications such as nucleotide transversions are still unachievable. Recently, a newly developed genome editing technology, known as Prime Editing, has been shown to be effective in mammalian genome engineering [22]. In this system, the sgRNA is extended to form a prime editing guide RNA (pegRNA) and SpCas9 nickase is fused with an engineered reverse transcriptase (Fig. 1D). The pegRNA serves as a template for the reverse transcriptase to introduce a DNA modification at the target site. Prime Editing represents the most comprehensive genome editor, allowing site-specific genomic insertions, deletions, as well as single-base transitions or transversions without introducing DSBs or requiring an exogenous donor DNA as a HDR repair template. However, the large size of Prime Editors limits their practicality with respect to in vivo applications.

Repair of CRISPR/Cas-mediated DNA Double-strand Breaks and Base Editing Induced DNA Mismatch

CRISPR/Cas-mediated DNA DSBs can be repaired by a variety of endogenous DNA repair pathways, including classical non-homologous end joining (C-NHEJ), homology-directed repair (HDR), and microhomology-mediated end joining (MMEJ). The C-NHEJ pathway repairs DSBs without a repair template and generates insertions and deletions (INDELs) (Fig. 2A). However, some studies applied the C-NHEJ pathway to precisely insert DNA fragments into post-mitotic cells, bringing precision to this error prone repair pathway [23]. The HDR pathway can precisely repair DSBs but is limited to S and G₂ phases of the cell cycle and requires a repair template, eliminating its application in post-mitotic cells, such as skeletal muscle and cardiomyocytes (Fig. 2B). MMEJ is an alternative NHEJ repair pathway and requires DNA microhomology present at the breakage point (Fig. 2C). MMEJ-mediated repair generates deletions with predictable outcomes. Application of the MMEJ pathway has been shown to precisely correct mutations caused by small duplications, serving as an alternative method for precise genome editing [24]. In contrast to CRISPR/Cas-mediated DNA DSBs, base editing induces cytidine or adenosine deamination, which are repaired by other pathways. Specifically, cytidine deamination induced by dCas9-CBE is repaired by the base excision repair (BER) pathway. When the dCas9 used in the base editing system is replaced by nCas9, the mismatch repair (MMR) pathway is triggered.

Fig. 2. DNA repair pathways involved in CRISPR/Cas-induced DNA double-strand break repair.

(A) Classical non-homologous end joining (c-NHEJ) is an error-prone DNA repair pathway, generating insertions or deletions (INDELs) (red) in the genome. Ku heterodimers (black circle) bind to the DNA breakage site and induce c-NHEJ. (B) Homology-directed repair (HDR) is a precise DNA repair pathway, which requires a donor template (magenta). (C) The microhomology-mediated end joining (MMEJ) pathway is a Ku-independent DNA repair pathway, which requires sequence homology (yellow) present at the DNA breakage site. MMEJ-mediated repair generates deletions (red) with predictable outcomes in the genome.

3. CRISPR/CAS CORRECTION OF MUSCULAR DYSTROPHIES

Duchenne Muscular Dystrophy

Duchenne muscular dystrophy (DMD) is the most common type of muscular dystrophy, affecting approximately 1 in 5,000 live male births [25, 26]. DMD is a recessive monogenic muscle disorder caused by mutations of the DMD gene residing on the X chromosome [27]. The DMD gene codes the dystrophin protein, one of the main components of the dystrophin-glycoprotein complex (DGC), which tethers the actin cytoskeleton to the inner surface of the sarcolemma [28] (Fig. 3). In the DGC, dystrophin serves as a shock absorber, maintaining cell membrane stability during muscle contraction. Complete loss of dystrophin expression leads to the most severe type of DMD, while in-frame truncation mutations in the spectrin-like repeat regions cause Becker muscular dystrophy (BMD), which is a mild form of muscular dystrophy. More than 7,000 mutations are associated with DMD. Among them, deletion mutations that disrupt the reading frame between adjacent exons are the predominant mutation type, representing approximately 70% of total cases. Other mutations, such as exon duplications, point mutations, and small deletions or insertions each account for approximately 10% of DMD cases [29]. DMD mutations are not evenly distributed in the dystrophin gene. The majority of DMD mutations are located in the hotspot regions, comprised of exons 2–20 and exons 45–55 [30].

