Identification of DAXX as a restriction factor of SARS-CoV-2 through a CRISPR/Cas9 screen

Interferon restricts SARS-CoV-2 replication in cell culture, but only a handful of Interferon Stimulated Genes with antiviral activity against SARS-CoV-2 have been identified. Here, we describe a functional CRISPR/Cas9 screen aiming at identifying SARS-CoV-2 restriction factors. We identify DAXX, a scaffold protein residing in PML nuclear bodies known to limit the replication of DNA viruses and retroviruses, as a potent inhibitor of SARS-CoV-2 and SARS-CoV replication in human cells. Basal expression of DAXX is sufficient to limit the replication of SARS-CoV-2, and DAXX over-expression further restricts infection. DAXX restricts an early, post-entry step of the SARS-CoV-2 life cycle. DAXX-mediated restriction of SARS-CoV-2 is independent of the SUMOylation pathway but dependent on its D/E domain, also necessary for its protein-folding activity. SARS-CoV-2 infection triggers the re-localization of DAXX to cytoplasmic sites and promotes its degradation. Mechanistically, this process is mediated by the viral papain-like protease (PLpro) and the proteasome. Together, these results demonstrate that DAXX restricts SARS-CoV-2, which in turn has evolved a mechanism to counteract its action. Here, Mac Kain and Maarifi et al. perform a functional CRISPR/Cas9 screen to identify SARS-CoV-2 restriction factors in A549 cells. They identify DAXX, a scaffold protein of nuclear bodies with diverse functions, that has anti-viral activity post SARS-CoV-2 entry, while SARS-CoV-2 has evolved a mechanism to counteract its action via PLpro-mediated proteasomal degradation.

S evere Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) is the causative agent of COVID-19 and the third coronavirus to cause severe disease in humans after the emergence of SARS-CoV in 2002 and Middle East Respiratory Syndrome-related Coronavirus (MERS-CoV) in 2012. Since the beginning of the pandemic, SARS-CoV-2 has infected more than 500 million people and claimed at least 6 million lives. While the majority of infected individuals experience mild (or no) symptoms, severe forms of COVID-19 are associated with respiratory failure, shock and pneumonia. Innate immune responses play a key role in COVID-19 pathogenesis: immune exhaustion 1 and reduced levels of type-I and type-III interferons (IFN) have been observed in the plasma of severe COVID-19 patients 2,3 . Imbalanced immune responses to SARS-CoV-2, with a low and delayed IFN response coupled to early and elevated levels of inflammation, have been proposed to be a major driver of COVID-19 4,5 . Neutralizing auto-antibodies against type-I IFN 6 and genetic alterations in several IFN pathway genes 7 have also been detected in critically ill COVID-19 patients. These studies highlight the crucial need to characterize the molecular mechanisms by which IFN effectors may succeed, or fail, to control SARS-CoV-2 infection.
Although SARS-CoV-2 has been described to antagonize the IFN pathway by different mechanisms involving the viral proteins ORF3b, ORF9b ORF6, and Nsp15 8 , detection of SARS-CoV-2 by the innate immune sensor MDA5 9,10 leads to the synthesis of IFN and expression of IFN Stimulated Genes (ISGs) in human airway epithelial cells 4 . IFN strongly inhibits SARS-CoV-2 replication when added in cell culture prior to infection 11,12 or when administered intranasally in hamsters 13 , suggesting that some ISGs might have antiviral activity 14 . Relatively few ISGs with antiviral activity against SARS-CoV-2, however, have been identified so far. For instance, spike-mediated viral entry and fusion is restricted by LY6E 15,16 and IFITMs 17,18 . Mucins have also been suggested to restrict viral entry 19 . ZAP, which targets CpG dinucleotides in RNA viruses, also restricts SARS-CoV-2, albeit moderately 20 . OAS1 has been recently identified in an ISG overexpression screen to restrict SARS-CoV-2 replication, through the action of RNAseL, both in cell lines and in patients 21 . Another overexpression screen identified 65 ISGs as potential inhibitors of SARS-CoV-2 22 , and found that BST-2 is able to restrict viral budding, although this activity is counteracted by the viral protein ORF7a. We hypothesize that additional ISGs with antiviral activity against SARS-CoV-2 remain to be discovered. Other antiviral factors that are not induced by IFN may also inhibit SARS-CoV-2: for instance, the RNA helicase DDX42 restricts several RNA viruses, including SARS-CoV-2 23 . While several whole-genome CRISPR/Cas9 screens identified host factors required for SARS-CoV-2 replication [24][25][26][27][28][29] , none focused on antiviral genes.
In this work, we performed a CRISPR/Cas9 screen designed to identify restriction factors for SARS-CoV-2, assessing the ability of 1905 ISGs to modulate SARS-CoV-2 replication in human epithelial lung cells. We report that the Death domain-associated protein 6 (DAXX), a scaffold protein residing in PML nuclear bodies 30 and restricting DNA viruses 31 as well as retroviruses 32,33 , is a potent inhibitor of SARS-CoV-2 replication. SARS-CoV-2 restriction by DAXX is largely independent of the action of IFN, and unlike most of its other known activities, of the SUMOylation pathway. Within hours of infection, DAXX relocalizes to sites of viral replication in the cytoplasm, targeting an early, post-entry step of the viral life cycle such as viral transcription or uncoating. We show that the SARS-CoV-2 papainlike protease (PLpro) induces the proteasomal degradation of DAXX, demonstrating that SARS-CoV-2 developed a mechanism to evade, at least partially, the restriction imposed by DAXX.

