Emerging Gene Correction Strategies for Muscular Dystrophies: Scientific Progress and Regulatory Impact

Zinc finger nucleases (ZFNs)

ZFNs are currently the highly advanced gene editing system. Initially, these were rationally designed taking advantage of key biological properties of zinc finger transcription factors such as their DNA sequence recognition function. Each zinc finger component is composed of ~30 amino acids and is capable of recognising 3 bp of DNA. The overall structure is arranged in tandem repeats of zinc finger motifs, hence allowing the recognition of longer sequences of DNA. The nuclease activity of ZFNs is conferred through fusion to FokI endonuclease’s catalytic domain. This design ensures that enzymatic and subsequent DNA cleavage activities are only targeted to sites recognised by the binding domains of ZFNs. The latter consist of Cys2His2 zinc finger structures, in which a single zinc atom is surrounded by 30 amino acids and is capable of recognising 3 bp of DNA. More often, however, three to six zinc finger units are assembled, which allows the recognition of 9- to 18-bp DNA sequences. This is usually regarded as acceptable for a single locus targeting in a human genome. From a structure-activity relationship, however, full functionality is not achieved unless FokI is presented in a dimerised form so as two DNA cleavage domains are dimerised around the target DNA sequence [19, 20]. In an attempt to maximise the biosafety profile of ZFNs, several methods have been employed to increase specificity. These include limiting the spacer length between the recognition sites of chimeric ZFN subunits [20]. In fact, ZFNs with shorter interdomain linkers connecting the Cys₂-His₂ zinc finger and the nuclease domains were shown to have a restricted activity with a 6-amino acid linker exerting the most selective activity at a target DNA site with a 6-bp spacer [21]. Another approach has been to generate obligate heterodimer nuclease domains with decreased off-target effects [22–24]. This is associated with a relatively weaker interaction between the heterodimer cleavage domains, which would necessitate stronger interactions between each monomer and the target site to achieve a site-specific cleavage while minimising cleavage at weakly bound nontarget sites. This low-affinity high-avidity approach has been proposed as a plausible mechanistic approach behind the site-specific cleavage activity of the newly engineered obligate heterodimer FokI domains. An alternative to the obligate heterodimer method has been the expression of autonomous ZFN pairs, the combined expression of which was as effective as obligate heterodimer ZFN domains at inducing targeted chromosomal deletion in mammalian cells, with reduced toxic effects that are often thought to be linked with unwanted individual ZFN subunits’ cross-reaction [25]. The enhancement of FokI’s enzymatic activity has also been reported, whereby an in vivo evolution-based method was employed to further increase the cleavage activity of FokI. The incorporation of the enhanced domain in heterodimer ZFN structures resulted in a potent product with improved overall cleavage profile [26].

The successful correction of different genetic mutations associated with various diseases, including sickle cell anaemia [27], haemophilia [28], α1-antitrypsin deficiency [29], X-linked severe combined immunodeficiency [30], and, more recently, HIV [31, 32], has led to its application in DMD, in which precise gene editing can be achieved by deleting targeted exons from the dystrophin gene. For instance, using extended Modular Assembly (eMA)/Context-Dependent Assembly (CoDA) methods, it was possible to generate several exon 51 targeted ZFNs. Two of which demonstrated a remarkable activity with mild off-target mutagenic effects in myoblast cells from DMD patients, harbouring a deletion of exons 48 to 50. When implanted into the hind limb of immunodeficient mice, the corrected myoblasts were capable of maintaining dystrophin expression in vivo, with a correct sarcolemmal localisation [33]. This proof-of-concept study has shown that the ZFN-based approach could potentially be adopted under a cell-based therapy approach for DMD, hence holding a promising faster translation into the clinic, considering that the approach is already in clinical trials for ex vivo cell modifications in HIV [32].

