key: cord-1008882-y1h35av7
authors: Rebendenne, Antoine; Roy, Priyanka; Bonaventure, Boris; Chaves, Valadão Ana Luiza; Desmarets, Lowiese; Rouillé, Yves; Tauziet, Marine; Arnaud-Arnould, Mary; Giovannini, Donatella; Lee, Yenarae; DeWeirdt, Peter; Hegde, Mudra; Garcia de, Gracia Francisco; McKellar, Joe; Wencker, Mélanie; Dubuisson, Jean; Belouzard, Sandrine; Moncorgé, Olivier; Doench, John G.; Goujon, Caroline
title: Bidirectional genome-wide CRISPR screens reveal host factors regulating SARS-CoV-2, MERS-CoV and seasonal HCoVs
date: 2021-05-27
journal: Res Sq
DOI: 10.21203/rs.3.rs-555275/v1
sha: b8673dabca944ef29a81219e9dab0311fc775927
doc_id: 1008882
cord_uid: y1h35av7

Several genome-wide CRISPR knockout screens have been conducted to identify host factors regulating SARS-CoV-2 replication, but the models used have often relied on overexpression of ACE2 receptor. Additionally, such screens have yet to identify the protease TMPRSS2, known to be important for viral entry at the plasma membrane. Here, we conducted a meta-analysis of these screens and showed a high level of cell-type specificity of the identified hits, arguing for the necessity of additional models to uncover the full landscape of SARS-CoV-2 host factors. We performed genome-wide knockout and activation CRISPR screens in Calu-3 lung epithelial cells, as well as knockout screens in Caco-2 intestinal cells. In addition to identifying ACE2 and TMPRSS2 as top hits, our study reveals a series of so far unidentified and critical host-dependency factors, including the Adaptins AP1G1 and AP1B1 and the flippase ATP8B1. Moreover, new anti-SARS-CoV-2 proteins with potent activity, including several membrane-associated Mucins, IL6R, and CD44 were identified. We further observed that these genes mostly acted at the critical step of viral entry, with the notable exception of ATP8B1, the knockout of which prevented late stages of viral replication. Exploring the pro- and anti-viral breadth of these genes using highly pathogenic MERS-CoV, seasonal HCoV-NL63 and −229E and influenza A orthomyxovirus, we reveal that some genes such as AP1G1 and ATP8B1 are general coronavirus cofactors. In contrast, Mucins recapitulated their known role as a general antiviral defense mechanism. These results demonstrate the value of considering multiple cell models and perturbational modalities for understanding SARS-CoV-2 replication and provide a list of potential new targets for therapeutic interventions.

7 d. Venn diagram comparing hits across screens conducted in Vero E6, A549, and Huh7 (or derivatives) 157 cells (ectopically expressing ACE2 and TMPRSS2 or not) 23-28 . The top 20 genes from each cell line are 158 included, with genes considered a hit in another cell line if the average z-score was > 3.

The additional, recently published genome-wide screens for SARS-CoV-2 host factors 161 have varied in the viral isolate, the CRISPR library, and the cell type ( Table 1) [23] [24] [25] [26] [27] [28] . We 162 acquired the read counts from all these screens and re-processed the data via the same 163 analysis pipeline to enable fair comparisons (see Methods); top-scoring genes were 164 consistent with the analyses provided in the original publications. Two screens, using 165 different CRISPR libraries, were conducted in A549 cells engineered to express ACE2; 166 comparison of these results showed a greater number of statistically significant hits in the 167

Zhang-Brunello dataset 28 compared to the Sanjana-GeCKO dataset 24 , but results were 168 generally consistent between the two, with 10 genes shared in the top 20 (Fig. S1c) . 169

Likewise, three groups conducted survival screens in related cell systems (Fig. S1d) : 170

Huh7 cells (Daelemans-Brunello 23 ); Huh7.5 cells (Poirier-Brunello 25 ), a derivative of 171

Huh7, which have biallelic loss-of-function mutation in the DDX58/RIG-I sensor; and 172

Huh7.5.1 cells, engineered to overexpress ACE2 and TMPRSS2 (Puschnik-GeCKO 26 ). 173

All three screens identified TMEM106B as a top hit, and we observed the best pair-wise 174 correlation between the two screens that used Huh7.5 and Huh7.5.1 cells (Fig. S1d) . 175 176 We next averaged gene-level z-scores and compared results across the Vero E6, A549, 177 and Huh7 cell lines. Examining the top 20 genes from each cell line, and using a lenient 178 z-score threshold of 3 to consider a gene a hit, we generated a Venn diagram to examine 179 their overlap (Fig. 1d) . By these criteria, only ACE2 and CTSL scored in all three models, 180 and 3 additional genes overlapped in two cell lines. Examining the cell-line specific hits, 181

in Vero E6 cells we continued to observe an enrichment of BAF proteins SMARCA4 and 182 DPF2 (Wei et al., 2021) ; notably, another nBAF complex member, ARID1A, also scored 183 in A549 cells. Genes scoring uniquely in A549 cells included several COMM domain-184 containing proteins, which have been implicated in NF-kB signaling 29 . Finally, Huh7 cells 185

showed specificity for EXT1 and EXT3L, genes involved in heparin sulfate biosynthesis, 186 as well as SLC35B2, which transports PAP, a substrate for intracellular sulfation. Overall, 187 8 these analyses suggest that individual cell models are particularly suited, in as yet 188 unpredictable ways, to probe different aspects of SARS-CoV-2 host factor biology. 

