Loss of the Nuclear Protein RTF2 Enhances Influenza Virus Replication

Viral infection triggers the secretion of type I interferons, which in turn induce the expression of hundreds of antiviral genes. However, the roles of these induced genes in controlling viral infections remain largely unknown, limiting our ability to develop host-based antiviral therapeutics against pathogenic viruses, such as influenza virus. Here, we performed a loss-of-function genetic CRISPR screen in cells prestimulated with type I interferon to identify antiviral genes that restrict influenza A virus replication. Besides finding key components of the interferon signaling pathway, we discovered a new restriction factor, RTF2, which acts in the nucleus, restricts influenza virus transcription, and contributes to the interferon-induced upregulation of known restriction factors. Our work contributes to the field of antiviral immunology by discovering and characterizing a novel restriction factor of influenza virus and may ultimately be useful for understanding how to control a virus that causes significant morbidity and mortality worldwide.

ized, especially in terms of their antiviral effect sizes, their targets, and their mechanisms of viral restriction (15). A deeper understanding of how virus restriction occurs naturally may inform novel therapeutic strategies against viruses that pose major health threats, such as influenza virus, which causes an estimated 3 million to 5 million cases of severe illness and 0.29 million to 0.65 million deaths worldwide annually (16,17).
This knowledge gap has been addressed in part by studies that systematically overexpressed each of 400 ISGs to determine their effects on virus replication, identifying dozens of factors with antiviral activity (15,18). However, this does not address the question of whether the individual ISGs are required, rather than sufficient, for restricting virus replication. Although loss-of-function screens have been done before, very few ISGs have been identified through such an approach (4,19). For instance, in the genome-wide small interfering RNA screen performed by Brass et al. to identify host factors that are involved in influenza A virus (IAV) infection, only 4 genes were reported to increase IAV replication and/or gene expression (4).
We hypothesized that a reason for not identifying more anti-IAV factors through such genome-wide screens is because the screens had not been designed and optimized specifically for such a purpose. Furthermore, the previously published IAVspecific screens were not performed in the presence of IFN treatment, and hence, it is possible that only restriction factors that have constitutively high expression levels and/or that exhibit very potent restriction effects would be identifiable. We thus hypothesized that if we were to tailor our screen conditions specifically to identify antiviral factors, such as pretreating cells with IFN prior to IAV infection, we could potentially identify novel host factors that restrict IAV infection in a systematic and unbiased way. Such a genome-wide screen not only could potentially identify novel antiviral effectors but also could identify cofactors or downstream effectors that are necessary for previously reported ISGs to function (such as ZMPSTE24, which is required for IFITM3 activity but which is not upregulated by IFN [20]).
In our study, we screened for factors that restrict IAV replication in response to interferon beta (IFN-␤) exposure in A549 lung epithelial cells and present evidence that replication termination factor 2 (RTF2) restricts influenza virus replication in the nucleus and affects the response to type I interferons.