Transcription activator-like effector nucleases (TALENs)

TALENs were originally isolated from Xanthomonas bacteria, plant pathogens capable of employing up to 40 effector proteins to circumvent eukaryotic cell defences during a host infection [34]. These are composed of repeated motifs of 33 to 35 amino acid residues, identical with the exception of the 12^th and 13^th residues, which are often known as the repeat-variable di-residues. The latter play a key role in TALEN’s DNA-binding specificity, whereby a different pair of amino acids would exhibit a specific binding to a corresponding nucleotide in the target sequence. For instance, the asparagine (Asn; N)-isoleucine (Ile; I) NI, the histidine (His; H)-aspartate (Asp; D) HD, the Asn-Asn NN or the Asn-lysine (Lys; K) NK, and the Asn-glycine (Gly; G) NG pairs preferentially bind to adenine, cytosine, guanine, and thymine, respectively. These constitute TALENs’ DNA-specific binding domains, which, like ZFNs, are conjugated to nonspecific FokI cleavage domains, hence directing them to the target site for gene editing and subsequent correction of the final protein product. A recognition sequence of 14 to 20 bp, together with an appropriately spaced FokI subunits separated by 12 to 19 bp, will ensure good genomic target recognition with maximal cleavage activity owing to efficient FokI dimerisation [35, 36].

Although not yet being as clinically advanced as ZFNs, the lower cost and ease of production of TALENs have generated an increasing interest in the technology. Recent work in DMD has shown the feasibility of the approach using optimised exon 51 targeted TALEN-encoding plasmids transfected into myoblast cells isolated from two different patients with deletions in exons 48 to 50. A satisfactory gene correction was observed with up to 12.7% and 6.8% of alleles confirmed to have indels in the two treated patient myoblast cell lines, respectively. This correction was further correlated with a good restoration of dystrophin expression at the expected predicted size of ~412 kDa compared to its expression in the wild-type isolated myoblasts [37]. Whole exome sequencing of the successfully corrected TALEN-treated cells revealed no insertions or deletions, except at the exon 51 target locus. Further analysis using the TALE-NT 2.0 Paired Target Site Prediction web server confirmed the nature of the observed single-nucleotide variants occurring. These were in fact related to expected genomic mutations that normally occur during cell clonal expansion, as none of these showed any similarity to the employed TALEN target site with spacers of 1 to 30 bases [37].

Besides the dystrophin gene as a main target in DMD, other genes have received comparable interest as targets for TALEN-based gene editing. Of these, myostatin (MSTN) has been the centre of attention as a member of the transforming growth factor-β family. This is mostly linked to its recognised role in muscle physiology, acting as a negative regulator of skeletal muscle mass. In fact, studies have shown that mutations in the MSTN gene cause an increase in skeletal muscle fibre numbers and sizes, which in turn lead to muscle hypertrophy without any alarming consequences [38–40]. These observations were recently translated into therapeutic interventions, whereby inhibiting the MSTN signalling pathway using pharmacological agents has shown real benefits in DMD but also in other muscle wasting pathologies such as sarcopenia and cancer cachexia [41–45]. A phase I study is currently under way to assess the safety of an anti-MSTN monoclonal antibody in advanced cancer patients with cachexia (NCT01524224). The same approach using a different pharmacological agent, BMS-986089 (Bristol-Myers Squibb), will soon be tested in a first clinical trial involving patients with DMD (NCT02515669). From a molecular therapeutic point view, MSTN gene editing provides a long-term control to switch off MSTN signalling in the wasted affected muscles. The approach was successfully demonstrated recently using a pair of TALEN-expressing plasmids targeting the human exon 2 locus, a highly conserved region within the coding sequence of the MSTN gene. Consistent with previous ZFN-related toxicity studies, the reported MSTN-TALEN targeted system was engineered using obligate heterodimers of the FokI domain to minimise off-target effects often seen with homodimer variants. Initial experiments in the HEK293 cell line revealed that the mutation induced by the TALEN approach was persistent with indels still detected 1 month after transfection using the T7E1 assay, a commonly used enzyme mismatch cleavage method for detecting mutations. A similar finding was reported in four different cell lines from different species including human, bovine, and murine cells. Further investigation in primary myoblast cultures derived from a dysferlin-deficient mouse model as well as from patients with dysferlinopathy confirmed an efficiency of 10.3% to 24.6% of gene editing after treatment [46]. Although most ZFN- and TALEN-based gene editing systems for muscle diseases to date have been engineered based on NHEJ mechanisms, proof-of-concept data are now emerging on TALEN-mediated homology-directed DNA repair (HDR). Targeted integration of dysferlin, a mutation of which is associated with limb-girdle muscular dystrophy (LGMD) type 2B and Miyoshi myopathy, has been shown following cotransfection of HEK293 with MSTN targeted TALEN-expressing plasmid and a donor plasmid expressing dysferlin tagged to an enhanced cyan fluorescent protein under a cytomegalovirus (CMV) promoter [46]. Although this approach is still at its early research and development stage, the results obtained in vitro remain promising and could offer the basis for future TALEN-corrected myoblast transplantation therapy for a number of severe muscular dystrophies.

