A cell-based platform for oxidative stress monitoring in motor neurons using genetically encoded biosensors of H2O2

Background Oxidative stress plays an important role in the development of neurodegenerative diseases: it either can be the initiator or part of a pathological cascade leading to the neuron’s death. Although a lot of methods are known for oxidative stress study, most of them operate on non-native cellular substrates or interfere with the cell functioning. Genetically encoded (GE) biosensors of oxidative stress demonstrated their general functionality and overall safety in various live systems. However, there is still insufficient data regarding their use for research of disease-related phenotypes in relevant model systems, such as human cells. Methods We applied CRISPR/Cas9 genome editing to introduce mutations (c.272A>C and c.382G>C) in the associated with amyotrophic lateral sclerosis SOD1 gene of induced pluripotent stem cells (iPSC) obtained from a healthy individual. Using CRISPR/Cas9, we modified these mutant iPSC lines, as well as the parental iPSC line, and a patient-specific SOD1D91A/D91A iPSC line with ratiometric GE biosensors of cytoplasmic (Cyto-roGFP2-Orp1) and mitochondrial (Mito-roGFP2-Orp1) H2O2. The biosensors sequences along with a specific transactivator for doxycycline-controllable expression were inserted in the “safe harbor” AAVS1 (adeno-associated virus site 1) locus. We differentiated these transgenic iPSCs into motor neurons and investigated the functionality of the biosensors in such a system. We measured relative oxidation in the cultured motor neurons and its dependence on culture conditions, age, and genotype, as well as kinetics of H2O2 elimination in real-time. Results We developed a cell-based platform consisting of isogenic iPSC lines with different genotypes associated with amyotrophic lateral sclerosis. The iPSC lines were modified with GE biosensors of cytoplasmic and mitochondrial H2O2. We provide proof-of-principle data showing that this approach may be suitable for monitoring oxidative stress in cell models of various neurodegenerative diseases as the biosensors reflect the redox state of neurons. Conclusion We found that the GE biosensors inserted in the AAVS1 locus remain functional in motor neurons and reflect pathological features of mutant motor neurons, although the readout largely depends on the severity of the mutation.


INTRODUCTION
Redox reactions are part of the cellular metabolism. Normally, reactive oxygen species (ROS), emerging as a byproduct of such reactions, are quickly neutralized by antioxidant systems [1]. In oxidative stress, due to excessive production or disturbed utilization, ROS accumulate, subsequently leading to the cell malfunction [2]. An increasing number of the ROS molecules alter protein structure, change properties of membranes due to lipid peroxidation and cause DNA damage, which allows considering oxidative stress as one of the major mechanisms of degenerative disorders and aging [3,4].
It is known that oxidative stress plays an important role in various pathologies, and its involvement in the development of neurodegenerative diseases is indisputable, although not always clear [5][6][7]. Oxidative stress takes a certain place in amyotrophic lateral sclerosis (ALS)a disorder characterized by the inevitable death of motor neurons resulting in progressive paralysis [8]. The first ALS-associated gene, SOD1, has been discovered in 1993 [9]. SOD1 encodes superoxide dismutase 1, the main component of the antioxidant system; thus, oxidative stress was proposed as the primary pathological mechanism of the disease [9]. Subsequent studies revealed that ALS has a much more complex etiology involving other genes and that only 10% of the cases are hereditary [10][11][12]. Moreover, it is considered now that not loss, but gainof-function of mutant SOD1 underlies the ALS development [13]. Nonetheless, signs of oxidative damage have been found in both patients and model organisms, regardless of the initial cause of the disease, suggesting a universal role of redox imbalance in motor neuron damage [14][15][16].
Redox studies are often conducted by measuring key molecules such as glutathione, hydrogen peroxide, NADP + /NADPH, and others [17,18]. Although dozens of chemical molecular probes have been developed for such analyses, most of them have low specificity, availability and interfere with the cellular processes [19,20]. Genetically encoded (GE) biosensors are free from these flaws and can be applied for the same measurement as the molecular probes. GE biosensors are protein-based and delivered inside the cell in the form of nucleic acid. The cell produces biosensor molecules as long as the coding sequence is available. Since the biosensor molecules are produced inside the cell, they do not have the problem of availability, and therefore can be applied not only in cell cultures but also in more complex model systems, such as a whole animal or plant [21][22][23]. The nature of the genetically encoded biosensors allows easy modifications, e.g. addition of the tags that direct the biosensor to specific cellular compartments (the nucleus, mitochondria, endoplasmic reticulum, and plasma membrane) [24,25]. Traditional methods for the GE biosensors research in cell culture apply their transient expression via 4 plasmid delivery or viral-mediated integration of the biosensors sequences [4]. The first approach provides a high-intensity signal but does not allow prolonged experiments. The second one, on the contrary, provides a stable expression of the biosensor but does not guarantee reliable results since randomly integrated biosensors can disturb genome function. Furthermore, some may also speculate that a high level of the biosensor's expression may alter cell functioning due to consumption of the target analytes [19]. Although many redox biosensors have been developed in recent years, only a few have been validated in model systems as suitable for studying disease-associated phenotypes [21,[26][27][28].
Here, we developed a platform for monitoring oxidative stress in motor neurons. We used GE biosensors of cytoplasmic (Cyto-roGFP2-Orp1) and mitochondrial (Mito-roGFP2-Orp1) H 2 O 2 ; a known marker molecule that reflects increased ROS production. The biosensors allow ratiometric measurement of hydrogen peroxide, and therefore relative oxidation in the corresponding compartments. The main advantage of the applied approach is that the biosensors' sequences were inserted in the "safe harbor" AAVS1 locus under the control of the doxycyclinedependent promoter, providing prolonged expression of the biosensors and preventing potential negative effect of off-target inserts. To investigate the functionality of the biosensors in these conditions, we generated isogenic induced pluripotent stem cell (iPSC) lines with two mutations affecting different parts of the ALS-associated SOD1 gene. We found that the Cyto-roGFP2-Orp1 and Mito-roGFP2-Orp1 biosensors remain functional and reflect H 2 O 2 levels in iPSCderived motor neurons (Fig 1a). Moreover, we showed that a combination of G128R/K129X mutations, affecting exon 5 of SOD1, results in rapid accumulation of H 2 O 2 and an aberrant response to exogenous H 2 O 2 in mature motor neurons, but it was not observed for the D91A mutation.

