Comparative CRISPRi screens reveal a human stem cell dependence on mRNA translation-coupled quality control

Abstract

The translation of mRNA into proteins in multicellular organisms needs to be carefully tuned to changing proteome demands in development and differentiation, while defects in translation often have a disproportionate impact in distinct cell types. Here we used inducible CRISPR interference screens to compare the essentiality of genes with functions in mRNA translation in human induced pluripotent stem cells (hiPS cells) and hiPS cell-derived neural and cardiac cells. We find that core components of the mRNA translation machinery are broadly essential but the consequences of perturbing translation-coupled quality control factors are cell type dependent. Human stem cells critically depend on pathways that detect and rescue slow or stalled ribosomes and on the E3 ligase ZNF598 to resolve a distinct type of ribosome collision at translation start sites on endogenous mRNAs with highly efficient initiation. Our findings underscore the importance of cell identity for deciphering the molecular mechanisms of translational control in metazoans.

Fig. 1. Comparative inducible CRISPRi screens identify essential components of the mRNA translation machinery in human cells.

a, Schematic of inducible CRISPRi cell line generation and screening. Right, gene counts for major functional groups of the human translation machinery included in the sgRNA library. Neo, neomycin resistance gene; mCh, mCherry; Puro, puromycin resistance gene; KRAB, Krüppel associated box; rtTA, reverse tetracycline-controlled transactivator; mKate, monomeric Kate2 protein. b, Schematic of the workflow of inducible hiPS cell differentiation into NPCs, neurons and CMs and the different types and durations of inducible CRISPRi screens. c, Principal component analysis of variance stabilizing-transformed count data for sgRNAs from DESeq2 (n = 2 biological replicates for CM screens; n = 3 biological replicates for inducible hiPS cell, inducible HEK293 cell, NPC and neuron screens). PC, principal component; diff, differentiation; sur, survival. d, Volcano plots of gene-level sgRNA log₂ fold change (mean of the top three sgRNAs per gene by magnitude) and P values (from comparisons of all sgRNAs targeting a given gene to all negative control sgRNAs) for each screening condition relative to matched uninduced (−Dox) controls. Genes with significant (two-sided Mann–Whitney test, P ≤ 0.1) positive or negative enrichment scores are shown in green and purple, respectively. e,f, UpSet plots showing overlap of genes with significant (two-sided Mann–Whitney test, P ≤ 0.1) negative enrichment scores in inducible hiPS cell, inducible HEK293 cell and NPC screens in comparison to common essential genes in cancer cell lines (DepMap) and genome-wide CRISPRi in the WTC11 hiPS cell line (e) and neuron differentiation or survival screens compared to a genome-wide CRISPRi screen in WTC11-derived i³Neurons (f). hiPSCi, inducible hiPS cell; HEK293i, inducible HEK293 cell. Source data

Fig. 2. Common and cell-type-specific effects of mRNA translation perturbations in human cells.

a, Violin plots of gene-level sgRNA log₂ fold changes (mean of the top three sgRNAs per gene by magnitude) in each screen separated into functional gene groups (r-proteins, translation factors, mRNA-associated proteins and ribosome-associated proteins). The number of genes with significant (two-sided Mann–Whitney test, P ≤ 0.1) enrichment or depletion in each screen is indicated above each violin. FC, fold change. b, Heatmaps of data in a. Significant (two-sided Mann–Whitney test, P ≤ 0.1) gene-level enrichment or depletion in each screen indicated by color. White, not significant in the respective screen. Data for genes with more than one annotated TSS (S1 and S2) were analyzed separately. Open and closed circles indicate nonessential and essential genes, respectively, in the common essential gene set from DepMap 23Q4 (ref. ) and genome-wide CRISPRi screens in WTC11 hiPS cells and H1 hES cells; x indicates the absence of data for alternative TSSs in DepMap. hESC, human embryonic stem cells; PIC, preinitiation complex; tRNA-iMet, initiator tRNA methionine; SKI, superkiller. Source data