Fig. 3. Muscle proteins involved in muscular dystrophies.

Genetic mutations causing muscular dystrophy generally affect the dystrophin-glycoprotein complex (DGC) while other mutations alter nuclear positioning, membrane repair, genome transcription, and RNA splicing. The main components of the DGC are the dystroglycan complex, sarcoglycan complex, and dystrophin. Mutations of dystrophin cause Duchenne or Becker muscular dystrophy (DMD or BMD); mutations of dysferlin cause limb-girdle muscular dystrophy (LGMD) type 2B; mutations of the sarcoglycan complex cause LGMD type 2C, 2D, 2E, and 2F; mutations in nuclear proteins cause other muscular dystrophies, including myotonic dystrophy type 1 and 2 (DM1 and DM2), and facioscapulohumeral muscular dystrophy type 1 (FSHD1).

Many spontaneous and engineered DMD animal models are in existence. The most widely used model is the mdx mouse, a spontaneously occurring mutation that has a C-to-T transition mutation in exon 23 [31, 32], resulting in a premature termination codon. Other mouse models such as mdx^2cv, mdx^3cv, mdx^4cv, mdx^5cv carry point mutations in intron 42, intron 65, exon 53 or exon 10, respectively [33]. However, single point mutations in the mdx series do not represent the predominant human DMD mutations. In order to address this limitation, several mouse models with exon out-of-frame deletion mutations within the hotspot regions have been generated by CRISPR/Cas9-mediated mutagenesis, including deletion of exon 43, 44, 45, 50, 51 and 52 [34–37]. In addition to mouse models of DMD, large animal models have been developed, including dog [38–44] and pig [45–48]. Unlike small rodent models, large animal models of DMD show early onset of cardiomyopathy and more severe dystrophic phenotypes. These large animal models are useful for testing CRISPR/Cas therapeutic gene editing in regard to body size and speed of disease progression.

CRISPR/Cas Correction of DMD

CRISPR/Cas-mediated genome editing in skeletal muscle and heart can correct various DMD mutations at the genomic level and restore the expression of functional truncated dystrophin. Several strategies for CRISPR/Cas-mediated correction of DMD have been developed, including exon deletion, exon skipping, exon reframing, and base editing (Table 1).

Table 1.

