Integrative Biomedical Research

1. Introduction

There is something quietly devastating about watching a disease unfold on a schedule. Duchenne muscular dystrophy (DMD) does exactly that. It is an X-linked recessive neuromuscular disorder, and it affects roughly 1 in 3,500 to 1 in 5,000 boys born worldwide - not a rare curiosity, but a recurring tragedy that clinicians and families have learned, painfully, to anticipate (Ali et al., 2024; Choi & Koo, 2021; Lim et al., 2020; Siddika et al., 2026). At the root of it lies a single gene. The DMD gene sits on chromosome Xp21.2-p21.1, and at roughly 2.2 to 2.4 megabases across 79 exons, it happens to be the largest gene in the human genome - a fact that is almost ironic, given how much can go wrong across that much genomic real estate (Chen et al., 2022; Łoboda et al., 2025; Saad et al., 2023). When everything works, that gene produces a 14-kilobase transcript encoding dystrophin, a 427-kDa rod-shaped protein tucked beneath the muscle cell membrane (Pickar-Oliver et al., 2021; Saad et al., 2023). Dystrophin is not glamorous. It does not catalyze anything. What it does, instead, is hold things together - linking the actin cytoskeleton inside the cell to the extracellular matrix outside it, through the dystrophin-associated protein complex, so that muscle fibers can absorb the mechanical stress of contraction without tearing themselves apart (Chen et al., 2022; Pickar-Oliver et al., 2021; Siddika et al., 2026).

Take dystrophin away, and that shock-absorbing function disappears with it. The sarcolemma becomes fragile and abnormally permeable, calcium floods into the cell, and a slow cascade begins: myocyte necrosis, chronic inflammation, and - over years - progressive fibrosis and fatty replacement of what used to be functional muscle (Brescia et al., 2020; Chen et al., 2022; Siddika et al., 2026). Clinically, this plays out with an almost cruel predictability. Boys typically show early motor delays between two and five years of age; independent walking is usually lost by ten to twelve; and death from cardiac or respiratory failure tends to arrive in the twenties or thirties, sometimes earlier (Laurent et al., 2024; Siddika et al., 2026).

Not every DMD mutation behaves the same way, and this distinction matters more than it might first appear. Large exonic deletions account for roughly 65-72% of cases, duplications for another 10-15%, and point mutations for the remainder, with two mutational hotspots - exons 45-55 and exons 3-9 - doing an outsized share of the damage (Chen et al., 2022; Erkut & Yokota, 2022; Siddika et al., 2026). The critical variable, though, is not simply where the mutation sits but whether it disrupts the reading frame. Out-of-frame mutations abolish dystrophin production almost entirely, producing the severe DMD phenotype; in-frame mutations, by contrast, still permit translation of an internally shortened but partly functional protein, yielding the comparatively milder Becker muscular dystrophy (BMD) (Mollanoori et al., 2020; Siddika et al., 2026). This single observation - that reading-frame restoration, rather than perfect sequence correction, can convert a severe disease into a mild one - has quietly shaped almost every therapeutic strategy discussed in this review.

For a long time, treatment for DMD meant corticosteroids - prednisone, deflazacort - which slow the rate of motor decline without touching the underlying genetic lesion, and which carry their own long-term costs in bone fragility, stunted growth, and metabolic disturbance (Ali et al., 2024; Chen et al., 2022; Łoboda et al., 2025). More recently, antisense oligonucleotide (ASO) drugs such as eteplirsen, golodirsen, viltolarsen, and casimersen have offered a genuinely new mechanism: they induce exon skipping at the RNA level, coaxing the splicing machinery to bypass problematic exons and restore the reading frame (Erkut & Yokota, 2022; Kuroda & Yokota, 2026; Łoboda et al., 2025). The trouble is that this correction is transient by design - it has to be re-administered weekly, for life, by intravenous infusion - and even then, dystrophin restoration in skeletal and especially cardiac muscle tends to remain modest, often under 5% of normal levels (Chen et al., 2022; Erkut & Yokota, 2022; Siddika et al., 2026). AAV-delivered micro-dystrophin gene replacement, exemplified by delandistrogene moxeparvovec, sidesteps the mutation-specific problem entirely by supplying a synthetic, heavily truncated dystrophin construct - but truncation means the protein necessarily lacks several structural repeats present in the wild-type molecule, and long-term durability data are still accumulating (Chen et al., 2022; Łoboda et al., 2025; Montagna et al., 2025).