Fig. 3. Differential resilience to perturbed ribosome rescue is not because of functional redundancy.

a, Schematic of expression constructs and workflows for growth competition and stalling reporter readthrough assays. b, Growth assays of cells transduced with the most potent sgRNA targeting ZNF598, ASCC3, HBS1L or PELO in the inducible hiPS cell screen (n = 2 biological replicates; line, mean). The percentage of GFP-positive (GFP⁺) cells was measured by flow cytometry (>10,000 cells per analysis) every four population doublings and normalized to GFP⁺ cell numbers in matched uninduced (−Dox) controls. GTP, guanosine triphosphate; GDP, guanosine diphosphate. c–e, Stalling readthrough of reporters containing (AAA)₂₀ (c), the XBP1 arrest peptide (d) and a no-stall control (e). The median fluorescence intensity for BFP and mOrange was quantified by flow cytometry (>20,000 cells per analysis) in the indicated cell lines transduced with the most potent sgRNA targeting ZNF598, ASCC3, HBS1L or PELO based on the inducible hiPS cell screen. The ratio of mOrange to BFP in knockdown cells was normalized to that for the same reporter in cells transduced with a nontargeting sgRNA (sgControl; n = 3 biological replicates; P values from an unpaired two-tailed t-test). UbC, ubiquitin C promoter; FLAG, Flag-tag. Source data

Fig. 4. Ribosome rescue prevents a cytotoxic ISR in human stem cells.

a, Global protein synthesis measurements in knockdown inducible hiPS cells or inducible HEK293 cells by HPG labeling for 30 min. Median fluorescence intensity was quantified by flow cytometry (>10,000 cells per analysis) and normalized to values from cells transduced with sgControl (n = 5 biological replicates; P values from an unpaired two-tailed t-test). b, LDH measurements in culture supernatants from knockdown inducible hiPS cells or inducible HEK293 cells. Values were normalized to supernatant from cells transduced with sgControl (n = 4 technical replicates and 4 biological replicates; P values from an unpaired two-sided t-test). c, Heatmap of mRNAs differentially expressed upon ZNF598 or HBS1L repression in inducible hiPS cells or inducible HEK293 cells (n = 2 biological replicates; two-sided Wald test with Benjamini–Hochberg correction, adjusted P ≤ 0.05; n = 3165). d, Heatmap of a data subset from c showing genes upregulated downstream of ATF4 within the ISR or downstream of ATF6 or IRE1α upon ER protein folding perturbations. e, Growth assays of inducible hiPS cells expressing an sgRNA targeting ZNF598 (day 6) or HBS1L (day 8) in the absence (−) or presence of GCN2i (A-92, 1.25 µM), PERKi (GSK2606414, 4 nM), p38i (SB203580, 1 µM) or ISRIB (50 nM) in comparison to uninduced controls (n = 3 biological replicates; P values from an unpaired two-tailed t-test). f, Growth assays of inducible hiPS cells expressing an sgRNA targeting ZNF598, HBS1L, GCN2 and PERK alone or in combination in comparison to uninduced controls (n = 5 biological replicates; P values from an unpaired two-tailed t-test). g, Immunoblot analysis of eIF2α phosphorylation, p38 phosphorylation and total eIF2α and p38 levels in inducible hiPS cells and inducible HEK293 cells (n = 4 biological replicates). Inducible hiPS cells treated with 2.5 µM tunicamycin (TM) for 2 h served as a positive control for induction of eIF2α phosphorylation; inducible hiPS cells treated with 0.05 mg L⁻¹ ANS for 15 min served as a positive control for induction of p38 phosphorylation. Rep, replicate. h, Quantification of signal intensity in g by densitometry (P values from an unpaired two-tailed t-test). Source data

Fig. 5. Defective ubiquitination by ZNF598 in human stem cells elicits ribosome pausing at start sites.