CoV-2 when knocked out in Vero E6 cells for this screen and the screen conducted by Wei et al. 2021 198 (Wilen; 27 ). The gene-level z-score and -log10(FDR) were calculated after averaging across conditions (of 199 note, the FDR value for ACE2 is effectively zero but has been assigned a -log(FDR) value for plotting Whole-genome knockout and activation screens to identify genes regulating SARS-CoV-211 2 replication in Calu-3 cells  212   Calu-3 cells, a lung adenocarcinoma cell line, are a particularly attractive model for  213 exploring SARS-CoV-2 biology, as they naturally express ACE2 and TMPRSS2. 214 Furthermore, we have previously shown that Calu-3 cells behave highly similarly to 215 primary human airway epithelia when challenged with SARS-CoV-2 30 . Additionally, they 216 are suited to viability-based screens, as they are highly permissive to SARS-CoV-2 and 217

show high levels of cytopathic effects upon replication, although the slow doubling time 218 of the cells (~5-6 days) presents challenges for scale-up. 219 220

To conduct genome-wide CRISPR KO and activation screens (Fig. 2a) , Calu-3 cells were 221 stably engineered to express Cas9 or dCas9-VP64, respectively. Calu-3-Cas9 cells 222

showed >94% Cas9 activity (Fig. S2a) and Calu-3-dCas9-VP64 cells transduced to 223 express sgRNAs targeting the MX1 and IFITM3 promoters induced expression to a similar 224 magnitude as following interferon treatment (Fig. S2b-c) . The more compact Gattinara 225 library 31 was selected for the knockout screen due to the difficulty of scaling-up this cell 226 line, while the Calabrese library was used for the CRISPRa screen 32 . Cells were 227 transduced with the libraries in biological triplicates at a low MOI, selected with puromycin, 228 and 15 to 18 days post-transduction, were either harvested for subsequent genomic DNA 229 extraction or challenged with SARS-CoV-2 at MOI 0.005, which led to >90% cell death in 230 3-5 days. The surviving cells were then cultured in conditioned media, expanded and 231 harvested when cell numbers were sufficient for genomic DNA extraction (see Methods). 232

The screening samples were processed and analyzed as above. 233

The knockout screen was most powered to identify proviral factors (Fig. S2d) , and the 235 top three genes were ACE2, KMT2C and TMPRSS2 (Fig. 2b) . Importantly, the latter did 236 not score in any of the cell models discussed above; conversely, CTSL did not score in 237 this screen. Interestingly, whereas the BAF-specific ARID1A scored in Vero E6 cells and 238 A549 cells, PBAF-specific components ARID2 (rank 5) and PRBM1 (rank 7) scored as 239

top hits in Calu-3. Additional new hits include AP1G1 (rank 4), AP1B1 (rank 9), and 

We next examined the CRISPRa screen ( Fig. 2c and S2e). In contrast to the knockout 284 screen, here we were able to detect both pro-and anti-viral genes; we speculate this is 285 due to the shorter length of time in culture post-SARS-challenge for the activation screens 286

(2 weeks, compared to 4 in the knockout screens). Assuringly, the top-scoring proviral 287 (sensitization) hit was ACE2. Several solute carrier (SLC) transport channels also scored 288 on this side of the screen, including SLC6A19 (rank 8), which is a known partner of ACE2 289 33 . Furthermore, SLC6A14 (rank 2) has been implicated in cystic fibrosis progression and 290

shown to regulate the attachment of Pseudomonas to human bronchial epithelial cells 34 . 291

On the antiviral side of the screen, a top scoring hit was LY6E (rank 10), which is a known 292 restriction factor of SARS-CoV-2 35 , further validating the ability of this screening 293 technology and cellular model to identify known biology. Additionally, MUC21 (rank 1), 294 MUC4 (rank 4), and MUC1 (rank 26) all scored; Mucins are heavily glycosylated proteins 295 and have a well-established role in host defense against pathogens 36,37 ; moreover, 296 MUC4 has been recently proposed to possess a protective role against SARS-CoV-1 297 pathogenesis in a mouse model 38 . Finally, we directly compared the knockout and 298 activation screens conducted in Calu-3 cells (Fig. S2g) . The only gene that scored in both 299

the knockout and activation screen, even using a lenient z-score threshold of >3, was 300 ACE2, emphasizing that different aspects of biology are revealed by these screening 301

