RNA dependent suppression of C9orf72 ALS/FTD associated neurodegeneration by Matrin-3

The most common genetic cause of amyotrophic lateral sclerosis (ALS) is a GGGGCC (G4C2) hexanucleotide repeat expansions in first intron of the C9orf72 gene. The accumulation of repetitive RNA sequences can mediate toxicity potentially through the formation of intranuclear RNA foci that sequester key RNA-binding proteins (RBPs), and non-ATG mediated translation into toxic dipeptide protein repeats. However, the contribution of RBP sequestration to the mechanisms underlying RNA-mediated toxicity remain unknown. Here we show that the ALS-associated RNA-binding protein, Matrin-3 (MATR3), colocalizes with G4C2 RNA foci in patient tissues as well as iPSC-derived motor neurons harboring the C9orf72 mutation. Hyperexpansion of C9 repeats perturbed subcellular distribution and levels of endogenous MATR3 in C9-ALS patient-derived motor neurons. Interestingly, we observed that ectopic expression of human MATR3 strongly mitigates G4C2-mediated neurodegeneration in vivo. MATR3-mediated suppression of C9 toxicity was dependent on the RNA-binding domain of MATR3. Importantly, we found that expression of MATR3 reduced the levels of RAN-translation products in mammalian cells in an RNA-dependent manner. Finally, we have shown that knocking down endogenous MATR3 in C9-ALS patient-derived iPSC neurons decreased the presence of G4C2 RNA foci in the nucleus. Overall, these studies suggest that MATR3 genetically modifies the neuropathological and the pathobiology of C9orf72 ALS through modulating the RNA foci and RAN translation.


Introduction
Amyotrophic lateral sclerosis (ALS) is a neurodegenerative disorder that is characterized by degeneration of upper and lower motor neurons, leading to progressive atrophy and weakness in skeletal muscles, eventually resulting in death due to respiratory failure [1,2]. An expansion of hexanucleotide GGG GCC (G4C2) repeat within the first intron of C9orf72 gene is the most common cause of familial ALS as well as frontotemporal dementia (FTD), commonly termed as C9-ALS/FTD [3,4]. While unaffected individuals carry less than 10 G4C2 repeats (3,4), C9-ALS/FTD patients carry anywhere between 30 and 1600 repeats [3,5].
Accumulation of repeat-bearing RNA in the neurons and glia is also observed in other neurological disorders caused by intronic repeat expansions, such as myotonic dystrophy and fragile X tremor ataxia syndrome, are also characterized by RNA accumulation or RNA foci [19,20]. The accumulation of expanded repeat-containing RNA transcripts, including G4C2 RNA foci, sequesters essential RNA-binding proteins (RBPs) [11,[21][22][23], thus leading to a potential loss-of-function of these proteins and, consequently, dysregulation of RNA metabolism [11,12]. Thus, identification of these RBPs might be critical for understanding the mechanism of C9orf72-mediated neurodegeneration, and possibly serve as therapeutic targets. Matrin-3 (MATR3), an ALS-associated RBP, has been shown to interact with the G4C2 repeat RNA in vitro [24,25] suggesting a potential functional relationship between them. MATR3 is an essential nuclear matrix protein that, in addition to maintaining the fibrogranular nuclear matrix network, is involved in regulating post-transcriptional RNA processing, including alternative splicing, mRNA stability and mRNA export [26][27][28][29]. Moreover, pathogenic mutations in MATR3 have been discovered in a subset of familial and sporadic ALS cases [30][31][32][33], and MATR3-positive cytoplasmic inclusions are widely found in post-mortem brain tissue of C9-ALS and sporadic ALS patients [30,34]. However, the disease-relevance and the molecular mechanisms underlying these observations is yet unclear.
Here, we demonstrate that MATR3 is an important component of C9orf72-associated disease neuropathology in C9-ALS patient postmortem brain tissues and iPSC motor neurons. We show colocalization between MATR3 and pathogenic G4C2 foci in C9-ALS patientderived neurons and in post-mortem brain tissues. Additionally, we found that MATR3 sub-cellular localization and levels are altered in C9-ALS patient neurons. Furthermore, we show that ectopic expression of MATR3 strongly suppressed G4C2-HRE toxicity in Drosophila, mediated by its RNA-binding domain. In addition, genetic manipulation of MATR3 levels modulated the RAN-translation product levels and RNA foci in mammalian cells and C9-ALS patient-derived motor neurons, respectively. Collectively, our data presents evidence that MATR3 is a novel genetic modifier of neuropathologies and motor dysfunctions associated with G4C2 repeats through modulating the RNA foci and RAN translation products.