CRISPR/Cas9 design and generation of SOD1 mutant iPSC lines
The guide RNAs (gRNAs) targeting sequences in exons 4 and 5 of the SOD1 gene and the AAVS1 locus were designed using the web-based tool http://crispr.mit.edu [31], with a selection of gRNAs with high-quality scores to avoid possible off-targets (Supplementary Table 2). We used CRISPR/Cas9 ribonucleoprotein (RNP) complexes to induce double-strand breaks in the target sites. The Alt-R® crisprRNA and tracrRNA were obtained from IDT (Integrated DNA technologies), and Cas9 protein was expressed in E. coli and purified according to the previously published protocol [32]. The RNP complexes were assembled according to the manufacturer's instructions before the cell transfection. For the introduction of the c.272A>C and c.382G>C mutations we used appropriate SOD1 RNP (20 pmol tracrRNA + 20 pmol crRNA (SOD1-4/SOD1-5) + 20 pmol Cas9) complexes mixed with 100 pmol of D91A ssODN (single-stranded oligodeoxynucleotide) donor or G128R ssODN donor (Supplementary Table 2). The cells were passed 24 hours before the transfection in the iPSC medium supplemented with Y-27632 (10 ng/ml). On the day of transfection, the cells were dissociated with TrypLE, strained through a 70 μm cell strainer (Miltenyi biotec), counted, centrifuged at 200g for 5 min, and resuspended in R buffer (Neon Transfection System, Invitrogen) according to the manufacturer's instructions. 10 ul of the suspension was taken to the electroporation by Neon Transfection System with the following impulse settings: 1100 V, 30 ms, 1 pulse. The cells were seeded onto feeder-coated 4 6 cm 2 dishes in the iPSC medium supplemented with Y-27632 (10 ng/ml). The next day, the cells were dissociated with TrypLE, strained through the cell strainer, and subcloned on three 96-well plates three cells per well for propagation and analysis. Genomic DNA of the survived clones was obtained and analyzed for the presence of the target mutations.

Screening of iPSC clones for the c.272A>C (D91A) and c.382G>C (G128R) substitutions
To detect the c.272A>C mutation, we designed primers for tetra-primer ARMS (amplificationrefractory mutation system) PCR screening using http://primer1.soton.ac.uk/primer1.html Table 2) [33] and performed touchdown 3-step PCR: annealing at 68-64 °C for 9 cycles, then at 64 °C for 21 cycles. The PCR products were analyzed in 2% agarose gel. Clones positive for mutant allele presence were further examined by Sanger sequencing (Supplementary Table 2). To detect the c.382G>C mutation, we designed a pair of primers that amplify the target locus of the SOD1 gene and two fluorescent probes targeting either wild-type or mutant sequence (Supplementary Table 2). Using LightCycler 480 (Roche), we analyzed the clones and selected those who had strong signals from the mutant-targeted probe. The target mutation was further confirmed by Sanger sequencing (Supplementary Table 2). The clones used in the experiments were characterized according to the Human Pluripotent Stem Cell Registry standards with the protocols described earlier [34].

