ALS 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 (G4C2) intronic hexanucleotide in the C9ORF72 gene is the most common cause of familial ALS and accounts for 10% of sporadic ALS cases1,2.

Although the precise underlying mechanisms of cell-type specificity and death is unknown in ALS, enhanced MN vulnerability to glutamate-mediated excitotoxicity leading to intracellular Ca2+ overload has been implicated3,4,5,6,7. Rodent studies have suggested that stoichiometry of AMPAR subunits is central to excitotoxicity noting that GluA1, 3, and 4 AMPAR subunits are Ca2+-permeable and that GluA2 subunit typically is Ca2+-impermeable8,9,10,11,12,13,14. Inefficient RNA editing of the GluA2 subunit, resulting in Ca2+-permeability, in sporadic ALS patient samples further highlights the importance of improved understanding of the role of AMPARs in ALS15,16.

A powerful approach to interrogate the consequences of G4C2 expansion in human MNs is the study of iPSC-derived MNs from ALS patients carrying the C9ORF72 mutation17,18. 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 studies19,20,21,22. In this study, we have used CRISPR/Cas9 technology to excise G4C2 repeats from ALS patient-derived MNs to enable direct study of the pathogenic role of G4C2 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.


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 G4C2 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 G4C2 repeats (Fig. 1a). First, gRNAs (1 and 2) along with Cas9 plasmid were transfected in HEK-293 cells, which have 3 × G4C2 repeats, to test if the gRNAs were functional. PCR amplification of G4C2 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 G4C2 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 G4C2 repeats (C9-1, up to ~638 repeats; C9-2, up to ~960 repeats; C9-3, up to ~760 repeats; Supplementary Fig. 1A). Repeat primed PCR and G4C2 locus PCR were used to screen individual single-cell-derived clones. PCR of G4C2 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 G4C2 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 G4C2 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 G4C2 repeats confirmed absence of repeat expansion in the positive clone (Fig. 1d). We further sequenced the G4C2 locus of both the alleles in the positive clones to determine the deleted sequence. C9-1Δ contains 2 × G4C2 repeats in both alleles; C9-2Δ contains 1 × G4C2 repeat in one allele and absence of G4C2 repeat in other allele; and C9-3Δ contains 2 × G4C2 repeats in one allele and 1 G4C2 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 tools23. 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).

Fig. 1
figure 1

Excision of the G4C2 repeat expansion. a Strategy for generating C9-Δ from iPSC harboring G4C2 repeat expansion mutation. Two gRNAs were designed flanking the G4C2 repeats in intron-1 of the C9ORF72 gene to generate a double-strand break (depicted by //). b PCR of the G4C2 locus in HEK-293 cells transfected with Cas9-gRNA-1 and 2 plasmids generated truncated PCR product (blue arrow) depicting both gRNAs are functional. c Repeat primed PCR and PCR flanking G4C2 locus showed absence of hexanucleotide repeat expansion in C9-Δ iPSC. d Southern blot analysis of C9ORF72 alleles in control, mutant C9ORF72, and C9-Δ iPSC lines. e Sanger sequencing of G4C2 locus in C9ORF72 gene-edited lines confirmed excision of hexanucleotide repeat expansion in mutant allele. G4C2 repeats in wild-type allele were also excised by CRISPR/Cas9 editing strategy (C9-1: 6 repeats, C9-2: 2 repeats, C9-3: 1 repeat)

