CRISPR/dCas9-based Scn1a gene activation in inhibitory neurons ameliorates epileptic and behavioral phenotypes of Dravet syndrome model mice

Dravet syndrome is a severe infantile-onset epileptic encephalopathy which begins with febrile seizures and is caused by heterozygous loss-of-function mutations of the voltage-gated sodium channel gene SCN1A. We designed a CRISPR-based gene therapy for Scn1a-haplodeficient mice using multiple guide RNAs (gRNAs) in the promoter regions together with the nuclease-deficient Cas9 fused to transcription activators (dCas9-VPR) to trigger the transcription of SCN1A or Scn1a in vitro. We tested the effect of this strategy in vivo using an adeno-associated virus (AAV) mediated system targeting inhibitory neurons and investigating febrile seizures and behavioral parameters. In both the human and mouse genes multiple guide RNAs (gRNAs) in the upstream, rather than downstream, promoter region showed high and synergistic activities to increase the transcription of SCN1A or Scn1a in cultured cells. Intravenous injections of AAV particles containing the optimal combination of 4 gRNAs into transgenic mice with Scn1a-haplodeficiency and inhibitory neuron-specific expression of dCas9-VPR at four weeks of age increased Nav1.1 expression in parvalbumin-positive GABAergic neurons, ameliorated their febrile seizures and improved their behavioral impairments. Although the usage of transgenic mice and rather modest improvements in seizures and abnormal behaviors hamper direct clinical application, our results indicate that the upregulation of Scn1a expression in the inhibitory neurons can significantly improve the phenotypes, even when applied after the juvenile stages. Our findings also suggest that the decrease in Nav1.1 is directly involved in the symptoms seen in adults with Dravet syndrome and open a way to improve this condition.


Introduction
Patients with Dravet syndrome usually show normal early development, but develop epileptic (tonic-clonic or clonic) seizures at 4-10 months after birth (Dravet, 2011). Epilepsy often begins with complex febrile seizures, and the non-febrile seizures follow in the later stage of the disease. Status epilepticus is common, and sudden unexpected death is not rare. In the second year of life, psychomotor regressions, such as severe intellectual disability and autistic features, also appear. Because epilepsy in Dravet syndrome is mostly drug-resistant, effective therapies are actively required. Heterozygous loss-offunction mutations in the SCN1A gene encoding the voltage-gated sodium channel alpha-1 subunit Nav1.1 have been reported in approximately 80% of Dravet patients (Lossin, 2009;Escayg and Goldin, 2010;Yamakawa, 2016). We and others have developed mouse models for Dravet syndrome, and suggested that Nav1.1 haploinsufficiency in the GABAergic neurons, especially in the parvalbumin-positive (PV+) ones, is the major cause of the epileptic seizures, sudden death, intellectual disability and autistic features of Dravet syndrome (Yu et al., 2006;Ogiwara et al., 2007;Han et al., 2012;Ito et al., 2013;Tatsukawa et al., 2018). Importantly, we further showed that Nav1.1 is also expressed in a distinct subpopulation of excitatory neurons and its haploinsufficiency in the excitatory neurons has a contrasting effect of ameliorating the epileptic seizures and sudden death in the Scn1a-deficient mice . This indicates that the therapeutic approaches may have to be specifically targeted to the inhibitory, rather than the excitatory neurons to secure their effectiveness and avoid unpreferable side effects.
Supplementing wild-type Nav1.1 proteins via viral gene delivery systems could be a therapeutic approach for Dravet syndrome. However, cDNA encoding Nav1.1 is too large (~6 kb) to be packaged into the currently-available viral systems including adeno-associated virus (AAV) or lentivirus (Wu et al., 2010;Kumar et al., 2001). One of the choices to treat Dravet syndrome is a CRISPR-ON system that specifically activates transcription of the target gene (Gilbert et al., 2013;Cheng et al., 2013;Maeder et al., 2013;Perez-Pinera et al., 2013). In this system,~100 nucleotides long guide RNAs (gRNAs) contain~20 nucleotide sequence complementary to the promoter regions of the target gene, and thereby direct the nuclease-dead (inactivated) Cas9 (dCas9) protein fused to a transcription activation domain to the promoter region. Activities of the first generation dCas9 transcriptional activators, such as dCas9-VP64 or -VP160 (Gilbert et al., 2013;Cheng et al., 2013;Maeder et al., 2013;Perez-Pinera et al., 2013) had a low efficiency. More potent second-generation transcriptional activators have been developed, such as Suntag in which the dCas9 with a repeating peptide array recruits multiple copies of antibody-fused VP64 (Tanenbaum et al., 2014), SAM in which the loops of guide RNA attracts multiple additional transcription factors MS2-p65-HSF1 (Konermann et al., 2015), or VPR in which the dCas-VP64 is combined with two additional transcriptional activation domains namely the NF-kB transactivating subunit p65 and viral transcription factor Rta (Chavez et al., 2015). Among these second generation activators, the dCas9-VPR seems advantageous for in vivo applications because this system is the simplest and requires a smaller number of DNA constructs than other systems. In this study, we therefore employed the dCas9-VPR to upregulate Scn1a transcription in the Scn1a-haplodeficient Dravet syndrome model mouse (Scn1a RX/+ ) which contains a heterozygous disease-causing mutation (R1407X) (Ogiwara et al., 2007). Our goal in the present work was to investigate whether increasing Scn1a expression specifically in inhibitory neuron would be sufficient to improve the phenotypes associated with the haploinsufficiency in Scn1a RX/+ mutant mice, potentially providing a proof of concept that such a strategy could be a viable therapeutic approach as well as further strengthening the idea that inhibitory neurons are the major actor in the disease. We show that the CRISP-ON system successfully enhances the transcription of Scn1a in the inhibitory, especially the PV+, neurons in Scn1a RX/+ mouse brain and significantly ameliorates their epileptic and behavioral phenotypes.

Study design
The primary objective of this study was to examine the CRISPR-ONdependent Scn1a gene activation as a therapy for the epileptic phenotypes including febrile seizures and seizure-associated sudden death of the Scn1a-haplodeficient Dravet syndrome model mice. The CRSIPR-ON system consists of the dCas9-VPR, Vgat-Cre and gRNAs in order to establish the inhibitory neuron-specific activation. gRNAs were designed for the mouse Scn1a and human SCN1A, and their efficiencies were evaluated initially in cell cultures by qRT-PCR and northern blot analyses. The gRNAs that were effective were then packaged into AAV and introduced into the model mice. The dCas9-VPR and Vgat-Cre alleles were introduced by mating with transgenic mouse lines. The febrile seizures of the CRISPR-ON-treated model mice were evaluated in a heat-controlled chamber. The accelerations of Nav1.1 expression and their histological distributions in the mice with improvements of febrile seizures were evaluated by qRT-PCR, western blot, and immunohistochemistry analyses. The control and treatment groups and the number of biological replicates (sample sizes) for each experiment are specified in the figures or their legends. We did not perform power analysis to predetermine the sample size. During the data analysis, the investigators were not blinded for the in vitro experiments. The febrile seizure test and ECoG recording were performed in a randomized blinded manner, and the randomization was decoded at the time of data analysis. The animals in the control and treatment groups were housed together to minimize the environmental differences and experimental bias.

Animal work statement
All animal experimental protocols were approved by the Animal Experiment Committee of the RIKEN Center for Brain Science. Mice were handled in accordance with the guidelines of the RIKEN Center for Brain Science Animal Experiment Committee. Food and water were available ad libitum, and the cages (less than 5 animals) were kept at 23°C and 55% humidity on a 12-h light/dark cycle with the lights off at 20:00.

