Identification of β2 microglobulin, the product of B2M gene, as a Host Factor for Vaccinia Virus Infection by Genome-Wide CRISPR genetic screens

Genome-wide genetic screens are powerful tools to identify genes that act as host factors of viruses. We have applied this technique to analyze the infection of HeLa cells by Vaccinia virus, in an attempt to find genes necessary for infection. Infection of cell populations harboring single gene inactivations resulted in no surviving cells, suggesting that no single gene knock-out was able to provide complete resistance to Vaccinia virus and thus allow cells to survive infection. In the absence of an absolute infection blockage, we explored if some gene inactivations could provide partial protection leading to a reduced probability of infection. Multiple experiments using modified screening procedures involving replication restricted viruses led to the identification of multiple genes whose inactivation potentially increase resistance to infection and therefore cell survival. As expected, significant gene hits were related to proteins known to act in virus entry, such as ITGB1 and AXL as well as genes belonging to their downstream related pathways. Additionally, we consistently found β2-microglobulin, encoded by the B2M gene, among the screening top hits, a novel finding that was further explored. Inactivation of B2M resulted in 54% and 91% reduced VV infection efficiency in HeLa and HAP1 cell lines respectively. In the absence of B2M, while virus binding to the cells was unaffected, virus internalization and early gene expression were significantly diminished. These results point to β2-microglobulin as a relevant factor in the Vaccinia virus entry process.


Introduction
Poxviruses are dsDNA virus whose replication cycle occur entirely in the cytoplasm of the host cell [1,2]. Among the Poxviruses, Vaccinia virus (VV)-which was long used as vaccine against Smallpox-is the prototype poxvirus, and is the best characterized member of the family [1].
VV entry into cells is a complex process that has been subject to numerous studies. The VV size -approximately 360 nm of diameter-has led to the assumption that its entry into the cell cannot happen through caveolin or clathrin vesicles [3]. VV attachment to the cell is thought to occur with glycosaminoglycans and laminin [4,5]. Also, several membrane proteins have been proposed to serve as VV entry receptors, co-receptors or functional signaling ligands [6][7][8], in a process that leads to VV internalization through fluid phase endocytosis [9][10][11][12][13] followed by membrane fusion. As an alternative route, direct membrane fusion at plasma membrane has also been found to be a way of VV entry to the cytoplasmic space [14][15][16][17]. Surprisingly, VV entry process varies significantly between different cell lines, virus strains and viral forms [3,16,[18][19][20]. Taken together, these observations point to the existence of several pathways leading to VV entry, which probably contribute to the broad tropism range of VV.
Genome-wide genetic screens have become an important tool to identify factors that are involved in biological processes. Several technologies, like RNA interference or insertional inactivation have been used to carry out high throughput screens. Recently, tools for the use of CRISPR/Cas gRNA libraries have allowed for a more targeted way of screening genes by inactivation, down-regulation or induction. Several high-throughput screenings using CRISPR/ Cas9 technology have been performed to target cellular factors involved in infection viruses like SARS-CoV-2, West Nile virus, Influenza A virus, Human cytomegalovirus and Zika virus among others, revealing many host cell factors involved in viral infection [21][22][23][24][25][26]. In order to unravel the complexity of virus-cell interactions in Poxvirus infection, both RNAi [27][28][29][30] and retroviral-mediated insertional mutagenesis DNA [31,32] have been performed. Overall, the most important hits from those studies include entry factors and cellular genes somehow involved in viral morphogenesis.
We have developed a genome-wide CRISPR/Cas9-based screening using a human genomewide sgRNA GeCKO library that depends on Cas9 expression and delivery of a sgRNA library into cells for target gene deletion [33,34]. The GeCKO library, designed to target all the genes of the human genome was used to obtain a pooled population of single-gene knock-outs that was then subject to Vaccinia virus infection. Through this approach, we have identified B2M as a pro-viral factor related to Vaccinia virus entry. Notably, B2M encodes the well characterized β 2 -microglobulin, a molecular chaperone known to form complexes with multiple partners, including HLA, HFE, FcRn, MR1 (Human Leukocyte Antigen, Hereditary hemochromatosis protein, Neonatal Fc Receptor and Major histocompatibility complex class Irelated gene protein). For those proteins, interaction with B2M is needed to reach the cell membrane [35][36][37][38]. We show herein that inactivation of B2M gene greatly impairs VV infection and viral production by affecting viral entry.

