Gabra2 is a genetic modifier of Scn8a encephalopathy in the mouse*

2.1. Animals

The conditional Scn8a^R1872W allele contains two copies of the last coding exon of Scn8a, a floxed wild‐type exon located upstream of an exon with the R1872W mutation. ¹⁰ The action of Cre recombinase deletes the wild‐type exon from the conditional allele and activates the R1872W mutant. The conditional allele was generated by transcription activatorlike effector nuclease knockin to a (C57BL/6JX SJL/J)F2 zygote. ¹⁰ The conditional allele was inserted into the C57BL/6J‐derived Scn8a locus on chromosome 15; the conditional line was bred from a single founder individual and maintained by >10 generations of backcrossing to strain C57BL/6J. The mutant line C3HeB/FeJ Scn8a^N1768D expresses the original SCN8A mutant, N1768D, ⁸ and was generated by >10 generations of backcrossing to strain C3HeB/FeJ. ²⁴ C57BL/6J Emx1^Cre/Cre mice with a Cre transgene inserted into the endogenous Emx1 locus (Jackson Laboratory catalog # 005628), and inbred strains C3HeB/FeJ, FVB/NJ, and DBA/2J were obtained from the Jackson Laboratory. Emx1^Cre mice express Cre recombinase in forebrain excitatory neurons beginning at E9.5. ²⁵ , ²⁶ Mice carrying both the conditional Scn8a^R1872W mutation and the Emx1^Cre are designated conditional knockin (cKI) mice. All animal experiments were approved by the University of Michigan and the Unit for Laboratory Animal Medicine at the University of Michigan in accordance with the National Institutes of Health Guide for the Care and Use of Animals (protocol # PRO00007986). Principles outlined in the ARRIVE guidelines and the Basel declaration https://www.basel‐declaration.org/ including the 3R concept were considered when planning the experiments.

2.2. Genotyping

The conditional allele of Scn8a was genotyped by polymerase chain reaction (PCR) using forward primer 5′‐GCACG TGCTG AAAAA GTGG‐3′ and reverse primer 5′‐CCTCC TCTTA CCGTG CAGAC‐3′, which yield a 414‐bp product from the wild‐type allele and a 448‐bp product from the conditional allele. Digestion of the PCR products with KpnI generates an intact 414‐bp fragment from the wild‐type allele and fragments of 262 bp and 186 bp from the conditional allele. The presence of EmxI^Cre was detected by PCR amplification of a 102‐bp product using forward primer 5′‐GCGGT CTGGC AGTAA AAACT ATC‐3′ and reverse primer 5′‐GTGAA ACAGC ATTGC TGTCA C T‐3′.

2.3. Targeted deep sequencing of the Scn8a transcript

Total RNA was prepared from whole brain by TRIzol extraction using a Direct‐zol RNA Miniprep Plus Kit (Zymogen R2071). cDNA was synthesized with the LunaScript RT SuperMix Kit (New England BioLabs). A 1011‐bp fragment of the Scn8a transcript containing the R1872W mutation was amplified from cDNA using forward primer 5′‐GGTCA TTCTC TCCAT TGTGG‐3′ and reverse primer 5′‐CCTCC ATTCT CCAGC TTGTT‐3′. Targeted deep sequencing of the PCR product was performed with the Illumina MiSeq system. Raw data were aligned to the mm10 reference genome using HISAT2 ²⁷ and analyzed using SAMtools mpileup. ²⁸ The number of sequence reads varied from 1100 to 6749 per sample, with the exception of one sample with 339 reads.

2.4. Quantitative RT‐PCR of Gabra2

Hippocampus was dissected from adult mice and cDNA was prepared as described above. The abundance of the Gabra2 transcript was assessed using the TaqMan gene expression assay (Mm01277042_m1) and normalized to Tbp as internal control (Mm01277042_m1). TaqMan assays were run on a StepOnePlus cycler (Applied Biosystems) using StepOne software v2.3 (Life Technologies). Relative transcript level was calculated by the ΔΔC _t method ²⁹ and compared with C57BL/6J brain.

2.5. Quantitative trait locus analysis

Genomic DNA was extracted from tail biopsy using GenElute Mammalian Genomic DNA Miniprep Kit (Sigma). Single nucleotide polymorphism (SNP) genotyping with the miniMUGA array ³⁰ was carried out at NeogenCorp (Lincoln, NE). The MiniMUGA array is designed to discriminate between most commonly used mouse populations and contains probes for >10 000 SNPs. Of these, 2792 were found to robustly discriminate between strains C57BL/6J and SJL/J and were retained for analysis. To achieve approximate normality, a log transformation was applied to the age at death (survival) phenotype. Data analysis was carried out using R/qtl2. ³¹ Briefly, we fit a linear mixed model at each of the 2792 retained loci, regressing the (transformed) phenotype of interest on the allele probabilities. The models also included sex as a fixed effect, and a random effect to control for the overall genetic similarity between samples at analyzed loci. The models were fit using maximum likelihood. Analysis was carried out on 111 affected F2 mice, the 60 mice described in the experiments below plus 51 mice evaluated for length of survival only.