Pathologic G4C2 RNA foci sequester MATR3 protein in C9-ALS patient neurons
Proteomic studies have suggested physical interaction between G4C2 RNA and MATR3 [24,25], however, the functional significance of this interaction in unclear. This led us to investigate relationship between MATR3 and G4C2 foci localization in neurons. To that end, we utilized C9-ALS patient-derived iPSCs that were further differentiated into motor neurons, hereby referred to as C9 iPSC-MN. Using fluorescent in situ hybridization (FISH) coupled with immunocytochemistry, we analyzed the nuclei of these cells and observed colocalization between G4C2 RNA foci and MATR3 in two independent C9-ALS patient-derived iPSC-MNs (Fig. 1a) (Additional file 1: Figure S1A). We also found that the apparent colocalization between G4C2 foci occurred with both punctate ( Fig. 1a) and diffuse (Additional file 2: Figure S2A) forms of MATR3 within the nuclei of C9 iPSC-MNs. Quantitative analysis revealed that about one-third of (See figure on next page.) Fig. 1 MATR3 colocalizes with pathogenic G4C2 RNA foci in C9-ALS iPSC-derived neurons and in post-mortem brain tissue. a Representative confocal image of colocalization (white arrows) between G4C2 RNA foci (green) and MATR3 protein (red) in C9-ALS patient-derived iPSCs differentiated into motor neurons (iPSC-MN), indicated by MAP2 staining (grey). Dotted-white box represents a single G4C2 foci that colocalized with MATR3 and represented in high magnification images to the right. b Quantification of percentage neurons that are positive for G4C2 foci (green bars) in two independent C9-ALS patient iPSC-MNs. Average % neurons that are positive for G4C2 foci in C9-ALS #1 = 58% and in C9-ALS #2 = 48%. Among them, over half of the neurons also showed co-localization between MATR3 and G4C2 RNA foci (red bars). n = 3 differentiations. c Quantification of percentage G4C2 foci per neuron that colocalized with MATR3. Average % G4C2 foci that colocalized with MATR3 in C9-ALS #1 = 75% and in C9-ALS #2 = 80%. n = 3 differentiations. d Representative images of immunohistochemical analysis of MATR3 signal (green) in motor cortex neurons from and C9-ALS patient tissue. RNA FISH analysis performed to examine co-localization between G4C2 RNA foci (red) and MATR3 (green) in nuclei of C9-ALS patient tissue cells. Maximum intensity projection (left) and single plane (right) representative images are shown. Inset within image on left represents overlay between Dapi (blue) and G4C2 foci (red). e Moderate (yellow) to strong (green) Pearson's coefficients indicate co-localization between RNA foci and MATR3 signal. n = 70 G4C2 RNA foci from 3 C9-ALS patient tissues. f Diagrammatic representation of (G4C2) × 10 pull-down assay. Nuclear lysates from HEK293T cells transiently transfected with either FLAG-MATR3 or FLAG-MATR3-ΔRRM2 were incubated with biotinylated G4C2 10 RNA. The protein-G4C2 RNA complex was pulled down with streptavidin and then separated by SDS-PAGE. g Immunoblot of biotin-G4C2 10 pull-down fraction probed for FLAG-MATR3 showed physical interaction between MATR3 and G4C2 RNA. The interaction was moderately diminished between MATR3-ΔRRM2 and G4C2 RNA. Error bars indicate S.E.M neurons exhibited colocalization between G4C2 RNA foci and MATR3 (Fig. 1b), within which close to 80% of foci colocalized with MATR3 (Fig. 1c). Conversely, across two C9-ALS iPSC MNs, 14% and 61% of MATR3 puncta, respectively, colocalized with G4C2 RNA, while a significant portion of MATR3 puncta occurred independent of G4C2 foci (Additional file 2: Figure S2B). To further validate G4C2 RNA and MATR3 colocalization in end-stage disease condition, we analyzed motor cortex sections of post-mortem brain tissues and found colocalization between G4C2 RNA and MATR3 in the neurons in C9-ALS patient brains (Fig. 1d, e). The neurons in the post-mortem brain tissue from healthy controls remained devoid of any G4C2-positive foci (Additional file 1: Figure S1B).
To confirm that colocalization is due to physical interaction between MATR3 and G4C2 RNA, we performed a biotinylated-G4C2 RNA pull down assay with nuclear lysates from HEK293T cells transiently expressing MATR3 (Fig. 1f ). Consistent with previous reports, MATR3 wildtype was successfully pulled down with biotinylated (G4C2) 10 , indicating that there is physical interaction between G4C2 RNA and MATR3 (Fig. 1g). Interestingly, the interaction was diminished by RNAbinding deficient variant, MATR3-ΔRRM2, that has a truncation in the primary RNA-binding domain of the protein (Fig. 1g). This suggests that MATR3 binds to G4C2 RNA through its RNA-binding domain and is subsequently sequestered into pathogenic G4C2 foci found in patient-derived tissues and iPSC-MNs.

