Highlights
We performed transcriptome analysis of hiPSC-derived ALS FUS mutant motoneurons
Bioinformatics analysis revealed cell adhesion pathways associated to sporadic ALS
miR-375 expression is affected by FUS mutation in motoneurons
Impairment of miR-375 alters RNA metabolism and increases pro-apoptotic factors
Summary
The FUS gene has been linked to amyotrophic lateral sclerosis (ALS). FUS is a ubiquitous RNA-binding protein, and the mechanisms leading to selective motoneuron loss downstream of ALS-linked mutations are largely unknown. We report the transcriptome analysis of human purified motoneurons, obtained from FUS wild-type or mutant isogenic induced pluripotent stem cells (iPSCs). Gene ontology analysis of differentially expressed genes identified significant enrichment of pathways previously associated to sporadic ALS and other neurological diseases. Several microRNAs (miRNAs) were also deregulated in FUS mutant motoneurons, including miR-375, involved in motoneuron survival. We report that relevant targets of miR-375, including the neural RNA-binding protein ELAVL4 and apoptotic factors, are aberrantly increased in FUS mutant motoneurons. Characterization of transcriptome changes in the cell type primarily affected by the disease contributes to the definition of the pathogenic mechanisms of FUS-linked ALS.
Introduction
The fatal neurodegenerative disease amyotrophic lateral sclerosis (ALS) is caused by loss of motoneurons (MNs) in the spinal cord and brain, leading to progressive muscle atrophy. Both sporadic ALS (sALS) and familial ALS (fALS) have been linked to a number of genes, including SOD1, C9ORF72, TDP-43, TAF15, and FUS/TLS(FUS) (Renton et al., 2014). The FUS gene encodes for an RNA-binding protein involved in RNA biogenesis and maturation (Lagier-Tourenne et al., 2012). Wild-type (WT) FUS is mainly localized in the nucleus, and a hallmark of FUS ALS patients is the presence of cytoplasmic inclusions containing mutated protein in the brain and spinal cord (Kwiatkowski et al., 2009, Vance et al., 2009). Many ALS-associated FUS mutations disrupt the function of the C-terminal nuclear localization signal (NLS). Therefore, defects in nuclear import, leading to aberrant cytoplasmic localization of FUS, have been proposed as the initial step in ALS pathogenesis (Bentmann et al., 2013, Dormann and Haass, 2011). Recent analyses in murine models suggest that loss of FUS function may not be sufficient to induce MN degeneration (Kino et al., 2015, Scekic-Zahirovic et al., 2016), and both loss of a nuclear function and gain of a toxic function in the cytoplasm have been proposed as contributing pathological mechanisms in FUS ALS (Ling et al., 2013).
FUS mutations are expected to exert a major change in the transcriptome. However, this has not been directly addressed in human MNs so far. Human induced pluripotent stem cells (iPSCs) represent a useful system to model ALS, as they can be differentiated into disease-relevant cell types (Sances et al., 2016). iPSCs carrying ALS mutations can be generated by reprogramming from patients (Boulting et al., 2011) or by gene-editing techniques, producing lines that differ only for the specific mutation (Kiskinis et al., 2014, Lenzi et al., 2015). Comparison of gene-edited mutant lines with their otherwise isogenic WT counterparts allows to assign any observed phenotypic difference to the ALS mutation, avoiding any bias due to different genetic backgrounds of non-isogenic mutants and controls.
