Poly-GR repeats associated with ALS/FTD gene C9ORF72 impair translation elongation and induce a ribotoxic stress response in neurons

Abstract

Hexanucleotide repeat expansion in the C9ORF72 gene is the most frequent inherited cause of amyotrophic lateral sclerosis (ALS) and frontotemporal dementia (FTD). The expansion results in multiple dipeptide repeat proteins, among which arginine-rich poly-GR proteins are highly toxic to neurons and decrease the rate of protein synthesis. We investigated whether the effect on protein synthesis contributes to neuronal dysfunction and degeneration. We found that the expression of poly-GR proteins inhibited global translation by perturbing translation elongation. In iPSC-differentiated neurons, the translation of transcripts with relatively slow elongation rates was further slowed, and stalled, by poly-GR. Elongation stalling increased ribosome collisions and induced a ribotoxic stress response mediated by ZAKα that increased the phosphorylation of the kinase p38 and promoted cell death. Knockdown of ZAKα or pharmacological inhibition of p38 ameliorated poly-GR–induced toxicity and improved the survival of iPSC–derived neurons from C9ORF72 ALS/FTD patients. Our findings suggest that targeting the RSR may be neuroprotective in patients with ALS/FTD caused by repeat expansion in C9ORF72.

INTRODUCTION

Amyotrophic lateral sclerosis (ALS) is a devastating adult-onset motor neuron degeneration disease for which no effective treatment is available (1). Frontotemporal dementia (FTD), the second most common form of dementia, causes behavior and language abnormalities (2). ALS and FTD share many pathological and genetic similarities. Around 40% of familial ALS and 25% of inherited FTD cases are caused by GGGGCC repeat expansion in the intron region of the C9ORF72 gene (3, 4). Through repeat-associated non-AUG (RAN) translation, five dipeptide repeat (DPR) proteins can be produced from the six reading frames of both sense [(GGGGCC)n; poly-GA, poly-GR, poly-GP] and antisense [(CCCCGG)n; poly-PA, poly-PR, poly-GP] repeat–containing transcripts (5–7).

Arginine-containing DPR proteins (R-DPRs: poly-GR and poly-PR) have been shown to be the most toxic by multiple studies (8–10). They affect the formation of membrane-less organelles and liquid-liquid phase separation dynamics (11). These changes can lead to nucleolar dysfunction (9, 11–13) and stress granule impairment (11, 13). Poly-GR and poly-PR also impact nucleocytoplasmic trafficking (14), axonal transport (15) and mitochondrial functions (16), etc. In particular, the pathology of poly-GR has been shown to correlate with neurodegeneration in C9ORF72-ALS/FTD patient tissues (22, 23), implicating the contribution of poly-GR to the disease pathogenesis. Several proteomic studies identified ribosomal proteins as interactors with poly-GR and poly-PR (11, 16, 24–27). Poly-GR has been shown to colocalize with ribosome proteins in the cytosol of the AAV-GR₁₀₀ mice and patient postmortem brain tissues (18, 25). Moreover, poly-GR expressing cells exhibited a decreased rate of protein synthesis (18, 24, 28). A study has unveiled the structure of R-DPRs binding to the ribosome exit tunnel by cryo-electron microscopy (29). These observations suggest that poly-GR may compromise the global translation system in C9ORF72-ALS/FTD patients. However, the exact mechanism of how the in vivo translatome is affected has not been explored.

Eukaryotic cells have multiple quality control and stress response pathways to respond to aberrant translational events. These pathways are triggered by endogenous and exogenous insults that lead to slow or stalled elongating ribosomes, such as defective mRNA, poor codon content, damaged nucleotides, nutrient starvation, or ribotoxin (30, 31). Slowed translating ribosomes increase the chance of collision, which can be detected by sensor proteins (e.g. ZNF598, ZAKα, GCN1) for activation of different translation surveillance pathways (30, 31). It has been reported that global ribosome collisions activate GCN2, which can phosphorylate eIF2α to activate the integrated stress response (ISR) pathway, an attempt to restore cellular homeostasis by shutting down the global translation. If collisions persist, activated ZAKα (MAP3K20) phosphorylates p38/JNK and stimulates the ribotoxic stress response (RSR), which elicits inflammation and apoptosis (31–33). Given the negative impact of poly-GR on translation, the potential roles of these signaling pathways in neurodegeneration need to be explored.

In this study, we demonstrate that poly-GR inhibits global translation elongation and increases ribosome stalling in induced pluripotent stem cell (iPSC)-differentiated neurons (iPSNs). We identified transcripts with relatively slow elongation rates that tend to be further stalled by poly-GR thus increasing the chance of ribosome collisions and sensitizing the neurons to the activation of the ZAKα-mediated RSR pathway. Knockdown (KD) of ZAKα by CRISPRi can inhibit p38 phosphorylation and improve neuron survival. Moreover, we found that both ZAKα reduction and p38 small molecule inhibitor treatment could improve the survival of C9ORF72-ALS/FTD patient-derived iPSNs. Our study reveals a molecular mechanism of poly-GR mediated toxicity on global translation and the ribosome stress pathway and identifies the RSR as a potential therapeutic target for treating C9ORF72-ALS/FTD.

RESULTS

Poly-GR associates with 60S large ribosome subunits and suppresses protein synthesis

To investigate how poly-GR affects translation, we constructed HeLa Flp-In cells stably expressing the GR₅₀ protein with randomized codons under the Tet-ON inducible promoter (Figure 1A). The small split NanoLuc tag, HiBiT (34), was fused to the N terminus and the FLAG tag was fused to the C-terminus of the construct for the quantification of reporter expression. GFP and GA₅₀ with the same tag was used as the negative control. After induction, GR₅₀, GFP and GA₅₀ had similar expression level (fig. S1, A and B). GR₅₀ and GA₅₀ were predominantly localized in cytoplasm (fig. S1C). Puromycin incorporation assay was used to test whether poly-GR affected total protein synthesis. Cells were incubated with puromycin, an analog of Tyrosyl-tRNA, to label newly synthesized proteins. The incorporation of puromycin into the growing peptide chain was then detected by immunoblotting with puromycin-specific antibody (35). HeLa reporter cells expressing GR₅₀ showed decreased puromycin incorporation compared with GFP or GA₅₀ control (Fig. 1, B and C), indicating a decrease in the newly synthesized protein. The expression of GR₅₀ reduced global protein synthesis, consistent with previous reports (18, 24, 28). To assess the potential interaction of poly-GR with translating ribosomes, we performed polysome profiling experiment. GR₅₀ co-sedimented with the 80S monoribosome and polyribosome fractions, whereas GFP or GA₅₀ was only found in the non-ribosome fractions (Fig. 1, D and E). Furthermore, by treating the cell lysates with EDTA, which can induce dissociation of mono- and poly-ribosomes (Fig. 1D; only 40S and 60S peaks remained), we observed the GR₅₀ was uniquely located in the 60S fraction (Fig. 1E). Altogether, these results indicate that poly-GR can associate with translating ribosomes through the interaction with the 60S subunits in cells, which is in line with the in vitro cryo-EM structure showing the binding of poly-GR with the 60S ribosome exit tunnel (29).

