{"paper_id":"04faf686-1c3f-438b-b815-e06554e7dd4f","body_text":"Decoding\nRNA Metabolism by RNA-linked CRISPR Screening in Human Cells\nPatrick J. Nugent 1,2, Heungwon Park 1, Cynthia L. Wladyka 3, Katharine Y . Chen 1,2, Christine Bynum 1,4, Grace\nQuarterman1,4, Andrew C. Hsieh 3,5, Arvind Rasi Subramaniam 1,6,†\n1 Basic Sciences Division and Computational Biology Section of the Public Health Sciences Division, Fred Hutchin-\nson Cancer Center, Seattle WA, USA\n2 Molecular and Cellular Biology Graduate Program, University of Washington, Seattle WA, USA\n3 Human Biology Division, Fred Hutchinson Cancer Center, Seattle WA, USA\n4 Department of Biology, Spelman College, Atlanta GA, USA\n5 Department of Medicine and Department of Genome Sciences, University of Washington, Seattle WA, USA\n6 Department of Biochemistry and Department of Genome Sciences, University of Washington, Seattle WA, USA\n† Corresponding author\nAbstract\nRNAs undergo a complex choreography of metabolic processes in human cells that are regulated by thousands\nof RNA-associated proteins. While the effects of individual RNA-associated proteins on RNA metabolism have\nbeen extensively characterized, the full complement of regulators for most RNA metabolic events remain unknown.\nHere we present a massively parallel RNA-linked CRISPR (ReLiC) screening approach to measure the responses\nof diverse RNA metabolic events to knockout of 2,092 human genes encoding all known RNA-associated proteins.\nReLiC screens highlight modular interactions between gene networks regulating splicing, translation, and decay\nof mRNAs. When combined with biochemical fractionation of polysomes, ReLiC reveals striking pathway-specific\ncoupling between growth fitness and mRNA translation. Perturbing different components of the translation and pro-\nteostasis machineries have distinct effects on ribosome occupancy, while perturbing mRNA transcription leaves\nribosome occupancy largely intact. Isoform-selective ReLiC screens capture differential regulation of intron reten-\ntion and exon skipping by SF3b complex subunits. Chemogenomic screens using ReLiC decipher translational\nregulators upstream of mRNA decay and uncover a role for the ribosome collision sensor GCN1 during treat-\nment with the anti-leukemic drug homoharringtonine. Our work demonstrates ReLiC as a versatile platform for\ndiscovering and dissecting regulatory principles of human RNA metabolism.\nIntroduction\nRNAs are carriers of genetic information, scaffolds for\nprotein complexes, and regulators of gene expression\ninside cells. RNAs undergo several metabolic events\nsuch as splicing, editing, localization, translation, and\ndecay during their intracellular lifecycle. RNA metabolic\nevents are executed by ribonucleoprotein complexes\ncomposed of RNA-binding proteins (RBPs), adapter\nproteins, and regulatory factors. Over 2,000 human\ngenes encode proteins that are part of ribonucleopro-\ntein complexes 1,2. Individual RNA-associated proteins\noften regulate the metabolism of hundreds of RNAs.\nMutations in RNA-associated proteins are associated\nwith many human diseases including cancer, neurode-\ngeneration, and developmental disorders 3,4. Thus, de-\ncoding the effect of RBPs and associated factors on\nRNA metabolism is critical for our understanding of post-\ntranscriptional gene regulation and molecular mecha-\nnisms underlying human disease.\nDespite extensive biochemical studies of RNA\nmetabolism and RBP function, we do not know the\nfull set of cellular factors that regulate specific RNA\nmetabolic events. This is because binding of RBPs\ncan increase, decrease or leave unchanged metabolic\nevents on their target RNA depending on their affinity,\nlocation, and other associated factors 5–8. Many RBPs\nalso associate with multiple ribonucleoprotein com-\nplexes and participate in several distinct RNA metabolic\nevents9. Conversely, protein factors that do not directly\nbind RNA can still affect RNA metabolism by regulating\nthe interactions between RNAs and RBPs, or by con-\ntrolling the cellular level and activity of RBPs 10. Hence,\nbiochemical studies of RBP-RNA interactions are insuf-\nficient to reveal the full spectrum of functional regulators\nof RNA metabolic events in cells.\nUnbiased genetic screening can identify cellular factors\nregulating RNA metabolism, but are limited in their cur-\nrent form. CRISPR screens using indirect phenotypes\nsuch as cell growth and fluorescent protein levels are dif-\nficult to engineer and interpret for many RNA metabolic\nevents11,12 due to potential false positives 13,14 and ge-\nnetic compensatory mechanisms 15. CRISPR pertur-\nbations followed by arrayed bulk RNA sequencing or\npooled single cell RNA sequencing can directly report\non RNA phenotypes 16,17. But these transcriptome-wide\n1\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted July 26, 2024. ; https://doi.org/10.1101/2024.07.25.605204doi: bioRxiv preprint \n\nsequencing\napproaches have limited flexibility to study\ndifferent types of RNA metabolic events, are biased to-\nwards highly expressed RNAs, and are costly and la-\nbor intensive to scale beyond a few dozen perturbations.\nThus, it has not been possible until now to genetically\ndissect the RNA-centric functions of human proteins at\nthe scale of reporter-based CRISPR screens and with\nthe ability to directly monitor diverse RNA metabolic\nevents.\nResults\nDevelopment of RNA-linked CRISPR screening in\nhuman cells\nWe reasoned that combining CRISPR-based perturba-\ntions with barcoded RNA readouts will provide a general\napproach to study the genetic control of different events\nin human RNA metabolism. Supporting the feasibility\nof this barcoding approach, RNA interference screens\nin human cells 18 and CRISPR interference screens in\nS. cerevisiae 19,20 have used barcoding to link pertur-\nbations to transcriptional readouts. However, lentiviral\ndelivery, commonly used for CRISPR screening in hu-\nman cells, will scramble sgRNA-barcode linkages due\nto template switching during reverse transcription 21–23\nand result in variable expression of RNA barcodes due\nto random genomic integration 24–26. T o avoid these lim-\nitations, we employed an iterative, site-specific integra-\ntion strategy to stably express SpCas9 (Cas9 hereafter),\nsgRNAs, and barcoded RNA reporters from a defined\ngenomic locus (Fig. 1A). First, we generated a clonal\nHEK293T cell line with a single attP ‘landing pad’ site\nfor the Bxb1 integrase 27,28 at the AAVS1 safe harbor lo-\ncus by Cas9-mediated homology-directed repair. Next,\nwe integrated a doxycycline-inducible Cas9 and an or-\nthogonal attP* site29,30 into the landing pad using Bxb1-\nmediated recombination. Finally, we integrated sgRNA\nand reporter RNA cassettes into the attP* site using\nBxb1-mediated recombination. We used fluorescent\nand antibiotic selection markers to enrich for cells with\nsuccessful integration events at each step, and we used\ninsulator elements to reduce transcriptional interference\nand promote long-term stable expression of integrated\ngenes (Methods). Using an EYFP fluorescent reporter,\nwe confirmed its uniform and stable expression after in-\ntegration (Fig. 1B, blue). After doxycycline addition, we\nobserved a progressive decrease in EYFP signal over 7\ndays that was specific to cells co-expressing an EYFP-\ntargeting sgRNA (Fig. 1B, yellow), validating our ability\nto robustly induce Cas9-mediated gene knockout.\nT o identify regulators of RNA metabolism, we targeted\n2,092 human genes encoding proteins annotated to\ninteract with RNAs or RNA-binding proteins in previ-\nous manual curation and RNA interactome surveys 1,2\n(Fig. 1C). We selected sgRNAs from the validated\nBrunello library31 and used a dual sgRNA design to max-\nimize knockout efficiency. We cloned the sgRNA pairs\nalong with random N 20 barcodes into a modular attB*-\nintegrating vector that allows insertion of arbitrary RNA\nreporters (Fig. S1A). Our final library targeted 2,190\ngenes with 4 sgRNA pairs per gene, and included posi-\ntive control sgRNA pairs targeting essential genes and\nnon-targeting sgRNA pairs as negative controls (T able\nS1). We linked the N 20 barcodes to sgRNAs by paired-\nend deep sequencing of the cloned plasmid library. We\nthen integrated this library into our attP* parental cell\nline, and counted barcodes in the genomic DNA and\ntranscribed RNA by deep sequencing (Fig. 1C, Fig.\nS1B). We recovered a median of 8 barcodes per sgRNA\npair (henceforth referred to as sgRNAs) with at least one\nbarcode for 99% of sgRNAs and 100% of all genes (Fig.\nS1C), thus capturing the diversity of our input library.\nT o test whether sgRNA-linked barcodes capture fitness\neffects, we sequenced and counted barcodes in ge-\nnomic DNA and mRNA at different time points after\nCas9 induction (T able S6). Barcode counts showed lit-\ntle systematic change on day 5 after Cas9 induction\n(Fig. 1D, left panel). However, on days 13 and 21 af-\nter Cas9 induction, barcode counts for a subset of sgR-\nNAs were strongly depleted in both the genomic DNA\nand mRNA in a highly correlated manner (Fig. 1D, mid-\ndle and right panels). The magnitude of depletion was\ncorrelated across distinct barcode sets for each gene\n(Fig. S1D), indicating the assay’s technical reproducibil-\nity. Barcodes in the genomic DNA corresponding to an-\nnotated essential genes (n = 745) were depleted 5.4–\n6.6 fold (median depletion at days 13 and 21) relative to\nother genes targeted in our library (n = 1401, Fig. 1E).\nBarcodes in the mRNA corresponding to the same es-\nsential genes were depleted 13.6–28.4 fold (median de-\npletion at days 13 and 21) relative to other genes tar-\ngeted in our library. The greater effect of essential gene\nknockout on mRNA relative to DNA might arise from\nthe decreased RNA content in slower growing cells 32.\nIn summary, our RNA-linked screening strategy accu-\nrately captures both the identity and fitness effect of\nCRISPR perturbations solely from sequencing of bar-\ncodes in mRNA and genomic DNA.\nReLiC identifies regulators of mRNA translation\nWe first applied ReLiC to study translation, an RNA\nmetabolic step that is not directly accessible to exist-\ning CRISPR screening methods. The traditional gold\nstandard for monitoring translation is polysome profil-\ning — ultracentrifugation of cell lysate through a density\ngradient to separate mRNAs based on their ribosome\n2\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted July 26, 2024. ; https://doi.org/10.1101/2024.07.25.605204doi: bioRxiv preprint \n\noccupancy33–35.\nWe sought to combine this classic bio-\nchemical fractionation technique with ReLiC screening\nto identify regulators of ribosome occupancy on mRNAs.\nWe used a spliced β-globin reporter 36,37 as a prototypi-\ncal model of a well-translated mRNA (Fig. 2A). First, to\nestimate the precision of our barcode-based measure-\nment, we inserted 6 random barcodes into the 3′ UTR\nof this reporter and stably integrated the barcoded re-\nporter pool into our attP* parental cell line. We counted\nthe RNA barcodes in monosome (one ribosome), light\npolysome (2-4 ribosomes), and heavy polysome (>4 ri-\nbosomes) fractions by sequencing. Over 75% of the β-\nglobin mRNA was in the light and heavy polysome frac-\ntions while the relative amount of the six barcodes varied\nless than 3% within each fraction (Fig. 2B). Next, we\ncloned the β-globin reporter into our ReLiC-RBP plas-\nmid library, integrated the library into the attP* cell line,\nand induced Cas9 for 7 days before fractionating cell\nlysates (Fig. 2A). The duration of Cas9 induction was\nchosen to allow for sufficient protein depletion while pre-\nventing loss of essential gene knockouts. After count-\ning sgRNA-linked barcodes in each fraction, we used\nMAGeCK38 to identify sgRNAs that significantly altered\nthe ratio of barcodes between heavy (H) or light (L)\npolysomes and monosomes (M) (T able S7). T o call a\ngene as a ‘hit’, we required that at least 3 sgRNAs for\nthat gene had consistent positive or negative effects on\npolysome to monosome barcode ratios 1 and controlled\nthe resulting false discovery rate (FDR) at 0.05 (T able\nS8).\nPolysome to monosome ratios for individual sgRNAs\nwere highly reproducible (r=0.92 and 0.80 for H/M and\nL/M, respectively) between replicate experiments (Fig.\n2C). We identified 304 and 126 gene knockouts that\ndecreased heavy polysome to monosome and light\npolysome to monosome ratios, respectively. 37 gene\nknockouts increased heavy polysome to monosome ra-\ntio, while none increased light polysome to monosome\nratio (Fig. 2D). The skewed distribution of gene hits with\nmore perturbations leading to a decrease in ribosome\noccupancy likely results from the efficient translation of\nβ-globin mRNA in unperturbed cells (Fig. 2B). Con-\nsistent with heavy polysome fractions containing better-\ntranslated mRNAs, heavy polysome to monosome ra-\ntios were more sensitive to perturbations with more\ngene hits and larger effect sizes than light polysome to\nmonosome ratios (Fig. 2D). We therefore focused on\nheavy polysome to monosome ratios for further analy-\nses.\nGene hits that decreased polysome to monosome ratios\nwere highly enriched for cytoplasmic ribosomal proteins\nand ribosome biogenesis factors (Fig. 2D, Fig. S2A).\nIndeed, 44 of the 54 large ribosomal (RPL) proteins and\n28 of the 36 small ribosomal (RPS) proteins were clas-\nsified as hits by MAGeCK (closed circles, Fig. 2E). As a\ngroup, knockout of large ribosomal proteins decreased\npolysome to monosome ratios more than knockout of\nsmall ribosomal proteins (Fig. 2E, median log 2 H/M: -\n3.17 vs -1.48 for RPL vs RPS). Similarly, knockout of\nlarge ribosomal subunit biogenesis factors decreased\npolysome to monosome ratios more than knockout of\nsmall ribosomal subunit biogenesis factors (Fig. 2E, me-\ndian log 2 H/M -1.82 vs -0.55 for large vs small subunit\nbiogenesis factors), though their overall effects were\nsmaller than knockout of corresponding ribosomal pro-\nteins.\nTranslation initiation factors were also enriched among\ngene hits that decreased heavy polysome to monosome\nratio (Fig. S2A), but their effects were more variable\n(Fig. 2F) and generally smaller than the effect of ribo-\nsomal protein depletion. Subunits of the EIF2, EIF2B,\nEIF3, and EIF4F initiation complexes all emerged as\ngene hits (closed circles, Fig. 2F). Some of the initiation\nfactor subunits that we did not classify as hits (open cir-\ncles, Fig. 2F) still had multiple sgRNAs that decreased\nheavy polysome to monosome ratio but fell just below\nour gene-level FDR threshold (EIF4G1 – FDR: 0.08) or\ndid not meet our stringent criterion of 3 distinct sgR-\nNAs with significant effects (EIF2S2, EIF4E – 2 sgR-\nNAs). In the case of the 12-subunit EIF3 and associated\nEIF3J, the seven subunits A,B,C,D,E,G,and I that we\ncalled as hits were the same ones that severely reduced\npolysome to monosome ratio and fitness when depleted\nby siRNA in HeLa cells 39. Aminoacyl-tRNA synthetase\nknockouts had mild and variable effects on ribosomal\noccupancy (Fig. 2F), presumably reflecting a balance\nbetween their direct effect on translation elongation and\nindirect effect on translation initiation through GCN2 and\nEIF2α phosphorylation 40–42.