MATR3 levels and subcellular localization altered in C9-ALS patient-derived iPSC-neurons
To investigate overall sub-cellular distribution of MATR3, we performed immunocytochemistry for endogenous MATR3 in C9-ALS iPSC-MNs (Fig. 2a). We observed a marked decrease in nuclear MATR3 immunoreactivity in the C9orf72 iPSC-MNs compared to that in control iPSC-MNs (Fig. 2a, b). Complementary to protein immunoreactivity levels, we also observed a concurrent decrease in MATR3 mRNA levels in C9-ALS iPSC-MNs compared to controls (avg. fold change = 0.47 for C9 #1 and 0.43 for C9 #2) (Fig. 2c). To verify that this phenomenon is not a consequence of global change in levels of nuclear proteins, we immunostained C9-ALS patient iPSC-MNs for Histone H3 and found that C9-ALS iPSC MNs exhibited no significant difference in nuclear Histone H3 levels (Additional file 3: Figure S3A, B). Concurrently, we determined that the mRNA levels of HIST1H3D, which encodes one of the Histone H3 variants, and found that mRNA levels of Histone H3 were also not altered in C9-ALS iPSC MNs compared to control iPSC-MNs (Additional file 3: Figure S3C). Overall, our results from patient post-mortem brains and iPSC-MNs suggest that disruption in levels and localization of MATR3 is specific to C9-ALS.
Next, we used immunohistochemistry to determine if the subcellular localization of MATR3 is affected in C9-ALS patient post-mortem brain tissue. In control non-ALS cases, MATR3 was predominantly nuclear in entorhinal cortex sections, with sparse cytoplasmic signal detected in a small percentage of cells (Fig. 2d). On the other hand, examination of C9-ALS patient postmortem brains revealed a significantly high percentage of cells with cytoplasmic immunoreactivity for MATR3 compared to that in control subjects (Fig. 2d, e). Interestingly, we found that in sporadic ALS patients, who don't carry the hexanucleotide repeat expansion mutation in C9orf72, MATR3 staining in the cytoplasm more was comparable to that observed in control tissues (Fig. 2d, e). This suggests that increased cytoplasmic MATR3 pathology in post-mortem brains might be exclusive to C9-ALS.

MATR3 is a strong suppressor of C9orf72 G4C2 HRE-mediated neurodegeneration in vivo
Our results so far indicated that G4C2-HRE in C9orf72 not only physically associates with MATR3, but also significantly impacts the levels and sub-cellular localization of MATR3. So, we sought to determine whether MATR3 and G4C2 HRE genetically interact in vivo in Drosophila. We utilized previously published Drosophila models of G4C2-HRE toxicity, particularly transgenic models expressing 3 repeats (3R), 30 repeats (30R), 36 repeats (36R) and 58 repeats (58R) of G4C2 [35][36][37]. When expressed in the eye, these models exhibit varying degrees of eye degeneration characterized by depigmentation, ommatidial fusion and/or loss, bristle disorganization, and in the case of 36R, necrotic patches (Fig. 3a). We found that ectopic expression of human transgenic MATR3 wildtype in these fly eyes significantly ameliorated repeat-induced eye degeneration, including amelioration of the necrotic patches caused by 36R (Fig. 3b). Phenotypic quantification of eye degeneration showed statistically significant suppression of toxicity across all G4C2-HRE models upon expression of MATR3 (Fig. 3c). Additionally, we observed that while G4C2 30R-mediated eye degeneration is exacerbated with aging to 30-days, expression of MATR3 in these flies continued to suppress the degeneration, even at later ages (Additional file 4: Figure S4A,C). Using RT-qPCR, we determined that the mRNA levels of G4C2 are not significantly changed upon expression of MATR3 (Fig. 3d), indicating that the suppression in eye phenotype is most likely not due to downstream effects of G4C2 expression.
To determine if ectopic expression of MATR3 in G4C2-HRE flies can suppress motor deficit, we assessed locomotion ability in adult flies expressing G4C2-30R in neurons. Expression of G4C2-30R alone results in severe motor degeneration in these animals, with only ~ 10% flies able to climb at all (Fig. 3f ). Ectopic expression of MATR3 in these flies significantly improved this motor function deficit (Fig. 3f ). Additionally, conditional expression of 30R in neurons markedly reduced survival in these flies, a phenotype which was also rescued by MATR3 expression (Fig. 3e). Owing to functional and structural similarity of MATR3 to other ALS-associated RNA-binding proteins with known RNA-recognition motifs, including TDP43, FUS and EWSR1, we asked if the eye degeneration of G4C2-30R could be suppressed by ectopic expression of TDP43, FUS and EWSR1. We found that expression of neither of these RBPs suppressed the eye degeneration in G4C2-30R flies (Additional file 5: Figure S5A, B). In fact, due to intrinsic eye degeneration that is induced by ectopic expression of these RBPs, particularly in the case of FUS and EWSR1, the eye degeneration compounded in the presence of G4C2-30R (Additional file 5: Figure S5A, B). Interestingly, both FUS and EWSR1 have been identified as binding partners of G4C2 RNA, suggesting that despite being a part of the shared G4C2-RBP proteome, MATR3 suppresses G4C2-HRE neurodegeneration through distinct mechanisms.