a, Schematic model of the consequences of ZNF598 depletion or ZNF598^RING expression in inducible hiPS cells. b, Polysome profiles (top) and immunoblot analysis of ZNF598 and uS5 in polysome gradient fractions (bottom) of inducible hiPS cells expressing sgControl, a ZNF598 sgRNA or ZNF598^RING (n = 1 biological replicate). c, Metagene profiles of ribosomal A-site occupancy from monosome footprints around CDS start and stop sites (n = 2 biological replicates). d, Volcano plot of differential ribosome pause sites upon ZNF598^RING expression in inducible hiPS cells (two-tailed Fisher’s exact test with Benjamini–Hochberg correction, adjusted P ≤ 0.01). e, Nucleotide (top) and amino acid (bottom) motif analysis of significantly increased pause sites in well-translated mRNAs (>0.5 footprints per codon in all samples; n = 3,421) in ZNF598^RING-expressing inducible hiPS cells. f, Volcano plot of differential ribosome pausing analysis upon ZNF598 knockdown (sgZNF598) in inducible hiPS cells as in d. g, Nucleotide (top) and amino acid (bottom) motif analysis of significantly increased pause sites in well-translated mRNAs (>0.5 footprints per codon in all samples; n = 2,463) in sgZNF598 inducible hiPS cells. h,i, GO term enrichment analysis of genes with significantly increased pause sites (one-tailed Fisher’s exact test with Benjamini–Hochberg correction, adjusted P ≤ 0.01) within the first five codons in ZNF598^RING-expressing inducible hiPS cells (h) and throughout the ORF in sgZNF598 inducible hiPS cells (i) filtered for TPM > 1 in RNA-seq from inducible hiPS cells. j, Distribution of monosome footprints (in reads per million (rpm)) along the H1-5 (left) and H3C2 (right) mRNA in control and ZNF598^RING-expressing inducible hiPS cells (n = 2 biological replicates). Significant differential pauses are indicated with red arrows. k, Representative histograms of cell-cycle analysis in inducible hiPS cells by DNA staining with EdU followed by flow cytometry. l, Changes in the fraction of cells in different cell-cycle phases calculated by flow cytometry analysis after EdU staining (n = 3 biological replicates; >10,000 cells per analysis; P values from an unpaired two-tailed t-test). Source data

Fig. 6. ZNF598 detects ribosome collisions during translation initiation.

a, Comparison of 5′ UTR length, CDS length and transcript abundance (by TPM in RNA-seq) in mRNAs with significantly increased start site pauses (in the first five codons of the respective CDS) in ZNF598^RING-expressing inducible hiPS cells (filtered for TPM > 1; n = 702), other well-translated mRNAs included in the pause site analysis (>0.5 footprints per codon, filtered for TPM > 1; n = 2,678) and remaining mRNAs with detectable expression in inducible hiPS cells (TPM > 1 in RNA-seq; n = 10,917). P values from a two-sided Wilcoxon test; NS, not significant (P > 0.01). Box plots: center line, median; box limits, upper and lower quartiles; whiskers, 1.5× the interquartile range. b, Representative density heatmaps of 42–68-nt footprints according to length and 5′ end position around CDS start (left) and stop (right) codons in control, ZNF598 knockdown (sgZNF598) and ZNF598^RING-expressing inducible hiPS cells (top to bottom) in one of two biological replicates. c, Polysome profiling (top) and immunoblot analysis of sucrose gradient fractions (bottom) from untreated inducible hiPS cells and after treatment with 0.05 mg L⁻¹ ANS for 15 min or a short (2.5 min) or long (2 hours) treatment with 2 µg ml⁻¹ HAR (n = 1 biological replicate from two independent experiments with similar results). d, Metagene profiles of ribosomal A sites from monosome footprints around CDS start and stop codons after treatment with 5 µM 4EGI-1 (n = 2 biological replicates). e, Global protein synthesis measurements by OPP labeling in control or ZNF598^RING-expressing inducible hiPS cells before and after treatment with 5 µM 4EGI-1. Median fluorescence intensity quantified by flow cytometry (>10,000 cells per analysis) was normalized to values from untreated inducible hiPS cells (n = 4 biological replicates; P values from an unpaired two-tailed t-test). Source data

Fig. 7. Model for ZNF598-mediated surveillance of translation initiation.