To expand the range of cell lines examined further, we also performed a knockout screen 303 with the Brunello library in another cell line naturally permissive to SARS-CoV-2 304 replication, the colorectal adenocarcinoma Caco-2 cell line. Here, however, the cells were 305 13 engineered to overexpress ACE2 in order to reach sufficient levels of CPE to enable 306 viability-based screening. Similar to Calu-3 cells, ACE2 and TMPRSS2 were the top 307 resistance hits (Fig. 2d, S2d and S2f), indicating that Caco-2 and Calu-3 cells, unlike 308 previously used models, rely on TMPRSS2-mediated cell entry, rather than the CTSL-309 mediated endocytic pathway, which did not score in this cell line (z-score=-0.2). 310

Assembling all the proviral genes identified across 5 cell lines, we observed a continuation 311

of the trend that screen results are largely cell line dependent (Fig. 2e) . and Caco-2 cells. We designed 2 sgRNAs to target these genes and generated polyclonal 318 knockout Calu-3 cell populations. In parallel, we generated 2 negative control cell lines 319 (coding non-targeting sgRNAs) and 2 positive control cell lines (ACE2 and TMPRSS2 320 KO). Two weeks post-transduction, knockout cell lines were challenged with SARS-CoV-321 2 bearing the mNeonGreen (mNG) reporter 39 and the percentage of infected cells was 322 scored by flow cytometry (Fig. 3a) . The knockout of about half of the selected genes 323 In parallel, we tested the impact of candidate knockout on SARS-CoV-2-induced CPEs. 331

Cells were infected with wild-type SARS-CoV-2 at an MOI of 0.005 and colored with 332 crystal violet when massive CPE was observed in the negative controls, ~5 days post-333 infection (Fig. S3a) . CPE analyses globally mirrored data obtained with mNG reporter 334 viruses, showing that the identified genes were bona fide proviral factors and not genes 335 14 the KO of which would only protect cells from virus-induced cell death. Encouragingly, 336 based on a recent scRNA-seq study 40 , the best-validated candidate genes, i.e. AP1G1, 337 AB1B1, AAGAB, KMT2C, EP300 and ATP8B1, were all well expressed in SARS-CoV-2 338 target cells from the respiratory epithelia (Fig. S3b) . Moreover, using RT-qPCR, we 339 observed that these genes were all expressed at slightly higher levels in primary human 340 airway epithelial cells (HAE) compared to Calu-3 cells (Fig. S3c) . 341

We then investigated the effect of these genes on other respiratory viruses. Noteworthy, 343 knockout had no substantial impact on the replication of a respiratory virus from another 344 family, the orthomyxovirus influenza A virus (IAV) strain A/Victoria/3/75 (H3N2) (Fig. 3b) . 345

In contrast, HCoV-NL63 replication was impacted by AP1G1, AP1B1 and EP300 KO, but 346 not by KMT2C or ATP8B1 KO (Fig. 3c) . Interestingly, seasonal HCoV-229E and highly 347 pathogenic MERS-CoV, which do not use ACE2 for viral entry but ANPEP and DPP4, 348 respectively, were also both strongly affected by AP1G1, and, to some extent, by AP1B1 349 and AAGAB KO (Fig. 3d-f) , showing a pan-coronavirus role of these genes. 

10 100

Calu-3

Vero E6

Caco-2 Screen:

10 100

HCoV-NL63 (RT-qPCR)

Relative infection efficiency

10 100

Relative infection efficiency

10 100

Relative infection efficiency donors for HAE cells), described in 30 , were analyzed by RT-qPCR using the indicated taqmans.

Next, we aimed to determine the life cycle step affected by the candidate KOs and we 389 examined the impact of the best validated candidate KO (i.e. with an effect >50% 390 decrease in mNG reporter expression, Fig. 4a ) on ACE2 global expression levels (Fig.  391   4b) . Immunoblot analysis revealed similar or higher expression levels of ACE2 in the 392 different KO cell lines in comparison to controls, with the exception of ACE2 and EP300 393 KO cells, which had decreased levels of ACE2. We then took advantage of recombinant 394

Spike Receptor Binding Domain (RBD) fused to a mouse Fc fragment, in order to stain 395 ACE2 at the cell surface (Fig. 4c ). Using this system, we did not observe a substantial 396 decrease in ACE2 at the plasma membrane, apart from ACE2 and EP300 KO cell lines, 397 as expected. 398

In order to assess the internalization efficiency of viral particles, we then incubated the 400 KO cell lines with SARS-CoV-2 at an MOI of 5 for 2h at 37°C, and treated the cells with 401