CRISPR-Cas genome editing changes the terms of the problem. Rather than intervening downstream, at the level of RNA or protein, it acts directly on genomic DNA - which means, at least in principle, a single treatment could produce a permanent, endogenously regulated correction rather than a therapy that has to be repeated indefinitely (Choi & Koo, 2021; Erkut & Yokota, 2022; Laurent et al., 2024). A single-guide RNA (sgRNA) directs a Cas endonuclease - most often Streptococcus pyogenes Cas9 (SpCas9) or the smaller Staphylococcus aureus Cas9 (SaCas9) - to a specific genomic site adjacent to a protospacer adjacent motif (PAM), where it introduces a targeted double-strand break (Chen et al., 2022; Erkut & Yokota, 2022). Because skeletal and cardiac myocytes are postmitotic, they repair these breaks predominantly through non-homologous end joining (NHEJ) rather than homology-directed repair, and this constraint has, somewhat unexpectedly, become an opportunity: NHEJ can be harnessed for single-cut exon skipping, double-cut deletion of entire mutational hotspots, or homology-independent targeted integration (HITI) of missing exons (Choi & Koo, 2021; Pickar-Oliver et al., 2021; Siddika et al., 2026). Newer, double-strand-break-free platforms - adenine and cytosine base editors, prime editing, and dCas9-based CRISPR activation of the autosomal paralog utrophin (UTRN) - extend this toolkit further still, offering ways to correct point mutations or compensate for dystrophin loss altogether without cutting the genome at all (Chen et al., 2022; Erkut & Yokota, 2022; Laurent et al., 2024).

It would be convenient if the story ended there - a permanent molecular fix, ready for the clinic. It does not. Proof-of-concept has been demonstrated repeatedly, across cell culture, rodent, canine, and porcine models of DMD, yet translating any of these strategies into humans runs into a set of stubborn, interconnected barriers around delivery, safety, and manufacturing scale (Ali et al., 2024; Chen et al., 2022; Laurent et al., 2024). Striated muscle is not a small target; it makes up well over 40% of total body mass, which means an effective in vivo therapy has to reach muscle tissue essentially everywhere in the body, not just at a local injection site (Chen et al., 2022; Łoboda et al., 2025). Recombinant AAV vectors - AAV8, AAV9, AAVrh74, and engineered myotropic variants such as MyoAAV - dominate this space because of their natural affinity for muscle, but they carry a hard physical packaging ceiling of roughly 4.7 kilobases, and SpCas9 alone already occupies about 4.1 kb before a guide RNA or promoter is even added (Chen et al., 2022; Happi Mbakam et al., 2022; Łoboda et al., 2025). Dual-vector systems, split-intein trans-splicing, and smaller Cas orthologs such as SaCas9 or Campylobacter jejuni Cas9 have all been proposed as workarounds, yet each still requires systemic vector doses in the range of 1 x 10^14 to 1.8 x 10^15 vector genomes per kilogram to achieve adequate body-wide transduction (Chen et al., 2022; Erkut & Yokota, 2022; Zhang et al., 2021).

Doses of that magnitude are not without consequence. High-dose systemic AAV administration has been associated with acute hepatotoxicity, dorsal root ganglion sensory neuron toxicity, complement activation, thrombotic microangiopathy, acute respiratory distress syndrome, and - in a small number of tragic clinical cases - fatal multi-organ failure (Ali et al., 2024; Łoboda et al., 2025; Laurent et al., 2024). Pre-existing neutralizing antibodies against AAV capsids, already present in a substantial fraction of the human population, block transduction outright and rule out re-dosing (Chen et al., 2022; Łoboda et al., 2025; Montagna et al., 2025). Persistent expression of bacterial-derived Cas proteins invites its own immune response, with cytotoxic T-cells capable of clearing the very myofibers that were just corrected (Chen et al., 2022; Choi & Koo, 2021; Happi Mbakam et al., 2022). And at the genomic level, prolonged nuclease activity raises the risk of off-target cleavage and of unintended capture of viral vector fragments at Cas9-induced break sites (Brescia et al., 2020; Chen et al., 2022; Choi & Koo, 2021). Non-viral alternatives - lipid nanoparticles, extracellular vesicles, and inorganic gold nanoparticles - sidestep much of this immunogenicity by delivering transient Cas ribonucleoprotein or mRNA rather than a persistent viral genome, though at the cost of comparatively poor delivery efficiency to deep skeletal muscle and rapid clearance by the liver and spleen (Farzi et al., 2026; Kuroda & Yokota, 2026).

Given how tightly these problems are linked - a molecular strategy is only as good as the vector that can deliver it, and a vector is only useful if it can be given safely at scale - this review sets out four connected objectives.

First, it evaluates the major CRISPR-Cas editing modalities available for DMD correction, comparing their molecular mechanisms, targeting efficiencies, and structural outcomes across single-cut and double-cut NHEJ, HITI/HDR knock-in, base editing, prime editing, and CRISPR activation (Chen et al., 2022; Erkut & Yokota, 2022; Siddika et al., 2026). Second, it critically examines viral and non-viral delivery vector platforms - AAVs, high-capacity adenoviral vectors, lipid nanoparticles, and extracellular vesicles - with attention to their transduction efficiency, tissue tropism, and packaging capacity for postmitotic skeletal, cardiac, and muscle stem cell populations (Brescia et al., 2020; Farzi et al., 2026; Łoboda et al., 2025; Siddika et al., 2026). Third, it investigates the safety, immunological, and genotoxic liabilities associated with systemic delivery, including dose-dependent organ toxicity, anti-AAV and anti-Cas immunity, off-target mutagenesis, and vector DNA capture at cleavage sites (Ali et al., 2024; Brescia et al., 2020; Chen et al., 2022; Łoboda et al., 2025). Finally, it identifies emerging scalability solutions and translational roadmaps - hyper-myotropic capsids, self-complementary AAV designs, transient non-viral platforms, targeted immunosuppression, and scalable manufacturing pipelines - that may help move CRISPR therapeutics from preclinical promise toward safe, accessible, and durable clinical treatment for patients with DMD (Ali et al., 2024; Chen et al., 2022; Farzi et al., 2026; Laurent et al., 2024; Siddika et al., 2026).

