The effector domain of gene editing platforms is a nuclease that induces DSBs at the target DNA sequence (Doudna, 2020; Li et al., 2020). Cas proteins possess intrinsic nuclease activity, whereas ZF nucleases (ZFNs) and TALE nucleases (TALENs) have been engineered by fusing the catalytic domain of the FokI nuclease to ZFs and TALEs, respectively (Figure 1A). FokI is a bipartite endonuclease that must dimerize to cleave the target sequence (Vanamee et al., 2001). ZFNs and TALENs therefore have two fused FokI domains binding opposite strands of adjacent sequences in reverse orientations, to promote FokI dimerization and genome restriction (Figure 1A). Spatial orientation and module spacing requirements decrease the probability of off-target cutting events. Site-specific DNA cleavage activates cellular DNA repair pathways, which then delete, insert or replace nucleotide sequences (Yeh et al., 2019). The two main DNA repair pathways for DSBs are the non-homologous end-joining (NHEJ) and homology-directed repair (HDR) pathways. NHEJ is error-prone, often introducing small insertions or deletions (indels), whereas HDR uses homologous sequences as a template, to ensure the correct repair of damaged DNA (Figure 1B). The NHEJ pathway is frequently used to inactivate toxic genes (Figure 1B). The introduction of indels at the 5′ end of the target gene results in frameshift mutations, generating premature stop codons. Other applications include the disruption of aberrant splicing sites or the deletion of large fragments of DNA through the creation of two DSBs in the same chromosome (Figure 1B). By contrast, the accuracy of the HDR pathway allows precise nucleotide insertions, deletions or substitutions at the target site (Figure 1B). This is achieved by using double- or single-stranded DNA templates containing the intended modification, flanked by homologous sequences. HDR can, thus, be used to correct both GOF and LOF mutations, for gene repair. HDR can also be exploited as an alternative approach to classical gene replacement, to improve control over the copy number of the gene of interest and to prevent insertional mutagenesis due to the random integration of viral vectors. HDR-mediated gene replacement involves the site-specific insertion of full transgenes (cDNA) at “safe harbor” locations, defined as sites within the genome at which the addition of sequences does not interfere with the neighboring genes and results in safe robust transgene expression (Figure 1B).

HDR-mediated gene editing is a promising approach for therapeutic applications, but it is generally less efficient than NHEJ and mostly restricted to the G2 and S phases of the cell cycle (Yeh et al., 2019). This imposes additional challenges for the application of HDR-based editing to post-mitotic cells and, therefore, to CNS disorders. Nishiyama and colleagues reported a high efficiency of HDR in the mouse brain (Nishiyama et al., 2017), but most groups have struggled to achieve such success with this approach. Several groups have proposed NHEJ-like strategies to overcome this limitation through precise gene editing in non-replicative cells by microhomology-mediated end-joining (MMEJ) (Yao et al., 2017b), homology-independent targeted integration (HITI) (Suzuki et al., 2016), and microhomology-dependent targeted integration (MITI) (Li et al., 2019). Other groups have explored HDR-like mechanisms, such as homology-mediated end joining (HMEJ) (Yao et al., 2017a) and single homology arm donor-mediated intron-targeting integration (SATI) (Suzuki et al., 2019). These techniques have yielded significantly higher rates of gene insertion in post-mitotic cells, although the mechanisms involved are not fully understood. Other groups have suggested approaches in which HDR repair is promoted by fusing the Cas9 nuclease to factors involved in the regulation of NHEJ/HDR pathways. For instance, p53-binding protein 1 (53BP1), which plays a major role in balancing NHEJ/HDR ratio, promotes DSB repair via the NHEJ pathway by preventing the DNA end resection required for HDR (Bunting et al., 2010). Cas9 fused to a dominant-negative 53BP1 enhances HDR and inhibits NHEJ in a target-specific manner, without modifying cellular DNA repair mechanisms overall (Jayavaradhan et al., 2019). Efforts have also been made to improve HDR by fusing Cas9 to RecA (RAD51 in eukaryotes), which plays a key role in homologous recombination (Cai et al., 2019; Kurihara et al., 2020), or by altering the conformational checkpoints for Cas9 binding to DNA (Kato-Inui et al., 2018).