Genetic screen reveals a set of hits potentially involved in VV infection
A loss of function, genome-wide screen was carried out by infecting a pooled population of cells harboring single gene KOs. The cell population was obtained by using a lentiviral library of 123,411 sgRNAs covering 19,050 genes and 1,864 miRNAs in the human genome [33]. HeLa cells constitutively expressing Cas9 were transduced with the lentiviral-sgRNA library, and passaged for 7-14 days in selective media. Of note, inactivation of genes essential for cell survival or growth are eliminated by this procedure, and therefore the screen is directed to the part of the genome that is non-essential in cell culture, which constitutes about 90% of the cellular genes [39,40].
We initially screened for gene KOs that could render the cells resistant to VV infection ( Fig  1). However, several experiments performed by infecting with VV strain WR under various multiplicities of infection resulted in complete cell death at 48-72 h.p.i. Even low m.o.i. resulted in no surviving cells, indicating that no single gene KO was sufficient to completely block infection. This result was consistent with a possible lack of complete resistance to infection by gene inactivation, in agreement with previous results [32]. Given that the replicative strain WR was used, we reasoned that even if some degree of resistance would be present in

B2M in Vaccinia virus infection
certain cells of the pool, the progeny virus in the cultures could lead to the death of initially surviving cells because of a second round of infection. In any event, based on these results, we concluded that probably only partial resistance could be achieved by single gene inactivations.
To search for pro-viral genes whose inactivation might lead to partial resistance to infection, we adapted our screen by introducing three main modifications: 1) we aimed at using large cell populations and controlled m.o.i. to infect only a portion of the cells, 2) we performed several consecutive rounds of infection to enhance the enrichment effect and 3) we used viruses with limited replication or production of extracellular virus to prevent the death of the surviving cells by infection with progeny virus.
For the ensuing experiments, viruses with limited spread ability in the culture were used. A virus deficient in genes A27L and F13L (V-ΔA27-ΔF13), which has a defect in the production of extracellular virus particles limiting its transmission [41,42]; and a virus deficient in the D4R gene (V-ΔD4), which is blocked in DNA replication of the viral genome and therefore is unable to generate infectious progeny virus [43]. In the case of the latter, a complementing HeLa cell line constitutively expressing VV D4 protein (HeLa-D4) was engineered to allow virus expansion in a similar way to previously reported [44]. Using those two virus mutants and different m.o.i.s, a number of experiments were completed (S1 Table), in which surviving cells were expanded, and analyzed. In parallel, uninfected cells were passaged for two weeks and used as control.
High throughput sequencing results for the sgRNA region amplified by PCR were subjected to bioinformatics analysis. Our computational analysis using MaGeCK [45,46] algorithm allowed us to identify genes that were enriched with respect to the initial cell pool. Fig 2A shows the analysis of a particular experiment and S2 Table shows the results from all experiments. Applying a False Discovery Rate (FDR) cutoff < 0.05 in MaGeCK analysis and log 2-fold change (LFC) > 0, hits from each experiment were selected (Fig 2B left panel). Considering that some gene knockouts can lead to enhanced growth and cell division over time, we carried out a control experiment in which the cell pool was passaged to assess the enrichment or depletion of those gene knockouts in the absence of infection (Fig 2B right panel), which was similar to the published screen results in public databases (i.e. DepMap Portal) [47]. We also used ScreenBEAM algorithm [48] (S3 Table) obtaining similar hits as with MaGeCK. A clustering analysis was performed but we could not detect an association of the hits obtained