Subtilisin A in order to eliminate the cell surface-bound viruses, followed by RNA 402 extraction and RdRp RT-qPCR to measure the relative amounts of internalized viruses 403 (Fig. 4d) . This approach showed that AP1G1, AP1B1, AAGAB and EP300 impacted 404 SARS-CoV-2 internalization to at least some extent, but not ATP8B1. We then used 405 vesicular stomatitis virus (VSV) particles pseudotyped with SARS-CoV-2 Spike, bearing 406 a C-terminal deletion of 19 aminoacids (hereafter named Spike del19) as a surrogate for 407 viral entry 41,42 , in comparison to VSV-G pseudotypes (Fig. 4e) . Of note, both ACE2 and 408 TMPRSS2 knockout specifically impacted Spike del19-VSV infection, confirming that the 409 pseudotypes mimicked wild-type SARS-CoV-2 entry in Calu-3 cells. We observed that 410 Spike del19-dependent entry was affected in most cell lines in comparison to VSV-G-411 mediated entry, with, again, the notable exception of ATP8B1 KO cells, implying a later 412 role for this gene. Analysis of SARS-CoV-2 RNA replication by RdRp RT-qPCR (Fig. 4f ) 413

18 and viral production in the cell supernatants by plaque assays (Fig. 4g) perfectly mirrored 414 the data obtained using the mNG reporter virus, apart from ATP8B1 KO cells. Indeed, in 415 the latter, there was only around 50% decrease in viral RNA replication or mNG reporter 416 expression, but more than one order of magnitude decrease in viral production, 417 suggesting a late block during viral replication (Fig. 4f-g) . Importantly, highly similar 418 results were obtained with MERS-CoV for AP1G1 and AP1B1, which had an impact 419 comparable to DPP4 receptor KO on viral production (Fig. 4h) . Moreover, as observed 420

for SARS-CoV-2, ATP8B1 KO also strongly impacted infectious MERS-CoV particle 421 production/release, whereas it did not impact infection as measured by Spike or dsRNA 422 intracellular staining ( Fig. 4h and 3e-f Log 10 TCID 50 /ml

Actin (Fig. S4b) . Interestingly, all these genes were 481 expressed to higher levels in HAE than Calu-3 cells, with the exception of CD44, which 482 was less expressed. Moreover, MUC21 was upregulated upon SARS-CoV-2 replication 483 in HAE and Calu-3 cells, as well as MUC4 in the latter (Fig. S4c-d) . 484

Looking at the antiviral breadth of the validated genes, we observed that the induction of 486 most of them had no impact on IAV infection (Fig. 5b) , with the exception of MUC4 and 487 MUC1, which decreased the infection efficiency by ~60-70%, as seen previously 37 , and 488 IL6R, with one of the 2 sgRNAs leading to 75% decrease in infection efficiency. 489

Interestingly, similarly to SARS-CoV-2, HCoV-229E appeared highly sensitive to the 490 increased expression of MUCs, IL6R, LY6E, CD44, but was less affected or not affected 491 at all by the other genes, such as PLAGL1 (Fig. 5c) .

by the 3 Mucins of interest and to some extent by PLAGL1, CD44, IL6R, LY6E and 493 ATAD3B, but not by the other candidates (Fig. 5d) . 

The mean and SEM of at least 4 (a) or 3 (b, c, d) 524 525

Next, we tested the impact on SARS-CoV-2 of some of the best candidates in naturally 526

permissive Caco-2 cells and in A549 cells engineered to ectopically express ACE2 ( 

We then explored the life cycle step affected by antiviral gene expression. The SARS-545

CoV-2 internalization assay, performed as previously, showed that most of the validated 546 genes, including those showing the strongest inhibitory phenotypes (namely MUC1, 547 MUC21, CD44, PLAGL1, IL6R, MUC4, and LYN) impacted viral internalization (Fig. 6a) . 548

The measure of viral entry using Spike del19-or G-pseudotyped VSV particles globally 549 mirrored the internalization data, and showed that G-dependent entry was as sensitive as 550

Spike del19-dependent entry to the induced expression of Mucins, IL6R or LYN (Fig. 6b) . 551

However, we observed that whereas CD44 and PLAGL1 had an impact on SARS-CoV-2 552 entry as measured by our internalization assay (as well as a number of other genes such 553 as TEAD3, but with milder effects), there was no effect of these genes on Spike del19-554 VSV pseudotypes, perhaps highlighting subtle differences in the mechanism of entry by 555 the latter compared to wild-type SARS-CoV-2. Moreover, LY6E induction had no 556 measurable impact on viral entry, either using the internalization assay or the VSV 557 pseudotype assay, contrary to what was reported before 35 . Differences in the 558 experimental systems used could explain the differences observed here and would 559 require further investigation. Finally, the impact of the best candidates on SARS-CoV-2 560 and MERS-CoV replication, measured by RdRp RT-qPCR (Fig. 6c) and plaque assays 561 (Fig. 6d) for SARS-CoV-2, or TCID50 for MERS-CoV (Fig. 6e) , recapitulated what was 562 observed with SARS-CoV-2 mNG reporter (Fig. 5a) and MERS-CoV Spike intracellular 563 staining (Fig. 5d) . 564 565 Noteworthy, the 3 Mucins of interest had the strongest impact on both SARS-CoV-2 and 566 MERS-CoV production (~2 log and ~1 log decrease, respectively, as compared to the 567 controls). The activation of IL6R, CD44, PLAGL1, and LYN also had a substantial impact 568 on SARS-CoV-2 replication (~1 log decrease or more, for at least 1 out of the 2 sgRNAs) 569 but had a globally milder impact on MERS-CoV replication, with LYN having no impact at 570 all ( Fig. 6d-e) . Whereas Mucins are well-known to act as antimicrobial barriers 53,54 , the 571 role of the other potent antiviral genes, such as IL6R, CD44 or PLAGL1, in limiting SARS-572