No product for therapeutic gene editing has yet been approved, but the first clinical trials based on this technology have demonstrated the safety of this approach (Schacker and Seimetz, 2019). However, as gene editing permanently modifies the DNA, several biosafety concerns have been raised concerning the induction of off-target DSBs and increases in genomic instability (Mills et al., 2003). Unlike DSBs, DNA single-strand breaks (SSBs) are common events under physiological conditions, and are less harmful than DSBs (Caldecott, 2008). Nickases were developed by mutating one of the catalytic sites of Cas9 (nCas9), such that only one strand of the DNA is cut (Doudna, 2020) (Figure 1A). Paired nickases targeting nearby sequences on opposing strands can create specific DSBs, while decreasing the chances of producing off-target DSBs (Dabrowska et al., 2018; Ge and Hunter, 2019). The use of SSBs and ssDNA repair templates to insert specific sequences has been explored as an alternative to DSB-mediated HDR (Rees et al., 2019). Nickase variants have improved the HDR:indel ratio, but, overall, this approach remains less efficient than DSB-mediated recombination.

Genome regulation offers additional therapeutic options through the modulation of gene expression at native loci. Gene expression is regulated by multiple factors, including both cis and trans elements, ultimately leading to the recruitment of RNA polymerases to promoter regions. Genome expression is also regulated by epigenetic marks, which determine chromatin accessibility state and comprise multiple elements, including the three-dimensional architecture of the DNA and histone or DNA modifications (Holtzman and Gersbach, 2018). An extensive list of possible histone modifications, including acetylation, methylation and phosphorylation, has been described, and all these processes can be altered to modulate gene expression (Holtzman and Gersbach, 2018). Epigenetic modifications, particularly for histone tails and DNA methylation status, have provided insight into the role of such changes in gene regulation and their contribution to disease. For instance, cytosine methylation (5C-methylcytosine) at CpG dinucleotides is usually enriched in silenced promoters (Weber et al., 2007; Kundaje et al., 2015) and has been implicated in genomic imprinting (Laan et al., 1999), whereas H3K9 acetylation is associated with active promoters (Ernst et al., 2011). For the alteration or restoration of gene expression profiles, ZFs, TALEs, and dCas proteins have been fused to scaffold transcriptional modulators or epigenetic modifiers (Figure 3A). Genome regulation strategies can be used to upregulate or repress gene expression by two different approaches: (1) transcriptional modulation through the recruitment of transcription factors and chromatin remodelers and (2) epigenome editing through the direct modification of epigenetic marks.

Transcriptional activation has been achieved through the tethering of ZFs, TALEs, and dCas9 to several copies of herpes simplex virus protein 16 (VP16), the transactivating domain of the NF-kB p65 subunit (p65), heat shock factor 1 (HSF1) and Epstein–Barr virus R transactivator (RTA) (Figure 3B). The targeting of multiple copies of transactivating domains to promoter regions was rapidly shown to have a synergistic activation effect. This led to the development of dCas-based second-generation activators, which can target multiple transactivating domains to a single locus. Chavez and coworkers evaluated the potency of several dCas9 activators in different cell lines and showed that the synergistic activator mediator (SAM) (Konermann et al., 2015), SUperNova Tagging (SunTag) (Tanenbaum et al., 2014) and the tripartite VP64-p65-RTA (VPR) (Chavez et al., 2015) systems were the most efficient at inducing gene activation (Chavez et al., 2016). These systems have been adapted to activate genes in vivo (Chew et al., 2016; Liao et al., 2017; Moreno et al., 2018; Zhou et al., 2018; Breinig et al., 2019; Savell et al., 2019; Zhan et al., 2019). Gene expression profiles can also be altered to reprogram cells to differentiate into particular cell types. Liao and coworkers reprogrammed hepatic cells into pancreatic-like beta cells, by activating the Pdx1 (Liao et al., 2017). They also improved DMD symptoms by activating the Utrn gene. Savell and coworkers demonstrated robust Fosb activation in several regions of the brain in vivo (Savell et al., 2019). Similarly, Zhou et al. demonstrated the in vivo genetic reprogramming of neurons in mouse brain by simultaneously activating the expression of Ascl1, Neurog2, and Neurod1 (Zhou et al., 2018). Breinig and coworkers recently altered the sgRNA length of a Cas12a-VPR variant to induce either gene activation or knockout in vivo (Breinig et al., 2019). This work has added an additional degree of complexity to these systems, allowing not only the targeting of multiple genes, but also a larger range of modifications. Artificial transcriptional repressors have also been generated by fusing the Kruppel-associated box protein (KRAB) domain to the DNA-binding platforms (Bailus et al., 2016; Zeitler et al., 2019) (Figure 3B). KRAB is a scaffold protein involved in recruiting KAP1/TIF1β corepressor complexes, which in turn recruit DNA methylases or histone modifier factors (Kim et al., 1996; Ying et al., 2015). The effects of KRAB on gene repression can be permanent or reversible, depending on developmental stage (Ying et al., 2015).

