Impairment of Mitochondrial Calcium Buffering Links Mutations in C9ORF72 and TARDBP in iPS-Derived Motor Neurons from Patients with ALS/FTD

Summary

TDP-43 dysfunction is common to 97% of amyotrophic lateral sclerosis (ALS) cases, including those with mutations in C9orf72. To investigate how C9ORF72 mutations drive cellular pathology in ALS and to identify convergent mechanisms between C9ORF72 and TARDBP mutations, we analyzed motor neurons (MNs) derived from induced pluripotent stem cells (iPSCs) from patients with ALS. C9ORF72 iPSC-MNs have higher Ca²⁺ release after depolarization, delayed recovery to baseline after glutamate stimulation, and lower levels of calbindin compared with CRISPR/Cas9 genome-edited controls. TARDBP iPS-derived MNs show high glutamate-induced Ca²⁺ release. We identify here, by RNA sequencing, that both C9ORF72 and TARDBP iPSC-MNs have upregulation of Ca²⁺-permeable AMPA and NMDA subunits and impairment of mitochondrial Ca²⁺ buffering due to an imbalance of MICU1 and MICU2 on the mitochondrial Ca²⁺ uniporter, indicating that impaired mitochondrial Ca²⁺ uptake contributes to glutamate excitotoxicity and is a shared feature of MNs with C9ORF72 or TARDBP mutations.

Graphical Abstract

Highlights

• •Mutations in C9orf72 and TDP-43 reduce Ca²⁺ buffering capacity in ALS motor neurons
• •High levels of Ca²⁺-permeable glutamate receptors link C9orf72 and TDP-43
• •Low mitochondrial Ca²⁺ uptake leads to glutamate-induced cell death
• •Imbalance of MICU1 and MICU2 reduces mitochondrial Ca²⁺ uptake

In this study, Dafinca, Talbot, and colleagues provide a mechanistic link between mutations in C9orf72 and TARDBP in induced pluripotent stem cell-derived motor neurons from ALS patients. They show both mutations cause increased expression of Ca²⁺-permeable receptors, reduced cytosolic Ca²⁺ buffering, and identify that perturbations in mitochondrial calcium uptake through an imbalance of MICU1 and MICU2 link the mutations.

Introduction

Amyotrophic lateral sclerosis (ALS) is a rapidly progressive and fatal neurodegenerative disorder. It is now clear that there is considerable genetic heterogeneity underlying ALS, with multiple genes independently acting as autosomal dominant risk alleles in approximately 10% of cases, while the majority of apparently sporadic cases do not have identified risk variants. However, despite etiological heterogeneity, 97% of cases are characterized by TDP-43 cytoplasmic aggregates at autopsy. A key unresolved question is, therefore, how different genetic factors converge on TDP-43 pathology. Because mutations in TARDBP can also lead to ALS with TDP-43 pathology, a comparison of genetic models based on these and other common mutations may reveal common pathways.

The commonest single genetic risk factor for ALS, and the related disorder frontotemporal dementia (FTD), is a (GGGGCC)_n expansion in intron 1 of the C9ORF72 gene, which accounts for 35%–40% of familial ALS cases, 5%–7% of sporadic ALS, and approximately 40% of familial FTD cases (DeJesus-Hernandez et al., 2011, Majounie et al., 2012, Renton et al., 2011). The number of (GGGGCC)_n hexanucleotide repeats is typically >1,000 in affected individuals compared with <30 in healthy controls, although the threshold for triggering ALS has not been precisely established (van der Zee et al., 2013, Dedeene et al., 2019).

Motor neurons (MNs) derived from patient iPSCs have shown a range of phenotypes, including defects in autophagy, sequestration of RNA-binding proteins, endoplasmic reticulum (ER) stress, changes in gene transcription, and altered neuronal excitability, demonstrating that these models are potentially powerful tools to explore pathways relevant to the pathogenesis of ALS (Dafinca et al., 2016, Donnelly et al., 2013, Sareen et al., 2013).

Calcium signaling is an essential mechanism for normal neuronal function, transducing electrical activity into intracellular molecular signals, and there is evidence that low cytosolic Ca²⁺ buffering together with high Ca²⁺ influx during neurotransmission contribute to the vulnerability of MNs in ALS. Excessive Ca²⁺ influx during glutamatergic neurotransmission has previously been reported in both rodent models as well as iPS-derived neurons as a result of increased Ca²⁺ permeability of AMPA receptors (Couratier et al., 1993, Rothstein et al., 1990, Rothstein et al., 1992, Selvaraj et al., 2018). The regulation of elements involved in Ca²⁺ transients is therefore critical to the normal functioning of neurons, and one of the main factors buffering Ca²⁺ during signaling is the uptake by mitochondria of significant amounts of Ca²⁺ from the cytosol.

In this study, using iPS-derived MNs with the C9ORF72 HRE, we have identified significant elevations in Ca²⁺ transients on depolarization and repeated firing in response to glutamate, and sustained high concentrations of cytosolic Ca²⁺. We demonstrate this is due to increased calcium permeability of AMPA and NMDA receptors, deficiency in mitochondrial Ca²⁺ uptake, and reduced calcium buffering capacity, all of which are reversed by genome editing using CRISPR/Cas9-mediated homologous recombination. We then compared these findings with iPS-derived MNs from ALS patients carrying the M337V mutation or I383T in the TARDBP gene, demonstrating that impairment of mitochondrial Ca²⁺ uptake is a common feature of ALS due to C9ORF72 mutations and TARDBP mutations, while low ER Ca²⁺ levels correlate with low Ca²⁺ response in TARDBP MNs on depolarization.

Results

Depolarization Induces High Intracellular Ca²⁺ in C9ORF72 MNs and High Recovery Times, Rescued by Genome Editing

Fibroblasts from three healthy controls and three different ALS/FTD patients carrying mutations in C9ORF72 were reprogrammed to induced pluripotent stem cells and differentiated to MNs using a previously published protocol, with modifications (Maury et al., 2015) (Figure S1). To analyze the response of C9ORF72 MNs to depolarization and their Ca²⁺ buffering capacity, the neurons were exposed to 50 mM KCl for 30 s during live imaging of intracellular Ca²⁺ labeled by Fura 2-AM (Figures 1A and 1B). We measured the amplitude of Ca²⁺ peaks elicited and found that C9ORF72 patient MNs released up to three times more Ca²⁺ into the cytosol compared with healthy MNs and 1.5 times more than the CRISPR/Cas9-corrected MNs (Figure 1C). Furthermore, we found that clearance of Ca²⁺ from the cytosol after depolarization was significantly delayed in C9ORF72 MNs compared with healthy and CRISPR/Cas9-corrected lines (Figure 1D).