CRISPR/Cas correction of muscular dystrophies in animal models

Animal model	Mutation type	Correction method	CRISPR/Cas type	AAV serotype	Reference
Myotonic dystrophy
HSALR mice	Human skeletal actin transgene with CTG repeats	Transcription repression	dSaCas9	ssAAV6	[71]
HSALR mice	Human skeletal actin transgene with CTG repeats	RNA targeting	RNA-targeting SpCas9	ssAAV9	[72]
DMSXL mice	Human DMPK gene with CTG repeats	Repeat deletion	SaCas9	ssAAV9	[73]
Duchenne muscular dystrophy
mdx mice	Ex23 nonsense mutation	Ex23 del	SpCas9	SSAAV9	[51]
mdx mice	Ex23 nonsense mutation	Ex23 del	SaCas9	SSAAV8	[52]
mdx mice	Ex23 nonsense mutation	Ex23 del	SaCas9	ssAAV9	[50, 53]
mdx mice	Ex23 nonsense mutation	Ex23 del	SaCas9	ssAAV8 or ssAAV9	[54]
mdx mice	Ex23 nonsense mutation	Ex21–23 del	SaCas9	ssAAVrh.74	[104]
mdx/Utr+/− mice	Ex23 nonsense mutation	Ex21–23 del	SaCas9	ssAAVrh.74	[105]
mdx4cv mice	Ex53 nonsense mutation	Ex52–53 del, HDR or Ex53 ref	SpCas9 or SaCas9	ssAAV6	[49]
Ex23 14bp-del or 1bp-ins	Ex23 out-of-flame	Ex23 ref	CjCas9	ssAAV9	[106]
ΔEx44 mice	Ex45 out-of-frame	Ex45 skip/ref	SpCas9	ssAAV9	[36]
ΔEx44 mice	Ex45 out-of-frame	Ex45 skip/ref	SpCas9	scAAV9	[58]
ΔEx50 mice	Ex51 out-of-frame	Ex51 skip/ref	SpCas9 or SaCas9	ssAAV9	[34, 95]
ΔEx43, ΔEx45, ΔEx52 mice	Ex44, Ex46, Ex53 out-of-frame	Ex44 skip/ref for ΔEx43 and ΔEx45, Ex53 skip/ref for ΔEx52	SpCas9	scAAV9	[35]
Humanized ΔEx52 DMD/mdx	Human Ex53 out-of-frame in mdx background	Ex47–58 del	SaCas9	ssAAV9	[107]
Ex20 nonsense mutation mice	Ex20 nonsense mutation	Base editing (A->G)	Base editing ABE	ssAAV9	[59]
ΔEx52 porcine model	Ex53 out-of-frame	Ex51 del	SpCas9	ssAAV9	[48]
ΔEx50 canine model	Ex51 out-of-frame	Ex51 skip/ref	SpCas9	ssAAV9	[57]

Utr; Utrophin

DMPK; dystrophia myotonica protein kinase

Ex; Exon

del; deletion

ins; insertion

skip; skipping

ref; reframing

SaCas9; Staphylococcus aureus Cas9

SpCas9; Streptococcus pyogenes Cas9

CjCas9; Campylobacter jejuni Cas9

ABE; Adenine base editor

ssAAV; single-stranded AAV

scAAV; self-complementary AAV

Out-of-frame deletion of a single or multiple exons accounts for ~70% of total DMD mutations (Fig. 4A). One strategy to correct such mutations is to delete the out-of-frame exon and restore the dystrophin open-reading-frame (ORF). Several studies successfully utilized this strategy in mice to correct DMD mutations in exon 23 and 53 [49–53]. For specific exon deletion strategies, two cooperative DNA DSBs flanking the targeted exon are required (Fig. 4B). However, there is a limitation with this strategy because if one DNA DSB is rejoined by NHEJ repair before the initiation of the second DNA DSB, the targeted exon cannot be completely excised. Moreover, sgRNA-mediated “double-cuts” may also introduce additional unwanted genomic modifications, including transversions and translocations [54].

Fig. 4. Strategies for CRISPR/Cas-mediated correction of DMD mutations.

(A) Schematic illustration showing the arrangement of exons 49–52 of the DMD gene in terms of their reading frame compatibility. This genomic region is used here as an example to highlight the strategies for CRISPR/Cas9 correction of DMD mutations. An out-of-frame deletion of DMD exon 50 results in splicing of exon 49 to exon 51. This creates a premature stop codon in exon 51 (red STOP sign). (B) Exon deletion is used to restore the DMD reading frame. Two sgRNAs targeting introns 50 and 51 will generate two DNA DSBs flanking exon 51. This leads to excision of exon 51 and subsequent splicing of exon 49 to exon 52. (C) Exon skipping is mediated by a single sgRNA, which targets the splice acceptor site of exon 51. The INDELs generated by NHEJ-mediated repair disrupt the splice acceptor sequence (AG dinucleotide) of exon 51, leading to splicing of exon 49 to exon 52. (D) Exon reframing is mediated by a single sgRNA targeting exon 51. The INDELs in exon 51 generated by NHEJ-mediated repair may restore the reading frame compatibility of exon 51 with exons 49 and 52. (E) Base editing can be applied to mutate the “AG” dinucleotide sequence present at the splice acceptor site of exon 51, leading to skipping of the out-of-frame exon 51.