C9ORF72 mutant MNs display RNA foci

Having corrected the mutation, we next generated, using a recently reported protocol24, 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, c; Supplementary Fig. 2A, B). Three principal hypotheses have been proposed to explain how mutant C9ORF72 is pathogenic: haploinsufficiency leading to reduced levels of C9ORF72 protein, RNA foci binding essential RNA-binding proteins, and direct toxicity of dipeptide repeats (DPR) generated by repeat-associated non-ATG (RAN) translation of G4C2 repeats25. To evaluate whether expansion results in loss of protein, western blot analysis was undertaken. This revealed comparable amounts of C9ORF72 protein expression in MNs derived from all control, mutant, and gene-corrected iPSC lines (Supplementary Figs. 2C, 8B). Intranuclear RNA foci in MNs are a characteristic pathological finding in post-mortem tissue of C9ORF72 patients1. Using fluorescent in situ hybridization we confirmed abundant intranuclear RNA foci in mutant MNs that were absent in controls and gene-corrected lines (Fig. 2d) and absent in RNAse-treated cultures. Recent reports have observed DPRs in mutant C9ORF72 material and indeed protein lysates from mutant C9ORF72 MNs expressed higher levels of GA, GP, and GR RAN peptides when compared to control and gene-corrected MNs (Fig. 2e). Collectively, these findings demonstrate a direct causal link between C9ORF72 mutation and the presence of intranuclear foci and RAN peptides. Next, we undertook survival analysis to determine if the C9ORF72 mutation confers selective cellular vulnerability on MNs under basal conditions. Survival, measured using either single-cell longitudinal analysis or absolute cell counts, showed no difference between control and mutant MNs (Supplementary Fig. 2D). Thus, although mutant MNs display pathological RNA foci and DPRs these do not impact upon cell viability or survival under basal conditions.

Fig. 2
figure 2

Excision of the G4C2 repeat expansion reverses RNA foci pathology in mutant MNs. a Schematic representation of MN generation and timings of various experiments performed. b Representative images of Isl1/2 and Tau immunostainings in 1-week-old MNs (scale bar: 50 µm). c Percentage of Isl1/2+ MNs from various iPSC lines (data represented as mean ± s.e.m., Con-1, N = 3; Con-2, N = 3; C9-1, N = 3; C9-1Δ, N = 3; C9-2, N = 3; C9-2Δ, N = 2; C9-3, N = 3; C9-3Δ, N = 3). d RNA fluorescent in situ hybridization showed mutant C9ORF72 MNs exhibit intra-nuclear RNA foci (scale bar: 5 µm). e Representative dot-blots for poly-GA, poly-GP, and poly-GR DPRs in mutant C9ORF72 MNs. RA retinoic acid, SAG smoothened agonist, DAPT N-[N-(3,5-Difluorophenacetyl)-L-alanyl]-S-phenylglycine t-butyl ester, BDNF Brain-derived neurotrophic factor

Mutant C9ORF72 MNs display no difference in excitability

We and others have previously shown that mutant C9ORF72 iPSC-derived MNs exhibit differences in excitability when using differentiation protocols which result in mixed cultures of neurons and glia17,26,27. In the present study using distinct protocols that produce neuronal only, MN-enriched cultures, whole-cell patch-clamp 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) or intrinsic properties (Supplementary Table 1) of MNs derived from control or mutant iPSCs. Frequency–current (F–I) relationships were also comparable between control and mutant C9ORF72 MNs (Fig. 3c, d; Supplementary Table 1). Collectively, these findings reveal no difference in excitability between control and mutant C9ORF72 MNs in MN-enriched cultures lacking glial cells and therefore implicate aberrant neuron–glia interactions in the functional deficits reported previously17,26,27.

Fig. 3
figure 3

Excitability of MNs derived from mutant C9ORF72 patient and control iPSCs. a Evoked firing types observed in iPSC-derived MNs. MNs displayed either multiple, single, or no action potentials upon current stimulation. b Percentage of cells showing each category of firing did not differ between MNs derived from control (Con-1, n = 31; Con-2, n = 53), C9ORF72 mutant (C9-1, n = 17; C9-2, n = 32; C9-3, n = 73), and gene-edited (C9-1∆, n = 37; C9-2∆,n = 34; C9-3∆, n = 84) iPSC lines. c Raw traces showing repetitive action potential firing (>5 action potentials) in response to current injection in MNs derived from control, C9ORF72 mutant, and gene-edited iPSC lines. d F–I relationships generated from repetitively firing MNs reveal no significant differences between MNs derived from control (Con-1, n = 8; Con-2, n = 23), C9ORF72 mutant (C9-1, n = 3; C9-2, n = 5; C9-3, n = 22), and gene-edited (C9-1∆, n = 3; C9-2∆, n = 4; C9-3∆, n = 39) iPSC lines (mean ± s.e.m. values are plotted in the graph). Calibration bars: a: 200 ms, 20 mV; c: 400 ms, 20 mV

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).

