Therapeutic strategy for spinal muscular atrophy by combining gene supplementation and genome editing

Defect in the SMN1 gene causes spinal muscular atrophy (SMA), which shows loss of motor neurons, muscle weakness and atrophy. While current treatment strategies, including small molecules or viral vectors, have shown promise in improving motor function and survival, achieving a definitive and long-term correction of SMA’s endogenous mutations and phenotypes remains highly challenging. We have previously developed a CRISPR-Cas9 based homology-independent targeted integration (HITI) strategy, enabling unidirectional DNA knock-in in both dividing and non-dividing cells in vivo. In this study, we demonstrated its utility by correcting an SMA mutation in mice. When combined with Smn1 cDNA supplementation, it exhibited long-term therapeutic benefits in SMA mice. Our observations may provide new avenues for the long-term and efficient treatment of inherited diseases.


INTRODUCTION
Spinal muscular atrophy (SMA) is a severe inherited neuromuscular disease caused by deletion or mutation of survival motor neuron gene 1 (SMN1), which contributes to numerous cellular processes and pathways, the most studied function is its role in snRNP assembly (1). SMA is characterized by the degeneration of lower motor neurons which leads to muscle weakness and atrophy. SMA occurs in approximately 1 in 11,000 newborns and represents the most common genetic cause of infant mortality (2). In human, there are two forms of SMNs. SMN1 is the primary gene for production of SMN protein. SMN2 is a paralog of SMN1, which differs only a few nucleotides and generates a truncated unstable protein with low levels of full-length SMN protein compared as SMN1. Although only 10-20% of the SMN2 gene product is fully functional, it can partially compensate for the loss of SMN1 and increased copy number of SMN2 is inversely correlated with disease severity in SMA patients (3). Today, there are several approaches to treat SMA and effectively prolong the life of patients. Antisense oligonucleotides (ASOs) targeted to RNA splicing of SMN2 gene produces a full-length mRNA and a functional SMN protein (4). ASOs have huge potential in SMA therapy but present some limitations including the difficulties crossing the blood-brain barrier. Another alternative approach for SMA is gene supplementation therapy, which delivers a fully functional copy of the SMN1 gene on the genome of self-complementary adeno-associated viral (AAV) vector packaged with serotype 9 (scAAV9). Recently, Mendell et al. showed that an one-time intravenous injection of high dose scAAV9-SMN1 resulted in improved motor function and extended survival in SMA patients (5).
However, this gene supplementation therapy cannot permanently restore endogenous SMN1 expression because the genome of the recombinant AAV vectors does not undergo genomic integration into the host genomic DNA. The AAV genomes persist mainly in an episomal form in the nucleus of the host cells, which eventually leads to dilution of the gene encoded by the AAV genome.
Therefore, new strategies for in situ correction of endogenous mutated sequences are needed for efficient and long-term improvement of SMA.
The CRISPR-Cas9 technology is a powerful genome-editing system which could be applied to potential cure for common inherited diseases (6,7). However, in situ gene correction of the SMA-causing mutation has not been reported, likely due to the inherent difficulties in accessing motor neurons within the spinal cord. A major hurdle is that genome editing in non-dividing cells, including motor neurons, is inefficient in the absence of the homology-directed repair (HDR) pathway used in conventional gene-repair approaches. Another problem is the low in vivo delivery of genome-editing tools into the spinal cord, which prevents sufficient genome correction to improve the disease phenotypes.
We recently developed a non-homologous end joining (NHEJ)-based homology-independent strategy for targeted integration (HITI) of transgenes in both dividing and non-dividing cells (8,9).
Since NHEJ is active throughout the cell cycle in a variety of cell types including neurons (10,11), HITI technology has the potential to provide a solution to some of the current challenges for efficient and long-term amelioration of SMA phenotypes. In here, by combining gene cDNA supplementation and genome editing using HITI (from now on, Gene-DUET), we show long-term survival and amelioration of SMA phenotypes. Beside improving current therapies for SMA, our observations have implications for the treatment of other inherited diseases especially for neurodegenerative disorders as well as neuromuscular diseases.