Exon skipping is a suitable strategy to correct mutations within the central rod domain of the dystrophin gene because this domain is mainly composed of spectrin-like repeats, which are tolerant of in-frame skipping. This strategy requires a single sgRNA targeting the splice acceptor or donor sequence. INDELs generated by NHEJ repair can disrupt splicing sequences, leading to exon skipping. Targeting the canonical splice acceptor signal which contains the “AG” dinucleotide sequence at the intron-exon boundary can be used for exon skipping (Fig. 4C). However, disruption of the canonical splicing acceptor signal sometimes may be insufficient to induce complete exon skipping since the splicing machinery may use other “AG” dinucleotide sequences present within the exon. When this occurs, the mutant exon will still be incorporated into the dystrophin transcript, rendering the exon skipping strategy ineffective.

Exon reframing is a CRISPR/Cas-mediated “single-cut” strategy to correct DMD mutations. In contrast to exon skipping in which a single sgRNA targets the intron/exon junction, the sgRNA used for exon reframing targets the exon itself (Fig. 4D). Recent studies discovered that SpCas9-induced DNA DSBs contains a single nucleotide 5’ overhang, which is prone to be filled with one nucleotide by the endogenous DNA repair machinery, leading to a high frequency of one nucleotide insertions [55, 56]. Therefore, exon reframing is an efficient strategy to repair mutations that can be reframed by a 3n+1 nucleotide insertion. Several studies applied the CRISPR/Cas9-mediated reframing strategy and successfully corrected exon 45, 51 and 52 out-of-frame mutations in mice and dogs [34–36, 57, 58].

The development of base editing for single-nucleotide modification brings accuracy and safety to genome editing because base editing does not generate DNA DSBs and hence its editing outcome is more predictable. Base editing is ideal to precisely correct point mutations, which account for ~10% of total DMD mutations [59]. In addition, base editing can also be applied to mutate the “AG” dinucleotide sequence present at the splice acceptor site, leading to skipping of the out-of-frame exon (Fig. 4E).

Facioscapulohumeral Muscular Dystrophy

Facioscapulohumeral muscular dystrophy (FSHD) is a dominant neuromuscular disorder primarily affecting the muscles of the face, shoulders, and upper arms. Two distinct molecular mechanisms cause FSHD, leading to classification of this disease into two subtypes (FSHD1 and 2). FSHD1 accounts for 95% of total FSHD cases and is caused by contraction of highly similar 3.3-kb repeat units, known as D4Z4 repeats on chromosome 4 [60] (Fig. 3). Patients with FSHD1 carry less than 10 copies of D4Z4 repeats, which induces DUX4 expression. DUX4 mRNA transcribed from the last D4Z4 repeat can be stabilized by a poly-adenylation signal present in a permissive haplotype. This leads to production of the double-homeobox transcription factor, DUX4. Under normal conditions, DUX4 is only expressed in the testis. In FSHD1 patients, DUX4 is expressed ectopically in skeletal muscle, which causes apoptosis and oxidative stress. Patients diagnosed with FSHD2 have a mutation in the SMCHD1 gene, leading to hypomethylation of CpG islands on the D4Z4 repeats, which subsequently induces DUX4 expression [61].

Several transgenic mouse models have been generated to study FSHD1 [62–64]. Ectopic overexpression of the human DUX4 gene in mouse skeletal muscle causes dystrophic and degenerative myopathies. However, DUX4 is restricted to primates and its transcriptional binding sites are not conserved in the murine genome. Therefore, discerning analysis and interpretation of the data is needed when using a mouse model to study FSHD disease progression and molecular mechanism.