CoV-2 entry remains to be elucidated. sgRNA-expressing cell lines were challenged with SARS-CoV-2 bearing a NLuc reporter 600 55 and the relative infection efficiency was analyzed by monitoring NLuc activity (Fig. 7a) . However, SARS-CoV-2-induced CPEs were increased in SLC6A14-induced cells 608 compared to the control, suggesting a late impact of this gene on viral replication and/or 609 an increase in cell death (Fig. 7b) . Interestingly, none of the identified proviral factors had 610 a positive impact on influenza A virus infection, with the notable exception of HNF1B, 611 which had a slight positive impact (Fig. 7c) . In contrast, all the identified proviral genes 612 28 had a positive impact on HCoV-NL63 infection (Fig. 7d) . We then studied the impact of 613 the candidates on HCoV-229E, using in parallel 2 sgRNAs targeting ANPEP as positive 614 controls (Fig. 7e) . Calu-3 cells are known to express low levels of ANPEP 56 , and, as 615 expected, ANPEP receptor induction greatly increased HCoV-229E infection in Calu-3 616 cells. Among the genes having a positive impact on SARS-CoV-2 and HCoV-NL63, only 617 TP73 induction had a positive effect on HCoV-229E infection (Fig. 7e) . In order to decipher the step(s) affected by the induction of the identified proviral genes, 636

we used SARS-CoV-2 internalization and VSV pseudotype assays (Fig. 8a-b) , as 637 previously. Using these 2 assays, we observed that induction of both HNF1B and NFE2 638 improved viral entry, but not TP73 or SLC6A19, which was surprising for the latter as it is 639 a known partner of ACE2 33 . In line with this, we observed that, despite differences in 640 ACE2 levels in the 2 negative control cell lines, induction of HNF1B and NFE2 seemed 641 to increase ACE2 expression, contrary to that of TP73 or SLC6A19 (Fig. S6a) . TP73 and 642 SLC6A19 induction, however, increased SARS-CoV-2 RdRp RNA amounts in infected 643 cells as well as infectious particle production, arguing for a post-entry impact on 644 replication ( Fig. 8c-d) . Interestingly, the pan-coronavirus cofactor TP73 (Fig. 7) was 645 particularly well expressed in ciliated cells from the respiratory epithelium, and SARS-646

CoV-2 infection in patients positively modulated its expression ( Fig. S6b; 40 ). TP73 is 647 known to be a pro-apoptotic transcription factor, inducing apoptosis upon DNA damage 648 and regulating DNA damage repair 57-59 . However, here we show that TP73 does not just 649 play a role in enhancing SARS-CoV-2-induced cell death, as its induction increases viral 650 replication and production. TP73 could be acting indirectly, through the induced 651 expression of SARS-CoV-2 cofactors. Interestingly, although expressed in a lower 652 percentage of cells as compared to TP73, HNF1B expression was also upregulated in 653 ciliated cells from COVID-19 patients compared to healthy controls 40 (Fig. S6b) . HNF1B 654 is a homeodomain containing transcription factor that regulates tissue-specific gene 655 expression positively or negatively, and HNF1B has been shown to modulate lipid 656 metabolism 60 , which might be related to its positive role on SARS-CoV-2 entry, in addition 657 to the observed increase of ACE2 expression. NFE2 is a transcription factor involved in 658 erythroid and megakaryocytic maturation and differentiation and, together with MAFK 659 (which was identified as an antiviral gene by our CRISPRa screen, Fig. 2C and 5a) , forms 660 a complex, which regulates various pathways 61 . Interestingly, genes regulated by MAFK 661 and NFE2 were both identified as differentially expressed upon SARS-CoV-1 replication 

Calu-3-Cas9 cells were stably transduced to express 2 different sgRNAs (g1, g2) per indicated gene and 682 selected for 10-15 days (parallel samples from Fig. 7-8) . The cells were lysed and expression levels of 683 ACE2 were analyzed, Actin served as a loading control. A representative immunoblot is shown. Simultaneously to our screens, similar bidirectional, genome-wide screens were 709 performed in Calu-3 cells by P. Hsu and colleagues 63 . Comparisons between our data 710 sets and theirs showed a very good overlap in the hits identified, both in the KO and 711 activation screens (Fig. S7) , with shared hits including host-dependency factors Adaptins 712 AP1G1 and AP1B1 as well as Mucins as antiviral proteins. Interestingly, ATP8B1, which 713 was identified in our Caco-2 KO screen, scored within the 25 best hits in Hsu and 

Identification of a Novel Coronavirus in Patients with Severe Acute 1161 Respiratory Syndrome

Coronavirus as a possible cause of severe acute respiratory syndrome. The 1163

Epidemiology and cause of severe acute respiratory syndrome (SARS

People's Republic of China

Isolation of a Novel Coronavirus from a Man with Pneumonia in Saudi Arabia

Clinical course and risk factors for mortality of adult inpatients with COVID-19

China: a retrospective cohort study

Human coronaviruses: what do they cause?