Fig. 1. Poly-GR associates with 60S large ribosome subunit and suppresses translation elongation.

(A) Schematic diagram of the GFP, GR₅₀, or GA₅₀ expressing reporter constructs that are stably engineered in HeLa Flp-In cells. (B) Puromycin incorporation assay of HeLa Flp-In reporter cells. Immunoblotting of metabolic labeling with puromycin, and ponceau staining as loading control. (C) Quantification of three independent biological replicates of the puromycin incorporation assay. Data are mean ± SEM. *P<0.05, one-tailed Mann-Whitney test. (D) Polysome profiles of HeLa Flp-In reporter cells expressing GFP, GR₅₀, or GA₅₀ with no treatment or treated with 20 mM EDTA. (E) Quantification of GFP, GR₅₀ or GA₅₀ expression across the polysome profiles in (D) by N-terminal HiBiT tag bioluminescence measurement. The relative levels in each fraction were calculated as percentage of the total levels from all the fractions. (F) Representative polysome profiles of HeLa Flp-In reporter cells for ribosome runoff assay before (0 min) and after harringtonine treatment (2 μg/ml, 2 min and 5 min). (G) Quantification of polysome (blue shade) to monosome (grey shade) ratio of (F) from four biological replicates (area under the curve). Data are mean ± SEM. *P<0.05, ANOVA with Bonferroni’s multiple comparisons.

Poly-GR suppresses translation elongation

To further investigate the role of poly-GR in global translation inhibition, we employed a ribosome runoff assay to study the ribosome translocation rate along mRNAs during elongation. Harringtonine treatment blocks the first round of peptide bond formation and thus effectively stops translation initiation, but the already initiated ribosomes continue to translate. Cycloheximide treatment stops translation elongation and freezes the ribosome on the mRNA transcripts. Therefore, treating cells with harringtonine followed by cycloheximide at different intervals will reveal how fast ribosomes run off the transcripts allowing for a calculation of relative elongation speed (36, 37). Under normal conditions, we observed a rapid increase of the 80S and decrease of the polysome fractions within 2 min of the harringtonine treatment, and most ribosomes completely ran off by 5 min (Fig. 1F). The ribosomal translocation rate is reflected in the ratio of polysome/80S peak area at an intermediate time point of the runoff assay from the polysome profile. We found the polysome/80S ratio was higher in the GR₅₀ cells than in the GFP cells at the 2 min runoff time point (Fig. 1, F and G), suggesting that the ribosomes run more slowly on the mRNAs in GR₅₀-expressing cells.

The effect of poly-GR on slowing down ribosome translocation was also confirmed by directly visualizing translation dynamics in live cells using the single molecule imaging of nascent peptides (SINAPS) technology (38, 39). In this system, mRNA was labeled with the MS2-MCP system, which was placed in the 3’ untranslated region (UTR). To limit mRNA mobility for long-term tracking, mRNAs were tethered to the cell membrane with MCP fused to the CAAX motif (40). The 24×SunTag epitope with AUG start codon was fused at the N-terminus of the NLuc open reading frame (ORF) (Fig. 2A). In the cell, the nascent protein containing the SunTag epitopes is immediately bound by a fluorescent nanobody, single chain variable fragment fused with super folder GFP (scFv-sfGFP) (41), stably engineered in the cells. The translation events were detected by SunTag signal that colocalizes with single RNA molecules. We performed a runoff experiment coupled with live cell imaging to quantify how fast the translation is completed on each RNA molecule with or without poly-GR. To monitor the effect of GR on translation in live cells, we added synthesized GR₂₀ to the culture media, which has been shown to penetrate the cells (12). Indeed, GR₂₀ was universally taken up by the cells (fig. S2). After a 30-min treatment, we added harringtonine to stop the translation initiation. Subsequently, the translation of single mRNA molecules was continuously imaged and tracked for 30 min (Fig. 2B and movie S1). As expected, harringtonine treatment led to a steady reduction of translation intensity over the time (Fig. 2C). We calculated the survival probability of the translation signal, which is the percentage of translating mRNA as a function of time after adding harringtonine. We observed that most of the translation signal on the AUG-SunTag-Nluc mRNA disappeared (ribosome runoff) within 10 min (Fig. 2C). However, for cells preincubated with GR₂₀ peptide, the survival curve was shifted to the right, indicative of a prolonged runoff time (Fig. 2D). The median runoff time for the NLuc reporter is 5 min, whereas it was 15 min in cells where GR₂₀ is expressed (Fig. 2E). These results provide direct evidence that poly-GR can slow down the translation elongation rate on mRNAs.

Fig. 2. Single molecule imaging of AUG-SunTag-NLuc reporter showed poly-GR slows down translation elongation.

(A) Diagram of the single molecule AUG-SunTag-NLuc reporter construct. (B) Snap shots from videos of ribosome runoff experiment for CTRL and GR₂₀ treated cells, respectively. Scale bar = 2 μm. Example translation intensity traces for (C) CTRL and (D) GR₂₀ treated cells after harringtonine treatment. (E) The survival curves of translation sites as a function of time after harringtonine treatment. Dashed line represents 50% of mRNAs with completed runoff (disappearance of translation signal on RNA). Shadow areas are 95% confidence bounds (Greenwood’s formula). CTRL: 10 cells, 125 TLS; GR₂₀: 8 cells, 178 TLS. ****P<0.0001, log-rank Mantel-Cox test.

Together, the results from both biochemical approaches and single molecule imaging experiment demonstrated that poly-GR affects protein synthesis by slowing down translation elongation.

Poly-GR influences ribosome translocation on specific mRNAs in i³Neurons

We further dissected the molecular mechanisms of poly-GR-mediated translational inhibition in human neurons using the CRISPRi-integrated i³Neuron system (CRISPRi-i³N) (42). The i³Neuron differentiation yields large quantities of highly homogenous neurons by induction of a neuronal transcription regulator neurogenin-2 (NGN2) engineered in the AAVS1 safe-harbor locus in iPSCs under a doxycycline-inducible promoter (43). We stably expressed either GFP, GA₅₀ control or GR₅₀ with the same tags as in HeLa reporters (Fig. 1A) in i³Neurons by lentiviral transduction (fig. S3, A and B). Similar to HeLa reporter cells, we observed a slight decrease of puromycin incorporation in GR₅₀ expressing i³Neurons compared with control cells (fig. S3C), confirming the reduction of protein synthesis. However, quantification of polysome/80S peak area in the ribosome runoff experiment did not show a significant difference in the translocation rate between GFP and GR₅₀ (fig. S3D).

It is possible that the translation elongation of a subset of mRNAs is preferentially influenced by poly-GR, which cannot be revealed by the overall polysome profile. We therefore assessed the transcriptome-wide RNA translation elongation by coupling RNA-seq with ribosome runoff profiling. We collected GFP or GR₅₀-expressing i³Neurons with 0-min or 3-min harringtonine treatment for polysome profiling and extracted mRNAs from the medium (3–7 ribosomes) and heavy (>7 ribosomes) polysome fractions as well as the input total RNA for high-throughput sequencing (Fig. 3A). We identified the genes that exhibit statistically significant variability across different conditions (data file S1). Unsupervised hierarchical clustering revealed that the distribution of these transcripts showed different response after ribosome runoff, yet the total input RNA levels were not changed (Fig. 3B). This suggests that GR₅₀ affects their translation dynamics without changing the total RNA expression level.