2. CRISPR-Cas Gene Editing for Duchenne Muscular Dystrophy: Mechanisms, Delivery, and Safety

What follows is organized around the same logic that seems to run through the field itself: first understand what CRISPR can, in principle, do to a mutated DMD allele; then ask how that molecular machinery actually gets into a patient's muscle; then confront what happens, biologically, when it does; and finally, look at the models researchers have relied on to answer all of the above. Four tables (Tables 1-4) and three schematic figures (Figures 1-3), presented together after the Conclusion, accompany this synthesis.

2.1 Molecular Mechanisms of CRISPR-Cas Editing Strategies

The mutational diversity described in Section 1.1 is, in a sense, the organizing constraint for everything in this subsection - different lesions call for genuinely different molecular tools, and Figure 1 traces how mutation type ultimately determines phenotype before any therapy is even considered.

Single-cut approaches are, in some sense, the most conservative entry point into this toolkit. A single sgRNA guides Cas9 to an exon-intron junction or to an exonic sequence itself, and the resulting NHEJ-mediated indel either disrupts a splice site - forcing the spliceosome to skip the offending exon - or, when placed within an out-of-frame exon, has roughly a one-in-three statistical chance of restoring the reading frame directly (Choi & Koo, 2021; Siddika et al., 2026). It is not a particularly elegant mechanism, relying as it does on the stochastic outcome of NHEJ repair, but its minimal vector footprint - just one guide RNA - makes it attractive from a packaging standpoint, and it has restored dystrophin to a striking degree in canine models (Table 1) (Chen et al., 2022; Choi & Koo, 2021).

Double-cut strategies push this logic further by pairing two sgRNAs that flank an entire mutational hotspot - most notably the exon 45-55 region, which alone accounts for correction potential in over 60% of DMD patient genotypes (Chen et al., 2022; Siddika et al., 2026). NHEJ then rejoins the two outer cut ends, deleting everything in between; done carefully, this can also generate 'hybrid exons' that preserve the structural phasing of the spectrin-like repeats in dystrophin's central rod domain, yielding a Becker-like protein with reasonable mechanical stability (Happi Mbakam et al., 2022). The obvious trade-off is that two simultaneous double-strand breaks carry a higher risk of large deletions, inversions, and chromosomal rearrangement than a single cut does (Brescia et al., 2020; Chen et al., 2022).

For missing exons rather than simple frame disruptions, homology-directed repair (HDR) would be the textbook solution - except that postmitotic skeletal and cardiac myocytes divide too rarely for HDR to operate efficiently (Pickar-Oliver et al., 2021; Siddika et al., 2026). Homology-independent targeted integration (HITI) was developed specifically to get around this limitation, co-delivering Cas9 with a donor template flanked by matching Cas9 cleavage sites so that NHEJ machinery - which remains active in non-dividing cells - inserts the donor directionally into the genome. This has enabled knock-in of a full human exon 52, or even larger multi-exon 'superexons,' restoring near full-length dystrophin in humanized mouse models (Pickar-Oliver et al., 2021; Siddika et al., 2026).

A separate branch of the field has tried to avoid double-strand breaks altogether. Base editors - adenine base editors (ABEs) and cytosine base editors (CBEs) - fuse a catalytically impaired Cas9 nickase to a deaminase enzyme, enabling direct A-to-G or C-to-T conversions without cutting both DNA strands, which can correct nonsense point mutations or disrupt splice sites to trigger exon skipping (Choi & Koo, 2021; Siddika et al., 2026). Prime editing goes a step further still, pairing a nickase-reverse transcriptase fusion with a prime editing guide RNA (pegRNA) to write essentially any small insertion, deletion, or substitution directly into the genome without a donor template or a double-strand break (Siddika et al., 2026). And for patients whose mutations resist all of the above, CRISPR activation (CRISPRa) offers a mutation-agnostic alternative: a catalytically dead Cas9 (dCas9) fused to transcriptional activators such as VP64 is targeted not at DMD itself but at its autosomal paralog, UTRN, upregulating utrophin to structurally compensate for the missing dystrophin (Choi & Koo, 2021; Siddika et al., 2026). Figure 2 summarizes how these five

Figure 1. Pathophysiological pathway linking DMD mutation type to clinical phenotype. Out-of-frame deletions, duplications, and point mutations (~65-72%, ~10-15%, and ~20-30% of cases, respectively) abolish dystrophin translation, precipitating sarcolemmal fragility, pathological calcium influx, myonecrosis, and progressive fibro-fatty replacement that defines the severe DMD phenotype. In-frame mutations instead preserve a truncated but partly functional dystrophin protein, producing the comparatively milder Becker muscular dystrophy phenotype. This frame-restoration logic underlies nearly every CRISPR correction strategy summarized in Table 1 (Chen et al., 2022; Mollanoori et al., 2020; Siddika et al., 2026).

