Human Motor Neurons With SOD1-G93A Mutation Generated From CRISPR/Cas9 Gene-Edited iPSCs Develop Pathological Features of Amyotrophic Lateral Sclerosis

Amyotrophic lateral sclerosis (ALS) is a fatal neurodegenerative disorder characterized by gradual degeneration and elimination of motor neurons (MNs) in the motor cortex, brainstem, and spinal cord. Some familial forms of ALS are caused by genetic mutations in superoxide dismutase 1 (SOD1) but the mechanisms driving MN disease are unclear. Identifying the naturally occurring pathology and understanding how this mutant SOD1 can affect MNs in translationally meaningful ways in a valid and reliable human cell model remains to be established. Here, using CRISPR/Cas9 genome editing system and human induced pluripotent stem cells (iPSCs), we generated highly pure, iPSC-derived MNs with a SOD1-G93A missense mutation. With the wild-type cell line serving as an isogenic control and MNs from a patient-derived iPSC line with an SOD1-A4V mutation as a comparator, we identified pathological phenotypes relevant to ALS. The mutant MNs accumulated misfolded and aggregated forms of SOD1 in cell bodies and processes, including axons. They also developed distinctive axonal pathologies. Mutants had axonal swellings with shorter axon length and less numbers of branch points. Moreover, structural and molecular abnormalities in presynaptic and postsynaptic size and density were found in the mutants. Finally, functional studies with microelectrode array demonstrated that the individual mutant MNs exhibited decreased number of spikes and diminished network bursting, but increased burst duration. Taken together, we identified spontaneous disease phenotypes relevant to ALS in mutant SOD1 MNs from genome-edited and patient-derived iPSCs. Our findings demonstrate that SOD1 mutations in human MNs cause cell-autonomous proteinopathy, axonopathy, synaptic pathology, and aberrant neurotransmission.


INTRODUCTION
Amyotrophic lateral sclerosis (ALS) is a progressive neurodegenerative disorder characterized by the gradual degeneration of motor neurons (MNs) leading to muscle weakness, atrophy, paralysis, and ultimately, respiratory failure and death (Le Verche and Przedborski, 2010). While mostly sporadic with unknown inheritance, about 10% of all ALS cases are familial. This molecular genetic pathology can provide important clues about the intrinsic and non-autonomous vulnerability of MNs (Boylan, 2015). Some familial ALS cases are linked to mutations in the superoxide dismutase 1 (SOD1) gene, encoding an antioxidant enzyme that functions as a homodimer, binding copper and zinc ions, to destroy superoxide radical (O 2 − ) in the body (Petrov et al., 2016). Despite being the first gene mutation identified in ALS (Rosen et al., 1993) and the extensive and important research on SOD1 and its putative disease mechanisms in ALS over the past decades, there are still no effective disease-modifying treatments for any form of ALS (Kawamata and Manfredi, 2010). Though the pathobiology of ALS is extraordinarily complex, historically research on disease mechanisms and experimental therapeutics has relied heavily on animal models that might not sufficiently replicate the human disease mechanisms and cellular neuropathology (Martin et al., 2007). Animal models, including SOD1-G93A transgenic mice, may not accurately model the genetics of ALS due to mutant gene copy numbers. Specifically, overexpression of mRNA or protein in some animal models and transfected cells make the models non-physiological, possibly producing phenotypes that are contrary to those observed from ALS patients with a single copy of a mutant SOD1 gene (Ludolph et al., 2010;Philips and Rothstein, 2015;Morrice et al., 2018). Moreover, overexpression of wild-type human SOD1 causes axonopathy and mitochondrial vacuolation in mice (Jaarsma et al., 2000;Fischer et al., 2004).
To overcome this limitation, human induced pluripotent stem cells (iPSCs), promising sources of vulnerable and diseaserelevant cell types, have enabled an opportunity to understand human disease modeling and mechanisms, and to explore human disease-relevant therapeutic development (Xu et al., 2013;Ohnuki and Takahashi, 2015). In addition, the evolution of genome editing systems allows creation of genetic modifications within cells with improved targeting efficiency and precision. Combination of these two approaches empowers investigation of human pathophysiology under the critically necessary human genomic background with all of the known and unknown genetic modifiers.
Here, we hypothesized that introduction of a diseasecausing SOD1-G93A mutation to human iPSCs will phenocopy spontaneously occurring human ALS neuropathology after their differentiation into MNs. To test the hypothesis, we utilized CRISPR/Cas9 genome editing technology on iPSCs and introduced a SOD1-G93A missense mutation by performing knock-in into the genome ( Figure 1A). By combining genome editing with stem cell differentiation approaches, we have generated highly pure MN lines that are "isogenic" to each other. With a patient-derived iPSC line harboring an A4V mutation as a positive control, we identified several disease phenotypes in these MNs relevant to ALS, including proteinopathy, axonopathy, synaptic pathology, and aberrant neurotransmission. Our findings demonstrate that genome edited iPSCs using CRISPR/Cas9-mediated targeted gene editing and their differentiation into MNs are important tools with their proper control cells to replicate and model ALS molecular, genetic, and neuronal pathology and to study mechanisms of disease in human ALS in cell culture.