Generation, selection, and screening of iPSC clones with target AAVS1 inserts
To insert Cyto-roGFP2-Orp1, Mito-roGFP2-Orp1 and transactivator in AAVS1, we used AAVS1 RNP (100 pmol tracrRNA + 100 pmol AAVS1 crRNA + 100 pmol Cas9) mixed with 5 μg of donor plasmids mix, containing equimolar amounts of transactivator donor (pAAVS1-Neo-M2rtTA, Addgene # 60843) + pCyto-roGFP2-Orp1-donor or pMito-roGFP2-Orp1-donor. The cells were prepared as it was described earlier and resuspended in R buffer. We mixed 100 μl of the cell suspension, RNP complexes and donor plasmids and performed transfection using Neon 7 Transfection System (2× reaction per experiment). The cells were then seeded onto feeder-coated 10 cm 2 dishes in the iPSC medium supplemented with Y-27632 (10 ng/ml) and maintained until small colonies formed, for 2-3 days prior to the selection. For the selection of subclones with the target biosensor and transactivator inserts, we supplemented the iPSC medium with puromycin dihydrochloride (Sigma-Aldrich) for 3 days. Then we replaced the antibiotic with neomycin sulfate (Sigma-Aldrich) and incubated the cells for 4-5 more days. Antibiotics concentrations were determined for each cell line by titration before the experiment. At the end of the selection, we added doxycycline hyclate (2 μg/ml, Sigma-Aldrich) and examined the remained clones for the presence of fluorescent signal from the biosensors' roGFP2 (reduction-oxidation sensitive green fluorescent protein 2) using the Nikon Eclipse Ti2-E (Nikon) microscope. The clones positive for the roGFP2 expression that survived double antibiotic selection were manually harvested into separate dishes for maintaining and analysis. We extracted genomic DNA from these iPSC clones and analyzed for the presence of the target and off-target inserts of the donor plasmids using PCR with specific primers (Supplementary Table 2).

Immunocytochemistry
Immunocytochemistry was performed on iPSCs and motor neurons (ChATdifferentiation day 20 and 28; ISL1 and MNX1differentiation day 28). The cells were fixed in 4% formaldehyde solution (Sigma-Aldrich) for 10 min at room temperature (RT), permeabilized with 0.5% Triton X-100 (Sigma-Aldrich) for 30 min at RT, and then incubated with blocking buffer (1% bovine serum albumin (BSA) in PBS, Sigma) for 30 min at RT. After, the cells were incubated with specific primary antibodies overnight at 4 °C. The appropriate secondary antibodies were added for 1.5-2 h incubation at RT. All antibodies were diluted in blocking buffer, and the cell nuclei were visualized with DAPI (1 μg/ml solution in PBS; Sigma-Aldrich). The antibodies and their dilution ratios are listed in the Supplementary Table 3. Micrographs were captured using either 8 Nikon eclipse Ti-E microscope (Nikon) and NIS Elements software or LSM-780 (Zeiss) microscope and ZEN black software.

Motor neuron differentiation
The iPSCs were seeded onto dishes coated with Matrigel-ESQ (Corning) in E8 (Gibco) medium and maintained in feeder-free conditions for at least 2 passages prior to differentiation. Motor neuron differentiation was performed according to the previously published 4-step protocol [35].
For neural patterning, the E8 medium was changed to basal neuronal differentiation medium

Preparation of live motor neuron samples for microscopy
Motor neurons have a low surface attachment, which makes prolonged microscopy experiments difficult. Therefore, in the biosensors experiments, the cells were seeded for maturation on the cell imaging coverglasses inside a layer of 33% Matrigel. To do so, we resuspended the cells in the 1.5× cold fourth step medium (supplemented with 15 ng/ml Y-27632), making a suspension with 1.5×10 5 cells/75 μl (1.5×10 5 cells/well). Then, we mixed 75 μl of the cell suspension with 35 μl of growth factor reduced Matrigel (Corning) and quickly applied 100 μl of the mix onto the surface of the chilled cell imaging coverglass, standing on a cold tube cooling rack turned upside down and covered with a paper towel. Using the tip of a pipette, we carefully spread the mix over the surface and left it on the cooling rack for 10 minutes to let the cells fall to bottom before the Matrigel polymerized. After, we transferred the coverglasses on top of the working surface and left them for another 10 minutes, and, then, the coverglasses were carried over to a CO 2 incubator for 1h for Matrigel stabilization. After 1 hour, we added 300-400 μl of the fourth step medium supplemented with Y-27632 (10 ng/ml) on top of the stabilized Matrigel layer ( Supplementary Fig. S1).

Reverse-transcription quantitative PCR (RT-qPCR)
Total RNA was extracted from iPSC and motor neurons with Trizol reagent (Invitrogen). The reverse transcription of 1 μg of total RNA was performed with 5x RT-buffer mix with M-MuLV-RH reverse transcriptase (Biolabmix) and random hexamer primer (Invitrogen) and diluted 1:

Flow cytometry analysis
To identify the proportion of motor neurons in the differentiation, we dissociated the cells with Accutase on the day 20 of the differentiation protocol, resuspended in cold PBS, and centrifuged at 400 g for 5 minutes (the same settings were used for all subsequent centrifugation steps). The pellet was resuspended in 1 ml cold 4% formaldehyde solution and incubated on ice for 10-15 minutes. After, we added 1 ml cold PBS, centrifuged the cells, discarded supernatant, resuspended the pellet in 1 ml ice-cold 100% methanol, and incubated it for 10-15 minutes on ice. Then, the pellet was washed twice with flow cytometry staining buffer (1% BSA, 0.2 μM EDTA, in PBS) and resuspended in it to 1×10 6 cells/ml concentration. 100 μl of the cell suspension was incubated with anti-ISL primary antibodies overnight at 4 °C. The cells were washed with the flow cytometry staining buffer and incubated with the secondary antibodies for 30 minutes at RT. Cells were analyzed using FACSAria (BD Biosciences). Unlabeled cells and isotype-labeled cells were used as controls.