AAV-mediated in vivo genome editing in spinal cord
AAV vector has emerged as one of the safest and most commonly used vectors for the delivery of therapeutic genes (12). In fact, AAVs have been used extensively in gene therapy clinical trials to patients with SMA, Duchenne muscular dystrophy (DMD) and X-linked myotubular myopathy patients (5,12,13). Over the past few decades, numerous engineered AAV capsids have been developed for effective gene delivery into various tissues, and current strategies for developing AAV vectors with tailored tropism are described (14). Recent report showed that engineered AAV-PHP.eB capsid enables efficient transduction in the central nervous system (15). Then, we first compared AAV-PHP.eB and AAV9 which is commonly used for clinical gene therapy of SMA patients. We injected GFP-expressing AAV (AAV-GFP) into neonatal wild-type (WT) mouse by intravenous injection and analyzed GFP expression in the brain, lung, heart, stomach, liver, spleen, pancreas, kidney, muscle and spinal cord 2 weeks later (Fig. 1, A and B, and fig. S1, A and B). The analyses of tissue sections showed that AAV-PHP.eB was more efficient in the spinal cord and brain than AAV9 (fig. S1, C to F). Notably, the GFP signals were abundantly merged with NeuN which is a motor marker in the spinal cord (Fig. 1C). These results demonstrate the advantage of AAV-PHP.eB for transduction into spinal motor neurons.
To evaluate the frequency of targeted gene knock-in in vivo, we previously used Ai14 mice carrying the CAG promoter at the Rosa26 loci ( Fig. 1D) (8). Since the HITI technology can perform in vivo knock-in in non-dividing cells represented by neural cells, we tested the in vivo efficacy of HITI in spinal cord by using Ai14-Cas9 mice which constitutively express Cas9. HITI-mediated GFP knock-in at the Rosa26 locus downstream of the CAG promoter was observed in the nucleus of the motor neurons and liver when the AAV-PHP.eB-Ai14-HITI including guide RNA (gRNA) expression cassette, GFPNLS-pA donor sandwiched by Cas9/gRNA target sequence and mCherry reporter were delivered into Ai14-Cas9 mice at postnatal day 0.5 (P0.5) (Fig. 1E and fig. S1, G and H). These results suggest that HITI-mediated genome editing is successful in the spinal motor neurons.
Targeted in vivo gene correction of SMA mice via HITI Next, to demonstrate the validity of the HITI technology for gene correction of SMA, we chose the SMA mouse (SMN2 +/+ ; SMNΔ7 +/+ ; Smn1 −/− ) as a disease model, which disrupts endogenous Smn1 gene (mSmn1) through lacZ reporter gene insertion into exon 2 and harbor two transgenic alleles of human SMN2 (16)(17)(18). To prevent the deleterious effect of CRISPR-Cas9-induced insertions/deletions (indels) in the endogenous exon, we targeted intronic sequences upstream of the exon2/lacZ in chromosome 13 ( Fig. 2A). Previously, we demonstrated that unknown recombination occurs within homologous sequence between genome and donor (19). To avoid such an unexpected event, we removed the homologous sequence on the donor by incorporating a portion of rat intron 1 of Smn1 including the splicing acceptor, codon-optimized mouse Smn1 cDNA (exons 2-8), and rat 3'UTR. The constructed pAAV-SMN1-HITI vector contains intron 1 targeting gRNA expression cassette and SMN1 donor sandwiched by two gRNA target sequences. We packaged the pAAV-SMN1-HITI and an AAV expressing Cas9 (pAAV-Cas9) with PHP.eB capsid, and systemically delivered the AAV-PHP.eB-SMN1-HITI and AAV-PHP.eB-Cas9 via intravenous injection into SMA mice at P0.5 (Fig. 2B). HITI-mediated gene knock-in was detected by PCR amplification only in treated tissue 2 weeks after injection (Fig. 2C). We also verified the corrected genome sequences in amplicons by Sanger sequencing (fig. S2A). Importantly, HITI-treated SMA mice showed phenotypic improvement compared to untreated SMA mice. HITI-treated SMA mice were able to walk independently whereas untreated SMA mice were unable to stand at 2 weeks old ( Fig. 2D and Movie S, 1 and 2). The mean survival and body weight were significantly increased in HITI-treated SMA mice than in untreated SMA mice (Fig. 2, E and F, and fig. S2B). However, the effect of HITI was not enough for SMA mice to survive more than 3 weeks even showing significant differences in behavior and survival analyses. These results suggest that HITI-mediated genome editing is successful in SMA mice and rescues SMA phenotypes, but is not sufficient for a therapeutic strategy for SMA.