Fig. 3. Poly-GR influences ribosome translocation on specific mRNAs in i³Neurons.

(A) Schematic diagram of ribosome runoff coupled with mRNA-seq assay. mRNA from input, medium and heavy polysome fractions were subjected to high-throughput RNA-seq. (B) Unsupervised hierarchical clustering and heatmap of all differentially expressed genes with distinct ribosome runoff patterns. Differentially expressed genes were detected via ANOVA analysis, with adjusted P-value < 0.05 from F-test. (C) KEGG pathway enrichment of genes in clusters 3 and 4 in (B). (D) Normalized read counts of ABCE1 in heavy fractions from runoff-RNA-seq shown by IGV. (E) qRT-PCR validation of ABCE1 from ribosome runoff assay. Data are mean ± SEM, N=6. *P<0.05, Kruskal-Wallis test. (F) Genes stalled by poly-GR were enriched with known stalling motifs. The number of known stalling motifs was counted for each gene. (G) Immunoblots of ABCE1 and quantification in i³Neurons expressing GFP, GR₅₀ or GA₅₀ for 10 days from three independent experiments. Data are mean ± SEM. **P<0.01, Kruskal-Wallis test. (H) Immunoblots of ABCE1 and quantification from 5 control and 5 patient iPSN lines. β-actin was blotted as internal control. Data are mean ± SEM. **P<0.01, two-tailed unpaired Mann–Whitney test. (I) qRT-PCR quantification of ABCE1 RNA levels from the same 5 control and 5 patient iPSN lines. Data are mean ± SEM. Two-tailed Mann-Whitney test. (J) Cumulative fraction of the protein levels comparing C9 versus CTRL groups from the AnswerALS proteomics dataset. Red line represents translation stalled genes from the runoff-seq data. Grey line represents all the unstalled genes from the proteomics database. P value was calculated by two-sided nonparametric K-S test.

We clustered the genes across all samples into eight distinct groups through K-means clustering (data file S1). Some genes (such as those in clusters 7 and 8) showed reduced distribution in polysome fractions after harringtonine treatment, indicating fast elongation as most ribosomes ran off the transcripts within 3 min after inhibiting the translation initiation. On the other hand, some genes (those in clusters 2–5) showed elevated distribution in heavy polysome fractions after harringtonine treatment, likely due to slow elongation rate or translation stalling on these transcripts. As expected, we noticed that transcripts with longer coding sequence (CDS) tend to run off more slowly (fig. S3E). Notably, mRNAs in clusters 3 and 4 showed an increased distribution in heave polysome fractions in GR₅₀-expressing i³Neurons than in the GFP control sample treated with harringtonine (Fig. 3B). These observations suggest that the elongation of ribosomes on these transcripts is slowed by poly-GR. It is noted that GR₅₀ did not affect the basal level of the RNA distribution in the polysome fractions without harringtonine treatment (Fig. 3B, GR₅₀-0min vs GFP-0min). As the steady state of mRNA association with polysomes correlates with translation initiation rate, these data suggest that poly-GR does not influence the translation initiation of these transcripts, but instead affects ribosome translocation during elongation. In particular, ribosomes on mRNAs with already slow elongation rates tend to be further hindered by poly-GR. Functional enrichment analysis identified these genes as highly enriched in pathways related to ALS, as well as to protein and RNA homeostasis-related functions (Fig. 3C), all closely related to disease pathogenesis in C9ORF72-ALS/FTD (44–46).

We examined and validated selective genes from clusters 3 and 4. ABCE1 (ATP binding cassette subfamily E member 1) is a co-translational quality control factor involved in the No-Go Decay (NGD) pathway, and also plays an important role in mitophagy (47, 48). The normalized read counts from heavy fractions showed higher peaks in GFP-3min samples than GFP-0min samples and this increase was even more profound in GR₅₀-3min compared to GR₅₀-0min samples (Fig. 3D). We validated by qRT-PCR quantification of the RNA level in the heavy fractions from GR₅₀ and GFP neurons before or after the ribosome runoff (Fig. 3E). Consistent with the sequencing result, GR₅₀ expression did not change the total expression level of ABCE1 mRNA, but increased its distribution in heavy fractions upon the ribosome runoff compared to the GFP control (Fig. 3E), supporting the slowed ribosome translocation. Other examples include HSP90AA1 (heat shock protein 90 alpha family class A member 1), STAU2 (Staufen double-stranded RNA binding protein 2), and ATXN10 (Ataxin-10), which all showed slowed runoff by poly-GR via both RNA-seq and qRT-PCR analysis (fig. S3, F and G).

We analyzed potential features of the mRNAs that are preferentially stalled by poly-GR. We screened the genes with known pausing motifs (49–51). These motifs were identified by multiple disome profiling studies, indicative of potential ribosome collision sites triggered by ribosome stalling. We found that the transcripts affected by poly-GR were enriched with the known stalling motif (Fig. 3F). Additionally, when performing a systematic scan for tri-amino acid (tri-AA), we observed a prevalence of pausing motifs containing arginine or lysine in genes impacted by poly-GR (fig. S3H). This indicates that the translation elongation of mRNAs containing potential stalling motifs is prone to be affected by poly-GR. Collectively, these data indicate poly-GR could modulate specific gene expression in neurons by slowing down or stalling their translation elongation.

We next tested whether the translation deficits could influence the protein expression level independent of the RNA changes. We examined the protein level of ABCE1 and found that indeed the protein expression was reduced in GR₅₀-expressing neurons (Fig. 3G), although its total mRNA level was not changed (Fig. 3E). We also observed similar results in iPSNs derived from C9ORF72-ALS/FTD patients and non-neurological controls. The ABCE1 protein was reduced in patient iPSNs while the mRNA level was not changed (Fig. 3, H and I). The poly-GR expression in patient iPSNs was verified by Meso Scale Discovery (MSD) immunoassays performed in a blinded manner (fig. S3I). We further expanded to the global analysis using the proteomic and transcriptomic data from Answer ALS (52), including nine C9ORF72-ALS/FTD lines and eight control lines of iPSNs. We found that the transcripts with increased stalling induced by poly-GR tends to show reduced protein levels in C9ORF72-ALS/FTD iPSNs compared to the ones without stalling (Fig. 3J). Eighty-eight genes were identified to have decreased protein expression levels without the corresponding mRNA reduction in C9ORF72 patient iPSNs (fig. S3J, and data file S2). We examined the protein and RNA levels of a few examples, including PML and KPNA1, which all contain stalling motifs (PPP, RRR, KKK). The protein levels were decreased in C9ORF72-ALS/FTD iPSNs, but the mRNA levels had no differences (fig. S3, K and L). In summary, while many factors could influence the protein abundance in patients, the translation defects could partially contribute to the protein reduction in C9ORF72-ALS/FTD iPSNs.