Figure 2. Five principal CRISPR-Cas molecular strategies evaluated for DMD correction, organized by DNA repair mechanism. Single-cut and double-cut approaches rely on error-prone non-homologous end joining (NHEJ); homology-independent targeted integration (HITI) and homology-directed repair (HDR) enable exon knock-in in non-dividing postmitotic cells; base and prime editing avoid double-strand breaks entirely; and CRISPR activation (CRISPRa) bypasses the DMD locus altogether by upregulating the autosomal paralog utrophin (UTRN). All five pathways converge on the same functional endpoint - restored or compensated dystrophin expression in striated muscle - with mechanism-specific evidence detailed in Table 1 (Chen et al., 2022; Choi & Koo, 2021; Pickar-Oliver et al., 2021; Siddika et al., 2026).

mechanistically distinct strategies converge on the same downstream goal, and the underlying evidence for each is compared systematically in Table 1.

2.2 Delivery Vector Platforms: Viral versus Non-Viral Systems

Having a molecular tool that works is, frustratingly, only half the problem. Because striated muscle makes up such a large share of total body mass, any editing strategy intended for systemic use needs a delivery vehicle capable of reaching muscle tissue essentially everywhere in the body, and this requirement narrows the field considerably (Chen et al., 2022; Łoboda et al., 2025). Table 2 and Figure 3 summarize the main platforms considered in the literature and the trade-offs each one carries.

Recombinant adeno-associated viruses - AAV8, AAV9, AAVrh74, and bioengineered myotropic variants like MyoAAV - remain, by a wide margin, the dominant in vivo delivery vehicle for muscle-directed gene editing, largely because of their naturally high tropism for skeletal and cardiac tissue (Łoboda et al., 2025; Siddika et al., 2026). Their packaging capacity, however, is not negotiable: roughly 4.7 kb, which is simply not enough room for SpCas9 (about 4.1 kb) plus a guide RNA and regulatory elements in a single particle (Chen et al., 2022; Łoboda et al., 2025). Researchers have worked around this in a few ways - splitting the payload across dual AAV vectors, using split-intein trans-splicing to reconstitute a full-length protein from two half-proteins, or simply switching to a smaller Cas ortholog such as SaCas9 (~3.2 kb) or Campylobacter jejuni Cas9 (Chen et al., 2022; Erkut & Yokota, 2022; Laurent et al., 2024). None of these solutions is free; each adds either genetic complexity or a somewhat lower editing efficiency relative to full-length SpCas9.

High-capacity adenoviral vectors (HC-AdVs) take a different approach entirely. Stripped of essentially all native viral coding sequence, they offer up to 36 kb of packaging space - enough for full-length SpCas9, multiple guide RNAs, and reporter constructs, all in a single particle (Brescia et al., 2020). Genetically retargeted variants bearing CD46-binding fiber motifs have achieved high transduction efficiency in human myogenic progenitor cells while remaining markedly less cytotoxic than earlier-generation adenoviruses, though they still lack the natural, evolved muscle tropism that makes AAV so effective (Brescia et al., 2020).

Non-viral platforms represent, in many ways, the opposite philosophy - trading some delivery efficiency for a substantially better safety profile. Lipid nanoparticles (LNPs), extracellular vesicles (EVs/exosomes), and inorganic gold nanoparticles (CRISPR-Gold) all deliver Cas9 as mRNA or as a pre-assembled ribonucleoprotein (RNP) complex, which is inherently transient and non-integrating (Andreana et al., 2021; Farzi et al., 2026; Kuroda & Yokota, 2026). This transience is, in fact, the main selling point: it drastically lowers immunogenicity and, because there is no persistent viral genome to raise neutralizing antibodies against, it opens the door to repeated dosing in a way that AAV simply cannot offer (Farzi et al., 2026; Kuroda & Yokota, 2026). The cost is systemic delivery efficiency - non-viral particles are preferentially cleared by the liver and spleen before they ever reach deep skeletal muscle, which remains their principal limitation for body-wide correction (Farzi et al., 2026).

2.3 Safety, Immunological, and Genotoxic Liabilities

It is tempting, reading the preclinical literature, to focus on efficacy and treat safety as a secondary concern - but the clinical record suggests otherwise, and Table 4 catalogs four interlocking liabilities that recur across nearly every systemic AAV-CRISPR program reported to date.