RNA-binding domain of MATR3 required to suppress G4C2 toxicity in vivo
We next sought to investigate which potential function of MATR3 is mediating suppression of G4C2-HRE toxicity. MATR3 has four known functional domains: two zinc finger motifs (ZF1 and ZF2), known to have DNA-binding and putative RNA-binding properties [29,38], and two tandem RNA recognition motifs (RRM1 and RRM2), with RRM2 domain demonstrated to exhibit more predominant RNA-binding activity ( Fig. 4a) [26]. To map the genetic interaction between G4C2 and MATR3 to the functional domains of the protein, we generated transgenic lines with each of the four functional domains deleted: ΔRRM1, ΔRRM2, ΔZF1 and ΔZF2 (Fig. 4b, Additional file 6: Figure S6A). MATR3 with truncation in either zinc-finger domains (ΔZF1 and ΔZF2) still retained its suppressive effect on G4C2-30R eye degeneration (Fig. 4c, d). On the other hand, deletion of RRM domains, particularly the RRM2 domain, had a detrimental effect in the ability of MATR3 to suppress G4C2-30R toxicity (Fig. 4c, d). This was even more evident in aged flies, where at day-30, expression of MATR3-ΔRRM2 was effective in suppressing G4C2 30R-mediated eye degeneration to the same extent as MATR3 (Additional file 4: Figure S4B). Expression of MATR3 deletion mutants alone in Drosophila eyes did not cause any external eye degeneration (Additional file 6: Figure S6B, Additional file 4: Figure S4B), confirming that the phenotypes observed in flies co-expressing G4C2-30R and MATR3-ΔRRM2 is not due to cumulative toxicities. Consistent with the eye phenotypes, in flies conditionally expressing G4C2-30R in adult neurons, ectopic expression of MATR3-ΔRRM2 was also unsuccessful in increasing their longevity (Fig. 4e). This suggests that the RNA-binding domain of MATR3 is required to suppress G4C2 HRE toxicity in vivo.

Ectopic expression of MATR3 suppresses RAN translation in mammalian cells
G4C2-HRE transcript potentially induces neurodegeneration through formation of nuclear RNA foci and through RAN-translation of the G4C2 RNA into toxic DPRs that form neuronal inclusions [6,39]. To investigate which of these pathogenic mechanisms is potentially altered by MATR3 expression, we used a reporter G4C2 construct, consisting of ~ 60 repeats (termed G4C2 60 ), with a fluorescent Dendra2 reporter in the GR-frame which is only detected if the transcript undergoes RAN translation. We co-transfected HEK293T cells with the G4C2 60 -Dendra2 reporter plasmid along with either FLAG or FLAG-MATR3 plasmid. Using RNA-FISH combined with ICC, we confirmed that overexpression of the FLAG-empty vector and G4C2 60 -Dendra2 reporter in HEK293T cells results in formation of G4C2 RNA foci in the nucleus, and importantly, that a subset of these cells (~ 59%) produced GR-Dendra2 RAN-translated product (Fig. 5a). We observed that overexpression of FLAG-MATR3 with G4C2 60 -Dendra2 significantly decreased in the percentage of G4C2-expressing cells that produced GR-Dendra2 RAN product (by ~ 58%) (Fig. 5a,  b). Ectopic expression of MATR3 had no impact on the Given that RRM2-deficient MATR3 was not able to suppress G4C2 toxicity in vivo, we investigated if RRM2 deletion results in a similar outcome in vitro. We found that overexpression of FLAG-MATR3-ΔRRM2 with MATR3 overexpression mitigates DPR production from G4C2 60 transcripts in an RNA-dependent manner. a Representative confocal images of HEK293T cells co-transfected with G4C2 60 -Dendra2 RAN translation reporter plasmid and either FLAG (empty vector), FLAG-MATR3 or FLAG-MATR3-ΔRRM2 plasmids. G4C2 60 expression resulted in formation of G4C2 RNA foci (red) and, further, the G4C2 60 transcripts also undergo RAN translation in a subset of cells to produce dipeptide repeats (DPRs), indicated by Dendra2 signal (green) that is in frame with polyGR (white arrows). b Quantification of the percentage of cells that produce GR-Dendra2 (green) RAN product showed that overexpression of FLAG-MATR3 (gray) significantly reduced RAN translation of G4C2 60 transcript to GR-Dendra2 RAN product. Overexpression of FLAG-MATR3-ΔRRM2 did not impact GR-Dendra2 production in these cells (One-way ANOVA) n = 15-18 per group. c Quantification of G4C2 60 -Dendra2 mRNA levels in cells expressing FLAG (empty vector) or FLAG-MATR3 showed no significant difference (One-way ANOVA) n = 5 per group. d Schematic of previously-published codon-optimized approach to produce dipeptide protein repeats by canonical translation of non-repeat RNA (i.e. not carrying GGG GCC repeats). e Representative images of Drosophila eyes and f quantification of external eye degeneration from flies expressing codon-optimized DPRs: GR36, PR36, GA36 and GP36. Flies expressing PR36 and GR36 developed significant external eye degeneration that is more severe is GR36 compared to PR36. Flies expressing either GA36 or GP36 did not show any external eye degeneration. Expression of MATR3 in these flies did not modify the phenotypes (Kruskal-Wallis test) n ≥ 50 per genotype. Error bars indicate S.E.M. **p value < 0.01 G4C2 60 -Dendra2 did not alter the percentage of G4C2expressing cells that produced GR-Dendra2 (Fig. 5a, b).
We also sought to test if MATR3 could suppress G4C2 toxicity through any mechanisms independent of G4C2 repeat-bearing RNA. To test this in vivo, we utilized codon-optimized dipeptide repeat (DPR) fly lines [37,40] that produce the end-product toxic peptides, however, bypassing the production of transcripts that carry the G4C2 repeat RNA (Fig. 5d). The codon-optimized expression of GR or PR dipeptide repeats (GR36 and PR36) in fly eyes results in severe degenerative phenotype (Fig. 5e, f ). We observed that ectopic expression of MATR3 in these flies did not alleviate the degenerative phenotype (Fig. 5e, f ). Additionally, codon-optimized expression of GA and PA dipeptide repeats (GA36 and PA36) does not result in any external eye degeneration in flies with or without MATR3 (Fig. 5e, f ). This suggests that ectopic expression of MATR3 does not impact posttranslational toxicity associated with C9-DPRs.