Clustered regularly interspaced short palindromic repeats (CRISPR)

Recent achievement in the field of gene editing has seen the emergence of CRISPR as a novel correction tool in biomedicine. CRISPRs were first described in prokaryotes as part of an RNA-guided adaptive immune system to protect against foreign intrusions such as viruses and plasmids. In response to these, bacteria and archaea are programmed to incorporate short sequences of foreign DNA at one end of a repeat element, which is often referred to as the CRISPR. The integrated inserts between the repeat elements of prokaryotic CRISPR-associated (Cas) loci hence confer a permanent future mechanism of defence against the past invaders [47]. Three types of CRISPR/Cas systems (I–III) have been described, with differing structural, functional, and mechanistic characteristics. Each system, however, shares a similar sequence arrangement of short repeated units of 30 to 40 nucleotides, separated by a unique “target” specific nonrepeated sequence (a spacer) of equal length [48, 49]. A leader sequence incorporating a promoter is often present, which initiates a unidirectional transcription of CRISPR sequences. Whereas types I and III systems are present in both bacteria and archaea, type II systems have been reported in bacteria only [50]. Type II systems that use a unique Cas protein referred to as Cas9 are of particular interest in medical application. The RNA-guided, nuclease-mediated genome editing of a type II CRISPR system is mediated by Cas9, a nuclease that is directed to the target DNA site by a single-guide RNA recognising a specific locus next to the protospacer adjacent motif. This subsequently creates a double-stranded break, which could be repaired by NHEJ or by a homology-directed mechanism of repair provided that an exogenous plasmid donor is codelivered [51].

