Cytoplasmic mislocalization of mutant FUS in knock-in mice and ALS fibroblasts

Using a series of novel antibodies that distinguish wild-type FUS from the mutant isoforms of the protein (Extended Data Fig. 2a–g), we observed diffuse cytoplasmic mislocalization of mutant, but not wild-type, FUS in P517L/WT and Δ14/WT animals, as well as strong nuclear staining for mutant FUS (Fig. ​(Fig.1f).1f). We next quantified the cytoplasmic versus nuclear distribution of wild-type and mutant FUS by subcellular fractionation and western blot analysis. Similar to the results of immunostaining experiments, we found that a minority (~26–45%) of wild-type FUS is cytoplasmic in the spinal cords of 1-year-old wild-type, P517L/WT and Δ14/WT animals (Fig. 1g,h). By contrast, most mutant FUS (~68–86%) is cytoplasmic in P517L/WT and Δ14/WT mutants (Fig. ​(Fig.1h).1h). Although mutant FUS does not appear to carry wild-type FUS into the cytoplasm, we asked if any of the RBPs known to interact directly or indirectly with FUS^22,23 were also mislocalized. Indeed, several of these RBPs (including TDP-43, hnRNP H and hnRNP U in P517L/WT and hnRNP U in Δ14/WT) are observed at a significantly higher proportion in the cytoplasm of mutant animals compared to wild-type controls (Fig. 1g,h). Taken together, these experiments show that mutant, but not wild-type, FUS is selectively mislocalized to the cytoplasm along with interacting RBPs in P517L/WT and Δ14/WT animals.

Open in a separate window

Extended Data Fig. 2

Variable reactivity of commercial antibodies with mutant FUS: characterization of mutant FUS expression and localization using novel mutant FUS-specific antibodies.

(a, b) Schematics of P517L and Δ14 mutant FUS proteins (top). Mutations are indicated by red asterisk for P517L and red C-terminal rectangle for Δ14. Bottom shows C-terminal amino acid sequences of FUS WT protein, FUS P517L and FUS Δ14 proteins, and the P517L and Δ14 synthetic peptides used for the generation of FUS-P517L and FUS-Δ14 antibodies (bottom). Red letters in the amino acid sequences indicate the residues specific to the FUS mutants. The locations of the epitopes for FUS-P517L and FUS-Δ14 antibodies are marked by horizontal bars below the C-termini of the corresponding schematized mutant proteins (bottom right). The murine and human C-terminal FUS amino acid sequences are conserved, and the illustrated mouse sequences also correspond to the human FUS P525L and Δ14 proteins. (c, d) Immunostaining of adult spinal cord sections of 1-year-old heterozygous mutant (P517L/WT and Δ14/WT) and WT/WT mice using FUS-Abcam[1–50] (FUS, white) and FUS-P517L or FUS-Δ14 (red) antibodies. Note the cytoplasmic staining with FUS-P517L andFUS-Δ14 antibodies in the animals that express the corresponding mutant FUS protein (upper panels) and the absence of staining of wild type FUS protein in WT/WT animals using either the P517L or the Δ14 antibodies (bottom panels). Scale bar: 20µm. (e) Representative images of spinal cord sections from WT/KO and P517L/KO animals show lack of immunoreactivity of three commercial antibodies with P517L mutant FUS protein. Scale bar=100µm. (f) Immunostaining of human cultured fibroblasts from an ALS-FUS^P525L patient (top) and a non-ALS control (bottom) using two distinct FUSWT antibodies (white) and the FUS-P517L(P525L) antibody (red). Identical confocal microscopy settings (laser power and gain) were used for imaging the top and the bottom panels. Scale bar: 50µm. (g) Denaturing immunoblot of brains of newborn mice of the indicated genotypes demonstrates variable affinities of five commercially available anti-FUS antibodies for mutant relative to wild type FUS. Note, for example, absence of visible bands for either P517L or Δ14 mutant FUS protein with the FUS-Abcam[1–50] antibody. In addition, absence of visible bands for WT protein confirms the specificity of FUS-P517L and FUS-Δ14 antibodies for mutant FUS. (h) Representative images of MNs from spinal cord sections of 2-year-old WT/WT (left), Δ14/WT (middle), and P517L/WT (right) animals stained with FUS-Abcam[1–50], FUS-Δ14, and FUS-P517L antibodies, respectively, show lack of obvious inclusions. Scale bar=5µm.

Source data

Dose-dependent toxicity of mutant FUS associated with RBP insolubility

The question of whether pathogenic mutations cause a functional deficiency of FUS or result in a toxic gain of function is critical to determining a therapeutic approach to ALS-FUS, by either silencing or restoration of lost function. FUS plays a critical role in development, as shown by the perinatal lethality of the elimination of FUS in knockout mice (KO/KO)²⁴. However, in previous studies¹², we showed that postnatal FUS deficiency is insufficient to cause MN degeneration, and we concluded, based on indirect evidence, that the toxicity of mutant FUS was a consequence of a gain of function. Here, using the P517L and Δ14 mutant FUS knock-in alleles in combination with the wild-type and knockout FUS alleles, we took a genetic approach to assess the effect of these mutations on FUS function and the effect of gene dosage on the degenerative MN phenotype that we observed in vivo in these knock-in animals. In a series of experiments, we crossed mice carrying Fus wild-type, mutant knock-in (P517L or Δ14) and knockout alleles to generate homozygous, heterozygous and hemizygous animals for each Fus allele. These experiments show that P517L and Δ14 mutant proteins retain partial FUS function and incompletely rescue the FUS knockout phenotype (Fig. 2a,b, Extended Data Fig. 3a,b and Table ​Table11).

Open in a separate window

Fig. 2

Dose-dependent toxicity and detergent insolubility of mutant FUS.

