C9ORF72 repeat expansion causes vulnerability of motor neurons to Ca2+-permeable AMPA receptor-mediated excitotoxicity

Mutations in C9ORF72 are the most common cause of familial amyotrophic lateral sclerosis (ALS). Here, through a combination of RNA-Seq and electrophysiological studies on induced pluripotent stem cell (iPSC)-derived motor neurons (MNs), we show that increased expression of GluA1 AMPA receptor (AMPAR) subunit occurs in MNs with C9ORF72 mutations that leads to increased Ca2+-permeable AMPAR expression and results in enhanced selective MN vulnerability to excitotoxicity. These deficits are not found in iPSC-derived cortical neurons and are abolished by CRISPR/Cas9-mediated correction of the C9ORF72 repeat expansion in MNs. We also demonstrate that MN-specific dysregulation of AMPAR expression is also present in C9ORF72 patient post-mortem material. We therefore present multiple lines of evidence for the specific upregulation of GluA1 subunits in human mutant C9ORF72 MNs that could lead to a potential pathogenic excitotoxic mechanism in ALS.

A LS is a relentlessly progressive and invariably fatal neurodegenerative disorder characterized by the selective death of MNs. Recognition that familial and sporadic ALS are largely phenotypically indistinguishable and share pivotal pathological features highlights the value of the study of familial ALS to shed mechanistic insight into ALS. Expansion of a GGGGCC (G 4 C 2 ) intronic hexanucleotide in the C9ORF72 gene is the most common cause of familial ALS and accounts for ∼10% of sporadic ALS cases 1,2 .
A powerful approach to interrogate the consequences of G 4 C 2 expansion in human MNs is the study of iPSC-derived MNs from ALS patients carrying the C9ORF72 mutation 17,18 . Generation of isogenic lines using gene editing technologies, where the only variable is the mutation of interest, thus allowing direct causality to be assigned to any phenotype, further extends the biological relevance of iPSC studies [19][20][21][22] . In this study, we have used CRISPR/Cas9 technology to excise G 4 C 2 repeats from ALS patient-derived MNs to enable direct study of the pathogenic role of G 4 C 2 repeats with particular reference to AMPAR-mediated excitoxicity. We show that mutant C9ORF72 MNs exhibit increased GluA1 AMPAR expression leading to enhanced MN-specific vulnerability to excitoxicity. This vulnerability is abolished by CRISPR/Cas9-mediated correction of the C9ORF72 repeat expansion in MNs. Finally, we demonstrate that AMPAR expression is also selectively dysregulated in spinal cord, but not cortical, post-mortem tissue from patients harboring C9ORF72 repeat expansions.

Results
Gene targeting and correction of the C9ORF72 mutation. A CRISPR/Cas9-mediated targeting strategy was used to correct the mutation in three independent C9ORF72 iPSC lines. Two CRISPR guide (g) RNAs (gRNA-1 and gRNA-2) were engineered to target either side of the G 4 C 2 repeats in the intron1 of the C9ORF72 gene. We selected gRNAs that have at least three mismatches to potential off-target sites, which are highly homologous to on-target sites. In the absence of a repair template the double-strand breaks were religated by non-homologous end joining resulting in deletion of G 4 C 2 repeats (Fig. 1a). First, gRNAs (1 and 2) along with Cas9 plasmid were transfected in HEK-293 cells, which have 3 × G 4 C 2 repeats, to test if the gRNAs were functional. PCR amplification of G 4 C 2 locus in control HEK-293 cells generated a 420 bp amplicon, whereas HEK-293 cells