Functional protein networks
Gene hits from multiple experiments were analyzed using the ReactomePA software to identify regulatory pathways (Figs 3 and 4 and S4 Table). Several high significance hits (FDR < 0.05 and LFC > 0) included genes previously described to play a role in the initial steps of VV infection such as ITGB1 or AXL [6,8], as well as genes encoding proteins involved in their downstream pathways. For instance, integrin alpha subunits (ITGA3, ITGAV and ITGA9) and actin cytoskeleton remodeling proteins (ACTG1, WASF2, ARPC2, ARPC3, ARPC4. . .) are related to the ITGB1/AXL and AKT activation pathway (Fig 4). Other of our strongest hits is CMIP, a poorly studied gene that is downstream-related to ITGB1 pathway [49,50]. Although weaker hits, two annexins were also identified with ScreenBEAM algorithm (ANXA2 and ANXA8L2), as was a member of the S100 family (S100A5) known to interact with annexins [51,52]. Annexins are commonly used to block phosphatidylserine due to its specific binding, and have been shown to reduce VV infectivity when pretreating VV mature virion with Annexin 5 [12].  S1 Table). Genes with FDR < 0.05 in at least 1 experiment are included in the analysis. Enrichment results after selection are shown in the left panel as a heatmap representing the gene rank obtained with MaGeCK. For comparison, the right panel shows the log2-fold change for each gene in non-infected control experiment. Dots to the right of the red dotted line indicate enrichment in non-Other hits were related to additional pathways known to be important during infection, like the proteasome-ubiquitin pathway. KCTD5 emerged as one of the strongest hits, although it showed some enrichment in uninfected controls. Remarkably, KCTD5 protein has been proposed to be an adaptor for cullin3 ubiquitin ligases and is known to modulate the activity of Rac1 protein, a GTPase involved in VV internalization [53,54].
Interestingly, B2M (the gene encoding β 2 -microglobulin) was among the best hits. When analyzing the top 1% genes in the MaGeCK rank by Reactome Pathway Analysis [55] the B2M-related 'Class I MHC mediated antigen processing & presentation' pathway was the most prominent (Fig 3 and S4 Table). Enriched genes associated to this pathway were B2M, UBE4A, CUL3, KLHL22, PSME1, UBE2F, RNF41, PSMC3, TRIM4, PSMA8, TRIP12, ASB17. All of them except B2M belong to proteasome/ubiquitin ligases pathway, whereas B2M is more related to post antigen-processing and class I MHC loading stage.
Given that B2M likely represented a novel player in VV entry, we reviewed the screening results to check if genes related to B2M were also positive in the screen. Notably, although with lower significance, TAPBP was detected in ScreenBEAM analysis (S3 Table). Other B2M related genes coding proteins like Antigen peptide transporter 2 (TAP2), Protein disulfideisomerase A6 (PDIA6) and BiP chaperone (HSPA5) had good scores, even though their respective FDRs (MaGeCK) or ranking criteria (ScreenBEAM) were below our initial cutoff. Those infected cultures for that particular gene inactivation. For instance, CMIP KOs are depleted from the pool during passages in the absence of infection, whereas CSK KOs are enriched. https://doi.org/10.1371/journal.ppat.1010800.g002

B2M in Vaccinia virus infection
results reinforce the notion that B2M and its related pathway are bona-fide pro-viral factors as revealed by the survival screen. B2M protein is known to bind a number of proteins that include HLA molecules, HFE or FcRn. We individually tested the scores for those genes but did not find any of those significantly enriched in the screening. Notably, published results on experimental infections of B2M deficient mice reported unexpected levels of resistance to VV [56,57], phenotypes that now might be reasoned to be influenced by B2M absence.

B2M KO impairs VV infection
The screen results point at B2M as a relevant novel factor in VV infection. To further study its role in VV infection, we obtained HeLa and HAP1 KO cell lines by CRISPR/Cas9-mediated gene-inactivation. Inactivation of B2M gene was confirmed by sequencing of the mutated region in HeLa and HAP1 cells lines, and loss of expression was followed by Western blot and immunofluorescence analysis in HeLa clones (Fig 5A and 5B). Low multiplicity infections (m. o.i. = 0.05) of those clones led to a decrease in progeny virus, indicating an effect of B2M absence on VV infection ( Fig 5C). Those results were confirmed in two independent cell clones (2.18-and 1.64-fold reduction). The reduction in virus production as a result of B2M gene inactivation was also confirmed in the background of a second cell line, HAP1 (Fig 5D  and 5E).
Since ITGB1 has been described to be involved in VV entry [6], we were interested in studying whether the function of B2M and ITGB1 are overlapping. Infection of ITGB1 KO, B2M KO, and B2M/ITGB1 double KO (DKO) cell clones in the HAP1 cells at low m.o.i. with a GFP-expressing VV revealed a significant reduction associated with B2M inactivation, and a smaller effect associated to ITGB1 inactivation, as measured by both viral production and GFP expression (Fig 5D and 5E). Concurrent inactivation of B2M and ITGB1 genes did not result in enhanced reduction when compared with B2M inactivation alone. We additionally tested if plaque size or number were affected by inactivation of B2M. For this, a GFP-expressing VV was used in a standard plaque assay in the HeLa B2M KO cells. At 24 hours after infection, there was a reduction of 60% approximately in the number of plaques ( Fig 6A). In addition, plaque size was clearly diminished (Fig 6B). Overall, these results indicate that VV replication was being negatively affected by B2M inactivation.