3.1. Backcrossing Scn8a^R1872W mice to strain C57BL/6J accelerates seizure onset

The conditional Scn8a^R1872W allele was generated on the (C57BL/6J × SJL/J)F2 strain background, and the F2 founder was backcrossed to strain C57BL/6J (B6). Mice from the N2 backcross generation were bred with B6.Emx1^Cre/Cre mice to generate N3 mice. In these mice, the R1872W mutant allele is activated in excitatory forebrain neurons beginning at E9.5. ²⁵ , ²⁶ The Scn8a^R1872W/+, Emx1^Cre/+ N3 mice exhibit generalized seizures beginning at 1 month of age, with median survival of 46 days (Figure 1A). ¹⁰ With increasing generations of backcrossing to strain C57BL/6J, seizure onset occurs earlier and life span decreases. For example, in the N8 backcross generation, the median survival of mutant mice was reduced to 22 days (Figure 1A). During backcrossing, the proportion of the genome derived from strain SJL/J is reduced from 50% in the F2 founder to an average of 0.2% in N8 mice, suggesting that loss of protective alleles from strain SJL/J might be responsible for the reduction in length of survival.

3.2. Prolonged survival of (C57BL/6J × SJL/J)F2 mutant mice

To detect the presence of protective alleles in strain SJL/J, we crossed wild‐type mice from this strain with N8 C57BL/6J Scn8a^R1872W mice to produce an F2 generation that was segregating alleles from both strains. The two‐generation breeding scheme is shown in Figure 1B. In the first cross, C57BL/6J congenic homozygous Scn8a^{R1872W/R1872W} mice from the N8 generation were bred with strain SJL/J to generate (B6 × SJL)F1 mice heterozygous for the conditional allele. In parallel, homozygous B6.Emx1^Cre/Cre mice were crossed with strain SJL/J to generate (B6 × SJL)F1 mice heterozygous for Emx1^Cre. The two types of F1 mice were crossed to produce the F2 generation. In agreement with Mendelian prediction, Scn8a^R1872W/+, Emx1^Cre/+ double heterozygotes accounted for 60 of 248 F2 offspring. The median life span of the (B6 × SJL)F2 mutants was increased to 66 days (Figure 1A). The F2 mice with genotype Scn8a^R1872W/+, Emx1^Cre/+ are designated cKI, and these mice are analyzed in the following experiments.

3.3. Seizure phenotypes in F2 cKI mice

The double heterozygous cKI F2 mice with genotype Scn8a^R8172W/+, EmxI^Cre/+ were monitored by observation for 8 hours per day, between 9 am and 5 pm. Three to 4 days prior to onset of convulsive seizures, the mice developed hindlimb clasping, myoclonic jerks, and freezing behavior. Convulsive seizures began with 10‐20 seconds of forelimb clonus with Straub tail followed by 5‐15 seconds of violent clonic movements that propelled the mice around the cage (see Video S1). One to five seizures were observed per 8 hours of monitoring. Seizures occurred daily for 1‐3 days after onset, followed by a seizure‐free period of 2‐14 days. A total of two to five cycles of seizures and seizure‐free periods were observed. Death occurred overnight in 59 of 60 mice. Mice were found in the morning with extended body and limbs indicative of death during a convulsive seizure. Two mice displayed kyphotic postures.

3.4. Age at seizure onset and length of survival in F2 cKI mice

N8 mice exhibit early seizure onset at 20‐30 days and short survival (Figure) 2A,B. In contrast, the majority of F2 mice exhibited onset at 40‐110 days of age (Figure 2C), with a broad distribution of life span from 20 days to >180 days (Figure 2D and Figure S1). At the extremes of the F2 distribution are five mice with survival of <30 days and five mice that are still surviving beyond 6 months of age.

There is a strong correlation between age at onset and length of survival in the F2 population (Figure 2E). There is little variation in duration of disease, the time between onset and death (Figure 3A). Duration was not correlated with onset, and the time for progression from seizure onset to death was similar in short‐lived and long‐lived mice (Figure 3B). The genetic variants segregating in this cross thus appear to primarily affect processes leading to development of the first seizure, rather than progression from first seizure to death.

Activation of the conditional allele requires Cre‐mediated deletion of exon 26a. The consistency of in vivo activation of the conditional allele was examined in F2 mice, with seizure onset varying between 20 and 180 days. An RT‐PCR fragment containing codon 1872 was amplified from whole brain RNA, and the number of mutant and wild‐type transcripts was determined by targeted deep sequencing. The proportion of transcripts containing the mutant sequence was 26 ± 2% and did not differ between mice surviving from 20 days to >100 days (Figure 2F). Long survival thus does not result from inefficient Cre‐mediated activation of the conditional allele.