Fluorescence intensity measurement
To measure the biosensors' fluorescence intensity, we obtained images of the MN using a Zeiss

Axon measurement
The immature motor neurons were seeded on the cell imaging coverglasses in a low density (1.5×10 4 cells/cm 2 ) and grown for 2 days in the fourth step medium supplemented with neurotrophic factor (NTF) cocktail: IGF1 (PeproTech, 10 ng/ml), CNTF (PeproTech, 10 ng/ml), BDNF (PeproTech, 10 ng/ml). Doxycycline (2 μg/ml) was added to induce the roGFP2 expression. RoGFP2 served as a label, marking the cellular contour and processes of the live neurons. We obtained mages for each cell line with the Nikon Eclipse Ti-2E microscope (20× objective, FITC channel). Using ImageJ, we manually measured the length of the longest processes (axons) of free-lying neurons with visible ends. Only axons with the length more than twice the size of the neurons' bodies were considered for the analysis. If the neuron had two long processes, the longest one was considered for the measurement. The mean length of the axons was calculated based on the data obtained from the differentiation of three separate iPSC clones for each genotype present.

Image acquisition
The general procedure for the redox biosensors measurement was described earlier by B.
Morgan and colleagues [36]. Although, we modified the protocol to be more suitable for cultured motor neurons.

Cells and solutions preparation
To measure H 2 O 2 utilization in real-time, we replaced the medium in the neurons with the neuronal deficit medium the day before the experiment, unless otherwise indicated. 1 hour before the experiment the medium has been removed from the analyzed wells (so as not to disturb the Matrigel layer with neurons inside), and replaced with warm HBSS + Ca 2+ , Mg 2+ . The cells were incubated in a CO 2 incubator for removal of the residual medium components from the Matrigel layer. Then, the old HBSS has been almost completely removed (with residual of ~ 50 μl/well) and replaced with the fresh HBSS (~ 300 μl/well).
The stock and working solutions were prepared freshly on the day of the experiment. The water stock solutions of 1 M DTT (Sigma-Aldrich) and 0.2 M diamide (Sigma-Aldrich) were made from a powder; 10 μM H 2 O 2 stock solution was made from hydrogen peroxide solution (30% w/w, Sigma-Aldrich).

Microscopy settings
We used a confocal Zeiss LSM-780 laser scanning microscope (Pan-Apochromat 20× objective) equipped with the 488 nm argon laser, the 405 nm UV diode laser, and a climate chamber connected with the temperature and CO 2 control modules. Cell imaging coverglass with the cells was placed on the stage without a lid, covered with a CO 2 cover, and left for 5-10 minutes for temperature equilibration. Tubes with the working solutions were also put inside the climate 13 chamber. During the equilibration, the cells were visualized with transmitted light to achieve a stable focus. Actual microscopy settings varied between the samples and required customization for obtaining a quality image. Table 1 describes the initial microscopy setup. An image did not contain oversaturated pixels and had a high noise-to-signal ratio. Gain 800 -Cyto-roGFP2-Orp1, 700 -Mito-roGFP2-

Biosensor calibration using DTT/diamide
To determine the states of the maximum oxidation and reduction possible for the biosensors, we treated the cells with diamide and DTT, respectively, using the following procedure: 15 The calibration procedure has been performed for every sample before the other experiments, and the values of the maximum oxidation/reduction were used for the dynamic range calculation and data normalization.

The basal H 2 O 2 level measurement and the measurement of H 2 O 2 utilization in real-time.
For the measurement of the Cyto-roGFP2-Orp1 and Mito-roGFP2-Orp1 signals in motor neurons, we applied the microscope settings determined during the calibration.

Excitotoxicity induction assay
To induce glutamate-mediated excitotoxicity, we incubated MN in the neuronal maintenance medium supplemented with 0.5 μM Retinoic acid, 0.15 μM Compound E, 20 μM monosodium glutamate (Sigma-Aldrich), and 100 μM l-trans-pyrrolidine-2,4-dicarboxylic acid (PDC, Sigma-Aldrich) for 5 days, changing the medium every other day. After 5 days of incubation, we obtained images of the treated MN and non-treated control MN. The data obtained at the end of the experiment were normalized to the starting oxidation values, measured before the glutamate addition, to describe the changes that emerged during the experiment.

Data normalization
The images were saved as 16-bit .tiff files and processed by ImageJ. For the analysis, the images were converted to the 32-bit format. Single pictures were split to the 405 nm (roGFP2 ox ), 488 nm (roGFP2 red ), and transmitted light channels. The intensity of the 405 and 488 channels was thresholded, and values below the threshold were set to "not a number" (NaN). A ratio image was created by dividing the 405 nm image to the 488 nm image (roGFP2 ox /roGFP2 red ), and the mean intensity of the resulting image was measured. The time series images were processed similarly. The files were imported in ImageJ as stacks, converted to the 32-bit format, split using the "Stacks-Shuffling-Deinterleave" plugin. The threshold was adjusted for both channels, and the values below the threshold were set to "not a number" (NaN  (1): The normalized data were used for visualization, comparison and statistical analysis.