In muscular dystrophy, the CRISPR/Cas9-based technology could theoretically be employed to correct germ-line DNA prenatally in one-cell zygotes or as a postnatal treatment. The first approach was recently tested in vivo in mdx zygotes. By injecting a single-guide RNA-guided CRISPR/Cas9 together with an appropriate HDR template, Long et al. demonstrated for the first time the feasibility of the approach to correct a nonsense mutation in exon 23 of the DMD gene [52]. Although the CRISPR/Cas9-treated mdx zygotes produced mice with varying degrees of gene correction, those showing 41% of gene correction by HDR and 83% correction by NHEJ repair mechanisms demonstrated a complete disease-free phenotype with normalisation of dystrophin expression at the histological level [52]. Despite showing some promising results in DMD, a successful germ-line DNA gene editing strategy would (if ever ethically permissible) require a good knowledge of the nature of mutation affecting the maternal disease carrier to allow a site-specific targeted correction. This prior knowledge, however, could be problematic in X-linked DMD in which it is estimated that one third of all mutations in the DMD gene arise spontaneously [53] and hence would not qualify for a Cas9 gene editing-based treatment. This would restrict the approach to the correction of well-characterised known point mutations, which only make 15% of DMD mutations. One alternative approach to consider would be a postnatal DMD gene correction. This was recently demonstrated using multiplexed single-guided RNAs to direct the Cas9 nuclease to mutations at exons 45 to 55 of the dystrophin gene [54], an approach that has the potential to correct more than 60% of mutations in DMD patients owing to the large hotspot mutation deletion achieved. Despite a good restoration of dystrophin expression both in vitro and in vivo following transplantation of the CRISPR/Cas9-treated DMD patient myoblasts into immunodeficient mice, the overall deletion efficiency, however, was less than that obtained following exon 51 deletion [54]. This suggests that the size of the targeted sequence is crucial in dictating Cas9-associated gene editing capacity to mediate an efficient repair by NHEJ or HDR, where a size-dependent decrease in nuclease-mediated gene deletion was previously noted [55]. Although proof-of-concept has eloquently been shown in research settings, CRISPR/Cas9-based gene editing in DMD would only be envisioned in the clinic should appropriate gene delivery methods are employed. Gene transfer systems capable of directing the elements of the CRISPR/Cas9 cassette to the diseased muscles in vivo are needed to ensure a satisfactory DMD gene editing and a full therapeutic outcome. For instance, AAV has previously been shown to be a safe and an effective gene transfer system in some clinical trials for gene replacement therapy [56]. In this regard, the AAV8 and AAV9 serotypes would be useful delivery tools for CRISPR/Cas9-mediated gene editing for DMD, owing to their reported efficient gene transfer to skeletal muscles and the heart following systemic administration [57, 58], which therefore could potentially result in a robust gene correction at the key target tissues harbouring the mutation in DMD patients. Recent studies have demonstrated the efficiency of AAV as a good platform tool for CRISPR/Cas9-mediated gene editing transfer to dystrophic muscles in DMD disease models in vivo. The approach was recently tested following administration by three different systemic routes, including intraperitoneal, intramuscular, and retro-orbital injections in postnatal mdx mice, and proved to be successful with an increasing dystrophin expression reported from weeks 3 to 12 after virus injection. This correlated with an overall improvement in skeletal muscle function [59]. Mutated exon 23 targeted deletion was also demonstrated to restore disease phenotype in both neonatal and adult DMD models in vivo by systemic and local delivery of a CRISPR/Cas9-expressing AAV [60]. A similar viral-mediated transfer of a CRISPR/Cas9 system coupled with paired guide RNAs flanking the mutated exon 23 in DMD proved that AAV is a good gene editing transfer system for local and systemic restoration of dystrophin expression in muscle cells but also myogenic stem cells in diseased muscle in vivo [61].

Oculopharyngeal muscular dystrophy

OPMD is an inherited autosomal dominant, slow-progressing, late-onset degenerative muscle disorder characterised by progressive eyelid drooping (ptosis), swallowing difficulties (dysphagia), and proximal limb weakness. Whereas the incidence in Europe is 1/100,000, the disease has been largely reported in the Bukhara Jew population in Israel (1/700) and the French Canadian population in Quebec (1/000) [62].

The underlying genetic defect behind OPMD is an abnormal expansion of a (GCG)n trinucleotide repeat in exon 1 of the PABPN1 gene, which leads to an expanded polyalanine tract at the N-terminal of the PABPN1 protein (12–17 repeats are often detected in mutant PABPN1 compared with only 10 repeats in the wild-type protein) [62].

From a molecular therapeutic point of view, OPMD is a good candidate for gene editing-based treatment. Previous studies have shown the importance of abrogating intranuclear inclusions (INIs) of mutant PABPN1, the main pathological hallmark of the disease, with neutralising agents capable of binding aggregated mutant proteins. Of these, intrabodies were shown to be effective in reducing mutant PABPN1-associated INIs and restoring a normal pattern of gene expression in a Drosophila OPMD model [63]. Although these show antiaggregate properties by tackling the disease phenotype at the protein level, they do not address the underlying molecular pathology behind it, hence necessitating a repeated administration approach to keep the disease under control. In this regard, gene editing could offer a long-term permanent therapeutic advantage. Allele-specific correction of expPABPN1 can be achieved by CRISPR/Cas9-mediated editing targeting the expanded GCG moiety. Unlike recently tested short hairpin RNA (shRNA)-based knockdown of PABPN1 [64], this gene level of correction could result in a permanent production of functional PABPN1 protein that is capable of correcting the OPMD phenotype at both histological (reduction of pathological aggregates) and molecular (abrogation of mutant PABPN1) levels. The need for an adjunct gene replacement therapy as is often the case in knockdown approaches [62] could no longer be necessary. However, thorough characterisation and mechanistic studies would need to be conducted both in vitro and in vivo to confirm CRISPR/Cas9 editing specificity towards the expanded mutant PABPN1 with no obvious off-target effects on the wild-type PABPN1 gene or indeed other genes within the treated tissues.