A CRISPR/dCas9-mediated gene repression strategy was used to reduce DUX4 expression in primary myoblasts isolated from a FSHD patient. In this strategy, the nuclease-dead dCas9 was fused with a KRAB transcriptional repressor and guided by a sgRNA to the promoter region of the DUX4 gene, leading to transcriptional repression [65]. However, this approach does not permanently block DUX4 expression, rendering its long-term therapeutic application ineffective. An alternative approach to use CRISPR/Cas9-based genome editing strategy to treat FSHD directs the nuclease-active Cas9 to the poly-adenylation site on the permissive haplotype and induces a DNA DSB. The subsequent NHEJ-mediated DNA DSB repair will generate INDELs and disrupt the poly-adenylation signal, converting the FSHD permissive haplotype to the nonpermissive haplotype [66] (Fig. 5A).

Fig. 5. Strategies for CRISPR/Cas-mediated correction of other muscular dystrophies.

(A) FSHD is caused by acquisition of a poly-adenylation signal (PAS) after the last D4Z4 repeat, leading to DUX4 transcript stabilization. Cas9 nuclease can be used to target the last D4Z4 repeat and NHEJ-mediated INDELs will subsequently disrupt the PAS sequence. (B) Myotonic dystrophy type I (DM1) is caused by trinucleotide CTG repeat expansion in the 3’ untranslated region (3’-UTR) of the DMPK gene. Two sgRNAs can be designed to generate two DNA DSBs that flank the CTG repeats, leading to deletion of the CTG repeats. (C) Myotonic dystrophy type 2 (DM2) is caused by a tetranucleotide CCTG repeat expansion in intron 1 of the CNBP gene. Two sgRNAs can be designed to generate two DNA DSBs that flank the CCTG repeats, leading to deletion of the CCTG repeats.

Myotonic Dystrophy

Myotonic dystrophy (DM) is an autosomal dominant disease characterized by progressive muscle weakness and wasting. Genetically, DM is caused by a repeat expansion in two different genes and can be classified into DM1 and DM2. DM1 is caused by expansion of a trinucleotide CTG repeat in the 3’ untranslated region (3’-UTR) of the DMPK gene, while expansion of a CCTG tetranucleotide repeat in intron 1 of the CNBP gene leads to DM2 [67] (Fig. 3). Both DM1 and DM2 cause sequestration of splicing factors, such as MBNL1 in the ribonuclear foci, leading to splicing abnormalities, RNA dysregulation and toxicity.

Many mouse models have been developed to study DM. For example, the transgenic DM300–328 mouse model carries a genetic fragment with expanded CTG repeats from the human DMPK locus [68]. Other models, such as the HSA^LR transgenic line carries hundreds of CTG repeats in the 3’-UTR of the human HSA gene [69]. These models recapitulate many pathological phenotypes seen in DM patients, including muscle weakness, myotonia, splicing defects. Therefore, these mouse models are valuable tools to study the molecular mechanisms and progression of DM.

Initial in vitro CRISPR/Cas-mediated correction of DM used the HDR pathway to knock-in multiple copies of poly-adenylation signals upstream of the trinucleotide repeat, thereby preventing inclusion of toxic trinucleotide repeats in the RNA transcripts [70]. Although these studies proved successful, HDR-based correction is not applicable in post-mitotic skeletal muscle. Several in vivo studies used CRISPR/Cas and its derivatives to correct DM in animal models (Table 1). The first study used the CRISPR/dCas system to repress transcription of the gene with expanded CTG repeats [71]. The second study applied a RNA-targeting SpCas9 to eliminate the RNA transcripts with expanded CTG repeats [72]. These two studies diminished DM pathology at the transcriptional or RNA level, while the mutation remains present in the genome. The third study used CRISPR/Cas9-mediated DNA DSBs to completely excise the CTG repeats from the DMPK locus [73]. This strategy requires the NHEJ pathway to repair the DM mutation, representing a promising therapy for DM (Fig. 5B and 5C). However, DM is a toxic gain-of-function muscle disorder and it remains uncertain what percentage of mutation correction at the genomic level is sufficient to provide therapeutic benefit in this disease.