Construction of the gRNA expression plasmids
For the in vitro cell culture experiments, each gRNA expression plasmids (pMLM3636-gRNA) was constructed by inserting each of the gRNA sequence into the MLM3636 plasmid (a gift from Dr. Keith Joung; Addgene #43860) as previously described (Maeder et al., 2013).

Evaluation of the gRNAs in cell culture
The Neuro2a cells and Hek293FT cells were grown in DMEM (Thermo Fisher Scientific) supplemented with 10% FBS. 1× 10 5 cells were plated in each well of a 6-well plate 2 days before transfection. The cells were grown until 80% confluency in each well. Two μg of SP-dCas9-VPR (Chavez et al., 2015) (a gift from Dr. George Church; Addgene #63798) which expresses dCas9-VPR and 500 ng of pMLM3636-gRNA, for the respective gRNAs, were delivered to each well with Lipofectamine LTX (Thermo Fisher Scientific), according to the manufacturer's instruction. For a mixture of five guide RNAs to test their synergistic effects, 100 ng of each of the gRNA plasmid (500 ng in total) was used. Forty eight hours after transfection, the cells were lysed for RNA purification.
Each PCR was done in triplicate. Relative quantification was performed using the modified ΔΔCt method. For the assays using the cultured cells, the no-DNA or vector DNA (MLM3636 plasmid without targeted gRNA sequence) transfected cells were used as the negative controls.

Northern blot analysis
Five μg of cellular and 20 μg of mouse brain RNA, prepared as described above, were separated by electrophoresis on a 1% agarose formaldehyde gel, and transferred to nylon membranes (positively charged; Roche Diagnostics, Indianapolis, IN, USA). The 615-bp 3′-UTR of Scn1a was used as a probe as described previously (Ogiwara et al., 2007). Labeling, hybridization, and detection of the probe was carried out using the DIG Northern Starter Kit (Roche Diagnostics).

AAV production and quantification
AAV particles were produced as described previously (Kobayashi et al., 2016) using a packaging plasmid AAV-PHP.eB (this plasmid was a gift from Dr. Viviana Gradinaru) (Chan et al., 2017). The PHP.eB serotype exhibits very efficient transduction of the central nervous system via intravenous injection in adult animals. The copy number of the viral genome (vg) was determined by quantitative PCR using the gRNA gene-specific primers Pr9F (5′-TTCCTTCACTCCAGATGTGA-3′) and Pr9R (5′-TGCGGCCAATTCGCCCTTTC-3′) for the AAV-4xU6-sgRNA. This primer pair was also used for the detection of the gRNA gene in the mouse brain by PCR analysis (Supplemental Fig. S5).

In vivo administration of the AAV vector
The AAV-PHP.eB particles containing the AAV-4xU6-sgRNA construct were resuspended in 150 μL of saline. The mice were anesthetized with inhalational anesthetic isoflurane (2%) and placed on a warm stand at 38°C. The AAV particles were then injected into the tail vein of the mice at a dose of 1.8 × 10 11 vg/mouse by using 29 gauge needle syringe.

Febrile seizure test (FS test)
FS tests were performed on CRISPR-ON-treated and control mice at 6, 8, 10, and 12 weeks after birth. The protocol of FS-test was reported previously (Tatsukawa et al., 2018). Briefly, the mice were placed onto a perforated horizontal partition in a hermetically closed Plexiglas box and heated by blowing hot air from below. Before increasing the air temperature, the mice were kept at 37°C for at least 3 min. We gradually elevated the body temperature by raising the temperature of the hot air by 0.5°C per 1 min and monitored the mouse behavior by video camera. The rectal temperature was continually monitored and temperature at baseline and seizure onset recorded. When a seizure commenced, the mouse was promptly moved to an ice-cold box until the normal body temperature was reached. To evaluate the gender difference in the FS-test, we used the Scn1a RX/+ mice in the 129 +Ter /SvJcl background (> N25).
To assess the severity of hyperthermia-induced seizures, we monitored adult mice (21 weeks-old males) in the above-described chamber at a fixed 43°C temperature. Mice behavior was video-monitored, with a cut-off time of 30 min. The latency to myoclonic, clonic (jerks and wild jumps) and generalized tonic-clonic seizures (GTCS) as well as the duration of GTCS were measured by an experimenter blind to the genotype and treatment of the mice.

Electrocorticographic (ECoG) recordings
Adult mice (23 weeks-old males) were subjected to surgery for ECoG recording. Stainless steel screws (1.1 mm diameter) serving as ECoG electrodes were placed over the bilateral somatosensory cortex ( ± 1.5 mm lateral to midline, 1.0 mm posterior to bregma) under 1-1.5% isoflurane anesthesia. A reference screw electrode was implanted in contact to the cerebellum (at midline, 1.5 mm posterior to lambda). A stainless steel wire bipolar electrode (100 μm diameter) was inserted in the cervical region of the trapezius muscle for electromyogram (EMG) recording. After at least one week of recovery from surgery, brain activity was recorded for 3 consecutive days (at a 256 Hz sampling rate) and analyzed off-line (SleepSign, Kissei Comtec, Japan). Each animal's behavior was continuously monitored using an infrared camera. ECoG/EMG recordings were analyzed manually offline by a reviewer blind to treatment and genotype. Abnormal events were quantified in every mouse from 12 h of recording composed of two 6-h windows selected during the dark phase from day 2 to day 3 and the light phase on day 3.

Open-field
The Open Field test is used to assess spontaneous exploratory behavior. Mice (6 wild-type, 6 Scn1a RX/+ and 5 CRISPR-ON treated Scn1a RX/+ ; 12-weeks old) were placed in a 50 cm × 50 cm square automated open-field homogeneously illuminated at 70 lx and allowed to freely explore for 30 min. Spontaneous activity was recorded and processed using manufacturer's tracking software (Time OFCR4; O'Hara & Co, Ltd., Tokyo, Japan). Traveled distance and percentage of time spent in the center area (defined as a 30 cm × 30 cm square area at the center of the open field) were automatically measured.

Elevated plus maze
The elevated plus maze is commonly used to investigate anxiety in rodents. It consists of two open arms (25 × 5 cm) crossing two enclosed arms of the same size with 15 cm high transparent walls and was placed 50 cm above the floor. To minimize the likelihood of animals accidentally falling from the apparatus, open arms were equipped with 3 mm high Plexiglas ledges. Luminosity was homogeneously set at 70 lx. Mice (6 wild-type, 6 Scn1a RX/+ and 5 CRISPR-ON treated Scn1a RX/+ ; 13-weeks old) were placed at the center of the maze, facing one of the open arms and were allowed to freely explore the maze for 10 min. Animals behavior was recorded and analyzed using manufacturer's tracking software (Time EP2; O'Hara & Co, Ltd., Tokyo, Japan).

Social behavior: 3-chambers task
We used the 3-chamber test to assess sociability and preference for social novelty, using a 43 cm × 63 cm transparent Plexiglas box, separated into 3 equivalent-sized (21 cm × 43 cm) chambers by transparent Plexiglas plates with 10 cm × 10 cm square openings allowing the mice to move freely from a chamber to another. One wire cylindershaped cage (10 cm diameter, 15 cm high) was placed at the center of the side chambers and were used to enclose 10 weeks old C57BL/6 J males during the test. These stranger mice were trained to be encaged prior to the main experiment. Tested mice (6 wild-type, 6 Scn1a RX/+ and 5 CRISPR-ON treated Scn1a RX/+ ; 16-weeks old) were first placed in the center chamber and allowed to freely explore the apparatus for a 10 min habituation period. A first stranger mouse was then enclosed randomly in one of the side-chambers and the tested mice were given 10 min exploration to assess sociability toward an unknown subject. In a final step a second stranger was enclosed in the opposite side-chamber and the tested mice were given another 10 min exploration time to assess their preference for social novelty. Mice behavior was video-recorded and interactions (number and duration) as well as time spent in the different chambers were analyzed by an experimenter blind to the genotype/condition.

