Wild-type FUS corrects ALS-like disease induced by cytoplasmic mutant FUS through autoregulation

Mutations in FUS, an RNA-binding protein involved in multiple steps of RNA metabolism, are associated with the most severe forms of amyotrophic lateral sclerosis (ALS). Accumulation of cytoplasmic FUS is likely to be a major culprit in the toxicity of FUS mutations. Thus, preventing cytoplasmic mislocalization of the FUS protein may represent a valuable therapeutic strategy. FUS binds to its own pre-mRNA creating an autoregulatory loop efficiently buffering FUS excess through multiple proposed mechanisms including retention of introns 6 and/or 7. Here, we introduced a wild-type FUS gene allele, retaining all intronic sequences, in mice whose heterozygous or homozygous expression of a cytoplasmically retained FUS protein (Fus∆NLS) was previously shown to provoke ALS-like disease or postnatal lethality, respectively. Wild-type FUS completely rescued the early lethality caused by the two Fus∆NLS alleles, and improved the age-dependent motor deficits and reduced lifespan caused by heterozygous expression of mutant FUS∆NLS. Mechanistically, wild-type FUS decreased the load of cytoplasmic FUS, increased retention of introns 6 and 7 in the endogenous mouse Fus mRNA, and decreased expression of the mutant mRNA. Thus, the wild-type FUS allele activates the homeostatic autoregulatory loop, maintaining constant FUS levels and decreasing the mutant protein in the cytoplasm. These results provide proof of concept that an autoregulatory competent wild-type FUS expression could protect against this devastating, currently intractable, neurodegenerative disease. Supplementary Information The online version contains supplementary material available at 10.1186/s13024-021-00477-w.


Background
Amyotrophic lateral sclerosis (ALS), the major adult onset motor neuron disease [1,2], is characterized by a progressive paralysis leading to death within a few years after onset. Mutations in FUS cause the most severe cases of ALS, with young onset and rapid disease progression [3,4]. FUS mutations are clustered in the Cterminal region of the protein, carrying a PY-nuclear localization sequence (NLS), responsible for its nuclear import. Truncating mutations have been described in ALS families, leading to complete loss of the PY-NLS, and cytoplasmic aggregation of FUS [5,6]. Studies in mouse models have demonstrated that cytoplasmic accumulation of FUS provokes motor neuron degeneration [7][8][9][10][11][12]. Indeed, heterozygous Fus knock-in mice with ALS-like truncating mutations develop mild, late onset muscle weakness and motor neuron degeneration, while haploinsufficient Fus knock-out mice do not show ALS related symptoms [10][11][12]. A successful therapeutic strategy for FUS-ALS may lie in reduction of the cytoplasmic FUS content, to avoid its toxic effects.
FUS levels are regulated by other RNA-binding proteins [13,14] and are tightly controlled by autoregulatory mechanisms [14][15][16]. Indeed, the addition of more than 20 copies of the complete human FUS gene to the mouse genome only slightly increases FUS protein levels, and does not lead to phenotypic consequences [8], showing the efficacy of this buffering system of FUS levels. Contrastingly, the saturation of FUS autoregulation, through overexpression of cDNA driven, autoregulatory incompetent, FUS expression, is highly toxic to neurons [9,17]. FUS autoregulation appears to involve at least three possible mechanisms, including exon skipping [15], intron retention [14] and microRNA [16], and recent evidence suggested that the major autoregulatory mechanism was retention of introns 6 and 7 [14]. Here, we tested the hypothesis that the expression of a wild-type FUS gene, carrying all regulatory elements necessary for autoregulation would engage autoregulation of the mutation carrying RNA, and subsequently decrease accumulation of FUS in the cytoplasm.