MATERIALS AND METHODS iPSC Culture
The institutional biosafety committee (JHU registration B1011021110) approved the use of human cells. The protocols met all ethical and safety standards for work on human cells. All iPSC lines used in this study are listed in Supplementary Table S1 and were characterized in previous studies (Wen et al., 2014;Li et al., 2015;Kim et al., 2020). The human iPSC lines were maintained on culture plates coated with Matrigel (Corning) in StemFlex medium (Gibco) and passaged every 4-6 days by EDTA dissociation buffer (Beers et al., 2012).

Mouse and Human Primary Cell Culture
Mouse cortical astrocytes were used to enhance attachment of iPSC-derived MNs for subsequent experiments including patch clamp recording. They were isolated from P3 to P4 CD1 mouse pups as described (Schildge et al., 2013) and cultured in Dulbecco's modified Eagle's medium (DMEM, Corning) supplemented with 10% FBS. Mouse embryonic fibroblasts (MEFs) were obtained from CF-1 mouse embryos at approximately 13.5 days gestation and cultured in DMEM with 10% fetal bovine serum (FBS, Hyclone), 1% GlutaMAX (Gibco), and 1% Minimum Essential Medium Non-Essential Amino Acids (MEM-NEAA, Gibco). MEFs were irradiated prior to culturing with iPSCs. Human primary astrocytes were purchased from Thermo Fisher Scientific. They were plated and cultured a week before MEA experiments in DMEM with 1% N2 supplement, and 10% FBS.

Alkaline Phosphatase Staining
An alkaline phosphatase stain (Sigma) was used to identify undifferentiated, pluripotent iPSCs ( Figure 1B). Briefly, iPSCs were washed twice with PBS and fixed in 4% paraformaldehyde solution for 10 min. The cells were washed again with PBS and incubated in SIGMAFAST BCIP/NBT (5-bromo-4-chloro-3-indolyl phosphate/nitro blue tetrazolium) solution (Sigma, 1 tablet in 10 mL) for 10-20 min at room temperature until the color disclosure. Solution was then removed and the cells were washed with PBS. Following the wash, cells were imaged by microscopy.

CRISPR/Cas9 Genome Editing and Validation
Introduction of SOD1-G93A mutation using CRISPR/Cas9 ( Figure 1A) was done using the indicated guide RNA and FIGURE 1 | Generation of the SOD1 +/G93A iPSC line and differentiation into MNs. (A) Schematic representation of CRISPR/Cas9-mediated genome editing. The PAM sequence and the mutation site are shown in red and blue, respectively. (B) Phase-contrast image of human iPSCs and alkaline phosphatase staining showing stem cell pluripotency. Scale bar, 200 µm. (C) Sequencing chromatogram demonstrating CRISPR/Cas9-mediated genome editing of SOD1 +/+ to SOD1 +/G93A . Homozygous nucleotide G in G93G and heterozygous nucleotides (G/C) in G93A are shown. (D) Schematic representation of homology-directed repair edit detection assay by ddPCR. A primer pair (red) amplifies the intended edit region. Mutant (FAM, blue) and wild-type (HEX, green) probes are designed to bind edited or unedited sequences, respectively. (E) Two-dimensional ddPCR scatter plot showing the four droplet clusters identified with a mutant and wild-type allele; FAM/HEX negative (gray), FAM positive (blue), HEX positive (green), and FAM/HEX positive (orange). (F) Concentrations of mutant and wild-type templates. (G) Quantification of SOD1-G93A mutation in the presence of wild-type DNA. CRISPR-induced G93A missense mutation was detected at frequency 50.2%. The error bars for (F,G) represents the Poisson 95% confidence intervals. (H) Schematic diagram showing the protocol used for MN directed differentiation from iPSCs. (I) Representative images of ISL1 + , Hb9 + , and ChAT + MNs on day 28. Hoechst (HO) was used to counterstain the cell nuclei. Scale bar, 20 µm. (J) Quantification of ISL1 + , Hb9 + , and ChAT + MNs on day 18-28. (K) Quantification of cells stained for both Hb9 and ChAT. (L) Representative images of SYP + presynaptic and PSD95 + postsynaptic elements, respectively, on TUJ1 + MNs on day 60. Scale bar, 5 µm. (M) Quantification of SYP + presynaptic and PSD95 + postsynaptic structures in MN cultures on day 30-60. All data for (J,K,M) are shown as means ± SEM. (N) Whole-cell patch clamp recording on a MN on day 28. Scale bar, 20 µm. (O) Representative voltage response showing repetitive action potentials to depolarizing current step. (P) Representative hyperpolarizing voltage response to negative current injection. **p < 0.01, ***p < 0.001, and **** p < 0.0001. donor DNA as described by us previously (Kim et al., 2020). This cell line, along with its isogenic wild-type control and patient-derived iPSCs with SOD1-A4V mutation, were used in this study. Detection of wild-type and mutant alleles and copy number determination of genomic SOD1-G93A mutations in edited iPSCs were done using digital droplet PCR (ddPCR) ( Figure 1D). Reaction mix was first prepared by adding isolated genomic DNA, 1X ddPCR supermix, 1X target (FAM-labeled) and wild-type (HEX-labeled) primers/probe (Supplementary  Table S2), MseI restriction enzyme, and H 2 O ( Figure 1D). This was then loaded into DG8 cartridge along with droplet generation oil. The cartridge was placed in the QX200 droplet generator (Bio-Rad) for droplet generation. After droplets were generated, they were transferred into a 96-well plate and thermal cycling was used to amplify the sample. The plate containing the amplicons in droplets was subsequently placed and run in the QX200 Droplet Reader (Bio-Rad). Data analysis was done using the Quantasoft software (Bio-Rad) with at least 10,000 droplets.
The verification of isogenicity of the mutant and the wild-type iPSC lines after the genome editing was done using short tandem repeat (STR) profiling analysis. Thirteen STP loci (CSF1PO, FGA, TH01, TPOX, VWA, D3S1358, D5S818, D7S820, D8S1179, D13S317, D16S539, D18S51, D21S11) and amelogenin (AMEL) for sex determination were interrogated. Patient-derived iPSCs with an A4V mutation served as a control. Genomic DNA was extracted from all cell lines (Qiagen) according to the manufacturer's instructions and the STR loci and the amelogenin locus were subsequently amplified using PCR primer pairs which were assessed previously (Azari et al., 2007). The 50 µL reaction mixture consisted of 100 ng genomic DNA, 1X standard Taq reaction buffer (NEB), 0.2 µM each primer, 200 µM dNTP, and 1.25 unit of Taq DNA polymerase (NEB). The PCR program started with an initial denaturation step at 95 • C for 3 min, followed by 35 cycles of denaturation at 95 • C for 30 s, annealing at 60 • C for 30 s, and extension at 72 • C for 45 s. The final extension step was at 72 • C for 2 min. This protocol allowed simultaneous genotyping of all STRs and AMEL in a single PCR amplification. To acquire better resolution for separation of the PCR products, they were loaded onto a 3.5% NuSieve GTG Agarose gels (Lonza). The size of the gels was 14 cm x 12 cm x 1 cm. DNA electrophoresis was performed for 6 h at 110 V and visualized by ethidium bromide staining and ChemiDoc Imager (Bio-Rad).
For off-target analysis, top seven candidates were selected based on COSMID web tool (Cradick et al., 2014) The list of oligonucleotide sequences and the summary of off-target analysis were summarized in our previous study (Kim et al., 2020). The genomic DNA was isolated from iPSCs and PCR amplification was performed at loci of the seven sites. PCR amplicons were checked for any possible off-targets by direct DNA sequencing.