Poly-GR sensitizes neurons to the ribosome collision-induced stress response

Increased ribosome stalling raises the chance of ribosome collision, which triggers various ribosome quality control (RQC) pathways. This involves the activation of the ribosome-associated E3 ligase ZNF598, which ubiquitinates small subunit proteins at the stalled ribosomes and recruits other RQC factors to dissociate the aberrant translation intermediates on the transcripts (53, 54). If not resolved, the prolonged stalling will induce global stress response signaling pathways (30, 31, 55). The long splicing isoform of MAP3K20 (ZAKα) has been identified as the sensor for stalled and/or collided ribosomes induced by UVB, ribotoxin, or translation inhibitors (33, 56). ZAKα in turn triggers downstream activation of stress-activated protein kinases (SAPKs, p38 and JNK) and GCN2-mediated eIF2α phosphorylation, leading to RSR and ISR respectively (30, 33). It is proposed that the initial cellular response to wide-spread ribosome collisions is to block global translation initiation through eIF2α phosphorylation, but that prolonged stress activates SAPK-mediated apoptosis pathway (30, 31).

Anisomycin is commonly used to inhibit translation elongation. It binds to the A site of the peptidyl transferase center on 60S subunit thus blocking peptide bond formation. Anisomycin has been used to induce ribosome collisions when applied at intermediate concentrations and has been shown to activate the eIF2α, p38 and JNK pathways in several cell lines (33). If applied at high concentration, every ribosome is halted, then no ribosome collision will happen. How human neurons respond to ribosome collisions was not reported previously. We first tested the ISR and RSR response in the control i³Neurons. We observed increased phosphorylation of p38 in response to intermediate concentrations of anisomycin (Fig. 4, A and B). This activation was in line with the increased ubiquitylation of RPS10 (Fig. 4, C and D), which is a known indicator of ribosome collisions downstream of ZNF598 (54). Notably, there was no increase of JNK or eIF2α phosphorylation (Fig. 4A). This suggests that the ribosome collision preferentially triggers RSR, especially p38 activation, rather than ISR in human i³Neurons.

Fig. 4. Poly-GR activates p38 MAPK through ZAKα-mediated ribotoxic stress pathway.

(A) Immunoblots for phosphorylated p38, JNK, eIF2α and total p38 in i³Neurons expressing GFP or GR₅₀, with no or increasing dose of ANS treatment (0, 0.004, 0.02, 0.1 μg/ml) for 20 min. (B) Quantification of phosphorylated p38 (p-p38) normalized to p38 from (A). N=5, Data are mean ± SEM. *P<0.05, two-way ANOVA, uncorrected Fisher’s LSD test. Both p-p38 bands are quantified. (C) Immunoblots for ubiquitylated RPS10 in i³Neurons expressing GFP or GR₅₀, with increasing ANS treatment (0, 0.004, 0.02, 0.1 μg/ml) as in (A). *non-specific band. (D) Quantification of ub-RPS10 normalized to total RPS10 from (C). N=3, Data are mean ± SEM. *P<0.05, two-way ANOVA, uncorrected Fisher’s LSD test. (E) Immunoblots for phos-tag gels detecting phosphorylated ZAKα in i³Neurons stably expressing WT or K45A mutant ZAKα. The cells were either treated with 0.1 μg/ml ANS or transduced with GFP or GR₅₀ lentivirus. (F) Quantification of phos-ZAK normalized to total ZAK from (F). N=3, Data are mean ± SEM. *P<0.05, **P<0.01, two-way ANOVA followed by Tukey’s post hoc test. (G) Immunoblots for phosphorylated p38, JNK, and eIF2α in i³Neurons with ZAKα knockdown or with exogenous expression of WT or K45A ZAKα. The cells were transduced with GFP or GR₅₀ lentivirus. (H) Quantification of p-p38 normalized to p38 from (G). N=3, Data are mean ± SEM. *P<0.05, **P<0.01, two-way ANOVA followed by Tukey’s post hoc test. (I) Immunoblots for phosphorylated p38, JNK, and eIF2α in the C9ORF72-ALS/FTD patient-derived iPSN and the isogenic control line. (J) Quantification of p-p38 normalized to p38 from (I). N=3, Data are mean ± SEM. *P<0.05, two-way ANOVA followed by Bonferroni’s multiple comparison.

We next examined whether poly-GR could influence the RSR pathway due to its effects on global translation elongation. The expression of GR₅₀ slightly increased the basal level of phospho-p38, and more drastically enhanced p38 activation upon treatment with an intermediate amount of anisomycin (Fig. 4, A and B), whereas the expression of GA₅₀ did not enhance p38 signaling compared with GFP (fig. S4, A and B). The ubiquitylation of RPS10 was also increased by poly-GR expression (Fig. 4, C and D) but not by poly-GA expression (fig. S4, C and D). Additionally, acute treatment of the neurons with increasing amounts of synthetic GR₂₀ peptide showed dosage-dependent induction of p38 phosphorylation (fig. S4, E and F), supporting the model wherein there are increased ribosome collisions in GR-expressing neurons. Together, poly-GR can sensitize the neurons to the ribosome collision-induced RSR pathway.

Poly-GR activates p38 MAPK through ZAKα-mediated RSR pathway

We next examined whether ZAKα mediates the ribosome collision-induced p38 activation in neurons and whether it is essential for the poly-GR–enhanced p38 phosphorylation. We generated iPSC lines expressing either the non-targeting control or ZAKα-targeting CRISPRi sgRNA and differentiated to i³Neurons. ZAKα was effectively knocked down (fig. S4G), and its reduction eliminated p38 activation upon the intermediate dose of anisomycin treatment (fig. S4, H and I). We expressed either wild-type (WT) ZAKα or ATP-binding defective mutant (K45A) in the ZAKα knockdown neurons. As expected, only the WT ZAKα enhanced the phosphorylation of p38 upon treatment with anisomycin, but the K45A mutant had minimal effect (fig. S4, H and I). We further directly assessed the ZAKα activation by measuring its auto-phosphorylation status, as its kinase activation is also important for its own phosphorylation (57). Anisomycin treatment induced a shift of the ZAKα band on the phos-tag gel, caused by the slower migration of the auto-phosphorylated protein (58), which did not occur in the K45A mutant (Fig. 4, E and F). Altogether, these data support a model wherein ZAKα kinase activity can be induced by ribosome collisions to stimulate the p38 phosphorylation through the RSR in human neurons.

We next used this platform to examine the poly-GR mediated ribosome stress response. Expression of poly-GR significantly increased the phosphorylation level of WT ZAKα but not the K45A mutant (Fig. 4E, lanes 3 and 6, and Fig 4F). This is in line with the increased abundance of phosphorylated p38 only in the WT ZAKα neurons (fig. S4, J and K). Of note, knockdown of ZAKα abolished p38 activation induced by GR₅₀ expression (Fig. 4, G and H). Similar results were also observed in the GR₂₀ peptide treated cells (fig. S4, L and M). Furthermore, we also examined the activation of the eIF2α, p38 and JNK signaling pathways in a pair of C9ORF72-ALS/FTD patient iPSN and its isogenic control in which the repeat expansion has been removed (59). The patient iPSNs showed significantly increased p38 activation at both basal level and upon ANS treatment compared to the control (Fig. 4, I and J). Collectively, these data support that the poly-GR activates p38 MAPK through ZAKα-mediated RSR pathway.