RESULTS
A genome-wide CRISPR screen for factors that restrict influenza virus reveals components of the type I interferon pathway but not individual interferonstimulated effector genes. To perform a genome-wide CRISPR screen, we generated A549 lung epithelial cells that stably express Cas9 (A549-Cas9 cells) and transduced these A549-Cas9 cells with the AVANA4 lentivirus library (21), which contains 74,700 single guide RNAs (sgRNAs) targeting 18,675 protein-coding genes, as well as 1,000 nontargeting sgRNAs. We added IFN-␤ to ϳ300 million library-transduced cells and after 24 h infected the cells with the influenza A virus A/Puerto Rico/8/1934 (PR8) at a multiplicity of infection (MOI) of 5. After 16 h, influenza virus-infected cells were flow sorted based on the level of viral hemagglutinin (HA) on the cell surface, and their genomic DNA was extracted for sequencing to identify enriched or depleted sgRNAs (Fig. 1A). To find antiviral factors, we identified sgRNAs that were enriched in cells that were within the top 10% of HA-expressing cells (infected, susceptible) versus cells with median viral HA expression in the lower peak (uninfected, protected) (Fig. 1B). The top three hits were interferon alpha/beta receptor 2 (IFNAR2), nonreceptor tyrosine-protein kinase 2 (TYK2), and Janus kinase 1 (JAK1), and further down the list were interferon alpha/beta receptor 1 (IFNAR1) and interferon regulatory factor 9 (IRF9), all of which are essential components in the IFN signaling pathway (22). These were further validated using additional single guides. However, we did not recover previously reported antiviral ISGs (23), such as the genes for IFITM3 (4-6), Mx1 (7), PKR (8-10), OAS3 (13), RNase L (14), RSAD2 (viperin) (24), TRIM22 (25), and ISG15 (26). This could be due to a combination of reasons, such as each gene having a small effect size in the context of  IFN-induced antiviral genes (as well as the expected reduction in effect size due to incomplete editing [27] and potential enrichment for nonedited cells that had higher fitness), as well as an inherent limitation of the assay that we chose (only genes acting on IAV life cycle steps up to HA trafficking would be recoverable). In contrast, since the IFN pathway components found in our screen are known to be required for inducing many ISGs at once, the impact of deleting them is to undo the antiviral effects of IFN. Using a more sensitive arrayed format to validate some of our primary screen hits, we found that of four well-studied antiviral effector genes, those for IFITM3, Mx1, PKR, and RSAD2, only the gene for IFITM3 showed an increase in infection compared to the nontargeting guides (Fig. 1C), while guides targeting IFN pathway components gave the expected increase in infection.
Validation of RTF2 as an antiviral factor. Among the top hits in our primary screen, RTF2 (also known as RTFDC1) was found to be an antiviral gene, with little being known about its cellular function (28) or role in infection. As further validation, we found that independent guides, which were designed to spread out across the entire RTF2 gene (Fig. 1D), led to higher HA levels in IAV-infected IFN-␤-pretreated A549 cells than in cells that received nontargeting guides (Fig. 1E), with the effect size being inversely correlated to RTF2 expression levels (Fig. 1F). In addition, overexpression of sgRNA-resistant RTF2 in RTF2-depleted cells restored protection against IAV infection ( Fig. 2A). Because gene editing with CRISPR-Cas9 does not lead to a loss of function in all cells of a polyclonal population, we isolated individual clones of cells treated with a strong guide against RTF2. Indeed, the effect size of knocking out RTF2 was larger in several clonal knockout lines in which RTF2 protein expression was low or undetectable (Fig. 2B). Subsequent experiments were conducted using clone 1 cells (RTF2-knockout [RTF2-KO] cells) and clone 1 cells expressing sgRNA-resistant RTF2 (RTF2-rescued cells), as such experimental data would be more reproducible due to the homogeneous nature of clonal cells. These clonal RTF2-KO cells had no detectable RTF2 protein and had a 2-bp deletion (chromosome 20 bp 56,474,701 and 56,474,702) leading to a frameshift in one allele and a 24-bp deletion (chromosome 20 bp 56,474,701 to 56,474,724) in the other allele (located within exon 3 of the RTF2 transcript ENST00000357348.9).
Besides monitoring the infection rate by measuring the percentage of HA-positive cells, we also monitored the production of infectious particles in the IAV-infected, IFN-␤-pretreated cells at 16 h postinfection, by collecting the cell culture supernatant and performing plaque assays on Vero cells. We noticed that a significantly higher virus titer was observed in RTF2-KO cells than in wild-type (WT) A549 cells and RTF2-rescued cells (Fig. 2C). Based on these data, we conclude that RTF2 has an antiviral role during IAV infection. Viral RNA levels are increased in RTF2-deficient cells at early time points. To monitor viral RNA levels, we used quantitative PCR (qPCR) to quantify IAV RNAs from infected cells and found that the nucleoprotein (NP) mRNA, cRNA, and viral RNA (vRNA) levels were higher in RTF2-KO cells than in wild-type A549 cells or RTF2-rescued cells (Fig. 2D). We also tracked the kinetics of infection by monitoring NP mRNA levels over time and noticed that the difference in NP mRNA levels between RTF2-KO cells and RTF2-rescued cells could be detected even at 4 h postinfection (hpi) (Fig. 2E). Our data suggest that depletion of RTF2 leads to increased replication during an early step in the IAV life cycle.