Gene-DUET strategy for SMA through gene supplementation and genome editing
The body weight of SMA mice at birth was lower than that of WT and heterozygous litters, suggesting that SMA phenotypes were already advanced at the time of treatment ( fig. S2C).
HITI-mediated gene correction by AAV and the resulting SMN1 protein production is likely to take a several days, which is too late to rescue SMA mice effectively. Previous reports also showed that the timing of initial treatment and high SMN1 expression was important for SMA therapy (20,21).
Based on these results, we established a new strategy, Gene-DUET, which is a combination of two strategies through wild-type cDNA supplementation and genome editing. Wild-type cDNA supplementation was accomplished by overexpressing mouse Smn1 (mSmn1) cDNA as previous reports (22). We redesigned the AAV vector (AAV-SMN1-DUET) containing the mSmn1 coding sequence (CDS) under the CMV promoter and SMN1-HITI as used in the previous construct in Fig. 2A. Similar to current gene supplementation therapy, only mSmn1 CDS would be expressed in the absence of Cas9 (Fig. 3A). In contrast, mSmn1 CDS and the gene-corrected mSmn1/SMN1 fusion gene can be co-expressed in the presence of Cas9 (Fig. 3B). We systemically delivered AAV-PHP.eB-SMN1-DUET with or without AAV-PHP.eB-Cas9 via intravenous injection into SMA mice at P0.5 (Fig. 3, C and D). The cDNA alone (without Cas9) or DUET (with Cas9) treated SMA mice looked very healthy compared to untreated SMA mice in general appearance 2 weeks after the injections (Fig. 2D, 3E and Movie S, 3 and 4). Dissection of tissues at 2 weeks showed that cDNA or DUET treatments improved the size of the spinal cord, brain, heart and muscle (Fig. 3, F and G, and fig. S3, A to E). We evaluated motor function in each treated SMA mice by analyzing the righting reflex test 2 weeks after the injections. Untreated WT and heterozygous Smn1 mice were able to right themselves quickly, whereas the SMA mice took longer or were unable to right themselves within 30 seconds. The cDNA-or DUET-treated SMA mice significantly showed improvement of righting reflex compared to untreated SMA mice in both male and female mice (Fig. 3H). Some of the treated SMA mice exhibited ear and digital necrosis and shorter tails due to necrosis starting at 5 weeks of age until loss as reported in previous reports ( fig. S3F) (23,24). These results suggest that both cDNA and DUET treatments dramatically improve the phenotypes of SMA mice.

Molecular restoration in SMA mice by cDNA and DUET treatments
We next performed RNA sequencing (RNA-seq) of spinal cord samples to understand the global transcriptional alterations by HITI, cDNA and DUET treatments in SMA mice compared to untreated SMA mice and control heterozygous mice. Principal component analysis (PCA) and heatmap analyses clearly segregated untreated SMA mice from the healthy heterozygous mice ( Fig.   4A and fig. S4A). The profile of cDNA-and DUET-treated SMA samples was placed close to that of heterozygous mice, suggesting restoration by treatments against SMA-induced molecular dysfunction. A functional gene enrichment analysis revealed that inflammatory pathways including p53 signaling pathway and cytokine-cytokine receptor interaction were up-regulated while motor neuron pathway including cholinergic synapse was down-regulated in SMA mice compared to heterozygous mice (Fig. 4, B and C, and fig. S4, B and C). These molecular changes were restored by both cDNA-and DUET-treatments but not restored by HITI treatment. (Fig. 4D). We performed quantitative reverse transcription PCR (RT-qPCR) and confirmed that activated p53 downstream genes including Gtse1, Ccng1, Perp and Sesn1 were significantly repressed in cDNA-and DUET-treated spinal cord compared to untreated SMA samples ( fig. S4D). These results suggest that both cDNA-and DUET-treatments restore the molecular dysfunction in the spinal cord of SMA mice.

Stable gene correction mediated by Gene-DUET
General appearance at 20 weeks old was not different between cDNA-and DUET-treated SMA mice ( Fig. 5A). In terms of body weight, cDNA-or DUET-treated SMA mice showed dramatic improvement compared to untreated SMA mice (Fig. 5B). Locomotor activity of cDNA-and DUET-treated SMA mice at 20 weeks old was similar to that of heterozygous mice, indicating that motor function was significantly improved even into adulthood ( fig. S5). The concern with neonatal AAV treatment is the loss of the transgene along with cell division during tissue growth. Previous report showed quick reduction of vector genome copies in the liver over a few weeks after neonatal AAV treatment (25). Indeed, GFP expression in the spinal cord was declined after 1 year compared to 2 weeks in AAV-PHP.eB-GFP injected WT mice (Fig. 1B, and fig. S6A). We also found a significant reduction in cDNA-derived exogenous mSmn1 expression in the spinal cord and liver at 1 year old compared to at 2 weeks old in cDNA-treated heterozygous mice ( fig. S6, B and C).
Theoretically, we thought that genome editing by HITI might be more advantageous than cDNA supplementation for permanent cure. To examine the efficiency of gene correction in DUET-treated SMA mice, we extracted genomic DNA from the treated mice and enriched the targeting regions by using customized probes that could hybridize and pull down the genomes around the Cas9 cleavage site (Fig. 5C). We performed the deep sequencing and calculated the editing efficiency in 20-and 40-weeks old DUET-treated tissues. The corrected sequences were detectable and stable in all tissues from these DUET-treated SMA mice (Fig. 5, D and E). These results suggest that gene correction by Gene-DUET strategy is stable for a long time in SMA mice. Crucially, both cDNA and DUET treatments significantly enhanced the survival of SMA mice compared to untreated SMA mice. More importantly, the mean survival of DUET-treated SMA mice was clearly improved over that of cDNA-treated SMA mice, especially in males (Fig. 5, F and G). These data suggest the synergistic effect of the Gene-DUET strategy by gene supplementation and genome editing.