Results
A restriction factor-focused CRISPR/Cas9 screen identifies genes potentially involved in SARS-CoV-2 inhibition. To identify restriction factors limiting SARS-CoV-2 replication, we generated a pool of A549-ACE2 cells knocked-out (KO) for 1905 potential ISGs, using the sgRNA library we previously developed to screen HIV-1 restriction factors 34 . This library includes more ISGs than most published libraries, as the inclusion criteria was less stringent (fold-change in gene expression in THP1 cells, primary CD4 + T cells or PBMCs ≥ 2). Therefore, some genes present in the library may not be ISGs per se in A549 cells. Transduced cells were selected by puromycin treatment, treated with IFNα and infected with SARS-CoV-2. Infected cells were immuno-labelled with a spike (S)-specific antibody and analyzed by flow cytometry. As expected 11,12 , IFNα inhibited infection by sevenfold (Fig. S1). Infected cells were sorted based on S expression (Fig. 1a), and DNA was extracted from infected and non-infected control cells. Integrated sgRNA sequences in each cell fraction were amplified by PCR and sequenced by Next Generation Sequencing (NGS). Statistical analyses using the MAGeCK package 35 led to the identification of sgRNAs significantly enriched or depleted in infected cells representing antiviral and proviral genes, respectively (Fig. 1b). Although our screen was not designed to study proviral genes, we did successfully identify the well-described SARS-CoV-2 co-factor cathepsin L (CTSL) 36 , validating our approach. USP18, a gene encoding a negative regulator of the IFN signaling pathway 37 , and ISG15, which favors Hepatitis C Virus replication 38 , were also identified as proviral ISGs. Core IFN pathway genes such as those encoding for the IFN receptor (IFNAR1), STAT1, and STAT2, were detected as antiviral factors, further validating our screening strategy. LY6E, a gene previously described to encode an inhibitor of SARS-CoV-2 entry 15,16 , was also a significant hit. Moreover, our screen identified APOL6, IFI6, DAXX and HERC5, genes that are known to encode proteins with antiviral activity against other viruses [39][40][41][42] , but had not previously been studied in the context of SARS-CoV-2 infection. For all these genes except APOL6, individual sgRNAs were consistently enriched (for antiviral factors) or depleted (for proviral factors) in the sorted population of infected cells, while non-targeting sgRNAs were not (Fig. 1c).

LY6E and DAXX display antiviral activity against SARS-CoV-2.
To validate the ability of the identified hits to modulate SARS-CoV-2 replication in human cells, we generated pools of A549-ACE2 knocked-out (KO) cells for different genes of interest by electroporating a mix of 3 sgRNA/Cas9 ribonucleoprotein (RNP) complexes per gene target. Levels of gene editing were above 80% in all of the A549-ACE2 KO cell lines, as assessed by sequencing of the edited loci (Table 1). As controls, we used cells KO for IFNAR1, for the proviral gene CTSL or for the antiviral gene LY6E, as well as cells electroporated with non-targeting sgRNAs/Cas9 RNPs (indicated here as WT). These different cell lines were then treated with IFNα and infected with SARS-CoV-2. Viral replication was assessed by measuring the levels of viral RNA in the supernatant of infected cells using RT-qPCR (Fig. 2a). In parallel, we titrated the levels of infectious viral particles released into the supernatant of infected cells (Fig. 2b). As expected, infection was significantly reduced in CTSL KO cells, confirming the proviral effect of this gene 36 . Among the selected antiviral candidate genes, only 2 had a significant impact on SARS-CoV-2 replication: LY6E (as expected), and to an even greater degree, DAXX. Both genes restricted replication in absence of IFNα, an effect which was detectable at the level of viral RNA (8-fold and 42-fold reduction of infection, respectively, Fig. 2a) and of infectious virus (15-fold and 62-fold reduction, Fig. 2b). Based on available single-cell RNAseq datasets, 43 DAXX is expected to be expressed in cell types physiologically relevant for SARS-CoV-2 infection such as lung epithelial cells and macrophages (Fig. S2).
