Extensive splicing changes in an ALS/FTD transgenic mouse model overexpressing cytoplasmic fused in sarcoma

Mutations in RNA-binding proteins (RBPs) such as TAR DNA-binding protein 43 (TDP-43) and fused in sarcoma (FUS) are associated with amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD). Recent evidence suggests that RNA dysregulation mediated by aberrant RBPs may play a critical role in neurodegeneration, but the underlying molecular mechanisms are largely unknown. In this study, we performed whole transcriptome profiling of various brain tissues of a transgenic (Tg) mouse model of ALS/FTD overexpressing the exogenous nuclear localization signal deletion mutant of human FUS (ΔNLS-FUS) to investigate changes associated with the early stages of ALS/FTD. Although there were not many differences in expression profiles between wild-type and Tg mice, we found that Sema3g was significantly upregulated in the frontal cortex and hippocampus of Tg mice. Interestingly, analysis of alternative splicing events identified widespread exons that were differentially regulated in Tg mice in a tissue-specific manner. Our study thus identified aberrant splicing regulation mediated by mutant FUS during the early stages of ALS/FTD. Targeting this aberrant splicing regulation represents a potential therapeutic strategy for ALS/FTD.


Sema3g is upregulated in the FCx and Hp tissues of Tg mice.
To examine the underlying mechanism of FUS proteinopathy during the early stage of ALS and FTD, we first compared the transcriptome data generated from 6-month-old wild-type (Wt) and ΔNLS-FUS Tg mice, in which exogenously overexpressed mutant FUS (ΔNLS-FUS) was mislocalized in the cytoplasm (see Supplementary Fig. S1) and significant motor deficits appear 9 . We therefore performed whole transcriptome analysis of FCx, Hp, and SC tissues isolated from Wt and Tg mice (n = 4). On average, 23.27 ± 2.30 × 10 6 reads from FCx tissue, 20.93 ± 2.99 × 10 6 from Hp, and 23.88 ± 1.93 × 10 6 from SC underwent quality control checks by FastQC and trimming by Trimmomatic. Hierarchical clustering and principal component analysis (PCA) showed a clear separation of each tissue, indicating good quality of the RNA samples and reads obtained (see Supplementary Fig. S2A). Surprisingly, no definite clustering pattern was detected between Wt and Tg mice, indicating that there were no major differences in transcriptome profiles between the two groups during early ALS/FTD (see Supplementary Fig. S2B-D).
Next, we examined the cell composition of each tissue sample by analyzing genes that are specifically expressed by neurons, motor neurons, astrocytes, oligodendrocytes, and microglia (see Supplementary Fig. S3 and Supplementary dataset 1: Cell composition). No significant changes in tissue-specific cell compositions were detected between Wt and Tg mice, indicating that gliosis and neuronal loss are not prominent in 6-month-old Tg mice.
We further probed for differentially expressed genes (DEG) that show a fold change (log2) ≥ 1 with adjusted p-values (p adj) < 0.05 between Wt and Tg mice in each tissue sample (see Supplementary dataset 2-4: Differentially expressed genes_FCx, Hp or SC). As shown in Table 1, only a few genes showed significant differences in expression levels; in Tg mice, Sema3g in FCx and Hp tissues was upregulated (Tg/Wt; 2.60 and 3.64 fold, respectively) and A830039N20Rik in FCx tissue was downregulated, while four genes in SC tissue were upregulated and another four (Hbb-b2, Hba-a2, Hba-a1, and Hbb-b1) downregulated (see Supplementary Table S1). We further analyzed Sema3g, Hbb-b2, Hba-a2, and Alas2 expression levels in four Tg and Wt mice by qRT-PCR and confirmed that Sema3g expression is significantly increased in FCx and Hp tissues of Tg mice (Tg/Wt; 5.56 and 6.37 fold, respectively; Fig. 1), but the decrease in expression levels observed for Hbb-b2, Hba-a2, and Alas2 in SC tissues was not significant (p = 0.105, p = 0.113, and p = 0.321, respectively).