Statistical analyses
Data are presented as the mean ± SEM. Statistical analyses were performed by KyPlot 5.0 (KyensLab Inc., Tokyo, Japan). The assay for gRNA in the cultured cells was analyzed by one-way ANOVA followed by multiple comparisons using Dunnett's test. Other data from three or four genotype groups was analyzed by one-way or two-way repeated measures ANOVA followed by multiple comparisons using Tukey or Tukey-Kramer method. P-value smaller than 0.05 was considered statistically significant.

Upregulation of Scn1a or SCN1A gene transcription by CRISPR-ON in vitro
Both the mouse Scn1a and human SCN1A genes have upstream and downstream promoter regions (Martin et al., 2007;Nakayama et al., 2010). We designed 11 gRNAs (six upstream and five downstream) in the mouse Scn1a (Fig. 1a) and eight gRNAs (four upstream and four downstream) in the human SCN1A (Fig. 1d) promoter regions on their forward strands. One gRNA (hmSC1U3) was compatible for both the mouse Scn1a and human SCN1A promoters. gRNAs were selected at 1 bp to 600 or 900 bp upstream from the respective transcription start sites (TSS), according to the previous studies (Cheng et al., 2013;Maeder et al., 2013;Konermann et al., 2015). The gRNAs consist of 110 nucleotides (nt) with 20-nt sequences complementary to the genomic sequences adjoining the protospacer adjacent motif (PAM: 5′-NGG-3′) (Supplemental Table S4; Supplemental Fig. S6 and S7). To avoid offtargeting effect, the targeting-sequences were selected not to be identical or highly homologous to other genomic regions by BLASTN search.
To evaluate the effect of the gRNAs in vitro the DNA constructs that express the mouse and human gRNAs, and dCas9-VPR were co-transfected into Neuro2a (Fig. 1b and c) and Hek293FT (Fig. 1e) cell lines, and evaluated by qPCR. Previous studies reported the synergistic effects of multiple gRNAs in the upregulation of gene transcriptions (Cheng et al., 2013;Maeder et al., 2013). We therefore tested each single gRNA separately and their mixture as well. Quantitative (q) RT-PCR and/or northern blot analyses revealed that the upstream gRNAs effectively increased Scn1a or SCN1A transcription, and a mixture of four gRNAs (mSC1U1, 2, 4 + hmSC1U3 for mouse; hSC1U1, 2, 4 + hmSC1U3 for human) showed further stronger effects, while downstream promoters were comparatively less effective in both the mouse ( Fig. 1b and c) and human genes (Fig. 1e). Interestingly, the promoting regions that we targeted to generate our gRNAs include segments that have been reported to be conserved non-coding sequences (CNS) with a high conservation across species (Martin et al., 2007).
In addition, to evaluate the specificity of our CRISPR-ON system, we analyzed the expression of the main sodium channel subunits (Scn2a, Scn3a, and Scn8a) by qPCR (Supplemental Fig. S8). A small yet significant decrease in Scn2a mRNA level was observed, but the Scn1a CRISPR-ON system did not enhance non-specifically Scn2a, Scn3a, or Scn8a genes expression in Neuro2a cells.

Generation of triple mutant mice and AAV vector of gRNAs to establish inhibitory neurons-specific CRISPR-ON in vivo
We designed an inhibitory neuron-specific CRISPR-ON of Scn1a to treat the Scn1a RX/+ Dravet syndrome model mice (Fig. 2a); the floxed-dCas9-VPR VPR/+ mice (see below) were mated with the Scn1a RX/+ mice to obtain the floxed-dCas9-VPR VPR/+ /Scn1a RX/+ mice. The resulting male mice were further interbred with Vgat-Cre Cre/+ female mice  by in vitro fertilization to obtain the floxed- dCas9-VPR VPR/+ /Vgat-Cre Cre/+ /Scn1a RX/+ triple mutant mice, and the pups were intravenously injected with the adeno-associated virus (AAV) which harbored the synergistically-effective four mouse gRNAs (mSC1U1, mSC1U2, hmSC1U3, and mSC1U4). Because of the possible aggravating effect of the increased Scn1a expression in the excitatory neurons , we restricted the expression of dCas9-VPR to the in inhibitory neurons in the mice in the protocol. Although the Nav1.1 haploinsufficiency in the PV + inhibitory neurons is the major basis for the Dravet syndrome as mentioned, Parv-Cre is rather late-onset and condition-dependent and therefore we used vesicular GABA transporter (Vgat)-Cre which is potent and stably expressed at high levels in the all inhibitory neurons .
To generate the floxed-dCas9-VPR VPR/+ mice, a targeting construct Ai-VPR was prepared using the SP-dCas9-VPR (Chavez et al., 2015) and the Ai9 (Madisen et al., 2010) constructs (Supplemental Fig. S2), in which the dCas9-VPR segment of SP-dCas9-VPR was amplified by PCR and inserted into and replaced the tdTomato segment of Ai9. The Ai-VPR targeting construct was then knocked-in into the Rosa26 locus, which has been used for achieving generalized expression safely (Chu et al., 2016), in mouse ES cells (Supplemental Fig. S3a). One ES cell line (9G) was selected to generate the floxed-dCas9-VPR knock-in mouse ( Fig. 2b; Supplemental Fig. S3 b and c). To evaluate the Cre-dependent dCas9-VPR expression in the mice, we generated a GABAergic neuron specific and constitutive dCas9-VPR expressing mouse line by crossing the floxed-dCas9-VPR mouse with a Vgat-Cre  and EIIa-Cre (Lakso et al., 1996) driver mice, respectively (Supplemental Fig. S4). The floxed-dCas9-VPR VPR/+ /EIIa-Cre Cre/+ mice (F2) showed severe growth retardation and died early postnatally (Supplemental Fig.  S9), indicating the toxicity of dCas9-VPR when expressed constitutively at the embryonic or early postnatal stages. In contrast, the floxed-dCas9-VPR VPR/+ /Vgat-Cre Cre/+ mice were born normally and were fertile. Western blot analysis with anti-Cas9 antibody showed that the dCas9-VPR protein was only observed in the Vgat-Cre Cre/+ positive floxed-dCas9-VPR VPR/+ mouse brains (Fig. 2c). Immunohistochemical analysis of the floxed-dCas9-VPR VPR/+ /Vgat-Cre Cre/+ mouse brain further showed that the distribution of the dCas9-VPR positive cells was similar to that of the Vgat-positive GABAergic neurons  in the olfactory bulb, cerebral cortex, hippocampus, striatum, cerebellum and medulla and pons, but the dCas9-VPR signals were not observed in the mice without the Vgat-Cre (Supplemental Fig. S10).
To obtain the floxed-dCas9-VPR VPR/+ /Vgat-Cre Cre/+ /Scn1a RX/+ triple mutant mice, we performed in vitro fertilization using the sperm of floxed-dCas9-VPR VPR/+ /Scn1a RX/+ and oocytes of Vgat-Cre Cre/+ mice. Fifty out of the 497 pups (10.1%) were triple mutant mice (Fig. 2d). Before the AAV-gRNAs injections, we evaluated the survivability of all the pups generated (Fig. 2d). 21.4% of Scn1a RX/+ mice died by P30, which is consistent to what we observed in the past for this model (22.2% lethality rate at P35 in an independent study) . Surprisingly, the survivability of the triple mutant mice was much lower than that of the floxed-dCas9-VPR VPR/+ /Vgat-Cre Cre/+ mice or that of the Scn1a RX/+ mice. The floxed-dCas9-VPR VPR/+ /Vgat-Cre Cre/+ mice also showed partial lethality. These observations suggest that the dCas9-VPR itself was toxic, and its expression in the inhibitory neurons aggravates the epileptic seizures and sudden death in the Scn1a-haplodeficient mice when expressed at the embryonic or early postnatal stages. Our previous work (Ogiwara et al., 2007) as well as other studies suggest that spontaneous seizures are the likely cause of death in pups before P30 but that past P30 the frequency of such seizures, and hence the lethality decreases. This is likely due to the recovery of PV-positive inhibitory neurons function observed in adult stage (Favero et al., 2018). As mice survive past P30 and advance toward adult stages, spontaneous seizures that lead to sudden death in juvenile mice became less in adult stages and it may be by the innate recover of action potential generation in PV interneurons and the apparent toxicity of dCas9-VPR in combination to the Scn1a mutation also decreases. Although it could be argued that this lethality might introduce a survivor bias for the following in-vivo studies, the fact that mice in the Scn1a haploinsufficient control groups are also affected by such effect and that animals in these control groups were selected randomly from a large pool of samples (9 mice out of 44 available for Scn1a RX/+ group for example) makes it unlikely that a significant bias is affecting the results and conclusions presented hereafter.
To prepare the AAV particles containing the gRNAs, we made a DNA construct (AAV-4xU6-sgRNA) that expresses the 4 gRNAs (mSC1U1, mSC1U2, hmSC1U3, and mSC1U4) using pAAV-MCS ( Fig. 2a; Supplemental Fig. S1). The construct was then packaged into the AAV-PHP.eB virus particles, which show highly efficient infection activity in neurons (ex.~70% cortical neurons), by intravenous injection into the mice (Chan et al., 2017). The AAV particles were then delivered into a subpopulation of the surviving pups generated by the in vitro fertilization of the floxed-dCas9-VPR VPR/+ /Scn1a RX/+ and Vgat-Cre Cre/+ mice ( Fig. 2a and d) via tail vein injection at postnatal 4 weeks (4w), which is the time after the onset of epileptic seizures in most of the Scn1a RX/+ mice (Ogiwara et al., 2007). None of the triple mutant mice that received the AAV-4xU6-sgRNA injection (n = 11) died until 12w, while two out of the 11 floxed-dCas9-VPR VPR/+ /Scn1a RX/+ mice with the AAV-4xU6-sgRNA injection died before 12w.