The first and most immediately dangerous is dose-dependent toxicity. Achieving meaningful body-wide transduction requires vector doses on the order of 1 x 10^14 to 1.8 x 10^15 vg/kg, and at that scale, human trials have reported acute hepatotoxicity, dorsal root ganglion sensory neuron damage, systemic complement activation, thrombotic microangiopathy, acute respiratory distress syndrome, and - in a small but sobering number of cases - fatal multi-organ failure (Ali et al., 2024; Łoboda et al., 2025; Laurent et al., 2024). Second, humoral and cellular immunity against the therapy itself poses a durability problem as much as a safety one: pre-existing neutralizing antibodies against AAV capsids are present in an estimated 30-50% of the human population, blocking initial transduction and ruling out any possibility of re-dosing, while memory T-cells against bacterial Cas proteins can mount a cytotoxic response that clears the very myofibers that were successfully edited (Chen et al., 2022; Happi Mbakam et al., 2022; Łoboda et al., 2025). Third, genotoxicity remains an open concern at the DNA level - Cas endonucleases can tolerate mismatches at near-cognate genomic sites, and persistent nuclease

Figure 3. Comparative delivery architecture for CRISPR cargo in Duchenne muscular dystrophy. Viral platforms (recombinant AAV, high-capacity adenovirus) achieve efficient muscle transduction but are constrained by packaging capacity or innate/adaptive immunogenicity; non-viral platforms (lipid nanoparticles, extracellular vesicles, CRISPR-Gold) substantially reduce immune risk and permit repeat dosing but face rapid hepatic and splenic clearance that limits deep-muscle delivery. All three vector classes converge on a shared translational bottleneck of systemic safety, manufacturing scalability, and long-term durability, elaborated further in Table 2 and Table 4 (Andreana et al., 2021; Brescia et al., 2020; Farzi et al., 2026; Łoboda et al., 2025).

expression raises the cumulative probability of off-target cleavage, chromosomal inversion, or larger structural rearrangement over time (Ali et al., 2024; Laurent et al., 2024). Fourth, and perhaps least anticipated, the double-strand breaks that Cas9 creates can act as 'sticky ends' that capture fragments of the recombinant viral vector itself, integrating pieces of AAV genome directly into the cleavage site - a phenomenon with genomic consequences that are not yet fully understood (Brescia et al., 2020; Chen et al., 2022).

2.4 Preclinical Animal and Cellular Models

None of the mechanisms or vectors described above would mean much without a credible way to test them, and the field has assembled a fairly deliberate hierarchy of models, summarized in Table 3, that trades convenience for physiological fidelity as one moves up the ladder.

Rodent models - the classical mdx mouse and various CRISPR-generated exon-deletion strains - sit at the accessible end of this spectrum: inexpensive, well-characterized, and easy to breed in numbers large enough for statistically powered studies, though their disease course is notably milder than the human condition, thanks to a comparatively robust capacity for muscle regeneration that partly masks the underlying pathology (Lim et al., 2020; Siddika et al., 2026). Humanized mouse models, which carry a full-length human DMD transgene engineered onto an mdx background, close part of that gap by allowing direct testing of human sequence-specific guide RNAs and donor templates in vivo (Pickar-Oliver et al., 2021). Large animal models - most notably the delta E50-MD dog and Golden Retriever muscular dystrophy (GRMD) canine lines, along with emerging porcine models - offer something rodents cannot: cardiac pathology, body mass, and systemic vector distribution kinetics that meaningfully resemble a human patient, making them the standard bridge before any first-in-human dosing decision (Ali et al., 2024; Laurent et al., 2024; Siddika et al., 2026). Finally, patient-derived induced pluripotent stem cells (iPSCs) and myoblasts provide a human genetic background for high-throughput guide RNA screening and off-target profiling, without the ethical or logistical burden of an animal study, though 2D or even 3D-engineered culture still cannot fully recapitulate whole-organ physiology (Brescia et al., 2020; Choi & Koo, 2021; Laurent et al., 2024).

3. Methods

Because this is a narrative synthesis rather than a formal systematic review, it does not carry a registered protocol - but reproducibility still matters, so the search and selection process is described here in enough detail that another reviewer could, in principle, retrace the same steps and arrive at a comparable evidence base.

3.1 Search Strategy and Information Sources

A structured literature search was conducted across PubMed/MEDLINE, Scopus, Web of Science, and Google Scholar for records published through early 2026, with no lower date limit imposed, given that foundational CRISPR-DMD studies date back to the mid-2010s and remain mechanistically relevant. Search strings combined controlled vocabulary (MeSH terms in PubMed, where available) with free-text keywords, connected using Boolean operators in the general form: ("Duchenne muscular dystrophy" OR "DMD") AND ("CRISPR" OR "CRISPR-Cas9" OR "gene editing" OR "genome editing" OR "base editing" OR "prime editing" OR "CRISPR activation") AND ("delivery" OR "AAV" OR "adeno-associated virus" OR "adenoviral vector" OR "lipid nanoparticle" OR "extracellular vesicle" OR "exosome" OR "non-viral" OR "immunogenicity" OR "safety" OR "toxicity"). Reference lists of retrieved reviews and primary studies were additionally hand-searched (a snowball or citation-chaining approach) to capture foundational preclinical reports that might not surface through keyword searching alone - which is, admittedly, a somewhat old-fashioned method, but one that reliably catches papers that database indexing tends to miss.