MATR3 modulates G4C2-HRE RNA foci in C9-ALS patient neurons
Collectively, our data suggests a G4C2 expanded repeat RNA-dependent genetic interaction between MATR3 and C9-ALS. To further validate this interaction in the context of C9-ALS disease neurons, we utilized siRNA (siMATR3) to significantly knockdown endogenous MATR3 in C9-ALS iPSC MNs (Additional file 7: Figure S7A, B). We observed that knocking down endogenous MATR3 resulted in a significant decrease in the percentage of neurons that exhibit G4C2 foci (Fig. 6b), and simultaneously, the number of G4C2 foci per individual neuron (Fig. 6a-c), consistently across two independent patient-derived C9-ALS iPSC MNs (Fig. 6b), confirming a disease-specific genetic interaction between MATR3 and G4C2 repeat RNA in these neurons.

Discussion
GGG GCC hexanucleotide repeat expansion in the first intron of C9orf72 gene leads to neurodegeneration through a combination of loss-of-function of endogenous C9orf72 protein, and gain-of-function of toxic repeatassociated RNA and protein aggregates [41]. The gainof-function RNA toxicity, that arises from bidirectional transcription of the repeats, can manifest in multiple ways. G4C2 and C4G2 repeat RNA form pathologic RNA aggregates, or RNA foci, within the brain and spinal cord of C9-ALS/FTD patients [11,12,[42][43][44][45]. These RNA foci directly interact with and sequester essential RNA-binding proteins and thereby compromise their normal function [23].
Colocalization and G4C2-pull down studies in C9-ALS brain tissues as well as iPSC-derived neurons have identified multiple RNA-binding proteins (RBPs) that interact with the G4C2 repeat RNA, including hnRNPfamily proteins (hnRNPA1, A2/B1, A3, K, H1, H3), ALYREF, ADARB2, RanGAP1 and MATR3 [7,24,25]. Meta-analysis of all proteins that make up the G4C2-RBP interactome has identified two independent groups that determined MATR3 to be one of the interactors of G4C2 RNA [7,24,25]. Consistent with these reports, we found that MATR3 binds to G4C2 RNA, mediated partially by its RRM2 RNA-binding domain. However, the disease relevance of this interaction was previously unknown. In this study, we identify for the first time colocalization between G4C2 RNA foci and MATR3 in C9-ALS/ FTD patient-derived brain tissue and iPSC-neurons, thus linking MATR3 to C9-ALS/FTD disease pathogenesis. RBPs have been shown to have differential binding affinities to G4C2 RNA depending on its higher-order secondary structure. For example, hnRNPH binds to both G-quadruplex and hairpin conformations of G4C2 [24], and on the other hand, SRSF1, nucleolin and Ran-GAP1 exhibit higher binding affinity to G-quadruplex [24,46]. Elucidation of differential binding affinities of MATR3 to G4C2-secondary structures are required to further characterize the nature this interaction. Interestingly, we also found that truncation of the RNA-binding domain of MATR3 does not completely abolish the interaction between MATR3 and G4C2 RNA, suggesting that there might be other factors mediating the interaction. In fact, a comparison of the protein interactome of MATR3 [47][48][49] and G4C2 RNA [7,24,25] reveals multiple common RNA-binding proteins including members of the hnRNP and DDX family of proteins as well as components of nuclear bodies including SFPQ and NONO. Thus, MATR3 could simultaneously be binding to G4C2 repeats directly and associate with other RNA-binding proteins that bind to G4C2, that multimerize to form the G4C2 foci. The results described in this study suggests protective effects of MATR3 in C9orf72-ALS by restoring RBP homeostasis.
Our data from post-mortem human brain tissues also indicate that sub-cellular localization of MATR3 is perturbed in C9-ALS neurons. MATR3 is a predominantly nuclear protein under physiological conditions, whereas, in C9-ALS patients, MATR3 exhibits significant cytoplasmic mislocalization. Disruption in MATR3 subcellular localization has previously been reported in other forms of ALS cases, including in C9-ALS and sporadic-ALS patient brain tissues [30,34]. However, contrary to previous observations in sporadic ALS patient tissues [34], we found that MATR3 localization in sporadic ALS brain tissues were comparable to that in control brains. The discrepancy between our observations could be potentially due to differences in patient demography (Caucasian patients in this study vs. Japanese patients in Tada et al. [34]) and difference in the age of death (avg. 58 years in this study vs. avg. 70 years in Tada et al. [34]). Additionally, the MATR3 antibodies used by both studies each recognize different epitopes which could also possibly explain the differences in our observations. In addition to perturbed sub-cellular localization, the levels of MATR3 are significantly reduced in C9-ALS neurons, interestingly at the transcriptomic level. Like many other RBPs, MATR3 is known to autoregulate its own levels [26,27,50]. Thus, reduced MATR3 levels in C9-ALS neurons could be a direct consequence of sequestration of MATR3 into G4C2 RNA foci. This raises the possibility that MATR3 sequestration in G4C2 foci might mimic a potential loss-of-function of endogenous MATR3, resulting in disruption of its native functions. Initially discovered as a component of the nuclear matrix, MATR3 is best known for its role in RNA metabolism, including in alternative splicing, mRNA stability and mRNA export [26-28, 48, 51, 52]. In fact, knocking down MATR3 has been reported to lead to transcriptome-wide changes in mRNA levels as well as alternative splicing events in mammalian cells [26,27]. In C9-ALS, widespread alterations of alternative splicing and alternative polyadenylation site usage has been identified in cerebellar tissues from patients [53]. Thus, depleted levels of MATR3 in C9-ALS patient neurons might be one of the mechanistic links to wide-scale splicing defects that contributes to neurodegeneration.