Figure 1.

C9ORF72 MNs Have High Ca²⁺ Transients and Recovery Times in Response to KCl

(A and B) Representative (A) images and (B) traces for Fura 340/380-loaded MNs after stimulation with 50 mM KCl.

(C) The amplitude of the first Ca²⁺ peak after stimulation with 50 mM KCl is higher in C9ORF72 MNs compared with two healthy controls and two edited lines of MNs (C9-Ed) (^∗p < 0.05, ^∗∗∗p < 0.001); n = 25–73 neurons per iPS clone from at least two independent differentiations.

(D) Recovery of Ca²⁺ levels to baseline is delayed in C9ORF72 MN (^∗∗∗p < 0.001); n = 25–127 neurons from two to four independent differentiations.

(E) Representative traces after ionomycin stimulation. (F) Total Ca²⁺ levels measured after stimulation with 5 μM ionomycin show significantly increased levels in all C9ORF72 MNs as well as in the C9-Ed compared with healthy controls (^∗∗∗p < 0.01); n = 25–78 neurons from two to three independent differentiations. Data are shown as mean ± SEM with one-way ANOVA and Dunnet's post hoc tests.

See also Figure S1.

We then investigated the total level of Ca²⁺ ions in the stores of C9ORF72 MNs using ionomycin, finding high levels of Ca²⁺ in both C9ORF72 patient MNs and C9-Ed corrected MNs compared with the healthy controls, indicating that the intracellular levels of Ca²⁺ may account for the higher Ca²⁺ transients we detect upon depolarization (Figures 1E and 1F).

These results show that C9ORF72 MNs release more Ca²⁺ into the cytosol after depolarization and have long delays in buffering cytosolic Ca²⁺, but these phenotypes can be reversed by excision of the expanded repeat in the CRISPR/Cas9-edited MNs.

Glutamate Induces High-Frequency/Low-Amplitude Ca²⁺ Waves in C9ORF72 MNs and Increases Their Susceptibility to Cell Death

Glutamate excitotoxicity, driven by a failure of reuptake of the neurotransmitter by adjacent astrocytes, is a potential mechanism of degeneration in ALS and may be implicated in neuronal death. To investigate the response of C9ORF72 MNs to glutamate exposure, we examined the amplitude and recovery of Ca²⁺ waves after stimulation with 100 μM glutamate for 30 s (Figure 2A). MNs from C9ORF72 patients generated significantly lower Ca²⁺ peaks compared with healthy and genome-edited controls (C9-Ed) (Figure 2B). Despite low release and intake of Ca²⁺, the C9ORF72 MNs required significantly longer time to recover to pre-stimulation cytosolic Ca²⁺ levels (Figure 2C).

Figure 2.

Glutamate Induces Frequent, Low Ca²⁺ Waves in C9ORF72 MNs and Ca²⁺ Levels Remain Increased in the Cytosol Contributing to Glutamate-Induced Cell Death

(A) Representative Fura 340/380 traces in response to glutamate stimulation.

(B) Amplitude of Ca²⁺ peaks in C9ORF72 MNs is lower than healthy controls and C9-Ed MNs (^∗∗p < 0.01, ^∗∗∗p < 0.001); n = 23–78 neurons from two independent differentiations.

(C) Recovery of Ca²⁺ levels to baseline is significantly increased in C9ORF72 MNs compared with healthy controls and is improved in C9-Ed MNs (^∗∗∗p < 0.001); n = 26–151 neurons from two to four independent differentiations.

(D) Representative Fura 340/380 traces in response to glutamate stimulation followed by CNQX and glutamate.

(E and F) (E) Ca²⁺ amplitude and (F) Ca²⁺ recovery times are significantly higher in C9orf72 MNs after blocking the AMPA receptors with CNQX (^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗p < 0.001); n = 11–24 neurons from two independent differentiations.

(G) Distribution of Ca²⁺ peaks elicited by glutamate stimulation in healthy controls, C9ORF72 MNs, and C9-Ed.

(H) Immunoblotting shows low levels of calbindin in C9ORF72 MNs compared with healthy and genome-edited C9-Ed (^∗∗p < 0.01). Data are shown as mean ± SEM with one-way ANOVA and Sidak's post hoc tests.

Using 100 μM kainate, we stimulated the AMPA/kainate receptors, but did not observe consistent differences between C9ORF72 patients and healthy controls (Figures S2A and S2B). To investigate the contribution of NMDA receptors to Ca²⁺ handling, the AMPA receptors were blocked by CNQX and we found that the amplitude of Ca²⁺ was higher in C9ORF72 MNs (Figures 2D and 2E), and that their recovery time to baseline Ca²⁺ levels was up to three times longer, indicating that the NMDA receptors may contribute to increased Ca²⁺ permeability in C9ORF72 MNs (Figure 2F).

One factor contributing to the delay in recovery is the increased excitability we detected in C9ORF72 MNs. After stimulation with glutamate, the majority of healthy cells elicited two Ca²⁺ peaks, while C9ORF72 MNs elicited up to seven peaks, with very few MNs showing only two peaks (Figure 2G). Prolonged exposure to lower levels of glutamate induced significantly increased cell death in C9ORF72 MNs (Figure S2C).

Calcium binding proteins have a critical role in controlling cytosolic Ca²⁺ concentration by transporting Ca²⁺ ions across membranes, modulating Ca²⁺ levels, and acting as Ca²⁺ sensors. We analyzed calbindin, calreticulin, calpain-1, and calnexin expression and found a significantly reduced expression of calbindin in all C9ORF72 MNs compared with controls (Figures 2G and S2D).

Therefore, despite a lower initial influx of Ca²⁺ in the cytosol in response to glutamate, C9ORF72 MNs show a higher frequency of Ca²⁺ waves and a long delay in clearing released Ca²⁺ from the cytosol, partially due to low levels of cytosolic calcium buffering, which is likely to contribute to glutamate-induced cell death.

TARDBP MNs Show Hypoexcitability on Depolarization and Long Recovery Times after Glutamate Stimulation

To assess whether alterations in Ca²⁺ handling are specific to C9ORF72 mutations, we studied the buffering capacity of TARDBP mutant iPS-derived MNs (Figure S3).