Recent evidence suggests that microRNAs (miRNAs) might play a role in MN diseases, such as ALS and spinal muscular atrophy (SMA) (Kye and Gonçalves, 2014). Conditional loss of the processing enzyme Dicer in postnatal MNs causes an SMA-like phenotype in mice (Haramati et al., 2010), suggesting that, collectively, miRNAs could be required for MN survival. Global downregulation of miRNAs was observed in MNs of sporadic ALS spinal cords isolated by laser-capture microdissection (Emde et al., 2015), and a protective role for individual miRNAs, such as miR-218 and miR-375, has been shown in MNs (Amin et al., 2015, Bhinge et al., 2016). Previous work from our lab and others has shown that proteins genetically linked to ALS play a role in miRNA biogenesis. FUS stimulates miRNA processing by facilitating the recruitment of the enzyme Drosha on miRNA genes (Morlando et al., 2012), and TDP-43 regulates miRNA biogenesis at multiple levels (Buratti et al., 2010, Di Carlo et al., 2013, Kawahara and Mieda-Sato, 2012). Moreover, overexpression of SOD1, TDP-43, or FUS proteins, either WT or carrying ALS mutations, impaired miRNA processing in the cytoplasm by inducing cellular stress (Emde et al., 2015). As most of the studies on the contribution of ALS-linked factors to miRNA biogenesis come from cell lines or non-MN cells, it is currently unknown which miRNAs could be affected by ALS mutations in MNs.
Here we took advantage of mutant FUSP525L and isogenic control iPSCs to generate and isolate pure human MNs, which were used for whole-transcriptome analysis by RNA sequencing (RNA-seq). We identified both long and short RNA species significantly altered in FUS mutant MNs. Changes in the transcriptome point to pathways previously associated to neurodegenerative diseases, such as those related to cell adhesion. Among differentially expressed miRNAs we focused on miR-375. Levels of miR-375 were decreased in mutant MNs, in which we observed an increase of a subset of its targets, such as p53 and the neural RNA-binding protein ELAVL4. We propose a pathological mechanism in which an impairment of miRNAs production, downstream of FUS mutations, would alter RNA metabolism and increase MN vulnerability via aberrant upregulation of pro-apoptotic targets.
Results
Differentiation and Isolation of iPSC-Derived MNs
We have recently reported the generation of iPSC lines devoid of mutations in the FUS gene (hereafter FUSWT) or carrying the severe P525L ALS-related mutation introduced by gene editing in both FUS alleles (hereafter FUSP525L) (Lenzi et al., 2015). We stably modified these lines by inserting a Hb9:GFP reporter into the AAVS1 locus (Wainger et al., 2014) (Figures S1A and S1B). To improve time and efficiency of MN generation from iPSCs, we took advantage of an optimized protocol based on neural induction and patterning of differentiating cells in adhesion conditions (Figure 1A). Using this protocol we consistently obtained MN progenitors in a 12 day time frame. A time course analysis of marker expression in differentiating FUSWT cells is shown in Figure 1B. By day 6, cells lose the pluripotency marker NANOG. Around this time point we detected a peak of expression for the neural progenitor gene PAX6, followed by a raise of pan-neuronal (TUJ1) and MN (ISL1) markers at day 9. The late MN gene CHAT was expressed at day 9, with a further increase until day 12. Levels of the MN progenitor marker HB9 peaked around day 9 and decreased at later time points. The mixed population of differentiated cells at day 12 was highly enriched of ISL1/2-positive neurons (Figure S1C). Live cell imaging and flow cytometry analysis indicated that 30%–40% of cells were GFP-positive at this time point (Figures 1C and 1D). Notably, expression of the reporter was detected only in retinoic acid and smoothened agonist (SAG)-treated cells, which induce a ventral spinal cord character, suggesting that the exogenous Hb9promoter was correctly turned on in the MN lineage (Figure 1E). We further validated the reporter system by gene expression analysis in distinct cell populations isolated using cell sorting (Figure S1D). Next-generation sequencing (RNA-seq) was performed on total RNA from GFP-positive and -negative cells collected after sorting. Cluster analysis based on gene expression data correctly separated the samples into GFP-positive and -negative ones (Figures S1E and S1F). GFP-positive samples had a similar expression profile, while GFP-negative ones were more dissimilar. A total of 1,631 genes were differentially expressed between GFP-positive and -negative cells, most of them having higher expression levels in GFP-negative samples (Figure S1G; Table S1). As shown in Figure 1F, levels of MN markers (in green) were higher in GFP-positive samples, with the exception of LHX1. Notably, the HB9 transcript was not significantly enriched in GFP-positive samples. This apparent inconsistency can be explained by the narrow window of time of HB9 expression: by the time point in which GFP protein reaches its maximum accumulation (day 12; Figure 1D) the endogenous HB9 mRNA is already downregulated (Figure 1B). Compared with GFP-positive cells, GFP-negative populations probably contained a fraction of MN progenitors at earlier stages, expressing high levels of HB9. As expected, ALS-related genes SOD1, C9ORF72, FUS, and TARDBP were not enriched in GFP-positive cells. Experimental validation of selected markers by real-time qPCR confirmed the RNA-seq results (Figure 1G). Purified GFP-positive cells were re-plated on laminin-coated dishes and upon further culturing acquired a neuronal morphology (Figure S1H). Immunostaining analysis indicated that sorted GFP-positive cells were highly enriched for ISL1/2-positive MNs (Figure 1H).