DISCUSSION
Gene therapy medicine, Zolgensma, for treating SMA by cDNA supplementation has been recently approved by FDA. However, the long-term efficacy of this treatment has not yet been determined.
Indeed, our data show that exogenous Smn1 expression is attenuated in spinal cord and liver cells one year after AAV injection. The development of HITI technology has enabled the correction of genomic mutations in non-dividing cells through the NHEJ pathway and here we provide a first demonstration of its utility in an SMA model in mice. Compared to cDNA supplementation, the in situ gene correction approach here reported is able to sustain stable and permanent Smn1 expression.
However, due to the widespread and profound phenotypic alterations that arise immediately after birth, gene correction alone was not sufficient to correct all of them although it is able to correct a subset of some of the phenotypic changes observed in SMA mice. The extent of cell and tissue phenotypic restoration and increase in survival time were observed when combining HITI-mediated genome editing and gene supplementation in SMA mice. We observed a more significant survival benefit in male mice than in female mice by Gene-DUET. This is likely due to sex differences in SMA phenotypes and tissue growth differences between male and female mice, with exogenous mSmn1 being retained more in female than in male mice (26). Previous report showed that sex-specific amelioration of SMA phenotype by an antisense oligonucleotide treatment (27). In fact, we also observed that the mean survival of cDNA-treated female SMA mice was two times longer than that of cDNA-treated male SMA mice.
Recently, the base editing combined with antisense oligonucleotide can ameliorate the SMA phenotypes (28), however that therapeutic effects depend on the copy number of endogenous SMN2, which may not be sufficient for SMA type 0 and type 1 patients. As we have shown in this work, the Gene-DUET supports therapeutic benefit for SMA mice compared to conventional gene supplementation therapy. In summary, our Gene-DUET strategy provides new exploratory avenues for the treatment of SMA in humans. This approach has a great potential for the field of genome-editing technologies that may be relevant for the treatment of inherited diseases, particularly neurodegenerative and neuromuscular diseases.

AAV production
AAV9 and AAV-PHP.eB viral particles were generated by or following the procedures of the Gene Transfer Targeting and Therapeutics Core at the Salk Institute for Biological Studies.

Facial vein AAV injection
The newborn (P0.5) mice were used for intravenous AAV injection as following previous report (29).
The AAV mixtures were injected via temporal vein of the newborn mice. After the injection, pups were allowed to completely recover on a warming pad and then returned to the home cage.

Intraspinal AAV injection
Neonatal pups were used for spinal cord AAV injection as following previous report(30). Briefly, P0.5 mice were anesthetized and 2 μl of AAV mixtures was slowly injected into the spinal cord using 33 Gauge Neuros syringe (65460-06, Hamilton,). After the injection, pups were allowed to completely recover on a warming pad and then returned to the home cage.

Genome extraction
Genomic DNA was extracted from cells and tissue samples using DNeasy Blood & Tissue Kits (69506, Qiagen) following the manufacturer's instruction.

RNA Analysis
Total RNA was extracted from cells and tissue samples using either TRIzol (Invitrogen) or RNeasy Kit (Qiagen) followed by cDNA synthesis using Maxima H Minus cDNA Synthesis Master Mix (Thermo Fisher). Real-time qPCR was performed using SsoAdvanced SYBR Green Supermix and analyzed using a CFX384 Real-Time system (Bio-Rad). All analyses were normalized based on amplification of mouse Gapdh.

Immunohistochemistry
Tissues were harvested after transcardial perfusion using ice-cold PBS, followed by ice-cold 4%

Image Capture and Processing of Primary Neurons
Immunocytochemistry samples of mice samples were visualized by confocal microscopy using a Zeiss LSM 710 Laser Scanning Confocal Microscope (Zeiss). Images were processed by ZEN2 Black edition software (Zeiss).

Righting reflex test
The righting reflex of untreated or treated mice was compared at postnatal 14 days. Mice were laid on their back and the time needed to flip over was recorded, with a maximum of 30 sec allowed.
Three trials were performed for each mouse and the shortest time was used for analysis.

Data and software availability
The accession number for the RNA-seq data reported in this paper is NCBI GEO: GSE207181.

Acknowledgments
We are grateful to Grace Chou, Tzu-Wen Wang, Ling Ouyang and Nasun Hah for next-generation

Competing interests:
We declare that none of the authors have competing financial or non-financial interests.

Data and materials availability:
Data supporting the findings of this study are available within the paper and its supplementary information files. RNA-sequence data was deposited in the Gene Expression Omnibus under the accession number GSE207181.