a, Kaplan–Meier survival curves for mice with the indicated combinations of Fus WT, P517L, Δ14 and null-knockout alleles. b, Median survival of selected genotypes plotted on logarithmic scale illustrates partial functionality and dose-dependent toxicity of mutant FUS. Increased survival of P517L/KO and Δ14/KO compared with KO/KO animals shows that mutant FUS protein is able to partially rescue the null phenotype and, thus, is functional. Comparison of P517L/P517L and P517L/Δ14 versus P517L/KO animals shows that further addition of mutant FUS protein decreases survival, consistent with dose-dependent toxicity of mutant FUS protein. Median survival was estimated for P517L/P517L (0.15 d) and KO/KO (0.2 d), as most newborn pups for these genotypes were found dead and, thus, could not be accurately quantified. The bar for P517L/KO is reproduced three times for comparison purposes. c, Immunoblot of sarkosyl-insoluble fractions from brains of newborn Fus WT/WT, heterozygous P517L/WT and homozygous P517L/P517L mice. Each lane corresponds to a separate animal. d, Immunoblot of sarkosyl-insoluble fractions from brains of 2-year-old Fus WT/WT and heterozygous P517L/WT and Δ14/WT mice. Each lane corresponds to a separate animal. e, Immunoblot of sarkosyl solubility fractionation of human brain stems of a non-ALS control and a patient with ALS-FUS P525L. f, Immunoblot of sarkosyl-insoluble fractions from human brain stem samples from a non-ALS control and a patient with ALS-FUS P525L. g, Relative abundance of inclusion isoforms of exons regulated by hnRNPH. The observed pattern is consistent with functional deficiency of hnRNPH, which promotes inclusion of exon 24 of ATXN2 (left) and inhibits the inclusion of exon 8 of hnRNPDL transcript (right). *P<0.05, **P<0.01 and ***P<0.001, using one-way ANOVA with Tukey’s post hoc test. n=4 animals per genotype. SI, sarkosyl insoluble; Sol, soluble (in hypotonic buffer); SS,sarkosyl soluble (in 1% sarkosyl and high salt).

Source data

Table 1

Summary of phenotypes of mutant FUS knock-in mouse lines

Genotype	Birth weighta	Suvival	MN loss	Muscle innervation	Other phenotypes
WT/WT	1.40±0.01 g n=183	Normal	Normal	Normal	NA
P517L/WT	1.37±0.01 g n=157	~90% survive to weaning	Onset at 1.5 years	TA denervation onset at 1 year	Malocclusion and rectal prolapse (all rare)
Δ14/WT	1.36±0.01 g n=173	~90% survive to weaning	Onset at 1.5 years	TA denervation onset at 1 year	Malocclusion and rectal prolapse (all rare)
WT/KO	1.34±0.02 g n=95	Normal	Normal	Normal	No known phenotype
KO/KO	1.00±0.02 g n=26	Perinatal lethal	Normal	Normal at birth	Hicks et al., 2000
P517L/KO	1.15±0.02 g n=14	21±12 d	NA	Normal at birth	Malocclusion, rectal prolapse and seizures
Δ14/KO	1.02±0.02 g n=13	Perinatal lethal; max survival ~1 week	NA	NA	NA
P517L/P517L	0.98±0.02 g n=26	Perinatal lethal	Normal at birth	Normal at birth	NA
Δ14/Δ14	1.09±0.01 g n=50	~20% survive to weaning	Onset at 1 year	NA	Malocclusion, rectal prolapse and seizures
P517L/Δ14	1.10±0.02 g n=44	Perinatal lethal; max survival ~3 weeks	NA	NA	NA
MN-Δ14/WT: c14/WT; ChAT-Cre	NA	NA	Onset at 1.5 years	NA	NA
MN-Δ14/Δ14: c14/c14; ChAT-Cre c14/Δ14; ChAT-Cre	NA	NA	Onset at 1 year	NA	NA
MN-P517L/Δ14: P517L/c14; ChAT-Cre	1.42±0.03 g n=28	~90% survive to weaning	Onset at 4 months	TA denervation onset at 4 months	Malocclusion, rectal prolapse and seizures (all rare)

Open in a separate window

^aMean ± s.e.m. n represents number of individual animals. NA, not available.

Open in a separate window

Extended Data Fig. 3

Decreased lifespan of mutant FUS heterozygous animals and absence of NMJ denervation and MN loss in newborn P517L/P517L mice.

(a) Box (median, 25^th and 75^th percentiles) and whiskers (2.5^th and 97.5^th percentiles) plot of birth weights of mice with the indicated combinations of Fus WT, P517L, Δ14, and null-knockout (KO) alleles. *P<0.05, **P<0.01 and ***P<0.001, using one-way ANOVA with Tukey’s post hoc test. N=13–183 animals per genotype as indicated in Table ​Table1.1. (b) Selected data reproduced from (a) to illustrate partial functionality and dose-dependent toxicity of mutant FUS. Increased birth weight of P517L/KO compared to KO/KO animals demonstrates that mutant FUS protein is able to partially rescue the null phenotype and thus is functional. Comparison of P517L/P517L versus P517L/KO animals demonstrates that further addition of mutant FUS protein decreases birth weight, consistent with dose-dependent toxicity of mutant FUS protein. Statistical significance was assessed in (a) as a part of a full analysis of all genotype groups. (c) Kaplan-Meier survival curves for successfully weaned WT/WT, P517L/WT, Δ14/WT, and WT/KO mice. Both heterozygous mutants but not WT/KO have significantly decreased median survival age compared to wild type controls. All possible pairwise comparisons were performed using Log-rank (Mantel-Cox) test and the resulting p-values were adjusted for multiple comparison using the Bonferroni correction. (d) NMJ staining of tibialis anterior (TA) muscle of newborn WT/WT (left) and homozygous P517L/P517L(right) animals using antibodies to pre-synaptic synaptophysin (green) and alpha-bungarotoxin (magenta, post-synaptic). Scale bar= 100µm. (e) Quantification of innervated NMJs in the WT/WT (black) and homozygous P517L (dark red) animals. N=3 for WT/WT and 4 for P517L/P517L groups respectively. (f) Immunofluorescence staining of lumbar level 4–5 (L4-L5) MNs in WT/WT (left) and homozygous P517L/ P517L (right) animals using anti-ChAT and anti-P517L antibodies (right inset panel). Scale bar=25µm for the left and middle panels, 100µm for the right panel. (g) Quantification of ChAT-positive neurons in lumbar levels 4–5 (L4-L5) in newborn WT/WT (black) and homozygous P517L/ P517L (dark red) animals. N=3 animals per genotype. For e and g, statistical significance was assessed using Welch’s t-test. Error bars represent SD.