Results
Wild-type FUS transgene rescues lethality and motor defects in Fus ΔNLS mice Human wild-type FUS transgenic mice (hFUS mice) expressing human FUS gene including its own human FUS promoter obtained from a BAC [8] were crossed with Fus ΔNLS mice [11] in a two-round mating ( Fig. 1 A). As previously described, Fus ΔNLS/ΔNLS mice (in absence of hFUS) die within the first hours after birth [11] and no homozygous mutant Fus ΔNLS mice were obtained at 1 month of age in the absence of hFUS (Fig. 1B). Contrastingly, expression of hFUS transgene completely rescued lethality of homozygous Fus ΔNLS/ΔNLS mice until adulthood ( Fig. 1 C). However, rescued homozygous Fus ΔNLS/ΔNLS mice displayed higher lethality throughout adulthood than wild-type littermate animals ( Fig. 1 C). Increased adult lethality was also observed in Fus ΔNLS/+ mice, with about a 30 % of death rate before 600 days of age (p = 0.0398, log rank, Fus +/+ vs. Fus ΔNLS/+ ), consistent with findings reported in another heterozygous knock-in model [10]. Nonetheless, most Fus ΔNLS/+ /hFUS mice survived until this age, and their survival rate was indistinguishable from non-transgenic normal mice (p = 0.33 Fus +/+ vs. Fus ΔNLS/+ /hFUS) or from single hFUS transgenic mice (Fig. 1D). The mild, late onset, muscle weakness observed in Fus ΔNLS/+ mice using inverted grid test [12], was rescued in Fus ΔNLS/+ /hFUS and in Fus ΔNLS/ΔNLS /hFUS mice ( Fig. 1E and Fig. S1A). Furthermore, hindlimb grip strength deficits associated with expression of Fus ΔNLS/+ were mildly and transiently improved in Fus ΔNLS/+ /hFUS females ( Fig. 1 F) but not in males ( Fig. 1 F). Indeed, in this test, the performance of hFUS transgenic mice decreased significantly in males after 10 months of age, thus confounding a potential protection ( Fig. 1 F and Fig. S1A). These protective effects were not caused or modified by changes in body weight as there were no significant changes in body weight across genotypes before 200 days of age. After this age, only Fus ΔNLS/ΔNLS /hFUS mice showed a mildly decreased body weight as compared to the wild-type and Fus ΔNLS/+ mice (Fig. S1B). Thus, wild-type human FUS significantly rescued lethality and, at least partially, motor deficits associated with cytoplasmically retained mutant FUS ΔNLS protein.