Unlike the recruitment of activating or repressing complexes/factors, epigenetic editing can modify epigenetic marks by targeting specific enzymes. Epigenetic activation has been achieved through site-specific DNA methylation by the DNA demethylase 10–11 translocation methylcytosine dioxygenase 1 (TET1) (Maeder et al., 2013; Choudhury et al., 2016a; Liu et al., 2016; Xu et al., 2016) or with enzymes promoting activating histone signatures, such as the histone acetylase core subunit p300 (Hilton et al., 2015) and histone methyltransferases (Cano-Rodriguez et al., 2016) (Figures 3C,D). For instance, Liu and coworkers described the in vivo demethylation of a methylation-sensitive Snrpn-GFP cassette in transgenic mice (Liu et al., 2016). Heterozygous mice carrying a paternal copy of the transgene do not express GFP, due to the methylated status of this copy of the gene. The authors reported the targeted active demethylation of the transgenic cassette and a 70% activation of GFP expression after lentiviral injections of dCas9-Tet1 into the brain. Rather than active DNA demethylation, Hilton et al. demonstrated that the fusion of p300 to the three DNA-binding platforms activated the expression of multiple endogenous genes (ilrn1, oct4, and myod1) through histone acetylation (Hilton et al., 2015). Finally, epigenetic repression has been achieved through direct DNA methylation (Bernstein et al., 2015), histone deacetylation (Kwon et al., 2017), and histone demethylation (Kearns et al., 2015) (Figures 3C,D).

Alzheimer’s disease (AD) is the main cause of dementia, affecting millions of people worldwide (Winblad et al., 2016; Dos Santos Picanco et al., 2018). One of the hallmarks of AD is the presence of scattered extracellular senile plaques, due to the accumulation of amyloid-β (Aβ) in the brain. Aβ is a secondary metabolite generated by the processing of amyloid precursor protein (APP) by β-secretase 1 (BACE1). Alternatively, APP may be processed via a non-amyloidogenic pathway involving α-secretases, leading to the generation of neuroprotective products (Richter et al., 2018). In a study of the treatment of a familial form of AD caused by the Swedish mutation of APP (APPsw), CRISPR-mediated NHEJ was used to inactivate the mutant APP (György et al., 2018). This can be achieved by designing sgRNAs targeting single nucleotide polymorphisms (SNPs) in the target sequence of the sgRNA (mismatch-based selectivity) or in the PAM (PAM-based selectivity). György and coworkers detected 1.3% indels in the APPsw allele after the hippocampal injection of a mismatch-based selective CRISPR/Cas9 system split into two AAV9 vectors (because of the limited capacity of AAV vectors of ∼4.8 kb) in Tg2576 mice (György et al., 2018). By contrast, Sun and coworkers used a non-allele selective CRISPR-mediated NHEJ strategy to push APP processing toward the non-amyloidogenic pathway (Sun et al., 2019). Based on evidence suggesting that deletion of the C-terminus of APP can mitigate Aβ generation (Koo and Squazzo, 1994) and reduce APP interactions with the BACE-1 enzyme (Das et al., 2016), the authors used CRISPR to generate C-terminally truncated APP, thereby circumventing the amyloidogenic processing of APP (Sun et al., 2019). In this study, APP truncation in WT and heterozygous APP-London human iPSC-derived neurons increased the production of the neuroprotective sAPPα and reduced the secretion of Aβ40/42 and the sAPPβ fragment. For in vivo studies in adult mice, the CRISPR-APP system was split into two AAV9 vectors and delivered to the dentate gyrus of WT mouse brains. The injection of CRISPR-APP reduces the level of the processed C-terminal fragments (CTFs) by half while having no or minimal impact on the total APP protein. No additional in vivo tests were performed to evaluate treatment efficacy in the context of AD (György et al., 2018; Sun et al., 2019), but these therapeutic strategies targeting the C-terminal part of APP are of interest because the aim was to attenuate pathological properties (Aβ generation) while potentially maintaining Duarte and Déglon Corrigendum: Genome Editing for CNS Disorders other physiological functions of APP. Another approach, developed by Park and coworkers, uses CRISPR-Cas9-loaded nanocomplexes targeting BACE1 in the 5XFAD and APP transgenic mouse models to reduce the generation of Aβ and improve AD symptoms (Park et al., 2019). Four weeks after CRISPR injection into the CA3 hippocampal region of 5XFAD mice, 45% of target sequences contained indels, and a 34% decrease in Bace1 expression was observed, revealing this method to be more efficient than the use of chemical BACE1 inhibitors. They also observed a decrease in Aβ plaque accumulation by a factor of more than two, together with a significant rescue of associative learning (fear conditioning test) and spatial working memory (Morris water maze) in the treated 5XFAD mice. These molecular and behavioral improvements were maintained for up to 12 weeks. Off-target evaluation by whole-genome sequencing (WGS), whole-exome sequencing (WES), Digenome-sequencing (Digenome-seq) and deep sequencing identified a few off-target mutations and small-scale chromosomal rearrangements.