RTF2's localization is important for its antiviral role. We hypothesized that RTF2 functions in the nucleus, because prior work showed that RTF2 localizes to replication forks in the nucleus (28) (and has to be removed from stalled replication forks to maintain genome stability) and that the RTF2 homolog in Schizosaccharomyces pombe inhibits replication restart to achieve efficient replication termination at the site-specific replication barrier, RTS1 (29). Through immunofluorescence staining of wild-type A549 cells and two different RTF2-KO cell clones (clones 1 and 7) with the endogenous anti-RTF2 antibody, we found that RTF2 appears to be localized mainly in the nucleus (Fig. 3A). This was corroborated by biochemical fractionation (Fig. 3B). In addition, live-cell imaging of wild-type A549 and RTF2-KO cells transduced to express mCherrytagged RTF2 showed a clear nuclear localization of the overexpressed RTF2 (Fig. 3C).
We then wondered if the nuclear localization of RTF2 is essential for its antiviral function. To do this, we first used the cNLS Mapper program (30) to identify potential classical importin-alpha/beta pathway-specific nuclear localization signals (NLSs) on RTF2. We found two potential NLSs in RTF2: a monopartite NLS (LEKKTKKPKKA) and a bipartite NLS (GATKRSIADSEESEAYKSLFTTHSSAKRSKE). Site-directed mutagenesis of the nucleotide sequences encoding the predicted monopartite NLS (by changing several positively charged residues to alanine residues) had no effect on the location of RTF2, but changing the sequences for the predicted bipartite NLS caused the overexpressed protein to be present in both the cytosol and the nucleus (Fig. 4A). This is perhaps because RTF2 is a small protein (33 kDa) that could still diffuse back into the nucleus upon overexpression despite having both predicted NLSs disrupted.
To test whether RTF2 loses its activity if it is mislocalized such that it cannot diffuse back into the nucleus, we generated an endoplasmic reticulum (ER)-retention mutant (KDEL) and a plasma membrane-targeted N-terminus-myristoylated mutant (Fig. 4B) and showed that the proteins were localized away from the nucleus (Fig. 4C). To test the effect of mislocalizing RTF2 during IAV infection, we quantified infection based on surface HA expression (Fig. 4D) and NP mRNA levels (Fig. 4E). By retaining RTF2 in the endoplasmic reticulum or by redirecting it to the inner leaflet of the plasma membrane, we disrupted RTF2's ability to confer antiviral protection. On the other hand, point mutations of the NLSs alone had no impact or a weak impact on RTF2 function in the rescued RTF2-KO cells. These data indicate that the RTF2 protein is localized to the nucleus and that relocalization to the plasma membrane or ER impairs its antiviral function.
Loss of RTF2 leads to increased viral polymerase activity and elevated levels of influenza virus mRNAs synthesized during primary transcription. Given the requirement of nuclear localization for the antiviral function of RTF2 and since IAV, unlike most RNA viruses, transcribes and replicates its genome in the nucleus, we wondered if RTF2 impacts IAV transcription and/or genome replication. To test this, we transfected wild-type A549, RTF2-KO, and RTF2-rescued cells with a plasmid harboring the gene for the firefly luciferase reporter driven by the viral RNA (vRNA) backbone, along with plasmids harboring the genes for PR8 polymerase subunits (PA, PB1, PB2) and NP (31). We found that RTF2-KO cells had a higher luciferase activity than the other cells, suggesting that cells that lack RTF2 are less able to restrict influenza virus transcription/ replication (Fig. 5A). We then tested whether rescuing RTF2-KO cells with the mislocalized mutants described above would result in luciferase activity lower or higher than that from rescuing cells with the NLS-intact form of RTF2. We found that overexpressing mislocalized RTF2 in the RTF2-KO cells did not reduce the higher luciferase signal seen in RTF2-KO cells relative to that seen in wild-type or RTF2-rescued cells (Fig. 5B), once again indicating that the nuclear localization of RTF2 is important for its function. In addition, we noted that the increased firefly luciferase reporter signal in RTF2-KO cells does not appear to correlate with increased transcription levels of transfected genes for IAV NP, PA, PB1, and PB2 (Fig. 5C), suggesting that the higher luciferase signal is indeed due to increased IAV polymerase activity.