Taken together, the combination of a short and efficient differentiation protocol and the stable insertion of the Hb9:GFP reporter allowed the effective isolation of a large number of purified, iPSC-derived, MNs.
Transcriptome Profiling in FUS WT and Mutant Human MNs
We have previously shown that ALS mutations in FUS do not impair differentiation (Lenzi et al., 2015, Lenzi et al., 2016). This observation was confirmed in MNs generated with the improved protocol described in Figure 1 from FUSWT and FUSP525L iPSCs. The fraction of GFP-positive cells was consistently comparable over six independent experiments (Figures 2A and 2B ). Moreover, GFP-positive cells with a FUSWT or FUSP525L background showed comparable levels of expression of MN markers (Figure S2A). Notably, in the Hb9:GFP FUSP525L line both the mutation and the reporter system have been introduced by gene editing. Therefore, besides the FUSP525L mutation, iPSC lines used for subsequent experiments were isogenic.
We took advantage of our iPSC system to systematically analyze the consequences of the FUSP525L mutation on the transcriptome in pure populations of in-vitro-derived human MNs. GFP-positive FUSWT and FUSP525L iPSCs were sorted at day 12 of differentiation and further cultured on laminin-coated dishes for an additional 7 days (day 12 + 7). Total RNA from three independent differentiation experiments was collected and analyzed by RNA-seq. Gene expression-based cluster analysis resulted in the segregation of samples into two distinct groups based on their FUSgenotype (Figures S2B and S2C). Out of 14,289 genes expressed in human MNs, 267 genes were upregulated and 244 downregulated in a consistent manner in FUSP525L compared with FUSWT (false discovery rate [FDR] < 0.05) (Table S1). The gene expression heatmap of differentially expressed genes is shown in Figure 2C. RNA-seq results were validated by real-time qPCR in at least three additional independent samples. Figure 2D shows validation of four representative genes among those downregulated (SCL17A8, NXPH2, CRIM1, and GRIN2A) and upregulated (TP53I3, CDH7, NNAT, and TSPYL5). Expression of the FUS transcript itself, as well as the mRNAs of other major ALS-related genes, was not significantly altered in FUS mutant MNs (Figure S2D).
We confirmed these results in an independent FUS WT and mutant isogenic pair. A second FUSP525L iPSC line (hereafter FUSP525L#2), generated as previously described (Lenzi et al., 2015), and its parental FUSWT iPSC line (hereafter FUSWT#2) were modified by inserting the Hb9:GFP reporter into the AAVS1 locus (Figures S3A–S3D). FUSWT#2 and FUSP525L#2 were then differentiated, sorted, and collected at day 12 + 7 for RNA analysis. All genes analyzed in the original isogenic pair resulted consistently altered in MNs obtained from these new iPSC lines (Figure 2D).