\nWe identified several gene knockouts outside the core\ntranslation machinery with decreased polysome to\nmonosome ratio (Fig. 2F). Four subunits of the CCR4-\nNOT complex (CNOT1, CNOT2, CNOT3, and CNOT7),\nwhich has been implicated in a wide range of RNA\nmetabolic processes 43, emerged as hits in our screen,\nwhich agrees with observations in S.cerevisiae strains\nlacking CNOT2 and CNOT3 homologs 44. Knockout of\nseveral subunits of the proteasome and the TRiC chap-\neronin complex led to substantially reduced polysome to\nmonosome ratios, comparable in magnitude to knock-\n1W\ne refer to ‘polysome to monosome ratio of barcode counts’ as ‘polysome to monosome ratio’ for brevity, but we emphasize the\ndistinction from the polysome to monosome ratio of A 260 as typically used in the polysome profiling literature.\n3\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted July 26, 2024. ; https://doi.org/10.1101/2024.07.25.605204doi: bioRxiv preprint \n\nout\nof core translation initiation factors (Fig. 2F). No-\ntably, these complexes did not arise as hits simply be-\ncause of their essentiality since knockout of other es-\nsential cellular complexes such as RNA polymerase II\nand splicing factor 3a/b (SF3) did not reduce polysome\nto monosome ratio (Fig. 2F). While neither the protea-\nsome nor the TRiC chaperonin complex has been di-\nrectly associated with translational regulation, they play\na critical role in maintaining cellular proteostasis by coor-\ndinating their activities with translational output45,46. Our\nresults reveal a reciprocal regulation of translation in re-\nsponse to changes in proteasomal and chaperonin ca-\npacity.\nPathway- and mechanism-specific effects of gene\nknockouts on ribosome occupancy\nRibosome occupancy on mRNAs is often correlated\nwith cellular growth rate, with slower growth accompa-\nnied by lower polysome to monosome ratio across differ-\nent growth conditions and organisms 39,47,48. Our mea-\nsurements of both barcode depletion after Cas9 induc-\ntion and polysome distribution of barcodes allowed test-\ning the generality of the relationship between ribosome\noccupancy and growth across ~2,000 different gene\nperturbations. Across all perturbations, decrease in\npolysome to monosome ratio was positively correlated\nwith barcode depletion in both mRNA and genomic DNA\nbut had a wide distribution (Fig. S2B). We then focused\non gene knockouts for ribosomal proteins, ribosome bio-\ngenesis factors, EIF3 subunits, proteasome, and RNA\npolymerase since these groups have several essential\ngenes with varying growth effects. Within each group,\ngene knockouts with lower polysome to monosome ra-\ntio also showed depleted mRNA and genomic DNA (Fig.\n2G, Fig. S2C). However, each gene group had charac-\nteristically distinct relationship between ribosome occu-\npancy and growth fitness as measured by barcode de-\npletion. Perturbing large ribosomal proteins and biogen-\nesis factors resulted in the largest decrease in polysome\nto monosome ratio relative to fitness, which was fol-\nlowed by perturbations of small ribosomal proteins and\nbiogenesis factors, and then EIF3 (Fig. 2G, Fig. S2C).\nPerturbing proteasomal subunits produced a smaller but\nstill significant decrease in ribosome occupancy, while\nperturbing RNA polymerase II subunits did not alter ri-\nbosome occupancy despite their significant effects on\ngrowth fitness (Fig. 2G, Fig. S2C). Hence, the cou-\npling between growth rate and ribosome occupancy in\nhuman cells is not invariant across all perturbations, but\ndepends on the pathway or the molecular process that\nis limiting growth.\nWe next examined the small group of gene knockouts\nthat increased the heavy polysome to monosome ra-\ntio in our screen (Fig. 2H, brown triangles, T able S8).\nThe translation factors EEF2 and EIF5A were among\nour top hits, consistent with their known role in pro-\nmoting translation elongation. Knockout of the canon-\nical elongation factors EEF1A1 and EEF1A2 also signif-\nicantly (P = 0.03-0.04) increased ribosome occupancy\nthough they fell just below our FDR threshold for calling\ngene hits (FDR = 0.08-0.09). Intriguingly, the ribosome-\nassociated quality control factor ASCC3 was the top\ngene hit for increased heavy polysome to monosome\nratio (log 2H/M = 0.62, FDR = 1e-4). Since ASCC3 is\ninvolved in splitting stalled ribosomes on mRNAs 49, its\npresence here suggests that even well-translated mR-\nNAs such as this β-globin reporter undergo some de-\ngree of ribosome stalling and quality control. Support-\ning this inference, knockout of the ribosome collision\nsensor ZNF598, which acts upstream of ASCC3 49, also\nincreased ribosome occupancy (log 2H/M = 0.19, FDR\n= 0.06, p = 0.007). In addition, knockout of METAP2,\nwhich removes methionine from the N-terminus of\nnascent polypeptides, increased ribosome occupancy\n(log2H/M = 0.22, FDR = 0.001, p = 3e-4), pointing to an\neffect of nascent peptide processing on the kinetics of\nmRNA translation.\nFinally, we asked whether differential effects of gene\nperturbations on ribosome occupancy as measured by\npolysome to monosome ratios are reflected in their cel-\nlular transcriptional response. Using a genome-scale\nPerturb-seq dataset 16, we correlated and clustered the\ntranscriptional profiles of translation factor perturbations\nthat had concordant or discordant effects on ribosome\noccupancy (Fig. 2I). Perturbations with concordant ef-\nfects on ribosome occupancy (Fig. 2H) did not show\na higher correlation with each other than with perturba-\ntions with discordant effects on ribosome occupancy. In\nfact, depletion of METAP2 and EIF2S1 (EIF2α), which\nare known to interact at a molecular level 50, had a\nmarkedly higher correlation in their transcriptional re-\nsponse even though these gene knockouts had dis-\ncordant effects on ribosome occupancy (Fig. 2H).\nThus, the effects on ribosome occupancy measured by\nReLiC are distinct from the downstream transcriptional\nresponses to these perturbations.\nIsoform-selective splicing screens using ReLiC\nWe next applied ReLiC to investigate regulators of alter-\nnative splicing, an RNA processing event that occurs on\nmost endogenous human mRNAs53. Existing screening\napproaches to study splicing require careful design of\nfluorescent protein reporters 13,54 and can result in high\nfalse positive and negative rates14. We reasoned that in-\nsertion of barcodes in the 3′ UTR will allow us to directly\nmeasure the ratio of different splice isoforms carrying\n4\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted July 26, 2024. ; https://doi.org/10.1101/2024.07.25.605204doi: bioRxiv preprint \n\nthe\nbarcode and thus capture the effect of the linked\nsgRNA perturbation in our ReLiC screens. T o test this\nidea, we used the same β-globin reporter 36,37 as in our\ntranslation screen (Fig. 3A). RNA-seq of cells stably ex-\npressing the β-globin reporter confirmed that the canon-\nically spliced β-globin mRNA with three exons is by far\nthe most abundant isoform with less than 1% of reads\nmapping to the two introns or to the splice junction be-\ntween exons 1 and 3 (Fig. 3B). We then performed three\nisoform-selective screens for regulators that increase in-\ntron 1 retention (i12), intron 2 retention (i23), or exon 2\nskipping (e13) (Fig. 3A). After harvesting RNA 1, 3, 5\nand 7 days post Cas9 induction, we selectively ampli-\nfied each isoform along with the barcode using primers\nthat anneal to the two introns or to the exon 1-exon 3\njunction (Fig. 3C) and counted barcodes by deep se-\nquencing. Similar to our polysome ReLiC screen, we\nused an FDR threshold of 0.05 and a minimum of three\nconcordant sgRNAs for calling gene hits.\nWe detected few or no gene hits one day after Cas9\ninduction for the three splice isoforms (Fig. S3A, Fig.\n3E), consistent with few proteins being depleted at this\nearly time point after their gene knockout. As the du-\nration of Cas9 induction increased, the three isoforms\nexhibited markedly distinct responses (Fig. S3A, Fig.\n3E). Three days after Cas9 induction, 18 gene knock-\nouts increased intron 2 (i23) retention while few or no\ngene hits were detected for the intron 1-retained and\nthe exon 2-skipped isoforms. This difference between\nisoform levels persisted at day 5, suggesting that splic-\ning of intron 2 is more sensitive to genetic perturbations\nthan the other two isoforms. A larger number of gene\nhits emerged for the intron 1-retained isoform by day 7\n(N = 101), while the number of gene hits for the intron 2-\nretained isoform remained similar between days 5 and\n7 (N = 62 and 54). Fewer gene knockouts increased the\nexon 2-skipped isoform (N = 22, 25 at days 5, 7) in com-\nparison to the two intron-retained isoforms at all time\npoints. Effect sizes of the gene hits were reproducible\nacross distinct barcode sets for each gene (Fig. S3B)\nand specific to each isoform (Fig. S3C).\nThe three isoform-specific screens identified both com-\nmon and unique sets of gene hits that were evident by\nautomated gene ontology analysis (Fig. 3D) and by\nmanual inspection (Fig. 3E). Gene hits in the two intron\nretention screens were dominated by core spliceosome\ncomponents and splicing-associated factors (yellow cir-\ncles and triangles, Fig. 3E). Spliceosome hits were dis-\ntributed throughout the splicing cycle starting from the tri-\nsnRNP complex that is required to form the catalytically\nactive spliceosome and included members from most\nknown spliceosomal subcomplexes 55–58. Our screen\nalso identified trans regulators of spliceosomal function\nsuch as CDK11B – a recently identified activator of the\nSF3b complex 59, and BRF2 – an RNA polymerase III\nsubunit required for transcription of U6 snRNA 60.\nRetention of intron 1 was promoted by an additional\ngroup of gene knockouts that were enriched for mRNA\ntranslation and nuclear RNA exosome factors (red and\nbrown triangles, Fig. 3D,E). Loss of ribosomal proteins\nand translation factors might inhibit nonsense-mediated\ndecay (NMD) of the intron 1-retained isoform. While\nretention of either intron 1 or intron 2 will generate a\npremature termination codon (PTC), only the intron 1-\nretained isoform will have a splice junction and an asso-\nciated exon-junction complex (EJC) downstream of the\nPTC, which is a well-known trigger for NMD 61–64. Con-\nsistent with a role for NMD, EJC components (MAGOH,\nEIF4A3, RBM8A) and RNA export factors (NCBP1,\nNCBP2) also emerged as hits only in the intron 1 reten-\ntion screen (Fig. 3E). Nevertheless, core NMD factors\nsuch as UPF and SMG proteins were not detected in\nany of the splicing screens, while the effect of nuclear\nRNA exosome components might be indirect through\ntheir role in ribosome biogenesis or RNA export 65,66.\nDifferential effects of SF3b complex subunits on\nsplicing\nIn contrast to intron retention, perturbations increas-\ning exon 2 skipping were enriched for a narrow set of\nsplicing factors. Components of the U2 snRNP , most\nnotably several members of the SF3 complex, were\namong the top hits (purple squares, Fig. 3D,E), suggest-\ning that their depletion allows some degree of splicing\nbut impairs the correct selection of splice sites. This\nis consistent with the subtle alterations in exon skip-\nping caused by disease-causing mutations in the SF3b\ncomplex67–69. Exon 2 skipping was also promoted by\nperturbing components involved in nuclear protein im-\nport (green squares, Fig. 3D,E), presumably through\ntheir effect on nuclear import of U2 snRNP proteins af-\nter their synthesis in the cytoplasm. Perturbing individ-\nual components of the 7-subunit SF3b complex 70 had\ndistinct effects on exon skipping and intron retention\n(Fig. 4A), even though all 7 subunits are essential for\ncell growth (Fig. S3D). Exon 2 skipping was greatly\nincreased upon loss of the subunits SF3B1, SF3B2,\nSF3B3, SF3B5, slightly increased by loss of SF3B7, and\nunaffected by loss of SF3B4 and SF3B6 (Fig. 4A). In-\ntron 2 retention was increased by loss of SF3B6 and\nSF3B7, while intron 1 retention was increased by loss\nof SF3B1, SF3B2, and SF3B5 (Fig. 4A). By contrast,\nloss of the activating helicase AQR increased the reten-\ntion of both introns 1 and 2 (brown markers, Fig. 4A).\nWe next examined how the differential effects of SF3b\n5\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted July 26, 2024. ; https://doi.org/10.1101/2024.07.25.605204doi: bioRxiv preprint \n\nsubunit\ndepletion on β-globin reporter splicing extend\nto endogenous mRNAs. T o this end, we generated\nHEK293T cell lines with the subunits SF3B5 and SF3B6,\nwhich affected distinct splicing events in our screen,\nindividually depleted through Cas9-mediated knockout.\nWe also targeted AQR, a top hit in both our intron re-\ntention screens, as a positive control and included a\nnon-targeting control sgRNA against firefly luciferase\n(FLUC). We performed RNA-seq 4 days after Cas9 in-\nduction to identify endogenous splicing events that are\nparticularly sensitive to the respective genetic pertur-\nbations. Loss of SF3B5 increased skipping of 45 an-\nnotated cassette exons by 10% or higher (Fig. 4B).\nFor some cassette exons, the exon skipped isoform in-\ncreased over 10-fold from less than 2% to 20-40% of\nthe total isoform fraction (Fig. 4C,D). Loss of SF3B6\nor AQR affected the skipping of less than 10 cassette\nexons at the same effect size, while all three splicing\nfactors increased aberrant retention of a similar number\nof distinct introns (Fig. 4B, Fig. S3D). Interestingly, in-\ncreased intron retention and exon skipping upon SF3B5\nloss occurred at distinct splice sites within the same tran-\nscriptional unit for genes such as RPL24 and RPL41\n(Fig. 4D). In summary, the differential effects of SF3b\nsubunits on splicing of the β-globin reporter extend to en-\ndogenous mRNAs with a subset of SF3b subunits play-\ning a more prominent role in regulating exon skipping.\nReLiC screen for regulators of mRNA quality control\nOur finding of ribosomal proteins and core translation\nfactors as hits in our screen for intron retention (Fig.