Effect of B2M deficiency on virus entry
We next investigated the stage of the replication cycle which was being affected in B2M KO cells. We anticipated an effect on an early step in the infection cycle, since viruses restricted to the early phase of the cycle (such as the D4-defficient virus) result in complete cell death (not shown). We therefore hypothesized that a block before that stage would be required for cell survival in the screen. Consequently, we focused on the virus entry process. We used an assay based on early luciferase expression [58], where an evaluation of the viral entry can be assessed by the luciferase activity, which is expressed under an early promoter and therefore, immediately after entry. Virus entry was significantly reduced as a result of B2M inactivation, both in HeLa and HAP1 cells (Fig 6C and 6D), indicating that inhibition of infection occurs at an early stage in the virus cycle, before the start of DNA replication.
Given the reported role of ITGB1 in VV infection, it could be reasoned that B2M inactivation might result in a decrease in ITGB1 levels, which in turn would inhibit VV infection. However, an indication that this might not be the case was that inactivation of B2M was more inhibitory than that of ITGB1 in our experimental setup ( Fig 6D). Also, since B2M is required for the transport of several membrane proteins from the ER to the plasma membrane, it could be argued that ITGB1 transport could be affected in the absence of B2M. To assess this possibility, we performed immunofluorescence with anti-ITGB1 antibodies in HeLa B2M KO cells (S4A Fig). Notably, ITGB1 plasma membrane levels in B2M KO cells were similar to that in parental HeLa cells, indicating that ITGB1 was transported efficiently to the plasma membrane independently of B2M. Therefore, we concluded that B2M KO effect on VV infection is not mediated by a decrease of ITGB1 in the cell surface.
An additional screening hit, AXL, has also been implicated in VV internalization [8,12,32,59]. However, we do not consider likely that the B2M effect on virus entry is mediated by AXL absence, since HAP1 cells do not express AXL [60]. This result was confirmed by Western blot, where the expression of AXL was detected in HeLa but not HAP1 cells (S4B Fig). This result is in agreement with AXL being an enhancer, albeit not an essential factor for infection [8,61].
VV produces different virus forms that differ in the number of membranes. In contrast with the prototypical intracellular mature virus (denoted IMV or MV) both extracellular virus and virus mutants lacking A26 can enter cells by direct fusion at the cell membrane [62,63]. Notably, inhibition by B2M KO of both extracellular virus and A26 deficient virus was smaller than that of intracellular virus (Fig 6E-6G), suggesting that entry by the endosomal route was involved.

B2M inactivation does not affect VV binding but delays internalization
In order to test if virus attachment was affected in B2M KO cells we performed a binding assay in HeLa B2M KO cells. To measure binding, purified VV particles incorporating a fluorescent protein fused to a virion protein constituent (VV A4-cherry) were incubated with cells and maintained at 4˚C to avoid virus internalization, and virus particles bound to cells were quantitated by confocal microscopy (Fig 7A and 7B). Results showed no significant differences between control and B2M KO cells indicating no correlation between B2M presence and VV attachment.
We next analyzed the effect of B2M absence on the efficiency of VV internalization. For this, after the initial virus adsorption period, cells were incubated at 37˚C to allow virus internalization and therefore, internalized virus particles were quantitated at different times ( Fig  8A). We used V-A4-cherry (for total virus) and an antibody targeting the VV Intracellular Mature Virion surface protein A27 to detect those viral particles that were not internalized ( Fig 8B). Virus internalization occurred progressively over time in HeLa cells, similarly to previous reports [9]. Comparison with HeLa cells deficient in B2M showed a consistent decrease in the internalization rate in B2M KO cells, an effect that was more pronounced at the intermediate tested times ( Fig 8A). These results suggest that a delay in virus internalization exists as a result of B2M inactivation, although some virus particles are eventually able to reach the cell interior.
We also tested whether B2M deficiency inhibited virus-induced membrane fusion. To isolate this activity from the internalization process, we tested the cell-cell fusion activity of uninfected cells mediated by cell-bound virus particles ("fusion from without"). Both in HeLa and in HeLa B2M KO cells, massive cell fusion was triggered by low pH treatment, but had basal levels at neutral pH (Fig 9).
The fact that virus internalization was diminished and/or delayed in B2M-deficient cells does not preclude additional effects later in infection. To ensure that later steps in infection were normal, we inspected cells at 18 hours post-infection by transmission electron microscopy. The different VV morphogenetic forms, including immature and mature virus forms, as well as enveloped viruses were present in the cells indicating no differences in the virus morphogenetic pathways (S5 Fig).