Bustos and coworkers investigated the potential of epigenome editing for AD by targeting the dlg4 gene, encoding the PSD95 protein (Bustos et al., 2017). PSD95 is a scaffolding protein present at the excitatory post-synaptic density, and is involved in the regulation and organization of post-synaptic synapses (Elias and Nicoll, 2007). Abnormal PSD95 expression has been described not only in AD, but also in other neurological disorders, such as Huntington’s disease (HD) and schizophrenia. The authors developed and validated in vitro several ZF-based epigenome modifiers targeting the proximal promoter region of dlg4/PSD95 for the activation or repression of PSD95 expression (Bustos et al., 2017). They demonstrated that alterations in expression were specifically associated with histone modifications rather than other changes, such as CpG methylation in DNA. The fusion of zinc fingers to the histone methyltransferase G9a (PSD95-6ZF-G9a) induced gene repression associated with an increase in the di- and tri-methylation of H3K9, whereas PSD95-6ZF-VP64 gene activation was coupled to H3 activation, probably through the recruitment of histone acetylases by the VP64 domain. PSD95-6ZF-VP64 was also shown to have neuroprotective effects. AβPPswe/PS-1 mice receiving AAV-PSD95-6ZF-VP64 injections into the hippocampus had higher levels of PSD95 expression and displayed a rescue of memory and spatial learning performances to normal aged-matched levels.

Parkinson’s disease (PD) is the second most common neurodegenerative disorder, affecting 2–3% of people under the age of 65 years (Poewe et al., 2017). PD patients display motor movement dysfunction, but also cognitive impairment, depression and dementia. At the cellular and molecular levels, PD is characterized by a striatal dopamine deficiency due to progressive neuronal loss in the substantia nigra, and by the formation of intracellular aggregates containing α-synuclein. Dopamine loss and basal ganglia circuitry disruption are well-defined features in PD, but this disease is extremely complex and driven by diverse molecular and neurophysiological mechanisms.

Several gene-based therapies for PD have been proposed, including the targeting of α-synuclein, cellular oxidation and the autophagy-lysosomal pathway (Poewe et al., 2017). Genome editing for PD has mostly been used for disease modeling in vitro (Safari et al., 2019). For instance, Kantor and coworkers induced the hypermethylation of CpG islands in SNCA intron 1 in iPS-derived dopaminergic progenitor neurons, through lentiviral transduction with a dCas9-DNMT3A system (Kantor et al., 2018). They observed a ∼ 25% decrease in α-synuclein protein levels and the rescue of mitochondrial-associated superoxide production and cell viability. They observed no overall change in the methylation status of the treated cells, identifying the dCas9-DNMT3A-mediated targeting of SNCA as a promising approach for PD treatment. Another potential therapeutic target is glial cell line-derived neurotrophic factor (GDNF), which has been shown to have neuroprotective effects and to improve Parkinsonian symptoms (Kordower et al., 2000; Tenenbaum and Humbert-Claude, 2017). Laganiere and colleagues used a ZF-p65 fusion to upregulate the expression of endogenous GDNF in a 6-OHDA rat model of Parkinson’s disease (Laganiere et al., 2010). They observed an increase in the number of TH-positive fibers in both the medial forebrain bundle and the substantia nigra after 7 weeks of AAV2-rGDNF-ZFP infusion (Laganiere et al., 2010). The rGDNF-ZFP-treated group performed better in the corridor test, the cylinder test and the drug-induced rotational test than the GFP-treated control. This study yielded promising results, but a clinical trial based on the direct infusion of GDNF into the putamen resulted in no significant improvement of Parkinson’s disease symptoms (Lang et al., 2006; Whone et al., 2019), raising questions about therapeutic efficacy of GDNF.