The partial loss of function caused by the P517L and Δ14 mutations could selectively affect specific activities of FUS or incompletely reduce all protein activities. In the latter case, two copies of the mutant allele would be expected to restore more of the normal FUS activity and further improve the knockout phenotype. Indeed, this is the case for Δ14, as the homozygous mutants (Δ14/Δ14) showed an improved survival compared to the hemizygous mutants (Δ14/KO) despite similar birth weights (Fig. 2a,b, Extended Data Fig. ​Fig.3a3a and Table ​Table1).1). By contrast, P517L/P517L animals were born at lower than expected Mendelian ratios and were smaller than P517L/KO animals, and none survived more than several hours after birth (Fig. 2a,b, Extended Data Fig. 3a,b and Table ​Table1).1). Surprisingly, P517L/Δ14 animals also had reduced survival compared to the P517L/KO animals (Fig. 2a,b and Table ​Table1).1). The exacerbation of the low birth weight and poor survival phenotype of the P517L/KO mutants by additional P517L protein (in P517L/P517L homozygous animals) or Δ14 protein (in P517L/Δ14 compound heterozygous mutants) shows dose-dependent gain-of-function toxicity of P517L and Δ14 mutant FUS. The increased severity of the P517L/P517L phenotype compared to the P517L/Δ14 phenotype suggests that P517L is more toxic than Δ14 mutant FUS, and this conclusion is further supported by decreased long-term survival of successfully weaned P517L/WT animals relative to Δ14/WT animals (Extended Data Fig. ​Fig.3c).3c). Notably, although each of these mutant alleles is sufficient to cause age-dependent, selective MN degeneration in the heterozygous mice, the decreased perinatal survival of P517L/P517L, P517L/Δ14 and Δ14/Δ14 animals appears to be unrelated to MN degeneration, as no denervation or MN loss was detected in these animals at postnatal day 0 (P0) (Extended Data Fig. 3d–g and Table ​Table11).

Mutant FUS toxicity might relate to its irreversible incorporation into cytoplasmic inclusions or detergent-insoluble protein aggregates, resulting in aberrant RNA processing^25,26. Although we do not observe obvious cytoplasmic FUS inclusions in MNs by immunofluorescent staining (Extended Data Fig. 2c,d,h), biochemical fractionation of brains of P0 animals reveals increased partitioning of FUS from the soluble to the detergent-soluble and detergent-insoluble fractions in P517L/WT and P517L/P517L animals compared to wild-type controls (Extended Data Fig. 4a,b). Indeed, side-by-side analysis of only the detergent-insoluble fractions from individual brains fractionated in the same experiments clearly shows the increase in detergent-insoluble FUS in P517L/WT relative to wild-type animals and a further accumulation of insoluble FUS in P517L/P517L relative to P517L/WT animals (Fig. ​(Fig.2c2c and Extended Data Fig. ​Fig.4c).4c). Given the insolubility of a broad range of related RBPs that we observed in the brain and spinal cord of patients with ALS/FTD carrying the C9orf72 expansion mutation and in sporadic ALS²⁷, we looked in the P517L/WT and P517L/P517L animals to see if FUS insolubility was associated with a similar increase in insoluble RBPs. We found increased partitioning of TDP-43 as well as hnRNP A1, hnRNP H, hnRNP U and other RBPs from the soluble to the detergent-soluble and detergent-insoluble fractions in newborn and 2-year-old P517L/WT animals (Fig. 2c,d and Extended Data Fig. 4c,d). To validate that the RBP insolubility in the FUS mutant mice reflects pathological changes in ALS-FUS, we performed the same fractionation on a postmortem brain sample from an ALS-FUS^P525L patient and showed insolubility of both total and P525L mutant FUS that was not seen in the age-matched, non-neurological control brain (Fig. ​(Fig.2e).2e). Moreover, we also observed increased partitioning of TDP-43, hnRNP A1, hnRNP H, hnRNP U and other RBPs from the soluble to the detergent-insoluble fractions of the FUS^P525L ALS brain, showing that, in patients with ALS-FUS, as in the mutant mice, FUS drives the insolubility of related RBPs, including TDP-43 (Fig. ​(Fig.2f2f).

Open in a separate window

Extended Data Fig. 4

Detergent insolubility of mutant FUS and related RNA-binding proteins.

(a) Immunoblot of sarkosyl solubility fractionation of brain from newborn FUS WT/WT, P517L/WT, and P517L/P517L mice. Sol = soluble (in hypotonic buffer), SS=sarkosyl soluble (in 1% sarkosyl and high salt), and SI=sarkosyl insoluble fractions. (b) Quantitation of protein in Sol (Soluble), SS (Sarkosyl Soluble), and SI (Insoluble) fractions. *P<0.05, **P<0.01 and ***P<0.001 comparing Soluble versus the sum of Sarkosyl Soluble and Insoluble fractions using one-way ANOVA with Tukey’s post hoc test. N=3 animals per group. (c) Quantitation of protein in sarkosyl insoluble fractions from newborn animals shown in Fig. ​Fig.3c,3c, expressed as log-ratio of wild type. (d) Quantitation of protein in sarkosyl insoluble fractions from 2-year-old animals shown in Fig. ​Fig.3d,3d, expressed as log-ratio of wild type. For c and d, *P<0.05, **P<0.01 and ***P<0.001 using one-way ANOVA with Tukey´s post hoc test. N=3 animals per genotype.

Source data

To test the functional significance of this increased insolubility of FUS and other RBPs, we looked in the mutant animals for evidence of aberrant splicing of their target pre-mRNAs, similar to that which we described previously in brain and spinal cord tissue from C9orf72-positive and gene-negative patients with ALS/FTD²⁷. As in those cases, RBP insolubility in the Δ14/WT and P517L/WT animals was also associated with alteration in splicing. For example, analysis of two targets of hnRNP H showed decreased inclusion of ATXN2 exon 24 and increased inclusion of hnRNPDL exon 8—consistent with decreased hnRNP H activity—in Δ14/WT and P517L/WT animals compared to wild-type controls (Fig. ​(Fig.2g).2g). P517L/WT showed a greater magnitude of detergent insolubility and splicing alterations compared to Δ14/WT animals (Fig. 2d,g and Extended Data Fig. ​Fig.4d),4d), once again suggesting increased toxicity of P517L over Δ14.

Taken together, these experiments reveal that normal FUS function is essential for perinatal mouse survival and that P517L and, to a lesser extent, Δ14 mutant FUS protein retain sufficient activity for a single copy of either allele to rescue the knockout phenotype, albeit partially. However, when expressed at higher levels in homozygous or compound heterozygous animals, mutant FUS protein is toxic in a dose-dependent manner, decreasing mutant survival. This toxicity is associated with increased insolubility of FUS, TDP-43 and related RBPs, many of which have been independently implicated in the pathogenesis of ALS. This insolubility leads to the consequent loss of function of these regulatory proteins and to the aberrant processing of their respective target RNAs.

Cell-autonomous expression of FUS Δ14 accelerates MN degeneration

Consistent with their highly pathogenic nature in patients with ALS-FUS, a single copy of the P517L or Δ14 allele is sufficient to cause progressive MN degeneration in the lifetime of heterozygous animals, providing a faithful mouse model of disease. As we showed, increasing the dose of mutant FUS in the P517L/P517L, P517L/Δ14 and Δ14/Δ14 mice results in increased toxicity, leading to perinatal lethality, seemingly due to widespread toxicity of mutant FUS in embryonic development. To avoid this developmental toxicity and restrict the dose-dependent effects of mutant FUS to MNs, we generated a Δ14/c14 mutant that carries the Δ14 allele and the conditional FUS c14 allele, which encodes wild-type FUS but is transformed into the Δ14 allele after Cre-dependent recombination (Fig. ​(Fig.1b).1b). We used the ChAT-Cre allele²⁸ to activate FUS c14 selectively in cholinergic cells so that only MNs are Δ14/Δ14 homozygous in this animal, whereas non-cholinergic cells remain Δ14/c14 and express only a single copy of the Δ14 mutant allele. To account for potential non-MN autonomous mutant FUS-dependent effects, we generated additional genotypes that are MN heterozygous or homozygous in a WT or Δ14/WT background (Fig. ​(Fig.3a3a).