VV colocalization with B2M
As the specific role of B2M in early VV infection is related with the early steps of infection, we tested the degree of colocalization of B2M and VV at cell membrane. Since viral entry is a sequential process in which different protein-protein interactions may occur over time [64,65], different time points were included in the experiment. After adsorption of fluorescent V-A4-cherry particles at 4˚C, cells were stained with antibodies specific for B2M and ITGB1. Confocal images of non-permeabilized cells revealed a partial colocalization of B2M and attached virus particles (Fig 10A). We found approximately 50% of VV particles colocalizing with B2M at 0, 15, 30, 45 and 60 min post infection (S6 Fig). To determine if these colocalization results were consistent, we estimated VV/ITGB1 colocalization, that has been described to occur, at least in HeLa cells at 0 min [6]. Our results showed that 38% of adsorbed VV particles colocalized with structures containing ITGB1. Also measured that 24% of B2M-positive VV titers were obtained by a standard plaque infectivity assay. Error bars indicate SD of three independent experiments. D) and E): HAP1 and derived cell lines were infected at m.o.i. = 0.05. At different time points, a fraction of supernatant was collected to determine extracellular virus production (D). Error bars indicate SD of three independent experiments. In the same cultures, fluorescence images were obtained at different times (E). p-values: ���� < 0.0001, ��� < 0.001, �� < 0.01, � < 0.05, ns > 0.05. https://doi.org/10.1371/journal.ppat.1010800.g005

B2M in Vaccinia virus infection
structures were also labeled with ITGB1, and that 20% of ITGB1-positive structures were also labeled with B2M. Those results indicate only partial colocalization of the two proteins with each other, and with the infecting VV particle, and show a substantially differential pattern for B2M and ITGB1 (Fig 10B).
Finally, we wished to determine if B2M was able to block virus infection by interacting directly with the virus particle. For that, purified VV particles were incubated with soluble

CRISPR/Cas9 genetic screen
In this work, we have used CRISPR/Cas9 technology to perform a pooled, genome-scale lossof-function screen with the aim of identifying cellular genes necessary for VV infection. From  some KOs in genes that perform functions related to the control of cell proliferation and apoptosis. This observation is common in screens of cellular genes [47]. The experimental approach to the problem of identifying cellular factors by this strategy was complicated by the considerable versatility and adaptability of VV infection. In fact, in our initial experiments, all cells showed cytopathic effect and suffered cell death over time, regardless of the multiplicity of infection used in the experiment. Failure to obtain even a single surviving cell suggests that a single KO mutation is not enough to confer complete resistance in the cells. This could be due to a number of different reasons. A first explanation might be that the virus is able to enter the cell using different, non-overlapping molecular pathways. Second, even if there is partial resistance, it may be difficult to isolate a few resistant cells within a large cell population of virus-producing cells. Indeed, since the bulk of the cells would remain susceptible they would generate high loads of progeny virus that may overcharge and kill the partially resistant cells. In this respect, it has been shown that even inactivated virus can induce cell death [66]. In similar approaches, high multiplicities of infection resulted in the death of the whole cell population, a difficulty already reported by Luteijn et al. [32].
In view of these considerations, we modified the screening protocols to exert a gradual enrichment of partially resistant cells. Those modifications included infection with viruses that have decreased spread or limited replication, and adjusting the multiplicity of infection to modulate the selective pressure. The screening design was directed at maintaining a controlled selective pressure in the form of virus-induced cell death in the cultures during extended periods. Importantly, this may favor the enrichment of cells harboring gene-KOs leading to enhanced cell growth, as we have seen over extensive passage of uninfected samples. Therefore, enrichment in our experiments may stem from increased resistance to infection or from faster cell growth. This was taken into account when analyzing the possible hits in the screening by comparing sgRNA changes in infected and control uninfected cultures (Fig 2B).
We have observed that viruses blocked in DNA replication, such as the V-ΔD4 virus mutant, efficiently promote cell death even though replication cycle is not completed. This