Wild-type FUS transgene decreases cytoplasmic accumulation of FUS in Fus ΔNLS mice
We then asked whether hFUS transgene altered levels of FUS in Fus ΔNLS mice. Consistent with previous results [18], total FUS levels increased in Fus ΔNLS/+ mouse brains as compared to Fus +/+ mice, and in Fus ΔNLS/+ /hFUS as compared to single hFUS transgenic mice ( Fig. 2 A-B). Consistent with previous results, the hFUS transgene on its own did not further increase total FUS proteins in wild type or Fus ΔNLS/+ mice. The increase observed in Fus ΔNLS/+ mice was not detected when an antibody targeting the NLS sequence (C-term FUS), absent from the FUS ΔNLS protein, was used, but was even more evident using an antibody targeting selectively mouse FUS ( Fig. 2 A-B and Source data for uncropped western blots). This increase in mouse FUS was normalized by the hFUS transgene. Human FUS levels remained unchanged across the three genotypes carrying hFUS.
We then asked whether increased FUS cytoplasmic levels were also rescued by the hFUS transgene and performed subcellular fractionation to obtain nuclear and   Fig. 3 A), and this was also observed in motor neurons using an antibody detecting total FUS using double FUS/ChAT immunofluorescence (Fig. 3B). No cytoplasmic staining was observed when using a C-terminal antibody ( Fig. 3 C), further confirming that the cytoplasmic staining is derived from mutant FUS protein. Indeed, specific immunolabelling of mouse FUS showed decreased overall signal in mice with hFUS transgene, and loss of cytoplasmic staining in Fus ΔNLS/+ /hFUS motor neurons (Fig. 3D). Interestingly, we observed significant nuclear staining for mouse FUS in Fus ΔNLS/ΔNLS /hFUS motor neurons despite the lack of NLS in mouse FUS in this genotype (Fig. 3D). Accumulation of cytoplasmic asymmetrically dimethylated (ADMA) FUS is a feature of FUS-ALS [5,6,19] patients which was recapitulated in the Fus ΔNLS/+ mice, as we previously reported [12]. Here, this significant increase in ADMA-FUS detected in Fus ΔNLS/+ cytoplasmic fractions, was largely prevented by the hFUS transgene in Fus ΔNLS/+ /hFUS mice ( Fig. 3 Fig. S2). We further investigated the three possible autoregulatory mechanisms that have been documented for FUS (Fig. S3). First, FUS protein is proposed to bind to its own pre-mRNA, leading to the splicing of exon 7, and the possible subsequent degradation of the abnormally Δexon 7 FUS mRNA through nonsense-mediated mRNA decay [15,20]. Interestingly, expression of hFUS transgene increased levels of the aberrantly spliced Fus Δexon 7 mRNA (Fig. 5 A and Fig. S4A). Secondly, increased FUS levels have recently been reported to lead to the retention of introns 6 and 7 in the mature mRNA, (See figure on previous page.) Fig. 1 hFUS transgene rescues lethality and motor deficits in Fus ΔNLS mice. A: Scheme of the breeding strategy. B: Representative genotyping results of 5 mice at 1 month of age. C-D: Kaplan Meier survival curve of the different genotypes either homozygous (C) or heterozygous for the ΔNLS mutation (D). Note that all Fus ΔNLS/ΔNLS mice die at birth, unless carrying a hFUS transgene. *, p < 0.05 Log Rank test; ***, p < 0.001 Log rank test. E: Age-dependent changes in the mean hanging time (s) and holding impulse (Ns) in the four-limb wire inverted grid test in Fus +/+ (+/+), and Fus ΔNLS/+ (Δ/+) mice with or without hFUS transgene. N = 10-28 per group. Mixed effect analysis, with 3 factors (Age, ΔNLS genotype and hFUS genotype). P < 0.001 for ΔNLS genotype, P < 0.001 for age, P < 0.001 for hFUS genotype. A significant protective interaction is observed between ΔNLS and hFUS genotypes (p = 0.0216, and p = 0.0366). Only 4 groups out of 5 are shown here for clarity. The whole dataset is shown in Fig S1. F: Hindlimb grip strength in female and male mice. Mixed effect analysis, with 3 factors (Age, ΔNLS genotype and hFUS genotype). For female mice, P < 0.001 for ΔNLS genotype, p = ns for age, p = ns for hFUS genotype. A significant protective interaction is observed between ΔNLS and hFUS genotypes (p = 0.0131). For male mice, P < 0.001 for ΔNLS genotype, p = ns for age, p = ns for hFUS genotype. No significant protective interaction is observed between ΔNLS and hFUS genotypes (p = 0.0512) We also observed prominent retention of human intron 7 in all samples derived from mice expressing the hFUS transgene ( Fig. 5D and Fig S4D), which is consistent with the strong conservation of introns 6 and 7 between species ( Fig S5). Thirdly, besides intron skipping and retention, FUS has also been reported to regulate its own levels through the stimulation of miR200 [16]. Another target of miR200 is ZEB1, whose expression is dependent upon levels of miR200 [21,22]. Here, Zeb1 expression appears unchanged in Fus ΔNLS/+ tissues, whether or not expressing the hFUS transgene ( Fig. S6), indirectly suggesting that this latter autoregulatory mechanism is not engaged in the effects mediated by the hFUS transgene.
To identify the predominant autoregulatory mechanism(s) contributing to reduction of mutant FUS by hFUS, i.e. intron retention and/or exon skipping, we performed RT-PCR using trios of oligonucleotides allowing (See figure on previous page.) Fig. 2 hFUS transgene corrects abnormal levels of FUS protein in Fus ΔNLS mice. A: Immunoblot analysis of FUS protein in total extracts from Fus +/+ (+/+) and Fus ΔNLS/+ (Δ/+) brains with or without hFUS transgene and of Fus ΔNLS/ΔNLS mice (Δ/Δ) with hFUS transgene at 22 months of age. Representative results using different antibodies revealing total FUS, the C-terminal (C-ter) NLS, mouse FUS, and human FUS. Vinculin was used as loading controls. Note that these immunoblots were performed on different membranes to avoid cross reaction between different antibodies and one representative Vinculin blot is shown. Uncropped western blots and their corresponding Vinculin western blots are provided in Source data. B: Quantification of immunoblotting experiments of panel A. Quantification of total, C-ter, mouse, human and ADMA-FUS protein levels in cytoplasmic and nuclear fractions of the indicated genotypes. N = 4-8. *p < 0.05, ***p < 0.001vs Fus +/+ , #, p < 0.05 and ###, p < 0.001 vs. indicated genotype by ANOVA followed by Tukey. C: Immunoblot analysis of FUS protein subcellular localization in cortex of Fus +/+ (+/+) and Fus ΔNLS/+ (Δ/+) mice with or without hFUS transgene and of Fus ΔNLS/ΔNLS mice (Δ/Δ) with hFUS transgene at 1 month of age. Representative results using different antibodies revealing total FUS, the C-terminal (C-ter) NLS, mouse FUS, human FUS and asymmetrically dimethylated arginine FUS (ADMA-FUS). SOD1 and HDAC1 are used as loading controls for cytoplasmic and nuclear protein extracts fractions, respectively. Note that these immunoblots were performed on different membranes to avoid cross reaction between different antibodies. Uncropped western blots and corresponding stain free gels are provided in Source data. D: Quantification of western blotting experiments of panel C. Quantification of total, Cter, mouse, human and ADMA-FUS protein levels in cytoplasmic and nuclear fractions of the indicated genotypes. N = 4-8. *p < 0.05, ***p < 0.001vs Fus +/+ , #, p < 0.05 and ###, p < 0.001 vs. indicated genotype by ANOVA followed by Tukey to calculate a percentage of intron retention or exon skipping relative to the total amount of Fus mRNA. As shown in Fig. 5E, and quantitated in Fig. 5 F and H (uncropped gels shown in Source data), a significant proportion of mouse Fus mRNA retained either intron 6 or intron 7 in wild-type tissues, and this proportion decreased with the ΔNLS transgene, consistent with inhibition of autoregulatory intron retention in these conditions [14]. Conversely, the large majority of mouse Fus mRNA appeared to retain intron 6 or intron 7 in the presence of hFUS transgene ( Fig. 5E-H, Fig. S4). In the same conditions, we were unable to detect a significant proportion of Δexon 7 mouse mRNA, suggesting that, while this species can be detected using RT-qPCR (Fig. 5E), it is quantitatively minor. Collectively, these data indicate that wild-type human FUS gene decreases expression of the endogenous Fus gene through increased retention of introns 6 and 7 leading to decreased production of toxic FUS ΔNLS protein, and subsequent alleviation of all the downstream consequences of the expression of cytoplasmically mislocalized mutant FUS.