ZNF598 prevents a cytotoxic ISR in human stem cells triggered by start site collisions.

Extended Data Fig. 1. Workflow for comparative genetic screens by inducible CRISPRi.

a, LFQ intensities of KRAB-dCas9 and mCherry protein levels in hiPSCi +/- doxycycline (Dox) measured by mass spectrometry (n = 3 biological replicates, top) and immunoblot analysis of HA-tagged KRAB-dCas9 in control hiPSCi and after treatment with 2 µM doxycycline for 2 days (bottom). b, Immunostaining for cell type-specific markers (green) and DAPI (blue). Scale bar: 10 µm. c, Quantification of % mCherry-positive cells by flow cytometry (n = 3 biological replicates) after KRAB-dCas9 induction with 2 µM doxycycline for two days (in hiPSCi and HEK293i), 5 passages (NPC), and for five days in differentiated neurons and CM. d, Quantification of cell doubling time (n = 2 biological replicates for HEK293i; n = 3 biological replicates for hiPSCi and NPC). e, Heatmaps of significant gene-level sgRNA log₂ fold change (FC) (two-sided Mann–Whitney test P ≤ 0.1) for cell identity markers. diff = differentiation, sur = survival, TSS = transcriptional start site. f, UpSet plot of overlap among genes with significant negative gene-level sgRNA log₂ FC (two-sided MannWhitney test P ≤ 0.1). g,h, Global protein synthesis measurements by O-propargyl-puromycin (OPP) labeling. (g) Median fluorescence intensity quantified by flow cytometry ( > 10,000 cells/ analysis); (h) Data from (e) normalized to a matched control not treated with OPP (-OPP + A647) and calculated as a fraction of the average signal in hiPSCi (n = 2 biological replicates for +OPP + A647; n = 1 biological replicate for -OPP + A647 and +OPP-A647). Source data

Extended Data Fig. 2. Validation of pooled CRISPRi screens.

a, Heatmaps of gene-level sgRNA log₂ FC for 16 genes selected for validation; white: not significant (two-sided Mann–Whitney test P > 0.1). b, Schematic of individual sgRNA expression construct. Puro: puromycin resistance gene. c, Correlation of single sgRNA phenotype from growth competition assays [log₂ (-/+Dox)] and log₂ FC of the same sgRNA in the pooled CRISPRi screen (Spearman’s R, P < 0.05) for the two most active sgRNAs (sgRNA1 and sgRNA2) for each gene in the screens. Solid lines denote linear regression model; grey shading denotes 95% confidence interval. d, Knockdown efficiency of genes from (a) compared to a non-targeting control (sgControl) measured by quantitative RT-PCR for the most active sgRNA from the screens (sgRNA1; n = 2 biological replicates, each with n = 3 technical replicates). e, Correlation plot of cell type specificity (hiPSCi log₂ FC minus HEK293i log₂ FC) calculated for individual sgRNAs from growth competition assays [∆log₂ FC (validation), n = 2 biological replicates] and the same sgRNAs in the pooled CRISPRi screen [∆log₂ FC (screen); n = 3 biological replicates]. Solid line denotes linear regression model; grey shading denotes 95% confidence interval. f, Heatmaps of significant (pairwise two-tailed t-test, FDR < 0.01) protein-level log₂ FC for screening targets in (a) measured by mass spectrometry in the indicated cell types in comparison to hiPSCi (n = 3 biological replicates; numbers in brackets indicate isoforms). g, Immunoblot analysis of ZNF598, ASCC3, HBS1L, and PELO in sgRNA-transduced cells after KRAB-dCas9 induction ( + Dox) and in matched uninduced controls (-Dox). Source data

Extended Data Fig. 3. mRNA expression levels for screen targets in hiPSC and hiPSC-derived cells.