B2M in Vaccinia virus infection
indicates that cell survival is not possible after the early stage of virus infection is reached. Because our screening was based on cell survival, it is not surprise that it was preferentially directed to the cell entry process. Vaccinia virus and Poxviruses in general are known to have a complex entry process into host cells, where they first bind in an unspecific manner to extracellular matrix components such as glycosaminoglycans (GAGs) and laminin, followed by a more specific attachment step [4,6,65]. During this process, signal activation pathways are triggered in the cell, leading to the internalization of the virus through fluid phase uptake and -to a lesser extent-direct fusion at plasma membrane [7,12,58]. With our screen approach, we identified several potential gene candidates that may play a role in early stages of infection. Statistically significant hits like ITGB1 and AXL, as well as some of their downstream pathway proteins such as Akt-pathway-related proteins, the small GTPases-RhoA, Rac1, Cdc42-, or actin cytoskeleton remodeling proteins [6,8,[67][68][69] were obtained. Additional gene hits were also related to these pathways. For instance, c-Maf activating protein (CMIP) was one of the strongest hits and is thought to participate in several cellular processes, including integrin expression through c-Maf activation, actin cytoskeleton remodeling and the Akt route [49,50,70]. Another hit was AhR (AHR), which is known to be bound to c-Maf [71,72]. We suggest that CMIP may have emerged as a hit in our screen due to the concurrence of the different actions where this protein is involved (Akt pathway, cytoskeleton, integrin expression. . .) [50,73,74]. Because CMIP is a poorly studied gene, it will be of interest to study its potential function in VV infection. Overall, this study contributes to expand the big picture of the likely cell-VV interactions during the first steps of infection.