In IFNα-treated cells, DAXX and LY6E KO led to a modest, but significant rescue of viral replication, which was particularly visible when measuring the levels of infectious virus by plaque assay titration (Fig. 2b), while the antiviral effect of IFNα treatment was completely abrogated in IFNAR1 KO cells, as expected (Fig. 2c). However, IFNα still had robust antiviral effect on SARS-CoV-2 replication in both DAXX KO and LY6E KO cells (Fig. 2c). This suggests that other ISGs likely contribute to the anti-SARS-CoV-2 IFN response. DAXX is sometimes referred to as an ISG, and was originally included in our ISG library, although its expression is only weakly induced by IFN in some human cell types 32,44 . Consistent with this, we found little to no increase in DAXX gene expression in IFNα-treated A549-ACE2 cells (Fig. S3). In addition, we tested the antiviral effect of DAXX on several SARS-CoV-2 variants that have been suggested to be partially resistant to the antiviral effect of IFN in A549-ACE2 cells 45 . Our results confirmed that Lineage B.1.1.7. (Alpha) and Lineage P1 (Gamma) SARS-CoV-2 variants were indeed less sensitive to IFN (Fig. 2d). DAXX, however, restricted all variants to a similar level than the original Lineage B strain of SARS-CoV-2 ( Fig. 2d), suggesting that while some variants may have evolved towards IFN-resistance, they are still efficiently restricted by DAXX. To determine whether DAXX restriction is specific to SARS-CoV-2 or also inhibits other RNA viruses, including coronaviruses, A549-ACE2 WT and DAXX KO cells were infected with SARS-CoV, MERS-CoV, and 2 RNA viruses belonging to unrelated families: Yellow Fever Virus (YFV) and Measles Virus (MeV), which are positive and negative strand RNA viruses, respectively. DAXX restricted SARS-CoV, but had no effect on the replication of YFV, MeV or MERS-CoV (Fig. 2e, f). Thus, our data suggests that DAXX restriction may exhibit some level of specificity. Non-infected, non-IFNα treated cells were harvested as a control. DNA was extracted from both cellular fractions and sgRNA loci amplification was carried out by PCR. Following NGS, bio-informatic analysis using the MAGeCK package was conducted. This figure was created with BioRender.com. b Screen results. By taking into account the enrichment ratios of each of the 8 different sgRNAs for every gene, the MAGeCK analysis provides a positive score for KO enriched in infected cells (i.e. restriction factor, represented in the top fraction of the graph) and a negative score for KO depleted in infected cells (i.e. proviral factors, represented in the bottom portion of the graph). Genes with an FDR < 0.05 are represented in black. 3 genes with a FDR > 0.05, but with a p value < 0.005 were additionally selected and are represented in red. c Individual sgRNA enrichment. For the indicated genes, the enrichment ratio of the 8 sgRNAs present in the library was calculated as the MAGeCK normalized read counts in infected cells divided by those in the original pool of cells and is represented in log2 fold change. As a control, the enrichment ratio of the 200 non-targeting control (NTCs) is also represented.
DAXX targets an early post-entry step. Next, we investigated which steps of the SARS-CoV-2 replication cycle were targeted by DAXX. To assess whether DAXX affects SARS-CoV-2 Spikemediated entry, we took advantage of a replication-competent Vesicular Stomatitis Virus expressing GFP (VSV*) and carrying the Spike protein instead of its G envelope (VSV*ΔG-S). This approach allows to study Spike-mediated viral entry without relying on lentiviral pseudotypes, which are likely to be targeted by DAXX 33 . We first ensured that DAXX was not affecting VSV replication in A549-ACE2 cells using the VSV* control virus (Fig. 3a). Flow cytometric analysis revealed that VSV*ΔG-S replicated at similar levels in WT and DAXX-KO cells (Fig. 3a), suggesting that DAXX does not inhibit the entry steps that are mediated by SARS-CoV-2 Spike.
To determine whether DAXX targets viral transcription, A549-ACE2 WT or DAXX KO cells were infected with SARS-CoV-2, and the intracellular levels of viral RNA were assessed at different time post-infection (Fig. 3c, d). At early time points (from 2 h to 6 h p.i.), the levels of viral RNA transcripts were similar in WT and DAXX KO cells, further suggesting that comparable amounts of SARS-CoV-2 RNA were entering cells in both cell lines. The levels of viral transcripts significantly increased starting at 8 h p.i., representing the initiation of viral transcription. The levels of viral RNA as detected by amplification of the 5ʹ UTR (Fig. 3c) were 6.4fold higher at 8 h; 4.1-fold higher at 10 h; and eightfold higher at 24 h post-infection in DAXX KO cells compared to WT cells. We observed a similar effect when using primers amplifying the RdRp region ( Fig. 3d) with levels of viral transcripts 1.7-fold and 3.5-fold A549-ACE2 knocked-out for the indicated genes were generated using a multi-guide approach, leading to pools of KO cells with a high frequency of indels. KO cells were pre-treated with 0 (circles) or 200 (triangles) U/mL of IFNα 24 h prior to infection with SARS-CoV-2 (at an MOI of 0.1). Supernatants were harvested at 72 h p.i. The mean ± SD of three independent experiments, each performed in three biological replicates, is shown. a For the titration of RNA levels, supernatants were heat-inactivated prior to quantification by qRT-PCR. Genome copies/mL were calculated by performing serial dilutions of a synthetic RNA with a known concentration. Statistics: 2-way ANOVA using Dunnett's test. Significant p values (below 0.05) are indicated on the graph. b For the titration of infectious virus levels by plaque assay, supernatants were serially diluted and used to infect VeroE6 cells. Plaques formed after 3 days of infection were quantified using crystal violet coloration. Together, these results suggest that while DAXX has no effect on viral entry, it restricts a post-entry step of the viral life cycle such as viral transcription or uncoating.