Myotonic dystrophies

DM is an autosomal dominant, slow-progressing inherited multisystem genetic disorder affecting the muscles (causing wasting), the eyes (leading to cataracts), and the heart (causing conduction defects). The disease is also characterised by metabolic disturbances (endocrine changes) and prolonged contraction of skeletal muscles (myotonia).

Two types of DM have been defined to date. These include DM type 1 (DM1) and DM type 2 (DM2), with DM1 being the most severe and most common form of DM. Overall incidence has been estimated at 3 to 15 per 100,000 in Europe with a higher prevalence in Iceland (1:10,000) and a reported incidence of as high as 1:500 in Quebec [65]. The disease is caused by a CTG trinucleotide expansion in the 3′-untranslated region of the DMPK gene. CTG repeats exceeding 37 are considered abnormal with larger repeats (below 400 repeats) leading to a more severe disease [66, 67]. Like most debilitating muscular pathologies, no curative treatment currently exists for DM. The use of pharmacological agents [68, 69] and antisense RNA-based therapies [70–72] have shown improvements in disease phenotype, but overall benefits on the long-term remain, in some cases, limited and full therapeutic efficacy is yet to be demonstrated in the clinic. Although still in its early infancy, gene editing for DM treatment is an appealing alternative to most pharmacological and gene therapy approaches reported to date. Correction of mutant DMPK at the gene level offers the possibility for a permanent modulation of pathological phenotypes and long-term disease control. A recent study has shown the feasibility of editing intron 9 in the DMPK gene in DM1 neural stem cells derived from human DM1 induced pluripotent stem cells (iPSCs) using a TALEN-mediated homologous recombination-expressing cassette integrated upstream of the CTG repeats [73]. A significant reduction in nuclear RNA foci, together with restoration of normal microtubule-associated protein τ (MAPT) and muscleblind-like (MBNL) splicing patterns, were observed [73]. Although transition into clinical use may be long and difficult, this proof-of-concept ex vivo study offers a rationale for the genetic correction of DM1-dervied stem cells as a potential autologous cell therapy for DM1 patients in the future.

Safety considerations for gene editing

The risk of inducing modifications at off-target genes, leading to unwanted side effects, is a major safety concern. The degree of off-target events and their clinical implications are crucial questions that need to be carefully addressed as part of a regulatory new investigational drug development plan. The primary concern is that modifications could be carcinogenic. It must also be remembered that (unless the therapy is based on a cell type that is selected and clonally expanded after modification) the off-target effects will be heterogeneous from cell to cell, so the analysis must encompass a suitably large (and potentially diverse) population of cells exposed to the modifying agent to evaluate if any cells may be adversely transformed even at low frequency. There may also be risks associated with changes that have cytotoxic effects, and if such effects are frequent in the same cells that undergo correct gene editing, the consequences could include loss of efficacy.

In this regard, rigorous quality assurance tests would need to be conducted to identify any insertions or deletions that could result from a gene editing NHEJ-based treatment. These include gene sequencing, mismatch cleavage assays based on CelI or T7 endonuclease I enzymes, and the tracking of indels by decomposition (TIDE) method. These tests could be used as quality assurance tools to determine the number and location of indels or substitution events occurring after gene editing treatment, more often based on in silico predictions. However, as previously noted with ZFNs, some mutations could occur near cryptic off-target sites that are not predictable in silico [74]. Care should also be taken when validating the assays employed for product characterisation purposes. This follows from previous observations whereby initial analysis by whole genome sequencing (WGS) in gene-corrected human iPSCs (hiPSCs) revealed a large number of indels, of which some were confirmed to be false positive [75]. This highlights some limitations of WGS and calls for a multitesting approach employing different analytical methodologies.