Gene expression heatmaps for screen targets in parental kucg-2 hiPSC and NPC, neurons and CM cultures differentiated from them (n = 2 biological replicates). Standardized Z scores were calculated from DESeq2-normalized RNA-Seq gene counts across samples (data from ref. ). Source data

Extended Data Fig. 4. Validation experiments with individual sgRNAs.

a,b, Growth assays of HEK293i or hiPSCi expressing a sgRNA targeting N4BP2 (a) or EDF1 (b) (n = 2 biological replicates; line: average) and knockdown efficiency measurements of target genes compared to a non-targeting control (sgControl) by quantitative RT-PCR (n = 2 biological replicates, each with n = 3 technical replicates). c, Growth assays of cells transduced with the second most potent sgRNA (sgRNA2) targeting ZNF598, ASCC3, HBS1L, or PELO in the hiPSCi screen (n = 3 biological replicates; line: average). The percentage of GFP-positive (GFP + ) cells was measured by flow cytometry ( > 10,000 cells/ analysis) every three population doublings and normalized to GFP+ cell numbers in matched uninduced (-Dox) controls. d, Growth assays of hiPSCi transduced with a sgRNA targeting HBS1L alone or in combination with a cDNA encoding wild-type or a GTPase-inactive HBS1L (H348A) (n = 3 biological replicates, P-values from unpaired two-tailed t-test). e, Heatmaps of significant (pairwise two-tailed t-test, FDR < 0.01) protein-level log₂ FC for SKI complex components in the indicated cell types in comparison to hiPSCi measured by mass spectrometry (n = 3 biological replicates). Source data

Extended Data Fig. 5. Analysis of stress pathway activation in different cellular contexts.

a, Growth assays of cells transduced with a sgRNA targeting EIF2S1 in HEK293i, hiPSCi, or NPC cells (n = 2 biological replicates; line: average) and knockdown efficiency measurements of target genes compared to a non-targeting control (sgControl) by quantitative RT-PCR (n = 2 biological replicates, each with n = 3 technical replicates). b, Venn diagrams of overlap between genes upregulated or downregulated in HEK293i or hiPSCi depleted for ZNF598 or HBS1L by CRISPRi (p-values from one-tailed Fisher’s exact test). c,d, Gene ontology (GO) enrichment in differentially expressed mRNAs using the “Biological Process” function in ClusterProfiler (one-tailed Fisher’s exact test with Benjamini-Hochberg correction, P-adj ≤ 0.01) in knockdown hiPSCi (b) or HEK293i (c). e, f, Flow cytometry analysis of eIF2α phosphorylation (d) and p38 phosphorylation (e) upon depletion of ZNF598 or HBS1L in hiPSCi in comparison to cells transduced with a non-targeting sgRNA (sgControl; n = 3 biological replicates; matched samples). hiPSC treated with 2.5 µg tunicamycin (TM) for 2 hours to induce the ISR or with 0.05 mg/l anisomycin (ANS) for 15 minutes to induce the RSR served as a positive control. g, Growth assays of hiPSCi expressing a non-targeting sgRNA (sgControl, day 6) in the absence (-) or presence of inhibitors of GCN2 (GCN2i, A-92, 1.25 µM), PERK (PERKi, GSK2606414, 4 nM), p38 (p38i, SB203580, 1 µM), or the ISR (ISRIB, 50 nM) in comparison to uninduced controls (n = 3 biological replicates). Source data

Extended Data Fig. 6. ZNF598^RING expression in human stem cells induces cytotoxicity but not through the ISR.