Open in a separate window

Fig. 3

Increased dosage of mutant FUS in MNs accelerates selective MN degeneration.

a, Table relating the indicated mouse genotypes to WT and Δ14 FUS protein expression in MN and non-MN cells. b, Numbers of ChAT-positive MNs at lumbar levels 4 and 5 in 1-year-old WT/WT (left), MN-Δ14/WT (middle, c14/WT; ChAT-Cre and Δ14/WT) and MN-Δ14/Δ14 (right, c14/c14; ChAT-Cre, Δ14/c14; ChAT-Cre and Δ14/Δ14) animals normalized to the wild-type controls. Each of the MN-Δ14/Δ14 genotypes shows fewer lumbar MNs in comparison to either of MN-Δ14/WT genotypes or WT/WT group, using one-way ANOVA followed by Fisher’s least significant difference. Data are shown as mean±s.d. n=3 animals per genotype. c, Density of Iba1-positive microglial cells at lumbar levels 4 and 5 in 1-year-old WT/WT (left, black), MN-Δ14/WT (middle, c14/WT; ChAT-Cre and Δ14/WT) and MN-Δ14/Δ14 (right, c14/c14; ChAT-Cre, Δ14/c14; ChAT-Cre and Δ14/Δ14) animals. Each of the MN-Δ14/WT genotypes show an increase in microglial density in comparison to WT/WT animals. Each of the MN-Δ14/Δ14 genotypes has increased microglial density in comparison to either of MN-Δ14/WT genotypes or WT/WT group. Statistical significance was determined by one-way ANOVA with Tukey’s post hoc test. Data are shown as mean±s.d. n=3 animals per genotype. d, Immunostaining of lumbar spinal cord sections of adult MN-P517L/Δ14 (P517L/c14; ChAT-Cre) animals using anti-ChAT (green), anti-FUS-WT (Abcam[1-50], yellow), anti-FUS-Δ14 (cyan) and mouse monoclonal anti-FUS-P517L (red) antibodies. Note anti-FUS-P517L staining in all cells, anti-FUS-Δ14 staining only in ChAT-positive cells and anti-FUS-WT staining only in ChAT-negative cells. Scale bar, 100µm. e, Percentage of completely innervated NMJs (that is, not partially or completely denervated) in TA (left) and soleus (right) muscles of CTRL (FUS WT-expressing control, black) and MN-P517L/Δ14 (P517L/c14; ChAT-Cre, green) animals. f, MN numbers at lumbar level 4 and 5 of CTRL (FUS WT-expressing control, black) and MN-P517L/Δ14 (P517L/c14; ChAT-Cre, green) animals normalized to the controls. g, Histogram of MN soma cross-sectional areas of CTRL (FUS WT-expressing control, black) and MN-P517L/Δ14 (P517L/c14; ChAT-Cre, green) animals. MNs with soma area ≥475µm² were classified as alpha-MNs, and MNs with soma area <475µm² were classified as gamma-MNs. Inset shows the numbers of ChAT-positive gamma-MNs (left) or alpha-MNs (right) at lumbar levels 4 and 5 in 2-year-old CTRL (FUS WT-expressing control, black) and MN-P517L/Δ14 (P517L/c14; ChAT-Cre, green) animals normalized to the controls. *P<0.05, **P<0.01 and ***P<0.001, using Welch’s t-test. Data are shown as mean±s.d. n=3 animals per group. For e and f, *P<0.05, **P<0.01 and ***P<0.001, using two-way ANOVA with Sidak´s post hoc test. Data are shown as mean±s.d. n=3 animals per group. CTRL, littermate c14/WT and/or c14/c14 animals depending on the breeding scheme.

To test whether an MN-restricted ‘double dose’ of mutant FUS accelerates ALS-related pathology, we compared mice expressing either one or two copies of mutant FUS in MNs at 1 year of age—a time point at which MN loss is not yet detected in heterozygous mutant FUS animals (Fig. ​(Fig.1c).1c). As expected, no MN loss was observed at 1 year in ‘MN-Δ14/WT’ mice expressing one copy of a Δ14 allele in MNs (Δ14/WT and c14/WT; ChAT-Cre) (Fig. 3a,b). By contrast, ‘MN-Δ14/Δ14’ mice expressing two Δ14 alleles in MNs (Δ14/Δ14, Δ14/c14; ChAT-Cre and c14/c14; ChAT-Cre) all showed a similar level of approximately 10% MN loss, indicating acceleration of MN degeneration by the expression of the second Δ14 allele in MNs (Fig. 3a,b). Similarly, an increase in the Iba1 microglial marker staining was observed in MN-Δ14/Δ14 animals relative to MN-Δ14/WT mice, showing that expression of the second Δ14 allele in MNs accelerates microgliosis (Fig. 3a,c). Time course analysis revealed that microgliosis is first detected at 6 months, and MN degeneration is first detected at 1 year, in MN-Δ14/Δ14 mice, showing that expression of an additional Δ14 allele in MNs of MN-Δ14/Δ14 mice accelerates the onset of both microgliosis and MN loss by 6 months compared to MN-Δ14/WT animals (Extended Data Fig. 5a–d).

Open in a separate window

Extended Data Fig. 5

Conditional c14 expression: NMJ denervation, selective motor neuron loss, gliosis and grip strength.