To further dissect RTF2's role in controlling IAV infection, we tested whether RTF2 specifically affects primary transcription by employing cycloheximide (CHX) to block protein translation. We tracked NP mRNA, cRNA, and vRNA levels after infecting CHXand IFN-␤-pretreated cells with IAV. As CHX treatment blocks the synthesis of new viral Cells were first suspended in HEPES-sucrose-Ficoll (HSF) buffer containing digitonin to extract the cytosolic proteins and washed once with HSF buffer, before lysing the nuclear pellet in RIPA buffer to release the nuclear proteins. TATA-binding protein (TBP), a nuclear protein, and tubulin, a cytosolic protein, were included as controls for the fractionation protocol. (C) Live-cell imaging of cells expressing either mCherry or the RTF2-mCherry fusion protein. WT A549 (top) and RTF2-KO cells (bottom) were transduced with lentiviruses that encode either FLAG-mCherry or RTF2-FLAG-mCherry and cultured for a week, before they were visualized under a wide-field epifluorescence microscope. proteins and, hence, halts genome replication and secondary transcription, the levels of NP cRNA and vRNA did not increase after infection in CHX-treated cells ( Fig. 5D and E). However, in CHX-treated cells, IAV NP mRNA levels (resulting from primary transcription) did go up slightly over time (Fig. 5E). More importantly, NP mRNA levels were higher in RTF2-KO cells than in RTF2-rescued cells, suggesting that RTF2 affects primary transcription. Altogether, these findings are consistent with the model that RTF2deficient cells allow more primary transcription of the virus.
However, it was unclear whether RTF2 also affects genome replication and/or RNA stability in general. To test the latter, we employed baloxavir, a PA endonuclease inhibitor that blocks cap snatching and, hence, the synthesis of mRNAs. We first infected IFN-␤-pretreated cells with IAV for 6 h, before administering baloxavir to prevent the additional synthesis of mRNAs and monitoring the RNA levels over time.
We found no evidence that influenza virus mRNA decays at a lower rate in RTF2-KO cells than in RTF2-rescued cells (Fig. 5F), suggesting that the higher mRNA level in RTF2-KO cells might be due to an increased synthesis rate.
In conclusion, RTF2's localization to the nucleus appears to be important for its function in controlling IAV. While we cannot rule out the possibility that IAV secondary transcription and genome replication are also affected by RTF2, our experiments showed that RTF2-KO cells lead to elevated primary transcript levels at a minimum, likely by affecting mRNA synthesis rather than degradation.