a, Immunoblot analysis of wild-type hiPSCi and hiPSC transduced with cDNA encoding HA-tagged wild-type (ZNF598^WT) or C29S/C32S (ZNF598^RING) ZNF598 (n = 2 biological replicates). b, Stalling readthrough of reporters containing an AAA-encoded stretch of twenty lysines (AAA₂₀) in control hiPSCi and hiPSCi transduced with ZNF598^WT or ZNF598^RING cDNA. The median fluorescence intensity for BFP and mOrange was quantified by flow cytometry ( > 20,000 cells/ analysis) and normalized to average values for the hiPSCi control (n = 3 biological replicates; P-values from unpaired two-tailed t-test). c, Global protein synthesis measurements by OP-Puromycin and AF647 detection. Median fluorescence intensity was quantified by flow cytometry ( > 10,000 cells/ analysis) and normalized to average values for the control (n = 6 biological replicates; p-values from unpaired two-tailed t-test). d, Lactate dehydrogenase (LDH) measurements in culture supernatants from wild type or ZNF598^RING-expressing hiPSCi two days after transduction. Values were normalized to supernatant from wild type hiPSCi (n = 4 technical replicates and 5 biological replicates; P-values from unpaired two-tailed t-test). e,f, Differential gene expression analysis in hiPSCi upon ZNF598 knockdown (sgZNF598) in comparison to a non-targeting sgRNA (sgControl), or in ZNF598^RING-expressing hiPSCi in comparison to untransduced hiPSCi (n = 2 biological replicates). (e) Heatmap of mRNAs differentially expressed in at least one context (two-sided Wald test; P-adj ≤ 0.05; n = 2289). (f) Heatmap of a subset from (e) showing genes associated with the ISR. g, Flow cytometry analysis of eIF2α (top) or p38 (bottom) phosphorylation upon ZNF598^RING expression in hiPSCi in comparison to cells transduced with a non-targeting sgRNA (sgControl; n = 2 biological replicates) on day 1-3 after transduction. h, Gene ontology (GO) enrichment in differentially expressed mRNAs from (e) using the “Biological Process” function in ClusterProfiler (one-tailed Fisher’s exact test with Benjamini-Hochberg correction, P-adj ≤ 0.01). Source data

Extended Data Fig. 7. ZNF598^RING increases ribosome occupancy at translation start sites in human stem cells.

a,b, Amino acid motif analysis of internal pause sites (excluding the first and last five codons of each CDS) in wild type, ZNF598 knockdown (sgZNF598), and ZNF598^RING-expressing hiPSCi (a) or HEK293i (b). c, Metagene profiles of ribosomal A sites from monosome footprints around CDS start and stop codons in wild type and ZNF598^WT-overexpressing hiPSCi (n = 2 biological replicates). d, Metagene profiles of ribosomal A sites from monosome footprints around CDS start and stop codons in wild type, sgZNF598 and ZNF598^RING-expressing HEK293i (n = 2 biological replicates). e,f, Nucleotide (top) and amino acid (bottom) motif analysis of significantly increased pause sites within (e) or excluding (f) the first five codons in ZNF598^RING-expressing hiPSCi (two-tailed Fisher’s exact test with Benjamini-Hochberg correction, P-adj ≤ 0.05). g, Volcano plot of differential ribosome pause sites upon ZNF598^RING expression in HEK293i (two-tailed Fisher’s exact test with Benjamini-Hochberg correction, P-adj ≤ 0.05) (left). Nucleotide (top) and amino acid (bottom) motif analysis of significantly increased pause sites in well-translated mRNAs ( > 0.5 footprints/codon in all samples, n = 1438) in ZNF598^RING-expressing HEK293i (right). h, Volcano plot of differential ribosome pausing analysis upon ZNF598 knockdown (sgZNF598) in HEK293i as in (g, left). Nucleotide (top) and amino acid (bottom) motif analysis of significantly increased pause sites in well-translated mRNAs ( > 0.5 footprints/codon in all samples, n = 1003) in sgZNF598 HEK293 as in (g, right). i, Violin plots (center line: median) of TPM values from RNA-Seq for genes with significantly increased A-site pausing within the first 5 codons in ZNF598^RING-expressing hiPSCi (n = 704) in sgControl HEK293i and hiPSCi (P-value from two-sided Wilcoxon test). j, Hierarchically clustered heatmap of scaled Z scores of normalized unique transcript counts (TPM) from RNA-Seq in sgControl HEK293i and hiPSCi (n = 2 biological replicates) for histone mRNAs with increased A-site pausing within the first 5 codons in ZNF598^RING-expressing hiPSCi (n = 41). k, Changes in the fraction of cells in different cell cycle phases calculated by flow cytometry analysis after EdU staining (n = 3 biological replicates, >10,000 cells/ analysis; P-values from unpaired two-tailed t-test). Source data