(a) Numbers of ChAT positive motor neurons at lumbar levels 4 and 5 in WT/WT (black) and MN-Δ14/WT (c14/WT; ChAT-Cre and Δ14/WT, purple and blue respectively) animals normalized to the wild type controls. (b) Numbers of ChAT positive motor neurons at lumbar levels 4 and 5 in WT/WT (black) and MN-Δ14/Δ14 (c14/c14; ChAT-Cre, Δ14/c14; ChAT-Cre, and Δ14/Δ14, red, green, and brown respectively) animals normalized to the controls. (c) Density of Iba1-positive microglial cells at lumbar levels 4 and 5 in WT/WT (black) and MN-Δ14/WT (c14/WT; ChAT-Cre and Δ14/WT, purple and blue respectively) animals. (d) Density of Iba1-positive microglial cells at lumbar levels 4 and 5 in WT/WT (black) and MN-Δ14/Δ14 (c14/c14; ChAT-Cre, Δ14/c14; ChAT-Cre, and Δ14/Δ14, red, green, and brown respectively) animals. (e) Percentages of NMJs classified as fully innervated (green), partially denervated (yellow), and fully denervated (red) in the tibialis anterior (left) and soleus (right) muscles in CTRL (FUS WT-expressing control) and MN-P517L/Δ14 (FUS P517L/c14; ChAT-Cre) animals. *P<0.05, **P<0.01 and ***P<0.001 indicates significant differences in fully denervated NMJs using two-way ANOVA with Sidak´s post hoc test. Data shown as mean±SD. N=3 animals per group. (f) Density of Iba1-positive microglial cells at lumbar levels 4 and 5 in CTRL (FUS WT-expressing control, black) and MN-P517L/Δ14 (FUS P517L/c14; ChAT-Cre, green) animals. (g) Percentage of cross-sectional area that is GFAP-positive at lumbar levels 4 and 5 in CTRL (FUS WT-expressing control, black) and MN-P517L/Δ14 (FUS P517L/c14; ChAT-Cre, green) animals. (h) Normalized grip strength of forelimbs (left) and hindlimbs (right) in CTRL (FUS WT-expressing control, black) and MN-P517L/Δ14 (P517L/c14; ChAT-Cre, green) animals normalized to the controls. *P<0.05, **P<0.01 and ***P<0.001 using two-way ANOVA with Sidak´s post hoc test. Data shown as mean±SD. N=5–28 per group at different time points. For a-d, *P<0.05, **P<0.01 and ***P<0.001 using two-way ANOVA with Tukey´s post hoc test. Data shown as mean±SD. N=3 animals per genotype. For f and g, *P<0.05, **P<0.01 and ***P<0.001 using two-way ANOVA with Sidak´s post hoc test. Data shown as mean±SD. N=3 animals per group. CTRL=littermate c14/WT and/or c14/c14 animals depending on the breeding scheme.

The conditional expression of Δ14 mutant FUS in MNs showed that the dosage of mutant protein determines the age of onset of MN degeneration in vivo. These experiments also revealed the cell-autonomous nature of this dose-dependent toxicity. The timing of microgliosis and subsequent MN loss in Δ14/WT animals was identical to that of c14/WT; ChAT-Cre mutants. The same was true in Δ14/Δ14, Δ14/c14; ChAT-Cre and c14/c14; ChAT-Cre mice, indicating that the MN degenerative phenotype is determined by the level of mutant Δ14 FUS expression specifically in MNs, without measurable contribution of mutant FUS expression in other cell types. Furthermore, these data suggest that microgliosis—and other possible, cell-non-autonomous contributors to MN degeneration—is a pathological response to mutant FUS-dependent MN pathology, as it occurs independently of mutant FUS expression in the responding microglia (Extended Data Fig. 5a–d).

Because our previous experiments suggested that mutant P517L is more pathogenic than Δ14, we suspected that co-expression of these two mutant alleles in MNs would further accelerate disease. We, therefore, generated P517L/c14; ChAT-Cre (‘MN-P517L/Δ14’) conditional compound heterozygote mutants that express both P517L and Δ14 mutant FUS only in MNs, on an otherwise P517L- and WT-expressing background. Immunofluorescent staining of spinal cords confirmed that, in MN-P517L/Δ14 mice, MNs express P517L and Δ14 (but not WT) FUS, whereas non-cholinergic cells express P517L and WT (but not Δ14) FUS (Fig. ​(Fig.3d).3d). As predicted, addition of the more pathogenic P517L allele led to NMJ denervation, microgliosis and astrogliosis at 3 months of age and MN degeneration at 4 months in MN-P517L/Δ14 mice (Fig. 3e,f and Extended Data Fig. 5f,g)—earlier than the corresponding changes in MN-Δ14/Δ14 animals (Extended Data Fig. 5b,d).

Despite this earlier onset, MN loss in MN-P517L/Δ14 mutants was limited to the same selectively vulnerable pools as in the heterozygous P517L/WT and Δ14/WT mice. The vulnerable TA muscle showed NMJ denervation, whereas the soleus muscle remained unaffected (Fig. ​(Fig.3e3e and Extended Data Fig. ​Fig.5e).5e). Here again, MN loss was limited to the large-diameter (alpha) MNs (Fig. ​(Fig.3g)3g) and ultimately plateaued at around 20% (Fig. ​(Fig.3f),3f), approximately the same level as in heterozygous P517L/WT and Δ14/WT mice (Fig. ​(Fig.1c),1c), suggesting depletion of the same subset of vulnerable MNs. Of note, denervation of the TA recovered in time: re-innervation of NMJs (Fig. ​(Fig.3e),3e), along with functional recovery from mild deficiencies in grip strength (Extended Data Fig. ​Fig.5h),5h), occurred by 1 year of age.

Overall, the data presented in Fig. ​Fig.44 show cell-autonomous toxicity of mutant FUS—as conditional expression of the Δ14 allele selectively in MNs is sufficient for MN degeneration and gliosis. This toxicity is dose dependent. Expression of two mutant FUS alleles in MNs accelerates the onset of NMJ denervation, MN loss and gliosis, with MN-P517L/Δ14 mice showing a particularly early disease onset. The selective degeneration of the same subpopulation of ALS-vulnerable MNs in MN-P517L/Δ14 mutants as in P517L/WT and Δ14/WT animals support the conclusion that MN-P517L/Δ14 mice are a disease-relevant model of early-onset ALS.

Open in a separate window

Fig. 4

Efficacy of the FUS ASO ION363 in FUS-ALS knock-in mice.