Inhibition of the RSR reduces the poly-GR mediated toxicity in i³Neurons

We next examined whether the activation of the p38 MAPK pathway contributes to the poly-GR induced toxicity in i³Neurons. The neuron survival rate was significantly reduced by GR₅₀ expression and moderately influenced by GA₅₀ (fig. S5, A and B). The result was further validated by propidium iodide (PI) staining of dead cells, as PI can only bind to the DNA where the plasma membrane has been compromised. A significantly increased PI signal was observed in both GR₅₀- and GA₅₀-expressing neurons compared to GFP-expressing neurons (Fig. 5, A to D). Another independent cell death quantification assay by measuring the LDH release to the culture media also showed the same trend (Fig. 5, E and F), confirming the DPR-induced cell death of i³Neurons. We expressed GFP, GR₅₀ or GA₅₀ in either the control or ZAKα CRISPRi neurons. ZAKα knockdown significantly improved neuron survival and reduced the cell death caused by poly-GR but not poly-GA (Fig. 5, A, B and E, and fig. S5A). These results support a model wherein the inhibition of a ZAKα-mediated RSR pathway can protect the neurons from specifically poly-GR-induced toxicity.

Fig. 5. Genetic knockdown of ZAKα or p38 inhibitor improves neuronal survival.

(A) Representative images of PI staining and bright field imaging of i³Neurons differentiated from control or ZAKα knockdown iPSCs. GFP, GR₅₀ or GA₅₀ were expressed on differentiation day 5 and images were taken on day 14. Scale bar = 60 μm. (B) Cell death quantification by PI staining of i³Neurons in (A). Each dot represents a biological replicate, >300 cells per replicate from total three replicates. Data are mean ± SEM. *P<0.05, **P<0.01, ANOVA with Tukey’s multiple comparison. (C) Representative images of PI staining of i³Neurons expressing GFP, GR₅₀ or GA₅₀ treated with or without VX-745. Scale bar = 60 μm. (D) Cell death quantification by PI staining of i³Neurons in (C). Each dot represents a biological replicate, around 300 cells/replicate from three replicates. Data are mean ± SEM. *P<0.05, ANOVA with Tukey’s multiple comparison. (E) Cell death quantification by LDH release assay in control or ZAKα knockdown cells expressing GFP, GR₅₀ or GA₅₀. Three independent experiments. Data are mean ± SEM. **P<0.01, ANOVA with Tukey’s multiple comparison. (F) Cell death quantification by LDH release assay in i³Neurons expressing GFP, GR₅₀ or GA₅₀ treated with or without VX-745. Three independent experiments. Data are mean ± SEM. *P<0.05, ANOVA with Tukey’s multiple comparison. (G) Diagram of i³Neuron differentiation and experiment timeline. (H) Representative images of Hoechst and PI staining of control and C9ORF72-ALS/FTD patient iPSNs in the glutamate-induced excitotoxicity assay. The iPSNs were infected with lentivirus expressing control or ZAKα shRNA ten days prior to the experiment. Scale bar = 30 μm. (I) Quantification of neuronal death by PI staining upon glutamate induced excitotoxicity in (F). Different shapes represent individual iPSN lines from 5 control and 5 patients. > 800 cells were quantified per line per condition. Data are mean ± SEM. **P<0.01, ***P<0.001, two tailed paired Student’s t test. (J) Representative images of control and patient iPSNs in the glutamate-induced excitotoxicity assay. Cells were pretreated with DMSO or 10 μM VX-475 4 hours prior to the experiment. Scale bar = 30 μm. (K) Quantification of neuronal death by PI staining in (H). Different shapes represent individual iPSN lines from 5 control and 5 patients. >800 cells per line per condition. Data are mean ± SEM. **P<0.01, ***P<0.001, two tailed paired Student’s t test.

We next examined whether direct targeting p38 using pharmacological method can have similar protective effect on GR-expressing neurons. Many p38 small molecule inhibitors have been developed for cancer and inflammatory diseases (60, 61). VX-745 (Neflamapimod), a selective and orally active p38α inhibitor with an IC₅₀ of 10 nM (62), was first developed for treating rheumatoid arthritis (63) but was later applied to therapy development for neurodegenerative diseases (64). In aged rat models, VX-745 treatment decreased IL-1β levels and improved behavioral performance in Morris water maze test (64). In a preclinical study, VX-745 improved behavior in novel object recognition and open field tests in a mouse model with lysosomal pathology and cholinergic degeneration (65). Because p38 activation is increased by poly-GR, we thus tested the effects of VX-745 on poly-GR induced neuronal death. We treated the i³Neurons with VX-745 after expressing GFP, poly-GR or poly-GA (Fig. 5G). We found that VX-745 could rescue the toxicity of poly-GR and reduce the neuronal death, but not the toxicity of poly-GA (Fig. 5, C, D and F, and fig. S5B). Altogether, these results show that the ZAKα-mediated RSR pathway contributes to poly-GR–induced neurotoxicity.

Inhibition of the RSR improves the survival of C9ORF72-ALS/FTD patient iPSNs

We next sought to determine whether inhibiting the ZAKα-mediated RSR is beneficial for neuronal survival of the C9ORF72-ALS/FTD patient iPSNs. We expressed an shRNA targeting ZAKα in differentiated patient iPSNs via lentiviral transduction which resulted in more than 85% reduction (fig. S5, C and D). C9ORF72-ALS/FTD iPSN is known to be sensitive to glutamate-induced excitotoxicity (66). Therefore, we treated day-32 control and C9ORF72-ALS/FTD patient-derived iPSNs with glutamate and measured cell death by PI staining. We observed significantly increased cell death in glutamate treated C9ORF72-ALS/FTD iPSNs whereas no such neuronal loss was detected in the control iPSNs (Fig. 5, H to K). The neurotoxicity induced by glutamate was ameliorated in the patient iPSNs when transduced with ZAKα shRNA (Fig. 5, H and I), suggesting the reduction of ZAKα can improve the neuron survival of C9ORF72-ALS/FTD patient-derived iPSNs. Similarly, pretreating cells with the p38 inhibitor VX-745 also decreased patient neuronal death induced by glutamate (Fig. 5, J and K). The rescue effect was not due to the decrease of poly-GR level, but rather through the downstream signaling, because we did not detect decreased poly-GR levels after either ZAKα knockdown or VX-745 treatment (fig. S5, E and F). Together, these data support that targeting the ZAKα-mediated RSR pathway can bring protective effect on the C9ORF72-ALS/FTD patient iPSNs.

DISCUSSION

This work uncovers that poly-GR impairs global translation by slowing down translation elongation, which increases ribosome stalling and collision, and sensitizes neurons to ZAKα-mediated RSR signaling. The findings align with published results on arginine-containing DPR toxicity on translation machinery (29, 67). We further identified that inhibition of the RSR pathway not only improves neuron survival in poly-GR–expressing cells, but also in C9ORF72-ALS/FTD patient–derived iPSNs. These results reveal molecular mechanisms of gene dysregulation by translation defects and provide insights on targeting RSR as a potential therapeutic strategy to explore for C9ORF72-associated ALS/FTD (Fig. 6).

Fig. 6. Working model of poly-GR mediated toxicity through translation impairment and RSR activation.