A subset of these genes was also affected in MNs obtained from a Hb9:GFP-modified heterozygous FUSP525L iPSC line, hereafter FUSP525L(ht) (Lenzi et al., 2015) (Figure 2D).
Functional Consequences of the FUSP525L Mutation in Human MNs
We next assessed whether the change in the transcriptome detected in FUS mutant cells might be informative of pathways potentially involved in ALS MN degeneration. Gene ontology (GO) term enrichment analysis with the FIDEA tool (D’Andrea et al., 2013) highlighted relevant categories that cooperate in pathways and distinct molecular functions. In particular, we noticed a striking enrichment for categories related to cell adhesion in genes differentially expressed in FUSP525L MNs (Figure 3A). Notably, “cell adhesion” is one of the categories consistently enriched in multiple studies focusing on FUS-bound transcripts and splice targets in mouse and human neurons (Orozco and Edbauer, 2013). GO term enrichment analysis of differentially expressed genes in laser-captured MNs from postmortem sporadic ALS patients also indicated “cell adhesion” (Batra et al., 2016). Moreover, functional enrichment analysis for genome-wide association studies involving several neurological diseases (including ALS) showed significant values for the “cell-cell adhesion” GO category (Guio-Vega and Forero, 2017). This category was also enriched in the analysis of differentially expressed genes in ALS fibroblasts carrying the C9ORF72 mutation (Kotni et al., 2016). Cell adhesion-related categories were enriched also when upregulated and downregulated genes were analyzed separately (Figure S4A). Interestingly, the Kyoto Enrichment of Genes and Genomes pathway “Neuroactive ligand-receptor interaction” was specifically enriched by the downregulated genes. Previous work found this pathway enriched in RNA profiling from C9ORF72 mutant fibroblasts (Kotni et al., 2016) and linked to neurological diseases such as Parkinson’s disease (Kong et al., 2015).
Seeking for ALS signature in FUS mutant MNs, we crossed our RNA-seq data with a recently published dataset from laser-captured MNs fom sALS patients (Batra et al., 2016, Kapeli et al., 2016). In these works, the authors found 2,346 upregulated and 955 downregulated genes in sALS patients. The variation of upregulated genes was more robust, since more stringent selection criteria had lower impact on their abundance. These two lists of genes were employed as two distinct gene sets in the context of a gene set enrichment analysis to evaluate whether they were over-represented in the set of genes deregulated upon FUS mutation. Notably, the set of genes downregulated in sALS patients was enriched among those downregulated in FUSP525L (FDR = 0.151), and the set of genes upregulated in sALS patients was significantly enriched among those upregulated in FUSP525L (FDR = 0.009) (Figure S4B). These results suggest a common signature in gene expression between in-vitro-derived FUSP525L MNs and sALS MNs. As suggested by FDR values, this tendency is stronger for the upregulated genes, consistent with the robustness of their deregulation in sALS patients.
We next investigated whether there is any relationship between differential gene expression and direct FUS binding in mutant MNs. The FUS interactome has never been characterized in purified human MNs; however, FUS CLIP-seq (crosslinking and immunoprecipitation, followed by high-throughput sequencing) data from human brain cortex are available (Lagier-Tourenne et al., 2012). We re-analyzed this dataset and evaluated the position of FUS binding clusters relative to the different regions of the pre-mRNAs, which are expressed in our MNs. FUS binding was enriched in the exonic regions of protein-coding genes, and in particular in the 3′ UTR, when compared with sequences located more (distal introns) or less (proximal introns) than 500 bp from the nearest exon-intron junction (Figure 3B). FUS-bound genes were classified as differentially expressed or not in FUSP525L MNs, and individual mRNA regions were analyzed separately. A significantly higher percentage of genes modulated in FUS mutant MNs was directly bound by FUS in the distal intron and 3′ UTR (Figure 3C). When the same analysis was repeated with the TDP-43-bound genes dataset (Tollervey et al., 2011), we did not observe a significant enrichment for the 3′ UTR set (Figure S4C). Taken together, these results suggest that mRNAs bound by FUS in the 3′ UTR might be more susceptible to changes in FUS activity due to ALS mutations. We validated differential expression of selected FUS-bound mRNAs in mutant MNs (TAF7, TAF15, GRIN2A, PCDH10, and CRIM1) (Figures 2D and 3D).