Evidence for Gastrointestinal Infection of SARS-CoV-2

SARS-CoV-2 Cell Entry Depends on ACE2 and TMPRSS2 and Is 1177 Blocked by a Clinically Proven Protease Inhibitor

Human coronavirus NL63 employs the severe acute respiratory 1179 syndrome coronavirus receptor for cellular entry

Angiotensin-converting enzyme 2 is a functional receptor for the SARS 1182 coronavirus

Crystal structure of NL63 respiratory coronavirus receptor-49 binding domain complexed with its human receptor

A pneumonia outbreak associated with a new coronavirus of probable bat 1187 origin

Efficient activation of the severe acute respiratory syndrome 1189 coronavirus spike protein by the transmembrane protease TMPRSS2

SARS coronavirus, but not human coronavirus NL63, utilizes cathepsin L 1192 to infect ACE2-expressing cells

Characterization of spike glycoprotein of SARS-CoV-2 on virus entry and its 1194 immune cross-reactivity with SARS-CoV

Inhibitors of cathepsin L prevent severe acute respiratory syndrome 1196 coronavirus entry

Human aminopeptidase N is a receptor for human coronavirus 229E

Dipeptidyl peptidase 4 is a functional receptor for the emerging human 1200 coronavirus-EMC

TMPRSS2 activates the human coronavirus 229E for cathepsin-1202 independent host cell entry and is expressed in viral target cells in the respiratory 1203 epithelium

The spike protein of the emerging betacoronavirus EMC uses a novel 1205 coronavirus receptor for entry, can be activated by TMPRSS2, and is targeted by 1206 neutralizing antibodies

Coronavirus biology and 1208 replication: implications for SARS-CoV-2

Lessons from the host 1211 defences of bats, a unique viral reservoir

Genome-wide CRISPR screening identifies TMEM106B as a proviral host 1213 factor for SARS-CoV-2

Identification of Required Host Factors for SARS-CoV-2 Infection in 1215 Human Cells

Genome-Scale Identification of SARS-CoV-2 and Pan-coronavirus 1217

Genetic Screens Identify Host Factors for SARS-CoV-2 and Common Cold 1219

Genome-wide CRISPR Screens Reveal Host Factors Critical for SARS-CoV-2 1221 Infection

A genome-wide CRISPR screen identifies host factors that regulate SARS-1223

CoV-2 entry

The STRING database in 2021: customizable protein-protein networks, 1225 and functional characterization of user-uploaded gene/measurement sets

SARS-CoV-2 triggers an MDA-5-dependent interferon response 1228 which is unable to control replication in lung epithelial cells

Genetic screens in isogenic mammalian cell lines without single cell 1231 cloning

Optimized libraries for CRISPR-Cas9 genetic screens with multiple 1233 modalities

ACE2 and gut amino 1235 acid transport

SLC6A14 Is a Genetic Modifier of Cystic Fibrosis That Regulates 1237

LY6E impairs coronavirus fusion and confers immune control of viral 1239 disease

Defensive Properties of Mucin 1241

Glycoproteins during Respiratory Infections-Relevance for SARS-CoV-2

The cell surface mucin MUC1 limits the severity of influenza A virus 1243 infection

Mucin 4 Protects Female Mice from Coronavirus Pathogenesis

An Infectious cDNA Clone of SARS-CoV-2

COVID-19 severity correlates with airway epithelium-immune cell 1249 interactions identified by single-cell analysis

A Vesicular Stomatitis Virus Replicon-Based Bioassay for the 1251 Rapid and Sensitive Determination of Multi-Species Type I Interferon

Measuring SARS-CoV-2 neutralizing antibody activity using pseudotyped 1253 and chimeric viruses

Critical role of type III interferon in controlling SARS-CoV-2 infection, 1255 replication and spread in primary human intestinal epithelial cells

The Role of

CD44: from adhesion molecules to signalling 1261 regulators

PLAGL1: an important player in diverse pathological processes

IL-6/IL-6 receptor system 1265 and its role in physiological and pathological conditions

YAP/TEAD3 signal mediates cardiac lineage commitment of human-induced 1268 pluripotent stem cells

Dual role of YAP and TAZ in renewal of the 1270 intestinal epithelium

Lesson on the Critical Balance of Kinase Signaling in Immunity. Front 1273 Immunol