Tg mice exhibit widespread differential alternative splicing events. To further explore the underlying mechanism of FUS proteinopathy, we investigated the differential alternative splicing events between Wt and Tg mice using the transcriptome data generated from the tissue samples. Analysis of alternative splicing events revealed that 788, 584, and 651 exons in the FCx, Hp, and SC tissue samples, respectively, were differentially regulated in Tg mice; we used a statistical threshold that captured as many significant changes as possible (FDR < 0.05; Fig. 2a, also see Supplementary dataset 5-7; Splicing changes_FCx, Hp, or SC). When we categorized the observed changes among the alternative splicing subtypes, namely alternative 5′ and 3′ splice sites, mutually exclusive exons, intron retention, and exon skipping, we found that the most frequent subtype was exon skipping across all tissues (Fig. 2a). Notably, 26 genes showed differential alternative splicing in all tissues (Fig. 2b), two of which (Slc35f3 and Zscan2) belonged to the same alternative splicing subtype and direction. We further investigated the pathways associated with these genes using Ingenuity Pathway Analysis (IPA; p < 0.05) and identified diverse pathways for each tissue sample with only three pathways (apelin cardiomyocyte signaling pathway, calcium signaling, Huntington's disease signaling) showing an overlap (Supplementary dataset 8: IPA_splicing.xlsx); this indicates that under mutant FUS (ΔNLS-FUS) overexpression, aberrant splicing regulation varies and is tissue-dependent. To explore our data further, we narrowed our study to 73 alternative splicing events in 72 genes under more stringent conditions (p < 0.001 and |IncLevelDifference| > 0.5) and identified 18 proteins with varying functions (RNA dysregulation, protein quality control, neurotransmitter, vesicular transport, microtubule regulation) that are possibly associated with the neurodegenerative pathology of ALS/FTD ( Table 2). www.nature.com/scientificreports www.nature.com/scientificreports/ To identify the role of RBPs in the regulation of the differential alternative splicing events identified, we analyzed known RBP motifs and binding sites around the differentially regulated alternatively spliced exons belonging to the exon-skipping subtype using rMAPS, which systematically generates RNA-maps depicting the spatial distribution of RBP motifs in the vicinity of these alternatively spliced exons (see Supplementary dataset 9-14; rMAPS. up or down_FCx, Hp or SC). The RNA-map generated for FUS, depicting the spatial distribution of the FUS motif around differential exon skipping events, showed FUS motif enrichment downstream (~50-125 bp) of the silenced and enhanced exons in FCx and Hp tissues, whereas enrichment of silenced exons was more dominant in SC tissue (Fig. 3). Notably, our analysis further revealed that other RBPs including those known to be associated with ALS and/or stress granule formation, i.e., FMR1, hnRNPA1, hnRNPA2B1, and Matrin-3, also showed enrichment around differentially regulated alternatively spliced exons in the Tg mice (Supplementary Fig. 4-6. Thus, our motif analysis suggests a potential role not just for FUS but for various RBPs in the aberrant splicing regulation observed resulting from mutant FUS expression in the Tg mice.

Discussion
Several studies have demonstrated that many RBPs with known mutations associated with ALS/FTD accumulate in the cytoplasm of the affected regions of the central nervous system and possibly contribute to the RNA dysregulation implicated in ALS/FTD pathogenesis [3][4][5][6]11 . Our previous study showed that the cytoplasmic mutant FUS aggregates observed in the ΔNLS-FUS Tg mice sequester mRNA and RNA transporters, leading to a disturbance in RNA/protein homeostasis in the dendrites, even before neuronal loss occurs 8,9 . Our current study provides valuable insights regarding the RNA dysregulation associated with the early stages of ALS/FTD before neuronal loss.
Whole transcriptome analysis identified Sema3g as significantly upregulated in FCx and Hp tissues isolated from the Tg mice. This is in agreement with the brain transcriptome profiles reported for sporadic ALS (NCBI Gene Expression Omnibus; accession number: GSE67196), which shows an increase in the expression of Sema3g in FCx 12 . Furthermore, transcriptome analysis performed during our previous study on the whole cerebrum of 1-year-old ΔNLS-FUS Tg mice also showed a 10-fold upregulation in Sema3g (Wt, 0.22 FPKM; Tg mice, 2.31 FPKM) 9 ; RNaseL, Kcnh7, Atp6ap1, Npl, and Hspa5 showed similar changes in expression levels to that of the previous study, but the differences were not significant. Interestingly, Sema3g was recently found involved in the regulation of synapse density and plasticity of hippocampal neurons through its neuronal neuropilin 2/plexin-A4 holoreceptor 13 . Sema3g-deficient mice are reported to show disturbances in hippocampal-dependent memory 13 . Moreover, our previous study on ΔNLS-FUS Tg mice showed a decrease in dendritic spine and synaptic density in the Hp, leading to an impairment of hippocampal memory retrieval 8 . These studies indicate a critical role for Sema3g in hippocampal synapse regulation in neurodegenerative diseases, and our current study suggests a possible role for Sema3g in the early stages of ALS/FTD. Therefore, we propose Sema3g as a potential biomarker for the early stages of ALS/FTD and a potential therapeutic target in various cognitive disorders.