To assess whether the changes in gene expression downstream of FUS ALS mutations are consistent with an animal model exhibiting ALS features, we crossed our data with the transcriptome profiling of a recently described mouse carrying targeted deletion of the PY-NLS (FusΔNLS/ΔNLS) (Scekic-Zahirovic et al., 2016). This mutation severely impairs the nuclear localization of FUS, mimicking the human P525L mutation and other truncations of the protein C-terminal. Among differentially expressed genes, 17 were downregulated and 8 were upregulated in both mutant mouse brain and human MNs (Figures 3E and S4D). The overlap was statistically significant both for upregulated and downregulated gene lists (p value for hypergeometric test <0.05). Interestingly, mRNA levels of TAF15, a FUS family member mutated in a subset of fALS patients (Couthouis et al., 2011), were significantly increased in both mouse (Scekic-Zahirovic et al., 2016) and human (Figure 3D). TAF15 was among the transcripts directly bound by FUS (Lagier-Tourenne et al., 2012). These findings suggest that a possible crosstalk between FUS and TAF15, hypothesized in murine models, might be conserved in human.
FUSP525L Mutation Impairs miR-375 Biogenesis in MNs by a Loss-of-Function Mechanism
Previous studies in cell lines have hypothesized a link between FUS mutations and the impairment of miRNA function (Emde et al., 2015, Morlando et al., 2012), but this was never directly investigated in human MNs. We performed miRNA profiling by small RNA sequencing (small RNA-seq) in FUSWT and FUSP525L MNs. Based on miRNA expression levels, samples could be clustered according to their FUSgenotype (Figures S5A–S5C). Of the 573 miRNAs expressed in human MNs (day 12 + 7), 21 resulted downregulated and 11 upregulated in FUSP525L mutant cells (FDR < 0.05) (Figure S5D; Table S1). We also analyzed to which extent the change in the miRNome is maintained at different time points. TaqMan array cards and small RNA-seq identified miR-375 and miR-484 as differentially expressed miRNAs at both day 12 and day 12 + 7 (Figure 4A; Table S2). These and other selected candidates were analyzed by real-time qRT-PCR in FUSP525L, FUSP525L#2, and FUSP525L(ht) MNs, and their isogenic WT controls (Figures 4B, S5E, and S5F). Among them, miR-375 was consistently downregulated in both homozygous FUSmutant lines over at least four independent experiments (Figure 4B).
miR-375 has been recently described as a protective miRNA in MNs (Bhinge et al., 2016). Consistent with previous observations (Bhinge et al., 2016), miR-375 was highly upregulated in MNs compared with undifferentiated iPSCs (Figure 4C). To gain insights into the mechanisms underlying miR-375 downregulation in FUSP525LMNs we transfected differentiating FUSWT iPSCs with small interfering RNAs (siRNAs) targeting FUS and analyzed gene expression at day 12 of differentiation. In these conditions, FUS levels were reduced to about 60% compared with scramble siRNA-transfected cells (Figure 4D). We have previously shown that FUS localization in the nucleus in homozygous FUSP525L cells is reduced to 50% of its normal levels, due to partial delocalization in the cytoplasm (Lenzi et al., 2015). Therefore, the decrease of FUS achieved by siRNA in WT cells is expected to reproduce possible loss of nuclear function effects of the mutation. FUS reduction did not affect differentiation, as assessed by ISL1, CHAT, and TUJ1 marker analysis (Figure 4D). Levels of miR-375 were significantly reduced upon FUS knockdown (Figure 4D). Conversely, expression of selected genes among the most upregulated (NNAT, TSPYL5, and TAF15) or downregulated (CRIM1 and SLC17A8) ones in FUSP525L MNs was not affected by FUS knockdown (Figure 4D).