Poly-GR slows down translation elongation and sensitizes neurons to a ZAKα-mediated ribotoxic stress response (RSR). Inhibition of this pathway improved the survival of neurons and may have therapeutic potential for C9ORF72-ALS/FTD patients. Images created with BioRender.

The polysome profiling results indicate that poly-GR binds to the ribosome, specifically to the 60S subunits, which is corroborated by the reported cryo-EM structure (29). Previous studies have shown that the translation of poly-GR itself from the GGGGCC RNA repeat could promote ribosome stalling and fail to be properly resolved by canonical RQC pathways (67, 68). Besides the cis-acting toxicity mechanism, we showed that the poly-GR could also impair global translation elongation in trans by runoff assays using both bulk polysome profile and single molecule imaging technique. Using ribosome runoff coupled with RNA-seq, we assessed the global effects and identified specific genes that undergo slowed translation elongation due to the presence of poly-GR in human neurons. These genes are enriched in ALS, synapse, and protein/RNA homeostasis-linked pathways. Their dysregulation likely contributes to the disease pathogenesis. We also found that the transcripts with stalling motifs and relatively low elongation rates tend to be further slowed down by poly-GR, suggesting the gene-specific susceptibility to this toxicity. It has been found that the protein expression levels do not always correlate with the RNA levels in C9 patient iPSNs. Besides the possibility of protein turnover differences, the translation defects could contribute to this discrepancy. The multi-omics analysis of patient iPSN and Drosophila studies also support this notion (69), in which, through an RNAi-based screening in Drosophila, disruption of translation-related genes play causal roles in the disease model.

Besides influencing the translation of specific transcripts, poly-GR also affects the stress signaling associated with the translation defects. We observed increased activation of ZAKα-p38 signaling pathway when treated with anisomycin in cells expressing poly-GR compared with GFP. Previous studies on the RSR pathway primarily focused on acute translation defects induced by translation inhibitors or UV damage (33, 56, 70). Our work suggests accumulation of the translation deleterious factor poly-GR, can increase the susceptibility of neurons to the stress pathway activation. Despite the low production of poly-GR in patients, when combined with additional insults to the translation pathway, the accumulated poly-GR in long term may eventually trigger RSR activation, potentially contributing to neuronal death. Additionally, some of the targets of poly-GR might contribute to the global translation defects. For example, we observed a reduction in the protein level of ABCE1 in C9ORF72-ALS/FTD patient neurons due to the GR-induced translation defects. ABCE1 is a co-translational quality control factor involved in the No-Go Decay (NGD) pathway and plays a role in mitophagy (47, 48). The impaired expression of ABCE1 may increase the accumulation of damaged mRNAs and faulty proteins, which contributes to the global stress signaling activation.

p38 activation has been implicated in many neurodegeneration diseases with different mechanisms (71–73). As a stress kinase, p38 can stimulate the release of proinflammatory cytokines, modulate neuronal plasticity (74) and endolysosomal function (75). In ALS models, p38 was previously reported to be activated by FUS^R521G, FUS^P525L and SOD1^G93A mutations (76–79). We now identified p38 is activated via the RSR pathway stimulated by poly-GR, revealing a novel molecular mechanism of p38 regulation in human neurons. Additionally, p38 has been shown to enhance the activation of microgliosis, accelerating the disease progression via the non-cell autonomous mechanism (80, 81). Whether the RSR pathway also contributes to the pathogenesis in microglia of C9ORF72-ALS/FTD needs further exploration.

The observed improvement in cell survival upon knockdown of ZAKα or treatment with p38 inhibitor VX-745 in GR-expressing neurons demonstrates that inhibition of the RSR pathway is sufficient to suppress the GR-mediated toxicity. Furthermore, we also showed that targeting ZAKα or p38 can bring beneficial effects in patient-derived iPSN models, highlighting the promise of targeting the RSR as a therapeutic strategy. It is worth noting that ZAKα activates RSR and p38 signaling by sensing the ribosome collision events. Reducing ZAKα will decrease the RSR response and cell death cascade activation triggered by the translation deficits, but will not directly rescue the translation abnormality. Multiple p38 inhibitors have been tested in neurons and showed improvement of neuronal survival. NMDA-induced excitotoxicity is reduced by the p38 inhibitor BIRB796 in rat cortical neuron cultures (82). SB203580 and semapimod inhibits the apoptosis of motor neurons in the SOD1^G93A mouse model (83). Currently, there are several brain-penetrating p38 inhibitors under development (84, 85). Among them, VX-745 has progressed into phase 2 clinical trials. It improved memory and learning and reduced the β-amyloid plaques in participants with mild AD (86). VX-745 is also reported to improve cognitive performance in people with dementia with Lewy bodies in a completed phase 2 clinical trial (65). These findings indicate the feasibility and premise of targeting p38 via VX-745 as a therapeutic strategy in C9ORF72-ALS/FTD patients.

In summary, our study identified a molecular mechanism as to how poly-GR can impair global translation and sensitize neurons to translational stress. Our findings support that the ZAKα-p38 signaling pathway is directly associated with GR toxicity and could be a potential therapeutic target for C9ORF72-ALS/FTD. There have been multiple studies reporting poly-GR mouse models that showed a spectrum of phenotypes with different severity, likely due to the variations of transgene or expression strategies (17–21). Nevertheless, it will be interesting to evaluate the rescue efficacy of approaches inhibiting ZAKα or p38 in vivo using C9ORF72 mouse models, especially the repeat-expressing models, in future studies.

MATERIALS AND METHODS

Plasmids

For HeLa Flp-In reporters, GFP, GR₅₀ or GA₅₀ with codon-optimized sequence (Genewiz) was cloned into pcDNA5-FRT-TO vector at the HindIII and NotI sites, with two consecutive FLAG tags at the 3’ terminus. The HiBiT-HA tag (GGTACCATGGTGAGCGGCTGGCGGCTGTTCAAGAAGATTAGCGGTTCAAGTGGATACCCATACGACGTCCCAGACTACGCTGGGTAC) was linked to the 5’ of APEX2 by PCR amplification, and the whole fragment was inserted upstream of the GFP, GR₅₀ or GA₅₀ at the KpnI and BamHI sites.

For GFP, GR₅₀ or GA₅₀ expression in i³Neurons, the above open reading frame (ORF) fragments (HiBiT-HA-APEX2-GFP/GR₅₀/GA₅₀-2×FLAG) were cut out from the pcDNA5-FRT-TO vector, and cloned into the lentivirus vector (modified from Addgene #52962 to have hygromycin resistance) via NheI and NotI. For lentiviral ZAKα-WT or ZAKα-K45A constructs, the ORF fragments were PCR amplified from Addgene plasmids #141193 and #141194, respectively, and inserted into the lentiviral vector (addgene#52962) via XbaI and BamHI.

For lentiviral CTRL shRNA (CCTAAGGTTAAGTCGCCCTCG) or ZAK shRNA (GCTTCTCTGGGATCACTCTAT) constructs, the oligos were annealed and inserted into the pLKO.1 backbone at the AgeI and EcoRI sites. The lentiviral constructs expressing CTRL sgRNA (GGACTAAGCGCAAGCACCTA) or ZAK sgRNA (GGAGGCCCCGCGCGCGACGA) were constructed following the protocol and the sequence provided in the previous study(87). Briefly, the oligos were annealed and inserted into the B9-Control vector at the BstXI and BlpI sites.