In addition to linking MATR3 with neuropathology in C9-ALS patient-derived neurons, we identified MATR3 as a strong suppressor of G4C2-HRE toxicity in vivo. Specifically, we found that deletion of the RNAbinding domain of MATR3 (MATR3-ΔRRM2) significantly reduced the ability of MATR3 to suppress G4C2 toxicity in vivo. Importantly, our data is consistent with a model of RNA-mediated suppression of G4C2 toxicity. Furthermore, utilizing a reporter for RAN translation, we showed that overexpression of MATR3 suppresses DPR production in human cells dependent on the RRM2 domain, suggesting that MATR3 interaction with G4C2 RNA is important for suppressing the C9-DPR toxicity. However, a caveat is that the cells overexpressing MATR3-ΔRRM2 also have endogenous wildtype MATR3 with, presumably, a fully functional RNA-binding domain. Thus, the true effects of RRM2 deletion upon DPR production would have to be further investigated in the context of MATR3 knockout or point mutations in the RNA-binding domain of endogenous MATR3 that debilitate its RNA-binding function. Nevertheless, in Drosophila models that produce DPR products through canonical translation of codon-optimized transcripts that don't bear the G4C2 repeats, MATR3 expression had no effect on the polyGR-or polyPR-mediated toxicities. This suggests that overexpression of MATR3 has no impact on post-translational toxicities conferred by DPR products in C9orf72-ALS, and that it is likely acting through upstream mechanisms which might be dependent on its RNA-binding function. Interestingly, we also demonstrated that in patient-derived neurons, modulating MATR3 levels directly impacts G4C2 RNA foci formation, raising a possibility that RNA foci and RAN translation/DPR mechanisms are linked through MATR3.
Deriving from the patient pathology showing physical-and colocalization-interaction between MATR3 and G4C2 RNA, and genetic interaction between MATR3 and G4C2, both in vivo and in C9-ALS patientderived neurons, we suggest two possible ways in which this interaction could be involved in C9-ALS pathobiology: (1) Decreased levels of nuclear MATR3 and/or cytoplasmic localization of MATR3 that could perturb its physiological functions due to alteration in the subcellular distribution; and, (2) nuclear sequestration of G4C2 RNA foci and suppression of RAN translation.
It is possible that loss of nuclear MATR3 and/or cytoplasmic localization of MATR3 in C9-ALS neurons could be promoting RAN translation of G4C2-HRE repeat RNA into toxic DPRs. We speculate that in cells expressing sufficient MATR3 or overexpressing MATR3, it interacts with G4C2 RNA to sequester the expanded repeat RNA within the nucleus, and inversely, decreases availability of G4C2-HRE RNA for RANtranslation. Therefore, MATR3 could be imparting a neuroprotective role to G4C2 RNA foci in C9-ALS. MATR3 may possibly aid in G4C2 RNA foci formation through mediating post-transcriptional modifications that increase the stability of G4C2 higher-order secondary structures and thus, serving to stabilize the foci. Our data also raises the possibility that MATR3 could be inhibiting nuclear export of G4C2 repeat RNA to the cytoplasm and this, in turn, reduces the production of RAN proteins. The benefit of inhibiting nuclear export of repeat expansion RNA has been previously explored in C9-ALS [39,54], as well as in other microsatellite expansion-mediated neurological disorders, including FXTAS [55], SCA31 [56] and myotonic dystrophy 2 [57]. Unbiased genetic screens in Drosophila models of C9-ALS, including ones used in this study, have identified that depletion of nuclear export adapters, such as ALYREF and SRFS1, is protective in these models [54,58,59]. Indeed, in patient-derived neurons, SRSF1 has been demonstrated to facilitate export of G4C2-HRE RNA, and furthermore, depletion of SRFS1 leads to an increase in the nuclear retention of G4C2 RNA, and concomitantly, a decrease in the production of RANtranslated toxic DPRs [54]. It is likely that MATR3 could be acting through the same or similar pathways to promote nuclear retention of G4C2-HRE RNA and subsequent sequestration into foci. Further investigation into the mechanisms that link MATR3-G4C2 RNA binding and culminating in RAN translation is needed to fully elucidate the potential of restoring normal levels of MATR3 and/or overexpression of MATR3 in suppressing C9-ALS neurodegeneration.