transfected with gRNAs and Cas9 generated both an unedited 420 bp amplicon (red arrowhead) and a truncated PCR amplicon (blue arrowhead) suggesting both gRNAs were able to target and delete G 4 C 2 repeats (Fig. 1b). Given that the gRNAs were functional, we next nucleofected Cas9 and gRNAs (1 and 2) into the mutant C9ORF72 iPSC lines that harbor G 4 C 2 repeats (C9-1, up to~638 repeats; C9-2, up to~960 repeats; C9-3, up tõ 760 repeats; Supplementary Fig. 1A). Repeat primed PCR and G 4 C 2 locus PCR were used to screen individual single-cell-derived clones. PCR of G 4 C 2 locus in mutant C9ORF72 iPSCs lines yielded a single amplicon of~420-450 bp from the wild-type allele (C9-1, 8 repeats; C9-2, 2 repeats; C9-3, 2 repeats). The mutant allele is not amplified due to larger amplicon size. Deletion of the repeat expansion by CRISPR/Cas9 yields additional PCR amplicons (Fig. 1c). Efficiency of deletion of G 4 C 2 repeats varied between the different iPSC lines ranging from 0.6-4.5% (C9-1, 2.27%; C9-2, 4.5%; C9-3, 0.6%). For each mutant C9ORF72 iPSC line, we selected 1 clone that was confirmed by repeat primed PCR to have a lack of G 4 C 2 repeats. All such lines are hereafter termed C9-1Δ, C9-2Δ, and C9-3Δ (Fig. 1c). In all three C9ORF72 repeat expansion edited lines the wild-type allele was also modified since we observed truncated PCR amplicons (Fig. 1c). Southern blot analysis using a probe outside the G 4 C 2 repeats confirmed absence of repeat expansion in the positive clone (Fig. 1d). We further sequenced the G 4 C 2 locus of both the alleles in the positive clones to determine the deleted sequence. C9-1Δ contains 2 × G 4 C 2 repeats in both alleles; C9-2Δ contains 1 × G 4 C 2 repeat in one allele and absence of G 4 C 2 repeat in other allele; and C9-3Δ contains 2 × G 4 C 2 repeats in one allele and 1 G 4 C 2 repeat in the other allele (Fig. 1e). This confirmed that the gene-corrected iPSC clone was karyotypically normal (Supplementary Fig. 1B) and retained pluripotency markers such as OCT3/4, TRA-1-60 ( Supplementary Fig. 1C). RNA-guided genome editing, although efficient, has higher off-target activity when compared to other genome editing tools 23 . Assessment of highest candidate off-target sites, two for gRNA-1 and two for gRNA-2, by Sanger sequencing confirmed no off-target activity for the chosen gRNAs in all three mutant C9ORF72 iPSC lines (Supplementary Fig. 1D).
C9ORF72 mutant MNs display RNA foci. Having corrected the mutation, we next generated, using a recently reported protocol 24 , MNs from two unrelated healthy controls (Con-1, Con-2) and from three patients harboring the C9ORF72 mutation (C9-1, C9-2, C9-3) and their isogenic pairs (C9-1Δ, C9-2Δ, C9-3Δ) (Fig. 2a). All lines efficiently differentiated into highly enriched neuronal cultures (>95% Tau + , devoid of astrocytes) that expressed the MN markers Isl1/2 (40-60%) and ChAT at 1 and 3 weeks post-plating, respectively (Fig. 2b,  Mutant C9ORF72 MNs display no difference in excitability. We and others have previously shown that mutant C9ORF72 iPSCderived MNs exhibit differences in excitability when using differentiation protocols which result in mixed cultures of neurons and glia 17,26,27 . In the present study using distinct protocols that produce neuronal only, MN-enriched cultures, whole-cell patchclamp recordings of MNs were undertaken to assess their firing properties (Fig. 3a). No significant differences were observed in the proportions of firing types (Fig. 3b)  Altered GluA1 subunit expression in mutant C9ORF72 MNs.