These results suggest that ALS mutations in the FUS gene might impair miR-375 production via a nuclear loss-of-function mechanism, while gain of toxic functions of the mutated protein might underlie the alteration of protein-coding genes expression.
miR-375 Target Genes in FUS Mutant MNs
To identify miR-375 targets in human MNs, we crossed mRNA expression data from the RNA-seq with predicted (TargetScan; Agarwal et al., 2015) and previously validated targets (miRTarBase; Chou et al., 2016). Eight mRNAs upregulated in FUSP525L MNs were also miR-375 predicted targets (Figure 5A; Table S3), including ELAVL4 (also known as HuD), which had been previously validated as a bona fidemiR-375 target (Abdelmohsen et al., 2010). The increased expression of ELAVL4and other predicted targets (PHLDA1, EBF3, B3GAT2) in mutant MNs was validated by real-time qPCR (Figure 5B). The ELAVL4 gene encodes for a neuron-specific RNA-binding protein (Bronicki and Jasmin, 2013). In mouse MNs, the transcript of the NRN1 (CPG15) gene is among known targets of ELAVL4, which promotes NRN1 mRNA accumulation by enhancing its stability (Akten et al., 2011). In line with this, we detected increased levels of NRN1 in FUSP525L MNs (Figure 5B).
We then noticed that the p53 gene (TP53) was a previously described target of miR-375 (Liu et al., 2013). TP53 mRNA was unchanged or slightly upregulated, respectively, in FUSP525L and FUSP525L#2 MNs (Figure 5B). In both lines, we detected an increase of p53 protein by western blot (Figure 5C). Among miR-375 targets, we noticed that EBF3 and PHLDA1 (also known as TDAG51) have been previously involved in the induction of apoptosis (Gomes et al., 1999, Zhao et al., 2006).
To assess whether changes in target mRNA levels could be explained by differences in miRNA levels, we transfected synthetic miR-375 mimics in FUSP525L differentiated cells. We observed a significant decrease of ELAVL4 and TP53 mRNA levels compared with control-transfected cells (mimics-C) (Figure 5D). Since levels of EBF3 and PHLDA1 were not decreased, we hypothesize that other mechanisms might contribute to their upregulation in FUS mutant cells.
Collectively, these findings suggest that dysregulation of miR-375 in FUS mutant MNs may have an impact on RNA metabolism and survival due to aberrant increase of ELAVL4 and p53, respectively.
Discussion
In this study we report the whole-transcriptome profile of in-vitro-derived human MNs with WT or mutant (P525L) FUS background. Our analysis was performed on two pairs of isogenic lines generated by gene editing. We therefore provide significant information on the changes in the transcriptome of the neuronal subtype primarily affected by ALS and due solely to the specific, ALS-linked, FUSP525L mutation. This represents a major advance beyond previous work, in which the transcriptome of mixed populations of progenitor cells (composed for the most part of non-MN cells and derived from non-isogenic FUS control/mutant iPSCs) was profiled by exon array (Ichiyanagi et al., 2016).
We found that numerous differentially expressed genes belong to functional categories related to the cell adhesion pathway. Interestingly, this correlates with transcriptome analysis of postmortem laser-capture microdissected MNs from ALS patients, genome-wide association studies on several neurological diseases and differential gene expression in C9ORF72 mutant cells (Aronica et al., 2015, Batra et al., 2016, Guio-Vega and Forero, 2017, Kotni et al., 2016). In the light of our findings, cell adhesion-related pathways might be considered as biomarkers and/or therapeutic targets in FUS ALS patients.