3.2 Eligibility Criteria

Records were considered eligible if they (a) were published in a peer-reviewed, English-language journal; (b) reported primary preclinical (in vitro, cell-based, or animal model) or clinical data on CRISPR-Cas-mediated genome editing, base editing, prime editing, or CRISPR activation directed at DMD or its animal/cellular disease models; or (c) were narrative or systematic reviews, mechanistic studies, or methodological papers directly relevant to CRISPR delivery vector engineering, immunogenicity, or genotoxicity in the context of muscular dystrophy. Studies addressing antisense oligonucleotide or small-molecule DMD therapies were retained only when needed for comparative context against gene editing approaches, consistent with the review's stated objectives. Conference abstracts without peer-reviewed full text, non-English publications, and preprints that had not undergone peer review at the time of screening were excluded, in keeping with standard narrative-review practice for evidence intended for clinical audiences.

3.3 Study Selection and Data Extraction

Titles and abstracts identified through the search strategy were screened for topical relevance against the eligibility criteria above; full texts of potentially relevant records were then retrieved and assessed in full. For each included study, the following data were extracted, where reported: editing modality and Cas ortholog or effector used; target exon(s) or mutation type; delivery vector and administration route; vector dose (vg/kg or equivalent); model system (in vitro, rodent, large animal, or human-derived); dystrophin restoration outcome (percentage of wild-type levels, where quantified); reported adverse events or immunological findings; and off-target or genotoxicity assessment methodology. Data were organized narratively by theme - editing mechanism, delivery platform, safety profile, and model system - rather than pooled quantitatively, since the heterogeneity of outcome measures, dosing regimens, and model species across the included literature precluded meaningful meta-analytic synthesis. This thematic organization directly informed the structure of Tables 1 through 4 and the accompanying schematic figures.

3.4 Quality Considerations and Limitations of the Approach

As a narrative rather than systematic review, this work does not apply a formal risk-of-bias instrument (such as SYRCLE's tool for animal studies) across every included record, and estimates of effect - dystrophin restoration percentages, indel efficiencies, and the like - are reported as originally published, without independent statistical re-analysis or correction for multiple comparisons across studies. Readers should weigh preclinical findings, particularly those from small-animal studies with limited sample sizes, with appropriate caution before extrapolating to expected human clinical performance. Where a finding is reported in only a single primary source, this is noted explicitly in the text rather than treated as independently replicated evidence.

4. A Synthesis of Molecular, Delivery, and Safety Findings Across the Included Literature

Pulling the included literature together, a fairly coherent picture emerges - not a simple one, but coherent - in which the choice of editing modality, delivery vector, and preclinical model are not independent decisions so much as three sides of the same translational problem. The findings below are organized accordingly.

4.1 Editing Efficiency Is Highly Mutation- and Modality-Dependent

Across the studies synthesized in this review (Table 1), no single CRISPR modality emerged as universally superior; rather, efficacy tracked closely with the specific mutation being targeted. Single-cut exon-skipping approaches, despite depending on the somewhat unpredictable outcome of NHEJ repair, restored dystrophin to as much as 92% of normal levels in cardiac muscle of canine DMD models when targeting exon 51 with a compact Cas ortholog (Table 1). Double-cut deletion of the exon 45-55 hotspot proved broadly applicable - relevant, in principle, to well over 60% of DMD genotypes - though multiplexed guide delivery via high-capacity adenoviral vectors achieved somewhat more modest allele correction, on the order of 42%, in human myoblasts (Table 1; Table 2). Homology-independent targeted integration, meanwhile, was the only approach capable of restoring near full-length dystrophin rather than an internally truncated version, but this came at the cost of requiring dual-vector co-delivery of Cas9, guide RNA, and donor template simultaneously, a logistical and packaging burden not shared by the single-cut strategies (Table 1). Base and prime editing outcomes were, frankly, more variable across studies - prime editing restored functional dystrophin expression to roughly 25-40% of healthy levels in human iPSC-derived myoblasts carrying an exon 51 deletion, a meaningful result, though still below full physiological restoration (Table 1).

4.2 Vector Choice Trades Delivery Efficiency Against Immunological Risk

Perhaps the clearest pattern to emerge from the delivery-focused literature (Table 2) is an almost linear trade-off: the vectors best at reaching muscle tissue tend to be the ones associated with the greatest immunological and toxicological risk, while the safest platforms are, so far, the least efficient at systemic muscle delivery. AAV9 and engineered myotropic variants achieved the highest reported in vivo transduction efficiency of any platform reviewed, but only at vector doses that, in human trials, triggered severe adverse events including thrombotic microangiopathy and, in the most serious reported cases,

Table 1: Overview of Major CRISPR-Cas Gene-Editing Strategies for DMD Correction. This table compares the six principal molecular editing approaches applied to DMD across their underlying DNA repair mechanism, the mutation type or genomic hotspot each best addresses, their key translational advantages and limitations, and a representative preclinical or clinical outcome with the corresponding citation. It is intended to help readers match a specific patient mutation profile to the editing strategy most likely to be effective, while highlighting the trade-offs each strategy still carries.