Discussion
In the current study, we show that providing a wild-type allele of the FUS gene is sufficient to rescue ALS-like phenotypes associated with cytoplasmically retained mutant FUS protein expression. Our result appears a priori paradoxical since the toxicity of FUS mutations was shown to be largely driven by cytoplasmic FUS [7][8][9][10][11][12], that is not expected to be directly compensated by the wild-type protein. Furthermore, overexpression of the wild-type protein was shown to be toxic to neurons [9,17,23].
The wild-type FUS transgene rescues phenotypes associated with the Fus ΔNLS mutation In this study, the hFUS transgene displayed broad protective effects against the lethality driven by the Fus ΔNLS allele in homozygosity. We previously showed that Fus ΔNLS/ΔNLS mice die at birth due to an inability to inflate lungs [11]. Here, this perinatal lethality was fully prevented by the hFUS transgene, and numbers of Fus ΔNLS/ΔNLS mice expressing the hFUS transgene were  obtained at the expected mendelian ratio, consistent with a full rescue. However, and in spite of surviving the perinatal period, a proportion of homozygous rescued mice died prematurely and abruptly at an adult age (about 20 % by one year of age). The only noticeable phenotype observed in these mice was an age-related decrease in body weight gain, but we did not observe prominent weakness, nor obvious ALS-related symptoms. A possible cause of death could be an increased sensitivity to epileptic seizures, as we recently showed that Fus ΔNLS/+ mice display increased spontaneous cortical neuronal activity [18]. Further work on these mice is required to finely characterize their phenotypes and pinpoint to the cause of their premature death. How can the hFUS transgene rescue the perinatal lethality of Fus ΔNLS/ΔNLS mice? Perinatal death of Fus ΔNLS/ΔNLS mice is similar to that of mice with a complete ablation of FUS. This suggests that the presence of FUS in the nucleus during development is required to bypass the perinatal period. Thus, it is likely that the hFUS transgene allows for the production of sufficient functional nuclear FUS to overcome the perinatal lethality of Fus ΔNLS/ΔNLS mice. Unexpectedly, a significant fraction of the mouse FUS protein was found in the nucleus of adult Fus ΔNLS/ΔNLS mice with hFUS transgene, albeit the endogenous FUS protein is completely truncated of the NLS. This is a priori surprising as the NLS is the major domain responsible for interaction with nuclear import receptors such as karyopherin ß2 [24,25]. There are at least two possible explanations to this observation. First, the human FUS protein could support the nuclear import of the mutant ΔNLS mouse protein as wild-type and mutant FUS interact with each other [26]. Second, nuclear import of FUS might be possible through NLS-independent mechanisms. Indeed, recent work has shown that ΔNLS mutants of FUS could still interact through RGG domains with karyopherin ß2 [24,25], as well as other nuclear import receptors [25] leading to significant nuclear import [24] .
Consistent with the protection offered in homozygous mice, the hFUS transgene prevented premature death and muscle weakness in Fus ΔNLS/+ mice. Importantly, the hFUS transgene had no effect per se on survival of wild-type mice. About 30 % of Fus ΔNLS/+ mice died before 2 years of age, which is consistent with previous findings reported in another knock-in model of FUS-ALS [10]. Further confirming the protection offered by the hFUS transgene, the motor defect of Fus ΔNLS/+ mice was also rescued. It should be noted however that the expression of the transgene led to mild motor defects, mostly in males in one of the tests (grip strength) used. This suggests that the protection offered by the hFUS transgene might be accompanied with toxicity appearing with age, echoing a recent report on the toxicity of viral overexpression of SMN in spinal muscular atrophy mice [27]. In all, our results demonstrate broad protective effects of the hFUS transgene on the deleterious phenotypes associated with either homozygous or heterozygous Fus ΔNLS mutation, yet not excluding some residual toxicity associated with the transgene expression.