a, Immunoblot probed with anti-FUS antibodies of brain and cervical or lumbar spinal cord protein lysates of 1-month-old WT/WT and P517L/WT animals treated with NTC or anti-FUS oligonucleotide (ION363). Each band in a row represents a separate animal. b, Quantitation of total, WT and P517L FUS protein levels shown in a. Welch’s t-test was used for comparisons of NTC versus ION363 within each genotype and neuroanatomical region. Data are shown as mean±s.d. n=3 animals per group. c, Immunoblot of brain sarkosyl-insoluble fractions of 1-month-old WT/WT and P517L/WT animals treated with NTC or anti-FUS oligonucleotide (ION363). Each lane represents a separate animal. d, Quantitation of protein in sarkosyl-insoluble fractions shown in c, expressed as log-ratio of P517L/WT ION363 for FUS-P517L and TDP-43 and as log-ratio of WT/WT NTC for all others. *P<0.05, **P<0.01 and ***P<0.001, using one-way ANOVA with Tukey´s post hoc test for comparison of four groups and Welch’s t-test for two groups. n=3 animals per group. Individual values and means are shown. e, Immunostaining of lumbar spinal cord sections of a 1-month-old ION363-treated animal with anti-ASO (red) and anti-ChAT (white) antibodies showing broad distribution of ION363 to MNs and other cells and predominantly nuclear localization. Scale bar, 100µm at ×20 and 20µm at ×100. f, Numbers of ChAT-positive MNs at lumbar levels 4 and 5 in NTC-treated FUS WT-expressing CTRL (black), ION363-treated FUS WT-expressing CTRL (gray), NTC-treated MN-P517L/Δ14 (dark red) and ION363-treated MN-P517L/Δ14 (light red) animals normalized to the NTC-treated FUS WT-expressing CTRL. g, Percentage of fully innervated NMJs in the TA muscles in NTC-treated FUS WT-expressing CTRL (black), ION363-treated FUS WT-expressing CTRL (gray), NTC-treated MN-P517L/Δ14 (dark red) and ION363-treated MN-P517L/Δ14 (light red) animals. h, Density of Iba1-positive microglial cells at lumbar levels 4 and 5 in NTC-treated FUS WT-expressing CTRL (black), ION363-treated FUS WT-expressing CTRL (gray), NTC-treated MN-P517L/Δ14 (dark red) and ION363-treated MN-P517L/Δ14 (light red) animals. For f–h, *P<0.05, **P<0.01 and ***P<0.001, using two-way ANOVA with Tukey’s post hoc test. Data are shown as mean±s.d. n=3 animals per group.

Source data

Antisense silencing of FUS in mice prevents MN degeneration

Our studies of the consequences of mutant FUS in this series of knock-in mice show a dose-dependent toxicity of mutant FUS that results in progressive degeneration of a select subpopulation of spinal MNs. This, together with our previous study using a conditional FUS knockout allele that showed that postnatal FUS deficiency in vivo does not result in MN degeneration¹², suggests that lowering levels of FUS protein in vulnerable neurons could be an effective strategy to treat patients with ALS-FUS. To test this idea, we used ION363, an ASO directed against the 6th intron of the FUS transcript that was developed by Ionis Pharmaceuticals to silence FUS in a non-allele-specific manner. ION363 was chosen from a series of candidate antisense therapeutics because the target intronic sequence is conserved between mouse and human, which would allow us to translate our findings in the FUS knock-in animals to clinical studies of patients with ALS-FUS.

First, to evaluate the efficacy of ION363 in lowering FUS levels in vivo in the mouse, we administered a single 20-μg dose of ION363 via ICV injection to newborn (postnatal day 1 (P1)) wild-type and P517/WT animals. Control animals received an identical injection of a non-targeted control (NTC) ASO. Western immunoblot analysis of brain and spinal cord of 1-month-old ASO-treated animals showed an overall decrease in both wild-type and mutant FUS protein to approximately 20–50% of the levels observed in matched controls (Fig. 4a,b).

We next analyzed whether detergent insolubility of RBPs, which is the biochemical signature associated with FUS toxicity (Fig. 2c–f), is affected by ION363 treatment. Indeed, a decrease in the levels of TDP-43, hnRNP A1, hnRNP U and UPF1 was observed in the detergent-insoluble fractions from ION363-treated versus NTC P517L/WT mice (Fig. 4c,d). Moreover, the ION363-induced decrease in wild-type FUS protein in wild-type mice also caused decreased partitioning of UPF1 and hnRNP A1 into the detergent-insoluble fractions (Fig. 4c,d), suggesting that wild-type FUS-dependent partitioning of some RBPs into detergent-insoluble subcellular compartments at a low level might reflect normal physiology.

Immunostaining of spinal cords with an antibody²⁹ to the modified backbone of the ION363 and NTC ASOs showed robust predominantly nuclear signal in all cells, including MNs (Fig. ​(Fig.4e4e and Extended Data Fig. 6a,b) at 1 and 4 months of age. However, by 6 months of age, analysis of anti-ASO immunostaining showed near background levels of signal for both ION363 and the NTC ASO (Extended Data Fig. 6a,b). The observed decrease in ION363 levels at 6 months of age was associated with a corresponding increase in FUS levels (Extended Data Fig. 6c,d).

Open in a separate window

Extended Data Fig. 6

FUS silencing by ION363 in FUS-ALS mice.

(a) Representative immunostaining of lumbar spinal cord sections from NTC-treated animals with an anti-ASO antibody. The staining shows widespread nuclear and cytoplasmic distribution at 1 and 4 months of age that disappears at 6 months. Scale bar=100µm. (b) Representative immunostaining of lumbar spinal cord sections from FUS ASO-treated animals with an anti-ASO antibody. The staining shows widespread nuclear distribution at 1 and 4 months of age that disappears at 6 months. Scale bar=100µm. (c) Immunoblot of lumbar spinal cord protein from 6-month-old WT/WT and P517L/WT animals treated with non-targeted control (NTC) or anti-FUS oligonucleotide (ION363). Each lane represents a separate animal. (d) Quantitation of FUS protein levels in lumbar spinal cord lysates from 6-month-old animals shown in (c). One-way ANOVA was performed to assess the differences in FUS levels. N=3 animals per group.

Source data

The model of disease that we propose for ALS-FUS based on our animal studies predicts that the decrease in FUS protein levels—and associated decrease in detergent insolubility of FUS and related RBPs—caused by ION363 would mitigate the MN phenotype in the FUS mutant mice. To test this prediction, we injected newborn (P1) MN-P517L/Δ14 and wild-type mice with a single ICV dose of 20 μg of ION363 or the NTC ASO and quantified MN loss in all experimental animals at 4 and 6 months of age. At 4 months, the NTC-injected MN-P517L/Δ14 mice lost an average of 12% of all lumbar MNs (consistent with untreated MN-P517L/Δ14 mice; Fig. ​Fig.3f),3f), but no MN loss was observed in ION363-treated MN-P517L/Δ14 mice (Fig. ​(Fig.4f).4f). Similarly, injections of ION363, but not NTC, completely prevented muscle denervation and microgliosis, both of which precede MN loss, in the MN-P517L/Δ14 mutants at 4 months (Fig. 4g,h). At 6 months of age, when ION363 concentration decreases and FUS levels normalize (Extended Data Fig. ​Fig.6d),6d), muscle denervation and microgliosis increased in ION363-treated MN-P517L/Δ14 mice to levels intermediate between those of NTC-injected MN-P517L/Δ14 and wild-type mice, suggesting the delayed onset of the disease pathology in this model. Nevertheless, no MN degeneration was observed in ION363-treated MN-P517L/Δ14 mice at 6 months after treatment (Fig. ​(Fig.4f4f).