Cross-talk between RTF2 and the IFN pathway. Since most of our experiments were conducted in the presence of IFN pretreatment, we wondered if IFN is required to observe RTF2's role in influenza virus restriction. We found no changes in infection when RTF2 was overexpressed (Fig. 6A), indicating that RTF2 alone is not sufficient to restrict viral replication. To investigate the dependence on IFN, we infected wild-type A549, RTF2-KO, and RTF2-rescued cells in the presence or absence of IFN and then measured the infection rates based on surface HA levels. We found that IFN pretreatment (which reduces overall infection rates) enhances the fold increase in infection between RTF2-KO cells and RTF2-rescued cells (Fig. 6B). We adopted two other complementary approaches-by using recombinant B18R (Fig. 6C), a decoy IFNAR protein produced by vaccinia virus to soak up IFN in the culture medium, and by using ruxolitinib (Fig. 6D), a JAK inhibitor-to block IFN signaling. All three approaches suggest that an intact IFN signaling pathway is required to observe the effect of RTF2 knockout (KO) on infection. A potential reason why the RTF2-KO phenotype is observed only in the presence of IFN may be that IFN activates RTF2 or upregulates an unknown cofactor of RTF2. This model remains to be tested. On the other hand, because RTF2 protein and mRNA levels were not upregulated in the presence of IFN (Fig. 6E to G), we hypothesized, alternatively, that RTF2 restricts influenza virus by positively regulating the response to IFN and that the loss of RTF2 would lead to a blunted IFN-induced antiviral response. Interestingly, we also noticed that although RTF2 expression appeared to be unaffected by IFN-␤ treatment alone, it was significantly reduced upon IAV infection (Fig. 6F and G).
To investigate the hypothesis that RTF2 positively regulates the response to IFN, we performed RNA sequencing to compare the transcriptomes of RTF2-KO and RTF2rescued cells in the presence of IFN. Principal-component analysis (PCA) showed that one dimension was associated with IFN exposure, while the other dimension was associated with IAV infection (Fig. 6H). While the transcriptomes of RTF2-KO and RTF2-rescued cells were quite similar in the absence of IFN exposure (and IAV infection), RTF2-KO cells showed lower levels of induction of ISGs than wild-type or RTF2-rescued cells (Fig. 6I). Among the reduced ISGs, we found the known influenza virus restriction factors IFITM1 and IFITM3 (Fig. 6J), which validated the findings at the protein level (Fig.  6K). Consistent with this observation, we found that the phosphorylation of STAT1, a major transcription factor responsible for upregulating ISGs during IFN signaling, was reduced in RTF2-KO cells (Fig. 6L) Fig. 7A and B) and higher VSV N mRNA levels (Fig. 7C).
Collectively, our data suggest that RTF2 contributes positively to the cellular response to IFN via the IFNAR pathway, as the loss of RTF2 results in lower levels of phosphorylated STAT1 (pSTAT1) and ISGs and impairment of the IFN-induced antiviral response.

DISCUSSION
In our screen, we used HA expression as a surrogate readout for viral infection and replication. There are inherent caveats to this approach. First, restriction factors that inhibit later stages of the life cycle, such as assembly, budding, and release, as well as subsequent rounds of infection (since we chose to omit trypsin from the cell culture medium), would be missed in our screen. Examples of such antiviral factors that could be missed include viperin and tetherin. In addition, potential host restriction factors that inhibit the expression of non-HA late genes of IAV, such as M2 and neuraminidase (NA), may also be missed. However, any restriction factors that target viral proteins involved in viral gene expression (such as polymerase subunits PA, PB1, and PB2 or an indirect regulator, such as NS1, which inhibits the cellular antiviral response and which traps cellular pre-mRNAs) should be captured in our screen. Any restriction factor that inhibits HA expression and/or that reduces HA stability should also be recoverable in our screen (since knocking out these hypothetical restriction factors would result in higher levels of HA expression). Interestingly, although our screen recovered genes involved in the IFN signaling pathway (the genes for IFNAR1, IFNAR2, JAK1, TYK2, and IRF9), we did not recover previously reported antiviral ISGs. This is in spite of the fact that both the IFITM3 and Mx1 proteins have been reported to be functional in A549  cells. We hypothesized that this might be due to the small effect size of each individual antiviral effector gene, as well as the redundancy of the many ISGs upregulated upon IFN exposure. Hence, knocking out a single ISG may not be sufficient for viruses to overcome the IFN-induced protection. In addition, despite puromycin selection eliminating nontransduced cells, the incomplete deletion of genes by CRISPR would further reduce the small effect sizes of each ISG and make it challenging to observe the enrichment of specific sgRNAs after flow sorting. Interestingly, this observation of not recovering classical ISGs has also been made in another screen recently described by Richardson et al. (19). Similar to our screen, their screen recovered genes in the IFN signaling pathway (such as the genes for IFNAR1, IFNAR2, JAK1, TYK2, and IRF9), as well as IFI6, but not classical ISGs, such as the genes for IFITM3 (4,15) and TRIM56 (32), which have previously been shown to inhibit yellow fever virus. Potential improvements to such screens in future could involve using a lower IFN dose, using a small validated sgRNA library, increasing the number of cells in the screen to detect smaller effect sizes, and performing a combinatorial screen (e.g., by single-cell approaches to test many combinations) where multiple antiviral genes are knocked out in the same cell.