The wild-type hFUS transgene mitigates disease through autoregulatory mechanisms Importantly our work demonstrates that the wild-type transgene activates FUS autoregulatory loop to mitigate the phenotype. A first possible protective mechanism could have been that the hFUS transgene rescues a loss of nuclear FUS. However, we did not observe any loss of nuclear FUS in Fus ΔNLS/+ mice. Alternatively, the hFUS transgene appears to indirectly protect from accumulation of mutant protein through the autoregulatory loop maintaining nuclear FUS levels [14,15,20] to avoid the toxicity of loss of nuclear FUS [28][29][30][31][32] or its excess [9,17,23]. We provide several lines of evidence demonstrating the engagement of FUS autoregulation upon expression of the hFUS transgene. First, Fus mRNA and protein levels are increased in Fus ΔNLS/+ mice, thereby compensating the proportion of FUS protein translated from the mutant allele and unable to enter the nucleus. Conversely, in single hFUS transgenic mice the addition of the exogenous FUS transgene is sufficient to decrease endogenous mouse Fus mRNA levels, consistent with previous studies [8,9]. Here, the addition of the hFUS transgene in Fus ΔNLS/+ mice rescued overexpression of endogenous Fus in Fus ΔNLS/+ mice, (See figure on previous page.) Fig. 5 hFUS transgene activates autoregulatory splicing in Fus ΔNLS/+ spinal cord. A-D: RT-qPCR results for endogenous Fus mRNA deleted of exon 7 (A), endogenous Fus mRNA retaining intron 6 (B), endogenous Fus mRNA retaining intron 7 (C) and exogenous FUS mRNA retaining intron 7 (D) in spinal cord at 1 month (left) or 22 months (right) of age. Note that the hFUS transgene activates autoregulatory exon 7 skipping as well as retentions of introns 6 and 7 in endogenous mRNA and retention of intron 7 in exogenous mRNA at 1 and 22 months of age. N = 4-8. *, p < 0.05, **p < 0.01, ***p < 0.001 vs. Fus +/+ , #, p < 0.05, ###, p < 0.001 vs. indicated genotype by ANOVA followed by Tukey. E-H: representative gel electrophoresis of RT-PCR assays identifying RNA species with or without intron 6 retention (E, G, upper panel), with or without intron 7 retention (E, G, middle panel), or with or without exon 7 skipping (E, G, lower panel) in spinal cord at 1(E) or 22 (G) months of age. We did not detect exon 7 skipped mRNA using these assays. Panels F and H show the percentage of intron 6 or 7 retention (intron + band intensity divided by the sum of intensities of intron + and intronbands, multiplied by 100), for 1 month (F) or 22 months (H) of age. N = 4-8. ***p < 0.001 vs. Fus +/+ , #, p < 0.05, ##, p < 0.01, ###, p < 0.001 vs. indicated genotype by ANOVA followed by Tukey and decreased mutant mRNA levels. Since this overexpression acts as a feed forward mechanism amplifying the cytoplasmic accumulation of FUS, avoiding this overexpression might on its own be sufficient to slow down the vicious cycle leading to phenotypes in Fus ΔNLS/+ mice. Consistently, our RT-PCR experiments demonstrate that a significant proportion of Fus mRNA retains introns 6 and 7, and that the percentages of intron retention in the endogenous Fus mRNA increase strikingly with the expression of the hFUS transgene. Importantly, the exogenous human transgene is also, on its own, subject to autoregulation in Fus ΔNLS/+ mice, despite the heterologous system. This is consistent with the strikingly high conservation of introns 6 and 7 of the FUS gene between species suggesting that autoregulation of FUS is critical for its functions (Fig. S5). Of note, the existence of autoregulation of the human transgene in the mouse model is a plausible explanation for the high toxicity of cDNA-based constructs devoid of required autoregulatory sequences, and the relative innocuity of genomic based constructs [8,9,17].
Our findings are in agreement with previous studies identifying retention of introns 6 and/or 7 as the major mechanism of Fus autoregulation [14]. Indeed, about half of the endogenous Fus transcript appears retaining either intron 6 or 7 in the cortex or spinal cord of wildtype mice. Consistent with the results of Humphrey and collaborators, the ΔNLS mutation leads to decreased retention of these two introns in both tissues. On the contrary, the introduction of the hFUS transgene leads to substantial retention of both introns, with nearly 90 % of endogenous Fus mRNA having intron 6 retained. The effect of the hFUS transgene appeared less marked on retention of intron 7, albeit this intron carries most of the FUS binding sites on the pre-mRNA [14,15,20]. In addition to the substantially increased intron retention upon hFUS transgene expression, we also observed enhanced exon 7 skipping. This mRNA species appeared however minor, as it was not observed using splicing assays, and required 4-6 supplementary PCR cycles to be detectable. While this suggests that exon skipping is a minor mechanism of FUS autoregulation in our in vivo model, our current results do not allow to completely exclude its contribution as Zhou and collaborators demonstrated that Δexon 7 Fus mRNA is subject to nonsense mediated mRNA decay (NMD) [15]. Nevertheless, other studies suggest that NMD is not involved in FUS autoregulation, at least in cultured cells [14]. Further, our study did not find evidence of altered miR141/200/ ZEB1 pathway [16] by measuring Zeb1 mRNA levels. Thus additional work is needed to fully evaluate the contribution of this pathway to in vivo FUS autoregulation. In all, our efforts are consistent with a predominant role of intron retention in FUS autoregulation in vivo, and further research is warranted to identify which intron is critical for this process.