β 2 -microglobulin gene (B2M)
Unexpectedly, B2M gene appeared as one of the strongest hits in the screen, implying B2M protein is an important player in VV infection. B2M is a molecular chaperone that is expressed in most mammalian cells and binds several protein ligands such as MHC-I, HFE or FcRn before their transport to the cell surface. We used different approaches to determine the role of B2M in VV infection. First, we found that B2M deficiency leads to a decrease in viral plaques, and a lower viral progeny titer. Second, B2M role is relevant during early infection stages, since early gene expression was clearly diminished (Fig 6). More precisely, we demonstrated that B2M absence does not affect VV binding but it does decrease internalization (Figs 7 and 8). Although internalization differences were around 25% between B2M KO and normal cells, these differences were much greater in early expression luciferase assays (around 70%). Several explanations may account for the difference: for instance, the existence of an unproductive internalization pathway in B2M KO cells, not leading to early transcription. Also, the internalization delay and defect in B2M KO cells may be indicative of the inefficacy of VV virions finding their signaling ligands at cell surface, therefore abrogating them to the unproductive pathway above mentioned.
ITGB1, which has also been shown to be involved in VV entry, was also a significant hit in the screen. However, several results point to B2M as acting independently of ITGB1. First, the effect of the inactivation of B2M in VV infection in HAP1 cells was stronger than that observed ITGB1 KO. Furthermore, depletion of B2M in the HAP1-ITGB1-KO line also resulted in a clear reduction of VV infectivity. This result shows that in the absence of ITGB1 and AXL (the latter not being expressed in HAP1 cells), B2M is still necessary for VV infection. Those results can be explained as B2M acting as a key player in alternative virus entry pathway.
Pathway Analysis of the data from screening experiments showed that Class I MHC mediated antigen processing & presentation pathway was the most enriched one, including many proteins related to the proteasome and ubiquitin system. In this respect, a screen using Monkeypox virus also pointed to several enriched ubiquitin genes and other B2M-related genes, such as TAPBP and CALR [31]. The ubiquitin system has been extensively studied for its importance during infection, and its relationship to B2M is hence intriguing in view of our results.
At this point, there is no clear mechanism to account for B2M role during infection. B2M requirement for VV infection might be the result of a direct involvement of this protein in the entry process. However, given that B2M forms protein complexes with a number of membrane proteins, one interesting possibility is that one of those complexes might act as a VV receptor or internalization factor. Nonetheless, we have analyzed our screening results searching for these B2M related genes, and we found no clear results. Gene redundancies, in particular regarding HLA genes, may obscure the screening results, that have the limitations imposed by the fact that only single gene inactivations are tested. Current experiments are directed at exploring if any of those B2M heterocomplexes is indeed involved in virus entry.
The known function of B2M does not offer an evident link with a mechanism of virus entry. However, its function in the initial steps of VV infection is not a singular case. Intriguingly, some similarities exist with Coxsackievirus A9 (CAV9) -an enterovirus B-whose internalization is also mediated by B2M [75][76][77] and which also makes use of at least two subunits of integrin; beta 6 and beta 3. Furthermore, CAV9 entry is mediated by dynamin, as is also the case with VV [78]. Recently, it has been found that the effect of B2M for Enterovirus B entry can be explained by the need of B2M in mediating the transport to the membrane of its receptor, the Human Neonatal Fc Receptor (FcRn) [79,80]. Expanding the similarities to the case of CAV9, we obtained ASAP1 as a hit (S3 Table), a GTPase activating protein involved in Arf1, Arf5 and Arf6 activation and cytoskeleton remodeling [81], the latter of which being known to mediate endocytosis for some viruses, like CAV9 [76]. As authors suggested, virions might undergo an unproductive entry pathway in B2M absence [76], a hypothesis that is consistent with our results. Due to the several similarities in cellular requirements of CAV9 and VV entry, it is likely that one of the entry mechanisms for VV is shared by CAV9. In any event, more research is needed to further clarify the details of the process.
We have shown that B2M deficiency does not affect cell-cell fusion mediated by virus particles, an event that is triggered at low pH (Fig 9). Also, we found that the effect is predominantly directed at intracellular virus forms, that are known to require low pH for membrane fusion. In contrast, extracellular virus and A26-deficient virus, which can induce membrane fusion at neutral pH and can enter cells directly at the cell membrane, were not affected by B2M inactivation (Fig 6). Those observations are consistent with a requirement of B2M in the internalization of incoming virus particles. The possible involvement of B2M in activating the signals leading to macropinocytosis will require further investigation.
One last interesting aspect of B2M involvement in virus entry is the relevance of B2M product (β-2 Microglobulin) in the immune response. The involvement of B2M as a proviral factor might have multiple consequences regarding the pathogenesis and immune response in animal infection. Interestingly, two reports in the literature found that B2Mdefficient mice were remarkably resistant to VV infection, despite the lack of cell-mediated immunity in those animals as a consequence of MHC-I deficiency [56,57]. This result suggests the B2M deficiency has also a VV blocking effect, even though those mice have a T cell immunodeficiency.
Overall, this report adds a new player to our picture of the complex interplay between cell and virus in the VV entry process. Delineation of the protein interactions and pathway activations taking place in this process will undoubtedly involve multiple players yet to be discovered.

Viruses
Vaccinia virus (VV) Western Reserve (WR) strain was obtained from the American Type Culture Collection (ATCC VR-119) and routinely propagated in BSC-1 cells. Previously described viruses V-GFP (recombinant WR expressing GFP) [82], and V-ΔA27ΔF13 (recombinant having the A27L and F13L genes deleted) are already described [41].
Virus infections were performed in media containing 2% FBS. Viral titers in cell lysates obtained after disrupting the cells by three rounds of freeze/thawing were determined by the standard VV plaque assay in BSC-1 cells [83].