Inhibitory neurons-specific CRISPR-ON ameliorates febrile seizures in Scn1a-haplodeficient mice
Scn1a-haplodeficient mice show increased susceptibility of hyperthermia-induced seizures, mimicking those in the patients with Dravet syndrome (Tatsukawa et al., 2018;Cao et al., 2012;Verbeek et al., 2013). We therefore subjected the floxed-dCas9-VPR VPR/+ /Vgat-Cre Cre/+ /Scn1a RX/+ triple mutant mice that received the AAV-gRNA injection (CRISPR-ON-treated Scn1a RX/+ mice) to the febrile seizure (FS) test, and the efficacies were measured at multiple time points (Fig. 3a). In the FS-test at 6w, all the mice with the wild-type allele for Scn1a (Scn1a +/+ ) did not show seizures at 44°C, which is the maximum temperature in the FS test (Fig. 3b). The CRISPR-ON-treated Scn1a RX/+ mice showed a tendency of increase in the temperature threshold for FS compared to the other control mice with the haplodeficient Scn1a allele (Scn1a RX/+ ), although the difference did not reach to a statistical significance at 6w. We further continued the FS test at 8w, 10w, and 12w on the three mouse lines, the CRISPR-ON-treated Scn1a RX/+ mice (n = 11), the floxed-dCas9-VPR VPR/+ /Scn1a RX/+ mice Fig. 2. Generation and evaluation of the floxed-dCas9-VPR VPR/+ /Vgat-Cre Cre/+ /Scn1a RX/+ triple mutant mice for CRISPR-ON. (a) Experimental design to establish CRISPR-ON in the mice. (b) Schematic diagram of the Rosa26 targeted floxed-dCas9-VPR gene. The transgene is driven by the ubiquitous CAG promoter and is interrupted by a loxP-stop (3 polyA signal)-loxP cassette to render the dCas9-VPR expression inducible by the Cre recombinase. Arrowheads indicate the targeting positions of the gRNAs for recombination. Arrows indicate the PCR primers for detection of the Cre/loxP recombination. (c) Western blot of total proteins extracted from one complete hemisphere of floxed-dCas9-VPR VPR/+ /Vgat-Cre Cre/+ , wild-type, Vgat-Cre Cre/+ , and floxed-dCas9-VPR VPR/+ littermates using anti-Cas9 antibody. Only the floxed-dCas9-VPR VPR/+ /Vgat-Cre Cre/+ mice showed the dCas9-VPR expression (arrows). GAPDH; control. (d) 12-week survival rates post-birth for the offspring from the mating of floxed-dCas9-VPR VPR/+ /Scn1a RX/+ and Vgat-Cre Cre/+ mice. At P30 (vertical gray line), monitoring of most groups were terminated as some groups were either partially or completely disposed for technical and logistic reasons. The floxed-dCas9-VPR VPR/+ /Scn1a RX/+ /Vgat-Cre Cre/+ triple mutant offspring exhibited a significant reduction in survivability. Mice in the Scn1a RX/+ and flox-dCas9-VPR/ Scn1a RX/+ groups also suffered significant losses, though to a lesser extent. After CRISPR-ON treatment, the triple mutant mice did not exhibit sudden death up to the end of the FS-test whereas additional losses were seen in the Scn1a RX/+ and flox-dCas9-VPR/ Scn1a RX/+ groups. with AAV-gRNA injection (n = 9), and the Scn1a RX/+ mice (n = 9). Notably, all the CRISPR-ON-treated Scn1a RX/+ mice survived until the FS test at 12w, but two of the floxed-dCas9-VPR VPR/+ /Scn1a RX/+ mice with AAV-gRNA injection and two of the Scn1a RX/+ mice died during the series of FS tests. Results of the mice that survived until the FS test at 12w are summarized in Fig. 3c. Interestingly, the temperaturethreshold of the febrile seizures for the CRISPR-ON-treated Scn1a RX/+ mice was significantly increased compared to that of the floxed-dCas9-VPR VPR/+ /Scn1a RX/+ mice with AAV-gRNA injection or that of the Scn1a RX/+ mice (Fig. 3c). We retrospectively confirmed the Vgat-Credependent recombination of the floxed-dCas9-VPR allele and the introduction of the gRNA expression vectors in these mice (Supplemental Fig. S5). Although these data were obtained on mice with mixed gender (see Fig. 3c legend), it is known that the patient's gender is not associated with the risk of Dravet syndrome and we also found that the Scn1a RX/+ mice did not show gender difference in the FS test (Supplemental Fig. S11); therefore, the gender of the mice should not affect the conclusions on the effects of CRISPR-ON treatments. Exposing mice to repetitive rounds of hyperthermia may have the potential to affect their sensitivity to hyperthermia in the tests at 8, 10, and 12 weeks, we however did not observe a significant drift in susceptibility to hyperthermia induced seizures throughout the successive tests in any of the groups, including wildtype controls that never displayed generalized seizures at any of the tested stages. The temperatures thresholds observed in the FS test in our Scn1a RX/ + mice (40.9°C) appear higher than in human cases in which most of patients develop seizures below 38.5°C (Verbeek et al., 2013) but such higher temperature threshold is a common pattern in Scn1a RX/+ studies across independent research groups (39.3°C (Oakley et al., 2009) and 40.6°C (Hsiao et al., 2016)). Importantly, the protocols used to induce hyperthermia, measure body temperature and detect seizures in mice seem to lead to different temperature thresholds observed between different research groups, whereas we did not observe significant variations between different batches of mice analyzed by different experimenters using identical systems in our laboratory. As a conclusion, despite higher temperature thresholds than observed in human cases, Scn1a mutant mice have a significantly higher susceptibility than their wild-type mice, none of which displayed seizures in our study, even at the cutoff temperature.
In order to investigate the impact of the CRISPR-ON treatment on seizure severity, we tested the mice by placing them in a chamber set at 43°C constant. None of the wildtype mice showed behavioral seizures during the 30 min period recorded (n = 3). In contrast, all Scn1a RX/+ mice displayed a sequence of myoclonic, clonic (jerks and wild jumps) and generalized tonic-clonic seizure (GTCS). Although the latency to the first stage of this sequence (myoclonic seizure) was not significantly longer in CRISPR-ON treated Scn1a RX/+ mice, there was a significant improvement in the latency to clonic seizures, wild-jumps and GTCS (Fig. 3d). The duration of GTCS was mildly, yet not significantly, shorter in CRISPR-ON treated Scn1a RX/+ mice than in the Scn1a RX/+ controls (Fig. 3e).
Taken together these results indicate that even though a full recovery was not reached, CRISPR-ON significantly ameliorated the epileptic phenotypes of the Scn1a-haplodeficient Dravet syndrome model mice.