Human post-mortem brain tissues
The patients and controls were sampled from a similar age range. At the time of death, the age of control individuals (n = 3) ranged between 54 and 76 years, C9-ALS patients (n = 6) ranged between 51 and 67 years and that of non-C9 ALS patients (n = 3) ranged between 46 and 58 years. Clinical and neuropathological information of C9-ALS, non-C9 ALS and age-matched nonneurological control post-mortem human brain tissues is listed in Table 1.

Immunohistochemistry
Formalin-fixed, paraffin-embedded human entorhinal cortex sections of ALS/FTD subjects were retrieved from the Neurodegenerative Brain Bank at the University of Pittsburgh, following protocols approved by the University of Pittsburgh Committee for Oversight of Research and Clinical Training Involving Decedents (CORID).

Quantification of cytoplasmic MATR3 staining
Quantification of the MATR3 IHC immunostaining was performed in a blinded fashion. Bright-field images from multiple fields of the entorhinal cortex were used for analysis. Briefly, cells with MATR3 signal (brown/DAB) in the cytoplasm were counted and plotted as a percentage of the total number of cells in the field.

Quantification of G4C2 FISH and MATR3 colocalization
For quantification of G4C2 RNA Foci and MATR3 colocalization in human tissue, Z stack images of 12-20 nuclei per patient from three C9 ALS/FTD patient motor cortex embedded paraffin sections were acquired using a Nikon A1R confocal microscope. Images were then deconvolved using 10 Landweber iterations on the Nikon element program. The single middle plane of the reconstructed RNA foci was then selected and analyzed for Matrin3 protein signal overlap used to calculate the Pearson's colocalization in Nikon elements. G4C2 RNA:MATR33 Pearson's coefficient values from 70 RNA foci were binned in values of 10 and colocalization was classified using < − 0.79 inverse colocalization; − 0.8 to 0.19 no colocalization; 0.20 to 0.59 moderate colocalization (yellow, Fig. 1e); 0.8-1 colocalized (green, Fig. 1e).

Patient-derived induced pluripotent stem cells (iPSCs)
Clinical information and respective sources of all control (healthy) and C9-ALS iPSC lines used in this study are listed in Table 2.

siRNA treatments
Two serial siRNA treatments were applied beginning 65-70 days after initiating MN differentiation. Each treatment was applied for 24 h. The second treatment began 4 days after application of the first siRNA treatment and FISH was performed on freshly fixed cells 4 days after the application of the second siRNA treatment.

Fluorescent in situ hybridization (FISH)
FISH analysis was performed as previously described [42]. Briefly, cells were cultured on 12 mm coverslips in 24 well plates, fixed in 4% PFA, and permeabilized in 0.3% Triton for 15 min. Before hybridization, cells equilibrated in 2× SSC + 50% formamide for 10 min at 60C and probe mixture was preheated for 10 min at 95C. For hybridization, 200 mL of hybridization buffer and 27 mL of probe mixture were added to each coverslip and cells were incubated for 1 h at 60C. Following hybridization, cells were incubated with 2× SSC + 50% formamide for 20 min at 65C, then again with fresh 2× SSC/50% formamide for 15 min at 65C, and finally with 1× SSC + 40% formamide for 10 min at 60C. Cells then underwent a series of washes at RT: 3× quick washes with 1× SSC, 2 × 5 min washes with 1× SSC, a 5 min wash with Trisbuffered saline, and a 5 min wash with Tris-glycine buffer. Cells were post-fixed in 3% PFA and then incubated with blocking buffer for 1 h at RT. Cells incubated with primary antibodies diluted in immunofluorescence buffer overnight at 4C and secondary antibodies diluted in immunofluorescence buffer for 20 min at RT. After secondary antibody incubation, cells were sequentially  Immunoresearch) and 5% heat-shocked BSA in Tris-buffered saline. Immunofluorescence buffer consisted of 2% heat-shocked BSA in Tris-buffered saline. DIG-labeled probe was detected with a fluorescein-conjugated sheep anti-DIG antibody (1:250, Roche). All buffers were made with RNase-free water or PBS.

Immunocytochemistry
Motor neurons, matured for 50 days, and transiently transfected HEK293T cells at 72-h post transfection, were used for immunocytochemistry. Cells were washed and fixed with 4% paraformaldehyde and blocked for 1 h in Phosphate-buffered saline (PBS) containing 0.2% Triton X-100 and 10% normal donkey serum for iPSC-MNs and 5% normal goat serum for HEK293Ts respectively (Jackson ImmunoResearch). Samples were then incubated overnight at 4 °C with primary antibodies. The next

Quantitative image acquisition and analysis
Images used for quantification were acquired at matched exposure times or laser settings and processed using identical settings. Quantifications were normalized within each respective experiment with n = 3 independent experiments unless otherwise specified in figure legends. Image acquisition was performed on a Leica DMI4000B laser scanning confocal microscope (Leica, Buffalo Grove, IL) or a Nikon W1 dual camera spinning disk confocal microscope (Northwestern University Center for Advanced Microscopy). Unless otherwise specified, image acquisition was performed through the z-dimension at 0.3-0.5 mm intervals and individual planes were projected into maximum intensity images.