However, despite not recovering any of the previously reported antiviral ISGs, we found that the loss of RTF2 led to the increased replication of influenza virus, at least in part through an increase in primary transcription and in part through a reduction in  the IFN response. This might explain why RTF2 was recovered in our screen when none of the other previously reported antiviral effectors, such as IFITM3, were identified.
While we do not know if RTF2 KO also affects early steps of infection, we found that RTF2-KO cells exhibited a higher rate of primary transcription (rather than a lower rate of degradation). RTF2-KO cells also displayed elevated luciferase activity in the minigenome reporter assay, an assay that is based on transfected plasmids instead of infection with live viruses and, hence, that bypasses all the earlier steps in the viral life cycle. Collectively, these suggest that, at a minimum, RTF2 affects influenza virus polymerase complex activity.
However, we have not resolved whether RTF2's restriction of influenza virus is due primarily to RTF2 being a direct antiviral effector, a global modulator of the IFN pathway, or both. It is increasingly appreciated that restriction factors could have dual functions (33), with examples such as IFITM3 acting as a feedback inhibitor of type I IFN induction (34), SAMHD1 inhibiting NF-B and IRF7 activation (35), and viperin enhancing Toll-like receptor signaling and type I IFN production in plasmacytoid dendritic cells (36). However, since RTF2 KO results in lower levels of pSTAT1 and, hence, lower levels of ISG upregulation, the depletion of RTF2 is likely to reduce important direct antiviral effectors (such as IFITM3) and, thus, increase the efficiency of multiple steps in the viral life cycle, including primary transcription. Because we demonstrated increased VSV infection in RTF2-KO cells, we hypothesize that RTF2 would restrict multiple viruses through its regulation of the IFN response.
Although RTF2's mechanism of action has yet to be worked out, our work advances the field by elucidating a novel cellular function of this understudied gene. Previously, Inagawa et al. reported that the Schizosaccharomyces pombe RTF2 homolog could be involved in site-specific replication termination by stabilizing stalled replication forks at the specific replication barrier RTS1 element (29), while Sasaki et al. reported that the Arabidopsis thaliana homolog may regulate pre-mRNA splicing in a ubiquitindependent fashion to affect embryonic development (37). The only study of RTF2 in human cells was done by Kottemann et al., reporting that RTF2 needs to be removed from stalled replication forks during induced DNA damage in order to maintain genome integrity (28). The one common thread across these studies is that RTF2 appears to affect the processing of nucleic acids. Future studies investigating whether RTF2 interacts with viral RNAs might reveal how RTF2 restricts virus infection, which may be testable in an in vitro replication assay with purified components. In addition, identifying RTF2's interaction partners, perhaps through immunoprecipitation-mass spectrometry, may also potentially shed light on the exact molecular mechanism of RTF2's antiviral function. Finally, the effect of RTF2 on DNA damage may also impact the IFN response through innate immune pathways triggered by DNA damage (38) and warrants further investigation. While progress has been made in understanding cellular antiviral defenses, the contributions of the many host antiviral factors to controlling specific viruses have yet to be determined. As each virus evolves a unique strategy to counteract host restriction factors, one may expect to find few pan-virus effectors (15,18) but, instead, may expect to find many effectors that have directly evolved to cope with different viruses. Elucidation of how RTF2 and other factors restrict viral replication will continue to provide insights into how our cells contend with the multitude of viruses.