To begin to understand the molecular consequences of C9ORF72 mutation on MNs, we next undertook RNA-Seq analysis. Hierarchical clustering and heat map analysis showed segregation between control (including MN from an isogenic line, C9-3Δ) and mutants (C9-1, C9-3) (Fig. 4a). The high degree of similarity between C9-3 and C9-3Δ further indicates comparison between isogenic control overcomes the transcriptional heterogeneity observed in different iPSC lines and observed differential expression is driven by the mutation (Fig. 4a). We filtered differentially expressed genes using the following stringent criteria: (1) genes must have significant differential expression in both independent mutant C9ORF72 lines (p < 0.05) when compared to independent control lines, (2) there must be significant difference between C9-3 and C9-3Δ (p < 0.05), and (3) no significant difference observed between C9-3Δ and controls (p > 0.05). Using this approach 85 upregulated and 80 downregulated genes, collectively 165, were identified (Fig. 4b, c, d; Supplementary Data 1, 2). Gene ontology analysis revealed that genes involved in RNA processing, protein targeting, nuclear transport, and ion transport were significantly upregulated in mutant C9ORF72 MNs. Among downregulated genes in mutant C9ORF72 MNs there was enrichment for genes involved in synaptic transmission, cell-cell signaling, and regulation of phosphate metabolism (Fig. 4f, g). The finding that the GluA1 subunit was significantly upregulated (>2-fold) in C9ORF72 MNs (red star, Fig. 4d) is consistent with multiple lines of evidence implicating dysregulation of glutamate receptor homeostasis in the pathogenesis of ALS 28 . RNA-Seq analysis of the other AMPAR subunits and their auxiliary subunits showed that only GluA1 mRNA was differentially expressed in mutant C9ORF72 MNs ( Supplementary Fig. 3). Subsequent RT-qPCR confirmed increased expression of GluA1 in all week 3 mutant C9ORF72 MNs with gene-corrected lines showing reversal to control levels. Moreover, we also observed increases in GluA3 mRNA expression in mutant C9ORF72 MNs for two out of the three lines examined (Fig. 5a). Western blot analysis also confirmed increased levels of GluA1 in all week 3 mutant C9ORF72 MNs with gene-corrected lines displaying levels comparable to controls (Fig. 5b, c; Supplementary Fig. 8A). Importantly, relative expression of GluA2 mRNA was not altered in any of the lines (Fig. 5a). As would be expected, given the neuronal maturation profile of AMPAR subunit composition 29 , we did not observe altered expression of Ca 2+ -permeable AMPAR subunits in week 1 MNs ( Supplementary Fig. 4A).
The increased proportion of Ca 2+ -permeable AMPARs in mutant MNs at week 3 could also be due to inefficient RNA editing of GluA2 subunit in addition to increased expression of non-GluA2 subunits. We found, however, that GluA2 subunits were efficiently RNA-edited in all mutant cultures ( Fig. 5d; Supplementary Fig. 4B). Together, these findings demonstrate that mutant C9ORF72 MNs exhibit an increased expression of Ca 2+ -permeable GluA1 AMPAR subunits, but not inefficient editing of the GluA2 subunit. Con-1 Mutant C9ORF72 MNs are more vulnerable to excitotoxicity.
To determine whether the observed changes in the AMPAR subunit expression profiles had physiological consequences we next examined whether week 3 mutant MNs exhibited enhanced relative functional expression of Ca 2+ -permeable AMPARs. MNs from all lines expressed robust AMPA-evoked currents that were potentiated by the allosteric modulator, cyclothiazide, and blocked by AMPAR antagonist CNQX ( Supplementary Fig. 5A). The presence of one or more edited GluA2(R) subunits in the AMPAR complex confers Ca 2+ -impermeability, reduced single-

Gene ontology overrepresentation on upregulated genes
Notable GO terms # of genes p-value Expression relative to control lines  C9-3 Color key and histogram 0 0 10 Count c d e f g ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-02729-0 channel conductance, and reduced sensitivity to channel-blocking polyamines 29,30 . We assessed the proportion of Ca 2+ -permeable AMPARs in MNs by measuring the block of AMPAR-mediated currents by NASPM, a selective blocker of Ca 2+ -permeable AMPARs (Fig. 6a, b), and estimated their mean single-channel conductance using non-stationary fluctuation analysis (Fig. 6c, d).