CRISPR-ON treatment reduces spontaneous spike discharges in Scn1a RX/+ mice
Because we identified a significant improvement of the susceptibility to hyperthermia-induced seizures using our CRISPR-ON strategy, we investigated further the impact of the treatment on the spontaneous epileptic phenotype. As the CRISPR-ON treatment started at P30, investigating spontaneous seizures occurring at the peak period, between P18 and P30 as commonly observed in Scn1a RX/+ pups, were not an appropriate approach. We however recorded the electrocorticographic (ECoG) signal in freely moving animals for 3 consecutive days using 24-27 weeks-old CRISPR-ON treated Scn1a RX/+ , non-treated Scn1a RX/ + and wild-type mice. Although spontaneous seizures are rare in adult Scn1a RX/+ mice, spontaneous generalized seizures were seen in one non-treated Scn1a RX/+ mouse, with 4 seizures occurring within a nearly 30 min window. The frequency of these events being low, they were not the most appropriate parameter to investigate the severity or duration of seizures in our CRISPR-ON treatment experiments. However, in addition to occasional spontaneous generalized seizures, abnormal electrophysiological patterns have been reported in Scn1a RX/+ mice with notably high-amplitude epileptiform discharge spikes (Richards et al., 2018). ECoG recordings revealed similar single spike discharges in our Scn1a RX/+ mice (Supplemental Fig. S12a) and we counted these abnormal spike discharges during 12-h periods. We did not observe any of these in wild-type recordings. Whereas all recordings from Scn1a RX/+ mice had abnormal spikes, the CRISPR-ON treatment led to a significant decrease in the frequency of these spike discharges (Supplemental Fig.  S12b). Interestingly, the animal that suffered generalized seizures displayed an extremely high number of these discharges (Supplemental Fig. S12b).

CRISPR-ON increases Scn1a expression predominantly in the PV+ inhibitory neurons in the Scn1a-haplodeficient mouse brain
To confirm and evaluate the upregulation of Scn1a expression in the brains of CRISPR-ON-treated Scn1a RX/+ mice, three mice were randomly-selected from each mouse line that survived until the FS test at 12w and wild-type mice that were subjected to FS test at 6w were sacrificed at 14w, and their Scn1a mRNA and Nav1.1 protein levels in membrane fraction were measured by using qRT-PCR (Fig. 4a) and western blot analysis (Fig. 4b and c), respectively. The amounts of Scn1a mRNA and the membrane-embedded Nav1.1 protein in the Scn1a RX/+ mice were almost half of those in the wild-type mice, as expected. Notably, the amount of Scn1a mRNA in the CRISPR-ON- Fig. 3. Improvements of the febrile seizures in the Scn1a-haplodeficient mice by the inhibitory neurons-specific CRISPR-ON. (a) Timeline for the febrile seizures test (FS test) on CRISPR-ON-treated Scn1a RX/+ Dravet syndrome model mice. (b) FS tests at postnatal week 6 (6w). Statistically significant difference was not detected among the 3 genotypes of mice with Scn1a RX/+ allele, though the threshold of febrile seizures (rectal temperature) of the floxed-dCas9-VPR VPR/+ /Scn1a RX/+ /Vgat-Cre Cre/+ triple mutant with AAV-gRNA injection (CRISPR-ON treated Scn1a RX/+ ) tends to be higher. One-way ANOVA and multiple comparison test using the Tukey-Kramer method. N.D.: seizure not detected. (c) Consecutive FS tests at 6w, 8w, 10w, and 12w. The mean threshold temperature is significantly higher in the CRISPR-ON-treated Scn1a RX/+ (male:5, female:6) compared to that of the untreated floxed-dCas9VPR VPR/+ /Scn1a RX/+ (male:3, female:4) or Scn1a RX/+ control (male:5, female:2) mice. The mice indicated by the numbers (#) were used for subsequent studies (Figs. 4a-c). Horizontal red bars indicate the average of temperatures (B and C). -: without injection. Tukey-Kramer method after two-way repeated measures ANOVA testing "time" and "experimental groups" as main effects. Wild-type controls (with and without VPR and Cre), displayed in panels b and c were not included in the statistical tests as no seizure was observed in these groups and no "temperature at generalized seizure onset" could be given to these samples. (d) Latency of hyperthermia-induced seizures. CRISPR-ON treatment extended the latency to develop clonic seizure (CS) (p = .044), wild jumps (p = .035), and generalized tonic-clonic seizure (GTCS) (p = .008) in Scn1a RX/+ mice placed at a constant temperature of 43°C. The latency to myoclonic seizures (MS) was not significantly different between the treated and non-treated groups. (e) Duration of GTCS. The duration of GTCS was 41% shorter, yet not significantly, in CRISPR-ON treated than non-treated Scn1a RX/+ mice (p = .09). Statistical significance was assessed using t-test. * p < .05, ** p < .01, *** p < .001. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.) T. Yamagata, et al. Neurobiology of Disease 141 (2020) 104954 treated Scn1a RX/+ mice was significantly larger than that in the Scn1a RX/+ mice (Fig. 4a) and the CRISPR-ON system did not lead to a non-specific significant change in expression for other sodium channel subunits (Supplemental Fig. S13). However, the membrane-embedded Nav1.1 protein amount in the CRISPR-ON-treated Scn1a RX/+ mice remained comparable to that in the Scn1a RX/+ mice ( Fig. 4b and c). We then analyzed the amount of Nav1.1 in total protein lysate from the brain of CRISPR-ON treated mice, and identified a more robust increase than in the membrane enriched fraction, even though the extent of this increase did not reach what was seen at the mRNA level ( Fig. 4d and e). The present study did not quantify the amount of Nav1.1 in individual neurons, however based on the intensity of the signal detected in the cell body in our immunohistological staining (see below), it is inferred that the amount of Nav1.1 is increased several-fold only in individual inhibitory neurons but not in excitatory ones. We investigated immunohistochemically the brain of the CRISPR-ON-treated Scn1a RX/+ mice at 14w without the FS tests ( Fig. 5a and b; Supplemental Fig. S14 and S15). Similar to the results in the wild-type mice (Ogiwara et al., 2007), the Nav1.1 immunosignals were observed at the thin neurites but not at the somata of the neurons in the Scn1a RX/ + mice ( Fig. 5a; Supplemental Fig. S14). In contrast, the dense Nav1.1 immunosignals were observed at the somata of the neurons in the CRISPR-ON-treated Scn1a RX/+ mice. The dense Nav1.1 immunosignals were mainly observed in the PV+ neurons in the neocortex, hippocampus, reticular thalamic nucleus (RTN), and cerebellum of the CRISPR-ON-treated Scn1a RX/+ mice ( Fig. 5b; Supplemental Fig. S15). In neocortex, 88% (72/82 cells) of somata with dense Nav1.1 signals were PV-positive, and 79% (72/91 cells) of somata of the PV+ neurons were positive for the dense Nav1.1 signals. In the hippocampus and RTN, most of the Nav1.1-overexpressed somata were also PV-positive.
In the cerebellum, the dense Nav1.1 signals were detected in the Purkinje cell somata but not in the stellate cell. Although the signals of Fig. 4. Upregulation of the Scn1a mRNA and Nav1.1 in the brain of CRISPR-ON-treated Scn1a RX/+ mice. (a) qRT-PCR analysis of the Scn1a mRNA. Significant increase was observed in the brain of the CRISPR-ON-treated Scn1a RX/+ mice (n = 3 mice for each genotype, RNA extracted from one complete hemisphere of 14weeks-old mice). (b and c) Western blot analyses of the membrane-embedded Nav1.1 from 4 mice of different genotypes (n = 3 mice for each genotype, proteins extracted from one complete hemisphere of 14-weeks-old mice). (d and e) Western blot analysis of Nav1.1 in total protein lysate extracted from one whole brain hemisphere of 14-weeks-old mice. The amount of Nav1.1 in the CRISPR-ON-treated Scn1a RX/+ mouse brain was significantly increased compared to Scn1a RX/+ control mice. TUBB: beta-tubulin. The mean of the Scn1a mRNA or Nav1.1 amount in Scn1a RX/+ mice was assigned as a value of 1 (a, c, e). Values are expressed as individual values (a, c) or mean ± standard error of the mean. Statistical significance was assessed using one-way ANOVA followed by Tukey's post-hoc multiple comparison test. ** p < .01, *** p < .001.
T. Yamagata, et al. Neurobiology of Disease 141 (2020) 104954 dCas9-VPR were abundantly observed in the GABAergic neurons of some regions such as the olfactory bulb and striatum (Supplemental Fig.  S8 b and g), the Nav1.1 signals were observed only in a few of those cells' somata (Supplemental Fig. S14). CRISPR-ON significantly enhanced Scn1a expression at the mRNA level in the olfactory bulb and the striatum, as well as in the neocortex (Supplemental Fig. S16). Immunohistochemistry analyses revealed a clear increased signal for Nav1.1 in the neocortex of CRISPR-ON treated mice compared to Scn1a RX/+ , while the signal intensity in hippocampus, cerebellum and olfactory bulb seemed mildly stronger (Supplemental Fig. S14). Nav1.1 signal was however below the detection level in the striatum of both treated and non-treated Scn1a haploinsufficient mice. Taken together, these results indicate that a significant increase in Scn1a expression at the mRNA level was detectable both at the scale of the whole brain or when looking at specific brain regions. In the whole brain, the Nav1.1 protein increase as a result of the CRISPR-ON treatment, though significant, did not reach the same extent as the mRNA. The limited increase of Nav1.1 in extracts enriched in membrane bound protein, together with the strong signal of Nav1.1 observed in neuronal somata suggest that a part of the protein produced does not reach the membrane surface and may accumulate in the cytoplasm. The CRISPR-ON-dependent increase in Nav1.1 expression was mostly observed in the PV+ GABAergic neurons where the Nav1.1 is originally expressed in the wild-type mice (Ogiwara et al., 2007).