\n3D,E) suggest that they promote the decay of aberrantly\nspliced mRNAs through the NMD pathway. However,\nprevious CRISPR screens for NMD using fluorescent\nprotein reporters recovered few ribosomal proteins and\ncore translation factors 71,72, presumably because these\ngenes are critical for protein expression. We reasoned\nthat sequencing mRNA barcodes using ReLiC provides\na general approach to identify regulators of mRNA qual-\nity control pathways independent of their effect on pro-\ntein expression. T o test this idea, we modified the β-\nglobin reporter from previous screens to add a prema-\nture termination codon (PTC) at position 39 in the sec-\nond exon (Fig. 5A), such that it is similar to previously\nused reporters for NMD36. At steady state, mRNA levels\nof the PTC-containing reporter were strongly reduced\nrelative to a reporter with a normal termination codon\n(NTC, Fig. S4A). T o measure mRNA effects specific to\nthe PTC and NTC reporters, we combined our ReLiC-\nRBP library with a dual barcoding strategy 19 to normal-\nize barcode counts for the reporter of interest relative to\nthat of the mCherry-puro selection marker within each\ncell (Fig. 5A). We harvested RNA 7 days after Cas9 in-\nduction and counted mRNA barcodes for the PTC and\nNTC β-globin reporters and the mCherry-puro marker\nby deep sequencing.\nOur dual barcode ReLiC screen recovered 90 gene hits\n(FDR < 0.05, 3 sgRNAs with concordant effects) whose\nknockout increased levels of the PTC reporter relative to\nthe mCherry-puro marker (Fig. 5B, Fig. S4C). We did\nnot observe any hits for the NTC reporter at the same\nFDR threshold, as we would expect given that both the\nNTC reporter and mCherry-puro marker encode mR-\nNAs with normal stability (Fig. 5B, Fig. S4C). Several\ncore components of the NMD pathway (UPF1, UPF2,\nSMG1, SMG5, SMG7, ETF1) were among the gene\nhits for the PTC reporter, indicating our ability to identify\nNMD-specific factors (Fig. 5B, pink circles). Other NMD-\nassociated factors such as SMG6 and EIF4A3 fell just\nbelow the FDR threshold but still significantly (MAGeCK\nP-value < 0.05) increased mRNA levels of the PTC re-\nporter (T able S8). Remarkably, a large proportion of\ngene hits for the PTC reporter encoded core factors in-\nvolved in various steps of mRNA translation (Fig. 5B,\nsquares, triangles, and diamonds; Fig. S4B). These\nincluded both small and large ribosomal proteins, ribo-\nsome biogenesis factors, translation initiation factors,\nand aminoacyl-tRNA synthetases. These translation-\nrelated hits are consistent with the known requirement\nof mRNA translation for NMD 73. Interestingly, several\ntranslation initiation factors in the EIF2, EIF2B, and EIF3\ncomplexes emerged as hits in our NMD screen, while\ngene knockouts encoding the EIF4F complex (EIF4A1,\nEIF4E, EIF4G1) did not increase PTC reporter levels\n(Fig. S4D). Notably, the lack of EIF4F hits in our NMD\nscreen was not simply due to variable knockout effi-\nciency since EIF4F components had a similar growth\ndepletion upon knockout as several EIF2, EIF2B, and\nEIF3 components (Fig. S4E). While the biochemical re-\nquirement for EIF4F in NMD remains unclear 74–76, our\ngenetic screen results suggest a limited in vivo role for\nEIF4F compared to EIF2, EIF2B, and EIF3 in regulating\nNMD.\nChemical and genetic modifier screens using ReLiC\nOur NMD screen also identified gene hits involved\nin ER and mitochondrial homeostasis (Fig. 5B, x\nmarkers). Since disruption of ER and mitochondrial\nhomeostasis are known to trigger phosphorylation of\nEIF2α by the kinases PERK and HRI, our ER- and\nmitochondria-related hits might arise from phosphory-\nlation of EIF2α upon their depletion. This is consis-\ntent with the known inhibition of NMD caused by phos-\nphorylation of EIF2α 77–79. T o directly identify regula-\ntors of NMD that act through EIF2α phosphorylation, we\nadapted ReLiC to perform a chemical modifier screen\n6\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted July 26, 2024. ; https://doi.org/10.1101/2024.07.25.605204doi: bioRxiv preprint \n\nusing\nthe small molecule ISRIB that renders translation\ninsensitive to EIF2α phosphorylation 80. After inducing\nCas9 for 6 days, we treated a ReLiC cell pool expressing\nthe PTC reporter with ISRIB or DMSO for 48 hours then\nharvested RNA and counted barcodes. We identified 30\ngene knockouts (FDR < 0.01) that decreased mRNA lev-\nels of the PTC reporter upon ISRIB treatment relative to\nthe DMSO control (Fig. 5C). These gene hits included\nseveral ER- and mitochondrially-localized proteins (Fig.\n5C, x markers), consistent with their knockout inhibiting\nNMD through EIF2α phosphorylation. Some of the IS-\nRIB screen hits were not identified in the original NMD\nscreen since they fell just below the FDR threshold (T a-\nble S8).\nKnockout of several aminoacyl-tRNA synthetases also\ndecreased PTC reporter levels upon ISRIB treatment\n(Fig. 5C, diamonds), suggesting that their depletion\ninhibits NMD through phosphorylation of EIF2α rather\nthan by decreasing translation elongation. T o test this\nhypothesis, we performed a genetic modifier screen us-\ning ReLiC to deplete the EIF2α kinase GCN2, which is\nactivated by uncharged tRNAs that accumulate upon in-\nhibition of aminoacyl-tRNA synthetases 40,41. We trans-\nduced the ReLiC cell pool with lentivirus expressing sgR-\nNAs targeting GCN2 or a non-targeting control, induced\nCas9 for 7 days, then harvested RNA and counted bar-\ncodes. We identified 12 gene hits (FDR < 0.01) that\ndecreased PTC reporter levels upon GCN2 depletion\n(Fig. 5D), out of which 10 were aminoacyl-tRNA syn-\nthetases (Fig. 5D, diamonds), confirming their action\nthrough GCN2-mediated EIF2α phosphorylation. T o-\ngether, the above experiments show that chemical and\ngenetic modifier screening using ReLiC can dissect the\npathways through which gene knockouts affect RNA\nmetabolic processes.\nGCN1 regulates cellular responses to the anti-\nleukemic drug homoharringtonine\nHomoharringtonine (HHT) is an FDA-approved\nchemotherapeutic that targets the ribosome and is\nused to treat chronic and acute myeloid leukemias 81.\nHHT binds to the large ribosomal subunit to arrest ini-\ntiating ribosomes at start codons and inhibit protein\nsynthesis82,83, but how cells respond to this translational\narrest is not well understood. Given ReLiC’s ability to\nidentify regulators downstream of both mRNA transla-\ntion and chemical perturbations, we sought to use this\napproach to probe the cellular response to HHT treat-\nment. T o this end, we performed ReLiC-RBP screens\nusing a simple reporter encoding EYFP (Fig. 6A). After\ninducing Cas9 for 7 days, we treated the cell pool with\n1 μM HHT or DMSO for 6 hours before harvesting RNA\nand counting reporter barcodes.\nUnlike our previous ReLiC screens where we uncov-\nered multiple gene hits and RNA metabolic pathways,\na single gene, GCN1, emerged as a clear hit (FDR <\n0.05) whose knockout increased EYFP reporter mRNA\nlevels during HHT treatment (Fig. 6B). GCN1 acti-\nvates the kinase GCN2 to trigger EIF2α phosphoryla-\ntion in response to amino acid limitation 84. GCN1 also\nbinds collided ribosomes on mRNAs85–87, which can trig-\nger both degradation of the nascent peptide and the\nmRNA88,89. However, since HHT arrests ribosomes at\nthe start codon, we would not expect amino acid limita-\ntion or ribosome collisions to occur under these condi-\ntions. Indeed, our ReLiC screen during HHT treatment\ndid not identify the uncharged tRNA sensor GCN2 or the\nribosome collision sensor ZNF598 and its downstream\neffectors GIGYF2 and DDX6 as hits (Fig. 6C). Since\nribosome collisions also trigger the ribotoxic stress re-\nsponse through the kinase ZAKα that was not included\nin our original screen 85, we measured p38 phosphory-\nlation in wild-type and GCN1-depleted cells. HHT treat-\nment increased p38 phosphorylation in GCN1-depleted\ncells while wild-type cells did not show a correspond-\ning increase. By contrast, treatment with the elongation\ninhibitor anisomycin potently triggered p38 phosophory-\nlation in both wild-type and GCN1-depleted cells (Fig.\n6D).\nRibosome collisions induced by elongation inhibitors\ntrigger upregulation of immediate early genes at the\nmRNA level 90. T o test if GCN1 regulates a similar\ngene expression program during HHT treatment, we\nperformed RNA-seq on wild-type and GCN1-depleted\ncells after 6 hours of HHT treatment and compared to\ncontrol conditions. HHT treatment caused widespread\nchanges in mRNA levels in both wild-type and GCN1-\ndepleted cells with ~225 up-regulated genes and ~450\ndown-regulated genes (> 2-fold change, p < 0.05). How-\never, a small group of 60 genes, which included the im-\nmediate early genes, showed differential up-regulation\nin the GCN1-depleted cells in comparison to wild-type\ncells (Fig. 6E). These included genes such as FOS,\nJUN, and ATF3, which were 2-4 fold up-regulated in wild-\ntype cells upon HHT treatment but were up-regulated\n25-50 fold in GCN1-depleted cells. Other genes such\nas MYC and TIMP3 that were mildly down-regulated in\nwild-type cells upon HHT treatment were instead 2-fold\nor more up-regulated in the GCN1-depleted cells. The\ntranscriptional upregulation of immediate early genes\nalong with increased p38 signaling in GCN1-depleted\ncells point to a potential role for GCN1 in mitigating ribo-\nsome collisions during HHT treatment.\nT o test for occurrence of ribosome collisions on en-\ndogenous mRNAs during HHT treatment, we first per-\n7\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted July 26, 2024. ; https://doi.org/10.1101/2024.07.25.605204doi: bioRxiv preprint \n\nformed\npolysome fractionation from both wild-type and\nGCN1-depleted cells after 1 hour of HHT treatment ( Fig\nS4F). Polysomes collapsed into monosomes upon HHT\ntreatment, and the disome peak was of comparable in-\ntensity and nuclease sensitivity in both wild-type and\nGCN1-depleted cells. Additionally, ribosome profiling\nafter 1 hour of HHT treatment showed no significant dif-\nferences in average ribosome occupancy on mRNAs\nbetween wild-type and GCN1-depleted cells (Fig. 6F).\nThus, ribosome collisions do not occur during HHT treat-\nment at a scale that is detectable by bulk biochemi-\ncal fractionation and do not alter global ribosome oc-\ncupancy on mRNAs. Nevertheless, highly expressed\nimmediate early genes such as JUN and MYC exhib-\nited extensive ribosome density throughout their the 5′\nUTR during HHT treatment (Fig. 6G), which was also\nrecapitulated by analysis of previous ribosome profil-\ning studies ( Fig S4 G). Furthermore, ribosomes initiate\nat multiple in-frame start codons even in the absence\nof HHT on mRNAs of several immediate early genes\nsuch as JUN, MYC, and JUND91–93. T ogether, these\nobservations suggest that collisions occur on these mR-\nNAs between upstream initiated ribosomes that have\ntransitioned to elongation and HHT-arrested initiating ri-\nbosomes at downstream start codons, which are then\nsensed by GCN1.\nDiscussion\nIn this study, we demonstrate ReLiC, an RNA-linked\nCRISPR screening platform for genetic dissection of di-\nverse RNA metabolic processes in human cells. ReLiC\nenables measuring the effect of thousands of gene per-\nturbations on mRNA translation, splicing, and decay –\nmolecular processes that are not readily accessible to\nexisting CRISPR screening methodologies. Our work\nreveals networks of molecular pathways, protein com-\nplexes, and individual proteins that mediate the effect of\ncis sequence elements and chemical perturbations on\nRNA metabolism. The resulting effects are consistent\nwith known molecular mechanisms and also provide\nnew insights into the interplay between RNA metabolic\nprocesses and cellular physiology.\nCombining ReLiC with biochemical fractionation reveals\ncharacteristic relationships between mRNA translation\nand other cellular processes. Knocking out proteaso-\nmal subunits decreases ribosome occupancy at a dis-\ntinct rate relative to growth fitness. The robustness of\nthis relationship hints at a rheostat that tunes the rate\nof global protein synthesis to match proteasomal capac-\nity, and could be mediated by shared cellular signaling\nor metabolic pathways 94,95. Conversely, the lack of ef-\nfect of RNA polymerase II depletion on ribosome occu-\npancy points to a tightly coordinated synthesis of the en-\ntire translation machinery at different rates of transcrip-\ntion. This decoupling between transcriptional capacity\nand ribosomal activity might enable human cells to main-\ntain optimal rates of protein synthesis across diverse cell\nstates and growth conditions, akin to bacteria 47,96.\nReLiC reveals the role of essential pathways and genes\nin RNA metabolism even when their knockout is dele-\nterious to cell growth. Chemical perturbations that ab-\nrogate protein expression can still be probed for their\ngenetic dependencies using ReLiC, as demonstrated\nby our identification of GCN1’s role during HHT treat-\nment. ReLiC captures differential effects of perturba-\ntions within the same protein complex such as between\nmembers of the SF3b complex and between large and\nsmall ribosomal proteins. Unlike biochemical strategies,\nReLiC identifies both direct effectors and indirect regula-\ntors of RNA metabolism, as exemplified by the identifica-\ntion of translation-related pathways across our screens\nfor ribosomal occupancy, splicing, and mRNA decay. In\ncontrast to single cell screening approaches, ReLiC can\nstraightforwardly combine CRISPR screening with bulk\nbiochemical readouts of RNA metabolism, thus provid-\ning a powerful framework to access and screen for RNA\nphenotypes such as localization 97, condensation98, and\nediting99. Further, ReLiC’s ability to selectively amplify\nand dissect rare RNA splicing events underscores its\nexquisite sensitivity and large dynamic range.\nWe anticipate that ReLiC can be extended to a broad\nrange of biological settings, genetic perturbations, and\nRNA types. Applying ReLiC to diverse cell types, cell\nstates, and disease models will reveal differences in\nRNA metabolism that underlie cellular heterogeneity\nand disease progression. While we have used Sp-\nCas9 to induce gene knockouts, alternative effectors\nlike base editors and prime editors can be readily incor-\nporated into our modular workflow to identify the role\nof specific protein domains or regulatory elements on\nRNA metabolism at nucleotide level resolution. Us-\ning non-coding, viral, and synthetic RNAs instead of\nmRNA reporters has the potential to unlock novel RNA\nregulatory mechanisms and therapeutic strategies. Fi-\nnally, expanding ReLiC from our RNA interactome-\nfocused library to all protein coding genes in the human\ngenome will illuminate new interactions between RNA\nmetabolism and other cellular processes.\nAuthor Contributions\nP .J.N. designed research, performed experiments, an-\nalyzed data, and wrote the manuscript. H.P . per-\nformed experiments. C.L.W. and A.C.H. assisted with\npolysome fractionation experiments. C.B., G.Q., and\n8\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted July 26, 2024. ; https://doi.org/10.1101/2024.07.25.605204doi: bioRxiv preprint \n\nK.Y\n.C. performed gene ontology analyses. A.R.S. con-\nceived the project, designed research, analyzed data,\nwrote the manuscript, supervised the project, and ac-\nquired funding.