Maximum oxidation and recovery rate calculation
The maximum biosensor's oxidation was calculated by subtraction of the initial normalized value of the roGFP2ox/roGFP2red ratio recorded at "0" time point from the normalized maximum oxidation value of the roGFP2ox/roGFP2red ratio obtained after H 2 O 2 addition (usually after 30 minutes of recording). The recovery rate was calculated by subtraction of the final normalized value of the roGFP2ox/roGFP2red ratio recorded at the end of the time series from the normalized maximum oxidation value with subsequent division of the resulting numbers to the time interval between these two time points (in hours).

Statistics
The data derived from at least three different clones of the same genotype were collected for the comparisons; except for measurements of H 2 O 2 utilization in starvation assay and Mito-roGFP2-Orp1 reaction to H 2 O 2 , where data were collected from the MN derived from one iPSC clone.
Analysis of the axon length was performed using one-way ANOVA with the Bonferroni

Introduction of the SOD1 D91A and G128R mutations in IPSCs of the clinically healthy donor
It is known that SOD1 has more than 140 mutations associated with ALS, and they define clinical features of the disease such as its manifestation age, rate of symptoms progression, presence of additional neurological symptoms, etc [37]. Several theoretical studies [38,39] hypothesized that severity of the symptoms is highly dependent on the position of a mutation in the sequence and its effect on the protein folding and stability. We designed two CRISPR/Cas9 guide RNA targeting the sequences in the exons 4 and 5 of SOD1 and two ssODN donor templates necessary for the introduction of c.272A>C and c.382G>C single nucleotide mutations that lead to the D91A and G128R substitutions, respectively, in the SOD1 polypeptide (Fig. 1b,   Supplementary Table 2). The c.272A>C mutation is known for its relatively mild character with late manifestation and long progression, while c.382G>C is characterized by extremely rapid development [40,41].
We introduced these mutations using CRISPR/Cas9, into a well-characterized control iPSC line (K7-4Lf), obtained earlier from a clinically healthy individual [30] (Supplementary Table 1) and recovered 66 clones for D91A variant and 124 clones for G128R variant. We screened those clones by tetra-primer ARMS PCR (for D91A) or qPCR with fluorescent probes (for G128R) and found 6 (9.1%) and 4 (3.2%) clones, respectively, presumably positive for the target mutations. The target mutations were further confirmed by Sanger sequencing. As a result, a number of clones with different SOD1 allelic variants were obtained (Fig. 1c). Since no homozygous clones were found, we chose clones with SOD1 D91A/del105 (SOD1-D91A) and SOD1 G128R/K129* (SOD1-G128R) variants for subsequent experiments. It was suggested that the destruction of one of the alleles will create a more severe phenotype and make biosensors' measurements more robust. These iPSC lines demonstrated features specific for pluripotent cells: they expressed specific genes and proteins (OCT4, NANOG, SOX2, and TRA1-60), positively stained for alkaline phosphatase, and were able to generate three germ layer derivatives. The cells also retained a normal 46, XX karyotype and were free of mycoplasma contamination ( Fig.   1d-e, Supplementary Fig. S2). The Tet-On expression system applied for the biosensors' expression consists of two elements [42]: the biosensor's sequence that follows the tetracycline-dependent promoter and the specific transactivator (rtTA, reverse tetracycline-controlled transactivator) essential for the controlled expression of the target genes. To deliver these elements in the cell's genome, we used biallelic target insertion in the safe harbor AAVS1 locus (the first intron of the PPP1R12C gene) via CRISPR/Cas9 (Fig. 2a). The donor plasmid containing the rtTA, homologous arms, and neomycin resistant gene for the selection was obtained from the vendor (Addgene) [43]. The biosensors' donor plasmids were designed and constructed in the lab [29]. These donor plasmids contained the tetracycline-dependent promoter, followed by the Cyto-roGFP2-Orp1 and Mito-roGFP2-Orp1 sequences, as well as ~800 base long homology arms, and puromycin resistance gene for the selection of the target clones (Supplementary Fig. S3). We delivered both donor plasmids along with the CRISPR/Cas9 RNPs in the IPSCs and selected them using appropriate media supplemented with neomycin and puromycin. Among the survived clones, we manually picked up and subcloned those who positively responded to the doxycycline (tetracycline 21 derivative) addition with the biosensors expression and screened the clones for the presence of target biallelic insertion in the AAVS1 locus by PCR (Fig. 2b-c). Further, clones positive for the target insertions were screened for additional off-target donor integration by PCR using specific pairs of primers that only detect non-integrating parts of the donor plasmids, implying that the presence of the PCR-product indicates an off-target insertion. We obtained from 3 to 17 separate iPSC clones positive for the target and negative for the off-target insertions for each cell line (Fig. 2d). All transgenic iPSC clones used further in the experiments were stained positive for the specific pluripotent cell markers SOX2 and SSEA4 (Supplementary Fig. 4).