When depolarized by 50 mM KCl for 30 s, TARDBP MNs showed significantly lower Ca²⁺ amplitudes in response to depolarization and significantly increased recovery times (Figures 3A–3D). The amplitude of Ca²⁺ peaks was up to two times higher in TARDBP MNs after stimulation with glutamate (Figures 3E and 3F), and no differences were detected in the responses after CNQX block of the Ca²⁺-permeable AMPA receptors between controls and TDP-43^M337V MNs. However, blocking the NMDA receptors with 10 μM memantine resulted in significantly higher Ca²⁺ levels in TDP^M337V compared with healthy controls, while no differences were observed between C9orf72 and healthy controls or C9-Ed (Figure S4A). Similar to C9ORF72 MNs, we recorded a significant delay in the recovery to baseline Ca²⁺ levels after glutamate exposure, up to two times longer than healthy controls (Figure 3G). The frequency of Ca²⁺ peaks was also higher in TDP-43^M337V MNs compared with controls (Figure 3H), but we did not detect differences in the expression of any calcium buffering proteins in TDP-43^M337V or TDP-43^I383T MNs (Figures S4B and S4C).

Figure 3.

TARDBP MNs Show Low Levels of Cytosolic Ca²⁺ on Depolarization and High Cytosolic Ca²⁺ after Glutamate Stimulation

(A and B) (A) Fluo-4 AM-loaded MNs responding to KCl and (B) representative Ca²⁺ traces.

(C) Depolarization with 50 mM KCl induces low Ca²⁺ waves in TARDBP iPS-derived MNs compared with healthy controls (^∗∗p < 0.01, ^∗∗∗p < 0.001) n = 33–112 neurons from two to four independent differentiations.

(D) Recovery to baseline Ca²⁺ levels is slower in TARDBP MNs compared with healthy controls (^∗∗∗p < 0.001); n = 33–51 neurons from two to four independent differentiations.

(E) Ca²⁺ traces after stimulation with 100 μM glutamate and CNQX.

(F) Higher amplitude Ca²⁺ waves are detected in TDP-43^M337V compared with the healthy controls (^∗∗∗∗p < 0.001); n = 30–142 neurons from two to four independent differentiations. No differences are detected after CNQX; n = 30–83 neurons from two independent differentiations.

(G) Recovery of baseline Ca²⁺ levels after stimulation with glutamate is higher in TARDBP MNs (^∗∗∗p < 0.001); n = 52–128 neurons from two to four independent differentiations.

(H) Distribution of the number of peaks elicited in response to glutamate stimulation. Data are presented as mean ± SEM with one-way ANOVA and Sidak's post hoc tests.

See also Figure S3.

These results confirm that, convergent with C9ORF72 mutations, TARDBP MNs demonstrate significant delays in buffering cytosolic Ca²⁺ and glutamate hyperexcitability. We also detect alterations in Ca²⁺ responses after depolarization, which show low levels in TARDBP and high levels in C9ORF72 MNs compared with healthy controls.

Ca²⁺-Permeable AMPA-R and NMDA-R Subunits Are Upregulated in C9ORF72 and TDP-43^M337V iPS-Derived MNs

Previous studies have shown that glutamate receptors are expressed at a higher level in C9ORF72 iPS MNs, along with age-dependent glutamate sensitivity and hyperexcitability (Bursch et al., 2019, Devlin et al., 2015, Donnelly et al., 2013, Sareen et al., 2013, Selvaraj et al., 2018). To analyze whether the expression of these receptors is altered and could account for increased vulnerability of C9ORF72 and TDP-43^M337V MNs to glutamate excitotoxicity, we used RNA sequencing data generated from 28-day-old C9-02 iPSC-derived MN cultures as well as from the edited clones derived from the same cell line (C9-Ed) and 30-day-old TDP-43^M337V MNs (three clones, 03-03, 03-04, and 03-06). RNA sequencing was performed in TDP-43^M337V and key results were validated in TDP-43^I383T MNs.

Differential expression analysis demonstrated a significant increase in the expression of AMPA receptor GRIA3 mRNA in C9ORF72 mutant iPSC-derived MNs, validated by qPCR (Figure 4A). Subunits 1 and 3 are known to increase the calcium permeability of this receptor and are balanced by the expression of AMPA receptor subunit 2, which has been shown to decrease calcium permeability. AMPA subunit 2 (GRIA2) was also found to be upregulated in C9ORF72 MNs compared with C9-Ed, suggesting a compensatory mechanism to balance the increase in Ca²⁺ permeability from GRIA1 and GRIA3. In TDP-43^M337V MNs analyzed at a similar time point, we detected significantly increased expression of AMPA receptor GRIA4, which is a Ca²⁺-permeable ligand-gated ion channel mediating excitatory synaptic transmission in the presence of L-glutamate (Figures 4B and S4B).

Figure 4.

Ca²⁺ Permeable Subunits of the Glutamatergic Receptors AMPA and NMDA Are Overexpressed in C9ORF72 MNs and TDP-43^M337V MNs

(A) RNA sequencing (RNA-seq) results in C9ORF72 and C9-Ed iPS-derived MNs for AMPA receptor subunits 1–4 (GRIA1–4) and qPCR validation of higher expression of GRIA1, GRIA2, and GRIA3 in C9ORF72 MNs (two or three patients pooled together) compared with C9-Ed (one or two clones) (^∗∗p < 0.01, Student's t test).

(B) RNA-seq of AMPA receptor subunits 1–4 (GRIA1–4) in TDP-43^M337V and qPCR validation of GRIA4 showing increased expression in patients versus healthy controls (^∗p < 0.05, Student's t test).

(C) RNA-seq results for NMDA receptor subunits 1–3A (GRIN1–3) and qPRC validation of GRIN2A elevation in C9ORF72 MNs (three patients) compared with C9-Ed (^∗p < 0.05, Student's t test).

(D) RNA-seq results for NMDA receptor subunits 1–3A in TDP-43^M337V MNs compared with healthy controls and qPCR validating high levels of GRIN1 and GRIN2B in TDP-43^M337V (^∗p < 0.05, Student's t test).

(E) RNA-seq results for metabotropic glutamate receptors 1–4 (GRM1–4) and qPCR validating increased expression of GRM2 and GRM4 in C9ORF72 MNs (two or three patients) compared with C9-Ed (^∗p < 0.05, Student's t test).

(F) RNA-seq results for GRM1–4 in TDP-43^M337V and qPCR validating increased expression of GRM3 in TDP-43^M337V compared with healthy controls (^∗∗∗∗p < 0.001, Student's t test). Normalized transcript per million (TPM) values are reported on the y axis and allow estimation of relative abundances of the transcription of each gene. Adjusted p values (p. adj) are derived from the differential expression analysis performed using DESeq2 and represent Benjamini-Hochberg corrected p values. Asterisks indicate p values below 0.05. qPCR data are presented as mean ± SEM. N = 3 independent differentiations for all iPS lines shown.

