27 28 29 Endogenous retroviruses (ERVs) are abundant and heterogenous groups of integrated 30 retroviral sequences that impact genome regulation and cell physiology throughout 31 their RNA-centered life cycle 1 . Failure to repress ERVs is associated with cancer, 32 infertility, senescence and neurodegenerative diseases 2-4 . Here, using an unbiased 33 genome-scale CRISPR knockout screen in mouse embryonic stem cells, we identify 34 m 6 A RNA methylation as a novel means of ERV restriction. Methylation of ERV mRNAs 35 is catalyzed by the complex of methyltransferase-like METTL3/METTL14 5 proteins 36 whose depletion, along with their accessory subunits, WTAP and ZC3H13, led to 37 increased mRNA abundance of Intracisternal A-particles (IAPs) and related ERVK 38 elements specifically, by targeting their 5'UTR region. Using controlled auxin-39 dependent degradation of the METTL3/METTL14 enzymatic complex, we showed that 40 IAP mRNA and protein abundance is dynamically and inversely correlated with m 6 A 41 catalysis. By monitoring mRNA degradation rates upon METTL3/14 double degron, we 42 further proved that m 6 A methylation destabilizes IAP transcripts. Finally, similarly to 43 m 6 A writers, triple knockout of the m 6 A readers YTHDF1, DF2 and DF3 6 increased IAP 44 mRNA abundance. This study sheds light onto a novel function of RNA methylation in 45 protecting cellular integrity by clearing reactive ERV-derived RNA species, which may 46 be especially important when transcriptional silencing is less stringent.repressors, by transducing cells with single guide (sg)RNAs against the KRAB-associated 80 protein 1 (KAP1) 11 (Extended Data Fig. 1d,e, Supplementary Fig. 1, Supplementary Table 2, 81 Supplementary Table 3).
4For the screen, we transduced IAPEz-reporter cells with a lentiviral genome-wide 83 sgRNA library at multiplicity of infection (MOI)= 0.2-0.3 12 (Fig. 1b). Frequencies of sgRNAs 84 upon blasticidin selection (5, 7 and 9 days) versus non-selected conditions were assessed 85 via deep sequencing and candidate genes were identified using Model-based Analysis of 86 Genome-wide CRISPR/Cas9 Knockout (MAGeCK) 13 . Efficiency of selection was evidenced 87 by drop-out of control intergenic sgRNAs and genes were ranked based on sgRNA P-values 88 ( Fig. 1c, Supplementary Table 4, Supplementary Table 5). Although genome-wide screens of 89 this magnitude typically suffer from low-statistical confidence, we identified several-but not 90 all-genes previously associated with ERV repression (Resf1, Trp53, Daxx, Atrx, Uhrf1, 91 Cbx1, Dnmt1) 14-17 among the top 100 hits. Obtaining an incomplete list of known regulators 92 is a common outcome of genome-wide screens which can be due to multiple factors 93 including time-dependent dropout of essential genes with strong effects on cell viability (such 94 as Kap1) 18 , heterogeneity in sgRNA efficiencies or limited representation of sgRNAs.
95Nonetheless, we were able to identify several novel candidates for IAP control
96( Supplementary Table 4), offering a foundation for futu...