\nAcknowledgements\nWe thank members of the Subramaniam lab, the\nBasic Sciences Division, and the Computational Bi-\nology Program at Fred Hutch for assistance with\nthe project and discussions and feedback on the\nmanuscript. The computations described here were per-\nformed on the Fred Hutchinson Cancer Center com-\nputing cluster. This research was funded by NIH R35\nGM119835 (A.R.S.), NSF MCB 1846521 (A.R.S.), NIH\nT32 GM008268 (P .J.N.), NIH R37 CA230617 (A.C.H.),\nNIH R01 CA276308 (A.C.H.), and NIH GM135362\n(A.C.H.). This research was supported by the Genomics\nand Flow Cytometry Shared Resources of the Fred\nHutch/University of Washington Cancer Consortium\n(P30 CA015704) and Fred Hutch Scientific Comput-\ning (NIH grants S10-OD-020069 and S10-OD-028685).\nThe funders had no role in study design, data collection\nand analysis, decision to publish, or preparation of the\nmanuscript.\nCompeting interests\nNone\nData, Code, and Material Availability\nAll high throughput sequencing data are publicly avail-\nable in the NCBI SRA database under BioProject PR-\nJNA1059490. SRA accession numbers with sample an-\nnotations are provided as supplementary table S5. All\nsoftware used in this study are publicly available as\nDocker images at https://github.com/orgs/rasilab/pa\nckages . All other data and analysis code are publicly\navailable at https://github.com/rasilab/nugent_2024 .\nMaterials and clarifications pertaining to this study can\nbe publicly requested at https://github.com/rasilab/nug\nent_2024/issues/new/choose.\nMaterials and Methods\nPlasmid construction\nPlasmids, oligonucleotides, and cell lines used in this\nstudy are listed in supplemental tables S2-S4. DNA se-\nquences of plasmids used in this study are available at\nhttps://github.com/rasilab/nugent_2024 . Unless speci-\nfied below, DNA fragments used for cloning were either\nexcised out by restriction digestion or amplified by PCR\nfrom suitable templates. Fragments were assembled to-\ngether using Gibson assembly 100, and transformed into\nNEB10beta cells. All constructs were verified by restric-\ntion digestion and Sanger or long read sequencing.\nLanding pad vector construction\nThe attP landing pad vector (pHPHS232) was cre-\nated by using the plasmid backbone, AAVS1 ho-\nmology arms, and cHS4 insulator from pASHS11\n(pAAVS1P-iCAG.copGFP101/Addgene 66577); the T et-\nresponsive promoter, attP, mT agBFP2, P2A, iCasp9,\nT2A, blasticidin S deaminase, and pCMV-rTTA from\npHPHS111 (Addgene 200630); NeoR from pHPHS27\n(mtk8b_LA_AAVS1_SA_neoR102 /Addgene 123742);\nand SV40pA from pHPHS5 (mtk4b_002_tSV40 102 / Ad-\ndgene 123843).\nThe attP* landing pad vector with Cas9 (pHPHS800)\nwas created using the plasmid backbone from\npYTK089103 (Addgene 65196); the cHS4 insula-\ntor from pASHS11 (Addgene 66577); the EF1α\npromoter from pHPHS3 (MTK2_007_pEF1α 102 /Ad-\ndgene 123702); attP* encoded on oAS1848; attB\nencoded on oAS1482/oAS1540; SpCas9-NLS-FLAG\nfrom lentiCas9-Blast 104 (Addgene 52962); T2A from\npPBHS126 (pRRL U6-empty-gRNA-MND-Cas9-t2A-\nBlast105); Hygromycin phosphotransferase (HPH)\nfrom pHPHS7 (MTK6_009 CMV-Hygro-bgPA 102 /Ad-\ndgene 123863); and SV40pA from pHPHS5 (Addgene\n123843).\nReporter plasmid construction\nA base vector for reporter cloning (pHPHS806) was\ncreated using the plasmid backbone from pYTK089\n(Addgene 65196); the cHS4 insulator, TRE3GV pro-\nmoter, and T2A-PuroR from pASHS11 (pAAVS1P-\niCAG.copGFP/Addgene 66577); EYFP-bGHpA from\npPBHS285106; attB* encoded on oAS1853/oAS1854;\nmCherry from pHPHS109 (Addgene 171598); and\nSV40pA from pHPHS5 (Addgene 123843).\nNext, reporters were cloned sequentially into the pH-\nPHS806 base vector. We wanted to add unique 6xN\nbarcodes in the 3′ UTR of all reporters to enable sample\npooling and multiplexing during sequencing. First, we\ncloned a series of reporters with unique barcodes in the\n3′UTR of the mCherry-puro reporters. A region of the\nPuroR cassette was digested out of pHPHS806 using\nBamHI/BsmBI and replaced by Gibson assembly with\nthe same region of PuroR amplified from pHPHS806 us-\ning oAS1292 and one of oAS1883-1886, each of which\nadds a unique 6xN barcode. The resulting plasmids\nwere referred to as pHPHS843-846.\npHPHS843-846 were used as backbones to clone the\nreporters of interest used for ReLiC screens after digest-\ning out EYFP with KpnI/AgeI and inserting EYFP with a\n6xN barcode or β-globin Norm and T er 39 reporters 36\n(referred to as NTC and PTC here). A 3xFLAG tag was\n9\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted July 26, 2024. ; https://doi.org/10.1101/2024.07.25.605204doi: bioRxiv preprint \n\nincluded\nupstream of the NTC and PTC reporters. The\nresulting plasmids were referred to as pHPHS853, pH-\nPHS926, and pHPHS927.\nPlasmid library construction\nFirst, a base vector for sgRNA cloning (pHPHS309) was\ncreated using the plasmid backbone from pYTK090 103\n(Addgene 65197), amplified using oAS1411/oAS1315;\nSV40pA from pHPHS5 (mtk4b_002_tSV40 102/Addgene\n123843), amplified with oAS1331/oAS1571; U6 pro-\nmoter and gRNA scaffold from pAS70 (Brunello library\nin lentiGuide-puro backbone 31/Addgene 73178), ampli-\nfied using oAS1406/oAS1386 and oAS1334/oAS1572,\nrespectively; a GFP dropout cassette from pYTK001 103\n(Addgene 65108), amplified using oAS1407/oAS1408;\nand a cassette encoding EcoRV and AscI re-\nstriction sites, an Illumina R1 sequencing primer\nbinding site, and a T7 promoter, amplified using\noAS1573/oAS1574/oAS1577/oAS1578. The R1 primer\nbinding and T7 sequences are for sequencing of sgRNA\ninserts at the EcoRV site and for in vitro transcription\nfrom genomic DNA, respectively; the AscI site allows\nfor insertion of reporters and barcodes.\nNext, the GFP dropout cassette was excised from pH-\nPHS309 by restriction digestion with BamHI/XhoI and\nreplaced with the custom RBP-targeting dual sgRNA li-\nbrary, which was synthesized by IDT as an oligo pool\noAS1899 (Supplementary T able S1) then amplified us-\ning oAS1612/oAS1613. Assembled plasmid pools were\ntransformed with high efficiency into NEB10Beta E. coli\nand referred to as pHPHS928.\nThe reporter barcodes were subsequently added to\nthe sgRNA library plasmid by Gibson assembly using\nthe plasmid backbone from pHPHS309, amplified us-\ning oAS1315/oAS1331; the RBP dual sgRNA library\nfrom pHPHS928, amplified using oAS1406/oAS1572;\nand a pair of 20xN barcode sequences, amplified\nusing oAS1573/oAS1574/oAS1575/oAS1576. Assem-\nbled plasmid pools were again transformed with high ef-\nficiency into NEB10Beta E. coli, bottlenecked to ~5x10 5\nbarcode pairs, and referred to as pHPHS932.\nNext, an AmpR cassette was inserted between the two\nsgRNAs in a two-step process. First, an AmpR vec-\ntor (pHPHS841) was created by Gibson assembly us-\ning the plasmid backbone from pYTK083 103 (Addgene\n65190), amplified using oAS1875/oAS1876; AmpR\nfrom pYTK083, amplified using oAS1877/oTB11; and\nthe mU6 promoter and tracRNAv2 separated by a\nHindIII site that was ordered as IDT gBlock oAS1878\nand digested with HindIII. The dual sgRNA library in pH-\nPHS932 has two BsmBI restriction sites in between the\ntwo sgRNAs that yield sticky ends that are compatible\nwith those generated from the BsaI and NcoI sites flank-\ning the AmpR cassette in pHPHS841. So, the AmpR\ncassette was then digested out of pHPHS841 using\nBsaI/NcoI and ligated into BsmBI-digested pHPHS932\nlibrary using T4 DNA ligase (Thermo). Ligated plasmid\npools were again transformed with high efficiency into\nNEB10Beta E. coli and referred to as pHPHS934.\nNext, the mCherry-puro reporters with unique 6xN bar-\ncodes were inserted into the pHPHS934 plasmid pool.\npHPHS934 was used as the plasmid backbone and\nwas digested with AscI, which cuts between the 20xN\nbarcodes. Barcoded mCherry-puro reporters were di-\ngested out of pHPHS853, pHPHS926, and pHPHS927\nusing NotI, which includes sequence fragments up-\nstream and downstream of the reporters that are ho-\nmologous to the free ends of AscI-digested pHPHS934\nfor Gibson assembly. Library diversity was maintained\nby transformation into high efficiency into NEB10Beta\nE. coli and the resulting plasmids were referred to as\npHPHS937, pHPHS938, and pHPHS940.\nFinally, the EYFP , β-globin PTC, and β-globin NTC\nreporters were inserted between the sgRNA cassette\nand the upstream 20xN barcode sequence in the pH-\nPHS937, pHPHS938, and pHPHS940 libraries with\nmCherry-puro reporters. pHPHS937-940 were di-\ngested with NotI, which cuts immediately upstream of\nthe upstream 20xN barcode sequence. Digesting pH-\nPHS853, pHPHS926, and pHPHS927 with NotI also\ncut out their EYFP , β-globin PTC, and β-globin NTC re-\nporters flanked by homologous sequences to the free\nends of NotI-digested pHPHS937, pHPHS938, and pH-\nPHS940. So, the digested reporters were directly incor-\nporated into pHPHS937, pHPHS938, and pHPHS940\nby Gibson assembly. Library diversity was maintained\nby transformation into high efficiency into NEB10Beta\nE. coli and the resulting plasmids were referred to as\npHPHS951, pAS243, and pAS244.\nThe above sequence of steps to create the final ReLiC\nlibraries is shown schematically in Fig. S1A.\nLentiviral sgRNA expression plasmid construction\nA vector expressing dual sgRNAs targeting GCN2\nwas created using the plasmid backbone from pH-\nPHS714 (pJRH051107/Addgene 171625), digested with\nBsmBI; and GCN2-targeting sgRNAs encoded on\noAS2037/oAS2038 that were PCR amplified to flank the\ndual sgRNA scaffold using pHPHS928 as a template.\nThe resulting plasmid was referred to as pAS194.\nsgRNA expression plasmid construction for stable\nintegration\nA base vector for cloning (pHPHS859) was pre-\npared by Gibson assembly using pHPHS309 as\n10\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted July 26, 2024. ; https://doi.org/10.1101/2024.07.25.605204doi: bioRxiv preprint \n\na\nbackbone and bGHpA- attB*-mCherry-T2A-puro\nfrom pHPHS809 as an insert. sgRNAs target-\ning AQR, SF3B5, SF3B6, GCN1 and FLuc en-\ncoded on oAS2137/oAS2138, oAS2155/oAS2156,\noAS2157/oAS2158, oAS2069/oAS2070, and\noPN748/oPN749, respectively, were PCR amplified\nto flank the dual sgRNA scaffold using pHPHS928\nas a template. These dual sgRNA PCR products\nwere inserted between the BamHI and XhoI sites of\npHPHS859 by Gibson assembly to make pAS298,\npAS307, pAS308, pAS232, and pHPHS913.\nFor RNA-seq of splicing factors, the β-globin reporter\nfrom pHPHS927 was inserted into pAS298, pAS307,\npAS308, and pHPHS913 by restricting all plasmids with\nNotI, which generates complementary homology arms\nthat were joined together by Gibson assembly to make\npAS310, pAS319-321. For RNA-seq during HHT treat-\nment, the EYFP reporter from pHPHS853 was inserted\ninto pAS232 and pHPHS913 after cutting all plasmids\nwith NotI and joined together by Gibson assembly to\nmake pAS251 and pAS254.\nCell culture\nHEK293T cells (RRID:CVCL_0063, ATCC CRL-3216)\nwere cultured in Dulbecco′s modified Eagle medium\n(DMEM 1X, with 4.5 g/L D-glucose, + L-glutamine, -\nsodium pyruvate, Gibco 11965-092) supplemented with\n10% FBS (Thermo 26140079) and passaged using\n0.25% trypsin in EDTA (Gibco 25200-056). Cells were\ngrown at 37C in 5% CO2. Cell lines were periodically\nconfirmed to be free of mycoplasma contamination.\nGeneration of landing pad cell lines\nT o generate an initial attP landing pad line, HEK293T\ncells were transfected with landing pad plasmid\n(pHPHS232) and pASHS29 (AAVS1 T2 CRISPR\nin pX330 108/Addgene 72833) using polyethylenimine.\nCells were selected with 10 μg/ml Blasticidin S, added\n96 hours post-transfection. Blasticidin selection was re-\nmoved after 4 days, and BFP expression was induced\nby adding 2 μg/ml doxycycline. 24 hours after doxycy-\ncline induction, the culture was further enriched for BFP+\ncells using a FACSAria II flow cytometer (BD). Clones\nwere isolated by limiting dilution into 96-well plates. Af-\nter isolating clones, two were pooled into a single cell\nline (hsPB126).\nT o integrate a Cas9 expression cassette with an or-\nthogonal attP* site into the initial attP landing pad\nclonal lines, hsPB126 was transfected with Cas9 land-\ning pad plasmid (pHPHS800) and Bxb1 expression plas-\nmid (pHPHS115) using TransIT-LT1 reagent (Mirus).\n72 hours post-transfection, hygromycin phosphotrans-\nferase (HPH) was induced by adding 2 μg/ml doxy-\ncycline, then cells were selected with 150 μg/ml Hy-\ngromycin B, added 96 hours post-transfection. After\n7 days, doxycycline and Hygromycin B were removed\nfrom cells and replaced with 10 μg/ml Blasticidin. Blas-\nticidin selection was ended after 7 days, and this poly-\nclonal cell line (hsPN266) was used for subsequent ex-\nperiments.\nIntegration of plasmid libraries into landing pad\nhsPN266 (HEK293T attP* Cas9 ) cells were seeded to\n60% confluency in one 15 cm dish per library. 20\nμg of attB*-containing reporter library plasmid (pAS243,\npAS244, pHPHS951) and 5 μg of Bxb1 expression vec-\ntor (pHPHS115) were transfected per 15 cm dish using\nTransIT-LT1 reagent (Mirus). Each library was trans-\nfected into a single 15 cm dish then expanded into four\n15 cm dishes 48 hours post-transfection. Cells were se-\nlected with 2 μg/ml puromycin, added 72 hours post-\ntransfection. Puromycin selection was ended after 4\ndays, and library cell lines (referred to as hsPN305,\nhsPN306, hsPN285) were contracted back into a sin-\ngle 15 cm dish. 24h after ending puromycin selection, 2\nμg/ml doxycycline was added to induce Cas9 expres-\nsion, and libraries were expanded into three 15 cm\ndishes – one each for RNA and gDNA harvests the next\nday plus a third for continued propagation. This split-\nting procedure was repeated every other day from the\npropagation dish, so harvests could be taking through-\nout the duration of the screen. At no point were cultures\nbottlenecked to fewer than 5x10 6 cells.\nLibrary genomic DNA extraction\nFor each harvest, reporter library genomic DNA was har-\nvested from one 50% confluent 15 cm dish of cells sta-\nbly expressing the ReLiC library. Genomic DNA was\nharvested using Quick-DNA Miniprep kit (Zymo), fol-\nlowing the manufacturer’s instructions, with 2.5 ml of\ngenomic DNA lysis buffer per 15 cm plate. 30 µg of\npurified genomic DNA from each library sample was\nsheared into ~350 nucleotide length fragments by soni-\ncation for 10 minutes on ice using a Diagenode Biorup-\ntor. Sheared gDNA was then in vitro transcribed into\nRNA (denoted gRNA below and in analysis code) start-\ning from the T7 promoter region in the insert cassette us-\ning the HiScribe T7 High Yield RNA Synthesis Kit (NEB).\nTranscribed gRNA was cleaned using the RNA Clean\nand Concentrator kit (Zymo).\nLibrary mRNA extraction\nFor each harvest, reporter library mRNA was harvested\nfrom one 50-75% confluent 15 cm dish of cells stably ex-\npressing the ReLiC library. T otal RNA was harvested by\nusing 3.5 ml of Trizol reagent (Thermo) to lyse cells di-\nrectly on the plate, and then RNA was extracted from\n11\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted July 26, 2024. ; https://doi.org/10.1101/2024.07.25.605204doi: bioRxiv preprint \n\nthese\nlysates using the Direct-zol RNA Miniprep kit\n(Zymo) following the manufacturer’s protocol. polyA+\nmRNA was extracted from total RNA using oligo dT25\nmagnetic beads (NEB). 