CRISPR-ON treatment in Scn1a RX/+ mice partially rescues behavioral phenotypes
In our previous work, we identified a subset of behavioral impairments in the Scn1a RX/+ mice, notably in the open field, elevated plus maze and 3-chambers tasks (Ito et al., 2013). In order to assess the effects of the CRISPR-ON treatment on behavior phenotypes, we injected floxed-dCas9-VPR VPR/+ /Vgat-Cre Cre/+ /Scn1a RX/+ triple mutant mice with the AAV-gRNA at postnatal day 30. These CRISPR-ON treated mice were submitted to a behavior testing battery starting 8 weeks after AAV injection, together with wild-type and Scn1a RX/+ controls. This behavior screening included mice that survived to P30, and although it could be argued that mice that survive could intrinsically different from the ones dying due to spontaneous seizures, as this would affect both CRISPR-ON treated and non-treated Scn1a RX/+ mice, it is unlikely to produce a significant bias. In the open field task, Scn1a RX/+ mice traveled significantly longer distances and spent significantly less time in the center of the field than their wild-type mice (Fig. 6a). CRISPR-ON treated Scn1a RX/+ mice showed an intermediate state between their wild-type and Scn1a RX/+ mice: traveled distance was significantly longer than wild-type and tended to be shorter than in the Scn1a RX/+ group, and the time spent in the center was significantly shorter than in wild-type yet significantly longer than Scn1a RX/+ animals. In the elevated plus-maze, Scn1a RX/+ mice visited the open arms significantly more frequently and tended to stay for a longer time than the wild-type mice, and although wild-type mice showed a significant preference for the closed arms over the open ones, this was not the case for Scn1a RX/+ animals (Fig. 6b). In the CRISPR-ON treated Scn1a RX/+ mice we observed a somewhat intermediate state between the patterns seen in wild-type and Scn1a RX/+ mice. The average number of entries and time spent in the open arms were in-between that of Scn1a RX/+ and wild-type animals, not differing significantly from either groups. Remarkably the significant preference for the closed arms over the open ones commonly seen in wild-type mice and absent in Scn1a RX/+ animals was recovered in the CRISPR-ON treated Scn1a RX/+ group. Taken together, these results indicate that the increased exploratory activity and thigmotaxis in novel environments seen in Scn1a RX/+ mice are partially, yet significantly improved in CRISPR-ON treated Scn1a RX/+ samples and suggest that signs of decreased anxiety in the elevated plus maze could be ameliorated.
We then used the three chambers task to assess social behavior in CRISPR-ON treated Scn1a RX/+ mice. In the sociability phase of the test, wild-type mice spent a significantly larger amount of time investigating T. Yamagata, et al. Neurobiology of Disease 141 (2020) 104954 (caption on next page) T. Yamagata, et al. Neurobiology of Disease 141 (2020) 104954 the cage entrapping a stranger animal than the empty one (Fig. 6c). Scn1a RX/+ and CRISPR-ON treated Scn1a RX/+ mice showed a tendency for a similar pattern, even though the statistical significance was not reached (p = .099 and p = .057 respectively). Even though the time spent investigating the stranger mouse appeared shorter in the Scn1a RX/ + and CRISPR-ON treated Scn1a RX/+ groups compared to wild-type mice, this difference did not reach the significance level. In the preference for social novelty phase of the task, wild-type control animals tended to prefer the stranger mouse over the familiar one, even though the significance level was not reached, whereas Scn1a RX/+ mice did not seem to discriminate the stranger mouse from the familiar one (Fig. 6d).
The preference for social novelty did not reach the significance level in the CRISPR-ON treated Scn1a RX/+ group. Further work with larger cohorts and/or systems with a higher efficiency may be required to investigate possible improvements in social behavior of these mice.