These pre-clinical data show that a single perinatal injection of ION363 results in a robust and persistent decrease in FUS protein levels for more than 4 months, and that, while the ASO perdures in the spinal cord, the onset of gliosis is delayed and MN loss is prevented, showing the therapeutic potential of FUS silencing—and of this specific FUS-ASO—in the treatment of patients with ALS-FUS.

ION363 (Jacifusen) drug

ION363 is a sodium salt of an oligonucleotide with a chemically modified backbone to improve its stability³³. Its full chemical name is O2′-(2-methoxyethyl)-P-thio-guanylyl-(3′->5′)-O2′-(2-methoxyethyl)-5-methyl-cytidylyl-(3′->5′)-O2′-(2-methoxyethyl)-adenylyl-(3′->5′)-O2′-(2-methoxyethyl)-adenylyl-(3′->5′)-O2′-(2-methoxyethyl)-5-methyl-uridylyl-(3′->5′)-O2′-(2-deoxy)-P-Thio-guanylyl-(3′->5′)-O2′-(2-deoxy)-P-Thio-thymidinyl-(3′->5′)-O2′-(2-deoxy)-5-Methyl-P-Thio-cytidylyl-(3′->5′)-O2′-(2-deoxy)-P-Thio-adenylyl-(3′->5′)-O2′-(2-deoxy)-5-Methyl-P-Thio-cytidylyl-(3′->5′)-O2′-(2-deoxy)-5-Methyl-P-Thio-cytidylyl-(3′->5′)-O2′-(2-deoxy)-P-Thio-thymidinyl-(3′->5′)-O2′-(2-deoxy)-P-Thio-thymidinyl-(3′->5′)-O2′-(2-deoxy)-P-Thio-thymidinyl-(3′->5′)-O2′-(2-deoxy)-5-Methyl-P-Thio-cytidylyl-(3′->5′)-O2′-(2-methoxyethyl)-adenylyl-(3′->5′)-O2′-(2-methoxyethyl)-5-methyl-uridylyl-(3′->5′)-O2′-(2-methoxyethyl)-P-Thio-adenylyl-(3′->5′)-O2′-(2-methoxyethyl)-5-methyl-P-Thio-cytidylyl-(3′->5′)-O2′-(2-methoxyethyl)-5-methyl-cytidine sodium salt. The sequence of ION363 corresponds to the 5′-GCAATGTCACCTTTCATACC-3′ sequence within the sixth intron of FUS gene. The molecular weight of ION363 is 7,477.4Da.

ION363 was manufactured by ChemGenes Corporation according to Good Manufacturing Practice standards. For the animal experiments, drug powder was reconstituted in sterile PBS to a total concentration of 4mg ml⁻¹. For the administration to the patient, ION363 was reconstituted in a USP compendial grade artificial cerebrospinal fluid to a total concentration of 11mg ml⁻¹. The appropriate drug volume for 20-mg, 40-mg and 60-mg doses was subsequently sterilized through a 0.2-µm filter into a sterile pyrogen-free syringe.

Mouse lines and procedures

All experiments involving live animals were approved by the Institutional Animal Care and Use Committee at Columbia University Irving Medical Center. C57Bl/6J (000664), ChAT-Cre (006410)²⁸, ChAT-CreΔneo (031661), protamine-Cre (003328)³⁴ and Pgk1-flpo (011065)³⁵ mouse lines were obtained from Jackson Laboratory. All strains were backcrossed to C57Bl/6J background for at least five generations. All animals were housed in a specific pathogen-free facility with ambient temperature of 18–23°C and 40–60% humidity and a 12-h light/dark cycle.

Mutant FUS knock-in mice were generated by targeting of the endogenous mouse FUS locus by homologous recombination. The targeting construct included a ~6.4 kb of C57BL/6 mouse strain genomic FUS sequence spanning exons 11–15, 3′ untranslated region (UTR) and downstream sequence. FUS genomic sequence was subcloned from BAC RP24-297F14 by gap repair recombine, and its border sequences (inclusive) were GGTCACCTCAAATAGTGAGTTTCATG on the 5′ end and CTTGTAGCTCAATTGGGTTGAAATA on the 3′ end. FRT-site flanked neomycin resistance cassette (NEO) was inserted 50bp downstream of the end of the 3′ UTR for positive selection of drug-resistant embryonic stem cells. For P517L point mutants, single base pair substitution was introduced in FUS exon 15. In the conditional FUS C14 construct, exon 14 was flanked by LoxP sites, and part of mouse exon 15 was replaced with human exon 15 sequence to accurately reproduce the novel human C-terminal Δ14 protein sequence when exon 14 is skipped and exon 15 is read out of frame in the FUS Δ14 mouse. Targeting constructs were electroporated into GFP-expressing LB10 mouse embryonic stem cells derived from a C57BL6/N mouse (GlobalStem, GSC-5003), and recombinants were detected by Southern blot.

FUS^KO (EUCOMM allele FUS^tm1d)¹² and Δ14 mouse lines were generated by first crossing FUS^FLOX (EUCOMM allele FUS^tm1c) or FUS C14 alleles to protamine-Cre line and subsequently crossing FUS^FLOX/WT or C14/WT; protamine-Cre-positive males to C57Bl/6J females.

All primer sequences used for genotyping are listed in Supplementary Table 1.

Motor function evaluation was performed using the grip strength test (Bioseb, BIO-GS3).

For ICV injections, newborn animals were first anesthetized on ice, and then 20μg of ION363 or NTC oligonucleotide in a total volume of 5μl was injected into the right lateral ventricle (2mm frontal and 1mm lateral to bregma) using a Hamilton Neuros syringe. After the injection, animals were placed on a heating pad and monitored until recovery.

To collect fixed tissue samples, the animals were transcardially perfused with PBS-heparin solution followed by 4% paraformaldehyde; brain, spinal cord and muscles (diaphragm, TA and soleus) were dissected.

PCR and RT–PCR

RNA from brain or spinal cord samples was isolated with TRIzol reagent (Invitrogen); its concentration was measured using NanoDrop 2000 (Thermo Fisher Scientific); and cDNA was generated with SuperScript IV reverse transcriptase (Invitrogen) according to the manufacturer’s protocol. GoTaq G2 Hot Start Master Mix (Promega) and PowerUp SYBR Green Master Mix (Thermo Fisher Scientific) were used to set up PCR or qPCR reactions, respectively. All primers used in this study are listed in Supplementary Table 1.

Immunofluorescence

Spinal cord samples were sectioned at 70-µm thickness on a Leica VT 1000S vibratome. TA and soleus muscles were sectioned at 35-µm thickness on a Leica CM 3050S cryostat. Diaphragms were stained as whole mount.