Fig. 4
figure 4

RNA-Seq analysis on C9ORF72 mutant MNs. a Heat map and dendrogram depicting hierarchical clustering of RNA sequencing reads from controls (Con-1, Con-2), C9ORF72 mutant (C9-1, C9-3), and C9-3Δ MNs. b Venn diagram depicting identification of differentially expressed genes between two mutant C9ORF72 MNs and control MNs. Central intersect represents 165 dysregulated genes compared to both control and gene-corrected MNs. c Scatter plot showing comparison of transcriptome reads of two independent C9ORF72 mutant MNs and C9-3Δ MNs to controls. Red crosses denote significantly upregulated genes and blue crosses denote significantly downregulated genes (p < 0.05, analysis performed using DESeq2). These dysregulated genes in C9ORF72 mutant MNs are reversed to control levels in C9-Δ MNs. d, e Graph depicting the spread of differentially regulated, upregulated (d), and downregulated (e) genes in mutant C9ORF72 MNs, red star denotes the levels of GluA1. f, g Gene ontology studies on upregulated (f) and downregulated genes (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 ALS28. 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 composition29, we did not observe altered expression of Ca2+-permeable AMPAR subunits in week 1 MNs (Supplementary Fig. 4A).

Fig. 5
figure 5

The C9ORF72 mutation is associated with increased expression of GluA1 AMPAR subunits. a Relative mRNA expression of AMPAR subunits in all cultures examined at week 3 after differentiation (data represented as mean ± s.e.m., Con-1, N = 5; Con-2, N = 6; C9-1, N = 8; C9-1Δ, N = 4; C9-2, N = 7; C9-2Δ, N = 6; C9-3, N = 7; C9-3Δ, N = 6). Statistical comparisons; one-way ANOVA with uncorrected Fisher’s LSD. Significant p-values for GluA1 C9-1 vs. Con-1: 0.0039, C9-2 vs. Con-1: 0.0360, C9-3 vs. Con-1: 0.0503, C9-1 vs. Con-2: 0.0130, C9-2 vs. Con-2: 0.0163, C9-3 vs. Con-2: 0.0385. p-values for GluA3 C9-1 vs. Con-1: 0.0218, C9-3 vs. Con-1: 0.0017, C9-1 vs. Con-2: 0.0256, C9-3 vs. Con-2: 0.0024. b Representative immunoblot showing elevated levels of GluA1 in C9ORF72 mutant MNs. c Quantification of GluA1 protein level in C9ORF72 mutant MNs (data represented as mean ± s.e.m., Con-1, N = 3; Con-2, N = 5; C9-1, N = 6; C9-1Δ, N = 3; C9-2, N = 3; C9-2Δ, N = 3; C9-3, N = 6; C9-3Δ, N = 6). d Representative gel picture showing efficient GluA2 Q→R RNA editing in week 3 cultures across all the lines. Efficient GluA2 RNA editing results in RFLP amplicons of 116 and 66 bp, see also Supplementary Fig. 4B. Band at 81 bp would be observed if inefficient editing of the GluA2 subunit was present

The increased proportion of Ca2+-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 Ca2+-permeable GluA1 AMPAR subunits, but not inefficient editing of the GluA2 subunit.

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 Ca2+-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 Ca2+-impermeability, reduced single-channel conductance, and reduced sensitivity to channel-blocking polyamines29,30. We assessed the proportion of Ca2+-permeable AMPARs in MNs by measuring the block of AMPAR-mediated currents by NASPM, a selective blocker of Ca2+-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 Ca2+-impermeable AMPAR population (Fig. 6b, d) and the maturation of AMPAR composition in native MNs29. 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 Ca2+-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 Ca2+-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 Ca2+-permeable AMPARs.