See also Figure S3.

In C9ORF72 MNs, significantly increased expression was seen in NMDA-R subunit 2A (GRIN2A), and a similar non-significant trend was observed in subunits 1 and 2B (GRIN1, GRIN2B) (Figure 4C). GRIN1 and GRIN2B were also significantly increased in TDP-43^M337V MNs and validated by qPCR (Figures 4D and S4B).

In addition to fast ionotropic receptors, glutamate also acts on metabotropic receptors (mGluR or GRM), which generate slower postsynaptic responses. GRM4 transcript expression was significantly increased in C9ORF72 iPSC-derived MNs compared with edited controls (Figure 4E). We also showed by qPCR a similar elevation in GRM2, both of which are believed to attenuate glutamate toxicity via a decrease in cAMP levels (Figure 4E). In TDP-43^M337V, we detected an upregulation of GRM3, which is also implicated in the inhibition of cAMP (Figure 4F).

These results show a consistent upregulation of Ca²⁺-permeable subunits of glutamate receptors in both C9ORF72 and TARDBP MNs, but suggest that the two mutations have different regulatory mechanisms for Ca²⁺ sensitivity.

Mitochondria Do Not Efficiently Clear Ca²⁺ from the Cytosol in C9ORF72 MNs due to Low Levels of MCU and MICU2

Increased cytosolic Ca²⁺ initiates a series of events, which include mitochondrial Ca²⁺ overload (Peng et al., 1998), generation of reactive oxygen species, and release of cytochrome c. To analyze whether deficiencies in Ca²⁺ uptake into mitochondria account for the observed delays in clearing cytosolic Ca²⁺, we investigated the dynamics of mitochondrial Ca²⁺ after stimulation with glutamate and caffeine of neurons loaded with Rhod 2-AM (Figure 5A). C9ORF72 mutant MNs exposed to 100 μM glutamate showed significantly reduced uptake of Ca²⁺ into mitochondria compared with healthy and C9-Ed MNs (Figures 5B and 5C). When stimulated with 20 mM caffeine, which opens ryanodine receptors on the ER, the uptake of Ca²⁺ into the mitochondria of C9ORF72 mutant MNs was reduced by half compared with healthy MNs and C9-Ed MNs (Figure 5D).

Figure 5.

Impairment of Mitochondrial Ca²⁺ Buffering Due to Reduced MCU and MICU2 Levels in C9ORF72 MNs

(A and B) (A) Representative image of Rhod 2AM-loaded neurons and (B) representative Ca²⁺ traces.

(C) Ca²⁺ uptake in the mitochondria is lower in C9ORF72 MNs compared with the C9-Ed MNs and healthy control MNs on glutamate stimulation; n = 15–62 neurons from two to three independent differentiations.

(D) Mitochondrial Ca²⁺ uptake in response to caffeine is significantly reduced in C9ORF72 MNs (^∗∗∗p < 0.001); n = 8–23 neurons from two to three independent differentiations.

(E) Representative Ca²⁺ traces recorded in response to glutamate stimulation and CCCP.

(F) Amplitude of Ca²⁺ released from mitochondria is significantly lower in C9ORF72 MNs (^∗∗p < 0.01, ^∗∗∗p < 0.001); n = 27–39 neurons from two independent differentiations.

(G and H) Immunoblotting of MCU (G) shows reduced levels in C9ORF72 MNs (H) (^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗p < 0.001).

(I–K) (I) RNA-seq showing MICU2 levels are downregulated in C9ORF72 iPS-derived MNs compared with C9-Ed and (J) immunoblotting confirming reduced expression of MICU2 in C9-02 and (K) qPCR confirming reduced MICU2 expression (^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗p < 0.001). Data are shown as mean ± SEM with one-way ANOVA and Sidak's post hoc tests.

See also Figure S5.

To confirm these findings, we analyzed total cytosolic Ca²⁺ during stimulation with glutamate followed by immediate addition of carbonyl cyanide m-chlorophenyl hydrazine (CCCP) to release mitochondrial Ca²⁺ (Figures 5E and 5F). The addition of the protonophore CCCP immediately after cell stimulation has been shown in many cell types to produce a large increase in cytosolic Ca²⁺ due to release of the Ca²⁺ accumulated by mitochondria (Herrington et al., 1996, Werth and Thayer, 1994). Using this method, we recorded significantly smaller Ca²⁺ peaks in the C9ORF72 patient MNs after CCCP stimulation compared with healthy control MNs, which released at least double the amount of mitochondrial Ca²⁺ (Figures 5E and 5F).

Mitochondria take up Ca²⁺ from the cytosol via the Ca²⁺ uniporter (MCU), which is a specific calcium pathway driven by the difference in electrical potential across the mitochondrial membrane. We found a significant decrease in its protein expression in C9ORF72 MNs (Figures 5G and 5H), which is not due to fewer mitochondria present in the C9ORF72 MNs verified using TOMM20 immunoblotting (Figure S5). Using RNA sequencing, we detected that one of the two gatekeeping proteins regulating MCU activity, MICU2, was significantly downregulated in C9-02 and rescued by the correction of the mutation in C9-Ed (Figure 5I), which we confirmed by immunoblotting and qPCR (Figures 5J and 5K).

These results suggest that the reduced uptake of Ca²⁺ from the cytosol after stimulation in C9ORF72 MNs is partly due to deficiency in mitochondrial Ca²⁺ buffering as a result of low expression of the mitochondrial Ca²⁺ uniporter and its regulatory MICU2.

TARDBP MNs Have Reduced Mitochondrial Ca²⁺ Uptake Capacity

Mitochondrial Ca²⁺ uptake was similarly assessed in TARDBP MNs. After stimulation with CCCP, we detected significantly lower levels of Ca²⁺ released from mitochondria in mutant lines compared with healthy controls (Figures 6A and 6B). RNA sequencing in TDP-43^M337V MNs showed a significant upregulation of the mitochondrial Ca²⁺ uptake regulator MICU1, which we confirmed by qPCR, with no differences in the expression of MICU2 (Figure 6C). The I383T mutation was not included in RNA sequencing and MICU1 levels were analyzed by western blotting (Figure S6A). In contrast to the C9ORF72 MNs, normal levels of MCU were detected in all TARDBP iPS-derived MNs (Figure 6D).

Figure 6.

TARDBP MNs Show Reductions in Mitochondrial Ca²⁺ Uptake with Increased Expression of MICU1

(A) Ca²⁺ traces after stimulation with glutamate followed by perfusion with CCCP.