Possible consequences for therapeutic strategies in FUS-ALS
Our results suggest that gene therapy to reintroduce the wild-type protein, while including sequences required for autoregulation, would enable the correction of molecular and behavioral phenotypes, meanwhile avoiding the toxicity of wild-type protein overexpression in FUS-ALS. Our work provides a proof of concept for a potential therapeutic strategy, albeit there are limitations to overcome before clinical translation. First, our study has been obtained in a heterologous system, with a human FUS transgene expressed in mouse cells. While the very high conservation of intronic sequences (Fig S5) gives hopes that a similar intervention should be protective also in human cells, an intermediate validation step using a human cell model is warranted. Second, our current results have been obtained using a complete FUS gene inserted through classical transgenesis in a locus independent of the mouse Fus gene. To translate these results in a therapeutically viable strategy, it would be first necessary to use a gene therapy vector, such as an adeno-associated virus (AAV), to provide the equivalent of our hFUS transgene. Thus, an important effort of sequence optimization is required to shorten the lead "therapeutic" construct, in order to allow introduction of a potential therapeutic sequence into the viral vector. Future research should thus aim at identifying the minimal sequence requirements for FUS autoregulation to ultimately engineer a small autoregulation competent expression construct.

Conclusions
Our results show that the phenotypes triggered by a cytoplasmically retained FUS protein associated to ALS can be rescued by a wild type FUS allele. The wild-type FUS allele activates the homeostatic autoregulatory loop triggering retention of introns 6 and 7 in the endogenous Fus mRNA, leading to decreased mutant protein load. Our work provides a proof of concept for a potential gene therapy strategy for FUS-ALS.