Although current assays would be valuable in identifying the location of suspected off-target mutations that could arise from NHEJ or HDR-based gene editing treatments, they do not, however, provide precise information on the ability of these mutations to cause carcinogenicity in the long term. Hence, additional functional studies are required to assess the significance of these off-target events at the molecular and cellular levels. For instance, an unwanted insertion in the middle of an essential gene sequence could have dramatic consequences. One should have enough knowledge on the biological and physiological functions of the affected gene to draw an informed decision on the overall safety of the proposed approach. Similarly, an off-target insertion into an enhancer or a repressor region would disturb genetic homeostasis resulting in an unwanted down-regulation or up-regulation of genes that come under the affected promoter. This could subsequently interrupt normal cellular activity with potential alteration of phenotype at histological and physiological levels. Although not all mutations could affect cellular proliferation, an unwanted “indel” mutation occurring within genes known to regulate cell replication or involved in programmed cell death raise an alarming concern from a safety point of view. This is due to potential risks arising from a compromised cellular viability leading to severe toxicity in healthy tissues or an uncontrolled replication that could eventually lead to tumour formation. These potentially serious consequences warrant careful considerations during early product development stages; hence, basic biological studies to further characterise the sites at which these mutations have occurred should be conducted as ad hoc validation studies to rule out any unwanted insertional mutagenesis and/or tumourigenic consequences following a gene editing-based treatment.

Although clinical safety experience with biological therapies based on gene editing is lacking, current product development and regulatory strategies should also draw from past and available human clinical trial data on retroviral and lentiviral vectors’ integration sites. A key safety parameter would be to satisfy the regulatory bodies that those detected off-target effects have been thoroughly characterised and occur at “low risk” sites unlikely to cause insertional oncogenesis following integration of foreign DNA. In fact, not all events at off-target genomic sites would result in serious side effects. For instance, those occurring at “extragenic” sites distant from essential gene regulatory sequences are less likely to cause harm than those affecting “intragenic” sites. Modifications to sites that do not fall within a gene transcription unit as well as those situated more than 50 kb away from the 5’ end of any gene and more than 300 kb from genes linked with cancer or microRNA sites are generally considered safer. Genomic sites not within ultraconserved regions and outside long noncoding RNAs would also represent low-risk sites in the event of off-target DNA integration [19, 76].

Furthermore, validation studies should not only focus on the affected off-target sites but also neighbouring sites by measuring the effect of these mutations on neighbouring gene expression. These studies should ideally be validated in vivo. The choice of appropriate animal models, however, remains a challenge and care should be taken when interpreting data [19, 76, 77].

Validation and choice of functional studies for assessment of toxicity

Although basic biological studies are valuable in assessing the extent of genome modification as a whole and increasing our understanding of different alterations taking place including those that are unlikely to result in clinical toxicity, they do not fill all the regulatory requirements gaps. Functional toxicity studies would, therefore, need to be conducted in parallel early during the preclinical stage of product development. Two paramount questions would need to be carefully addressed to satisfy regulators’ concerns on serious toxicity issues that are specific to genome editing-based medicines. These include cytotoxicity and genotoxicity. Tumourigenicity is the most important safety consequence to consider, but germ-line effects are also not expected to be permissible. Cytotoxicity may be caused by (i) the vector itself (including expression of viral vector antigens and potential persistence of the vector in cells), (ii) gene editing machinery (and any persistence thereof, especially DNase), (iii) off-target gene editing, or (iv) on target gene editing (possible if the editing is either not accurately restoring the wild-type or is generating new antigenicity and/or genetic instability).