Our study also identified a large number of differentially regulated, alternatively spliced exons in the 6-month-old ∆NLS-FUS Tg mice, which corresponds to the early stages of ALS/FTD. Pathway analysis indicated that these exons were associated with different pathways in a tissue-specific manner. Previous studies on FUS-knockout mice demonstrated a major role of FUS in regulating specific subtypes of alternatively spliced exons 14 ; 77% of cassette exons increased their inclusion in the knockout brain 14 ; however, the rate was only 54.1%    www.nature.com/scientificreports www.nature.com/scientificreports/ The previous study as well as the current experiments showed that the mutant FUS protein is strictly localized in the cytoplasm, whereas splicing events occur in the nucleus (see Supplementary Fig. S1) and no evident changes in expression level or nuclear localization of endogenous FUS 8,9 were observed, suggesting that the aberrant splicing observed in ΔNLS-FUS Tg mice may be due to an indirect effect, and that mutant FUS may be interfering with other factors involved in splicing regulation in a toxic gain-of-function manner. Furthermore, recent evidence showed that the mutant FUS aggregates sequester mRNA and associated proteins including spliceosomes in the cytoplasm 8,9,[15][16][17] . Therefore, we investigated the role of other RBPs in regulating the differential alternative splicing events identified. As rMAPS analysis revealed various RBPs enriched around the cassette exons of differentially regulated exon skipping events, we propose that the mutant FUS possibly interferes with other RBPs, perturbing the splicing machinery and resulting in aberrant splicing regulation and RNA dysregulation. www.nature.com/scientificreports www.nature.com/scientificreports/ There are several limitations associated with the study. First, we cannot exclude the possibility that the induced expression of mutant FUS may itself affect neuronal functions. However, in our previous study, we confirmed that the expression level of exogenous FUS was ~80% of that of endogenous FUS and that nuclear localization of endogenous mouse FUS was not affected by mutant FUS expression 9 . Nonetheless, Tg mice overexpressing even wild-type FUS were reported to show pathologies associated with motor neuron degeneration reminiscent of ALS 18,19 . Recent genomic analysis identified four mutations in the 3ʹ untranslated region of FUS-which increases the expression of native FUS-that are associated with sporadic and familial ALS 20,21 . Therefore, it was suggested that increased expression of even wild-type FUS can affect RNA regulation and cause ALS-associated symptoms 3,20,21 . Furthermore, in previous studies, other RNA-binding proteins (FMR1, staufen, and G3bp) and various mRNA were trapped in the cytoplasmic ΔNLS-FUS aggregates, suggesting that these aggregates directly lead to RNA dysregulation without interference from endogenous FUS 8,9 . Although we cannot rule out that cytoplasmic aggregates of ΔNLS-FUS sequester undetectable levels of endogenous FUS, our data indicate that mislocalized mutant FUS results in aberrant splicing in a toxic gain-of-function manner.
Second, our transcriptome analysis was limited to 6-month-old Tg mice that developed FTD-like behavioral deficits and had decreased dendritic spine and synaptic density starting from 9-13 weeks of age 9 . It would be interesting to examine whether these Tg mice show the observed aberrant splicing regulation at 2-3 months of age. Third, in ALS patient-specific induced-pluripotent stem cells, including ALS patient-derived VCP(R155C and R191Q), SOD1(A4V), and FUS(R521G) mutations, increased intron retention is a dominant feature during early neural differentiation 22,23 , which is in contrast to the most frequent exon skipping type detected in our study. Although we could not exclude that each ALS mutation has different effects on splicing regulation, it is also possible that aberrant splicing events in ALS are altered depending on neural differentiation stage and maturation. Finally, we were unable to confirm the elevated sema3G protein levels as well as protein-level changes resulting from aberrant splicing in patient samples. We aim to examine these levels in patient CSF in future studies. This is crucial as secreted sema3G is a potential biomarker for evaluating disease stage and its downstream factors, neuropilin-2/PlexinA4 receptor and Rac1, may also represent therapeutic targets in ALS/FTD treatment.