Extended Data Fig. 8. Analysis of long ribosome footprints and cellular responses to start site stalling.

a, Venn diagram of overlaps between 5´ TOP motif mRNAs (n = 97) and transcripts with significantly increased A-site pauses in the first 5 codons in ZNF598^RING-expressing hiPSCi (n = 704), and comparison of 5´ UTR lengths and TPM levels from RNA-Seq in both groups (p-value from two-sided Wilcoxon test). Box plots: center line, median; box limits, upper and lower quartiles; whiskers, 1.5× interquartile range. b, Comparison of ribosome recruitment scores for 5´ UTRs quantified by direct analysis of ribosome targeting (DART) for histone mRNAs with start site pauses (n = 29), other mRNAs with start site pauses (n = 351), other well-translated mRNAs (n = 1340), and other detectable mRNAs (n = 3791). Box plots: center line, median; box limits, upper and lower quartiles; whiskers, 1.5× the interquartile range. c, Density heatmaps of 50-80 nt ribosome footprints according to length and 5’ end position around CDS start (left) and stop sites (right) for a second biological replicate. d, Density heatmaps of 50-80 nt ribosome footprints according to length and 3’ end position around CDS start (left) and stop sites (right) for a second biological replicate. e, Metagene profiles of 50-80 nt ribosome footprints around CDS start and stop sites in hiPSCi (n = 2 biological replicates). f, Polysome profiles and immunoblot analysis of uS10 and eS10 in sucrose gradient fractions from control HEK293i or after a short (2.5 min) treatment with 2 µg/ml homoharringtonine (HAR) (n = 1 biological replicate). g, Metagene profiles of 50-80 nt ribosome footprints around CDS start and stop sites in ZNF598-depleted hiPSCi treated with GCN2 inhibitor (GCN2i, A-92, 1.25 µM; n = 2 biological replicates). h, Polysome profiles (from Fig. 6c) and immunoblot analysis of uS3 and uS5 in sucrose gradient fractions from control hiPSCi, after short (2.5 min) or long (2 hours) treatment with 2 µg/ml homoharringtonine, or after treatment with 0.05 mg/l anisomycin (ANS) for 15 minutes (n = 1 biological replicate). Membranes from Fig. 6c were stripped and re-probed with antibodies against uS3 and uS5. i, Global protein synthesis measurements by OPP labeling in control hiPSCi and/or after treatment with 6.25, 12.5, and 25 µM 4EGI-1 for 1 day. Median fluorescence intensity was quantified by flow cytometry ( > 10,000 cells/ analysis) and normalized to the average value in controls (n = 2 biological replicates). Source data

Extended Data Fig. 9. Gating strategies for flow cytometry.

a, Representative gating strategy for growth assays (Figs. 3b, 4e, 4f and Extended Data Fig. 2c, 2e, 4a, 4b, 4c, 4d, 5a). b, Representative gating strategy for stalling reporter assays (Fig. 3c,d,e and Extended Data Fig. 6b). c, Representative gating strategy for OPP assays, phospho-staining, and mCherry expression assays (Figs. 4a, 6e and Extended Data Figs. 1c, 1g, 1h, 5e, 5f, 6g, 8i). d, Representative gating strategy for EdU assays (Figs. 5k, 5l and Extended Data Fig. 7k).