Eloquent studies and approaches would have to be designed to address these issues. At preclinical stage, this could include viability assays based on GFP-positive (GFP⁺) cells, whereby the treated cells are cotransfected with the tested nuclease-expressing construct and a GFP-expressing plasmid. This would allow investigators to track and quantify any observed decline in the GFP⁺ cell population as a result of a nuclease-related toxicity. The approach would also be valuable in dose escalation studies when deciding on optimal dosage administration for subsequent first-in-human clinical studies. Monitoring of clonal changes has also been used as an informative way to assess genotoxicity in vitro in cells pretagged with unique short sequence identifiers to allow tracking of changes in starting clone dynamics over a period of time [78]. Similarly, other approaches have relied on fluorescently tagged cell cycle indicators in cell lines such as HeLa FUCCI cells as a complementary in vitro validation method to assess the genotoxicity effects of genome editing at the cell cycle level [79].

However, it is important to bear in mind that measuring cytotoxicity and genotoxicity using approaches based on reporter genes and tagging systems is restricted to preclinical assessments and cannot be employed at later stages of product development when assessing potential toxicity in human clinical trials. Hence, the need for clinical toxicity assays that are fit for purpose should not be neglected. Furthermore, correlation of these assays with clinical outcomes is yet to be demonstrated. For this reason, all functional toxicity studies conducted in vitro should use cellular models and gene transfer methods that are similar to those intended to be applied in clinical settings to fulfil some of the safety regulatory requirements.

Vector considerations for in vivo and ex vivo gene editing

Potential cytotoxic effects do arise not only from on-target and off-target effects but also from the employed gene transfer system itself. In this regard, the type of vector (viral or nonviral) and the gene (nuclease) transfer approach (direct in vivo delivery versus ex vivo) adopted will have to be implemented earlier during product development and considered as part of the constructed regulatory strategy for the nuclease-based medicine under consideration.

For severe neuromuscular disorders in which there is a widespread distribution of affected muscles, AAV vectors and in particular serotypes 8 and 9 are seen as ideal systems for gene editing-based nuclease transfer to diseased muscle tissues owing to their relatively high tropism for skeletal muscle cells and the heart [57, 58] as well as their documented safety profile in the clinic following recent approval of Glybera^® in Europe [80]. From a product development point of view, this can be achieved through an HDR-based gene editing construct packaged within a recombinant AAV. Although technically this is achievable with ZFNs due to a relatively small monomer insert size of ~ 1.2 kb, this might not be the case for Cas9 nucleases. With an insert size of ~4.1 kb, this could be an issue when it comes to packaging a bicistronic construct harbouring a donor DNA template [81], considering the limited packaging capacity of rAAVs. Taking into account the EMA’s guidelines on quality, nonclinical and clinical issues related to the development of recombinant adeno-associated viral vectors (EMEA/CHMP/GTWP/587488/2007 Rev. 1) [82], caution should be taken during the manufacturing stage of the vector to avoid the possibility of producing AAV particles whose packaged DNA is greater than that of wild type virus. Technically, this could be overcome by splitting the Cas9 gene between two vectors. However, this markedly increases the developmental and regulatory complexity, with each product warranting a full characterisation and a thorough assessment of approaches employed during manufacturing, quality control, and preclinical evaluation (including choice of animal models for testing, vector persistence and tropism, reactivation of virus infection, and germ-line transmission) in addition to clinical studies. The latter would have to be based on a data-driven dose selection for each vector used and show a comprehensive picture of virus biodistribution and shedding, immunogenicity profile, and germ-line transmission considering the permanent gene modification achieved with gene editing-based nucleases and the risks associated if the virus expressing these genes accidentally infect germ-line cells. A long-term follow-up is therefore highly recommended and should not be neglected as part of a complete regulatory plan for a gene editing-expressing AAV vector.

Similarly, the use of genetically modified and nonmodified cells including myoblast cells and myogenic stem cells for the treatment of monogenic inherited neuromuscular diseases has been well validated in preclinical and clinical studies [83–86] and it is anticipated that these therapies would become normal treatment modalities in the clinic in the future.