Huntington’s disease (HD) is a neurodegenerative disorder caused by an inherited dominant CAG trinucleotide expansion mutation on the HTT gene. In vivo genome editing strategies for HD have explored NHEJ-mediated gene inactivation (Merienne et al., 2017; Monteys et al., 2017; Yang et al., 2017) and the transcriptional repression of HTT (Zeitler et al., 2019). Yang et al. used two separate AAVs expressing SpCas9 and two sgRNAs targeting the flanking regions of the CAG repeat in a non–allele-specific manner in the HD140Q-KI mouse model (Yang et al., 2017). The injection of neuron-specific AAV-Cas9-HTT resulted in the efficient transduction of medium spiny neurons, significantly decreasing the accumulation of both mutant (mHTT) and WT HTT in the striatum of 9-month-old homozygous and heterozygous HD140Q-KI mice. The treated heterozygous mice performed better in the rotarod, beam and grip strength tests. Although no deleterious effects of depleting both mutant HTT copies from homozygous HD140Q-KI mice were detected (Yang et al., 2017), it still remains a matter of debate whether disruption of the normal physiological functions of WT HTT lead to harmful effects at adult stages (Liu and Zeitlin, 2017). With this in mind, Monteys and coworkers designed a PAM-based strategy targeting a SNP for specific inactivation of the mutant HTT allele (Monteys et al., 2017). They demonstrated the allele selectivity of the chosen sgRNAs in vitro in fibroblasts from human HD patients and showed efficient HTT exon-1-targeted deletion following the injection of allele-selective AAV1 CRISPR-HTT into BACHD transgenic mice. This treatment halved the levels of human mHTT mRNA in the striatum. However, it should be noted that heterozygous BACHD transgenic mice have about five tandem copies of the human mHTT gene and two copies of the endogenous mouse WT gene (Gray et al., 2008). In these studies, the spCas9 was constitutively expressed. The stable and permanent expression of nucleases eventually leads to higher levels of on-targeting editing, but it also increases the occurrence of off-target events and immunogenic responses. We have tried to overcome this problem by developing the self-inactivating KamiCas9 system, for transient Cas9 expression (Merienne et al., 2017). This system is based on a lentiviral vector with a larger cloning capacity than AAV. It is composed of the Cas9 nuclease, a sgRNA targeting HTT and a second sgRNA targeting the translation start site of the Cas9 nuclease. High on-target efficiency and inactivation of the Cas9 nuclease over time are ensured by the use of a strong PolIII promoter (H1) to drive the sgHTT and a weak PolIII promoter (7sk) to drive the sgCas9. We demonstrated high levels of exogenous hHTT-82Q (20–35%) and Cas9 (∼40%) editing following the injection of LV-KamiCas9 and hHTT-82Q into mouse striatum. Western blot analysis of striatal samples from mice receiving LV-KamiCas9 injections revealed an almost-complete absence of the Cas9 protein after 2 months.

Garriga-Canut et al. (2012) attempted to design a CAG copy number-dependent ZF-based transcription repressor exclusively targeting the mHTT allele. This first tool established the proof-of-principle for HTT repression in vivo, decreasing mHTT mRNA levels by about 30% in the brains of R2/6 mice receiving AAV1-ZF-Kox1 injections. Despite the achievement of selective repression in vivo, the mutant allele in R6/2 mice contains 115–160 repeats, a number not consistent with the degree of CAG expansion in most HD patients. Zeitler and coworkers recently generated a second-generation ZF-KRAB that preferentially recognizes pathogenic CAG repeats, and demonstrated highly significant mHTT suppression with wild-type allele preservation in patient derived-iPSCs (Zeitler et al., 2019). They observed beneficial behavioral effects in R6/2 mice for 7 weeks after the intrastriatal injection of AAV-ZF-KRAB, and demonstrated the absence of inflammation or adverse effects of long-term expression in mouse brain.

Amyotrophic lateral sclerosis is a neurodegenerative disease caused by the progressive neurodegeneration of both upper and lower motor neurons (Rowland and Shneider, 2001). Muscle atrophy begins in adult patients with ALS and progresses to total paralysis and, eventually, death. Approximately 2% of ALS cases result from a dominant mutation of the SOD1 gene. Gaj et al. (2017) mitigated ALS symptoms and improved the survival of a mouse model of ALS, G93A-SOD1 mice, containing 25 copies of the human mutant SOD1, by disrupting the human SOD1 gene with the Staphylococcus aureus Cas9 (SaCas9). The CRISPR system was packaged into a single AAV9 variant (double-tyrosine mutant) shown to enhance gene transfer to the CNS (Petrs-Silva et al., 2009; Dalkara et al., 2012). The authors demonstrated efficient neuronal transduction of the ventral horn of the spinal cord, with up to 74% of motor neurons expressing the nuclease, after systemic injections in neonatal transgenic mice (Gaj et al., 2017). Western blot analysis revealed a 2.5- to 3-fold decrease in mutant SOD1 protein levels, but sequencing data showed that only a small fraction of the total human SOD1 transgenes had been edited (0.2–0.4%). This discrepancy may reflect the large numbers of glial cells in the gray matter of the spinal cord, which were not efficiently transduced, or differences in SOD1 expression in transduced and non-transduced regions of the spinal cord. Regardless of this divergence, the onset of disease in animals treated with SaCas9-SOD1 was delayed by 33 days, and survival was 28–30 days longer than in the control. In age-matched mice, the editing of SOD1 improved rotarod performance, prevented weight loss and reduced muscular atrophy. The treatment was unable to slow the progression of the disease after its onset, but end-stage tissue analysis in SaCas9-SOD1-treated mice revealed the presence of ∼50% more motor neurons. SOD1 inclusion bodies were observed in astrocytes, suggesting that glial cell targeting might be required to slow the progression of the disease, since these cells have been shown to play a role in disease progression (Boillée et al., 2006; Yamanaka et al., 2008).