Cell culture

HeLa Flp-In cells, 293T cells and U-2 OS cells stably expressing AUG-SunTag-NLuc reporter(88, 89) were grown in DMEM supplemented with 10% (v/v) FBS, 100U/mL penicillin and 100 μg/mL streptomycin. HeLa Flp-In cell lines expressing GFP/GR₅₀ reporters were generated as described before(90). The transgene expression was induced with 2 μg/ml tetracycline (Sigma, T7660) for 2 days. 293T cells were used to package lentiviruses expressing GFP, GR₅₀, ZAKα WT/K45A, CTRL/ZAK shRNA or CTRL/ZAK sgRNA. All cells were maintained at 37°C with 5% CO2.

The i³Neuron iPSC line was obtained from the laboratory of Michael Ward, NIH (42). The iPSCs were grown in E8 medium (Thermo Fisher, A15169–01) on Matrigel-coated plates (Corning, 354277). iPSCs were differentiated into iPS neurons following the protocol (43). Briefly, iPSC cells were first seeded and cultured in neuronal induction medium for 3 days. Then cells were dissociated and seeded on 24-well plate pre-coated with poly-L-ornithine at density of 3×10⁵. Cells were maintained in BrainPhys Neuronal Medium with a half-media change every other day. To generate stable CRISPRi i³Neurons, i³N iPSCs were infected with lentivirus expressing either non-targeting control or ZAK-targeting sgRNA. Puromycin (1 μg/ml) selected for transduced cells. No obvious toxicity was observed during neuron differentiation when knocking down ZAK. To overexpress ZAKα WT/K45A, the i³Neurons with ZAK sgRNA were transduced with lentivirus carrying ZAKα-WT or ZAKα-K45A on differentiation day 5 at an MOI of 1. For expressing the GFP, GR₅₀ or GA₅₀ in i³Neurons, cells were transduced with lentiviral GFP/GR₅₀/GA₅₀ on differentiation day 10 at an MOI of 1 and harvested on day 14, or as indicated in the diagram for individual experiments.

Peripheral blood mononuclear cell (PBMC)-derived iPSC lines from C9ORF72-ALS/FTD patients and non-neurological disease controls were obtained from the Answer ALS repository at Cedars-Sinai iPSC Core (table S1). iPSCs were maintained in mTeSR medium (StemCell Technologies, 85850) on Matrigel-coated plates. iPSCs were differentiated into spinal motor neurons as previously described(52).

Cell culture medium was refreshed 2 hours before treatment. Anisomycin (Sigma-Aldrich, A9789) was dissolved in ethanol. Synthetic GR₂₀ peptide (Genscript, SC1208) was dissolved in water. Cells were treated as indicated in the figures.

Neuronal toxicity assays

For toxicity assay in i3Neurons, iPSC differentiated neurons stably expressing CTRL or ZAK sgRNA were transduced with lentivirus expressing GFP or GR₅₀ on day 5 in the BrianPhys stage. On day 14, propidium iodide (PI) staining was performed by adding to the media at 1μg/mL for 30 min. Bright field and PI images were taken by Nikon Eclipse TS100 microscope. For VX-745 treatment experiment, 4 μM VX-745 or DMSO was added to the medium on day 6, fresh drug was added when changing medium every other day. PI staining and bright field images were taken on day 14. PI-positive cells were quantified by Fiji. CytoTox 96^® Non-Radioactive Cytotoxicity Assay (Promega, G1780) was used as an alternative method to quantify cell death by measuring the lactate dehydrogenase (LDH) release to the cell media. For this assay, 50 μl media was used according to the manufacturer’s protocol.

For glutamate-induced excitotoxicity in patient-derived iPSNs, five lines of CTRL and C9-ALS iPSN were used. On day 12 of differentiation, cells were plated on Falcon 8-well culture slide (Corning, 354118), with daily media change to remove dead cells. For testing the rescue effect of ZAK knockdown, cells were infected with lentivirus expressing CTRL or ZAK shRNA on day 22. For testing the rescue effect of VX-745, cells were pretreated with 4μM VX-745 for 2 hrs before the experiment. On the day of experiment (day 32), cells were treated with 10 μM L-glutamate for 4 hrs. Then the cells were stained with Hoechst33342 (1 μg/ml) and PI (1 μg/ml) for 30 min for the quantification of total and dead cells. Images were taken with Zeiss LSM900 confocal microscope and then quantified by Fiji.

Immunoblotting

Cells were collected in cold PBS and lysed in RIPA buffer containing phosphatase inhibitor cocktail A and B (Bimake, B15002) and protease inhibitor cocktail (Millipore Sigma, 11697498001). The lysates were incubated on ice for 10 min, followed by centrifugation at 14,000 rpm at 4°C for 10 min. The supernatants were heated in Laemmli loading buffer, resolved by SDS-PAGE and transferred to nitrocellulose membrane. Membranes were blocked with 5% nonfat dry milk in TBST for 1 hour at room temperature. Blots were washed and then incubated with indicated primary antibodies diluted in 5% BSA in TBST overnight at 4°C. Blots were then washed and incubated with secondary antibodies at room temperature for 2 hours. The membranes were developed using Clarity Western ECL detection reagents (Bio-Rad, 1705061) or ECL Select Western Blotting detection (GE Healthcare, RPN2235), and imaged on a Bio-rad ChemiDoc Imaging System.

For puromycin incorporation assay, cells were incubated with 10 μM puromycin (Sigma, P8833) for 20 min, and then immediately collected and lysed in RIPA buffer containing protease inhibitors. The protein lysates were quantified and subjected to western blot analysis. Puromycin incorporation was normalized to total protein loading (ponceau S staining).

For phos-tag gel immunoblotting, cell lysates were resolved in 6% SDS-PAGE with 10 μM Phos-tag Acrylamide (Wako, AAL-107) and 20 μM ZnCl₂, and transferred to PVDF membrane overnight. The rest procedure is as above and membranes were developed using SuperSignal West Pico Plus ECL substrate (Thermo Fisher Scientific, 34580) and imaged on a Bio-rad ChemiDoc System.

Primary antibodies used: puromycin (Millipore, MABE343), p38 (CST, 9212), p-p38 (CST, 9211), p-eIF2α (CST, 9721), p-JNK (CST,4668), GAPDH (CST, 2118), β-actin (CST, 3700), ZAKα (Bethyl Laboratory, A301–993A-T), ZAK (Proteintech, 14945–1-AP), RPS10 (LSBio, LS-C335612), ABCE1 (ABclonal, A9135). Secondary antibodies: goat anti-rabbit and goat anti-mouse IgG HRP-conjugated (Cytiva, #NA934 and #NA931). The densitometry of the immunoblots was quantified using Fiji.

Measurement of soluble poly-GR in iPSC-derived neurons

The soluble poly-GR levels in iPSC-derived neurons were assessed using the Meso Scale Discovery (MSD) Immunoassay as before (91). Briefly, cells were lysed using Tris-based lysis buffer, and lysates were normalized to equal concentrations before being loaded into duplicate wells. MSD assays were conducted in a blinded manner.