Nevertheless, our study identified extensive splicing changes mediated by mutant FUS in brain tissues during the early stages of ALS/FTD and confirmed the notion that dysfunction in RNA processing is a key aspect of neurodegeneration associated with ALS/FTD. Although further study is needed to validate whether aberrant splicing regulation leads to abnormal changes in protein levels and cellular functions linked to neurodegeneration, our study indicates that strategies targeting this aberrant splicing regulation may be a useful approach for the treatment of the motor/cognitive deficits of ALS/FTD.

Materials and Methods
Mice. Generating ΔNLS-FUS Tg mice has been described in detail in our previous study 9 ; briefly, an NLS deleted mutant of human FUS cDNA was inserted into the pTSC21k vector containing the murine Thy1.2 expression cassette. The Tg mice were backcrossed for more than five generations with C57BL6/J mice. All mice used in this study were individually housed in ventilated micro-isolation cages under controlled temperature, humidity, lighting, and ventilation conditions with free access to pellet food and sterilized water during the experiments. DNA was isolated from ear clips and genotyping was performed using the following primer pair: 5′-AAGAAGACCTGGCCTCAAACG-3′ and 5′-TATCCCTGGGGAGTTGACTG-3′. Animal research protocols employed in this study were approved by the Committee on Animal Care and Use, Keio University (09217- [4]) and KAN Research Institute, Inc. and were conducted according to the Animal Experimentation Guidelines of Keio University School of Medicine and KAN Research Institute, Inc. immunohistochemical staining. Mice were deeply anesthetized with the MMB combination anesthetic (i.e., 0.3 mg/kg of medetomidine hydrochloride, 4.0 mg/kg of midazolam, and 5.0 mg/kg of butorphanol tartrate) and then perfused intracardially with freshly prepared 2% paraformaldehyde in phosphate buffer for 10 min. Brains were dissected and soaked in the same fixative for 2 h on ice. For cryoprotection, the tissue pieces were placed into 20% sucrose solution for 4 h and 25% sucrose solution overnight, after which they were quickly frozen using liquid nitrogen and stored at −80 °C. Serial 7-µm-thick sections were cut using a cryostat (Leica Biosystems, Wetzlar, Germany). The sections were blocked with 1% BSA/10% normal donkey serum/0.5%Triton X-100/M.O.M. Blocking Reagent in PBS for 1 h at room temperature, followed by incubation with the primary anti-c-Myc 9E10 (1:100; sc-42; Santa Cruz Biotechnology, Dallas, TX) and FUS polyclonal (1:100; PA5-23696; Thermo Fischer Scientific, Waltham, MA) antibodies overnight. The sections were then incubated with the secondary antibodies Cy3 anti-mouse IgG (1:100; 715-166-151; Jackson ImmunoResearch Laboratories, West Grove, PA) and Alexa Fluor 488 anti-rabbit IgG (1:100; 711-546-152; Jackson ImmunoResearch Laboratories) for 1 h at room temperature. The embedded samples were imaged with a laser scanning confocal microscope (LSM700; Carl Zeiss, Oberkochen, Germany).

RnA extraction.
Total RNA was isolated from FCx, Hp, and lumbar SC tissues from 6-month-old mice [Wt mice, n = 4 (two males and two females); Tg mice, n = 4 (two males and two females)] using TRIzol reagent (15596-018; Invitrogen, Carlsbad, CA) according to manufacturer's instructions. RNA samples were digested with the RNase-free DNase Set (79254; Qiagen, Hilden, Germany) to remove genomic DNA and further purified using the RNeasy Kit (74104; Qiagen). RNA sample quality was determined using the Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA) using RNA 6000 NanoChips (5067-1511; Agilent Technologies). RNA samples with an RNA integrity number (RIN) higher than 7 were used for further analysis, including qRT-PCR and RNA sequencing.