DAXX restriction is mediated by its D/E domain but is
SUMOylation-independent. DAXX encodes a small scaffold protein that acts by recruiting other SUMOylated proteins in nuclear bodies through its C-terminal SUMO-Interacting Motif (SIM) domain 46 . The recruitment of these factors is required for the effect of DAXX on various cellular processes such as transcription and apoptosis, and on its antiviral activities 32,[47][48][49] . DAXX can also be SUMOylated itself 50 , which may be important for some of its functions. To investigate the role of SUMOylation in DAXX-mediated SARS-CoV-2 restriction, overexpression assays using WT and two previously described mutated versions of DAXX 51 were performed (Fig. 4a). Fifteen lysine residues have been mutated to arginine in the first mutant (DAXX 15KR), which is unable to be SUMOylated. The second mutant is a truncated version of DAXX that lacks its C-terminal SIM domain (DAXXΔSIM) 48 and is thus unable to interact with its SUMOylated partners. A549-ACE2 were refractory to SARS-CoV-2 infection upon transfection with any plasmid, precluding us from using this cell line. The experiments were performed in 293T-ACE2 cells, which are permissive to SARS-CoV-2 18 and easy to transfect. In order to quantify the antiviral effect of overexpressed DAXX WT and mutants, we assessed the number of transfected cells that were positive for the Spike protein by flow cytometry. Western blot (Fig. S4a) and flow cytometry ( Fig. S4b) analyses showed that DAXX WT and mutants were expressed at similar levels, with around 40 to 50% cells expressing the HA-tagged constructs. DAXX WT, 15KR and ΔSIM efficiently restricted SARS-CoV-2 replication. Indeed, at 24 h p.i., the proportion of infected cells (among HA-positive cells) was reduced by 2 to 3-fold as compared to control transfected cells (Fig. 4b). This effect was less pronounced but still significant at 48 h p.i. (Fig. 4c). Moreover, DAXX overexpression led to a significant reduction of the levels of intracellular viral RNA (Fig. S5), in line with our earlier results showing that DAXX targets viral transcription (Fig. 3c, d). Together, these results show that DAXX overexpression restricts SARS-CoV-2 replication in a SUMOylation-independent mechanism.
DAXX was recently described as a protein chaperone able to solubilize protein aggregates and unfold misfolded proteins 52 . We investigated whether this activity was required for SARS-CoV-2 restriction using a DAXX mutant lacking the D/E domain, critical for this chaperone activity 52 . The DAXXΔD/E mutant, while expressed at similar levels than WT DAXX in transfected cells (Fig. S4c), was unable to restrict SARS-CoV-2, as assessed by Western Blot analysis on Spike levels (Fig. 4d) and RT-qPCR analysis detecting intracellular SARS-CoV-2 RNA (Fig. 4e). Taken together, these results suggest that DAXX dampens SARS-CoV-2 replication through a SUMOylation-independent mechanism that likely involves its chaperone activity. SARS-CoV-2 infection triggers DAXX re-localization. DAXX mostly localizes in PML nuclear bodies 30 , whereas SARS-CoV-2 replication occurs in the cytoplasm. We reasoned that DAXX may re-localize during the course of infection in order to exert its antiviral effect. We first examined the effect of DAXX overexpression on the replication of SARS-CoV-2-mNeonGreen 53 by microscopy. DAXX overexpression starkly reduced the number of infected cells (Fig. 5a, b), confirming our flow cytometry data (Fig. 4). Using double staining for HA-tagged DAXX and SARS-CoV-2, we found that most of the DAXX-transfected cells were negative for infection, and conversely, that most of the infected cells did not express transfected DAXX (Fig. 5c), confirming that DAXX imposes a major block to SARS-CoV-2 infection. Next, we infected 293T-ACE2 cells with SARS-CoV-2 and used highresolution confocal microscopy to study the localization of endogenous DAXX (Fig. 5d). As expected 30 , DAXX localized in discrete nuclear foci in non-infected cells. Strikingly, SARS-CoV-2 infection induced the re-localization of DAXX in the cytoplasm, as early as 6 h post-infection, although some nuclear foci were still detected. At 24 h post-infection however, DAXX was completely absent from nuclear bodies, and was found almost exclusively in the cytoplasm, in close association with dsRNAs, likely representing SARS-CoV-2 replication sites. These results suggest that early events following SARS-CoV-2 infection trigger the cytoplasmic translocation of DAXX.