(B) Mitochondrial Ca²⁺ release after CCCP is significantly lower in TARDBP MNs (^∗∗∗p < 0.001); n = 22–75 neurons from two to three independent differentiations.

(C) RNA-seq shows upregulation of MICU1 and validation by qPCR (^∗p < 0.05).

(D) Immunoblotting shows no significant differences in the expression of MCU between healthy controls and TDP-43 MNs.

(E and F) MitoTracker imaging shows higher intensity in TDP-43^I383T (E), but no significant difference in TDP-43^M337V (F). Data are presented as mean ± SEM with one-way ANOVA and Sidak's post hoc tests.

See also Figure S6.

To investigate the deficits in mitochondrial function in comparison with the C9ORF72 MNs, we analyzed mitochondrial membrane potential using MitoTracker (Figure 6E). Our previous work had shown that C9ORF72 iPS-derived MNs have significant morphological alterations and reduced potential of the mitochondria, but in TDP-43^M337V MNs assessment of mitochondrial membrane potential revealed no differences between the healthy controls and patient MNs (Figure 6F).

These results show that impaired mitochondrial Ca²⁺ buffering is a common pathomechanism between C9ORF72 and TARDBP mutations and that an imbalance between the regulatory proteins MICU1 and MICU2 plays an important role in the reduction of mitochondrial Ca²⁺ uptake.

C9ORF72 and TARDBP MNs Have Low ER Ca²⁺ Release

The regulation of Ca²⁺ signaling is one of the most important functions of ER-mitochondria crosstalk (Filadi et al., 2017). We previously reported high levels of Ca²⁺ in the ER of C9ORF72 iPS-derived MNs (Dafinca et al., 2016). To investigate ER Ca²⁺ homeostasis in TARDBP MNs and C9ORF72 MNs, the neurons were perfused with carbamylcholine to activate the acetylcholine receptors and stimulate ER Ca²⁺ release through IP₃R (Figure 7A). Both TARDBP MNs and C9ORF72 MNs responded with significantly lower release of ER Ca²⁺ compared with healthy controls (Figure 7A). The levels of IP₃Rs were not different between the genotypes, suggesting that the differences in Ca²⁺ release in TARDBP MNs are mainly due to low levels of ER Ca²⁺ stores that we detected after stimulation with 10 μM thapsigargin (Figures 7B, S6B, and S6C). We also investigated whether altered regulation of ER Ca²⁺ induces ER stress, but we did not detect increases in GRP78/BiP in TDP-43^M337V MNs (Figure S6D). Furthermore, we analyzed the localization of TDP-43 to investigate a potential mechanism where cytoplasmic mislocalization contributes to the similar phenotypes observed in C9ORF72 and TARDBP MNs, but we did not detect mislocalization in any of the patient lines at baseline (Figure S7).

Figure 7.

ER Ca²⁺ Release Is Impaired in TARDBP and C9ORF72 MNs

(A) Ca²⁺ traces for TDP-43^M337V, TDP-43^I383T, C9-02, C9-Ed, and healthy MNs after stimulation with 100 μM carbachol, and quantification of the peaks shows reduced Ca²⁺ amplitudes in TARDBP and C9ORF72 MNs (^∗∗p < 0.01, ^∗∗∗p < 0.001, ^∗∗∗∗p < 0.0001); n = 20–111 neurons from two to four independent differentiations.

(B) Measurement of ER Ca²⁺ levels shows significantly reduced levels of Ca²⁺ in TARDBP MNs (^∗∗p < 0.01, ^∗∗∗p < 0.001); n = 61–86 neurons from three independent differentiations.

(C) Immunoblotting shows a significant increase in the expression of Bcl-2 in TDP-43^M337V (^∗p < 0.05). Data are presented as mean ± SEM with one-way ANOVA and Dunnett's post hoc tests.

See also Figures S6 and S7.

ER Ca²⁺ release is under the regulation of the antiapoptotic Bcl-2 family of proteins and we have previously shown that its levels were significantly increased in a cell model (HEK293) that expressed the entire genomic locus of TARDBP with M337V (Mutihac et al., 2015). Consistent with our previous findings, human iPS-derived MNs from the TDP-43^M337V patient showed significantly increased levels of Bcl-2 expression (Figure 7C), confirming that ER Ca²⁺ homeostasis is disrupted in TDP-43^M337V MNs and that low ER Ca²⁺ may contribute to low Ca²⁺ release on depolarization in TDP-43^M337V MNs.

Discussion

In a previous phenotypic characterization of C9ORF72 iPS-derived MNs, we described increased expression of markers of stress granule formation, activation of cellular death pathways (cleaved caspase-3, cytochrome c), and mitochondrial dysregulation (Dafinca et al., 2016). In this study, we focused on the physiological implications of the C9ORF72 mutation in iPS-derived MNs from patients compared with isogenic control lines generated by CRISPR/Cas9 where the pathogenic hexanucleotide expansion was removed, as well as to iPS-derived MNs from patients carrying the M337V or I383T mutation in TARDBP.

High glutamatergic synaptic activity in ALS is believed to lead to calcium overload and eventually MN death (Cleveland and Rothstein, 2001, Pasinelli and Brown, 2006). Recent evidence shows increased Ca²⁺ permeability of AMPA receptors in C9ORF72 iPS-derived MNs (Selvaraj et al., 2018), and influx of Ca²⁺ through AMPA-Rs and NMDA-Rs may be a signal for activating the intrinsic stress response and contributing to triggering non-AUG-dependent translation in C9ORF72 cellular models (Westergard et al., 2019). In line with recent research, we detected increased excitability to glutamate stimulation and increased susceptibility to glutamate-induced cell death. Despite low glutamate-induced Ca²⁺ amplitudes, C9ORF72 MNs show a higher sensitivity to extracellular glutamate by responding with Ca²⁺ waves for a prolonged period of time due to increased expression of Ca²⁺-permeable AMPA-R subunits (GRIA1 and GRIA3) and NMDA-R subunits (GRIN2A). Low glutamate-induced Ca²⁺ amplitudes may be explained by increased expression of GRIA2, which is a Ca²⁺-impermeable AMPA subunit that has been shown to affect AMPA-induced Ca²⁺ peaks similarly in FUS iPS-derived MNs (Bursch et al., 2019), and which may be a compensatory mechanism activated by the MNs. However, there is a clear shift in the balance between Ca²⁺-permeable subunits and Ca²⁺-impermeable subunits, which also increases the recovery times in the C9ORF72 MNs and allows Ca²⁺ entry at non-physiological levels, which is cleared at a lower rate due to low levels of the cytosolic calcium buffering protein calbindin.