Table 2: Comparison of Delivery Vector Platforms for In Vivo and In Vitro DMD Gene Editing. This table summarizes six viral and non-viral delivery systems used to transport CRISPR cargo into muscle tissue, comparing their packaging capacity, tissue tropism and delivery efficiency, immunogenicity and toxicity profile, capacity for repeat dosing, and supporting citations. It is intended to clarify why no single vector currently satisfies all translational requirements simultaneously, and to show how delivery efficiency and safety tend to trade off against one another across platforms.

fatal outcomes (Table 2; Table 4). High-capacity adenoviral vectors offered markedly larger packaging capacity - enough to deliver full-length Cas9 and multiple guides in a single particle - and lower cytotoxicity than earlier adenoviral generations, but still lacked the innate muscle tropism that makes AAV so effective without genetic retargeting (Table 2). Non-viral platforms sat at the opposite end of the risk spectrum: lipid nanoparticle and extracellular vesicle delivery of Cas9 mRNA or ribonucleoprotein achieved comparatively low but non-zero systemic muscle delivery, restoring dystrophin in roughly 4-5% of muscle fibers in one reported exon 44 deletion mouse model, while essentially eliminating the risk of viral genome integration and permitting repeated dosing that AAV simply cannot support (Table 2).

4.3 Safety Liabilities Cluster Around Four Recurring Mechanisms

Synthesizing the safety literature (Table 4) surfaced four recurring, and often interacting, liabilities rather than a single dominant risk. High-dose systemic AAV toxicity was the most acute and clinically visible, directly implicated in the trial pauses and regulatory holds referenced throughout the included literature. Pre-existing and treatment-induced immunity - both humoral, against AAV capsids, and cellular, against bacterial Cas proteins - emerged as the primary obstacle to durable, repeatable treatment, since roughly 30-50% of the population already carries neutralizing anti-AAV antibodies before ever receiving a dose. Genotoxicity, encompassing both off-target cleavage and the unexpected capture of viral vector DNA fragments at Cas9-induced break sites, was reported less frequently in absolute terms but was consistently flagged by regulatory-facing studies as requiring unbiased genome-wide screening (such as GUIDE-seq or CIRCLE-seq) before any human dosing decision. A fourth, more insidious pattern - the gradual dilution of edited dystrophin expression as unedited muscle satellite cells regenerate new, uncorrected myofibers over years of normal muscle turnover - appeared across multiple long-term follow-up discussions, suggesting that durability, not just initial efficacy, deserves more attention than it has typically received (Table 4).

4.4 Preclinical Models Show a Fidelity-Accessibility Trade-Off

Finally, the evidence drawn from animal and cellular models (Table 3) reflected a consistent trade-off between practical accessibility and physiological fidelity to human disease. Rodent models remained the workhorse for initial guide RNA screening and mechanistic proof-of-concept, precisely because of their low cost and short breeding cycle, but their comparatively mild phenotype and robust regenerative capacity appeared to systematically understate the severity of human DMD pathology. Large animal models - canine and porcine, in particular - more closely mirrored human cardiac involvement, systemic vector pharmacokinetics, and immune responses, and were consistently positioned in the literature as the necessary bridge before first-in-human dosing, despite their considerably higher cost and logistical burden. Human iPSC-derived and patient myoblast models occupied a useful middle ground, offering direct relevance to human genetic sequence without the ethical or resource demands of a large animal study, though even 3D-engineered culture systems could not fully reproduce whole-organ hemodynamics or systemic immune surveillance.

5. Reconciling Molecular Promise with Delivery Reality

Taken as a whole, this body of literature tells a story that is more nuanced than either the early enthusiasm around CRISPR or the more recent caution following high-profile trial complications would suggest on their own. The editing science, broadly speaking, works - not perfectly, and not for every mutation, but well enough across multiple independent modalities (Table 1) that molecular correction is no longer the primary obstacle to treating DMD. Delivery is.

5.1 The Packaging Problem Is Not Going Away Soon

It is worth pausing on just how persistent the AAV packaging constraint has proven to be. Nearly a decade of engineering effort - dual vectors, split inteins, smaller Cas orthologs - has produced workable solutions, but none that fully escapes the underlying trade-off between packaging capacity and the natural muscle tropism that makes AAV clinically attractive in the first place (Table 2). High-capacity adenoviral vectors solve the size problem outright, yet the literature reviewed here suggests they have not yet matched AAV's efficiency at reaching muscle without additional retargeting engineering (Table 2; Table 4). It may be that no single vector platform will end up being the answer; the more likely path, based on the trajectory of this literature, involves matching vector choice to clinical context - perhaps AAV-based single-cut editing

Table 3: CRISPR-Engineered Animal and Cellular Models of Duchenne Muscular Dystrophy. This table catalogs six model systems used to evaluate CRISPR-based DMD therapeutics, from rodents through large animals to human-derived cells, detailing the genetic modification used to create each model, its key pathological features, its primary research applications, and a balanced summary of its advantages and limitations. It is intended to guide selection of an appropriate model for a given stage of preclinical development, from initial guide RNA screening through to pre-clinical safety and dosing studies.