Comparison of weeks 1 and 3 control MNs showed a reduction in both mean NASPM block and single-channel conductance that is consistent with a transition to a predominantly Ca 2+ -impermeable AMPAR population (Fig. 6b, d) and the maturation of AMPAR composition in native MNs 29 . In contrast, mutant C9ORF72 MNs at week 3 retained both a high sensitivity to NASPM block and an elevated single-channel conductance consistent with persistent expression of Ca 2+ -permeable AMPARs (Fig. 6b, d). Strikingly, the maturation of AMPAR composition in C9-1Δ, C9-2Δ, and C9-3Δ MNs reflected that of control MNs. Importantly the retention of Ca 2+ -permeable AMPARs in mutant C9ORF72 MNs was not due to a slower maturation of AMPAR composition as prolonging culture times for a further 2 weeks did not reveal any changes in the biophysical properties of AMPARs (Fig. 6b, d). Furthermore, we did not observe any impairment in the maturation of intrinsic properties or excitability of MNs across all lines, indicating that AMPAR-mediated-induced firing is unlikely to be a differential source of excitotoxicity in our cultures. These data demonstrate that the C9ORF72 mutation increases persistence in the relative expression of Ca 2+ -permeable AMPARs. Given the alteration in AMPAR composition we next examined whether C9ORF72 patient-derived MNs showed increased vulnerability to AMPA-induced excitotoxicity. Exposure to AMPA resulted in greater levels of cell death in week 3, but not week 1, mutant C9ORF72 MN cultures vs. control and C9-Δ cultures (Fig. 6e, f; Supplementary Fig. 5B, C). Importantly, enhanced susceptibility of mutant C9ORF72 MNs to excitotoxicity was not due to elevated AMPAR current flux as evidenced by no differences in AMPA-mediated current densities in each of the cell lines at week 3 ( Supplementary Fig. 5D). Excitotoxicity experiments performed in the presence of joro toxin (JTX), a selective blocker of Ca 2+ -permeable AMPARs 31 significantly reduced cell death of week 3 mutant MNs (Fig. 6f). These data therefore indicate the C9ORF72 mutation is associated with a higher susceptibility to excitotoxicity due an increase in relative expression of Ca 2+ -permeable AMPARs through altered AMPAR subunit expression.
To determine whether persistent expression of Ca 2+ -permeable AMPAR was selective to MNs we next generated cortical neurons from Con-1, C9-3, and C9-3Δ iPSC lines ( Supplementary Fig. 6A, B, C) 32,33 . The extent of NASPM block of AMPAR-mediated currents and the estimated single-channel conductance of AMPARs in C9-3 cortical neurons, unlike MNs, did not demonstrate an elevated expression of Ca 2+ -permeable AMPARs ( Supplementary Fig. 6D, E). Taken together these findings are consistent with a MN-specific cellular pathophysiology caused by the presence of mutant C9ORF72.
Elevated GluA1 expression in spinal cord of ALS patients. To evaluate whether findings in iPSC-derived neurons are recapitulated in pathological samples from ALS patients, AMPAR subunit expression was next examined by RNAscope®. Postmortem cervical spinal cord and prefrontal cortex (BA9) sections from three C9ORF72 mutation-carrying ALS patients and three age-matched controls with no known neurodegenerative disease were treated with fluorescence in situ hybridization (FISH) probes against AMPAR subunits and the positive control enzyme PPIB ( Supplementary Fig. 7). In agreement with previous reports of control samples displayed negligible GluA1 expression in adult MNs (Fig. 7a) 34,35 . However, high expression of GluA1 RNA was observed in the anterior horn neurons of all ALS patients (Fig. 7b). No difference was found in GluA2, 3, and 4 RNA expression between ALS and control spinal cord samples (Fig. 7c). Conversely, in prefrontal cortex sections, expression of all AMPAR RNA transcripts was similar between ALS and control samples (Fig. 7c, d). These data confirm in human postmortem tissue that the C9ORF72 expansion is associated with an increase in GluA1 RNA expression in spinal cord MNs but not in prefrontal cortical neurons.

Discussion
The ability to study human disease-causing mutations under genetically defined conditions allows causal rather than correlative associations to be ascribed. Using CRISPR technology to correct the C9ORF72 mutation from three patient-derived iPSC lines we show that the C9ORF72 mutation is associated with an increase in GluA1 AMPAR subunit expression, functional expression of Ca 2+ -permeable AMPARs, and MN vulnerability to excitotoxicity. C9ORF72 patient post-mortem tissue also demonstrates a marked selective upregulation in GluA1 RNA expression in spinal cord MNs. Both iPSC and pathological data further reveal that dysregulation of AMPAR expression is not found in cortical neurons and thus is MN specific.