Angelman syndrome is a neurological disorder caused by a genetic UBE3A deficiency resulting in intellectual disability, ataxia and seizures (Laan et al., 1999). The paternal Ube3a allele is specifically silenced by a brain-specific antisense transcript (Ube3a-ATS). LOF mutations in the maternal allele therefore lead to UBE3A deficiency. Bailus and coworkers developed a ZF-KRAB repressor targeting the transcription start site of Ube3a-ATS (Snurf/Snrpn promoter), to overcome the paternal imprinting of the Ube3a gene (Bailus et al., 2016). The systemic injection of TAT-S1-linked UBE3a-6ZF-KRAB repressor partially rescued Ube3a expression levels in the hippocampus and cerebellum of a mouse model of Angelman syndrome. However, this therapeutic approach may require multiple treatments, because the repressor function of the KRAB domain has been shown to be transient (Gilbert et al., 2014; Ying et al., 2015).

MECP2 encodes a nuclear protein involved in the transcriptional and post-transcriptional regulation of many genes (Cheng and Qiu, 2014). Duplication or triplication of Xq28 leads to MECP2 GOF mutations mostly affecting boys (Ramocki et al., 2009). This syndrome is characterized by intellectual disability, poor speech development, motor dysfunction and anxiety. Yu and coworkers reported that the normalization of MeCP2 levels in the medial prefrontal cortex of adult MECP2 transgenic mice through CRISPR/Cas9-mediated NHEJ can reverse the social recognition deficit (Yu et al., 2020). The CRISPR system was packaged into two AAV particles (SpCas9 + sgRNA), which were stereotaxically injected into the mouse brain. Immunostaining and western blotting 6 weeks after treatment showed that MeCP2 protein levels had almost halved. Despite improvements in social recognition behavior, the treatment had no effect on locomotor activity, or heightened anxiety-like behaviors, suggesting that different brain areas or neural circuits may contribute to the diverse aspects of the syndrome.

Fragile X syndrome (FXS) is the most common single-gene form of autism spectrum disorders (ASDs), for which there is currently no effective treatment (Kaplan and McCracken, 2012). It is caused by a trinucleotide CGG repeat expansion in the 5′ UTR of the fragile X mental retardation 1 (FMR1) gene, encoding the fragile X mental retardation protein (FMRP) (Dölen and Bear, 2008; Persico and Napolioni, 2013). This mutation inactivates the gene, due to hypermethylation of the expanded repeats and heterochromatin formation. Excessive mGluR5 signaling has been observed not only in FXS, but also in other ASDs (Silverman et al., 2012). Lee and coworkers explored the CRISPR-mediated disruption of metabotropic glutamate receptor 5 (mGluR5) as a mean of counteracting FXS by delivering RNPs SpCas9 or Cas12a targeting the mGluR5 to the striatum of Fmr1-knockout mice (a mouse model of FXS) (Lee B. et al., 2018). The editing tool was delivered with CRISPR-gold technology, which combines gold nanoparticles conjugated with oligonucleotides and the endosomal disruptive polymer PAsp(DET), for the transfer of RNPs into cells by endocytosis (Lee et al., 2017). The indel frequency was 14.6%, and a 40–50% decrease in mGluR5 mRNA and protein levels was observed. In addition, behavioral analysis revealed that mGluR5-CRISPR-Gold rescued the excessive digging and exaggerated repetitive jumping behaviors of treated mice.