Alternative Protocol for MN Differentiation
Differentiation induction of cholinergic neurons from human iPSCs was carried out using the Quick-Neuron TM Cholinergic-SeV Kit (Elixirgen Scientific, Inc., Baltimore, MD). The protocol was used because it is overall more rapid in generating differentiated MNs. This differentiation of human iPSCs into cholinergic neurons is driven by a Sendai virus (SeV)-based vector (ID Pharma Co., Ltd., Tokyo, Japan) which allows forced overexpression of neural inducing factors (Goparaju et al., 2017) in a temperature-dependent manner (Ban et al., 2011). Briefly, human iPSCs were infected with the SeV vector and incubated at 33 • C, 5% CO 2 for 2 days in differentiation medium for 2 days. Then, the cultures were transferred to 37 • C, 5% CO 2 to inactivate the SeV vector and further incubated for 16-18 h. Finally, immature neurons were passaged onto a coated microelectrode array plate, CytoView MEA (Axion Biosystems), after being mixed with human primary astrocytes (Thermo Fisher Scientific). This protocol has been shown to yield cholinergic MNs as identified by islet1 and Hb9 staining in combination with choline acetyltransferase (ChAT) staining (Goparaju et al., 2017).

MEA Recording
Approximately 60,000-80,000 wild-type control or ALS mutant MNs along with 15,000-20,000 human primary astrocytes in 10 µL differentiation medium were plated on each well of a poly-l-ornithine-coated CytoView MEA 48-well plate. Glia-MN co-culture was used to allow for better uniform dispersal of MNs throughout the cultures for improved global recording within the wells. After 1-3 h of allowing the cells to settle, additional 300 µL of the medium was added to each well and further incubated at 37 • C, 5% CO 2 . Neuronal activity was recorded using Maestro Pro MEA system and AxIS Navigator software (Axion Biosystems). To exclude the possibility of the MEA signals being artifacts, tetrodotoxin (TTX) was used as a negative control to block action potentials. Following baseline recording, vehicle (H 2 O) and two different concentrations of TTX (final concentrations 2.5 and 25 nM) were added sequentially to the wells and the activity was recorded for 600 s after each treatment. Additionally, recordings were made on wells with only astrocytes plated to determine if astrocytes are contributing to activities. All the collected data were digitized and analyzed with NeuralMetric Tool and AxIS Metric Plotting Tool (Axion Biosystems).