Discussion
In this study, we showed that the inhibitory neurons-specific upregulation of Scn1a expression by CRISPR-ON effectively ameliorated the febrile seizures, sudden death and behavioral deficits of Scn1a-haplodeficient mice.
Our previous studies showed that even though Nav1.1 is expressed in PV+ as well as somatostatin-positive (SST+) interneurons, the Nav1.1 haploinsufficiency in PV+ inhibitory neurons is the major basis for the epileptic seizures and sudden death in mice with Scn1a-haplodeficiency while SST+ may have at most a marginal impact (Ogiwara et al., 2007;Ogiwara et al., 2013;Tatsukawa et al., 2018). Moreover, while the Nav1.1 haploinsufficiency in inhibitory neurons results in a more severe phenotype than observed in standard Scn1a RX/+ mice, the additional haploinsufficiency in excitatory neurons exerts an ameliorating effect, suggesting that this would also occur in the case of patients with Dravet syndrome . These observations suggest that the therapeutic approaches, such as the introduction of the intact SCN1A cDNA or mRNA into patients, should specifically target the GABAergic cells, especially the PV+ cells, rather than targeting the whole body or all neurons, in order to maximize the therapeutic effects and avoid toxic side-effects. However, this idea is still not widely accepted in the research community. For example, hyperexcitability was reported in excitatory neurons derived from iPS cells of Dravet syndrome patients suggesting that the Nav1.1 haploinsufficiency in the excitatory neurons cause their hyper-excitabilities and results in the epileptic phenotypes Jiao et al., 2013). Other studies also using neuronal differentiation of iPS cells derived from patients of Dravet syndrome have shown that GABAergic neurons had deficient action potential generation and excitability whereas excitatory neurons were not significantly affected (Higurashi et al., 2013;Sun et al., 2016). Although iPS cells-based studies might not be fully mature yet, our previous work supported the latter hypothesis as Nav1.1 haplodeficiency in the inhibitory but not in the excitatory neurons reproduced the epileptic seizures in mice . Our present study further supports our hypothesis that the Nav1.1 haploinsufficiency in the GABAergic neurons is a primary basis for Dravet syndrome.
Significant improvements to electrophysiological deficits in cultured Scn1a RX/+ cortical neurons following an activation of transcription of Scn1a have recently been reported (Colasante et al., 2020). Surprisingly, they identified only one single gRNA targeting the downstream promoter to lead to efficient transcription increase. This was however not the case in our study. In contrast we found several gRNAs in the upstream promoting region to increase significantly the expression of Scn1a and discovered that a mix of these gRNAs offered a multiplicative efficiency in cultured cells. These differences may be due to the different cell types, gRNAs sequences or activator used in the two studies. Interestingly however, the work by Colasante and collaborators also shows a milder expression increase in-vivo at the mRNA level than invitro and also reveals only a mild improvement in the sensitivity to hyperthermia-induced seizures. Our present study, using a slightly different approach thus complements these observations and offers an in depth investigation of various phenotypes characterizing Scn1a RX/+ mice, treating mice after the juvenile stage and focusing on phenotypes observable in adult mice.
Scn1a RX/+ mice show recurrent spontaneous generalized seizures after P18 (Ogiwara et al., 2007), but the frequency of these spontaneous seizures becomes very low in the adult mice (Yu et al., 2006;Ito et al., 2013). In addition, the high lethality observed from P14 to nearly P30 is likely creating a bias as the most severely affected animals dying and the least severely affected ones surviving to adulthood. For these reasons, it is particularly difficult to use spontaneous generalized seizures as a marker to investigate the efficiency of a treatment in adult Scn1a RX/ + mice. We however observed a significant improvement in the susceptibility to hyperthermia-induced seizures, with CRISPR-ON treated mice having a higher temperature threshold to generalized seizures as well as a longer latency to seizure onset and a tendency for shorter seizure duration when placed in a constant elevated temperature environment than Scn1a RX/+ mice. ECoG however proved useful as abnormal single spike discharges can be observed consistently in all of the Scn1a RX/+ mice [37, present study]. Of note, we observed generalized seizures in one Scn1a RX/+ animal and recordings in this sample showed a very large number of single spike discharges, whereas no wild-type animals displayed any of these. This supports the idea that these discharges have a direct connection with the apparition of seizures in Scn1a RX/+ mice. The significant decrease in these abnormal spikes Fig. 6. Inhibitory neurons-specific CRISPR-ON treatment partially recovers behavioral phenotypes in Scn1a-haplodeficient mice. (a) In the open field task, tracking images revealed a strong thigmotaxis in Scn1a RX/+ mice compared to wild-type (WT) mice whereas CRISPR-ON treated Scn1a RX/+ animals seemed intermediate. The traveled distance was significantly longer in the Scn1a RX/+ group whereas CRISPR-ON treated Scn1a RX/+ mice traveled significantly longer distances than WT mice but tended to travel shorter distances than Scn1a RX/+ mice (genotype-time interaction: p = .279; time: p = 8.0E-3; genotype: p = 6.6E-5). Scn1a RX/+ mice spent significantly less time in the center than WT mice whereas CRISPR-ON treated Scn1a RX/+ mice showed an intermediate state significantly differing from both groups (genotype-time interaction: p = .856; time: p = .026; genotype: p = 1.3E-9). (b) In the elevated-plus maze, tracking traces of wild-type mice appeared stronger in the closed arms (CA) than the open arms (OA), whereas traces were comparable in all arms for the Scn1a RX/+ group and an intermediate state was seen in CRISPR-ON treated Scn1a RX/+ samples. Scn1a RX/+ mice visited the open arms more often than their WT mice, whereas CRISPR-ON treated Scn1a RX/+ mice appeared intermediate, not differing significantly from either group (one-way ANOVA p = .040, post-hoc Scn1a RX/+ vs WT p = .032; CRISPR-ON treated Scn1a RX/+ vs WT p = .296; CRISPR-ON treated Scn1a RX/+ vs Scn1a RX/+ p = .497). WT mice spent significantly more time in the closed arms than the open ones (p = 9.6E-4) but this was not the case in the Scn1a RX/+ group (p = .196). This significant preference for the closed arms was recovered in CRISPR-ON treated Scn1a RX/+ mice (p = 1.5E-5). Although the time spent in the open arms appeared mildly longer in the Scn1a RX/+ group, this effect did not reach the significance level (p = .508). OA: open arm, Ce: center, CA: closed arm (c) In the sociability phase of the 3-chambers task, WT mice spent significantly more time investigating the stranger's cage (Str) than the empty one (Emp)(p = 2.8E3). Scn1a RX/+ and CRISPR-ON treated tended to show a similar pattern, but the statistical significance was not reached (p = .099 and p = .057 respectively). Although the time spent investigating the stranger mouse seemed shorter in the Scn1a RX/+ and CRISPR-ON treated groups, this did not reach the significance level (p = .392). (d) In the preference for social novelty phase of the task, WT and CRISPR-ON treated Scn1a RX/+ mice spent a minor yet not significant larger amount of time investigating the stranger mouse (Str) than the familiar one (Fam), whereas Scn1a RX/+ mice did not seem to discriminate the two (p = .177, p = .175 and p = .836). Values are expressed as mean ± standard error of the mean. Statistical significance was assessed using two-way ANOVA followed by oneway ANOVA post-hoc tests (a) or one-way ANOVA followed by Tukey method (b-d) with significance set at * p < .05, ** p < .01 and *** p < .001.
suggests that the increase in Nav1.1 expression in inhibitory neurons induced by CRISPR-ON helps suppress abnormal brain activity and supports the improvements seen in the susceptibility to hyperthermiainduced seizures.