TARDBP MNs carrying the M337V mutation showed upregulation of Ca²⁺-permeable AMPA-R subunits (GRIA4) and NMDA-R subunits (GRIN1 and GRIN2B) and a similar high frequency of Ca²⁺ waves in response to glutamate, demonstrating a similar imbalance toward increased Ca²⁺ permeability. Blocking NMDA receptors resulted in higher Ca²⁺ levels in TARDBP compared with C9ORF72, indicating that Ca²⁺ permeability of the AMPA receptors is higher in TARDBP than C9ORF72.

Although depolarization induced hyperexcitability in C9ORF72 MNs and hypoexcitability in TARDBP MNs, the recovery times for both C9ORF72 and TARDBP were significantly higher than healthy controls. We showed that these Ca²⁺ responses could be linked to the levels of ER Ca²⁺ stores, as we measured low ER Ca²⁺ levels in TARDBP MNs. Furthermore, we found that both ALS mutations were associated with reduced release of Ca²⁺ from the ER stores through IP₃R. Release through IP₃R is essential for ER-mitochondria transfer of Ca²⁺ ions that contribute to the regulation of tricarboxylic acid cycle and ATP production. Previous research has shown that there is decreased contact between the ER and mitochondria in cells carrying FUS and TARDBP mutations (Stoica et al., 2014), and, together with low ER Ca²⁺ release, could account for mitochondrial dysfunction in both mutations.

We have shown previously that the mitochondrial membrane potential is significantly reduced in the C9ORF72 MNs (Dafinca et al., 2016), indicating a reduced capacity to take up cytosolic calcium. In line with our previous findings, we detect low mitochondrial Ca²⁺ uptake in C9ORF72 MNs compared with both healthy and corrected MNs. Ca²⁺ entry into the mitochondria is mediated by the Ca²⁺ uniporter channel complex that contains the mitochondrial Ca²⁺ uniporter, a Ca²⁺ selective pore, and its associated regulatory proteins, MICU1 and MICU2 (Payne et al., 2017). A contributing factor to the low uptake in C9ORF72 MNs is the reduced expression of the MCU along with reduced expression of MICU2. MICU2, together with MICU1, is located on the outer mitochondrial membrane, where it modulates and regulates Ca²⁺ affinity of the MCU channel in a complex functional interaction (Pendin et al., 2014). Although there is contradictory data in the literature with regard to their exact roles, it is generally accepted that MICU2 plays an important role in inhibiting Ca²⁺ uptake during resting state in the presence of low cytosolic Ca²⁺ levels, leading to the stabilization of mitochondrial Ca²⁺ levels (Patron et al., 2014). Consistent with our results in iPS-derived MNs, it has previously been shown that downregulation of MICU2 and MCU results in a net inhibition of Ca²⁺ uptake (Pendin et al., 2014, Plovanich et al., 2013). Therefore, the delayed buffering capacity of C9ORF72 MNs could be explained by a reduced expression of MICU2 and MCU.

In contrast, TARDBP MNs show upregulation of MICU1, which provides an essential protective mechanism by limiting MCU-mediated Ca²⁺ uptake at low levels of cytosolic Ca²⁺ to limit mitochondria becoming constitutively loaded with Ca²⁺ under resting conditions (Mallilankaraman et al., 2012). Optimal stoichiometry between MICU1 and MICU2 was shown to be necessary for efficient gatekeeping of the MCU channel and mediating Ca²⁺ uptake (Payne et al., 2017). Both mutations affect the stoichiometry of MICU1 and MICU2 with the same overall result of high MICU1 compared with MICU2. Therefore, it is evident that dysregulation of the balance in these gatekeeping proteins on the mitochondrial Ca²⁺ uniporter is essential to controlling the physiological uptake of Ca²⁺ in the mitochondria and that this phenomenon is a convergent pathomechanism in ALS MNs.

In our C9ORF72 MNs, we detected several physiological deficits in Ca²⁺ homeostasis which were reversed by the excision of the repeat expansion in CRISPR/Cas9 MNs. The cytosolic levels of free Ca²⁺ ions are significantly higher upon stimulation and we show that this is at least partly explained by reduced mitochondrial buffering of Ca²⁺ due to low expression of MCU and reduced levels of MICU2, reduced availability of the calcium binding protein calbindin, and increased expression of Ca²⁺-permeable receptors. We show that some of these phenotypes are common to MNs derived from TARDBP patients, such as high expression of Ca²⁺-permeable glutamate receptors and reduced mitochondrial Ca²⁺ buffering, but differences in excitability are also evident between the two genotypes which can partly be accounted for by increased or reduced ER Ca²⁺ levels, respectively. The differences and similarities in Ca²⁺ homeostasis we identify here between C9ORF72 and TARDBP mutations are an important and relevant aspect of disease. We observe these phenotypes in the absence of TDP-43 cytoplasmic mislocalization, suggesting that mislocalization may be a consequence of these dysregulations. Overall, our findings suggest that targeting dysregulation of Ca²⁺ homeostasis may be one strategy to preserve MN viability in ALS.

Experimental Procedures

Generation of hiPSC Lines

All iPSC lines were derived from skin biopsy fibroblasts, collected under ethical approval granted by the South Wales Research Ethics Committee (WA/12/0186) in the James Martin Stem Cell Facility, University of Oxford, under standardized protocols (see Table S2). Fibroblasts and derived iPSC lines tested negative for mycoplasma (MycoAlert, Lonza, UK, www.lonza.com/). OXTDP-01 clones were derived using the Sendai virus-based reprogramming system CytoTune™ (Klf4, Oct4, Sox2, c-Myc genes in individual viruses); OXTDP-03 clones were derived using Cytotune 2.0 (polycistronic vector Klf4–Oct3/4–Sox2, cMyc, and Klf4 separate viruses) (Life Technologies, Rockville, MD, http://www.lifetech.com, used according to the manufacturer's instructions). Transduced fibroblasts were plated for generation of iPSC clones, and clones were picked, expanded, and banked as described previously (Dafinca et al., 2016).

Quality Control Assessment of hiPSC Lines

Genome integrity and identity of derived iPSC clones versus the original fibroblast pool was assessed by Illumina Human CytoSNP-12v2.1 beadchip array (300,000 markers) or Illumina HumanOmniExpress-24 (700,000 markers) and analyzed using KaryoStudio and GenomeStudio software (Illumina).

iPSC lines were assessed for expression of pluripotency markers by flow cytometry as described previously (Dafinca et al., 2016), using a FACS Calibur (Becton Dickinson, Franklin Lakes, NJ) and analyzed using FlowJo software. The following antibodies (clone, isotype control, supplier) were used: TRA-1-60 (330614, IgM-488, BioLegend), NANOG (D73G4, IgG-647, Cell Signaling Technology).