Lyn regulates inflammatory responses in Klebsiella pneumoniae infection via the 1275 p38/NF-κB pathway

Lyn kinase regulates egress of flaviviruses in autophagosome-derived 1277 organelles

The Role of the Cell Surface Mucin MUC1 as a Barrier to Infection 1279 and Regulator of Inflammation

Mucins in the 1281 mucosal barrier to infection

A nanoluciferase SARS-CoV-2 for rapid neutralization testing and screening of 1283 anti-infective drugs for COVID-19

Infection of human alveolar macrophages by human coronavirus strain 1285 229E

p73 induction after DNA damage is 53 regulated by checkpoint kinases Chk1 and Chk2

p73 protein regulates DNA damage repair

Inhibition of hepatocyte nuclear factor 1b induces hepatic steatosis through 1291

DPP4/NOX1-mediated regulation of superoxide

Small Maf proteins (MafF, MafG, MafK): History, structure 1294 and function

Integrative 1296 Network Biology Framework Elucidates Molecular Mechanisms of SARS-CoV

Genome-wide, bidirectional CRISPR screens identify mucins as critical 1299 host factors modulating SARS-CoV-2 infection

The Role of the Clathrin Adaptor AP-1: Polarized Sorting 1302 and Beyond

The type 4 subfamily of P-type ATPases, putative 1306 aminophospholipid translocases with a role in human disease

BBA) -Molecular Basis of Disease 1741

Multilevel proteomics reveals host perturbations by SARS-CoV-2 and 1309 SARS-CoV

TMEM41B and VMP1 are scramblases and regulate the distribution of 1311 cholesterol and phosphatidylserine

Tocilizumab in Hospitalized Patients with Severe Covid-19 Pneumonia

Expression of SARS-CoV-2 entry factors in lung epithelial stem cells and its potential 1316 implications for COVID-19

Improved vectors and genome-wide libraries for 1318 CRISPR screening

Genome-scale CRISPR-Cas9 knockout screening in human cells

The interferon-inducible isoform of NCOA7 inhibits endosome-mediated viral 1322 entry

Growth of human hepatoma 1324 cells lines with differentiated functions in chemically defined medium

Robust hepatitis C virus infection in vitro

Reverse genetics of SARS-related coronavirus using vaccinia 1329 virus-based recombination

Engineering a Replication-Competent, Propagation-Defective Middle East 1331 Respiratory Syndrome Coronavirus as a Vaccine Candidate

A SIMPLE METHOD OF ESTIMATING FIFTY PER CENT 1333 ENDPOINTS12

Generation of SARS-CoV-2 Spike Pseudotyped Virus for Viral 1335 Entry and Neutralization Assays: A 1-Week Protocol

Detection of 2019 novel coronavirus (2019-nCoV) by real-time RT-1338 PCR

Replication of human coronaviruses SARS-CoV, HCoV-NL63 and 55

HCoV-229E is inhibited by the drug FK506

Diagnosis value of SARS-CoV-2 antigen/antibody combined testing 1342 using rapid diagnostic tests at hospital admission

Interestingly, we observed that most of the identified genes impacted the early phases of 728 the replicative cycle. This observation was true for both the host dependency factors and 729 the antiviral inhibitors, presumably emphasizing the fact that viral entry is the most critical 730 step of the viral life cycle and probably, as such, the most easily targeted by natural 731 defenses. Among the host-dependency factors essential for viral entry, the Adaptin 732 AP1G1 and, to a lower extent, Adaptin AP1B1 and their partner AAGAB, surprisingly 733 played a crucial role. The AP-1 complex regulates polarized sorting at the trans-Golgi 734 network and/or at the recycling endosomes, and may play an indirect role in apical sorting 735 64 . Interestingly, AAGAB has been shown to bind to and stabilize AP1G1, and in AAGAB 736 KO cells, AP1G1 is known to be less abundant 65 , which may suggest a role of AAGAB 737 via the regulation of AP-1 complex here. Our data showed that the KO of AP1G1, AP1B1 738 or AAGAB impacted SARS-CoV-2 entry, while not affecting ACE2 expression at the cell 739surface. In line with this observation, the KO of these factors also impacted MERS-CoV 740and HCoV-229E, which use different receptors. However, all these coronaviruses use 741 TMPRSS2 for Spike priming in Calu-3 cells, therefore a possible explanation could be 742 that the AP-1 complex might be important for surface expression of TMPRSS2 (a 743 hypothesis that we have so far been unable to test, due to the lack of specific TMPRSS2 744 antibodies). Alternatively, the AP-1 Adaptins might be important for the proper localization 745 of other plasma membrane components, which play a role in SARS-CoV-2 attachment 746 and/or entry. 747 748 Our analysis revealed that another cofactor affecting viral entry, EP300, which is a histone 749 acetyltransferase, was most likely having an indirect effect on SARS-CoV-2 replication, 750 by regulating ACE2 expression. The fact that EP300 impacted HCoV-NL63 but not HCoV-751 229E or MERS-CoV reinforced this hypothesis. This was also true for two proviral factors 752 identified through our CRISPRa screens, HNF1B and NFE2. In contrast, proviral factor 753 TP73 had no effect on ACE2 expression or viral entry, and actually impacted the 4 754 coronaviruses we tested here, suggesting the potential regulation of pan-coronavirus 755 factor(s) by this transcription factor. 756 757 34 An exception among the proviral genes that we characterized was ATP8B1, the only one 758 acting at a late stage of the viral life cycle. ATP8B1 belongs to the P4-Type subfamily of 759ATPases (P4-ATPases) transporters, which are flippases translocating phospholipids 760 from the outer to the inner leaflet of membrane bilayers 66 . ATP8B1 has been shown to 761 be essential for proper apical membrane structure and mutations of this gene have been 762 linked to cholestasis. The fact that ATP8B1 was important for both SARS-CoV-2 and 763 MERS-CoV replication highlighted a potentially conserved role for coronaviruses and it 764 would be of high interest to understand the underlying molecular mechanisms. 765Interestingly, ATP8B1 and its homologous ATP8B2 were recently identified as binding-766 partners of SARS-CoV-2 ORF3 and M, respectively 67 , suggesting that the virus might 767 subvert their functions. Of note, TMEM41B, an integral protein of the endoplasmic 768 reticulum known to regulate the formation of autophagosomes, lipid droplets and 769 lipoproteins, was recently shown to be both an essential coronavirus cofactor 25 and a 770 phospholipid scramblase whose deficiency impaired the normal cellular distribution of 771 cholesterol and phosphatidylserine 68 . Whether ATP8B1 could play a similar role in 772 coronavirus replication remains to be determined. 773