Mouse models and genotyping
Mouse experiments were approved by local ethical committee from Strasbourg University (CREMEAS) under reference number 2,016,111,716,439,395 and all experimental procedures performed in San Diego were approved by the Institutional Animal Care and Use Committee of the University of California, San Diego. Transgenic mice were generated as described in [11,12] and [8], were bred in Charles River animal facility and housed in the Faculty of medicine from Strasbourg University with 12/12 hours of light/dark cycle (light on at 7:00 am) under constant conditions (21 ± 1°C; 60 % humidity) and with unrestricted access to food and water.
Mice were weaned and genotyped at 21 days by PCR from tail biopsy, or at death if occurring before 21 days of age.
The following primer sequences were used to genotype mice: hFUS Heterozygous Fus ΔNLS/+ knock-in mice, lacking the PY-NLS, were crossed with mice expressing human wild type FUS from a complete, autoregulatory competent, human gene to obtain following genotypes: Fus +/+ , Fus ΔNLS/+, Fus ΔNLS/ΔNLS , Fus +/+ /hFUS, Fus ΔNLS/+ /hFUS, Fus ΔNLS/ΔNLS /hFUS. The genetic background of all mice used in this study is C57Bl6/J. Breeding steps were performed in parallel in both laboratories. 76 mice of the F2 generation were generated in Strasbourg, and 110 mice of the F2 generation were generated in San Diego.

Mouse behavior Survival
Survival was studied during the first hours after birth and dead new born mice were genotyped. Mice surviving the post-natal period were genotyped at 21 days and followed weekly until death or euthanized using ketamine-xylazine when they reach the following endpoints: auto-mutilation, weight loss greater than 10 % of the initial weight and when they could not turn around again within 10 s after being laid on their side.

Inverted grid
Mice were habituated for 30 min in the test room prior testing. Motor performance was assessed weekly as described previously [12] from 1 month until 22 months of age. The wire grid hanging time (or "hang time") was defined as the amount of time that it takes the mouse to fall down from the inverted grid and was measured visually with a stopwatch. The procedure was repeated 3 times during 5 min with 5 min break between tests. All mice were returned to their homecage after completing the test. The holding impulse corresponds to hanging time normalized with mouse weight and gravitational force.

Grip test
Grip strength was measured using a Grip Strength Meter (Columbus Instruments, Columbus, OH) on cohorts (N = 12-30) made up of approximately the same number of males and females. Mice were allowed to grip a triangular bar only with hind limbs, followed by pulling the mice until they released; five force measurements were recorded in each separate trial.

Histological techniques
Mice aged of 22 months were anesthetized with intraperitoneal injection of 100 mg/kg ketamine chlorhydrate and 5 mg/kg xylazine then perfused with PFA 4 %. After dissection, spinal cord was included in agar 4 % and serial cuts of 40 μm thick were made with vibratome.

RNA extraction and RT-qPCR
Total RNA was extracted from spinal cord and frontal cortex using TRIzol® reagent (Life Technologies). 1 µg of RNA was reverse transcribed with iScript™ reverse transcription (Biorad, 1,708,841). Quantitative polymerase chain reaction was performed using Sso Advanced Universal SYBR Green Supermix (Bio-Rad 1,725,274) and quantified with Bio-Rad software. Gene expression was normalized by calculating a normalization factor using actin, TBP and pol2 genes according to GeNorm software [55].
Primer sequences are provided in Table S1. ) and stained with ethidium bromide using standard procedures. For quantification, we quantified individually the signal intensities of the two bands, and computed a % of intron retention as such: (intensity of Intron + band )/ (intensity of Intron + band + intensity of Intron-band)*100. We did not quantify a percentage of exon 7 skipping as the exon 7 skipped product was below the detection threshold of the assay.

Statistics
All results from analysis are presented as mean ± standard error of the mean (SEM) and differences were considered significant when p < 0.05. Significance is presented as follows: * p < 0.05, ** p < 0.01, and *** p < 0.001. For comparison of two groups, two-tailed unpaired Student's t -test was used in combination with Ftest to confirm that the variances between groups were not significantly different. For longitudinal analysis of behavioral data, results were analyzed using a mixed effect analysis with three factors (ΔNLS genotype, hFUS genotype and age) as indicated in the figure legends. Data were analyzed by using the GraphPad Prism version 8.0.