Motor neurons derived from the IPSCs with mutant SOD1 show impaired axon growth
We utilized a previously described protocol of highly efficient motor neuron differentiation (Fig.   3a) [35] that allowed us to obtain iPSC-derived motor neurons with a characteristic morphology within 30 days. For each type of iPSC (K7-4Lf, iALS, SOD1-D91A, and SOD1-G128R), we differentiated three separate clones. All motor neurons derived from the iPSC were positively stained for ChAT, ISL1, and MNX1 and expressed mRNA of these proteins on comparable levels ( Fig. 3b-c, Supplementary Fig. S5). We analyzed MN differentiation efficiency on the day 20 of differentiation by counting the ISL + cells using flow cytometry. The efficiency ranged between 89 and 95% with the 91.9±7% ISL + cells for K7-4Lf, 91.1±2% for SOD1-D91A, 89.3±3.8% for SOD1-G128R, and 94.7±1.5% for iALS (Fig. 3d).
To characterize MN obtained, we measured axonal length of the derived MN on the day 21 of differentiation in low-density culture. The mean length of the axons in SOD1-D91A and iALS MN was 90.7±42.6 μm and 90.4±41.9 μm, respectively, which was significantly lower than in control K7-4Lf MN (107.8±45.5 μm). The mean axon length in SOD1-G128R (67±30.6 μm) was even lower than in SOD1-D91A and almost forty percent lower compared to the control K7-4Lf.
This suggests the presence of the pathological effect of the mutations introduced in SOD1, as well as a different level of severity of these mutations (Fig. 3e).

The biosensors' expression declines during differentiation
Although the AAVS1 site is located in the intron of a transcriptionally active gene and was described previously as suitable for stable expression of transgenes [44], we have discovered that the motor neurons did not always retain detectable fluorescence level of the biosensors at the terminal stages of the differentiation, and this did not depend on the particular cell line used for the differentiation (Fig. 3f-g). Analysis of the expression level of rtTA and roGFP2 in the MN 23 that lost the biosensor's signal revealed that the terminally differentiated MN expressed mRNA of the rtTA at the same level as the corresponding IPSCs from which they were obtained, while the expression of the biosensor's roGFP2 was decreased by two orders, suggesting that the biosensor's promoter was selectively inhibited (Fig. 3h). It is known that the differentiation process is accompanied by chromatin remodeling [45]. Since the rtTA expression is constitutive, while biosensors require tetracycline-derivatives to activate transcription, we suggested that the active state of the promoter during differentiation prevents its inhibition. We have been supplementing the differentiation medium with doxycycline every other day from the first day of differentiation to keep the biosensor's promoter in an active state. This resulted in a more stable expression of the biosensors' mRNA on a level comparable to the IPSC (Fig. 3h) as well as in a high level of the fluorescence intensity.

Target insertion of the genetically encoded biosensors of H 2 O 2 in the AAVS1 locus does not affect their basic properties
The main principle of the ratiometric H 2 O 2 biosensors is based on the roGFP2 ability to change its fluorescent properties. Two cysteine residues positioned close to the roGFP2 chromophore form a disulfide bond under oxidation, leading to structural changes that influence the protein fluorescence [46]. Thus, the reduced roGFP2 (roGFP2red) have excitation maximum at 490 nm, the oxidized (roGFP2ox)at 400 nm, with the emission maximum at 510 nm for both forms. It is possible to obtain a signal from each form separately by exciting the biosensor at two wavelengths; the intensity of these signals, in turn, allows estimating relative oxidation of the roGFP2, i.e. the proportion of the oxidized molecules in the cytoplasm (for Cyto-roGFP2-Orp1) or mitochondria (for Mito-roGFP2-Orp1). Fusion of the roGFP2 to the thiol peroxidase Orp1 creates a redox relay in which H 2 O 2 specifically oxidizes Orp1, and then this state is passed to the roGFP2, with subsequent Orp1 reduction (Fig. 4a) [46]. Since the reaction is reversible, the biosensors reflect not only the steady-state level of H 2 O 2 in the cell but also dynamic changes in H 2 O 2 caused by certain events.
For the ratiometric H 2 O 2 biosensors, dynamic rangethe difference between the fully oxidized and fully reduced biosensorreflects the scale of the signals that can be distinguished in the 26 experiment (Fig. 4b), and according to the literature, this number varies from 3 to 8 [23,36,47].
We measured the dynamic range of the Cyto-roGFP2-Orp1 and Mito-roGFP2-Orp1 biosensors in every MN sample used in the work (Fig. 4c). To do that, we added to the cells either oxidizing agent diamide or reducing agent DTT. Afterward we recorded changes in the biosensors' signal in real-time. The biosensors quickly reacted to the addition of the oxidizing and reducing agents, reaching the plateau 8-10 minutes after the chemicals were added (Fig. 4d-e). The calculated dynamic range was 4±0.33 for Cyto-roGFP2-Orp1 and 3.3±0.61 for Mito-roGFP2-Orp1 and stayed within acceptable values.