Limb-girdle Muscular Dystrophy

Limb-girdle muscular dystrophy (LGMD) is a group of autosomal neuromuscular diseases caused by mutations in a broad range of genes. LGMD is characterized by atrophy and weakness of proximal muscles in the hip and shoulder joints. However, the severity and disease progression in each subtype of LGMD varies. For example, LGMD2B is caused by mutation of the DYSE gene, which encodes the large transmembrane protein dysferlin [74]. Dysferlin is critical for membrane repair in skeletal muscles and heart. Mice deficient in dysferlin develop dystrophic phenotypes, including muscle necrosis, inflammation, and fatty replacement of muscle. The dysferlin gene exceeds the maximal packaging capacity of an AAV vector, posing challenges to conventional gene replacement therapy. Some studies used a dual-AAV approach to split the dysferlin cDNA into two segments and packaged each segment into an independent AAV vector [75–77]. The full-length dysferlin gene is then reconstituted either by trans-splicing or homologous recombination.

Sarcoglycanopathies are another group of LGMDs, which are caused by mutations of the sarcoglycan complex. The sarcoglycan complex is composed of α-, β-, γ-, δ-, ε-, and ζ-sarcoglycan. Recessive mutations of the α-, β-, γ-, and δ-sarcoglycan cause LGMD2C, 2D, 2E, and 2F, respectively [78, 79]. The sarcoglycan complex is a key component of the dystrophin glycoprotein complex, which is necessary for maintaining muscle membrane stability and integrity. Similar to patients affected with DMD or BMD, patients affected with sarcoglycanopathies develop skeletal muscle degeneration and cardiomyopathy. Small rodent models of sarcoglycanopathies have been developed [80–85]. In addition, two canine models have been reported to carry mutations in sarcoglycan α and δ, manifesting phenotypes seen in LGMD2C and 2F [86, 87]. These animal models recapitulate many of the pathophysiological phenotypes seen in human patients, and hence serve as reliable animal models of LGMD.

Using induced pluripotent stem cells (iPSCs) derived from LGMD patients, CRISPR/Cas9-mediated in vitro genome editing has been used to correct mutations causing LGMD2B, 2C, and 2D [88, 89]. However, these proof-of-concept studies utilized the HDR DNA repair pathway to correct these mutations, which is not feasible in post-mitotic skeletal muscle and heart. Nevertheless, newer technologies such as CRISPR/dCas9-mediated base editing or prime editing allow precise genome editing without the need for HDR and are under development as strategies to repair LGMD mutations in vivo.

4. CHALLENGES OF THERAPEUTIC GENOME EDITING

Limitations of Animal Models

Animal models of muscular dystrophies have been developed to study disease progression and molecular mechanisms and are valuable tools for optimizing gene therapy. However, genetic differences among different species pose challenges for direct therapeutic translation of animal-based discovery to patients. For example, the D4Z4encoded DUX4 retrogene in FSHD is unique to the primate lineage. To generate a FSHD1 animal model, a DUX4 transgenic mouse model was generated containing the entire human FSHD1 locus insertion. However, this engineered DUX4 mouse model does not develop a dystrophic muscle phenotype, offering limited use as a mouse model to evaluate genetic therapies for FSHD1 patients.

While human and mouse dystrophin orthologs are highly similar in terms of exon composition and amino acid sequence, due to codon degeneracy, the human and mouse dystrophin genes vary at the genomic level. Since CRISPR/Cas-mediated genome editing is guided by the sequence-specific sgRNA, human-specific sgRNAs may not target the mouse genome and in turn cannot be tested in DMD mouse models. As such, genomic sequence variation between humans and mice inevitably interferes with sgRNA screening and optimization for clinical translation. To address these issues, two humanized DMD mouse models were generated by deleting exon 45 and 52 of the human DMD gene and stably integrating them in mouse chromosome 5 [90, 91]. When backcrossed to mdx mice, these humanized DMD mouse models display dystrophic phenotypes, providing informative tools for studying antisense oligonucleotide or CRISPR/Cas-mediated gene therapy. However, a recent study reported that the human DMD transgene in these humanized DMD mouse models is present twice per locus and is organized in a tail-to-tail orientation, complicating the genetic background [92]. A plausible solution is to generate X-linked humanized DMD models in which parts of the mouse Dmd gene are replaced by the human ortholog.