Traumatic CNS injuries and stroke are very common causes of disability, and the treatments currently available are very limited. CNS trauma involves an initial mechanical injury, which is followed by a cascade of molecular and cellular phenomena, ultimately leading to neuronal death by apoptosis. Genome editing therapy strategies have focused on VEGF, which is a neuroprotective factor that favors endothelial cell proliferation and blood vessel formation (Shweiki et al., 1992). These studies used engineered ZFs targeting the proximal promoter of VEGF fused to the transactivating domain of the NF-kB p65 subunit. Michael Fehlings’s laboratory has demonstrated an increase in the number of blood vessels and angiogenesis, a decrease in neurodegeneration and an improvement of behavioral outcomes in a rat model of SCI following the intraspinal microinjection of AdV-ZFP-VEGF and AAV2-ZFP-VEGF activators (Liu et al., 2010). The timing of treatment for traumatic damage is an important parameter for clinical application. Beneficial effects have been shown following the administration of AdV-ZFP-VEGF 24 h after injury (Figley et al., 2014). In addition, Siddiq et al. (2012) used the unilateral fluid percussion injury model in rats to demonstrate the neuroprotective and angiogenic effects of ZFP-VEGF delivery to the cortex or hippocampus by intracerebral injection. Treatment did not improve performance in the Morris water maze or balance beam latency experiments relative to control, but the treated group performed significantly better than controls in the rotarod test.

GM2-gangliosidoses are autosomal recessive disorders caused by the deficiency of a lysosomal enzyme, β-hexosaminidase, resulting in the accumulation of GM2 gangliosides. Biallelic LOF mutations of the Hex α-subunit (HEXA) or Hex β-subunit (HEXB) genes lead to Tay-Sachs disease and Sandhoff disease, respectively. Ou and coworkers recently used a cross-correction strategy based on liver-targeted HDR-mediated CRISPR editing to restore the function of β-hexosaminidase in the brain, in a Sandhoff mouse model (Ou et al., 2020). They injected a dual AAV system consisting of AAV8-SaCas9 and AAV8-HEXM-sgRNA targeting the albumin safe harbor locus into neonatal Sandhoff mice, to introduce, via HDR, the coding sequence of a modified human Hex μ subunit (HEXM) able to process GM2 gangliosides (Karumuthil-Melethil et al., 2016). Four months after the systemic delivery of this sequence, levels of MUGS and MUG activity in the brain were significantly higher than those in untreated Sandhoff mice. Mice receiving the AAV8-HEXM-sgRNA alone displayed no such increase in MUGS and MUG activities, indicating an absence of HEXM expression from the episomal donor template vector. In addition, treated mice performed better in the rotarod test and one in three mice had lower levels of neuronal lysosomal accumulation, indicating that hepatocyte editing can lead to neurological improvements. Indeed, the HEXM variant has been reported to improve gangliosidosis in both the Sandhoff and Tay-Sachs models (Karumuthil-Melethil et al., 2016; Osmon et al., 2016), suggesting that this strategy may provide protection against both disorders.

About 20% of the 100 or so alleles associated with deafness result from GOF mutations (Müller and Barr-Gillespie, 2015). DFNA36 is a progressive hearing loss disease caused by dominant mutations of the tmc1 gene, leading to the neurodegeneration of sensory hair cells. This disease is of particular interest due to the existence of an orthologous mouse mutation, Beethoven (Bth), which also causes hearing loss in mice (Zhao et al., 2014). Two recent reports described the use in vivo of allele-specific CRISPR-mediated NHEJ as a therapeutic strategy for DFNA36 (Gao et al., 2018; György et al., 2019). Gao and coworkers used SpCas9 together with a sgRNA matching the mutant allele, but not the WT allele, to knockout the mutant allele (Gao et al., 2018). They delivered RNP complexes bound to cationic lipids and, even though the targeting of the mutant allele was highly selective (96% of mutant/WT), the frequency of indels was low (1.8%). Nevertheless, the treatment was sufficiently effective to promote hair cell survival, particularly for inner hair cells (IHCs), and to improve cochlear function significantly between the frequencies of 8 and 23 kHz (Gao et al., 2018). However, at 8 weeks, an analysis of cochlear function in treated Tmc1Bth/ + mice revealed less evident improvements relative to the control, suggesting that higher levels of mutant gene inactivation might be required to stop neurodegeneration, or that the small proportion of WT alleles inactivated might neutralize the benefits of mutant knockout over time. This strategy resulted in allele-specific editing, but PAM-based strategies are generally more selective, as demonstrated by György and coworkers (György et al., 2019). They used the SaCas9-KKH variant to edit the mutant allele in a PAM selective manner (György et al., 2019). The SaCas9-KKH/sgRNA treatment via AAV-Anc80L65 was more selective that the treatment used in the previous study, with no detectable indels in the WT allele and a frequency of 2.2% indels for the mutant allele. At the age of 6 months, SaCas9-KKH/sgRNA-treated mice had significantly higher survival rates for both inner hairy cells and outer hair cells (OHCs), with normal hair bundle morphology in all cochlea, except for the OHCs in the basal region, which were absent. The authors also demonstrated the stable maintenance of low thresholds of auditory brainstem responses for up to 40 weeks. Finally, GUIDESeq analysis detected no genome-wide off-target events in Tmc1WT/WT fibroblasts, further highlighting the potential interest of AAV-SaCas9-KKH-sgTmc as a therapeutic strategy for DFNA36 hearing loss.