Plasmids. The pXPR101 plasmid used to generate A549-Cas9 cells was provided by the Broad Institute Genetic Perturbation Platform. Individual sgRNAs were cloned into pLentiCRISPR-V2 (Addgene accession number 52961) for follow-up studies. For rescue experiments, the Cas9 gene in pXPR101 was replaced by codon-mutated versions of RTF2. Additional site-directed mutagenesis was performed to generate mislocalized forms of RTF2. For the ER-retention mutant, a signal peptide (MLLSVPLLLGLLGL AAAD) (39) was added to the N terminus, and a KDEL motif was added to the C terminus of the NLS mutant. For the plasma membrane-targeting mutant, a myristoylation motif (GCIKSKRKDNLNDDGVD) (40) was added to the N terminus of the NLS mutant. The plasmids used in the luciferase reporter assay were previously described (41).
Genome-wide CRISPR screen. We transduced ϳ120 million Cas9-positive A549 cells with the AVANA-4 lentiviral library (21) at a low MOI to achieve a 30% infection rate (such that most cells received only a single sgRNA) and an average of a 500-fold coverage of the library after selection. At 24 h postransduction, cells were subjected to 8 days of puromycin selection to remove nontransduced cells. On day 9 post-lentiviral transduction, ϳ300 million cells were subjected to 200-U/ml IFN-␤ treatment for 24 h and then infected with the influenza A virus PR8 at an MOI of 5 for 16 h. At 16 h postinfection, the cells were washed with PBS and stained with primary anti-influenza A virus HA antibody (catalog number AB1074; EMD Millipore) and secondary Alexa Fluor 488-conjugated anti-goat IgG (catalog number A11055; Thermo Fisher), before fixation with 4% formaldehyde. HA-positive and HA-negative cells were sorted by fluorescence-activated cells sorting (FACS) and harvested for genomic DNA using a Qiagen blood and tissue extraction kit according to the manufacturer's protocol. PCR of guide DNA (gDNA) was performed in 100-l reaction mixtures to attach the sequencing adaptors and barcode samples. Each reaction mixture consisted of 50 l of gDNA plus water, 40 l of PCR master mix, and 10 l of a uniquely barcoded P7 primer (stock at a 5 M concentration). The master mix was comprised of 75 l Ex Taq DNA polymerase (Clontech), 1,000 l of 10ϫ Ex Taq buffer, 800 l of the deoxynucleoside triphosphates provided with the enzyme, 50 l of P5 stagger primer mix (stock at a 100 M concentration), and 2,075 l water. PCR cycling conditions were an initial 1 min at 95°C, followed by 30 s at 94°C, 30 s at 52.5°C, and 30 s at 72°C for 28 cycles and a final 10-min extension at 72°C. Samples were purified with Agencourt AMPure XP SPRI beads according to the manufacturer's instructions (catalog number A63880; Beckman Coulter) and sequenced on a HiSeq2000 sequencer (Illumina). The screen results were analyzed with the CRISPR gene scoring tool developed at the Broad Institute, using both the negative binomial distribution (STARS [21] software, version 1.1) and the hypergeometric distribution.
Virus infection. A549 cells were inoculated with 150 l (12-well plate) or 2 ml (T75 flask) of influenza A virus or vesicular stomatitis virus at an MOI of 5 for 1 h at 37°C in serum-free DMEM. The cells were then rinsed once and replaced with fresh serum-free DMEM supplemented with 1% BSA for the length of the respective experiments. Infection was subsequently monitored by FACS or quantitative reverse transcription (RT)-PCR. For experiments involving CHX treatment, cells were exposed to either dimethyl sulfoxide (DMSO) or 50-g/ml CHX for 2 h toward the end of the 24-h IFN pretreatment, prior to infection; CHX was retained in the infection and cell culture medium until the cells were harvested. For IFN dependence experiments, cells were pretreated with 0.5-g/ml IFNAR decoy B18R or 5 M JAK inhibitor ruxolitinib for 2 h prior to IFN-␤ treatment, and the respective blockers were left in the medium throughout the duration of the subsequent 24-h IFN treatment. For the RNA stability experiment, 50 nM baloxavir was added to the cells at 6 hpi.