Extraction of Misfolded and Aggregated SOD1
Approximately 6 × 10 6 cultured MNs for each line were used. Cells were first washed with PBS, collected by using a cell scraper, and harvested by centrifugation for 5 min at 200 g, 4 • C. RIPA buffer with protease inhibitors was added to the pellet and the cells were incubated for 20 min on ice. During the incubation, probe sonication was performed briefly for 2-3 times to disrupt cell membranes and homogenize the samples. The lysates were centrifuged for 10 min at 14,000 g, 4 • C and the supernatant (RIPA-soluble fraction) was collected. The pellet was washed with PBS and centrifuged again with the same condition. Urea buffer (8 M Urea, 4% CHAPS, 40 mM Tris, 0.2% BioLyte 3/10 ampholyte, 2 mM tributylphosphine) with protease inhibitors was then added to the pellet. After incubation for 30 min at room temperature, the samples were centrifuged for 10 min at 14,000 g, 4 • C and supernatant (Urea-soluble fraction) was collected for SDS-PAGE and western blotting.

MN Cell Body Size, Axon Length, and Process Branch Point Measurements
MNs spheroids, derived from treatment with small molecules (Figure 1H), on day 18 were individualized into single cells with Accutase, plated on coverslips in 24 well plates, and grown for additional 2, 4, and 6 days in cell culture. Cells were fixed for 10 min with 4% paraformaldehyde at each time point and immunostained with antibodies (Supplementary Table S3) to neuronal markers including TUJ1, MAP2, and Tau. Cell body sizes of MNs of each cell line at 4 days after plating were measured using ImageJ. For axon length measurement, an axon of a cell was first identified by a neurite that is Tau-positive but MAP2-negative. Using ImageJ, the length of the neurite was quantified. Only those neurons with axons that had more than twice the length of cell body were selected and measured. To count the number of branch points of both dendrites and axons, tracings of individual cells were first performed. This was done semi-automatically by using the NeuronJ plugin of ImageJ with fluorescence images taken 4 days after plating.

Quantification of Synapse Size and Density
MNs on day 60, a time at which synapses are well developed and prominent (Figure 2M), were fixed and immunostained with the anti-synapsin and anti-PSD95 antibodies (Supplementary Table S3). Fluorescence images of MNs were taken and analyzed using ImageJ. The area of each puncta was measured for quantifying synapse size. Images used for analyses were thresholded with a constant value for each channel. For synaptic density, the number of puncta per 100 µm was measured using a plugin, Puncta Analyzer.

Measurement of Cellular ATP Level
Total cellular levels of ATP of iPSC-derived MNs were measured using ViaLight Plus kit (Lonza) according to manufacturer's instruction. Dissociated 20,000 cells were plated in each well of Matrigel-coated 96-well plates. On the day of measurement, cells were first washed with PBS and lysed for 20 min in cell lysis reagent on a shaker to extract ATP. The cell lysates were then transferred to a 96-well clear bottom white polystyrene microplate (Corning) and ATP monitoring reagent plus was added to generate luminescent signal. After 1-2 min, the plate was placed in LMax II luminometer (Molecular Devices) and the ATP levels were measured with SoftMax Pro software (Molecular Devices).

Immunofluorescence Staining
Cells were fixed in 4% paraformaldehyde for 10 min at room temperature and washed three times with PBS. Following fixation, cells were permeabilized with 0.2% Triton X-100 for 10 min, blocked in 10% donkey serum in PBS for 1 h, and incubated in primary antibodies overnight at 4 • C. Cells were then washed in PBS, incubated with secondary antibodies, and stained with Hoechst 33258 DNA dye for nuclear visualization. See Supplementary Table S3 for a full list of antibodies.

Western Blotting
Protein concentrations were first determined by Pierce BCA Protein Assay Kit (Thermo Fisher Scientific) and samples were loaded and separated by SDS-PAGE. For RIPA-and UREA-soluble fractions, 25 µg of proteins from RIPA-soluble fractions and equivalent volumes of proteins from UREA-soluble fractions were used. Proteins were transferred onto nitrocellulose membranes, washed with a blocking buffer, and probed with primary antibodies overnight at 4 • C. The primary antibodies used are listed in Supplementary Table S3. The membrane was washed three times in the blocking buffer for 5 min and probed with secondary antibody conjugated with HRP (1:10,000, Invitrogen) for 1 h. Bands were detected with Pierce ECL western blotting substrate (Thermo Fisher Scientific).

Microscopy and Image Acquisition
Cells on coverslips were mounted in ProLong Gold Antifade Mountant (Thermo Fisher Scientific, P36930) and confocal images were taken and analyzed using Leica TCS SP8 microscope and LAS X software (Leica, Germany). Fluorescence images were obtained by Keyence BZ-X700 fluorescence microscope. For phase-contrast images, Nikon Eclipse TS100 was used.

Statistical Analysis
Statistical analyses were performed by using Student's t-test and one-way analysis of variance using GraphPad Prism. All data are represented as mean ± SEM. Statistical significance was considered when p-value was less than 0.05 ( * p < 0.05, * * p < 0.01, * * * p < 0.001, and * * * * p < 0.0001).