Clearance of Sendai virus from reprogrammed cell lines was confirmed by polymerase chain reaction (PCR). RNA was extracted and reverse transcribed as described previously (Dafinca et al., 2016) with fibroblasts as negative control, and fibroblasts infected with Sendai reprogramming viruses 5 days previously as positive control. PCR was carried out according to the Cytotune manufacturers' instructions, using actin β as a positive control for each line (actin β primers, 92 bp, Eurogentec), and products were visualized by running on a 1.5% agarose gel with ethidium bromide, and 1 kB HyperLadder (Bioline https://www.bioline.com) primers: Cytotune 2.0 primers: SeV F: GGA TCA CTA GGT GAT ATC GAG C; R: ACC AGA CAA GAG TTT AAG AGA TAT GTA TC; KOS F: ATG CAC CGC TAC GAC GTG AGC GC; R: ACC TTG ACA ATC CTG ATG TGG; Klf4 F: TTC CTG CAT GCC AGA GGA GCC C; R: AAT GTA TCG AAG GTG CTC AA; c-Myc F: TAA CTG ACT AGC AGG CTT GTC G; R: TCC ACA TAC AGT CCT GGA TGA TG; Actin F AGGCATCCTCACCCTGAAG; R TCCATGCCCAGGAAGGAAG; 2-log ladder (New England Biolabs).

CRISPR/Cas9 Nuclease-Mediated Gene Editing of the G4C2 Repeat Expansion

Correction of (G4C2)₁₀₀₀ repeat expansion in C9ORF72 is described elsewhere (manuscript under review). In brief, to correct the mutation, double-nicking CRISPR/Cas9 and homology-directed repair were used on the C902-02 iPS line to remove 1,000 G₄C₂ repeats and replace them with 2 G₄C₂ repeats using 2 guide RNAs and a double-stranded DNA plasmid donor template. A Puro/TK selection cassette was also inserted in the donor template plasmid to facilitate the isolation of the corrected clones. The gRNAs, donor plasmid, and Cas9 were transfected into the C902-02 iPSCs using Neon nucleofection (Thermo Fisher Scientific). Several days after transfection, the puromycin selection was started and clones were picked and screened. After successful integration of the donor template, Cre/loxP-mediated excision was used to remove the Puro/TK selection cassette from the edited lines. The absence of the repeat and presence of the remaining loxP site in the targeted alleles was confirmed by sequencing and Southern blot analysis.

MN Differentiation and Analysis

MN differentiation was performed using a previously published protocol (Maury et al., 2015), with some modifications. In brief, the iPSCs were grown on Geltrex in mTeSR 1 supplemented with mTeSR supplement (STEMCELL Technologies) and antibiotics (Life Technologies). Induction was started when iPSCs were 90% confluent in DMEM/F12/Neurobasal medium 1:1, 1× N2 supplement, 1× B27 supplement, ascorbic acid (0.5 μM, Sigma), β-mercaptoethanol (50 μM), compound C (1 μM), and Chirr99201 (3 μM, Tocris Bioscience). On day 2, the induction medium was supplemented with all-trans retinoic acid (1 μM, Sigma) and Smoothened agonist (500 nM). On day 6, we removed from the medium Chirr99201 and compound C. On day 9, the neural precursors were split 1:3 using Accutase (Life Technologies), and Rock inhibitor (Tocris Bioscience) was added for 24 h. The neural precursors were maintained until day 19 in this medium and changed to the basal medium supplemented with brain-derived neurotrophic factor (10 μM, Life Technologies), glial cell line-derived neurotrophic factor (10 μM, Life Technologies), and laminin (500 ng/μL). DAPT (10 μM, Tocris Bioscience) and Rock inhibitor were added for 7 days and withdrawn for the rest of the maturation. Analysis was performed on day 30 of the differentiation protocol in all experiments.

Calcium Imaging

Calcium transients in iPS-derived MNs were measured using the fluorescent probe Fura 2-AM, Fluo 4-AM, or Rhod 2-AM according to the manufacturers protocol (Life Technologies). Neurons were incubated with 4 μM Fura 2-AM dissolved in 0.2% dimethyl sulfoxide (Sigma-Aldrich) with 0.04% pluronic acid in Hank's balanced salts solution (HBSS) supplemented with HEPES, CaCl₂, and MgCl₂ for 1 h at 37°C. Single-cell Ca²⁺ images were taken by fluorescence microscopy at room temperature. The dye was excited at 340 and 380 nm, and the emission fluorescence was collected at 510 nm. Images were taken every 30 s in a time-lapse fashion and a baseline was recorded for 30 s before stimulation. Between stimulations, cells were washed for 5 min with HBSS supplemented with HEPES. Neurons were incubated with 20 μM Fluo 4-AM in 0.2% dimethyl sulfoxide and 0.04% pluronic acid for 30 min at room temperature in HBSS supplemented with HEPES, CaCl₂, and MgCl₂. The dye was excited at 488 nm and images were taken every 30 s in a time-lapse fashion with a baseline recorded for 30 s before stimulation.

The stimuli used for calcium release were 50 mM KCl (Sigma), 1 mM L-glutamic acid (Sigma), 4 μM CCCP (Tocris Bioscience), 20 mM caffeine (Tocris Bioscience), 100 μM CNQX (Tocris Bioscience), and 10 μM memantine (Tocris Bioscience). Analysis of fluorescence intensity was performed using ImageJ and FIJI. Fluorescence measurements are expressed as a ratio (ΔF/F_o) of the mean change in fluorescence (ΔF) at a pixel relative to the resting fluorescence at that pixel before stimulation (F_o). For each recording, the soma of neuronal looking cells was selected as a region of interest and up to 25 neurons per recording were selected. The fluorescence intensity was plotted for each neuron against frame number and the y values representing F₀ (start of response) and F₁ (peak response) were manually copied. The recovery time was analyzed from the x axis values representing the frame number at peak fluorescence and the frame number when fluorescence intensity returned to F₀.

Mitochondrial and ER Ca²⁺ Imaging

Neurons were loaded with 4 μM Rhod 2-AM for 45 min at 37°C. The neurons were imaged as described above, using excitation at 567 nm. Image analysis was performed using FIJI and GraphPad Prism, as described above.