30-50 μg of total RNA was used\nas polyA selection input for total barcode counting li-\nbraries from each sample while 10-12 μg was used as\ninput for splicing or polysome fraction barcode counting\nlibraries. 4 μl of oligo dT25 beads were used per 1 μg\nof total RNA input.\nmRNA and genomic DNA barcode sequencing\n100-500 ng of polyA-selected mRNA or in vitro tran-\nscribed gRNA from each library was reverse transcribed\ninto cDNA using SuperScript IV reverse transcriptase\n(Thermo) following the manufacturer’s protocol. For RT ,\nwe used a primer that binds downstream of the 20xN\nreporter barcode: either oPN777 for mRNA barcode 1,\noPN731 for gRNA barcode 1, or oPN779 for mRNA bar-\ncode 2. oPN777 and oPN779 contain a 7 nt UMI. Li-\nbraries for sequencing total levels of barcode 1 or bar-\ncode 2 in each sample were performed in a single step.\nFor both barcodes, a 100-200 μl PCR was performed\nusing Phusion polymerase (Thermo) for 20-25 cycles\nwith cDNA template comprising 1/5th of the final volume,\nand oPN776 was used as a constant reverse primer\nthat binds the Illumina P5 sequence present on oPN777\nand oPN779. Indexed forward primers that bind a con-\nstant region upstream of each barcode were used to\nenable pooled sequencing of different samples (one of\noPN730, oPN738, oPN809, oPN815-822, or oJY1-14\nfor Barcode 1 or one of oPN734, oPN739, or oPN823-\n825 for Barcode 2). All of these reactions generated a\n181 bp amplicon that was cut out from a 2% agarose\ngel and cleaned using the Zymoclean Gel DNA Recov-\nery Kit (Zymo).\nFor splicing screens, two rounds of PCR were per-\nformed. Round 1 was performed as a 50 μl PCR for\n30 cycles, again with cDNA template comprising 1/5th\nof the final volume and oPN776 as a constant reverse\nprimer. The forward primer for Round 1 was chosen\nbased on the measured splicing event: oPN841 for\nintron 1 retention, oPN789 for intron 2 retention, or\noAS2029 for exon 2 skipping. These generate 532, 302,\nand 286 bp amplicons, respectively, which were cut out\nfrom a 2% agarose gel and cleaned using the Zymo-\nclean Gel DNA Recovery Kit (Zymo), eluting in 15 μl.\nRound 2 PCR was then essentially the same as the\nsingle-step PCR for total Barcode 1 sequencing, except\nreactions were 20 μl, used 4 μl of cleaned Round 1 prod-\nuct as template, and proceeded for 5 cycles.\nLibraries were sequenced on an Illumina NextSeq 2000\nusing custom sequencing primers. Custom primers for\nBarcode 1 were oAS1701 for Read 1, oPN732 for Index\n1, oPN775 for Index 2, and oPN731 for Read 2. Custom\nprimers for Barcode 2 were oPN735 for Read 1, oPN737\nfor Index 1, oPN778 for Index 2, and oPN736 for Read\n2. Read lengths varied between sequencing runs with\n10% phiX spiked in.\nsgRNA insert-barcode linkage sequencing\nsgRNA insert-barcode linkages were determined at the\nstep right after barcodes were added to the cloned\nsgRNA plasmid pool, prior to adding AmpR between\nthe sgRNAs. A 422 bp amplicon containing both\nsgRNAs and 20xN barcodes was generated from 1.5\nng of pHPHS932 plasmid by 10 cycles of PCR us-\ning oKC196/oPN726 primers and Phusion polymerase\n(Thermo). This product cut out from a 1.5% agarose\ngel and cleaned using the Zymoclean Gel DNA Recov-\nery Kit (Zymo). This sample was sequenced on an Illu-\nmina NextSeq 2000 using custom sequencing primers:\noAS1701 for Read 1 (26 cycles), oKC186 for Index 1 (6\ncycles), oAS1702 for Index 2 (20 cycles), and oKC185\nfor Read 2 (75 cycles).\nCRISPR-Cas9 mediated GCN2 knockout for modi-\nfier screen\nHEK293T cells were seeded to 60% confluency in a 10\ncm dish. Cells were transfected with 5 μg of lentivi-\nral transfer plasmid encoding sgRNA targeting GCN2\n(pAS194) or a nontargeting control (pHPHS714), 4 μg\nof psPAX2 (Addgene #12260), and 1 μg of pCMV-VSV-\nG (Addgene #8454) using Lipofectamine 3000 reagent\n(Thermo). Virus was harvested 48 h post-transfection,\nfiltered using a 0.45 micron syringe filter (Genesee), and\nimmediately used to transduce hsPN283 cells that were\nseeded to 25% confluency in a 15 cm dish. doxycycline\nwas added to 2 μg/ml at the same time as transduction\nto induce Cas9 expression, and this culture was main-\ntained from this point and harvested as described in “In-\ntegration of plasmid libraries into landing pad”.\nCRISPR-Cas9 mediated gene knockout for RNA-seq\nhsPN266 (HEK293T attP* Cas9 ) cells were seeded\nto 80% confluency in a 6-well dish. 1.6 μg\nof attB*-containing dual sgRNA + reporter plasmid\n(pAS251,254,310,319-321) and 400 ng of Bxb1 expres-\nsion vector (pHPHS115) were transfected per well using\nLipofectamine 3000 reagent (Thermo). Each construct\nwas transfected into a single well of the 6-well dish then\nexpanded into a 10 cm dish 48 hours post-transfection.\nCells were selected with 2 μg/ml puromycin, added\n72 hours post-transfection. Puromycin selection was\nended after 4 days on these cell lines (referred to as\nhsAS103,112-114,309,313). After ending puromycin se-\nlection, 2 μg/ml doxycycline was added to induce Cas9\nexpression. Cells were grown in 6-well plates in the\n12\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted July 26, 2024. ; https://doi.org/10.1101/2024.07.25.605204doi: bioRxiv preprint \n\npresence\nof doxycycline for 4 days then harvested for\nRNA-seq.\nPolysome profiling\nAfter Cas9 induction, 293T cells expressing ReLiC li-\nbraries were passaged for 6 days. On day 6, lysates\nwere prepared from each library at 30% confluency in a\n15 cm dish. Cultures were treated with 100 μg/ml cy-\ncloheximide for 5 minutes prior to harvest, then cells\nwere trypsinized (including 100 μg/ml cycloheximide)\nand pelleted at 300xg for 5 min. Cell pellets were lysed\non ice in 300 μl of polysome lysis buffer (10 mM Tris-\nHCl pH 7.4 (Ambion), 132 mM NaCl (Ambion), 1.4 mM\nMgCl2 (Ambion), 19 mM DTT (Sigma), 142 μg/ml cy-\ncloheximide (Sigma), 0.1% Triton X-100 (Fisher), 0.2%\nNP-40 (Pierce), 607 U/ml SUPERase-In RNase Inhibitor\n(Invitrogen)) with periodic vortex mixing. Lysates were\nclarified by centrifugation at 9300xg for 5 min and su-\npernatants were transferred to fresh tubes. This total\nlysate was split into two parts: 50 μl for total mRNA\nisolation, and 250 μl for polysome profiling. For each\nsample, the 250 μL lysate fraction was layered onto a\n10%–50% (w/v) linear sucrose gradient (Fisher) contain-\ning 2 mM DTT (Sigma) and 100 μg/mL heparin (Sigma).\nThe gradients were centrifuged at 37,000 rpm for 2.5\nh at 4°C in a Beckman SW41Ti rotor in Seton 7030 ul-\ntracentrifuge tubes. After centrifugation, samples were\nfractionated using a Biocomp Gradient Station by up-\nward displacement into collection tubes, through a Bio-\nRad EM-1 UV monitor (Bio-Rad) for continuous mea-\nsurement of the absorbance at 260 nm. 820 μl of TRI-\nzol Reagent (Invitrogen) were added to each RNA frac-\ntion. T otal (input), monosome-associated (fraction 4\nand 5), low polysome-associated (fractions 6-9), and\nhigh polysome-associated (fractions 10-13) mRNA sam-\nples were isolated from TRIzol (Invitrogen) using the\nDirect-zol RNA Miniprep Plus Kit (Zymo Research) with\nDNaseI treatment according to manufacturer’s direc-\ntions.\nT o examine whether GCN1 affects the level of RNAse-\nresistant disomes during HHT treatment, polysome pro-\nfiling was performed with four different samples: 293T\ncells expressing sgGCN1 and sgFLUC from “CRISPR-\nCas9 mediated gene knockout for RNA-seq” after 1\nweek of Cas9 induction, treated for 1 hour with 1 μM\nHHT or DMSO. Polysome profiling was performed sim-\nilar to the Polysome ReLiC screen, but with the follow-\ning modifications. Each sample was harvested from one\n10-cm dish at 70% confluency. Prior to loading onto su-\ncrose gradients, lysates were incubated with or without\nthe addition of 1 U of micrococcal nuclease per μg of\nRNA and 5 μM CaCl 2 at room temperature for 1 hour.\nMicrococcal nuclease digests were quenched by addi-\ntion of 5 μM EGTA prior to loading on sucrose gradients.\nRNA-seq\nRNA was isolated using the Direct-zol RNA Miniprep kit\n(Zymo). Sequencing libraries were generated with the\nNEBNext Ultra II Directional RNA Library Prep Kit (NEB)\nand sequenced on a NextSeq 2000 (Illumina) with 2x50\ncycle paired-end reads.\nRibosome profiling\nRibosome profiling was performed with four different\nsamples: 293T cells expressing sgGCN1 and sgFLUC\nfrom “CRISPR-Cas9 mediated gene knockout for RNA-\nseq” after 1 week of Cas9 induction, treated for 1 hour\nwith 1 μM HHT or DMSO. For each sample, we used\none 15-cm plate of cells, seeded to ~40% confluence\nat harvest. Ribosome profiling protocol was adapted\nfrom109 with the following modifications. For sample har-\nvesting, we removed media from each plate and flash\nfroze samples by placing the plate in liquid nitrogen and\ntransferred to −80 °C until lysis. We performed nucle-\nase footprinting treatment by adding 80 U RNase I (In-\nvitrogen AM2294) to 25 μg of RNA. We gel-purified ribo-\nsome protected fragments with length between 26 and\n34 nucleotides using RNA oligo size markers. We used\npolyA tailing instead of linker ligation following previous\nstudies110,111. Libraries were sequenced on an Illumina\nNextseq 2000 in 50bp single end mode.\nImmunoblot analysis\nsgGCN1 and sgFLUC cell lines used for RNA-seq were\nincubated with drugs at indicated concentrations for 30\nor 60 min before harvest. Homoharringtonine (Biosynth,\nFH15974) and anisomycin (RPI, A50100) were dis-\nsolved in DMSO. Cells were rinsed with PBS and lysed\nin RIPA buffer. Lysates were kept on ice during prepa-\nration and clarified by centrifugation at 15,000 rpm for\n10 min. After clarification, supernatants were boiled\nin Laemmli loading buffer containing DTT , and West-\nern blots were performed using standard molecular bi-\nology procedures Proteins were resolved by 4%–20%\nCriterion TGX protein gels (Bio-Rad) and transferred\nto PVDF membranes using a Trans-Blot Turbo transfer\nsystem (Bio-Rad). Membranes were blocked with 5%\nBSA (Thermo) in TBST and incubated with primary anti-\nbodies overnight at 4°C with gentle rocking. Blots were\nwashed with TBST , then incubated with secondary anti-\nbodies diluted in TBST + 5% BSA for 1 hr at RT with\ngentle rocking. Membranes were washed again with\nTBST , developed using SuperSignal West Femto Max-\nimum Sensitivity Substrate (Thermo), and imaged on a\nChemiDoc MP imaging system (Bio-Rad).\n13\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted July 26, 2024. ; https://doi.org/10.1101/2024.07.25.605204doi: bioRxiv preprint \n\nFlow\ncytometry\nAfter dissociating cells from culture dishes, they were\npelleted and resuspended in Dulbecco’s phosphate-\nbuffered saline (Gibco 14190-144) supplemented with\n5% FBS. Forward scatter (FSC), side scatter (SSC),\nBFP fluorescence (BV421), YFP fluorescence (FITC),\nand mCherry fluorescence (PE.T exas.Red) were mea-\nsured for 10,000 cells in each sample using a BD FACS\nSymphony or Fortessa instrument.\nqRT-PCR\nPlasmids that express the β-globin PTC and NTC re-\nporters along with mCherry-Puro (pHPHS926 and pH-\nPHS927) were integrated into hsPN266 cells by trans-\nfection using TransIT-LT1 reagent and pHPHS115 Bxb1\nexpression plasmid. After Puromycin selection, cells\nwere grown in the presence of 2 μg/ml doxycycline\nfor 4 days then RNA was harvested from both sam-\nples from a 6-well plate at 30% confluency. cDNA\nwas prepared from 500 ng of total RNA using random\nhexamer primers and Maxima RT enzyme (Thermo).\ncDNA reactions were diluted 1:10, then 4 μl of diluted\ncDNA were used as template in 20 μl qPCR reac-\ntions using Phusion polymerase (Thermo) and SYBR\nGreen (Thermo) run on a QuantStudio5 thermocycler\n(Thermo). Reactions were performed as 3 biological\nreplicates using oPN719/oPN731 for the β-globin re-\nporters or oPB466/oPB467 for mCherry.\nComputational analyses\nPre-processing steps for high-throughput sequencing\nwere implemented as Snakemake 112 workflows run\nwithin Singularity containers on an HPC cluster. Python\n(v3.9.15) and R (v4.2.2) programming languages were\nused for all analyses unless mentioned otherwise. All\nsoftware used in this study are publicly available as\nDocker images at https://github.com/orgs/rasilab/pa\nckages.\nBarcode to insert assignment\nRaw data from insert-barcode linkage sequencing are\nin FASTQ format. Barcode and sgRNA insert sequences\nwere extracted from corresponding reads and counted\nusing awk; sgRNA inserts and corresponding barcodes\nwere omitted if the sequenced sgRNA insert was not\npresent in the designed sgRNA library (oAS1899). The\nremaining barcodes were aligned against themselves\nby first building an index with bowtie2-build with de-\nfault options and then aligning using bowtie2 with op-\ntions -L 19 -N 1 --all --norc --no-unal -f. Self-\nalignment was used to exclude barcodes that are linked\nto distinct inserts or ones that are linked to the same\ninsert but are aligned against each other by bowtie2\n(presumably due to sequencing errors). In the latter\ncase, the barcode with the lower count is discarded\nin filter_barcodes.ipynb. The final list of insert-\nbarcode pairs with a minimum of 5 reads is written as\na comma-delimited .csv file for aligning barcodes from\ngenomic DNA and mRNA sequencing below.\nBarcode counting in genomic DNA and mRNA\nRaw data from sequencing barcodes in genomic DNA\nand mRNA are in FASTQ format. Barcode and UMI\nsequences were extracted from corresponding reads,\ncounted using awk, and assigned to reporters based\non their unique 6xN identifier. Only distinct barcode-\nUMI combinations where the barcode is present in\nthe filtered barcodes .csv file from linkage sequenc-\ning are retained. The number of UMIs per bar-\ncode and associated insert are written to a .csv file\nfor subsequent analyses in R. Only barcodes with a\nminimum of 20 UMIs were used for analysis. Bar-\ncode counts from pairs of samples were used to run\nMAGeCK38 with --additional-rra-parameters set to\n--min-number-goodsgrna 3. sgRNAs without a mini-\nmum of 20 UMI in one of the compared samples were\nset to 20 UMI counts before running MAGeCK.\nRNA-seq analyses\nRaw reads were aligned against the human genome\n(GRCh38) along with transcript annotations from En-\nsembl (v108, Homo sapiens ). Only primary chromo-\nsomes (1-22, X, MT) were used for sequence align-\nment and downloaded from https://ftp.ensembl.or\ng / p u b / r e l e a s e - 1 0 8 / f a s t a / h o m o _ s a p i e n s / d n a /.\nTranscript annotations were downloaded from h t t p s :\n/ / f t p . e n s e m b l . o r g / p u b / r e l e a s e - 1 0 8 / g t f / h o m o\n_ s a p i e n s / H o m o _ s a p i e n s . G R C h 3 8 . 