Fig. 6
figure 6

Mutant C9ORF72 MNs exhibit Ca2+-permeable AMPARs receptors and greater vulnerability to excitotoxicity. a Sample whole-cell voltage-clamp recordings conducted at −74 mV depicting the block of AMPA (10 μM)-mediated currents (in the presence of cyclothiazide (10 μM)) by Ca2+-permeable AMPAR-selective blocker NASPM (3 μM) in week 5 MNs derived from a control, mutant C9ORF72, or C9-Δ lines. Scale bars: 5 s, 50 pA. b Mean ( ± s.e.m.) percentage NASPM block of AMPA currents in all MNs examined at week 1, 3, and 5 (data for week 1/week 3/week 5; Con-1, n = 6/10/6, N = 3/3/1; Con-2, n = 14/12/6, N = 2/2/2; C9-1, n = 9/9/14, N = 2/1/3; C9-1Δ, n = 5/8/5, N = 2/2/2; C9-2, n = 9/12/12, N = 2/3/3; C9-2Δ, n = 8/10/10, N = 3/4/4; C9-3, n = 7/11/11, N = 3/3/3; C9-3Δ, n = 10/10/7, N = 3/3/3) after differentiation (statistical comparisons; one-way ANOVA with Bonferroni’s post hoc test). c Representative non-stationary fluctuation analysis recordings of AMPAR-mediated currents in the AC-current (upper) and DC-current modes (lower) of week 5 MNs derived from a control, mutant C9ORF72, or the C9-Δ lines. Current-variance plots (bottom right) generated from the recordings presented were used to derive the γ. Scale bars; 5 s, 40 pA. d Mean estimated AMPAR γ in all lines examined at week 1, 3, and 5 (data week 1/week 3/week 5; Con-1, n = 15/15/12, N = 4/4/3; Con-2, n = 13/15/8, N = 2/2/2; C9-1, n = 10/11/21, N = 2/3/3; C9-1Δ, n = 7/6/8, N = 2/2/2; C9-2, n = 15/15/13, N = 3/3/2; C9-2Δ, n = 11/12/7, N = 3/3/2; C9-3, n = 12/18/19, N = 4/3/4; C9-3Δ, n = 17/23/8, N = 4/4/2) after differentiation (statistical comparisons; one-way ANOVA with Bonferroni’s post hoc test). e Representative images of Calcein AM-stained neurons derived from control, C9, and C9-Δ lines before and after 24-h treatment of AMPA (100 μM) in the presence of cyclothiazide (100 µM). Scale bar: 50 µm. Cyclothiazide is an AMPAR-selective allosteric potentiator used to prevent rapid AMPAR desensitization. f Mean percentage cytotoxicity of MNs treated with AMPA, AMPA + JTX, and AMPA + CNQX (data represented as mean ± s.e.m., Con-1, N = 6; Con-2, N = 5; C9-1, N = 5; C9-1Δ, N = 5; C9-2, N = 9; C9-2Δ, N = 5; C9-3, N = 5; C9-3Δ, N = 5) and at least 100 cells per experiment, Greek letter phi denotes significant difference (two-way ANOVA with uncorrected Fisher's LSD) in cell death with JTX treatment  when compared to AMPA treatment

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 Ca2+-permeable AMPARs31 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 Ca2+-permeable AMPARs through altered AMPAR subunit expression.

To determine whether persistent expression of Ca2+-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 Ca2+-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®. Post-mortem 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 post-mortem tissue that the C9ORF72 expansion is associated with an increase in GluA1 RNA expression in spinal cord MNs but not in prefrontal cortical neurons.

Fig. 7
figure 7

AMPAR subunit expression in post-mortem spinal cord of healthy patients and ALS patients with C9ORF72 repeat expansion. a Representative RNAscope® images showing the expression of AMPAR GluA1—four transcripts in the anterior horn of cervical spinal cord sections from an ALS patient (C9ORF72) and an age-matched healthy patient and reveals increased expression of GluA1 in ALS patient sample. Scale bar, 25 µm. b Quantification of the mean ± s.e.m. expression levels of AMPAR transcripts in spinal cord samples from three independent patients and controls. Note the lack of GluA1 transcript expression in control spinal cord samples and strong expression in patient samples. c Representative RNAscope images showing the expression of AMPAR GluA1—four transcripts in the prefrontal cortex from an ALS patient (C9ORF72) and an age-matched healthy patient. Scale bar, 25 µm. d Quantification of the mean ± s.e.m. expression levels of AMPAR transcripts in prefrontal cortex samples from three independent patients and controls showing equivalent expression of all transcripts