Quantification of nuclear MATR3 and nuclear Histone H3 immunoreactivity
The nuclear region of interest was defined by DAPI and the mean signal intensity of MATR3 as the mean pixel intensity per mm 2 as calculated by the NIS-Elements Advanced Research v5 software (Nikon software, Northwestern University Center for Advanced Microscopy). Fold change was determined by averaging the values from all control neurons and normalizing to this average.

Quantification of G4C2 RNA foci
FISH images were acquired in a blinded manner using a Plan Apo l 60× 1.45 NA oil immersion objective through the z dimension at 0.3 mm intervals and set to equal contrast. Images were cropped to single neuron dimensions and analyzed by individual plane for nuclear G4C2 foci in a blinded manner by the primary operator in one sitting. We defined "G4C2 foci" as signals that met the following criteria: (a) diameter > 0.8 µm (b) signal intensity > 1.5 fold higher than diffuse nuclear signal (c) punctate morphology and (d) nuclear localization.

MATR3 and G4C2 colocalization analysis
The assessment of the % of G4C2 RNA foci that colocalized with MATR3 was performed by identifying nuclear G4C2 foci as described above and tested for the presence or absence of total MATR3 signal (punctuate and diffuse). The assessment of the % of MATR3 puncta that colocalized with G4C2 RNA foci was performed by identifying nuclear MATR3 puncta (based on size and contrast relative to overall nuclear MATR3 signal intensity) and tested for the presence or absence of a G4C2 RNA focus.

Drosophila assays
The detailed list of Drosophila lines used in this study and their respective sources are outlined in Table 3. All Drosophila stocks were cultured on standard dextrose media on a 12-h light/dark cycle.

Eye degeneration quantification
Expression of transgene in Drosophila eyes was driven using the GMR-gal4 driver at 28 °C. One day old F1 adults (or 30-day old adults for aging experiments) were collected and external images of the eye were using Leica M205C microscope equipped with a Leica DFC450  [53,57,58,60].

Adult survival and motor dysfunction
For conditional pan-neuronal expression, G4C2-30R and UAS-MATR3 lines were crossed with inducible driver, ElavGS-Gal4. Day 1 adults from the F1 progeny were collected every 24 h and moved to standard media mixed with 20 mM RU486 to induce transgene expression in adults. The adults were cultured at 28 °C and death was recorded every other day. Motor function was assessed on day-30 of their lifespan. Locomotion was assessed using the RING assay, as previously described [61] with a few modifications. Briefly, flies were transferred, without anesthetization, into plastic vials and placed in the RING apparatus. The vials were tapped down against the bench and the climbing was recorded on video. Quantifications were performed manually by a third party in a blinded manner.
Biotinylated G4C2 RNA pull-down G4C2 10 oligomer biotinylated at the 5′ end was synthesized by Integrated DNA Technologies Inc. (IDT) and RNA pull-down was performed as previously described [25]. Nuclear lysates from HEK293T cells were extracted using the NE-PER nuclear cytoplasmic extraction kit (ThermoScientific 78833) following manufacturer instructions. 0.5 mg of lysates made up to 500 µL with in RNA structure buffer (10 mM Tris pH 7, 0.1 M KCl, 10 mM MgCl 2 ) were pre-cleared with 50 µL of pre-washed Dynabeads Streptavidin Magnetic Beads (Thermofisher 11205D) at 4 °C overnight on a rotary shaker. 4 µg of biotinylated RNA was heated to 90 °C for 2 min in RNA structure buffer (10 mM Tris pH 7, 0.1 M KCl, 10 mM MgCl2) and incubated at for 20 min at RT. Biotin-G4C2 10 RNA oligomer was added to the pre-cleared nuclear extract and incubated at 4 °C for 4 h. 50 µL of pre-washed Dynabeads Streptavidin Magnetic Beads was added to each reaction and incubated at 4 °C for 1 h. After incubation with the nuclear extracts, beads were washed 5 times in wash buffer (10 mM Tris-Cl, pH 7.5; 1 mM EDTA, 2 M NaCl). Co-precipitated proteins were eluted from bound biotinylated RNAs in wash buffer and boiled in 1× NuPage LDS-Sample buffer (Invitrogen NP0007) at 95 °C for 5 min. Co-precipitated proteins were detected by Western Blot. All buffers were freshly supplemented with 0.5 mM DTT, 0.1 U/mL RNase inhibitor, 0.1 µg/µL tRNA and 5 mM EDTA.

RNA preparation and RT-quantitative PCR (qPCR)
RNA was isolated using RNeasy Mini Kit (Qiagen) according to manufacturer instructions. cDNA synthesis was performed using the iScript Select cDNA Synthesis Kit (BioRad; 170-8897) and was subsequently run using the Bio-Rad iQ ™ Supermix (170-8862) on a 96-well plate (Applied Biosystems, #4306737) on Applied Biosystems StepOnePlus Real-Time PCR system. Primer pairs and probes were designed for each target of interest and housekeeping control α-tubulin (for Drosophila) and GAPDH (for human) using Integrated DNA Technologies PrimeTime qPCR Assay (www.idtdn a.com). The comparative Ct method was used for analyzing the fold change differences [62]. The list of primer-probe sequences are listed in Table 4.