SARS-CoV-2 PLpro induces proteasomal degradation of DAXX. Next, we asked whether this relocalization of DAXX following infection destabilizes the protein. Western blot analysis revealed that SARS-CoV-2 infection induces a marked decrease of total DAXX expression levels in infected cells (Fig. 6a). In contrast, SARS-CoV-2 infection had no effect on DAXX mRNA levels (Fig. S6). Importantly, the decrease in DAXX protein levels is likely not attributed to a global host protein expression shut down, as the levels of Lamin B, HSP90, Actin, GAPDH, Tubulin, TRIM22 and RIG-I were unchanged upon infection (Fig. 6a). The fifteen lysine residues of DAXX 15KR have been mutated to arginine. DAXXΔSIM lacks the 732-740 C-terminal residues. Both mutants were described in. 48 DAXXΔD/E is lacking its 414-505 domain and has been described in 52  These results suggest that DAXX may be specifically targeted by SARS-CoV-2 for degradation. SARS-CoV-2 papain-like protease (PLpro) is a possible candidate for this activity, as it cleaves other cellular proteins such as ISG15 54,55 , and ULK1 56 . Moreover, PLpro of foot-and-mouth disease virus (FDMV) degrades DAXX 57 . To investigate this possibility, we treated cells with GRL0617, an inhibitor of SARS-CoV-2 PLpro 55 ; MG132, a welldescribed proteasome inhibitor; or Masitinib, an inhibitor of SARS-CoV-2 3CL protease 58 . These inhibitors had minimal effects on cell viability at the selected concentrations (Fig. S7). Strikingly, GRL0617 treatment partially restored DAXX expression (Fig. 6b), especially at the highest concentration. Similarly, MG132 also prevented DAXX degradation in SARS-CoV-2 infected cells. In contrast, Masitinib treatment had no effect on DAXX levels. These results suggest that PLpro, but not 3CL, targets DAXX for proteasomal degradation. Consistently, GRL0617 treatment also restored DAXX subcellular localization to nuclear bodies (Fig. 6c). As expected, GRL0617 treatment also inhibited the production of SARS-CoV-2 proteins, such as Spike (Fig. 6b), and may thus have an indirect effect on DAXX levels by inhibiting SARS-CoV-2 replication itself. However, the fact that Masitinib also inhibits Spike production but does not restore DAXX expression suggested that DAXX degradation is not an unspecific consequence of a reduced viral replication but rather a specific activity of PLpro. To investigate further the direct contribution of PLpro to DAXX degradation, we assessed the impact of overexpressing a panel of individual SARS-CoV-2 proteins in 293T-ACE2 cells on DAXX levels. We included in the analysis mCherry-tagged SARS-CoV-2 Non-structural proteins (Nsp 59 , which were not expressed from a lentiviral vector that may be targeted by DAXX antiviral activity 33 . This included Nsp3 (which encodes PLro), Nsp4, Nsp6, Nsp7, Nsp10, Nsp13 and Nsp14. All proteins were expressed at similar levels (Fig. S8a). Only the overexpression of Nsp3 led to DAXX reduced expression (Fig. 6d,  Fig. S8b). This effect was dose-dependent (Fig. 6e, Fig. S8c, d), and was abrogated when cells were treated with GRL0617 (Fig. 6f,  Fig. S8e). Taken together, these results indicate that PLpro directly induces the proteasomal degradation of DAXX.

Discussion
The whole-genome CRISPR/Cas9 screens conducted to date on SARS-CoV-2 infected cells mostly identified host factors necessary for viral replication [24][25][26][27][28][29] and did not focus on antiviral genes, as did our screen. Three overexpression screens, however, identified ISGs with antiviral activity against SARS-CoV-2 16,21,22 . In the first one, Pfaender et al. screened 386 ISGs for their antiviral activity against the endemic human coronavirus 229E, and identified LY6E as a restriction factor inhibiting both 229E and SARS-CoV-2. Our screen also identified LY6E as a top hit (Fig. 1), further validating the findings of both studies. Four additional genes had significant p-values in both Pfaender et al. and our work: IFI6, HERC5, OAS2 and SPSB1 ( Supplementary Data 1 and  2). We showed that knocking-out LY6E and DAXX only partially rescued SARS-CoV-2 replication in IFN-treated cells (Fig. 2), suggesting that other IFN effectors active against SARS-CoV-2 remain to be identified. For instance, other proteins, such as IFITMs, inhibit SARS-CoV-2 viral entry [17][18][19] . In the second screen, Martin Sancho et al. tested 399 ISGs against SARS-CoV-2. Among the 65 antiviral ISGs identified, they focused on BST-2, which encodes a protein targeting viral budding. BST-2 was not a significant hit in our screen ( Supplementary Data 1 and 2). This discrepancy is likely due to the fact that our screen relies on the sorting of S-positive cells, and is therefore unable to detect factors restricting the late stages of the viral replication cycle. The most recent overexpression screen assessed the contribution of 539 human and 444 macaque ISGs in SARS-CoV-2 restriction, and further characterized the role of OAS1 in sensing SARS-CoV-2 and restricting its replication through RNASEL. While we did not identify OAS1 or RNASEL in our screen (Supplementary Data 1 and 2), we did identify hits in common with this screen, including IFI6 and OAS2 (that were also identified by Pfaender et al.). Of note, DAXX was absent from the ISG libraries used by these overexpression screens, which explains why it was not previously identified as an antiviral gene for SARS-CoV-2. Our sgRNA library, by including 1905 genes, targeted a wider set of ISGs and "ISG-like" genes, including genes like DAXX that are not (or only weakly) induced by IFN in some cell types 32,44 . One potential caveat to our screen is that it compared IFN-treated infected cells to non-infected untreated cells (rather than IFN-treated noninfected cells). Although we validated this approach in previous studies 34,60 , it may cause enrichment of genes induced by IFN, but not antiviral against SARS-CoV-2 per se. Interestingly, IFN has a stronger effect on DAXX gene expression levels in cells from other mammals such as bats. 61 Future studies may investigate whether DAXX orthologs of different species are also able to restrict SARS-CoV-2 and whether DAXX participates in IFNmediated viral restriction in these species.