Rescuing RTF2-KO cells. To rescue RTF2 expression in RTF2-KO cells, an XPR101_rescue plasmid expressing a FLAG-tagged codon-mutated version of RTF2 was used. Cells were selected with 1-g/l puromycin and 10-g/l blasticidin for 8 days. Expression of rescued RTF2 was confirmed via Western blotting.
Flow cytometry. Trypsin was used to lift the adherent cells, before neutralization with DMEM supplemented with 10% FBS. The cells were then rinsed once with cold phosphate-buffered saline (PBS)-1% BSA before being stained with anti-influenza A virus HA (catalog number AB1074; EMD Millipore) in PBS-1% BSA (1:200) for 30 min on ice, rinsed with cold PBS twice, and stained with the secondary antibody, Alexa Fluor 488-conjugated donkey anti-goat IgG (catalog number A11055; Thermo Fisher), at a 1:500 dilution for 30 min on ice. After staining, the cells were rinsed in PBS twice, fixed with regular 6-well plates in DMEM. Cells were imaged on a wide-field epifluorescence microscope (Applied Scientific Instrumentation), with the images acquired using Micro-Manager software. The acquisition settings and exposure times were kept consistent across experiments.

Luciferase reporter assay for influenza A virus replication.
To measure viral polymerase activity, we utilized a vRNA-luciferase reporter system. Briefly, A549 cells were transfected with a vRNA reporter plasmid expressing firefly luciferase under a viral untranslated region. The cells were also transfected with influenza A virus PA, PB1, PB2, and NP and Renilla luciferase plasmids. At 24 h posttransfection, the cells were lysed and mixed with Dual Glo substrate (Promega) according to the manufacturer's protocol. Luminescence was measured and quantified using a Synergy H1 multimode microplate reader (BioTek).
Biochemical fractionation. The cell fractionation protocol was adapted from the Ficoll-digitonin protocol previously described (44). Briefly, 1.5 ϫ 10 6 trypsinized WT A549 cells were washed with PBS twice and resuspended in 600 l of HEPES-sucrose-Ficoll (HSF) solution (20 mM HEPES-KOH, 6.25% Ficoll, 0.27 M sucrose, 3 mM CaCl 2 , 2 mM MgCl 2 ) with 50-g/ml digitonin and EDTA-free protease inhibitor cocktail (Roche). The cells were kept on ice for 10 min, before they were spun down at 1,000 ϫ g for 3 min. The supernatant was collected and then further centrifuged at 15,000 ϫ g for 10 min to generate the cytosol fraction. The nucleus pellet from the first centrifugation was rinsed with HSF buffer once and spun down again at 1,000 ϫ g for 3 min. The supernatant was collected as the washed fraction, and the pellet was then lysed in 50 l of RIPA buffer with EDTA-free protease inhibitor cocktail (Roche) on ice for 20 min. The lysed nuclei were then centrifuged at 10,000 ϫ g for 10 min, before the supernatant was collected as the nuclear fraction for Western blot analysis. The antibodies used for this experiment included mouse anti-alpha-tubulin antibody (clone DM1A; catalog number 14-4502-82; Thermo Fisher), anti-TATA-binding protein (catalog number 8515s; Cell Signaling Technology), and mouse anti-RTFDC1 antibody (clone 1E8; catalog number LS-C340588; LifeSpan Biosciences).
Statistics and schematics. Data were tested for statistical significance with GraphPad Prism software. A t test, one-way analysis of variance (ANOVA), and two-way ANOVA were performed with their respective multiple comparisons, as indicated in the figure legends. All data are represented as the mean Ϯ standard deviation. The images in the schematics were created using BioRender software.
Data availability. The RNA sequencing data are available in the Gene Expression Omnibus (GEO) database under accession number GSE146403.