Genome Editing Efficiency and Off-target Effects

Muscular dystrophies are caused by either loss or gain of function mutations in the human genome. The dystrophin gene in DMD and many genes in LGMD are examples of loss of function mutations, while contraction of D4Z4 repeats in FSHD1 and expansion of tri- or tetranucleotide repeats in DM are examples of gain of function mutations. The goal of CRISPR/Cas-mediated correction of muscular dystrophies is to restore the function of mutated genes or to eliminate toxic gene products. No report exists showing in vivo genome editing with 100% efficiency; hence the first challenge in translating CRISPR/Cas genome editing for therapeutic purpose is to improve editing efficiency.

Three strategies have been applied to improve CRISPR/Cas genome editing efficiency. The first strategy is to minimize the number of sgRNAs used in genome editing. Conventionally, two sgRNAs are required to excise an out-of-frame exon from the DMD gene [51–53]. However, this “double-cut” approach is not efficient because productive editing occurs only when two cooperative DSBs are generated around the target genomic region. If the first DNA DSB is repaired by NHEJ, then the second DNA DSB alone cannot excise the entire out-of-frame exon. In contrast, a single sgRNA-induced single DNA DSB at the 5’-end of an out-of-frame exon is sufficient to restore the DMD ORF. This is because the INDELs generated by the “single-cut” approach can either reframe the out-of-frame exon or disrupt the splicing signal, leading to exon skipping. Several studies demonstrated that the single sgRNA-mediated “single-cut” strategy is highly efficient in correcting DMD mutations [34–36, 57, 58].

A second strategy used to improve genome editing efficiency is to optimize the ratio of sgRNA and Cas9 nuclease. Several studies showed that sgRNA abundance is a rate limiting factor in in vivo genome editing. This occurs because of a preferential depletion of sgRNA AAV vector after systemic delivery of AAV-packaged CRISPR/Cas genome editing components [36, 50]. Highly efficient genome editing can be achieved by delivering more sgRNA AAV vectors than Cas9 AAV vectors.

A third strategy used to improve genome editing efficiency is to use self-complementary AAV as a delivery vector for in vivo genome editing. The conventional AAV genome is single-stranded, which is prone to DNA degradation. The self-complementary AAV packages a double-stranded DNA and is resistant to DNA degradation. In addition, self-complementary AAV does not require second-strand synthesis, which is a rate-limiting step for gene expression. Recent studies utilized self-complementary AAV as a vector for sgRNA delivery and demonstrated highly efficient in vivo genome editing in a mouse model of DMD [35, 58].

Complete restoration of normal levels of dystrophin is not achievable for therapeutic gene editing because CRISPR/Cas-mediated gene editing of skeletal muscle is not 100%. Studies in patients with Becker muscular dystrophy (BMD) have estimated that ~15% of normal levels of dystrophin protein could provide therapeutic benefit [93]. Moreover, a recent study showed that very low residual dystrophin protein expression (0.5–5%) can cause a shift in disease phenotype from DMD toward BMD [94]. In most reported studies of CRISPR/Cas-mediated gene editing of DMD, dystrophin restoration across various skeletal muscle groups is over 20% [34, 36, 57, 58, 95]. Therefore, CRISPR/Cas-based therapeutic gene editing represents a promising therapy for DMD.