Polysome profiling and runoff assay

For the runoff experiment, the cells were treated with 2 μg/ml harringtonine (Abcam, 141941) for 0, 2 or 5min, followed by immediate addition of 100 μg/ml cycloheximide (Sigma, C1988) for 10 min. Cells were directly lysed in lysis buffer [20 mM HEPES (pH 8), 150 mM KCl, 5 mM MgCl2, 1% Triton X-100, 1mM DTT, phosphatase inhibitor cocktail (Thermo Scientific, PI88667), EDTA-free Protease inhibitor cocktail (Thermo Scientific, A32955)]. Lysates containing 100 μg total RNA were run through 10–50% sucrose gradients using Beckmann Coulter SW41 Ti rotor at 40,000 rpm at 4 °C for 2 hours. Gradients were fractionated using a Biocamp piston gradient fractionator. The absorbance at 260 nm was recorded. Equal amounts of sample from each fraction was used for luciferase detection using Nano-Glo HiBiT Lytic Detection kit (Promega, N3040). Equal volumes from each fraction were also trichloroacetic acid (TCA)–precipitated and subjected to SDS-PAGE, immunoblotting, and Nano-Glo HiBiT blotting detection (Promega, PRN2410). Quantification of the polysome/80S ratio was done by measuring the area under the curve of 80S and polysome peaks in individual sample.

RNA-seq coupled with the ribosome runoff profiling

In i³Neurons, polysome profiling follows the same procedure as with HeLa, except the runoff treatment with harringtonine is 0 and 3 min. RNA from total lysate was quantified by Qubit HS RNA kit (Thermo Fisher Scientific, Q32852) and 10% was saved to extract input RNA by TRIzol (Ambion, 15596018). The remaining lysates (around 20 μg total RNA) were run through 10–50% sucrose gradient. Fractions were collected and combined into medium (3–7 ribosomes) and heavy (>7 ribosomes) fractions. RNA was extracted by TRIzol LS reagent (Ambion, 10296028). Purified RNA was quantified by Quant-iT RiboGreen RNA kit (ThermoFisher Scientific, R11490) and 200 ng from each sample was used for poly-A selection and RNA-seq library preparation following the manufacturer’s instruction (NEB, E7765). The libraries were quantified with Qubit dsDNA HS kit (ThermoFisher Scientific, Q32851) and examined by Fragment analyzer. Libraries were pooled and sequenced for 150bp paired-end reads. The raw reads from fastq files were first processed by default quality-filter using Trimmomatic (92). Cleaned reads were mapped to human GRCh38/hg38 and quantified by RSEM(93). Estimated gene counts (>5 in all samples) were used for subsequent analysis. ANOVA was used to identify the list of genes exhibiting statistically significant variability across condition (F test and Bonferroni correction, padj<0.05). Robust K-means clustering was performed with R package dicer (94). GO enrichment analysis was done by Enrichr (95). Mapped reads in bigwig format were visualized on IGV. Motif counting and transcript feature analysis was done by in-house scripts (https://github.com/nicolearnstocode/RunoffSeq).

Proteomic data analysis

Proteomic data were analyzed following the previous pipeline (96). Briefly, the data matrix was obtained from the Answer ALS database. The iPSN lines used in the analysis are listed in table S2. Proteins with missing values in over four samples were removed. Protein levels were normalized to the AE8 iPSN batch control line within each batch. Only genes with both proteomic and transcriptomic data were retained. The proteomic data matrix was divided by the normalized read count from the transcriptomic matrix for each gene and individual line. The candidate list was defined by the Wilcoxon test, p<0.05. Z-scores calculated from proteomic data and transcriptomic data for each gene comparing control and C9 patient samples were used for the heatmap.

RNA extraction and RT-qPCR

RNA was extracted from iPSN cells with TRIzol. The cDNA was synthesized using High-Capacity cDNA Reverse transcription kit (ThermoFisher Scientific, 4368813). qPCR of target genes was performed using PowerUp SYBR Green master mix (ThermoFisher Scientific, A25776). All RT-qPCR reactions were performed with at least three biological replicates for each group and two technical replicates on the CFX96 real-time PCR detection system (Bio-Rad). Expression values of each gene were normalized to RPLP0 mRNA as the internal control. Inter-group differences were assessed by two-tailed Student’s t-test. For RT-qPCR validation of the hits from ribosome runoff assay, equal amount of CLuc RNA was added to each sample prior to RNA extraction, and the target RNA levels were normalized to CLuc mRNA. All the primer sequences are listed in table S3.

Single molecule imaging of ribosome runoff in live cells

U-2 OS cells stably expressing AUG-SunTag-NLuc reporter (88) were seeded in 35 mm cover glass (Cellvis, D35–20-1.5-N) and grown overnight. 500 μM 3-Indoleacetic acid (Sigma, I2886) was added into the culture medium the night before and throughout the experiment. Cells were incubated with 100 nM Halo dye (JF646) (97) for 30 min and washed to remove excess dyes, and the medium was changed to Leibovitz’s L-15 media (Gibco, 21083–027) supplemented with 10% FBS before imaging. Tokai Hit stage top incubator was used to control the humidity and keep the sample at 37 °C during imaging session. Cells were treated with 2 μM GR₂₀ peptide for 30 min, and the translation sites were identified during this time. Half of the warm imaging media was mixed with harringtonine (Cayman Chemical, #15361) and added back to the imaging dish to a final concentration of 3μg/mL. Live imaging was started 1 min exactly after adding drug. A snapshot of the RNA and protein channels were acquired every 10 sec for a total of 30 min.

The live cell imaging analysis followed the previous protocol (88). Single particles of RNA or translation sites were identified with Airlocalize (98) and tracked with u-track (99). In the ribosome runoff experiment, we measured the time it took for each translating mRNA to completely lose its translation intensity starting from harringtonine addition. In practice, runoff occurs when the intensity of translation signal fell below 10% of its maximum intensity. We did not include the molecules if translation site (TLS) and the corresponding mRNA signal disappeared together, since these events may be due to the loss of the mRNA. We also discarded the TLS with low initial protein signal in the first four frames (below 10% of maximum intensity), as they were not translating at the beginning of the assay. GraphPad Prism was used to plot the results of run-off assays as Kaplan-Meier curves. We used Greenwood’s formula to calculate the 95% confidence bounds. Log-rank Mantel-Cox test was used to examine the statistical significance between the survival curves.

Statistical analysis

We used GraphPad Prism 8 software and R (4.1.1) for the statistical analyses. Shapiro-Wilk test was used to check normality. For RNA-seq coupled with ribosome runoff profiling analysis, ANOVA was used to identify the list of genes exhibiting statistically significant variability across condition (F test and Bonferroni correction, p_adj<0.05). For analyzing proteomics data from Answer ALS, we used Wilcoxon test. For live cell imaging analysis, Log-rank Mantel-Cox test was used to examine the statistical significance between the survival curves. For Western blot quantification, at least three independent experiments were done. For comparing multiple groups, ANOVA with Bonferroni’s multiple comparison or Tukey’s post hoc test was used with normal distributed data, Kruskal-Wallis test was performed for not normally distributed data. For two groups comparison, Mann-Whitney test was performed with data that is not normal distributed, and t test was used for normally distributed data.