We identify DAXX as a potent antiviral factor restricting the replication of SARS-CoV-2, acting independently of IFN (Fig. 3). DAXX fulfills all of the criteria defining a bona fide SARS-CoV-2 restriction factor: knocking-out endogenous DAXX leads to enhanced viral replication (Fig. 2), while over-expression of DAXX restricts infection (Figs. [4][5]. While DAXX had no effect on Spike-mediated viral entry (Fig. 3), it led to a stark reduction in the levels of viral transcripts at 8 h post-infection, in the levels of Spike protein at 16 h post-infection (Fig. 3). This suggests that DAXX blocks a post-entry step of the viral life cycle such as viral transcription. DAXX co-localizes with viral replication sites (Fig. 5) and SARS-CoV-2 antagonizes DAXX to some extent, as evidenced by the proteasomal degradation of DAXX induced by PLpro (Fig. 6). Although DAXX expression is not upregulated by IFNα in A549 cells (Fig. S3), basal levels of expression are sufficient for its antiviral activity, as has been shown for other potent restriction factors. Publicly available single-cell RNAseq analyses (Fig. S2) indicated that DAXX is expressed in cell types targeted by the virus in vivo, such as lung epithelial cells and macrophages. Interestingly, DAXX exhibited some degree of specificity in its antiviral activity, as unrelated viruses such as YFV and MeV, as well as the closely related MERS-CoV were not sensitive to its action, in contrast to SARS-CoV and SARS-CoV-2 (Fig. 2). Future work will determine which viral determinants are responsible for the specific antiviral activity of DAXX.
DAXX is mostly known for its antiviral activity against DNA viruses replicating in the nucleus, such as adenovirus 5 (AdV5) 62 and human papillomavirus (HPV) 63 . Most of these viruses antagonize PML and/or DAXX, which interacts with PML in nuclear bodies 30 . We show here that DAXX is also able to restrict SARS-CoV-2, a positive sense RNA virus that replicates in the cytoplasm. Recent studies have shown that DAXX inhibits the reverse transcription of HIV-1 in the cytoplasm 32,33 . Within hours of infection, DAXX subcellular localization was altered, with DAXX accumulating in the cytoplasm and colocalizing with incoming HIV-1 capsids 33 . Here, we observed a similar phenomenon, with a rapid re-localization of DAXX from the nucleus to cytoplasmic viral replication sites (Fig. 5), where it likely exerts its antiviral effect. Early events in the replication cycle of both HIV-1 and SARS-CoV-2, such as fusion between viral and cellular membranes, or virus-induced stress, may thus trigger DAXX relocalization to the cytoplasm. DAXX seems to inhibit SARS-CoV-2 by a distinct mechanism than HIV-1: whereas the recruitment of interaction partners through the SIM-domain is required for the effect of DAXX on HIV-1 reverse transcription 32 , it was not the case in the context of SARS-CoV-2 restriction. This result was unexpected, since DAXX has no enzymatic activity and rather acts as a scaffold protein recruiting SUMOylated partners through its SIM domain 51 . Other DAXX functions, such as interaction with the chromatin remodeler ATRX 30 or its role as a chaperone protein 52 are, however, SIM-independent. This last activity was recently shown 52 to be dependent on the D/E domain (amino acids 414 to 505). We found that DAXX anti-SARS-CoV-2 activity also depends on this domain (Fig. 4), suggesting that it restricts SARS-CoV-2 replication through an original mechanism involving protein refolding. Future work will determine whether DAXX binds and refolds viral proteins to hamper viral replication, or acts through binding and folding another host factor.
Our results suggest that SARS-CoV-2 evolved a mechanism to antagonize DAXX restriction, with PLpro inducing its degradation via the proteasome (Fig. 6). This antagonism, however, is not complete, since knocking-out DAXX expression enhanced SARS-CoV-2 replication (Fig. 2). Another possibility is that DAXX, by acting early in the viral life cycle (i.e. as soon as 8 h p.i., Fig. 3) exert its antiviral effect before the expression of PLpro. Proteins expressed by other viruses are also able to degrade DAXX: for instance, the AdV5 viral factor E1B-55K targets DAXX for proteasomal degradation 62 , and FDMV PLpro cleaves DAXX 57 . We showed in Fig. 2 that several SARS-CoV-2 variants and SARS-CoV were sensitive to DAXX, but MERS-CoV was not. Thus, it will be interesting to test whether PLpro from these different coronaviruses differ in their ability to degrade DAXX, and whether this has an impact on their sensitivity to DAXX restriction. Future research may also establish whether PLpro induces the degradation of DAXX through direct cleavage, or whether it acts in a more indirect way, such as cleaving or recruiting cellular cofactors. Such investigations may be relevant for the development of PLpro inhibitors 64 : indeed, in addition to directly blocking SARS-CoV-2 replication, PLpro inhibitors may also sensitize the virus to existing antiviral mechanisms such as DAXX restriction.