We also crossed our data with the transcriptome analysis from a recently reported murine model (Scekic-Zahirovic et al., 2016). Among genes commonly upregulated in both species we noticed TAF15, which is mutated in some ALS patients (Couthouis et al., 2011). TAF15 is a member of the FET family of RNA-binding proteins and is closely related to FUS. Similarly to FUS and TDP-43, upregulation of WT TAF15 causes neurodegeneration in Drosophila (Couthouis et al., 2011). Notably, TAF15 levels were slightly increased in FUS null mice (Kino et al., 2015, Scekic-Zahirovic et al., 2016), but not in human MNs with reduced FUS levels (Figure 4D). This discrepancy might be explained by residual FUS protein in our RNAi experiment. A shift in TAF15 solubility and its localization into FUS-positive inclusions has been observed in frontotemporal lobar degeneration, but not in ALS (Neumann et al., 2011). Therefore, it remains unclear whether TAF15 upregulation in FUS mutant MNs contributes to ALS pathogenesis.
Changes in mRNA levels may be a direct consequence of the FUS mutation. Despite the fact that specific information on FUS-bound transcripts in human MNs is still missing, FUS CLIP-seq data from human brain cortex are available (Lagier-Tourenne et al., 2012). Our results suggest a possible direct effect of FUS mutations on those transcripts that are targeted by FUS on their 3′ UTR. Interestingly, levels of several transcripts were altered in the same direction in homozygous and heterozygous FUSP525L MNs. The P525L mutation is inherited in an autosomal dominant way in ALS patients, and homozygous individuals have never been reported. Unexpectedly, in spinal cord tissue from heterozygous FUS ALS patients a near absence of nuclear FUS was observed in MNs, suggesting that both mutant and WT FUS can be sequestered within cytoplasmic inclusions (Vance et al., 2013). As the nuclear fraction of FUS in homozygous and heterozygous P525L cells is, respectively, 80% and 50% (Lenzi et al., 2015), the homozygous mutant could be regarded as representative of a more advanced stage of the disease. Mechanistically, toxic functions of mutant FUS might underlie the alteration of at least a subset of mRNAs, as suggested by knockdown experiments.
FUS mutations might also affect gene expression indirectly, for instance by altering miRNA levels. We found that miR-375 levels were consistently lower in MNs derived from two independent FUSP525L lines. Bhinge et al. (2016) recently showed that miR-375 is enriched in human spinal MNs derived from embryonic stem cells. Interestingly, miR-375 overexpression protected MNs from DNA damage-induced apoptosis. Mechanistically, miR-375 may exert this function by targeting p53 (Bhinge et al., 2016, Liu et al., 2013). We show here that p53 and other pro-apoptotic miR-375 predicted targets are upregulated in FUS mutant MNs. Notably, spinal MNs of ALS patients have been shown to display higher levels of p53 protein, which could be involved in apoptosis-mediated neuronal death (Martin, 2000, Ranganathan and Bowser, 2010).
In FUSP525L MNs we observed 40%–50% reduction in miR-375 levels. This is strikingly similar to the situation observed by Bhinge et al. (2016) in SMA MNs, in which miR-375 levels were reduced to 40%. Therefore, miR-375 dysregulation represents a common feature between two different MN diseases. While the molecular basis of miR-375 downregulation in SMA MNs remains currently unknown, we propose that ALS mutations causing FUS exclusion from the nucleus impair miR-375 production via a loss-of-function mechanism.
In addition to p53-induced apoptosis, miR-375 dysregulation might have an impact on other crucial pathways. Its decrease in FUSP525L MNs is mirrored by an increase of its target ELAVL4, an RNA-binding protein playing multiple roles in neural development, function, and degeneration (Bronicki and Jasmin, 2013). ELAVL4 regulates the stability and/or localization of several target transcripts. We propose that FUS, through the miR-375/ELAVL4 axis, is hierarchically at the top of a regulatory network with a broader role in the RNA metabolism of MNs.
In conclusion, transcriptome analysis of isogenic FUSWT and FUSP525L in-vitro-derived human MNs provided an insight on target genes and pathways altered by ALS mutations in the cell type primarily affected by the disease.