Tissue sections were incubated in primary antibodies (Supplementary Table 2) diluted in 5% normal donkey serum in Tris-buffered saline with 0.5% Triton X-100 (TBS-T) overnight at 4°C. The sections were then washed in TBS-T three times and incubated with the corresponding secondary antibodies (Supplementary Table 3) for 1h. After three washes in TBS-T, the sections were mounted in aqueous medium (Fluoromount G, Southern Biotech) and imaged using a Leica SP8 confocal microscope. Images were analyzed using LAS X and ImageJ software packages.

Western blotting

Tissue lysates were mixed with 4× sodium dodecyl sulfate (SDS) sample buffer, incubated at 100°C for 5min and run on Novex WedgeWell Tris-Glycine gels (4–20%, 8–16% or 10–20%) and transferred onto nitrocellulose membranes (Thermo Fisher Scientific) using mini-blot modules (Thermo Fisher Scientific). The membranes were then blocked in Pierce Protein-Free T20 (TBS) Blocking Buffer (Thermo Fisher Scientific) and incubated with the corresponding primary antibodies (Supplementary Table 2) overnight at 4°C. Then, the membranes were washed in TBS with 0.1% Tween 20 three times and incubated with HRP-conjugated secondary antibodies (Supplementary Table 3) for 1h. After three washes in TBS-Tween 20, the membranes were treated with SuperSignal West Pico or SuperSignal West Femto (Thermo Fisher Scientific) and imaged using an iBright CL1000 chemiluminiscence imaging system (Thermo Fisher Scientific).

Nucleo-cytoplasmic fractionation

This protocol was adapted from Shiihashi et al.³⁶. Frozen tissue samples were weighed and homogenized in 10μl per 1mg of tissue of cytoplasmic buffer (10mM HEPES, 10mM NaCl, 1mM KH₂PO₄, 5mM NaHCO₃, 5mM EDTA, 1mM CaCl₂, 0.5mM MgCl₂). After incubating the homogenized tissue on ice for 10min, 0.5μl mg⁻¹ of 2.5M sucrose was added and centrifuged at 6,300g for 10min. The supernatant was collected as the cytoplasmic fraction, and the pellet was washed three times with the nuclear buffer (10mM Tris pH 7.4, 300mM sucrose, 1mM EDTA, 0.1% Nonidet-40), homogenized and centrifuged at 1,000g for 5min. The pellet was resuspended in 5μl per 1mg of tissue of immunoprecipitation buffer containing 2% SDS as the nuclear fraction.

Subcellular fractionation with sarkosyl

This protocol was adapted from Conlon et al.³⁷. Tissue samples were weighed and, for every milligram of tissue, 18μl of ice-cold soluble buffer (0.1M MES pH 7, 1mM EDTA, 0.5mM MgSO₄, 1M sucrose) containing 50mM N-ethylmaleimide (NEM), 1mM NaF, 1mM Na₃VO₄, 1mM PMSF and Halt protease and phosphatase inhibitor cocktail (Thermo Fisher Scientific) was added. Samples were homogenized by several passages through needles with decreasing diameter (20-gauge, 23-gauge and 27.5-gauge needles). An equivalent volume of the homogenate (700μl) was collected and centrifuged at 50,000g for 20min at 4°C. The supernatant was collected as a soluble fraction. The pellet was resuspended in 700μl of RAB buffer (100mM MES pH 6.8, 10% sucrose, 2mM EGTA, 0.5mM MgSO₄, 500mM NaCl, 1mM MgCl₂, 10mM NaH₂PO₄, 20mM Na, 1mM PMSF, 50mM NEM) containing 1% of N-lauroylsarcosine (sarkosyl) and Halt protease and phosphatase inhibitor cocktail (Thermo Fisher Scientific), vortexed for 1min at room temperature and incubated at 4°C overnight with end-over-end rotation. The samples were then centrifuged at 200,000g for 30min at 12°C, and the supernatant was collected as the sarkosyl-soluble fraction. The pellet was resuspended in 700μl of RAB buffer and passed 3–5 times through a 27.5-gauge needle to fully disperse the pellet, creating the sarkosyl-insoluble fraction.

Immunohistochemistry

Five-micron tissue sections were deparaffinized in xylene and rehydrated in ethanol solutions with decreasing concentrations. Antigen retrieval was performed by autoclaving the slides in Tris-EDTA buffer (10mM Tris pH 9, 1mM EDTA, 0.05% Tween 20) for 10min. The slides were then washed in TBS and incubated with primary antibodies (Supplementary Table 2) in TBS solution with 0.01% Triton X-100 and 5% normal goat serum overnight at 4°C. The next day, the slides were washed three times in TBS and incubated with HRP-conjugated secondary antibodies (Supplementary Table 3). After three washes in TBS, the slides were developed with DAB Substrate Kit (Thermo Fisher Scientific) for 10min and counterstained with Mayer’s hematoxylin for 2min. The slides were then immediately rinsed in tap water for 5min, dehydrated in ethanol solutions with increasing concentrations followed by xylenes and mounted with Organo/Limonene Mount (Millipore Sigma).

Statistics and reproducibility

Statistical analysis was performed using the GraphPad Prism 9.0.2 software package. All experiments were performed in at least three technical replicates and using at least three animals per group where applicable.

This work was supported by the National Institute of Neurological Disorders and Stroke (grant no. 5R01NS106236 to N.A.S.) and the Tow Foundation. Support for the FUS ASO (ION363) expanded access program was provided by Project ALS and the ALS Association. Additional funding was generously provided by Nancy Perlman and Tom Klingenstein and from the Judith and Jean Pape Adams Charitable Foundation. J. Chen provided expert technical assistance. We thank S. Brenner-Morton for generating FUS mutant-specific antibodies; A. Nemes for help with mouse embryonic stem cell targeting; and E. Conlon for technical expertise and guidance with biochemical fractionation. We thank F. Bennett and Ionis Pharmaceuticals for providing access to ION363 and for their support in moving this experimental therapeutic into the clinic. All data presented in the manuscript were generated independently of Ionis Pharmaceuticals. Special thanks to R. Smiley, S. Aziz-Zaman, S. Messina and other members of the Columbia University ALS Center research staff for the management and execution of the FUS ASO expanded access program. Thanks to R. Brown, J. Watts, T. Yu, L. Black, D. Walk, V. Estess and E. Fleming from Project ALS for their many contributions to this effort. We are grateful to S. Przedborski, L. Pellizzoni, H. Phatnani, E. Lowry, P. St. George Hyslop and R. Axel for their comments on the manuscript and for helpful discussions. This work is dedicated to the memory of Jaci and Alex Hermstad.