RESULTS
Genome-Edited Human iPSCs, Patient-Derived iPSCs, and Isogenic Control iPSCs Are Efficiently Differentiated Into Highly Pure MNs As G93A is the most widely studied SOD1 mutation in animal models including transgenic mice (Gurney et al., 1994;Pan et al., 2012) and A4V is one of the most common and aggressive SOD1 mutation in North America (Chen et al., 2014;Fay et al., 2016), we selected iPSC lines harboring these mutations for our study.
We introduced the SOD1-G93A missense mutation by CRISPR/Cas9 genome editing ( Figure 1A) into a healthy control iPSC line (C3-1) (Wen et al., 2014). The pluripotency of iPSCs was verified by alkaline phosphatase staining prior to genome editing ( Figure 1B). Genome editing was carried out by electroporation of cells with Cas9 nuclease and a guide RNA that specifically targets wild-type SOD1 allele along with a singlestranded donor oligonucleotide (ssODN) (Figure 1A). After isolating and expanding single clones, a heterozygous SOD1-G93A mutation was validated in cells by Sanger sequencing ( Figure 1C) and off-target analysis was done to make sure the mutation was introduced only at the locus of our interest (Kim et al., 2020). To further confirm the copy number of wild-type and mutant alleles, we performed droplet digital PCR ( Figure 1D). Figures 1E-G, each allele was present as one single copy at a ratio near 1:1. The isogenicity of the two iPSC lines was verified by STR profiling analysis. The amelogenin locus for sex determination confirmed same sex. The 13 STR loci (CSF1PO, FGA, TH01, TPOX, VWA, D3S1358, D5S818, D7S820, D8S1179, D13S317, D16S539, D18S51, D21S11) were compared for patterns and sizes of PCR products of both cell lines with DNA gel electrophoresis. All 14 amplicons of one cell line matched with that of the other cell line, suggesting the unique identity of the two cell lines (Supplementary Figure S1). As a comparator to this mutant line, a patient-derived iPSC line with SOD1-A4V mutation (GO013) was used (Li et al., 2015). The SOD1-A4V line had PCR product patterns distinct from the other lines. Patientderived iPSCs with SOD1-G93A mutation were unavailable.

As shown in
Using these iPSC lines, we used directed differentiation with small molecule morphogens to generate human MNs (Kim et al., 2020). iPSCs were differentiated efficiently into MNs within 28-30 days (Figure 1H). The designation of MN was established by the multipolar morphology, relatively large cell body size and large open nucleus and by immunofluorescence staining using three different MN markers, including an early MN marker, ISL1, and mature MN markers, Hb9 and ChAT (Figure 1I). At 18-21 days of differentiation, more than 80% of the cell population was ISL1 and Hb9 positive (84.4 ± 3.7% and 90.1 ± 2.6% for ISL1 and Hb9 positive cells, respectively) and by day 28-31, we obtained 93.3 ± 1.6% ChAT-expressing MNs (Figure 1J). Because Hb9 also identifies subsets of spinal interneurons (Chang and Martin, 2011), we also co-labeled the cells with Hb9 and ChAT and 86.7 ± 3.5% of cells were positive for both markers (Figure 1K). The MN cultures also showed robust positivity for the presynaptic bouton marker synaptophysin (SYP) and the postsynaptic marker postsynaptic density protein 95 (PSD95) (Figure 1L). The average size of presynaptic puncta was generally larger than that of postsynaptic puncta, and both SYP + presynaptic puncta and PSD95 + postsynaptic puncta became more prominent in sizes and numbers over time (Day 30: 0.252 ± 0.007 µm 2 and 0.201 ± 0.011 µm 2 for SYP + and PSD95 + puncta sizes, respectively, Day 60: 0.492 ± 0.015 µm 2 and 0.277 ± 0.007 µm 2 for SYP + and PSD95 + puncta sizes, respectively), indicating maturation of synapses in the MN culture ( Figure 1M).
We performed whole-cell patch clamp recording on differentiated human MNs and confirmed that the cells were electrophysiologically functional neurons ( Figure 1N). Figures 1O,P, the neurons fired repetitive action potentials in response to depolarizing current step injections, and displayed hyperpolarization followed by post-rebound action potential in response to negative current injections. These data demonstrate that the human iPSC-derived differentiated neurons are functional MNs as defined by morphology, immunophenotyping, and electrophysiology.