B-27 supplement deprivation affects mitochondrial level of H 2 O 2 regardless of the genotype
The basal level of H 2 O 2 reflects redox balance and general condition of the cell. To obtain information about the redox state of the MN, we performed live imaging on the differentiation day 29. We did not observe any differences between K7-4Lf and SOD1-D91A MN: relative oxidation level of the cytoplasm and mitochondria was the same for both types of neurons (Fig.   5a). No differences were also detected between these two cellular compartments.  . 5b). To correct the observed phenotype, we added a combination of neurotrophic factors 28 (NTF) to the culture medium during SOD1-G128R MN maturation (differentiation days [19][20][21][22][23][24][25][26][27][28][29]. This resulted in significant decrease in cytoplasmic oxidation to the normal level. The mitochondrial level of H 2 O 2 , however, was not affected by the NTF addition (Fig. 5c).
Standard culture conditions are not always beneficial for the redox studies because of the protective influence of the standard medium. We removed the B-27 supplement from the medium 24 h before live imaging of the MN and measured the cytoplasmic and mitochondrial H 2 O 2 levels (Fig. 5d). We discovered that B-27 deprivation did not influence the cytoplasmic  (Fig 5d).  (Fig. 6c).
Culturing of SOD1-G128R MN with the NTF reduced oxidation of the cytoplasm, but it did not manage to keep it to the normal level (Fig. 6d). The mitochondrial level of H 2 O 2 , again, was unaffected by the NTF addition.  (Fig. 7a). The overall reaction of the cells was similar regardless of the culture conditions; the MN expressing Cyto-roGFP2-Orp1 demonstrated oxidation followed by slow reduction, reflecting the change in the cytoplasmic H 2 O 2 level. However, the reaction of MN cultured in the B-27-deprived medium before the experiment was more prominent. We detected a higher value of the maximum biosensor oxidation and faster reduction compared to the non-starved MN (Fig. b-c). Since the cellular reaction was affected by the components present in the standard medium, we conducted further measurements of the dynamic response on the cells that were starved before the experiment.
Using the parameters of the experiment established earlier, we recorded the reaction of MN expressing the Mito-roGFP2-Orp1 biosensor to 10 μM H 2 O 2 in real-time. We did not detect any changes in response to the exogenous H 2 O 2 ; an oxidation value for the Mito-roGFP2-Orp1 biosensor remained constant during imaging, suggesting that the mitochondrial H 2 O 2 level was also constant (Fig. 7d). The addition of H 2 O 2 in higher concentrations (25 μM and 50 μM) induced mitochondrial oxidation but, as it was found earlier, damaged the neurons. This fact forced us to abandon the measurement of the dynamic response for the Mito-roGFP2-Orp1 sensor in further experiments (Fig. 7e). Next

Motor neurons expressing the Cyto-roGFP2-Orp1 biosensor accumulate H 2 O 2 in the cytoplasm due to glutamate-induced excitotoxicity
Glutamate excitotoxicity is one of the major mechanisms of the ALS development [49].
Excessive activation of the glutamate receptors leads to an increased Ca 2+ influx, subsequent mitochondrial dysfunction, and apoptosis [50]. It is known that this process is accompanied by the increased ROS production, which connects it with the oxidative stress [51]. To test whether the Cyto-roGFP2-Orp1 and Mito-roGFP2-Orp1 biosensors can reflect redox imbalance caused by the excitotoxicity, we incubated the MN with monosodium glutamate and the glutamate reuptake inhibitor (PDC) for 5 days and measured the cytoplasmic and mitochondrial H 2 O 2 levels. Since SOD1-G128R MN (both control and experimental) died shortly after the beginning of the experiment due to the reduced viability, the measurement was conducted only for K7-4Lf, SOD1-D91A, and iALS MN. We discovered that the glutamate treatment induced the accumulation of H 2 O 2 in the cytoplasm, but not in the mitochondria, despite the known connection of mitochondrial dysfunction with the excitotoxicity (Fig. 8a-b). We observed mitochondrial oxidation in SOD1-D91A MN in both control and glutamate treated samples, although we were unable to determine if the oxidation was a hallmark of the SOD1 mutation or a technical artifact. The oxidation of the cytoplasm in K7-4Lf MN treated with glutamate was 28% higher, in iALS MNby 41% higher, in SOD1-D91A MNby 31.5% higher compared to the 35 non-treated sample (Fig. 8a). We observed that iALS MN responded to the glutamate treatment more prominently than control MN. However, we did not find the same for SOD1-D91A (Fig.   8a), which makes us suggest that this effect was not caused by the SOD1 mutation.
Further, we investigated how incubation with the monosodium glutamate affected dynamics of H 2 O 2 utilization in the cytoplasm of the Cyto-roGFP2-Orp1-expressing MN. We found that MN treated with glutamate had reduced recovery rate after the H 2 O 2 addition compared to the nontreated sample (Fig. 8c-g). We also observed the tendency towards the slower recovery of the mutant MN (36% slower for SOD1-D91A MN, 38% slower for iALS MN, 23% slower for K7-4Lf MN), however, it was insignificant.