Toxicity of the dCas9-VPR has previously been reported in drosophila (Ewen-Campen et al., 2017). Here, we showed that the mice with the constitutional expression of dCas9-VPR by the EIIa-Cre driver are perinatally lethal, indicating that the dCas9-VPR is also toxic in mice. Although the floxed-dCas9-VPR VPR/+ /Vgat-Cre Cre/+ mice were born normally and were fertile, the survivability of the floxed-dCas9-VPR VPR/+ /Vgat-Cre Cre/+ /Scn1a RX/+ triple mutant mice was much lower than that of the floxed-dCas9-VPR VPR/+ /Vgat-Cre Cre/+ mice or that of Scn1a RX/+ mice. These observations indicate that even if the dCas9-VPR expression is restricted to the inhibitory neurons, it is still toxic and the toxicity becomes obvious when it is expressed in the Scn1a RX/+ mice, which have functional impairments in the inhibitory neurons (Ogiwara et al., 2007). The constitutive Cas9-expressing mice have been reported to be fertile, had normal litter sizes, presented no morphological abnormalities, and were able to breed to homozygosity (Platt et al., 2014). In addition, no side effect has been reported in the dCas9-VP64-expressing mice (Matharu et al., 2019), and therefore the toxicity of the dCas9-VPR may derive from p65 or Rta in the VPR element but not from the dCas9 and VP64. Though previous studies suggested the toxicity of the dCas9-VPR could be alleviated by decreasing the expression level of the activator in drosophila (Ewen-Campen et al., 2017), it remains important to provide a system with a high efficiency. Future studies will be important to generate a system with an enhanced efficiency while containing eventual toxic effects, either by modifying the activator regions or by using different versions of the Cas9 protein.
Even though the dCas9-VPR is toxic in mice, after the AAV-gRNA injection at 4w, none of the floxed-dCas9-VPR VPR/+ /Vgat-Cre Cre/ + /Scn1a RX/+ triple mutant mice (n = 11) died until the FS test at 12w, while some of the floxed-dCas9-VPR VPR/+ /Scn1a RX/+ or Scn1a RX/+ mice continued to die during that period. Together with the significant improvements of febrile seizures in the CRISPR-ON-treated Scn1a RX/+ mice, these observations indicate that the therapeutic effect of CRISPR-ON overcame the toxic effect of the dCas9-VPR in the Scn1a RX/+ mice and guarantee the benefit of CRISPR-ON as a therapeutic approach for Dravet syndrome.
The size of the dCas9-VPR used in our study being close to the size of Nav1.1, they share the same limitation in terms of packaging in AAV particles. Recent studies have however brought new versions of Cas9 proteins that are very promising for the fact that they can be packaged directly in standard AAV particles such as Nme-Cas9 family (Edraki et al., 2019) and SaCas9 (Ma et al., 2018). These new shorter Cas9 proteins will prove helpful in developing new targeting vectors without requiring the expression of a transgene. Coupling with inhibitory neuron specific promoters such as Dlx5/6 would provide the cell-type specificity as recently reported (Colasante et al., 2020). We believe further work regarding the activator sequence "VPR" or searching for shorter and more efficient activators will also prove useful to alleviate side-effects and improve the overall efficiency of the system, with the ultimate goal to produce an inhibitory specific, high-efficiency dCas9 small enough to be packages in AAV particles while guide RNAs could be provided either within the same viral construct or on an independent one.
In addition to severe epilepsies, patients of Dravet syndrome develop cognitive deficits, ataxia, hyperactivity and autistic features. In previous studies, we have identified patterns mimicking these phenotypes in Scn1a RX/+ mice, with notable abnormalities in the spontaneous exploratory behavior and deficits in social related tasks (Ito et al., 2013). In a recent work, we observed similar, yet milder phenotypes in PV-Cre driven conditional knockout mice, suggesting that haploinsufficiency in inhibitory cells could be driving these behavioral abnormalities (Tatsukawa et al., 2018). In the present work, enhancing the expression of Scn1a using a CRISPR-ON strategy specifically targeting cells expressing VGAT allowed a partial, yet significant improvement of animals behavior. The strong hyperactivity and thigmotaxis observed in the open field, as well as the characteristic signs of decreased anxiety in the elevated plus maze observed in Scn1a RX/+ mice was improved to a state intermediate between Scn1a RX/+ and wild-type animals. Moreover, although the use of a complex triple mutant design limited the number of samples available, signs of improvement were seen in the 3-chambers social task, with a likely recovery of the preference for social novelty that is usually completely lost in Scn1a RX/+ mice. CRISPR-ON treatment is thus a promising strategy to improve phenotypes consequent a haploinsufficiency of Scn1a, not only for febrile seizures but also on a behavioral standpoint. Further work aiming at increasing the efficiency of the treatment will prove very useful to hopefully bring abilities of Scn1a RX/+ mice back to a level comparable to wild-type mice. This is especially important given the fact that the CRISPR-ON technique used in the present work is rather invasive and should thus be more effective and/or provide fewer damaging side-effects than existing treatments to represent a sufficient value as a new therapeutic approach. Even though the present work shows that targeting only inhibitory neurons may be sufficient to improve some of the phenotypes, the CRISPR-ON treatment does not have to be restricted to inhibitory cells, but special care should however be taken to prevent a massive expression of Nav1.1 that would go far beyond the normal level as several reports revealed deleterious gain of function of Nav1.1 in particular in cases of hemiplegic migraine (Cestèle et al., 2013;Dhifallah et al., 2018).
Our in-vivo study revealed that the CRISPR-ON treatment led to a significant increase of Scn1a mRNA expression to a level close to wildtype controls. In contrast the extent of Nav1.1 increase in whole brain total protein extracts, though significant, was comparatively milder. Our approach, using triple mutant mice expressing dCas9-VPR under the control of a Vgat-Cre driver means that when gRNAs are provided using an AAV virus, the resulting CRISPR-ON transcription activation will affect all VGAT expressing inhibitory neurons. Our previous work however demonstrated that Nav1.1 expression is restricted to PV+ and SST+ neurons, derived from the medial ganglionic eminence, but not in vasoactive intestinal peptide or reelin positive neurons derived from the caudal ganglionic eminence (Yamagata et al., 2017). The CRISPR-ON treatment may thus activate the transcription of Scn1a in all inhibitory neurons whereas the translation may be restricted to neurons that originally express Nav1.1, i.e. PV+ or SST+ neurons. The fact that the strong Nav1.1 signal is only observed in a few cells in comparison to the number of cells expressing the dCas9 protein supports this hypothesis, even though we do not exclude that the CRISPR-Cas9 system itself may also interfere with the translation even in cells that are meant to express Nav1.1. In addition, we observed a strong Nav1.1 signal in the somata of PV+ neurons mostly, though Nav1.1 protein is usually localized in the axon initial segment of these cells (Ogiwara et al., 2007;Yamagata et al., 2017). There may thus be an accumulation of Nav1.1 in the cytoplasm which could be inactive, even though more work is required to determine the dynamics of the transport to the membrane from this pool of cytoplasmic Nav1.1. Assuming that part of the Nav1.1 produced thanks to the CRISPR-ON treatment accumulates in the cytoplasm, it is not excluded that it gets degraded in a feedback or control system to prevent an excess of protein. This would also contribute to the fact that the increase by the CRISPR-ON treatment at the protein level is not as high as it is at the mRNA level. These limitations may explain why the epileptic and behavioral phenotypes are only partially improved in CRISPR-ON treated mice. We believe that further work on improving the efficiency of Nav1.1 translation and more importantly to enhance its transport to the membrane constitute the main axis of improvement for the CRISPR-ON therapeutic approach.
In our study, CRISPR-ON was effective when introduced into the Scn1a RX/+ mice at postnatal 4 weeks (4w), which is after the onset of the epileptic seizures in most of the mice (Ogiwara et al., 2007). This indicates that the upregulation of Nav1.1 in the Scn1a RX/+ mice can treat their epilepsy and recover their behavioral abnormalities even after the juvenile stage. Recently, the rescue of obesity in the Sim1-and Mc4r-haplodeficient mice by CRISPR-ON has been reported (Matharu et al., 2019). Although the usage of transgenic introductions of Cre and dCas9-VPR genes and rather modest improvements in seizures and abnormal behaviors hamper direct clinical application, these results indicate that CRISPR-ON has a promising potential as a therapy for diseases with haplodeficiencies of disease-responsible genes. Future work aiming at improving efficiency and transgenic system-free delivery methods will provide critical advance toward a use in patients.