SpCas9 has been preferentially used for in vivo genome editing of a variety of DMD mouse models, including out-of-frame deletion of exon 43, 44, 45, 50, 51 and 52 [34–37]. However, due to the large protein size of SpCas9, it cannot be packaged with the sgRNA component in a single AAV vector. Therefore, a dual AAV vector system is required to separately deliver the SpCas9 nuclease (AAV-SpCas9) and its sgRNA (AAV-sgRNA). However, the dual AAV delivery system significantly increases the AAV production cost and the total AAV particles administered to patients, bringing economic and safety concerns to therapeutic genome editing. One strategy to address this limitation is to replace SpCas9 with another Cas9 ortholog that is smaller in size, such as SaCas9 [49]. This will permit the AAV vector to accommodate both the Cas9 nuclease and its sgRNA in one vector, potentially improving genome editing efficiency.

In addition to improving genome editing efficiency, off-target effects are another concern in CRISPR/Cas-based therapeutic genome editing. DNA DSBs induced by Cas9 are mediated by base pairing of the sgRNA with the target DNA. It has been reported that Cas9 is capable of inducing DNA DSBs in the presence of a mismatched sgRNA-DNA duplex, leading to off-target effects [96–98]. Several approaches can be used to reduce off-target effects, including screening sgRNAs with high on-target scores and using engineered SpCas9 with high-fidelity [18–20].

Long Term Benefits of CRISPR/Cas Therapeutic Genome Editing

Skeletal muscle is a post-mitotic and terminally differentiated tissue, which is ideal for CRISPR/Cas therapeutic genome editing. A recent study in the DMD mdx mouse model demonstrated that sustained genome editing and dystrophin expression was observed for 18 months after systemic injection of AAV-delivered SaCas9 [50]. However, stem cells residing in skeletal muscle can undergo de novo myogenesis or fuse with pre-existing myofibers [99]. It remains unknown whether a contribution of unedited muscle stem cells to CRISPR/Cas edited myofibers will dilute out the genome editing outcome in skeletal muscle. In contrast, human adult cardiomyocytes have a very low turnover rate [100, 101]. Therefore, CRISPR/Cas therapeutic genome editing in the human heart should provide long-term clinical benefit.

Recent studies showed that AAV serotype 9 can effectively transduce muscle stem cells [102, 103], bringing hope for delivery of CRISPR/Cas genome editing components to stem cell compartments. However, transgene expression tested in these two studies was driven by a ubiquitous promoter, which may cause unwanted gene expression in other tissues or organs. Therefore, developing a tissue-specific promoter with activity in both mature myofibers and muscle stem cells may further advance CRISPR/Cas therapeutic genome editing towards clinical application.

5. CONCLUSIONS AND FUTURE PERSPECTIVES

Greater than 400 genes have been identified to cause neuromuscular disorders [4]. Among these, over 40 genes cause a variety of muscular dystrophies, which significantly affect the quality of life and ultimately shorten life expectancy of patients. Dystrophin was first cloned and shown to cause DMD in 1987 [27]. In the same year, the CRISPR array was discovered in the genome of Escherichia coli, although its function remained elusive until this past decade [11]. From the initial proof-of-concept studies in cell based in vitro editing to large animal based in vivo editing, the CRISPR/Cas system has been shown to be effective in correcting diverse mutations at the genomic level and restoring protein function. Another challenge for clinical translation and commercialization of CRISPR/Cas therapeutic genome editing is the manufacturing cost of the delivery vector. AAV is the most common delivery vector for gene therapy. However, current production costs for systemic AAV treatment are significantly higher than for other treatments. Zolgensma, an FDA-approved AAV-based gene therapy drug to treat SMA, costs over two million dollars per patient per treatment.

Over the past five years, the scientific field has witnessed a significant advancement in CRISPR/Cas genome editing. In addition to the original CRISPR/Cas system, many other genome editing platforms have been developed, including RNA targeting, base editing, and prime editing. Despite challenges, CRISPR/Cas remains an efficient, precise and accurate tool for genomic manipulation. The CRISPR/Cas system has revolutionized the gene therapy field and represents a promising therapeutic approach to treat the underlying genetic causes of diverse muscular dystrophies.