1 0 8 . g t f . g z,\nand subset using awk to include only transcripts on\nprimary chromosomes. Reference index for align-\nment using STAR v2.7.11a with options --runThreadN:\n36, --runMode: genomeGenerate, --sjdbGTFfile:\ngtf file from above, --limitSjdbInsertNsj:\n3000000, --genomeFastaFiles fasta files from\nabove. Alignment was performed using STAR with\noptions --runThreadN: 36, --runMode: alignReads,\n--alignSJoverhangMin: 300, --alignSJDBoverhangMin:\n6, --outSAMmultNmax: 1, --quantMode: GeneCounts,\n--readFilesCommand: zcat. All annotated splice\njunctions were extracted from the GTF file using\nextract_splice_site_annotations.py which closely\nfollowed the script hisat2_extract_splice_sites.py\nfrom HISAT2. Start and end coordinates of the spliced-\nout intron were used to designate splice junctions. An-\nnotated cassette exons were identified as those splice\njunction coordinates that contain exactly 1 exon within\nthem, exactly 2 introns within them, and either the 5’ or\n14\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted July 26, 2024. ; https://doi.org/10.1101/2024.07.25.605204doi: bioRxiv preprint \n\nthe\n3’ end of the enclosed introns being the same as the\n5’ or the 3’ end of the parent splice junction. Number\nof reads aligning to each splice junction was extracted\nfrom the STAR output file SJ.out.tab. T o quantify exon\nskipping, we used a percent spliced out metric (100 -\npercent spliced in) since most of the cassette exons\nthat were skipped were fully included in unperturbed\ncells. Percent spliced out was calculated using junc-\ntion reads aligning to the skipped isoform junction (mini-\nmum threshold of 2 reads) divided by the sum of junction\nreads aligning to the skipped isoform junction and the\njunction reads aligning to the flanking included isoform\njunctions (minimum threshold of 100 reads summed\nacross the two flanking junctions). Number of reads\naligning to each intron was calculated using the align-\nments file and splice junction annotation file using the\nGenomicRanges function findOverlaps with a minimum\noverlap of 10nt and a minimum threshold of 100 reads.\nThe intron read count was normalized to per nt by mul-\ntiplying by a factor of read_length / ( intron_length\n+ read_length) to account for the difference in length\nbetween introns. Percent spliced in was calculated\nas the normalized intron read count divided by the sum\nof the normalized intron read count and the read count\nfor the annotated splice junction with the intron splice\nout. Gene counts file from STAR was used to perform\ndifferential expression analysis using DESeq2 1.38.0.\nRibosome profiling analyses\nPolyA adapters were trimmed from sequencing reads\nusing cutadapt 4.4 with parameters -a AAAAAAAAAA\n--minimum-length=22 --match-read-wildcards.\nTrimmed reads were aligned to ribosomal RNA con-\ntaminant sequences (NCBI accession NR_003287.2,\nNR_003286.3, NR_023363.1, and NR_003285.2) using\nbowtie 1.3.1 with default parameters. Trimmed reads\nthat did not align to ribosomal RNA were aligned against\nhuman transcripts (MANE v1.3) using bowtie 1.3.1\nwith parameters --norc --no-unal --sam . Aligned\nreads were converted to BAM format, sorted and in-\ndexed using samtools 1.16.1 . Aligned reads between\n27 nt and 33 nt were trimmed by 13nt from their 5′\nend to identify the location of the P-site. The location\nof the P-site relative to the annotated start codon of\neach transcript was calculated using the start codon\nannotation in MANE v1.3. Reads assigned to each\nlocation relative to the start codon of all transcripts\nwere summed and normalized by the maximum value\nacross all locations to calculate the metagene ribo-\nsome P-site profile. P-site profile for individual genes\nwas calculated by summing reads assigned to each\nlocation along the unique MANE transcript for that\ngene. Analysis of previous ribosome profiling stud-\nies was performed using the same pipeline as above\nwith the following modifications. We used the ribosome\nprofiling data from harringtonine- or lactimidomycin-\ntreated samples corresponding to SRA accession\nnumbers SRR1802157, SRR1802156, SRR1802155,\nSRR1802151, SRR1802150, SRR1802149, SRR1802136,\nSRR1802135, SRR1802134, SRR1802133, SRR1333394,\nSRR4293695, SRR4293693, SRR1630828, SRR1630830,\nSRR1630829, SRR6327777, SRR9113062, SRR9113063,\nSRR2732970, SRR2954801, SRR2954800. The se-\nquences CTGTAGGCACCATCAAT and AGATCGGAAGAGC were\nused from adapter trimming. P-site counts across all\nsamples were summed to calculate the profile for indi-\nvidual genes.\nPerturb-seq analyses\nNormalized bulk expression profiles from genome-wide\nPerturb-seq data 16 were downloaded from Figshare as\nthe file K562_gwps_normalized_bulk_01.h5ad. Data\nwere subset to include only the genes of interest, tran-\nscripts with infinite expression were removed, and Pear-\nson correlation coefficients between all pairs of ex-\npression profiles were calculated using the R function\ncorr.test.\nGene ontology analyses\nThe gene_summary.txt output from MAGeCK was or-\ndered either by pos|fdr for positive fold changes or by\nneg|fdr for negative fold changes and input into the\nweb interface of GOrilla 113. Enriched cellular processes\nand components were manually curated for representa-\ntive GO terms with minimal overlap of genes.\nChemicals\nReagent Source Identifier\nISRIB Sigma SML0843\nHomoharringtonine Biosynth FH15974\nAnisomycin Research\nProducts International A50100\nHygromycin B Research Products International H75020\nPuromycin dihydrochloride Research Products International P33020\n15\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted July 26, 2024. ; https://doi.org/10.1101/2024.07.25.605204doi: bioRxiv preprint \n\nAntibodies\nReagent Source Identifier\np38\nMAPK Cell Signaling\nT echnology\n8690; RRID:AB_10999090\nPhospho-p38 (Thr180/Tyr182) Biolegend 690201; RRID:AB_2801132\nGCN1 Bethyl A301843AT ; RRID:AB_1264319\nGoat Anti-Rabbit IgG (H\nL)-HRP Conjugate\nBio-Rad 1721019; RRID:AB_11125143\nGoat Anti-Mouse IgG (H\nL)-HRP Conjugate\nBio-Rad 1721011; RRID:AB_11125936\n16\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted July 26, 2024. ; https://doi.org/10.1101/2024.07.25.605204doi: bioRxiv preprint \n\nFigure\n1\nDevelopment\nof RNA-linked CRISPR (ReLiC) screening in human cells.\nA. Strategy for genomic integration of Bxb1 attP landing pad, SpCas9, and dual sgRNA and barcoded RNA reporters. Unlabeled white\nrectangles represent cHS4 insulator sequences. attP and attP* refer to orthogonal recombination sites for the Bxb1 integrase that differ\nby a single nucleotide mismatch and undergo recombination only with their corresponding attB and attB* sites. Genetic elements are not\ndrawn to scale.\nB. Validation of Cas9 activity. sgEYFP and sgCTRL are single guide RNAs targeting EYFP or a non-targeting control, respectively. Each\nhistogram represents fluorescence of 10,000 cells as measured by flow cytometry. ‘Days post Cas9’ refers to days after addition of\ndoxycycline to induce Cas9 expression.\nC. Strategy for ReLiC sgRNA library design and validation. sgRNAs and barcodes were iteratively cloned as shown in Fig. S1A and\nintegrated into the genome as shown in Fig. 1A.\nD. Correlated change in barcode frequency between genomic DNA and mRNA after Cas9 induction . Each point corresponds to fold-\nchange in mRNA or genomic DNA (gDNA) barcode counts for a single gene between day 1 and days 5, 13, or 21 post Cas9 induction.\nFold-changes are median-centered across sgRNA pairs (sgRNAs henceforth) in the library, and the gene level fold-changes are median\nvalues across all detected sgRNAs for that gene. r refers to Pearson correlation coefficient between mRNA and genomic DNA log 2 fold-\nchanges.\nE. Essential genes are depleted in genomic DNA and mRNA after Cas9 induction. Histogram of fold-change in mRNA or genomic DNA\ncounts for all genes targeted in the ReLiC library. Essential genes were defined as genes annotated as pan-essential in the DepMap\ndatabase (n = 745). All other genes targeted in our library were classified as non-essential (n = 1401).\n17\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted July 26, 2024. ; https://doi.org/10.1101/2024.07.25.605204doi: bioRxiv preprint \n\nFigure\n2\nPolysome\nReLiC identifies regulators of mRNA translation.\nA. Strategy for combining ReLiC and polysome fractionation. Lysates from cell pools expressing the ReLiC-RBP library with a β-globin\nreporter were fractionated on a 10-50% sucrose gradient to separate polysomes from monosomes. Absorbance at 260 nm (A 260) was\nused to monitor ribosomal RNA signal along the gradient during fractionation. RNA extracted from monosome (M), light polysome (L), and\nheavy polysome (H) fractions was used to count reporter barcodes by deep sequencing.\nB. Reporter distribution across polysome fractions. Points correspond to relative mRNA level in each fraction for distinct 3′ UTR barcodes\n(n=6) for the β-globin reporter.\nC. Correlation between replicates. Points represent individual sgRNAs in the ReLiC library. Polysome to monosome ratios are median-\ncentered across sgRNAs in the library. r refers to Pearson correlation coefficient.\n(continued on next page)\n18\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted July 26, 2024. ; https://doi.org/10.1101/2024.07.25.605204doi: bioRxiv preprint \n\n(continued\nfrom previous page)\nD. Gene hits that alter polysome to monosome ratio. Each point corresponds to a gene targeted by the ReLiC library. Horizontal axis\nindicates median of polysome to monosome ratios across all detected sgRNAs for each gene. Vertical axis indicates gene-level P-value\nfrom MAGeCK. Number of genes with FDR < 0.05 and decreased or increased polysome to monosome ratio are indicated with N and the\nindividual genes are highlighted in dark grey triangles. All other genes are shown as light grey circles.\nE. Change in polysome to monosome ratio for ribosomal protein and ribosome biogenesis genes. Closed circles are genes that we call\nas gene hits (FDR < 0.05 with 3 or more concordant sgRNAs). Open circles are genes that do not pass our gene hit threshold.\nF .Change in polysome to monosome ratio for protein groups and complexes. Closed and open circles denote gene hits and non-hits\nsimilar to E.\nG. Comparison of ribosome occupancy and mRNA depletion. Points correspond to genes belonging to one of the highlighted groups.\nShaded areas correspond to 95% confidence intervals for a linear fit of log 2 polysome to monosome ratio to log 2 mRNA depletion within\neach gene group.\nH. mRNA ratios between polysome fractions for individual translation factors. Each point corresponds to a distinct sgRNA pair for that\ngene. Grey bars denote median log 2 ratio across all detected sgRNA pairs for that gene.\nI. Correlation of expression profiles upon depletion of translation factors as measured by Perturb-seq. Bulk expression profiles are from a\nprevious genome-scale Perturb-seq (multiplexed perturbation and single cell RNA-seq) study 16. r refers to Pearson correlation coefficient.\nEEF1A1, EEF1A2, and ZNF598 depletion did not have significant expression correlation with any of the other depletions, so they are\nexcluded to visualize differences between the remaining factors.\n19\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted July 26, 2024. ; https://doi.org/10.1101/2024.07.25.605204doi: bioRxiv preprint \n\nFigure\n3\nIsoform-specific\nsplicing screens using ReLiC.\nA. Schematic of ReLiC splicing screens. A β-globin reporter with 3 exons (e1, e2, e3) separated by two introns was used for ReLiC splicing\nscreens. After ReLiC library integration and Cas9 induction, RNA was harvested at different time points. Barcodes linked to different\nisoforms corresponding to retained intron 1 (i12), retained intron 2 (i23), skipped exon 2 (e13), or all isoforms (total) were amplified by\nPCR and counted by deep sequencing. Location of RT primer and PCR primers used for PCR amplification of barcodes for each isoform\nare shown as black arrows. Splicing phenotype for each gene was calculated as the log 2 ratio of barcode counts for each isoform to the\ntotal barcode counts using MAGeCK. Isoform ratios are median values across all sgRNAs for each gene after median-centering across\nall sgRNAs in the library.\nB. Relative abundance of reporter splice isoforms. T op panel shows RNA-seq read count at each nucleotide position of the 1.7kb β-globin\nreporter. Bottom panel shows the different splice isoforms and the read counts mapping to each splice junction or intron.\nC. Selective amplification of barcodes linked to splice isoforms. Agarose gel lanes show RT-PCR products of expected size for the different\nisoforms: total: 181bp, i12: 532bp, i23: 302bp, and e13: 286bp.\nD. Gene ontology analysis. Selected cellular processes and components enriched among gene hits on day 7 after Cas9 induction. Markers\nare sized according to the fold enrichment of the GO term. GO terms with FDR > 0.05 are indicated by dashes.\nE. Identity of gene hits. Each point corresponds to a gene targeted by the ReLiC library. Different panels correspond to days after\nCas9 induction (horizontal) and isoform screens (vertical). Marker shape denotes isoform identity and marker color denotes one of five\nhighlighted gene groups. Genes with FDR < 0.05 and belonging to one the highlighted groups are listed in the legend.\n20\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted July 26, 2024. ; https://doi.org/10.1101/2024.07.25.605204doi: bioRxiv preprint \n\nFigure\n4\nDifferential\neffects of SF3b complex subunits on splicing.\nA. Relative reporter isoform levels upon SF3b complex perturbations. Splicing phenotypes are shown for genes encoding SF3b complex\nsubunits and the helicase AQR. AQR is shown as a positive control hit for intron retention. FDR < 0.05 is indicated by large marker, and\nFDR ≥ 0.05 is indicated by small marker.\nB. Change in endogenous splicing isoforms upon SF3b complex perturbations. RNA-seq was performed 4 days after inducing Cas9 in\ncells expressing sgRNAs targeting SF3B5, SF3B6, AQR, or a non-targeting FLUC control. Change in intron retention or cassette exon\nskipping were calculated across all ENSEMBL-annotated transcripts, and ranked by decreasing magnitude of change with respect to the\nFLUC control sample.\nC. Examples of endogenous isoform changes. Read counts for RPL41 and RPL24 loci are shown for the RNA-seq from B. Specific\nretained introns and skipped exons are highlighted in green and blue rectangles, respectively. Schematics at the bottom correspond to\nENSEMBL isoforms with the highlighted retained intron and skipped exon events.\nD. Quantification of isoform fraction for the endogenous intron retention and exon skipping events in C. Note that the RNA-seq coverage\nat the skipped exon in C reflects the magnitude of exon inclusion and not the magnitude of exon skipping.\n21\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted July 26, 2024. ; https://doi.org/10.1101/2024.07.25.605204doi: bioRxiv preprint \n\nFigure\n5\nDissecting\nco-translational quality control using chemogenomic ReLiC screening.