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 Ca2+-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 DPRs25. These processes are not mutually exclusive. We observed RNA foci and DPRs in mutant MNs whose expression was reduced or abolished by gene correction17,18,36. 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 findings17,18,37,38,39. 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 DPRs40. 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 hyperexcitability27, then progress to a hypoexcitable state17. In contrast, here we have used a modified protocol for the generation of highly enriched neuronal cultures that contains negligible numbers of glial cells24,41. 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 glial-mediated 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 excitability42.

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 AMPAR-mediated, but not NMDAR, excitotoxicity3,4,43. 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 Ca2+-permeable to Ca2+-impermeable AMPARs which reflects increased GluA2 containing AMPARs; the dominant functional determinant of AMPAR-mediated Ca2+-permeability29. Indeed, this maturation is recapitulated in this study with control and gene-corrected iPSC MNs showing a shift from predominant Ca2+-permeable to Ca2+-impermeable AMPAR from 1 to 3 weeks in culture. In contrast, mutant C9ORF72 MNs display persistent high levels of Ca2+-permeability at 3 and 5 weeks in culture. The blockade of AMPA-mediated MN cell death by JTX treatment, a selective blocker of Ca2+-permeable AMPARs, confirms that C9ORF72 mutation susceptibility to excitotoxicity is due to an increase in relative expression of Ca2+-permeable AMPARs. Our findings are consistent with an earlier study that showed an enhanced vulnerability of iPSC-derived mutant C9ORF72 MNs to glutamate-mediated excitotoxicity18.

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 Ca2+-permeable16,44. Our finding of efficient RNA editing does not support reduced editing as an explanation for increased Ca2+-permeability in mutant C9ORF72 MNs and is consistent with findings in native neurons and SOD1 familial cases45,46. Alternative explanations for increased Ca2+-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 subunits8,10,11,47,48. 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 Ca2+-permeable AMPAR identity in our cultures. Although GluA1 upregulation has previously been reported in the spinal cord and hippocampus of hSOD1G93A mice11,49, 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 Ca2+-impermeable AMPARs, indicating that mutant C9ORF72-dependent increases in Ca2+-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 regulation37,50. Furthermore, our earlier report that oligodendrocytes generated from the same mutant C9ORF72 patient iPSC lines used in this study express predominantly Ca2+-impermeable AMPARs further highlights MN-specific consequence of the C9ORF72 mutation51. Finally, in post-mortem studies, we show for the first time that mutant C9ORF72 spinal cord but not cortical neurons show selective upregulation of GluA1 transcript by quantitative RNAscope compared to age-matched control samples. We therefore propose that the mutant C9ORF72-dependent increase in Ca2+-permeable AMPAR-mediated excitotoxicity is MN specific.

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 pathology18. 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 unknown52,53.

Our RNA-Seq study uncovered a mutant C9ORF72-associated increase in GRIA1 (GluA1) expression, which led to the hypothesis of increased Ca2+-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 AMPARs54, 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-4355,56,57. SERP1 (RAMP4) is induced by ER stress via a XBP-1-dependent process58 whose role is to reduce protein stress by stabilizing proteins to facilitate glycosylation and promote translocation of proteins into the ER59,60,61. HSP90B1 (GRP94, GP96) is another ER chaperone involved in the unfolded protein response (UPR), and a ATF6 target62. Collectively, the induction of these genes point to a degree of ER stress and activation of the UPR. Conversely, increased ER Ca2+ levels, elevated GRP78/BiP63, and vulnerability to tunicamycin-induced ER stress/UPR64 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 transport65 and even be exacerbated by increased GluA2-lacking AMPARs, since excessive Ca2+ influx can influence ER stress66. Also of interest, GluA1 surface expression is known to be promoted by the UPR67, 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 post-mortem material, suggests a possible mechanism for MN vulnerability to AMPAR-mediated excitotoxicity.