Electrophysiological Profiling of Mutant MNs Identifies Impairments in Functional Networks
Because individual mutant MNs showed defects in axons and synapses, we tested whether electrophysiological functional network activity was correspondingly aberrant. We used microelectrode arrays (MEA) that allow repeated and simultaneous extracellular recordings of cells in a non-invasive manner (Hofmann and Bading, 2006;Massobrio et al., 2015;Odawara et al., 2016). MNs differentiated using the alternate protocol were characterized for cell type and enrichment by immunofluorescence staining of MAP2, Hb9, and ChAT, and their quantification is shown (Supplementary Figures S4A,B). Wild-type control or ALS mutant MNs along with human primary astrocytes were plated and cultured in MEA 48-well plate, covering the electrodes in each well (Figures 5A,B). Comparison of the total number of actions potentials and spike frequencies showed significantly less spontaneous firing of action potentials (21935.3 ± 6370.3, 42708 ± 2507.9, and 96907.3 ± 12988.3 for SOD1 +/A4V , SOD1 +/G93A , and SOD1 +/+ MNs, respectively, p < 0.01, one-way ANOVA) with lower mean firing rates (Figures 5C-E; 1.5 ± 0.4 Hz, 3.1 ± 0.2 Hz, and 6.7 ± 0.9 Hz for SOD1 +/A4V , SOD1 +/G93A , and SOD1 +/+ MNs, respectively, p < 0.01, one-way ANOVA) in the mutants during the recording episode. Network bursting of these MNs was also interrogated. Network bursting, an intermittent and synchronized network-wide bursts of action potentials characterized by brief periods of intense and multiple spikes (Fardet et al., 2018), is especially important as it provides a more comprehensive profile of electrophysiological activities of MN pools. The total number of network bursts was significantly smaller in the mutants (3.7 ± 0.3, 6.0 ± 1.2, and 14.3 ± 2.0 for SOD1 +/A4V , SOD1 +/G93A , and SOD1 +/+ MNs, respectively, p < 0.01, one-way ANOVA) and their bursting did not occur as frequently as the wild-type control (0.0041 ± 0.0003 Hz, 0.0067 ± 0.0013 Hz, and 0.0159 ± 0.0023 Hz for SOD1 +/A4V , SOD1 +/G93A , and SOD1 +/+ MNs, respectively, p < 0.01, oneway ANOVA), which corresponded to the less number of spikes and lower mean firing rate detected from individual MN pools (Figures 5C,F,G). These results suggest that the mutant MNs have impaired ability of firing action potential and reduced synchronized network firing. Interestingly, however, once network bursting is engaged, mutant MNs showed remarkably longer duration of bursting (Figure 5H; 16.1 ± 3.7 s, 10.7 ± 2.1 s, and 4.6 ± 0.6 s for SOD1 +/A4V , SOD1 +/G93A , and SOD1 +/+ MNs, respectively, p < 0.05, one-way ANOVA).
Although not reaching significance, the average number of spikes per network burst in the mutants was also higher than the wild-type (Figure 5I; 3646.3 ± 1165.1, 3874.2 ± 423.8, and 2825.0 ± 223.8 for SOD1 +/A4V , SOD1 +/G93A , and SOD1 +/+ MNs, respectively). To confirm that the signals detected are not due to artifacts, we treated the cells with TTX and recorded the plate. After the baseline recording, cells were treated with vehicle (H 2 O) and then with two different concentrations of TTX (final concentrations 2.5 and 25 nM) sequentially. The activities of MNs diminished drastically at 2.5 nM TTX treatment and with 25 nM TTX, no activities were detected, suggesting that the signals are solely from the neuron Na + channel-driven action potentials ( Figure 5J). In addition, in order to exclude the possibility that the astrocytes are contributing to the network activities, we recorded wells plated with only astrocytes and did not see any signals (data not shown).