Generation of multi-guide gene knockout cells. 3 sgRNAs per gene were designed (Supplementary Table 2). 10 pmol of NLS-Sp.Cas9-NLS (SpCas9) nuclease (Aldevron #9212) was combined with 30 pmol total synthetic sgRNA (10 pmol for each sgRNA) (Synthego) to form RNPs in 20 µL total volume with SE Buffer (Lonza #V5SC-1002). The reaction was incubated at room temperature for 10 min. In total, 2 × 10 5 cells per condition were pelleted by centrifugation at 100 × g for 3 min, resuspended in SE buffer and diluted to 2 × 10 4 cells/µL. 5 µL of cell solution was added to the pre-formed RNP solution and gently mixed. Nucleofections were performed on a Lonza HT 384-well nucleofector system (Lonza #AAU-1001) using program CM-120. Immediately following nucleofection, each reaction was transferred to a 96-well plate containing 200 µL of DMEM 10% FBS (5 × 10 4 cells per well). Two days post-nucleofection, DNA was extracted using DNA QuickExtract (Lucigen #QE09050). Cells were lysed in 50 µL of QuickExtract solution and incubated at 68°C for 15 min followed by 95°C for 10 min. Amplicons were generated by PCR amplification using NEBNext polymerase (NEB #M0541) or AmpliTaq Gold 360 polymerase (ThermoFisher #4398881) and the primers indicated in Supplementary Table 3. PCR products were cleaned-up and analyzed by Sanger sequencing. Sanger data files and sgRNA target sequences were input into Inference of CRISPR Edits (ICE) analysis https://ice.synthego.com/#/ to determine editing efficiency and to quantify generated indels. 73 Percentage of alleles edited is shown in Table 1 (n = 3).
Hit validation. In total, 2.5 × 10 4 A549-ACE2 KO cells were seeded in 96-well plates 18 h before the experiment. Cells were treated with IFNα and infected as described above. At 72 h post-infection, supernatants and cellular monolayers were harvested in order to perform qRT-PCR and plaque assay titration. Infectious supernatants were heat-inactivated at 80°C for 10 min. For intracellular RNA, cells were lysed in a mixture of Trizol Reagent (Invitrogen #15596018) and PBS at a ratio of 3:1. Total RNA was extracted using the Direct-zol 96 RNA kit (Zymo Research #R2056) or the Direct-zol RNA Miniprep kit (Zymo Research #R2050). For SARS-CoV-2 detection, qRT-PCR was performed either directly on the inactivated supernatants or on extracted RNA using the Luna Universal One-Step RT-qPCR Kit (NEB #E3005E) in a QuantStudio 6 thermocycler (Applied Biosystems) or in a StepOne Plus thermocycler (Applied Biosystems). The primers used are described in Supplementary Table 4. Cycling conditions were the following: 10 min at 55°C, 1 min at 95°C and 40 cycles of 95°C for 10 s and 60°C for 1 min. Results are expressed as genome copies/mL as the standard curve was performed by diluting a commercially available synthetic RNA with a known concentration (EURM-019, JRC). For SARS-CoV and MERS-CoV, qRT-PCR were performed using FAM-labelled probes (Eurogentec) and the Superscript III Platinum One- Step qRT-PCR System (Thermo Fisher Scientific, #11732020). The cycling conditions were the following: 20 min at 55°C, 3 min at 95°C and 50 cycles of 95°C for 15 s and 58°C for 30 s. The primers used are described in Supplementary Table 4 SARS-CoV-2, SARS-CoV and MERS-CoV infection assays. A549-ACE2 cells were infected by incubating the virus for 1 h with the cells maintained in DMEM supplemented with 1 μg/ml TPCK-trypsin (Sigma #4370285). The viral input was then removed and cells were kept in DMEM supplemented with 2% FBS. For 293T-ACE2 cells, infections were performed without TPCK-trypsin. MERS-CoV and SARS-CoV infections were performed in DMEM supplemented with 2% FBS and cells were incubated 1 h at 37°C 5% CO2. Viral inoculum was then removed and replaced by fresh DMEM supplemented with 2% FBS. All experiments involving infectious material were performed in Biosafety Level 3 facilities in compliance with Institut Pasteur's guidelines and procedures. When indicated, SARS-CoV-2 infected cells were stained for intracellular Spike levels as described below.
Yellow fever virus and measles virus infection assays. Cells were infected with YFV (at an MOI of 0.3) or MeV-GFP (at an MOI of 0.2) in DMEM without FBS for 2 h in small volume of medium to enhance contacts with the inoculum and the cells. After 2 h, the viral inoculum was replaced with fresh DMEM 10% FBS 1% P/ S. FACS analysis were performed at 24 h p.i. Cells were fixed and permeabilized using BD Cytofix/Cytoperm (Fisher Scientific, # 15747847) for 30 min on ice (all the following steps were performed on ice and centrifuged at 4°C) and then washed tree times with wash buffer. Cells infected with YFV were incubated with the pan-flavivirus anti-Env 4G2 antibody for 1 h at 4°C and then with Alexa 488 anti-mouse IgG secondary antibodies (Thermo Fisher, #A28175) for 45 min at 4°C in the dark. Non-infected, antibody-stained samples served as controls for signal background. The number of cells infected with MeV-GFP were assessed with the GFP signal, using non-infected cells as controls. Data were acquired with an Attune NxT Acoustic Focusing Cytometer (Life technologies) and analyzed using FlowJo software.