Cloning of CRISPR/Cas9 constructs

Cas9 and gRNA were cloned into an expression plasmid pSpCas9(BB)-2A-GFP (pX458) following a published protocol68. Web tool ( was used to design CRISPR gRNA sequences and identify potential homologous off-target sites. Sequences of gRNAs are as follows: gRNA-1 5′-AACTCAGGAGTCGCGCGCTAGGG-3′, gRNA-2 5′-GGCCCGCCCCGACCACGCCCCGG-3′.

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% CO2.

Gene editing of iPSCs

iPSC were dissociated into single-cell suspension using 1 × Accutase (Sigma). 8 × 105 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 G4C2 repeats by G4C2 locus PCR and repeat primed PCR using an ABI 3130xl genetic analyzer.

Repeat primed PCR

The C9ORF72 repeat expansion mutation was confirmed in both iPSCs and MNs by repeat primed PCR using Qiagen multiplex PCR kit. PCR components are as follows: 1 × Multiplex Mastermix, 7% dimethylsulfoxide, 0.6 M Betaine, 7.6 µM FAM-labeled RP Fw (5′-CTGTAGCAAGCTCTGGAACTCAGGAGTCG-3′), 3.6 µM RP Rev: (5′-TACGCATCCCAGTTTGAGACGCCCCGGCCCCGGCCCCGGCCCC-3′), 11.6 µM Tail Rev (5′-TACGCATCCCAGTTTGAGACG-3′), 8 mM 7-deaza-2′-dGTP, and 200 ng DNA. Cycling conditions were performed as per the manufacturer’s recommendations with annealing temperature of 68 °C for 15 cycles, followed by 60 °C for a further 20 cycles. PCR products were separated on an ABI 3130xl analyzer (Life Technologies) and data were analyzed using GeneMarker software (Soft Genetics).

A G4C2 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 G4C2 Fw: 5′-CAAGGAGGGAAACAACCG-3′ G4C2 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 described69. 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).

MN differentiation

MN differentiation from iPSCs was performed using established protocol24 with minor modifications. iPSCs were dissociated into single cells using 1 × Accutase, 2 × 106 iPSCs were neuralized as a suspension culture using SB-431542 (20 µM), LDN-193189 (0.1 µM), and CHIR-99021 (3 µM) in N2/B27 media (0.5 × Neurobasal, 0.5 × Advanced DMEM/F12, 1 × Antibiotic-Antimycotic, 1 × Glutamax, 100 µM beta-mercaptoethanaol, 1 × B27, 1 × N2, and 10 µM ascorbic acid). On day 2, neural spheres were simultaneously patterned to spinal cord identity by treating with retinoic acid (RA, 0.1 µM) and smoothened agonist (SAG, 500 nM) along with SB-431542, LDN-193189, and CHIR-99021 in N2/B27 media for additional 5 days. On day 7, spheres continued to be cultured in RA and SAG and, also, brain-derived neurotrophic factor (BDNF, 10 ng/ml) in N2/B27 media to generate MN progenitors. On day 9, MN progenitors were cultured in the previous media in addition to DAPT (10 µM) for additional 5–7 days. At day 14–16, MN spheres were dissociated using 0.05% Trypsin-EDTA and plated on to a poly-ornithine, laminin (5 µg/ml), fibronectin (10 µg/ml), Matrigel (1:20)-coated dishes. Dissociated MNs were cultured in NB media (1 × Neurobasal, 1 × Glutamax, 1 × non-essential amino acids, 100 µM beta-mercaptoethanol, 1 × B27, 1 × N2, RA (1 µM), ascorbic acid (2.5 µM), BDNF (10 ng/ml), glial-derived neurotrophic factor (10 ng/ml), ciliary neurotrophic factor (10 ng/ml), insulin-like growth factor (10 ng/ml)). Cells were treated with Uridine (1 µM)/5-fluoro-2′-deoxyuridine (U/FDU) on day 1 to remove residual proliferating cells. Media was changed every 2–3 days.