There are three major hypotheses as to how C9ORF72 repeat expansion might be pathogenic: loss of, as yet still unknown, normal C9ORF72 function, sequestration of RNA binding proteins by RNA foci, and direct toxicity mediated by DPRs 25 . These processes are not mutually exclusive. We observed RNA foci and DPRs in mutant MNs whose expression was reduced or abolished by gene correction 17,18,36 . These findings, consistent with the idea that RNA-binding proteins sequestered by RNA foci in mutant C9ORF72 MNs dysregulate RNA metabolism, are supported by the differential transcriptional findings between mutant, control, and corrected MNs revealing an enrichment for genes involved in RNA metabolism, stress response, and nuclear transport. Collectively our data are consistent with recent studies including those from human pathological findings 17,18,37-39 . Moreover, our RNA-Seq data reveal new candidate genes that will require future validation and functional assessment. These include genes involved in synaptic transmission, protein targeting, and cell-cell signaling. In view of recent findings, it would be of interest to determine whether the increased expression of GluA1 is related to pathological RNA foci and/or DPRs 40 . Furthermore, it will be of considerable interest to study whether other genetic cause of ALS results in AMPAR dysregulation. In mixed cultures of neurons and glia derived from C9ORF72 repeat expansion patient iPSCs, MNs first exhibit intrinsic hyperexcitability 27 , then progress to a hypoexcitable state 17 . In contrast, here we have used a modified protocol for the generation of highly enriched neuronal cultures that contains negligible numbers of glial cells 24,41 . Under these conditions we found no difference in MN intrinsic excitability. A likely explanation for the lack of changes in excitability in the present study is that glialmediated non-cell autonomous mechanisms underlie changes in MN function. This hypothesis is supported by recent findings in murine cell culture models where exposure of wild-type MNs to conditioned medium derived from astrocytes expressing mutant SOD1 alters ion channel function and MN excitability 42 .
Multiple lines of evidence suggest that dysregulated glutamate homeostasis plays a role in ALS. These include patient studies showing reduced expression of glial glutamate transporter, enhanced CSF glutamate levels, and in vitro and in vivo studies that reveal selective MN vulnerability to glutamate or AMPARmediated, but not NMDAR, excitotoxicity 3,4,43 . AMPAR subunit composition is dynamic and changes in development and is dysregulated in disease. In early post-natal life, there is a switch from predominantly Ca 2+ -permeable to Ca 2+ -impermeable AMPARs which reflects increased GluA2 containing AMPARs; the dominant functional determinant of AMPAR-mediated Ca 2+ -permeability 29 . Indeed, this maturation is recapitulated in this study with control and gene-corrected iPSC MNs showing a shift from predominant Ca 2+ -permeable to Ca 2+ -impermeable AMPAR from 1 to 3 weeks in culture. In contrast, mutant C9ORF72 MNs display persistent high levels of Ca 2+ -permeability at 3 and 5 weeks in culture. The blockade of AMPA-mediated MN cell death by JTX treatment, a selective blocker of Ca 2+ -permeable AMPARs, confirms that C9ORF72 mutation susceptibility to excitotoxicity is due to an increase in relative expression of Ca 2+ -permeable AMPARs. Our findings are consistent with an earlier study that showed an enhanced vulnerability of iPSC-derived mutant C9ORF72 MNs to glutamatemediated excitotoxicity 18 .
One of the proposed mechanisms of MN death in ALS is through inefficient post-transcriptional editing of the Q/R site within the GluA2 subunit with the resulting AMPAR becoming Ca 2+ -permeable 16,44 . Our finding of efficient RNA editing does not support reduced editing as an explanation for increased Ca 2+ -permeability in mutant C9ORF72 MNs and is consistent with findings in native neurons and SOD1 familial cases 45,46 . Alternative explanations for increased Ca 2+ -permeability include disease-related alterations in AMPAR subunit composition. Experimental evidence in support of this idea is provided by findings in the SOD1 model of ALS, although these reports do not agree on the precise identity of dysregulated AMPAR subunits 8,10,11,47,48 . Against this background, we show multiple lines  of evidence that demonstrate upregulation in the GluA1 AMPAR subunit in C9ORF72 mutant MNs. The maintenance of functional expression of the GluA1 subunit strongly correlates with an increased Ca 2+ -permeable AMPAR identity in our cultures. Although GluA1 upregulation has previously been reported in the spinal cord and hippocampus of hSOD1G93A mice 11,49 , our study now demonstrates potential functional consequences of increased GluA1 in an experimental ALS model. Furthermore, our data demonstrate that mutant C9ORF72 cortical neurons possess predominantly Ca 2+ -impermeable AMPARs, indicating that mutant C9ORF72-dependent increases in Ca 2+ -permeable AMPARs are selective to MNs. Moreover, our findings are consistent with recent reports from C9ORF72 FTD patient iPSC neurons and post-mortem cortices that also do not show altered AMPAR subunit transcript regulation 37,50 . Furthermore, our earlier report that oligodendrocytes generated from the same  Importantly, our in vitro and post-mortem data suggest that GluA1 subunit upregulation may be contributory to selective MN death in ALS. Notwithstanding our observation of selective upregulation of GluA1 in post-mortem mutant C9ORF72 spinal cord MNs, we are unable to determine whether this is causal or a consequence of a separate disease pathology. It is also important to note that our study cannot exclude other AMPAR-related mechanisms that could potentially play a role in mutant C9ORF72-related MN pathology 18 . Furthermore, the degree to which other mechanisms may concurrently contribute to excitotoxicity via elevated glutamate levels at synapses, such as changes in glial glutamate transporter expression remains unknown 52,53 .