DISCUSSION
We tested the hypothesis that human iPSCs expressing a heterozygous SOD1-G93A mutation generated by CRISPR/Cas9  100 µm). Each dot indicates a dendrite analyzed. All data shown as mean ± SEM. *p < 0.05, **p < 0.01, ***p < 0.001. genome editing and their differentiation into MNs could be used as a faithful cell system that develops autonomous ALS-relevant pathological phenotypes. The evidence indicates that this hypothesis was correct. The human genome-edited SOD1-G93A MNs, as well as MNs generated from a patient-derived iPSC line with SOD1-A4V, developed robust disease phenotypes including proteinopathy, structural attrition, axonopathy, synaptic pathology, and functional defects.
The human iPSCs harboring a SOD1-G93A missense mutation were generated by us using CRISPR/Cas9 for the first time (Kim et al., 2020). The cell line expresses a single allele of mutant SOD1-G93A, thus at physiological disease-relevant levels, with the corresponding human genetic background. This work is significant because we showed that the MNs derived from the cell line develop ALS-related disease features. The detailed phenotypic characterization demonstrated in this study is original and this novel cell line supports the needs for studying ALS mechanisms in a human cell system. It is especially important as the SOD1-G93A variant is one of the most common mutations in ALS and known to have relatively rapid disease progression (Kato, 2008). In addition, this human cell line can be a good comparison to the SOD1-G93A transgenic mouse (Gurney et al., 1994) which is the most widely used and characterized rodent model of ALS. However, despite the decades-spanning use of rodents in ALS research and preclinical therapeutics, the translation of this work on transgenic mice to clinical practices has been ineffective. A possible explanation for this failure is the high copy numbers of human SOD1-G93A expressed in transgenic animals. The SOD1-G93A mice are known for their severe mitochondrial damage including structural and biochemical pathology within MNs in vivo (Bendotti et al., 2001;Sasaki et al., 2004;Chang and Martin, 2009) and in primary cell culture (Chang and Martin, 2016). However, we did not observe obvious changes in mitochondrial morphology or protein levels for VDAC1 or SOD2 in human SOD1-A4V or SOD1-G93A MNs. Moreover, ATP production was enhanced significantly in the human mutant MNs compared to wild-type MNs throughout their maturation (Supplementary Figure S5). From this aspect, our CRISPR/Cas9 engineered iPSCderived human MNs are a new and alternative tool that could generate novel information on ALS pathogenesis that is distinct from data gleaned from SOD1 mice.
As with primary neural cell cultures (Lesuisse and Martin, 2002), modeling human diseases with iPSCs has several technical challenges. Heterogeneous cells in culture including unwanted mixed neuronal cell types and poorly differentiated immature cells, and partially dying cells all contribute to variability. With cells of different origin, there are genetic differences between individuals (Chen et al., 2014). In ALS-related cell culture studies, for example, having MNs and other neuronal cell types as well as glia bearing the same ALS gene mutation confounds interrogation of MN autonomous pathophysiology. We employed directed differentiation of highly pure MNs using the principles of embryogenesis and small molecule morphogens to mitigate confounders of cell heterogeneity, thusly, identifying disease phenotypes specifically in human MNs. However, for morphological and electrophysiological studies it is often necessary to grow the MNs on feeder layers of normal astrocytes for optimal cultures. In this case, we used wild-type mouse astrocytes so these cells would not be confused with human MNs morphologically, and for MEA we used wild-type human astrocytes, and they did not contribute the MEA activity, shown to be dose-dependently blocked by TTX. However, the inherent limitation here is the reductionism to a single human cell type in culture devoid of the complexity of the human CNS with a heterozygous gene mutation being expressed in a variety of cell types. The re-establishment and renewal of MN autonomous concepts for degeneration using human MNs derived from iPSC lines could provide significant new information regarding relevant cell-type therapeutics for ALS. Our observations suggesting MN autonomous mechanisms of degeneration in human ALS SOD1 mutants are consistent with previous studies on transgenic mice showing that neuronspecific expression of human mutant SOD1 is sufficient to induce MN degeneration in mice (Jaarsma et al., 2008;Wang et al., 2008), while astrocyte-specific expression of mutant SOD1 did not cause disease (Gong et al., 2000). However, mouse cell culture experiments have shown that mutant SOD1 in astrocytes can precipitate degenerative changes in wild-type MNs and worsen degenerative in MNs harboring mutant SOD1 (Nagai et al., 2007). Additional cell culture experiments are needed using our human SOD1-A4V or SOD1-G93A MNs in co-culture with astrocytes to further explore if non-neuronal cells aggravate the pathology in MNs.
A prominent phenotype discovered in our mutant MNs was misfolded and aggregated SOD1. It was seen diffusely in the cell body and processes, including axons, and as discrete cytoplasmic inclusions (Figure 2). This morphological visualization of misfolded SOD1 aggregates in processes and distinct inclusions within MN cell bodies under basal condition has not been shown before in iPSC-derived MNs. These morphological findings were supported by biochemical corroboration with detection of SOD1 in the urea-soluble fraction. SOD1 aggregation and inclusion formation are salient pathologies seen in human ALS (Forsberg et al., 2019). Our finding that misfolded and aggregated SOD1 is present spontaneously in iPSC-derived MNs is novel because previously insoluble SOD1 protein was found only when the proteasome was inhibited artificially in iPSC-derived MNs (Kiskinis et al., 2014). This supports the idea that accumulation of misfolded and aggregated SOD1 is a spontaneous pathology occurring in the presence of endogenously functioning proteasome. Thus, a possible driver of ALS could be intrinsic to proteasome dysfunction. These results hint that therapeutic targeting of the proteasome with modulating agonists could be relevant to ALS. This idea can be tested in our new human iPSC cell line.
Because we observed misfolded and aggregated SOD1 in MN processes including axons, we examined our cell cultures for axonal pathology, which is a dominant feature of ALS neuropathology (Carpenter, 1968;Delisle and Carpenter, 1984;Hirano et al., 1984;Fischer et al., 2004). We found that SOD1 +/A4V and SOD1 +/G93A MNs develop prominent axonopathies identified as axonal length truncation, smaller number of branch points, and axonal swelling in the mutants when compared to the wild-type MNs (Figure 3). These findings are original and potentially significant. Mutant MNs with misfolded and aggregated SOD1 and axonopathy identified here are meaningful because of their possible relationship to perturbations in intracellular trafficking, but pursuit of such requires live-cell imaging. Previous studies have suggested that misfolded SOD1 disrupts ER-to-Golgi trafficking driven by COPII vesicles (Burk and Pasterkamp, 2019). Also, it has been shown that mutant SOD1 inhibits kinesin-and dynein-mediated axonal transport (Huai and Zhang, 2019). As axonal transport is an essential cellular process responsible for the movement of molecular cargos including lipids, mitochondria, proteins, and cellular organelles in neurons (Burk and Pasterkamp, 2019), its defect has been implicated in ALS pathogenesis. Impaired transport increases stalling of cellular cargos along the axon, resulting in axonal swelling (Fassier et al., 2013). Axonopathy is also critical to MNs because they have long axons that innervate skeletal muscle fibers; sciatic nerve motor axons can be up to 1 meter in length. The aberrant axonal branching may disadvantage neuronal network formation and signal transmission (Suzuki et al., 2020). Shorter axonal length and diminished branching in the mutants could affect recruitment of myofibers during muscle contractions leading to weakness and fatigue. Neuromuscular junction (NMJ) number and integrity could also be compromised by these axonal perturbations (Kalil and Dent, 2014). As dismantling of NMJs perhaps plays a critical role in the onset of ALS (Lanuza et al., 2019), once reliable and quantitatively robust new human MNs cell co-culture systems with skeletal muscle cells are fashioned they could be important for exploring the dismantling of the NMJ and identifying therapeutics for rescuing distal axon disease phenotypes.
Dysregulation of synapses is thought to play a role in ALS (Wang et al., 2009;Jiang et al., 2019). By interrogating the synaptic compartment, we found, with time in culture and achievement of stable mature MNs, larger postsynaptic puncta sizes and higher synaptic puncta density in the mutants. These synaptic phenotypes, regarding synapse size and density, are novel discoveries for human iPSC-derived MNs. Synaptic phenotypes in human MN cell culture are relevant because they appear to be associated with the ALS disease progression (Sasaki and Maruyama, 1994;Shiihashi et al., 2017;Starr and Sattler, 2018). Consequently, in addition to axonopathy per se, the synapse has been receiving great attention as a promising therapeutic target for the disease (Murray et al., 2010;van Zundert et al., 2012;Casas et al., 2016;Cantor et al., 2018).
The larger postsynaptic puncta size was demonstrated by PSD95 which is a major member of the membrane-associated guanylate kinase family and known for regulating glutamate receptors at the synapse. The size and intensity of PSD95 puncta correlates with synaptic development and maturity of glutamatergic excitatory synapses (El-Husseini et al., 2000). In iPSC-derived MNs from C9ORF72 ALS/FTD patients, cell surface levels of NMDA receptor NR1 and the AMPA receptor GluR1 were increased and these receptors were accumulated at postsynaptic densities (Shi et al., 2018). Overexpression of TDP-43 A315T in mouse cortical neurons also showed elevated levels of GluR1 (Jiang et al., 2019). We also found that SYN + PSD95 + synaptic puncta density was higher in the mutants. This finding mirrors results in which glutamatergic excitatory synaptic inputs and dendritic spine densities were increased in presymptomatic TDP-43 Q331K mice, consistent with the concept of excitotoxicity mechanisms in ALS (Fogarty et al., 2016). Our results on synaptic dysregulation in combination with other systems of ALS suggest that mutant SOD1 MNs exhibit advanced maturation of synapses compared to wild-type MNs. This abnormal development of synapses might be linked to a sentinel earlier cellular and molecular mechanism of ALS.
Lastly, our MEA results could be related to defects that we observed in the axons and synapses of the mutant MNs. Network bursting of the mutant MNs has possible relevance in generation of rhythmic motor patterns (Obien et al., 2014). Central pattern generators (CPGs), neural networks that drive rhythmic movements such as walking, chewing, and breathing without sensory feedback, have been correlated with ALS and dysfunction of CPGs has been found in ALS patients (Aydogdu et al., 2011). Mutant MNs having less number of network bursting with lower network burst frequency implies that their circuit formation locally or with other cell types in the central nervous system is abnormal, possibly due to pathologies or dysregulations that we found in the axons and synapses. Also, of note is that mutants showed longer duration with more numbers of spikes in a network burst.
ALS has aging as a key risk factor, with the symptoms usually developing between the ages of 55-70 (Martin et al., 2017). In Parkinson's disease, similarly with aging as a strong risk factor, patients administered levodopa showed reduced burst duration as well as shorter amplitudes on electromyography (EMG), which correlated with the motor improvement (Tinkhauser et al., 2017). Similar results were shown by quantifying age-related differences in burst activity measured by EMG. Interestingly, muscles of old adults (adults > 70 years old) exhibited longer burst duration with higher mean amplitudes over an 8 h recording when compared to young adults (adults < 26 years old) (Theou et al., 2013). With this precedent in mind, our results on ALS MNs suggest that perturbations in MNs bursting patterns in cell culture could provide early prodromal mechanistic insight.