For ER Ca²⁺ imaging, MNs were loaded with 20 μM Fluo 4-AM in 0.2% DMSO and 0.04% pluronic acid for 30 min at room temperature in HBSS supplemented with HEPES, CaCl₂, and MgCl₂. For ER Ca²⁺ release, neurons were recorded for 30 s under continuous perfusion of HBSS followed by 30 s of 50 μM carbachol in HBSS depleted of CaCl₂ and MgCl₂. The neurons were then washed in HBSS with CaCl₂ and MgCl₂. For measurement of ER Ca²⁺ levels, the neurons were loaded on to the microscope and washed in HBSS depleted of CaCl₂ and MgCl₂, and 10 μM thapsigargin were added directly to the coverslip.

Image analysis was performed using FIJI and GraphPad Prism, as described above.

Immunofluorescence

Mature MNs were fixed with 4% paraformaldehyde-PBS for 15 min and incubated with 10% donkey serum for blocking for 1 h at room temperature. The cells were incubated overnight at 4°C with rabbit anti-TOMM20 (Abcam 78547, 1:500), mouse or rabbit anti-β-III Tubulin (Tuj1) (Covance MMS-435P, PRB-435P; 1:1,000), mouse anti-SMI-32 (Covance 801701, 1:1,000). After washing with 0.1% Triton X-/PBS three times for 10 min, the samples were incubated with Alexa Fluor 488-, Alexa Fluor 568-, and Alexa Fluor 647-conjugated donkey anti-rabbit, anti-mouse, or anti-goat secondary antibodies (Life Technologies A32766, A10042, and A32849) for 1 h at room temperature. Nuclei were stained with DAPI. Fluorescence was visualized using a confocal microscope Zeiss LSM.

Protein Extraction and Western Blotting

Cells were lysed in RIPA buffer (Sigma), sonicated, and incubated on ice. After centrifugation at 3,000 × g for 10 min, the supernatant was retained and protein concentration was quantified using the bicinchoninic acid assay (Sigma). Protein was loaded and resolved on SDS-PAGE (4%–12% Tris-glycine gel) using a Mini-PROTEAN electrophoresis system (Life Technologies) under constant current and transferred to a methanol-activated polyvinylidene difluoride membrane (Immobilon-P) using the Novex Blot system under constant voltage (100 mA for 70 min). Blots were transferred to blocking buffer (Tris-buffered saline [TBS], 0.1% Tween 20, 5% skimmed milk) for 1 h at room temperature and incubated in TBS + 0.1% Tween 20 + 1% milk with primary antibodies. Primary antibodies used were mouse anti-MCU (Abcam ab219827, 1:500), mouse anti-calbindin (Abcam ab108404; 1:1,000), rabbit anti-calreticulin (Abcam ab2907; 1:500), mouse anti-calnexin (Abcam ab112995, 1:500), rabbit anti-calpain-1 (Abcam ab39170, 1:500), rabbit anti-inositol-1,4,5-triphosphate receptor (Abcam ab5804, 1:300), and mouse anti-β-actin (Sigma A5441, 1:5,000). Horseradish peroxidase-conjugated anti-mouse IgG or anti-rabbit IgG (Life Technologies, G1040 and G1234) were used as secondary antibodies and the signal was visualized using an ECL or ECL Plus detection system (Millipore) on a Bio-Rad ChemiDoc. The integrated optical density of each band was measured in ImageJ and expression was normalized to β-actin levels in the same blot for comparative expression assessment.

RNA Sequencing

RNA sequencing and analysis was performed as described elsewhere (manuscript under review). In brief, RNA was extracted from samples from whole MN cultures at day 28. Library preparation was performed using the paired-end Illumina TruSeq sample preparation kit and samples were run on the Illumina HiSeq 4000 sequencer for 75 cycles. Samples underwent extensive quality control, and two samples with significant biases were removed from the analysis. A total of four samples derived from cell line C9-02-02, three samples from cell line C9-Ed-02-01, and three samples from cell line C9-Ed-02-02 were used in this analysis. Three samples derived from two healthy control lines (180 and 840) were also used. A counts table and normalized abundances in the form of transcripts per million were obtained using Salmon. Differential expression analysis was performed using DESeq2 taking into account operator bias in the linear model. Benjamini-Hochberg corrected p values were reported for the differential expression analysis.

The data discussed in this publication have been deposited in NCBI's Gene Expression Omnibus and are accessible under GEO series accession number NCBI: GSE147544 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE147544) and NCBI: GSE139144 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE139144).

qPCR

RNA sequencing results were validated on an extended sample cohort for the C9ORF72 MNs (two genome-edited samples, three control samples, six C9-02-02 and C9-02-03 samples, one C9-04-01, and one C9-01-06, coming from three independent differentiations). For validation of RNA sequencing of TDP-43^M337V MNs, we used five healthy controls and three M337V clones (TDP-03-03, TDP-03-04, and TDP-03-06) from three independent differentiations. RNA extraction was performed using the RNeasy Mini Kit (QIAGEN), following the manufacturer's instructions, and cDNA was prepared using the High-Capacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific). qPCR was performed using the Fast SYBR Green Master Mix (Thermo Fisher Scientific) using 0.2 μL of cDNA and 250 μM primers (Table S1). Samples (10 μL) were run in triplicate on the LightCycler 480 (Roche). Three reference genes were used for all reactions (TOP1, GAPDH, and ATP5B) and the geometric mean of all three genes was used as the reference for gene expression estimation. Relative abundance estimation was carried out using the 2^-ΔΔCt method.

Statistical Analyses

At least two independent differentiations were performed for every experiment shown and data pooled from at least three independent cultures. For TDP-43 MNs, results from three clones of one TDP-M337V patient are grouped together (TDP-03-03, TDP-03-04, and TDP-03-06) and two clones (01-01 and 01-02) are grouped for TDP-I383T. For groups of three or more, ANOVAs were performed with relevant corrections for multiple comparisons. All statistical analyses were done using GraphPad Prism 8.

Author Contributions

R.D. conceived the idea, designed, performed, and interpreted most of the experiments. P.B. performed some of the data acquisition and analysis (immunoblotting). L.F. performed qPCR validations, analysis, and immunoblotting. A.C. performed some of the data acquisition, interpretation, and iPS differentiations to motor neurons. J.S. performed and analyzed RNA sequencing results. N.A.A. generated the genome-edited iPS line and reprogrammed fibroblasts to iPS. C.S. aided with qPCR validations. J.V. and S.A.C. reprogrammed fibroblasts to iPSCs. R.D. and K.T. interpreted the results and wrote the manuscript.

Published: April 23, 2020