Our RNA-Seq study uncovered a mutant C9ORF72-associated increase in GRIA1 (GluA1) expression, which led to the hypothesis of increased Ca 2+ -permeable AMPA receptor currents and elevated vulnerability to AMPAR-mediated excitotoxicity. However, other genes whose expression are perturbed may also contribute to this phenotype. Our data also revealed a C9ORF72 mutation-associated reduction in PICK1 expression. PICK1 regulates the surface trafficking of GluA2-containing AMPARs 54 , and so a reduction in its expression may modify the proportion of GluA2-lacking AMPARs at the surface. Further genes within this data set may contribute to, or reflect other aspects of mutant C9ORF72-associated cellular phenotypes. For example, we observe a C9ORF72 mutation-associated increase in HSPB8, SERP1, and HSP90B1. The small heat shock protein HSPB8/ HSP22 plays a role in protein quality control and is capable of promoting the clearance of aggregates of C9ORF72 dipeptides, as well as mutant SOD1 and TDP-43 [55][56][57] . SERP1 (RAMP4) is induced by ER stress via a XBP-1-dependent process 58 whose role is to reduce protein stress by stabilizing proteins to facilitate glycosylation and promote translocation of proteins into the ER [59][60][61] . HSP90B1 (GRP94, GP96) is another ER chaperone involved in the unfolded protein response (UPR), and a ATF6 target 62 . Collectively, the induction of these genes point to a degree of ER stress and activation of the UPR. Conversely, increased ER Ca 2+ levels, elevated GRP78/BiP 63 , and vulnerability to tunicamycin-induced ER stress/UPR 64 have been observed in C9ORF72 mutant motor neurons. This may be triggered by the presence of C9ORF72 dipeptides and/or C9ORF72 hexanucleotide-induced disruption of nucleocytoplasmic transport 65 and even be exacerbated by increased GluA2-lacking AMPARs, since excessive Ca 2+ influx can influence ER stress 66 . Also of interest, GluA1 surface expression is known to be promoted by the UPR 67 , raising the possibility of positive feedback on excitotoxic signaling.
In summary, this study has combined patient iPSC lines and CRISPR/Cas9 technologies to demonstrate that the C9ORF72 repeat expansion recapitulates key pathological features of mutant C9ORF72 ALS and, with C9ORF72 ALS patient postmortem material, suggests a possible mechanism for MN vulnerability to AMPAR-mediated excitotoxicity.
Derivation of iPSCs. Dermal fibroblasts from patient and control individuals were obtained under full Ethical/Institutional Review Board approval at the University of Edinburgh. Fibroblasts were reprogrammed to iPSCs by either Sendai virus or retrovirus expressing OCT4, SOX2, C-MYC, and KLF4. iPSCs were maintained in Matrigel (BD Biosciences)-coated plastic dishes in E8 medium (Life Technologies) at 37°C and 5% CO 2 .
Gene editing of iPSCs. iPSC were dissociated into single-cell suspension using 1 × Accutase (Sigma). 8 × 10 5 cells were nucleofected with 2 µg each of Cas9-gRNA1 and Cas9-gRNA2 plasmid using the Amaxa 4D nucleofector system (program CA137) following the manufacturer's instructions. Transfected cells were plated onto Matrigel (BD)-coated dishes in the presence of E8 medium supplemented with ROCK inhibitor (10 µM). Upon confluence, cells were dissociated again into single cells and plated down at low density (2000 cells/10 cm dish) for clonal analysis. Individual clones were picked and screened for deletion of G 4 C 2 repeats by G 4 C 2 locus PCR and repeat primed PCR using an ABI 3130xl genetic analyzer.