DNFA36 results from GOF mutations of the tmc1 gene, whereas LOF mutations in both tmc1 alleles result in the autosomal recessive congenital DFNB7/B11 hearing loss disorder. Gene disruption approaches are not suitable for the treatment of DFNB7/B11. Yeh and coworkers explored the use of a base editing strategy to correct the tmc1 alleles in Tmc1^Y182C/Y182C mice (Yeh et al., 2020). They reported 2.3% base editing in a bulk organ of Corti at P14 after the injection of a dual AAV system encoding AID-BE into the inner ear at P1. Tmc1 is expressed only in hair cells. The authors therefore analyzed base editing at the RNA level, to improve the quantification of editing in these cells. They observed ∼ 50% editing in the cDNA, but these results must be interpreted with caution because they may not reflect the editing at the DNA level. The treatment of mice resulted in the preservation of hair bundle morphology and a restoration of the mechanotransduction current in the sensory hair cells. There were 46% more hair cells in the treated mice 4 weeks after injection, with a progressive decrease in cell numbers thereafter, until 6 weeks. The decrease in cell survival was followed by a decline in hearing function, suggesting that more efficient base editing is required to prevent the degeneration of hair cells over time.

Retinitis pigmentosa (RP) is an inherited disorder and the most common cause of progressive vision loss (Kalloniatis and Fletcher, 2004). It is defined by an initial progressive loss of rod photoreceptors, followed by cone photoreceptor degeneration. One form of RP results from a biallelic LOF mutation in the PDE6B gene, introducing a premature stop codon. Cai et al. used HDR-mediated CRISPR editing to correct the mutation (Cai et al., 2019). In this study, the authors developed an improved CRISPR system for HDR (Cas9/RecA) consisting of a sgRNA with MS2 aptamers for the recruitment of MS2-RecA fusion proteins to the target site to promote recombination between the cleavage site and a ssDNA donor template. The potential of this tool to repair the PDE6B gene was evaluated by electroporating the retinas of WT and rd1 mice with four plasmids (SpCas9 + sgRNA-MS2apt + MS2-RecA + ssTemplate) at P0. A 2% restoration of PDE6B WT protein levels was observed in Cas9/RecA-treated mice, whereas no wild-type PDE6B protein was detected in Cas9-treated mice (SpCas9 + sgRNA-MS2apt + ssTemplate), indicating that Cas9/RecA enhances HDR efficiency. Cas/RecA treatment at P0 rescued both rod and cone photoreceptors, but the degree of rescue was 1.8- and 1.6-fold lower, respectively, when mice were treated at P3, suggesting that the loss of photoreceptor proliferation had a negative effect on HDR-mediated correction. In addition, an analysis of visual function and pupillary light reflexes revealed that Cas9/RecA partially rescued the pupillary light reflexes of rd1 mice, demonstrating beneficial effects of treatment.

Leber congenital amaurosis type 10 (LCA10) is an autosomal recessive condition causing early blindness in infancy (Stone, 2007; Stone et al., 2017). It is defined by LOF mutations of both CEP290 alleles. The IVS26 point mutation creating a new splice donor site is the most frequent defect. It alters transcript splicing and generates a premature stop codon in the processed mRNA. Maeder and coworkers recently reported an exhaustive drug dosing study of the use of AAV5-SaCas9-mediated NHEJ to correct the IVS26-driven aberrant CEP290 splicing in retina photoreceptor cells (EDIT-101) (Maeder et al., 2019). The proposed strategy induced a cleavage on either side of the mutation, with a pair of sgRNAs used to delete or invert the fragment containing the IVS26 mutation. The authors evaluated the kinetics and dose response of the editing system in the retina of CEP290 IVS26-KI mice and cynomolgus monkeys, in which maximum editing rates of 21.4 and 27.9%, respectively, were obtained. They also demonstrated ocular tolerability in all animals, except those without immunosuppression regimens, which displayed mild inflammation. This report resulted in the first approved preclinical study of CNS genome editing for clinical trial continuation in humans (NCT03872479). The cep290 cDNA is ∼7.5 kb long, a size well-beyond the capacity of the AAV vectors used for gene replacement. This approach demonstrates the therapeutic potential of gene repair for counteracting CNS disorders without the need to provide exogenous WT transgenes.