CONCLUSION
In conclusion, our findings demonstrate that genome edited iPSCs using CRISPR/Cas9-mediated targeted gene editing and their differentiation into MNs provide important cellular tools to study mechanisms of disease in human ALS, including proteinopathy, axonopathy, synaptic pathology, and electrophysiological defects. This work can provide needed new insight into human cell-relevant therapeutic targets.

DATA AVAILABILITY STATEMENT
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

AUTHOR CONTRIBUTIONS
BK and LM conceived and designed the experiments, analyzed the data, and wrote the manuscript. JR assisted in stem cell maintenance and motor neuron differentiation. YJ took fluorescence images and performed ATP assay. JK conducted whole-cell patch clamp recordings and analyzed the data. All authors approved the final manuscript.

FUNDING
This work was supported by the NIH NS034100 and NS052098.

ACKNOWLEDGMENTS
The authors are thankful for Wei-Kai Huang at the University of Pennsylvania for technical advice on iPSC culture and MN differentiation. We thank Junaid Afzal at the University of California San Francisco for providing guidance on genome editing. The authors also thank Bloomberg Flow Cytometry and Immunology Core at Johns Hopkins University for technical assistance with the ddPCR. We also thank Bumpei Noda and Tetsuya Tanaka for collaborating and providing technical capability in MEA.