A G 4 C 2 flanking PCR for screening targeted clones was also performed using Qiagen multiplex PCR kit. The PCR reaction was set up as per manufacturer's guide lines using the following primers G 4 C 2 Fw: 5′-CAAGGAGGGAAACAA CCG-3′ G 4 C 2 Rev: 6-FAM-5′-GGAAAGCAAGGAAGAGGC-3′. PCR products were separated on an ABI 3130xl analyzer (Life Technologies) and data were analyzed using GeneMarker software (Soft Genetics).
Southern blot. Southern blot analysis was used to analyze the C9ORF72 repeat expansion mutation as described 69 . DNA was isolated using Wizard® genomic DNA purification kit (Promega). 20 µg of DNA was digested overnight with Xba endonuclease, separated on a 0.8% agarose gel electrophoresis, gel was subsequently depurinated, denatured, and transferred overnight on to a positively charged nylon membrane (Roche). The membrane was UV cross-linked and hybridized overnight at 47°C with a digoxigenin (DIG)-labeled PCR probe. PCR probe was synthesized from genomic DNA using PCR DIG probe synthesis kit (Roche) using following primers South Fw: 5′-CTTGCAGATCAAAAGGCACA-3′ and South Rev: 5′-TTGACGCACCTCTCTTTCCT-3′. Following hybridization, membrane was washed and labeled with anti-DIG antibody and developed using CDP-Star reagent as per manufacturer's recommendation (Roche).
Quantitative RT-PCR. qPCRs were performed following previously optimized protocol 32 . Briefly, RNA from week-1 and week-3-old MNs were isolated using RNeasy kit (Qiagen). cDNA was synthesized from 250 ng of total RNA using DyNAmo TM cDNA synthesis kit (Thermo Scientific). Real-time quantitative PCR reactions were set up using DyNAmo ColorFlash SYBR Green qPCR kit (Thermo Scientific) and performed on CFX96 system (BioRad). Primers (forward and reverse) and annealing temperature used in this study are as follows: GRIA1: RNA sequencing. Total RNA from iPSC-derived MNs was assessed for quality (Agilent Bionalyzer) and quantity (Invitrogen Qubit) before library preparation. Illumina libraries were prepared from 1 µg of total RNA using TruSeq RNA Sample Prep Kit v2 with a 10 cycle enrichment step as per the manufacturer's recommendations. Final libraries were pooled in equimolar proportions before Illumina sequencing on a HiSeq 2500 platform using 75 base paired-end reads.
Reads were mapped to the primary assembly of the human (hg38) reference genome contained in Ensembl release 83. Alignment was performed with STAR, version 2.4.0i 70 . Tables of per-gene read counts were generated from the mapped reads with featureCounts version 1.4.6-p2 71 . Differential expression analysis was then performed using DESeq2 72 (R package version 1.10.0).
Immunocytochemistry. Cells were cultured depending on the experiment, washed once with PBS to remove media components, and fixed with 4% PFA for 20 min at room temperature. Cells were washed three times with PBS and blocking buffer containing 3% goat serum and 0.25% Triton X-100 in TBST (TBS-Tween) subsequently added for 30 min. Primary antibodies diluted in blocking buffer were then added and incubated overnight at 4°C. Cells were then washed three times with TBS-Tween, and corresponding secondary antibodies coupled with fluorophores were added and incubated at room temperature for 1 h. Coverslips were washed and mounted on glass slides with FluorSave TM reagent and imaged using Zeiss observer Z1. Images for cell count and neurite outgrowth were analyzed using ImageJ. Antibodies used in this study are as follows: Electrophysiology. Experiments investigating excitability were performed as previously described 26 . In short, whole-cell patch-clamp recordings were made in the current-clamp mode to determine firing properties of MNs. Recordings were performed using electrodes filled with internal pipette solution contained (in mM): 140 potassium methane-sulfonate, 10 NaCl, 1 CaCl 2 , 10 HEPES, 1 EGTA, 3 Mg 2 -ATP, and 0.4 Na 3 -GTP, pH 7.2-7.3 (adjusted with KOH; osmolarity adjusted to 300 mOsm with sucrose). Cells were bathed in an extracellular recording comprising (in mM): 127 NaCl, 3 KCl, 2 CaCl 2 , 1 MgCl 2 , 26 NaHCO 3 , 1.25 NaH 2 PO 4 , and 10 