DISCUSSION
In the present work, we established a valid approach for studying neurodegeneration in cellbased models using a combination of methods such as CRISPR/Cas9 genome editing, genetically encoded biosensors, and iPSC differentiation. Nowadays, GE biosensors are more often applied for research of different physiological and pathological processes [47,52,53] where they replaced molecular probes. In the case of redox processes, it provided information about dynamics of components of redox balance, e.g., hydrogen peroxide or GSH/GSSG ratio, in different cell compartments and tissues [23,47,54,55]. Since biosensors molecules are delivered inside the cells as plasmids for temporal expression or as viral vectors for sequence integration in the genome, the robustness of biosensors-related studies often relies on the efficiency of delivery and place of transgene integration. Here, we tried to improve the existing approach for biosensors experiments by target insertion of the cytoplasmic and mitochondrial H 2 O 2 biosensors in combination with their controlled expression. We have generated a number of IPSC lines with the insertions of both the biosensor and transactivator essential for doxycycline-dependent expression, using the simultaneous delivery of CRISPR/Cas9 RNPs and donor plasmids.
We have discovered that a single copy of the biosensor is enough for producing a sufficient signal. Moreover, it allows achieving a similar dynamic range between samples, leading to more accurate data processing [56]. However, the mean dynamic range measured in the study for 37 cytoplasmic and mitochondrial biosensors was comparable with the lowest numbers observed previously in other experiments [23,47]. This indicates that the dynamic range is directly connected with the biosensor's expression level, and more powerful promoters can improve this system [57]. The original design of the research suggested controllable induction of biosensors only on particular days during differentiation to avoid an excessive amount of fluorescent protein in the cell. However, we found a substantial decrease in fluorescence signal when it was induced only at the terminal stages of MN differentiation. Following measurement of mRNA expression confirmed a significant drop of the expression level of the biosensor, but not transactivator suggesting that maintaining the promoter in an active state prevents it from silencing [58].
Regular supplementation of doxycycline during differentiation helped to sustain a high level of the biosensors' expression in MN. However, we still observed clones that partially lost their fluorescence and became mosaic, as well as clones with generally low fluorescence intensity (data not shown). All these suggest that chromatin remodeling occurring during differentiation affects transgenes expression in the AAVS1 locus and sets a restriction on the type of transgenes that can be inserted into AAVS1. Although this problem can be overcome for biosensors [52], transgenes that are not supposed to be constantly active, such as cell-specific reporters, will not work well due to their promoter silencing [59,60]. This fact underlines the necessity to discover new safe harbors in the human genome and test them not only in terms of safety but the ability to sustain a desirable level of transgene expression.  [50]. This effect has not been shown earlier, probably because previous studies did not focus on cell recovery and used large amounts of H 2 O 2 for biosensor oxidation [23]. Possible insufficiently high dynamic range also can be the reason since it does not allow detecting small changes in H 2 O 2 concentration. This also can explain the lack of 39 mitochondria response to the glutamate: the power of the induced stress was not enough to provoke detectable changes in mitochondrial H 2 O 2 , but it didn't mean that mitochondria were not affected. We believe that the application of more advanced versions of H 2 O 2 biosensors with higher sensitivity and resolution in the future may be beneficial in such a study [65].
Correction of the mutation is more commonly used for to generate isogenic cell models [66,67].
In this work, we did not aim to discover the effect of particular SOD1 mutations but rather to test how the approaches used here can be applied for cell models. We tried to introduce the mutations in different parts of the SOD1 gene, leading to the D91A and G128R substitutions in the polypeptide [40,41]. And since no homozygous variants have been found, we selected the lines that have one allele with the target mutation and the other with large deletion (SOD1-D91A) or premature termination codon (SOD1-G128R) to maximize potential damage and make pathological phenotype more perceptible.
We found that MN derived from the SOD1-G128R iPSC line demonstrated significant differences in functioning that were detected by the biosensors. The most apparent is the higher level of cytoplasmic and mitochondrial oxidation that was found in the MN. Moreover, SOD1-G128R MN reacted more prominent than others to the stress tests such as starvation, induced by B-27 deprivation and culturing in the antioxidant-deprived medium. Notably, immature SOD1-G128R MN did not show differences in cytoplasmic H 2 O 2 , suggesting that the pathological phenotype requires time to develop, as it happens in ALS patients, who usually develop the symptoms after a certain age [68]. Neurotrophic factors affected cytoplasmic but not mitochondrial H 2 O 2 level, although it became less apparent in the antioxidant-deprived medium.
This suggests that SOD1-G128R MN functioning only relies on a highly protective culture system, and these MN are unable to sustain their viability without additional help. An elevated level of mitochondrial H 2 O 2 regardless of the NTF addition shows the presence of unresolved mitochondrial malfunction that could not be corrected at that point. Presumably, mitochondrial dysfunction of SOD1-G128R MN resulting from the presence of mutant SOD1 in the cell leads to the redox imbalance and H 2 O 2 accumulation inside the mitochondria at first [69]. Then, when the H 2 O 2 level reaches the limit of buffering capacity in mitochondria, H 2 O 2 enters the cytoplasm, probably due to the mitochondria death and fragmentation, initiating apoptosis. The presence of mitochondrial dysfunction also explains the slower axon growth since the lack of energy due to the dysfunction forces the cell to focus on survival rather than rebuilding the cytoskeleton [70].
Despite the significant, albeit not so prominent, decrease in axonal growth rate observed for

CONCLUSION
We developed a cell-based platform that allowed us to study redox balance in the live motor neurons with the detection of basal oxidation of the cytoplasm and mitochondria and the dynamic response of the cells to the stressors. Using isogenic iPSCs, we confirmed that mutations affecting different parts of the SOD1 sequence have a different impact on motor