\nA. Dual barcode strategy for measuring reporter mRNA levels. Red octagons represent location of stop codons along the β-globin reporter.\nN20 barcodes are added to the 3′ UTR of both the reporter and the mCherry-puro control. Reporter mRNA levels are calculated as the\nratio of barcode counts for the reporter to the mCherry-puro control. Reporter mRNA levels represent median values across all sgRNAs\nfor each gene and are median-centered across all sgRNAs in the library after log 2 transformation.\n(continued on next page)\n22\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted July 26, 2024. ; https://doi.org/10.1101/2024.07.25.605204doi: bioRxiv preprint \n\n(continued\nfrom previous page)\nB. Gene hits from dual barcode NMD screen. Genes with increased reporter mRNA level are classified as hits if they have FDR < 0.05\nas calculated by MAGeCK. Hits within one of the six highlighted gene groups are listed in the legend. Genes are arranged alphabetically\nalong the x-axis. The lower right panel shows reporter mRNA level of the highlighted hits for the PTC reporter. Markers for gene hits are\njittered along the x-axis to reduce overlap.\nC. Chemical modifier screen with ISRIB. Cell pool expressing the ReLiC-RBP library with the PTC reporter from a was treated with 200 nM\nISRIB or DMSO for 48 hours after Cas9 induction for 5 days. mRNA fold-change is calculated by normalizing the barcode counts for each\nsgRNA in the ISRIB-treated sample to the corresponding counts in the DMSO-treated sample, and median-centered across all sgRNAs.\nGenes with lower mRNA level in the ISRIB-treated sample and FDR < 0.01 as calculated by MAGeCK are classified as hits. Marker colors\nand shapes denote the highlighted gene groups from B.\nD. Genetic modifier screen with GCN2 depletion. Cell pool expressing the ReLiC-RBP library with the PTC reporter from a was transduced\nwith lentivirus expressing a GCN2-targeting sgRNA or a control sgRNA, followed by Cas9 induction for 7 days. mRNA fold-change is\ncalculated by normalizing the barcode counts for the GCN2 sgRNA sample to the corresponding counts in the control sgRNA sample,\nand median-centered across all sgRNAs. Genes with lower mRNA level in the GCN2 sgRNA sample and FDR < 0.01 as calculated by\nMAGeCK are classified as hits. Marker colors and shapes denote the highlighted gene groups from B.\n23\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted July 26, 2024. ; https://doi.org/10.1101/2024.07.25.605204doi: bioRxiv preprint \n\nFigure\n6\nGCN1\nregulates cellular responses to the anti-leukemic drug homoharringtonine.\nA. Chemogenomic ReLiC screen using homoharringtonine (HHT). ReLiC-RBP cell pool with an EYFP reporter was treated with 1 μM HHT\nor DMSO for 6 hours after Cas9 induction for 7 days.\nB GCN1 regulates mRNA levels upon HHT treatment. Each point represents a gene in the ReLiC-RBP library. Ratio of mRNA barcode\ncounts for the reporter dre calculated between the HHT treatment and the DMSO-treated control, and are median-centered across all\nsgRNAs. Genes with increased reporter mRNA ratio are classified as hits if they have FDR < 0.05 as calculated by MAGeCK.\nC. mRNA level changes upon HHT treatment for factors known to resolve ribosome collisions. Points represent ratios between reporter\nbarcode counts during HHT treatment compared to DMSO treatment for individual sgRNAs targeting each gene. P-values comparing the\nindicated perturbations to cells expressing the nontargeting Nluc control sgRNA are from a two sample t-test: ** (0.001 < P < 0.01), ns (P\n> 0.05).\nD. Immunoblots for phosphorylation of p38 in HEK293T cells +/- GCN1. Cells were treated with homoharringtonine (1 μM), anisomycin\n(10 μM), or DMSO for 1 hour. Anisomycin (ANS) is a positive control for ribosome collision-induced p38 phosphorylation.\nE. GCN1-dependent changes in endogenous mRNA expression during HHT treatment. RNA-seq was performed 8 days after inducing\nCas9 in cells expressing dual sgRNAs targeting GCN1 or a non-targeting FLUC control. Prior to harvest, cells were treated with HHT\n(1 μM) or DMSO for 6 hours. Each points corresponds to a gene and represents the ratio of mRNA levels between HHT and DMSO\ntreatment. Black highlighted points correspond to immediate early genes (IEGs), which are also shown separately in the lower panel.\nF .Metagene alignment of ribosome P-site density in 5′ UTR and CDS region across all transcripts. Ribosome profiling was performed on\n+/- GCN1 cells after harringtonine (1μM) or DMSO treatment for 1 hour.\nG. Ribosome P-site density in 5′ UTR and CDS region of JUN and MYC transcripts. X-axis indicates position along the transcript in\nnucleotides.\n24\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted July 26, 2024. ; https://doi.org/10.1101/2024.07.25.605204doi: bioRxiv preprint \n\nSupplementary\nFigures\nFigure S1\nReLiC\nlibrary design and validation.\nA. Depiction of cloning scheme for ReLiC library and reporters.\nB. Distribution of barcode read counts for sgRNA pairs in mRNA and genomic DNA.\nC. Number of unique barcodes linked to each sgRNA in ReLiC library.\nD. Correlation between distinct barcode sets in ReLiC fitness screens. Each point represents a unique sgRNA pair from the ReLiC RBP\nlibrary. For each sgRNA pair, individual linked barcodes were randomly partitioned into two sets of equal size (or to within a barcode for\nodd number of detected barcodes). r refers to Pearson correlation coefficient between the barcode sets.\n25\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted July 26, 2024. ; https://doi.org/10.1101/2024.07.25.605204doi: bioRxiv preprint \n\nFigure\nS2\nPolysome\nReLiC screen for regulators of mRNA translation.\nA. Gene ontology analysis of perturbations that decrease heavy polysome to monosome ratio.\nGene ontology analysis performed using GOrilla 113 and a subset of enriched terms representative of specific gene classes are shown.\nB. Comparison of heavy polysome to monosome ratio with growth fitness measured by mRNA and genomic DNA barcode seqencing 13\ndays after Cas9 induction for all gene knockouts.\nC. Comparison of heavy polysome to monosome ratio with growth fitness measured by genomic DNA barcode sequencing for gene\nknockouts in specific groups. Points correspond to genes targeted in the ReLiC-RBP library.\nShaded areas correspond to 95% confidence intervals for a linear fit of polysome to monosome ratio to growth fitness within each gene\ngroup.\n26\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted July 26, 2024. ; https://doi.org/10.1101/2024.07.25.605204doi: bioRxiv preprint \n\nFigure\nS3\nIsoform-specific\nsplicing screen using ReLiC.\nA. Number of gene hits that increase the level of the indicated reporter isoform on indicated days after Cas9 induction.\nB. Correlation between barcode sets. For each sgRNA, individual linked barcodes were randomly partitioned into two sets, as in Fig. S1D.\nEach point represents a unique gene that was classified as a hit either with barcode Set A or barcode set B. r refers to Pearson correlation\ncoefficient between barcode sets.\nC. Correlation between relative levels of different mRNA isoforms. Values represent Pearson correlation coefficients for pairwise compar-\nison between the two barcode sets in B.\nD. Depletion of genomic DNA barcodes corresponding to SF3b complex subunits after Cas9 induction.\n27\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted July 26, 2024. ; https://doi.org/10.1101/2024.07.25.605204doi: bioRxiv preprint \n\nFigure\nS4\nDissecting\nmRNA quality control using ReLiC.\nA. Validation of β-globin NMD reporters. Relative reporter mRNA levels measured by qPCR (n=3). Y-axis represents -ΔΔC t value of\nindicated reporter mRNA relative to mCherry-Puro control mRNA.\nB. Gene ontology analysis of perturbations that increase PTC reporter mRNA levels.\nC. Volcano plot of reporter mRNA levels with dual barcode screen.\nEach point corresponds to a gene targeted by the ReLiC library. Marker shape and color denotes one of highlighted gene groups. Genes\nwith FDR < 0.05 and belonging to one of the highlighted groups are listed in the legend.\nD. PTC reporter levels for individual translation initiation complex subunits. Points denote mean and error bars denote standard deviation\nacross sgRNAs for each gene. P-values are as calculated by MAGeCK.\nE. Growth fitness after depletion of translation initiation complex subunits.\n(continued on next page)\n28\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted July 26, 2024. ; https://doi.org/10.1101/2024.07.25.605204doi: bioRxiv preprint \n\n(continued\nfrom previous page)\nF .Polysome profiles of GCN1-depleted and control cell lines after HHT treatment.\nCells were treated with 1 μM HHT or DMSO for 1 hour prior to lysis. Polysome lysates were digested with 1 U micrococcal nuclease / μg\nof RNA prior to sucrose gradient sedimentation to isolate RNAse-resistant monosomes and disomes.\nG. Ribosome P-site density on JUN and MYC mRNAs from previous ribosome profiling studies using harringtonine or lactimidomycin to\narrest initiating ribosomes.\n29\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted July 26, 2024. ; https://doi.org/10.1101/2024.07.25.605204doi: bioRxiv preprint \n\nSupplementary\nTable Descriptions\nS1: sgRNA pairs and genes targeted in the ReLiC-RBP library\nS2: Plasmids used for this study\nS3: Oligonucleotides used for this study\nS4: Cell lines used for this study\nS5: SRA accession numbers\nS6: Read counts for sgRNAs\nS7: MAGeCK output for sgRNA comparisons\nS8: MAGeCK output for gene comparisons\n30\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted July 26, 2024. ; https://doi.org/10.1101/2024.07.25.605204doi: bioRxiv preprint \n\nReferences\n1. Gerstberger, S., Hafner, M. & Tuschl, T . A census of human RNA-binding proteins . Nat Rev Genet 15, 829–845 (2014).\n2. Hentze, M. W., Castello, A., Schwarzl, T . & Preiss, T . A brave new world of RNA-binding proteins . Nature Reviews Molecular Cell Biology\n19, 327–341 (2018).\n3. Kapeli, K., Martinez, F . J. & Yeo, G. W. Genetic mutations in RNA-binding proteins and their roles in ALS . Hum Genet 136, 1193–1214\n(2017).\n4. Gebauer, F ., Schwarzl, T ., Valcárcel, J. & Hentze, M. W. RNA-binding proteins in human genetic disease . Nat Rev Genet 22, 185–198\n(2021).\n5. Sun, S., Zhang, Z., Fregoso, O. & Krainer, A. R. Mechanisms of activation and repression by the alternative splicing factors RBFOX1/2 .\nRNA 18, 274–283 (2012).\n6. Backlund, M. et al. Plasticity of nuclear and cytoplasmic stress responses of RNA-binding proteins . Nucleic Acids Res 48, 4725–4740\n(2020).\n7. Van Nostrand, E. L. et al. A large-scale binding and functional map of human RNA-binding proteins . Nature 583, 711–719 (2020).\n8. Ray, D. et al. A compendium of RNA-binding motifs for decoding gene regulation . Nature 499, 172–177 (2013).\n9. Schneider-Lunitz, V., Ruiz-Orera, J., Hubner, N. & van Heesch, S. Multifunctional RNA-binding proteins influence mRNA abundance and\ntranslational efficiency of distinct sets of target genes . PLoS Comput Biol 17, e1009658 (2021).\n10. England, W. E. et al. An atlas of posttranslational modifications on RNA binding proteins . Nucleic Acids Res 50, 4329–4339 (2022).\n11. Przybyla, L. & Gilbert, L. A. A new era in functional genomics screens . Nat Rev Genet 23, 89–103 (2022).\n12. Genolet, O., Ravid Lustig, L. & Schulz, E. G. Dissecting Molecular Phenotypes Through FACS-Based Pooled CRISPR Screens . Methods\nMol Biol 2520, 1–24 (2022).\n13. Gonatopoulos-Pournatzis, T . et al. Genome-wide CRISPR-Cas9 Interrogation of Splicing Networks Reveals a Mechanism for Recognition\nof Autism-Misregulated Neuronal Microexons . Molecular Cell 72, 510–524.e12 (2018).\n14. Scarborough, A. M. et al. SAM homeostasis is regulated by CFIm-mediated splicing of MAT2A . eLife 10, e64930 (2021).\n15. El-Brolosy, M. A. et al. Genetic compensation triggered by mutant mRNA degradation. Nature 1 (2019) doi: 10.1038/s41586-019-1064-z.\n16. Replogle, J. M. et al. Mapping information-rich genotype-phenotype landscapes with genome-scale Perturb-seq. Cell (2022)\ndoi:10.1016/j.cell.2022.05.013.\n17. Kowalski, M. H. et al. CPA-Perturb-seq: Multiplexed single-cell characterization of alternative polyadenylation regulators.\n2023.02.09.527751 (2023) doi: 10.1101/2023.02.09.527751.\n18. Herholt, A. et al. Pathway sensor-based functional genomics screening identifies modulators of neuronal activity . Sci Rep 8, 17597\n(2018).\n19. Muller, R., Meacham, Z. A., Ferguson, L. & Ingolia, N. T . CiBER-seq dissects genetic networks by quantitative CRISPRi profiling of\nexpression phenotypes. Science 370, (2020).\n20. Alford, B. D. et al. ReporterSeq reveals genome-wide dynamic modulators of the heat shock response across diverse stressors . eLife\n10, e57376 (2021).\n21. Sack, L. M., Davoli, T ., Xu, Q., Li, M. Z. & Elledge, S. J. Sources of Error in Mammalian Genetic Screens . G3 (Bethesda) 6, 2781–2790\n(2016).\n31\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted July 26, 2024. ; https://doi.org/10.1101/2024.07.25.605204doi: bioRxiv preprint \n\n22.\nHill, A. J. et al. On the design of CRISPR-based single cell molecular screens . Nat Methods 15, 271–274 (2018).\n23. Xie, S., Cooley, A., Armendariz, D., Zhou, P . & Hon, G. C. Frequent sgRNA-barcode recombination in single-cell perturbation assays .\nPLoS One 13, e0198635 (2018).\n24. Jordan, A., Defechereux, P . & Verdin, E. The site of HIV‐1 integration in the human genome determines basal transcriptional activity and\nresponse to T at transactivation. The EMBO Journal 20, 1726–1738 (2001).\n25. Ellis, J. Silencing and variegation of gammaretrovirus and lentivirus vectors . Hum Gene Ther 16, 1241–1246 (2005).\n26. Ramezani, A. & Hawley, R. G. Strategies to Insulate Lentiviral Vector-Expressed Transgenes . Methods Mol Biol 614, 77–100 (2010).\n27. Russell, J. P ., Chang, D. W., Tretiakova, A. & Padidam, M. Phage Bxb1 integrase mediates highly efficient site-specific recombination in\nmammalian cells. BioT echniques40, 460–464 (2006).\n28. Xu, Z. et al. Accuracy and efficiency define Bxb1 integrase as the best of fifteen candidate serine recombinases for the integration of DNA\ninto the human genome . BMC Biotechnology 13, 87 (2013).\n29. Ghosh, P ., Kim, A. I. & Hatfull, G. F . The Orientation of Mycobacteriophage Bxb1 Integration Is Solely Dependent on the Central Dinu-\ncleotide of attP and attB . Molecular Cell 12, 1101–1111 (2003).\n30. Jusiak, B. et al. Comparison of Integrases Identifies Bxb1-GA Mutant as the Most Efficient Site-Specific Integrase System in Mammalian\nCells. ACS Synth. Biol. 8, 16–24 (2019).\n31. Doench, J. G. et al. Optimized sgRNA design to maximize activity and minimize off-target effects of CRISPR-Cas9 . Nature Biotechnology\n34, 184–191 (2016).\n32. Lin, J. & Amir, A. Homeostasis of protein and mRNA concentrations in growing cells . Nat Commun 9, 4496 (2018).