3.5. Genetic mapping of loci affecting survival

Genetic mapping of the survival phenotype was performed on the 60 F2 mice with the cKI genotype Scn8a^R1872W/+, Emx1^Cre/+ described above and 51 mice with the same genotype from a second F2 cohort comprised of subsequent litters from the same cross. A total of 111 F2 mice are included in the quantitative trait locus (QTL) analysis. The mapping results are presented as a Manhattan plot in Figure 4. The significance thresholds were derived using the empirical distributions of the lod scores based on 10 000 permutations. The 0.05 and 0.1 quantiles of these distributions are shown as black and red dotted lines, respectively. One major locus with a significant lod score (P < .05) was mapped to chromosome 5 (Figure 4).

In addition to the chromosome 5 QTL, two loci on chromosome 1 were significant at the .1 level, and a marginal peak was observed on chromosome 19. To further investigate the lod peaks that did not quite achieve significance at the .05 level, we also fit the models controlling for the allele probabilities associated with the peak on chromosome 5 as fixed effects; our conclusions did not change in this sensitivity analysis.

3.6. Gabra2: A candidate gene in the chromosome 5 QTL

To define the maximum length of the QTL on chromosome 5, we examined the SNP haplotypes of the five mice with survival time of <30 days. These mice were homozygous for BB alleles in the interval that includes UNC9382069 at 67 Mb on chromosome 5 and S6R053314276 at 82 Mb. Within this interval, the Gabra2 gene is located at 71 Mb. C57BL/6J and closely related strains are known to carry a single base deletion at position −3 in the splice acceptor site of exon 5 in Gabra2. Homozygosity for this variant reduces the abundance of the Gabra2 transcript in hippocampus to 25% of wild type. ²³ To determine whether the splice site variant is segregating in our F2 cross, we amplified genomic DNA from C57BL/6J and SJL/J mice from our colony and sequenced the Gabra2 splice acceptor site. The thymine residue that is deleted in strain C57BL/6J is marked with an arrow above the SJL/J chromatograph in Figure 5A. The chromatograms identify the wild‐type allele in strain SJL/J and confirm the deletion in strain C57BL/6J.

To determine whether the deletion is associated with a difference in Gabra2 expression in the parental strains, we carried out RT‐PCR of hippocampal RNA. The level of Gabra2 transcript is 4.5‐fold higher in strain SJL/J than in C57BL/6J (Figure 5B). This is consistent with other strains carrying the wild‐type allele. ²³ , ³²

To evaluate the physiological effect of the Gabra2 variant, we examined the relationship between Gabra2 genotype and seizure phenotypes. The median survival of Gabra2^B/B mice was 53 days, significantly shorter than the 70 days and 79 days of BS and SS mice, respectively (P < .05; Figure 5C and Figure S2). The survival of BS and SS mice did not differ significantly (P = .67), demonstrating dominant expression of the wild‐type Gabra2 allele (Figure S2).

Seizure onset was earlier in Gabra2^B/Bhomozygotes (median = 27 days) compared with BS and SS mice (median = 45 and 48 days, respectively; P < .05; Figure 5D and Figure S2). The difference between BS and SS mice was not significant (P = .08). These findings indicate that the hypomorphic Gabra2 allele in strain C57BL/6J contributes to the more severe phenotype of earlier onset and shorter survival. The two Gabra2^B/B mice exceeded 50 days, suggesting that these individuals inherited other protective loci from strain SJL/J. (the previous sentence should begin: "The age at seizure onset for two...")

3.7. Additional protective modifiers in strain SJL/J

The congenic line C3HeB/FeJ Scn8a^{N1768D 24} carries the patient mutation N1768D, which causes a defect in channel inactivation similar to that of R1872W with a less severe patient phenotype. ¹¹ , ¹² Strain C3HeB/FeJ does not carry the Gabra2 splice site variant (Figure S3). The abundance of brain Gabra2 transcript in strain C3H is comparable to strain SJL/J (Figure 5B). C3HeB/FeJ Scn8a^N1768D/+ mice exhibit 50% lethality at 6 months of age ²⁴ (Figure 6A).

To confirm the presence of non‐Gabra2 modifier(s) in strain SJL, we tested the ability of SJL to rescue the N1768D Scn8a variant on the C3H strain background. In the (C3H × SJL)F1 cross, both parental strains contribute a wild‐type Gabra2 allele. Most of the (C3H × SJL) F1‐Scn8a^N1768D/+ mice survived beyond 6 months (Figure 6A). These data demonstrate that strain SJL/J carries one or more protective modifier(s) other than Gabra2, and that these modifier(s) protect the N1768D allele of Scn8a.

To determine whether the protective modifier(s) are present in other inbred strains, we tested strains FVB/NJ and DBA/2J. These strains did not rescue the mutant mice in the cross with C3HeB/FeJ Scn8a^N1768D/+ (Figure 6B,C). Thus, the protective non‐Gabra2 modifier(s) that are present in strain SJL are not present in strains FVB/NJ and DBA/2J. This information will be useful for future identification of the additional modifier(s) in strain SJL.