Table 4: Key Biological, Safety, and Regulatory Bottlenecks in Clinical Translation of DMD Gene Editing. This table identifies six interlocking translational challenges that currently constrain the clinical development of CRISPR-based DMD therapeutics, linking each to its underlying biological cause, its observed clinical impact, emerging mitigation and engineering strategies, and its current regulatory or development status, together with supporting citations. It is intended to serve as a practical checklist of the safety and scalability issues that any translational program must address before, and during, first-in-human dosing.

for mutation-specific, one-time correction, and non-viral RNP delivery for applications where repeated dosing or lower immunogenicity matters more than maximal efficiency.

5.2 Immunogenicity May Be the More Tractable Problem, Paradoxically

Somewhat counterintuitively, the immunological barriers documented in Table 4 may be more addressable in the near term than the packaging constraint. Transient immunosuppression protocols, plasmapheresis prior to dosing, and epitope-engineered 'immunosilenced' Cas variants are all active areas of engineering effort, and non-viral delivery sidesteps much of the anti-capsid antibody problem by design (Table 2; Table 4). The deeper concern - cellular immunity against bacterial Cas proteins - seems harder to fully eliminate as long as SpCas9 or SaCas9 remain the workhorse effectors, though muscle-restricted promoters and transient expression windows offer partial mitigation. Whether these engineering solutions will prove sufficient to permit true re-dosing, rather than merely reducing the severity of a first administration, remains, honestly, an open question that this literature has not yet fully answered.

5.3 Durability and the Satellite Cell Problem Deserve More Attention

One finding that struck us as somewhat under-discussed relative to its clinical importance is the gradual dilution of corrected dystrophin expression as unedited satellite cells regenerate new myofibers over years of normal muscle turnover (Section 4.3; Table 4). A treatment that looks highly effective at six or twelve months post-dosing could, in principle, show declining benefit a decade later - and because DMD is a pediatric-onset disease with a young patient population and a long expected disease course, this durability question is not a minor technicality. Engineering AAV capsids with tropism for muscle satellite cells specifically, rather than mature myofibers alone, appears in the literature as one proposed solution, though it remains largely at a conceptual or early preclinical stage.

5.4 Model Selection Shapes What We Think We Know

It is also worth being honest about how much of the encouraging efficacy data summarized in Table 1 comes from rodent models whose disease course, as noted in Section 4.4, systematically understates human DMD severity. This is not a criticism of the studies themselves - rodent models remain indispensable for rapid, statistically powered screening - but it does argue for interpreting headline efficacy numbers (such as the 92% dystrophin restoration reported in canine cardiac muscle, or the near-90% base-editing correction in human iPSC cardiomyocytes) with an awareness of which model produced them (Table 1; Table 3). Large animal and human-derived data, where available, generally paint a more conservative, though still genuinely promising, picture than mouse data alone.

5.5 Toward a Realistic Translational Roadmap

Synthesizing across all four themes, a plausible near-term translational pathway would likely combine several partial solutions rather than waiting for one decisive breakthrough: hyper-myotropic AAV capsids (or their non-viral equivalents) to lower the effective vector dose needed for systemic correction; DSB-free base or prime editors, where mutation type permits, to reduce genotoxic risk relative to double-strand-break-dependent NHEJ strategies; transient, targeted immunosuppression around the time of dosing rather than open-ended immune suppression; and manufacturing pipelines capable of producing clinical-grade vector at the scale a systemic, body-wide therapy actually requires - a logistical challenge that receives comparatively little attention in the molecular literature but that several included sources flag as a genuine rate-limiting factor for equitable access (Table 4). None of this guarantees success, and the field has already seen how badly things can go when dosing outpaces safety data. But the underlying science, at least based on the literature synthesized here, does not appear to be the limiting factor anymore. Delivery, safety, and scale are.

5.6 Limitations of This Study

This synthesis has the limitations inherent to any narrative review: it does not apply formal risk-of-bias scoring across included studies, it draws on preclinical outcome measures that vary considerably in how they were quantified across labs, and it necessarily reflects the literature available at the time of the search rather than the field's likely near-future state, which is evolving quickly (see Section 3.4). Readers seeking a fully systematic, meta-analytic synthesis of quantitative editing or delivery efficiency outcomes should treat this review as a starting point for that more targeted undertaking rather than a substitute for it.

6. Conclusion

CRISPR-Cas genome editing has, over roughly a decade, shifted DMD therapeutics from symptom management toward the real possibility of permanent, DNA-level correction. This review synthesized molecular strategies, delivery vector platforms, safety liabilities, and preclinical models to show that editing efficacy, while genuinely mutation-dependent, is no longer the primary constraint - delivery is. Packaging limits, systemic immunogenicity, off-target genotoxicity, durability of correction, and manufacturing scalability remain closely linked barriers to clinical translation. Continued progress in hyper-myotropic vector engineering, transient non-viral ribonucleoprotein delivery, double-strand-break-free precision editors, and targeted immunosuppression will likely determine whether CRISPR therapeutics reach DMD patients safely, durably, and equitably in the years ahead.

References