Entry assays. Cells were seeded at 1 × 10 5 cells per well in 24-well plates in DMEM with 1% FBS. The next day, cells were infected with VSV* (MOI 0.001) or VSV*ΔG-S (MOI 7) in DMEM without FBS. The virus suspension was removed after 2 h and replaced with DMEM with 1% FBS. 16 h p.i., cells were washed once with PBS, trypsinized and subsequently fixed in 4% PFA. Fixed cells were washed once with PBS and analyzed by flow cytometry. The percentage of infected cells was identified based on GFP expression.
Overexpression assay. In total, 2 × 10 5 293T-ACE2 cells were seeded in a 24-well plate 18 h before the experiment. Cells were transfected with 500 ng of plasmids expressing HA-DAXX WT, HA-DAXX 15KR and HA-DAXXΔSIM plasmids, using Fugene 6 (Promega # E2691), following the manufacturer's instructions. HA-NBR1 was used as negative control. After 24 h cells were infected at the indicated MOI in DMEM 2% FBS. When indicated, cells were treated with 10 mM of remdesivir (MedChemExpress #HY-104077) at the time of infection. For flow cytometry analysis, cells were fixed with 4% formaldehyde and permeabilized in a PBS 1% BSA 0.025% saponin solution for 30 min prior to staining with corresponding antibodies for 1 h at 4°C diluted in the permeabilization solution. Samples were acquired on a BD LSR Fortessa and analyzed using FlowJo. Total RNA was extracted using a RNeasy Mini kit and submitted to DNase treatment (Qiagen). RNA concentration and purity were evaluated by spectrophotometry (NanoDrop 2000c, ThermoFisher). In addition, 500 ng of RNA were reverse transcribed with both oligo dT and random primers, using a PrimeScript RT Reagent Kit (Takara Bio) in a 10 mL reaction. Real-time PCR reactions were performed in duplicate using Takyon ROX SYBR MasterMix blue dTTP (Eurogentec) on an Applied Biosystems QuantStudio 5 (ThermoFisher). Transcript levels were quantified using the following program: 3 min at 95°C followed by 35 cycles of 15 s at 95°C, 20 s at 60°C, and 20 s at 72°C. Values for each transcript were normalized to expression levels of RPL13A. The primers used are indicated in Supplementary Table 4.
Microscopy immunolabeling and imaging. 293T-ACE2 cells were cultured and infected with SARS-CoV-2 as described above. When indicated, cells were treated with 50 μM of GRL0617 (MedChemExpress #HY-117043) at the time of infection. Cultures were rinsed with PBS and fixed with 4% paraformaldehyde (electronic microscopy grade; Alfa Aesar) in PBS for 10 min at room temperature, treated with 50 mM NH4Cl for 10 min, permeabilized with 0.5% Triton X-100 for 15 min, and blocked with 0.3% BSA for 10 min. Cells were incubated with primary and secondary antibodies for 1 h and 30 min, respectively, in a moist chamber. Nuclei were labeled with Hoechst dye (Molecular Probes). Images were acquired using a LSM700 (Zeiss) confocal microscope equipped with a 63X objective or by Airyscan LSM800 (Zeiss). Image analysis and quantification was performed using ImageJ (Fiji) v2.1.
Western blot. 293T-ACE2 cells were transfected with the indicated plasmids or treated with the indicated concentrations of GRL0617; with 10 µM of Masitinib (MedChemExpress #HY-10209); or with 10 µM of MG132 (SIGMA #M7449), an inhibitor of the proteasome and infected with SARS-CoV-2. Cell lysates were prepared using RIPA lysis and extraction buffer (ThermoFisher #89901). Protein concentration was determined using Bradford quantification. Proteins were denatured using 4X Bolt LDS Sample Buffer (Invitrogen) and 10X Bolt Sample Reducing Agent (Invitrogen). 40 µg of proteins were denatured and loaded on 12% ProSieve gel and then subjected to electrophoresis. Gels were then transferred (1 h, 90 V) to Western blotting membranes, nitrocellulose (GE Healthcare #GE10600002) using Mini Trans-Blot Electrophoretic Transfer Cell (Biorad #1703930EDU). Membranes were blocked with 5% BSA in PBS (blocking buffer) and incubated with primary antibodies diluted in blocking buffer. Membranes were washed and incubated with secondary antibodies diluted in blocking buffer. Chemiluminescent acquisitions were performed on a Chemidoc TM MP Imager and analyzed using Image Lab v6 software (Bio-Rad Laboratories).
Flow cytometry. For flow cytometry analysis, all cells were fixed with 4% formaldehyde. For intracellular staining, cells were permeabilized in a PBS 1% BSA 0.025% saponin solution for 30 min prior to staining with corresponding primary antibodies for 1 h at 4°C and then secondary antibodies for 45 min at 4°C, diluted in the permeabilization solution. Acquisition was done with BD Fortessa and Attune NxT cytometers. Data was analyzed with the FlowJo software (Treestar Inc., Oregon, USA v10.8.1).