Quantitative RT-PCR

qPCRs were performed following previously optimized protocol32. 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 DyNAmoTM 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: 5′-TGCTTTGTCGCAACTCACAGA-3′, 5′-GGCATAGACTCCTTTGGAGAAC-3′, 64 °C; GRIA2: 5′-CATTCAGATGAGACCCGACCT-3′, 5′-GGTATGCAAACTTGTCCCATTGA-3′, 58 °C; GRIA3: 5′-ACCATCAGCATAGGTGGACTT-3′, 5′-GGTTGGTGTTGTATAACTGCACG-3′, 58 °C; GRIA4: 5′-TTCCGAGCAGCGTGCAAATA-3′, 5′-GCATTGGGGCTGGTGTTATGA-3′, 58 °C; C9orf72: 5′-TGTGACAGTTGGAATGCAGTGA-3′, 5′-GCCACTTAAAGCAATCTCTGTCTTG-3′, 59 °C; ACTB: 5′-GTTACAGGAAGTCCCTTGCCATCC-3′, 5′-CACCTCCCCTGTGTGGACTTGGG-3′, 59 °C.

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 2.4.0i70. Tables of per-gene read counts were generated from the mapped reads with featureCounts version 1.4.6-p271. Differential expression analysis was then performed using DESeq272 (R package version 1.10.0).


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 FluorSaveTM 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: MAP2 1:1000 (Sigma M9942), SMI312 1:1000 (Abcam ab24574), Tau 1:2000 (DAKO A0024), ISL1/2 1:50 (DSHB 39.4D5), ChAT 1:250 (EMDMillipore AB144P).

Western blot analysis

MNs were cultured for 3 weeks and lysed directly with 2 × Laemmli buffer (125 mM Tris HCl pH 6.8, 4% SDS, 10% β−mercaptoethanol, 20% glycerol, and 0.004% bromophenol blue) and boiled for 10 min at 99 °C. Proteins were then run on a SDS-Poly acrylamide gel electrophoresis, blotted onto a nitrocellulose membrane. Membranes were then probed with respective primary and secondary antibodies and developed using ECL or ECL Advance (GE Healthcare). Blots were scanned and densitometry analysis was performed using ImageJ. Antibodies used GAPDH (1:5000), GluA1 (1:1000), C9ORF72 (1:2000).

RNA fluorescence in situ hybridization

FISH analysis was performed using an Alexa546-conjugated (GGCCCC)4 oligoneucleotide probe (IDT). Briefly, cells on glass coverslips were fixed in 4% paraformaldehyde for 15 min, permeabilized in 70% ethanol at 4 °C overnight, incubated with 50% formamide diluted with 2X SSC for 10 min at room temperature and hybridized for 2 h at 45 °C with the probe (0.16 ng/µl) in hybridization buffer consisting of 50% formamide, 10% dextran sulfate, yeast transfer RNA (1 mg/ml), salmon sperm DNA (1 mg/ml). The cells were washed twice with 50% formamide diluted with 2 × SSC for 30 min at 37 °C and once with 2X SSC at room temperature for 30 min. Immunocytochemistry was performed as described above.


Experiments investigating excitability were performed as previously described26. 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 CaCl2, 10 HEPES, 1 EGTA, 3 Mg2-ATP, and 0.4 Na3-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 CaCl2, 1 MgCl2, 26 NaHCO3, 1.25 NaH2PO4, and 10 D-glucose (equilibrated with 95% O2 and 5% CO2, 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 decribed33 using electrodes filled with (in mM): 155 K-gluconate, 2 MgCl2, 10 Na-HEPES, 10 Na-PiCreatine, 2 Mg2-ATP, and 0.3 Na3-GTP, pH 7.3, 300 mOsm. Cells were bathed in an extracellular recording comprising (in mM): 152 NaCl, 2.8 KCl, 10 HEPES, 2 CaCl2, 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;

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 detail73 with the exception that the AC-coupled signal was filtered with a 1–1200 Hz band-pass 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 protocol74. 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 analysis

Statistical analysis was performed using GraphPad Prism software. Data are represented as mean ± s.e.m. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, one-way ANOVA with Bonferroni’s multiple comparisons test or uncorrected Fisher’s LSD test.

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.