D-glucose (equilibrated with 95% O 2 and 5% CO 2 , pH 7.45; osmolarity, 310 mOsm). Recordings were low-pass filtered at 4 kHz and digitized at 50 kHz using pClamp software (Axon Instruments). F-I relationships were generated by recording the responses to a series of 1 s duration current steps of 5-65 pA (10 pA increments). For other electrophysiological experiments, recordings were performed as decribed 33 using electrodes filled with (in mM): 155 K-gluconate, 2 MgCl 2 , 10 Na-HEPES, 10 Na-PiCreatine, 2 Mg 2 -ATP, and 0.3 Na 3 -GTP, pH 7.3, 300 mOsm. Cells were bathed in an extracellular recording comprising (in mM): 152 NaCl, 2.8 KCl, 10 HEPES, 2 CaCl 2 , 10 glucose, pH 7.3, 320-330 mOsm. Current and voltage measurements were typically low-pass filtered online at 2 kHz, digitized at 10 kHz and recorded to computer using the WinEDR V2 7.6 Electrophysiology Data Recorder (J. Dempster, Department of Physiology and Pharmacology, University of Strathclyde, UK; www.strath.ac.uk/Departments/PhysPharm/). Non-stationary fluctuation analysis of slowly rising whole-cell currents evoked by AMPA were used to estimate the AMPAR unitary single-channel current. The analysis of AC and DC currents was performed as described previously in detail 73 with the exception that the AC-coupled signal was filtered with a 1-1200 Hz bandpass frequency.
All experiments were carried out at room temperature. Patch electrodes were manufactured from borosilicate glass to have resistances of 3-6 MΩ. Recordings were made with a series resistance less than 25 MΩ. Reported potential values are corrected for liquid junction potential.
Excitotoxicity assay. MNs (25,000 per well of a 96-well plate) were cultured for 1 or 3 week/s. Before the assay was performed, cells were further cultured overnight with minimal essential media containing neurobasal, B27, and N2. Live cells were labeled with 4 µM Calcein-AM (Invitrogen) for 5 min and imaged using Zeiss Observer Z1. Cells were then treated in the presence of cyclothiazide (100 µM) with either AMPA (100 µM) or AMPA/CNQX (30 µM) or AMPA/JTX (1 µM) for 24 h. Following treatment, cells were labeled again with Calcein AM and reimaged. Cell counts before and after treatment were analyzed using ImageJ.
Human post-mortem tissue analysis. Formalin-fixed paraffin-embedded cervical spinal cord and prefrontal cortex (BA9) from three ALS patients carrying the C9ORF72 repeat expansion and three age and sex-matched controls were sectioned at 5-μm thickness on to superfrost slides. BaseScope reagents (Advanced Cell Diagnostics) were used as per manufacturer's guidelines according to the original protocol 74 . In brief, following deparafinisation, tissue was incubated with hydrogen peroxide for 10 min at room temperature and target antigen retrieval was performed by submerging slides in BaseScope 1 × target retrieval reagent at 99°C in a Braun Multiquick FS 20 steamer for 15 min. The tissue was then permeabilised using BaseScope protease III at 40°C for 30 min. Probe hybridization was then performed by incubating the slides with four drops of custom-designed BaseScope probe for AMPAR subunits, negative control probe (DapB) or positive control (PPIB) probe for 2 h at 40°C. Following successive probe amplification steps, transcripts were detected using the BaseScope RED detection kit and slides were counterstained using 50% Gills haematoxylin and 0.02% ammonium water. The slides were then cleared in xylene and mounted with a 24 × 50 mm coverslip using two drops of VectaMount mounting medium. Sections were then imaged at 80× magnification on a NanoZoomer and relative number of transcripts was quantified by counting the dots observed in the cell manually by two independent neuropathologist. Three random fields were chosen and ten cells from each field were counted (in total 30 cells per post-mortem sample) and averaged for each patient. Data were plotted as mean number of transcripts (dots per cell) per post-mortem sample.
Data availability. The data that support the findings of this study are available from the corresponding author upon reasonable request.