Although gene edited-based cell therapies for HIV and leukaemia are seeing a rapid positive progress in clinical trials and are rapidly extending to other debilitating diseases, a smooth and safe progress of these life-changing advanced therapies in the market would require fulfilments of a number of criteria from a regulatory point of view. The ultimate goal is to ensure that these therapies would reach the severely affected individuals at minimal health and safety risks with regards to the treated patients themselves as well as third parties and the environment in large. According to the EMA’s guidelines, a multifactorial risk assessment approach needs to be considered taking into account the origin of cells involved, the type of vector employed during the genetic modification procedure, the manufacturing process, the noncellular components used as part of the formulated product, and the intended specific therapeutic use of the final product.

Lentiviral vectors have long been the vector of choice for the ex vivo engineering of stem cells and muscle progenitor cells for the treatment of muscular dystrophies [87–89]. This adds an additional layer of regulatory requirements from quality, nonclinical and clinical perspectives. The manufacture of viral vectors to the GMP quality standards required for an approvable medicinal product is expensive and often inefficient, but developers are helped by the guidance issued in these areas by the EMA (for example, “Guideline on the quality, nonclinical and clinical aspects of gene therapy medicinal products (EMA/CAT/80183/2014)”) [90], as well as the monograph of the European Pharmacopoeia (Ph. Eur. 5.14). A thorough product characterisation is required as part of this process that would include different but complementary methods based on molecular, biological, and immunological assays with the aim of assuring the identity, purity, and potency of the produced genetically modified cells as a final product. Major considerations for viral vectors used in vivo are the biodistribution, potential for generation and/or shedding of infective virus, potential for germ-line modification, and the impact of interaction with the recipient’s immune status. Clinical use of viral vectors and genetically modified cells should also comply with applicable regulations for genetically modified organisms.

Complexities arise from the genetic manipulation (gene correction) as well as the differentiation status and capacity of the modified cells, which could result in a mosaic cell population. Furthermore, intrinsic variations between cells as a result of donor differences give rise to massive batch variations in the final product. From a regulatory perspective and in line with the EMA’s guidelines, nonclinical and clinical studies need to be conducted with cell medicinal products that are well characterised. The manufacturing process needs to be robust and quality control focused capable of maintaining consistency and reproducibility of the final cell-based product. For this, all starting materials need to be well defined and carefully documented. For treatment of muscle diseases, it is often reasonable to administer cell-based products that are in a differentiated state, which may pose less tumourigenic risks. However, one cannot exclude the existence of a subpopulation of cells in an undifferentiated proliferative state. For iPSCs, for example, it is empirical to conduct additional testing of cell transformation and tumour formation during the early manufacturing stage of the product as a precaution measure. This would often need to be combined with the selection of appropriate markers during critical manufacturing steps for assuring a defined stage of differentiation that is intended for therapeutic use.

For any cell-based product that has undergone a substantial ex vivo manipulation, a robust process validation process is therefore paramount and should include a combination of genetic stability, tumourigenicity, and phenotypic profile assessments of both wanted and unwanted cell populations at all critical stages of manufacturing to ensure safety requirements are met. The mosaic nature of cell-based medicinal products often complicate their identity and purity-related characterisation. Most often, it is reasonable to accept that purity does not always equate homogeneity. Truly selective markers that could accurately map and distinguish different cell types and differentiation stages are yet to be identified, which often render product characterisation a challenging task for most developers. Nevertheless, one should not underestimate the importance of a thorough demonstration of product consistency as a minimum requirement for characterisation purposes.

The ability to track any cell-based therapy following administration in patients is crucial for clinical monitoring purposes. However, limitations in current medical methodologies do not allow a full biodistribution profile to be drawn from human studies. Hence, the importance of thoroughly designed biodistribution studies in nonclinical models should take into account the multistep biodistribution characteristics of cell products, including migration, niche, engraftment, differentiation, and persistence, together with reliable in vivo tracking methods such as the use of marker genes or labelled cells. Current European regulatory requirements do not give exemptions when the risk profile of the cell-based product under investigation is subject to a safety concern or when its route of administration (such as intravenous delivery) warrants a special attention. For this, noninvasive methods based on clinically accepted tracers should be considered and their use should be justified when conducting biodistribution studies in human trials.