\n33. Warner, J. R., Knopf, P . M. & Rich, A. A multiple ribosomal structure in protein synthesis . Proc Natl Acad Sci U S A 49, 122–129 (1963).\n34. Noll, H., Staehelin, T . & Wettstein, F . O. Ribosomal Aggregates Engaged in Protein Synthesis : Ergosome Breakdown and Messenger\nRibonucleic Acid Transport . Nature 198, 632–638 (1963).\n35. Gierer, A. Function of aggregated reticulocyte ribosomes in protein synthesis . Journal of Molecular Biology 6, 148–IN8 (1963).\n36. Zhang, J., Sun, X., Qian, Y . & Maquat, L. E. Intron function in the nonsense-mediated decay of beta-globin mRNA: indications that\npre-mRNA splicing in the nucleus can influence mRNA translation in the cytoplasm. RNA 4, 801–815 (1998).\n37. Coulon, A. et al. Kinetic competition during the transcription cycle results in stochastic RNA processing . eLife 3, e03939 (2014).\n38. Li, W. et al. MAGeCK enables robust identification of essential genes from genome-scale CRISPR/Cas9 knockout screens . Genome Biol\n15, 554 (2014).\n39. Wagner, S., Herrmannová, A., Šikrová, D. & Valášek, L. S. Human eIF3b and eIF3a serve as the nucleation core for the assembly of eIF3\ninto two interconnected modules: the yeast-like core and the octamer . Nucleic Acids Res 44, 10772–10788 (2016).\n40. Wek, S. A., Zhu, S. & Wek, R. C. The histidyl-tRNA synthetase-related sequence in the eIF-2 alpha protein kinase GCN2 interacts with\ntRNA and is required for activation in response to starvation for different amino acids . Mol Cell Biol 15, 4497–4506 (1995).\n41. Dong, J., Qiu, H., Garcia-Barrio, M., Anderson, J. & Hinnebusch, A. G. Uncharged tRNA activates GCN2 by displacing the protein kinase\nmoiety from a bipartite tRNA-binding domain . Mol Cell 6, 269–279 (2000).\n42. Darnell, A. M., Subramaniam, A. R. & O’Shea, E. K. Translational Control through Differential Ribosome Pausing during Amino Acid\nLimitation in Mammalian Cells . Molecular Cell 71, 229–243.e11 (2018).\n43. Collart, M. A. The Ccr4‐Not complex is a key regulator of eukaryotic gene expression . WIREs RNA 7, 438–454 (2016).\n32\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted July 26, 2024. ; https://doi.org/10.1101/2024.07.25.605204doi: bioRxiv preprint \n\n44.\nPanasenko, O. O. & Collart, M. A. Presence of Not5 and ubiquitinated Rps7A in polysome fractions depends upon the Not4 E3 ligase .\nMol Microbiol 83, 640–653 (2012).\n45. Harper, J. W. & Bennett, E. J. Proteome complexity and the forces that drive proteome imbalance . Nature 537, 328–338 (2016).\n46. Frydman, J. Mechanism and Function of the Eukaryotic Ring-Shaped Chaperonin TRiC/CCT . Biophysical Journal 102, 428a (2012).\n47. Balakrishnan, R. et al. Principles of gene regulation quantitatively connect DNA to RNA and proteins in bacteria . Science 378, eabk2066\n(2022).\n48. Metzl-Raz, E. et al. Principles of cellular resource allocation revealed by condition-dependent proteome profiling . eLife Sciences 6,\ne28034 (2017).\n49. Juszkiewicz, S., Speldewinde, S. H., Wan, L., Svejstrup, J. Q. & Hegde, R. S. The ASC-1 Complex Disassembles Collided Ribosomes .\nMolecular Cell 79, 603–614.e8 (2020).\n50. Datta, B., Chakrabarti, D., Roy, A. L. & Gupta, N. K. Roles of a 67-kDa polypeptide in reversal of protein synthesis inhibition in heme-\ndeficient reticulocyte lysate . Proc Natl Acad Sci U S A 85, 3324–3328 (1988).\n51. Black, D. L. Mechanisms of Alternative Pre-Messenger RNA Splicing . Annu. Rev. Biochem. 72, 291–336 (2003).\n52. Reixachs‐Solé, M. & Eyras, E. Uncovering the impacts of alternative splicing on the proteome with current omics techniques . WIREs RNA\n13, e1707 (2022).\n53. Kelemen, O. et al. Function of alternative splicing . Gene 514, 1–30 (2013).\n54. Moore, M. J., Wang, Q., Kennedy, C. J. & Silver, P . A. An Alternative Splicing Network Links Cell Cycle Control to Apoptosis . Cell 142,\n625–636 (2010).\n55. Arenas, J. E. & Abelson, J. N. Prp43: An RNA helicase-like factor involved in spliceosome disassembly . Proceedings of the National\nAcademy of Sciences 94, 11798–11802 (1997).\n56. Martin, A., Schneider, S. & Schwer, B. Prp43 Is an Essential RNA-dependent ATPase Required for Release of Lariat-Intron from the\nSpliceosome*. Journal of Biological Chemistry 277, 17743–17750 (2002).\n57. Fourmann, J.-B. et al. Dissection of the factor requirements for spliceosome disassembly and the elucidation of its dissociation products\nusing a purified splicing system . Genes Dev. 27, 413–428 (2013).\n58. Wahl, M. C. & Lührmann, R. SnapShot: Spliceosome Dynamics I . Cell 161, 1474–e1 (2015).\n59. Hluchý, M. et al. CDK11 regulates pre-mRNA splicing by phosphorylation of SF3B1 . Nature 609, 829–834 (2022).\n60. Carrière, L. et al. Genomic binding of Pol III transcription machinery and relationship with TFIIS transcription factor distribution in mouse\nembryonic stem cells . Nucleic Acids Research 40, 270–283 (2012).\n61. Cheng, J., Belgrader, P ., Zhou, X. & Maquat, L. E. Introns are cis Effectors of the Nonsense-Codon-Mediated Reduction in Nuclear mRNA\nAbundance. Molecular and Cellular Biology 14, 6317–6325 (1994).\n62. Thermann, R. et al. Binary specification of nonsense codons by splicing and cytoplasmic translation . The EMBO Journal 17, 3484–3494\n(1998).\n63. Le Hir, H., Gatfield, D., Izaurralde, E. & Moore, M. J. The exon–exon junction complex provides a binding platform for factors involved in\nmRNA export and nonsense‐mediated mRNA decay . The EMBO Journal 20, 4987–4997 (2001).\n64. Lykke-Andersen, J., Shu, M.-D. & Steitz, J. A. Communication of the Position of Exon-Exon Junctions to the mRNA Surveillance Machinery\nby the Protein RNPS1 . Science 293, 1836–1839 (2001).\n65. Schilders, G., Raijmakers, R., Raats, J. M. H. & Pruijn, G. J. M. MPP6 is an exosome-associated RNA-binding protein involved in 5.8S\nrRNA maturation. Nucleic Acids Research 33, 6795–6804 (2005).\n33\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted July 26, 2024. ; https://doi.org/10.1101/2024.07.25.605204doi: bioRxiv preprint \n\n66.\nHiraishi, N., Ishida, Y . & Nagahama, M. AAA-ATPase NVL2 acts on MTR4-exosome complex to dissociate the nucleolar protein WDR74 .\nBiochemical and Biophysical Research Communications 467, 534–540 (2015).\n67. Darman, R. B. et al. Cancer-Associated SF3B1 Hotspot Mutations Induce Cryptic 3’ Splice Site Selection through Use of a Different\nBranch Point. Cell Rep 13, 1033–1045 (2015).\n68. DeBoever, C. et al. Transcriptome sequencing reveals potential mechanism of cryptic 3’ splice site selection in SF3B1-mutated cancers .\nPLoS Comput Biol 11, e1004105 (2015).\n69. Alsafadi, S. et al. Cancer-associated SF3B1 mutations affect alternative splicing by promoting alternative branchpoint usage. Nat Commun\n7, 10615 (2016).\n70. Cretu, C. et al. Molecular Architecture of SF3b and Structural Consequences of Its Cancer-Related Mutations. Molecular Cell 64, 307–319\n(2016).\n71. Zhu, X., Zhang, H. & Mendell, J. T . Ribosome Recycling by ABCE1 Links Lysosomal Function and Iron Homeostasis to 3’ UTR-Directed\nRegulation and Nonsense-Mediated Decay . Cell Rep 32, 107895 (2020).\n72. Zinshteyn, B., Sinha, N. K., Enam, S. U., Koleske, B. & Green, R. Translational repression of NMD targets by GIGYF2 and EIF4E2 . PLOS\nGenetics 17, e1009813 (2021).\n73. Belgrader, P ., Cheng, J. & Maquat, L. E. Evidence to implicate translation by ribosomes in the mechanism by which nonsense codons\nreduce the nuclear level of human triosephosphate isomerase mRNA. Proc Natl Acad Sci U S A 90, 482–486 (1993).\n74. Ishigaki, Y ., Li, X., Serin, G. & Maquat, L. E. Evidence for a Pioneer Round of mRNA Translation: mRNAs Subject to Nonsense-Mediated\nDecay in Mammalian Cells Are Bound by CBP80 and CBP20 . Cell 106, 607–617 (2001).\n75. Durand, S. & Lykke-Andersen, J. Nonsense-mediated mRNA decay occurs during eIF4F-dependent translation in human cells . Nat Struct\nMol Biol 20, 702–709 (2013).\n76. Rufener, S. C. & Mühlemann, O. eIF4E-bound mRNPs are substrates for nonsense-mediated mRNA decay in mammalian cells . Nat\nStruct Mol Biol 20, 710–717 (2013).\n77. Gardner, L. B. Hypoxic Inhibition of Nonsense-Mediated RNA Decay Regulates Gene Expression and the Integrated Stress Response .\nMol Cell Biol 28, 3729–3741 (2008).\n78. Wang, D. et al. Inhibition of Nonsense-Mediated RNA Decay by the Tumor Microenvironment Promotes Tumorigenesis . Mol Cell Biol 31,\n3670–3680 (2011).\n79. Goetz, A. E. & Wilkinson, M. Stress and the nonsense-mediated RNA decay pathway . Cell Mol Life Sci 74, 3509–3531 (2017).\n80. Sidrauski, C. et al. Pharmacological brake-release of mRNA translation enhances cognitive memory . eLife 2, e00498 (2013).\n81. Gandhi, V., Plunkett, W. & Cortes, J. E. Omacetaxine: a protein translation inhibitor for treatment of chronic myelogenous leukemia . Clin\nCancer Res 20, 1735–1740 (2014).\n82. Fresno, M., Jiménez, A. & Vázquez, D. Inhibition of translation in eukaryotic systems by harringtonine . Eur J Biochem 72, 323–330\n(1977).\n83. Ingolia, N. T ., Lareau, L. F . & Weissman, J. S. Ribosome Profiling of Mouse Embryonic Stem Cells Reveals the Complexity and Dynamics\nof Mammalian Proteomes . Cell 147, 789–802 (2011).\n84. Marton, M. J., Crouch, D. & Hinnebusch, A. G. GCN1, a translational activator of GCN4 in Saccharomyces cerevisiae, is required for\nphosphorylation of eukaryotic translation initiation factor 2 by protein kinase GCN2 . Mol Cell Biol 13, 3541–3556 (1993).\n85. Wu, C., Roy, B., He, F ., Yan, K. & Jacobson, A. Poly(A)-Binding Protein Regulates the Efficiency of Translation T ermination . Cell Reports\n33, 108399 (2020).\n86. Pochopien, A. A. et al. Structure of Gcn1 bound to stalled and colliding 80S ribosomes . Proc. Natl. Acad. Sci. U.S.A. 118, e2022756118\n(2021).\n34\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted July 26, 2024. ; https://doi.org/10.1101/2024.07.25.605204doi: bioRxiv preprint \n\n87.\nFedry, J. et al. Visualization of translation reorganization upon persistent ribosome collision stress in mammalian cells . Mol Cell 84,\n1078–1089.e4 (2024).\n88. Oltion, K. et al. An E3 ligase network engages GCN1 to promote the degradation of translation factors on stalled ribosomes. Cell (2023)\ndoi:10.1016/j.cell.2022.12.025.\n89. Müller, M. B. D., Kasturi, P ., Jayaraj, G. G. & Hartl, F . U. Mechanisms of readthrough mitigation reveal principles of GCN1-mediated\ntranslational quality control . Cell 0, (2023).\n90. Sinha, N. K. et al. EDF1 coordinates cellular responses to ribosome collisions . eLife 9, e58828 (2020).\n91. Hann, S. R., King, M. W., Bentley, D. L., Anderson, C. W. & Eisenman, R. N. A non-AUG translational initiation in c-myc exon 1 generates\nan N-terminally distinct protein whose synthesis is disrupted in Burkitt’s lymphomas . Cell 52, 185–195 (1988).\n92. Short, J. D. & Pfarr, C. M. Translational Regulation of the JunD Messenger RNA * . Journal of Biological Chemistry 277, 32697–32705\n(2002).\n93. González-Sánchez, A. M., Castellanos-Silva, E. A., Díaz-Figueroa, G. & Cate, J. H. D. JUN mRNA Translation Regulation is Mediated by\nMultiple 5’ UTR and Start Codon Features. 2023.11.17.567602 (2023) doi: 10.1101/2023.11.17.567602.\n94. Vabulas, R. M. & Hartl, F . U. Protein Synthesis upon Acute Nutrient Restriction Relies on Proteasome Function . Science 310, 1960–1963\n(2005).\n95. Jiang, H.-Y . & Wek, R. C. Phosphorylation of the α-Subunit of the Eukaryotic Initiation Factor-2 (eIF2α) Reduces Protein Synthesis and\nEnhances Apoptosis in Response to Proteasome Inhibition . J. Biol. Chem. 280, 14189–14202 (2005).\n96. Scott, M., Gunderson, C. W., Mateescu, E. M., Zhang, Z. & Hwa, T . Interdependence of Cell Growth and Gene Expression: Origins and\nConsequences. Science 330, 1099–1102 (2010).\n97. Fazal, F . M. et al. Atlas of Subcellular RNA Localization Revealed by APEX-Seq . Cell 178, 473–490.e26 (2019).\n98. Glauninger, H. et al. Transcriptome-wide mRNA condensation precedes stress granule formation and excludes stress-induced transcripts.\n2024.04.15.589678 (2024) doi: 10.1101/2024.04.15.589678.\n99. Xiao, Y .-L. et al. Transcriptome-wide profiling and quantification of N6-methyladenosine by enzyme-assisted adenosine deamination. Nat\nBiotechnol 1–11 (2023) doi:10.1038/s41587-022-01587-6.\n100. Gibson, D. G. et al. Enzymatic assembly of DNA molecules up to several hundred kilobases . Nat Methods 6, 343–345 (2009).\n101. Cerbini, T . et al. Transcription Activator-Like Effector Nuclease (TALEN)-Mediated CLYBL T argeting Enables Enhanced Transgene Ex-\npression and One-Step Generation of Dual Reporter Human Induced Pluripotent Stem Cell (iPSC) and Neural Stem Cell (NSC) Lines .\nPLOS ONE 10, e0116032 (2015).\n102. Fonseca, J. P .et al. A T oolkit for Rapid Modular Construction of Biological Circuits in Mammalian Cells . ACS Synth. Biol. 8, 2593–2606\n(2019).\n103. Lee, M. E., DeLoache, W. C., Cervantes, B. & Dueber, J. E. A Highly Characterized Yeast T oolkit for Modular, Multipart Assembly . ACS\nSynth. Biol. 4, 975–986 (2015).\n104. Sanjana, N. E., Shalem, O. & Zhang, F . Improved vectors and genome-wide libraries for CRISPR screening . Nat Methods 11, 783–784\n(2014).\n105. Gray, E. E. et al. The AIM2-like Receptors Are Dispensable for the Interferon Response to Intracellular DNA . Immunity 45, 255–266\n(2016).\n106. Burke, P . C., Park, H. & Subramaniam, A. R. A nascent peptide code for translational control of mRNA stability in human cells . Nat\nCommun 13, 6829 (2022).\n107. Hamilton, J. R. et al. T argeted delivery of CRISPR-Cas9 and transgenes enables complex immune cell engineering. Cell Rep 35, 109207\n(2021).\n35\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted July 26, 2024. ; https://doi.org/10.1101/2024.07.25.605204doi: bioRxiv preprint \n\n108.\nNatsume, T ., Kiyomitsu, T ., Saga, Y . & Kanemaki, M. T .Rapid Protein Depletion in Human Cells by Auxin-Inducible Degron T agging with\nShort Homology Donors . Cell Reports 15, 210–218 (2016).\n109. Ingolia, N. T ., Brar, G. A., Rouskin, S., McGeachy, A. M. & Weissman, J. S. The ribosome profiling strategy for monitoring translation in\nvivo by deep sequencing of ribosome-protected mRNA fragments . Nat Protoc 7, 1534–1550 (2012).\n110. Ingolia, N. T ., Ghaemmaghami, S., Newman, J. R. S. & Weissman, J. S. Genome-Wide Analysis in Vivo of Translation with Nucleotide\nResolution Using Ribosome Profiling . Science 324, 218–223 (2009).\n111. Lee, S. et al. Global mapping of translation initiation sites in mammalian cells at single-nucleotide resolution . PNAS 109, E2424–E2432\n(2012).\n112. Köster, J. & Rahmann, S. Snakemake—a scalable bioinformatics workflow engine . Bioinformatics 28, 2520–2522 (2012).\n113. Eden, E., Navon, R., Steinfeld, I., Lipson, D. & Yakhini, Z. GOrilla: a tool for discovery and visualization of enriched GO terms in ranked\ngene lists. BMC Bioinformatics 10, 48 (2009).\n36\n.CC-BY 4.0 International licenseavailable under a \nwas not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprint (whichthis version posted July 26, 2024. ; https://doi.org/10.1101/2024.07.25.605204doi: bioRxiv preprint","source_license":"CC-BY-4.0","license_restricted":false}