Among the best antivirals we identified through our CRISPRa screens, the well-known 775 antimicrobial defenses, membrane-associated Mucins played a broad and potent role at 776 limiting coronavirus entry. Interestingly, these Mucins were upregulated in COVID-19 777 patients 40 . Additionally, we showed that induced expression of two other membrane 778 proteins, CD44 and IL6R, could also limit SARS-CoV-2 viral entry. Both these proteins 779 are classically seen as important players during immune responses, being involved 780 mainly in adhesion/trafficking and pro-inflammatory processes, respectively. Interestingly, 781 CD44 has also been demonstrated to serve as a platform that brings other membrane 782 receptors together with actin cytoskeleton, possibly within lipid rafts 44 

The lentiviral vector expressing ACE2 (pRRL.sin.cPPT.SFFV/ACE2, Addgene 145842) 806has been described 30 . The pLX_311-Cas9 (Addgene 96924) and pXPR_BRD109, which 807 express Cas9 and dCas9-VP64, respectively, have been described 32 . LentiGuide-Puro 808 vector was a gift from Feng Zhang 71,72 (Addgene 52963) and we have described before 809 For CRISPR-Cas9-mediated gene disruption, Calu-3, Caco-2 and A549-ACE2, cells 831 stably expressing Cas9 or dCas9-VP64 were first generated by transduction with LX_311-832 37 Cas9 or XPR_BRD109, respectively, followed by blasticidin selection at 10 µg/ml. WT 833Cas9 activity was checked using the XPR_047 assay (a gift from David Root, Addgene 834 107145) and was always >80-90%. dCas9-VP64 activity was checked using the 835 pXPR_502 vector expressing sgRNA targeting IFITM3 and MX1 ISG promoters. Cells 836 were transduced with guide RNA expressing LentiGuide-Puro or XPR_502 (as indicated) 837and selected with antibiotics for at least 10 days. Coulter, A63880). Prior to sequencing the sample was quantitated by qPCR and diluted 894 to 2nM. 5 µL of the sample was then further diluted and denatured with 5 µL 0.1N NaOH 895 and 490 µL HT1 buffer (Illumina). Samples were sequenced on a HiSeq2500 HighOutput 896 (Illumina) with a 5% spike-in of PhiX. 897 898

For each published screen, corresponding authors provided raw read counts. For the 900 screens conducted in this paper, guide-level read counts were retrieved from sequencing 901 data.We log-normalized read counts using the following formula: We calculated residuals from this spline and z-scored these values at the guide-level 915 (anchors package). We calculated gene-level z-scores by averaging across guides and 916 conditions, and p-values were combined across conditions using Fisher's method. Genes 917 were filtered by number of guides per gene, which was generally one guide fewer or 918 greater than the median number of genes per gene for that library (e.g. for Brunello 919 screens, which has a median of 4 guides per gene, we applied a filter of 3 to 5 guides per 920 gene). This guide-filtering step accounts for any missing values in the file compiling data 921 40 across all screens (all_screens_v3.xlsx). We then used these filtered gene-level z-scores 922 to rank the genes such that the rank one gene corresponded to the top resistance hit. The