Generation of recombinant and mutant Vaccinia viruses
Recombinant virus V-A4-cherry was generated as previously described [78]. Briefly, BSC-1 cells (6-well plate format, 10 6 cells/well), were infected at a multiplicity of infection (m.o.i.) of 0.05 PFU/cell with virus WR. After 1 h viral adsorption, virus inoculum was removed and cells were then transfected with 2 μg of plasmid pm-cherry-A4L (kindly provided by Wen Chang) using FuGeneHD (Promega) according to the manufacturer's recommendations. After 2 to 3 days, cells were harvested, and recombinant virus was isolated by three consecutive rounds of plaque purification. Viral stocks were generated in BSC-1 cells and viral titers were determined by standard plaque assay (PFU/mL). V-ΔD4 virus was isolated following the protocol already described [43,44] by infection/ transfection of BSC-1 cells with WR virus and plasmid pΔD4R-dsRed. pΔD4R-dsRed contains the dsRed gene downstream of a synthetic early/late promoter between two DNA segments acting as recombination flanks for the D4L locus. At 72h the culture was collected and several successive rounds of plaque purifications in BSC-1 cells were performed by selecting large red plaques (representing single recombinants). The resulting virus was amplified in the complementing cell line expressing the D4R gene (HeLa-D4R). The amplified stock was plaque purified in the HeLa-D4R line. The size phenotype for each clone was checked by plaquing in parallel in the BSC-1 line and the HeLa-D4L line. Virus clones that produced large plaques in the HeLa-D4 cell line and single infected cells in BSC-1 were selected.
V-e.Luc VV recombinant expressing firefly luciferase (Renilla reniformis) under the control of an early promoter was constructed by inserting an early-Luc cassette in plasmid pA-E-Luc into the A27L locus using a plaque selection procedure [41]. Recombinant VV expressing nanoLuc (V-e-nLuc) from an early promoter has been previously described [41]. An A26-deficient VV expressing nanoLuc was isolated as follows. Plasmid pA-E-nLuc, in which the flanking sequences at the A27L locus correspond to 400 nucleotides at the end of the A25L gene and 400 nucleotides at the beginning of the A26L gene [41], was digested with BamHI and NheI enzymes and subsequently religated to obtain plasmid pA-E-nLuc-A26-. By this procedure, approximately 200 nucleotides were removed from the A26L sequence. to obtain the was used into the A27L locus using a plaque selection procedure [41].

Cell-Cell fusion assay
Cell fusion was performed by mixing HeLa cells stably expressing fluorescent proteins obtained by lentiviral transduction. To obtain HeLa cell lines stably expressing GFP, lentiviral stocks were derived from plasmid pRRL.sin18.CMV-eGFP, derived from the pRRLsin18. CMVeGFPWpre series [84][85][86] was kindly provided by Juan Carlos Ramirez.
To obtain a lentiviral plasmid for mScarlet fluorescent protein expression, the Scarlet coding sequence was PCR amplified from plasmid pRB-mScarlet [87] using oligonucleotides len-tiScarlet5': CCGTCTAGAGGATCCACCATGCGTACGAAGGGCGAGGAGGCAGTGAT-CAAGGAG and LentiScarlet3': CGCGAATTCTTCGAAACTAATTGTACAGCTCGTCCATGCCGCCGCCGGT, subsequently digested with XbaI and EcoRI and inserted into the backbone of lentiviral plasmid LentiCas9-Blast, digested with XbaI and EcoRI to remove the coding sequence of the Cas9-Blasticidin cassette.

B2M in Vaccinia virus infection
cell clones were visible at certain dilutions from which lentiviral titers were estimated. Cells transduced with an m.o.i. of 3-5 were cloned by limit dilution in MW96 wells and individual clones were harvested and expanded for further analysis. Cas9 expression was verified by immunofluorescence and western blotting using anti-flag antibody. A cell clone with the highest Cas9 expression was termed HeLa-Cas-ML, expanded and used for subsequent experiments.
HeLa and B2M KO cell lines stably expressing GFP and Scarlet were obtained by transduction with lentiviral stocks obtained from pRRL.sin18.CMV-eGFP or pLenti-Scarlet with titers of 5x10 6

CRISPR Knock-Out (GeCKO) v2.0 libraries: lentivirus stock and derived cell lines
The Human GeCKO v2.0 library, a gift from Feng Zhang, in the two lentiviral vector version (Addgene 1000000049), was used [33,34]. The lentiviral library LentiGuide-puro consists of specific sgRNA sequences for gene knock-out covering the gene set of the human genome. The library is supplied as two half-libraries (A and B). When combined, the libraries contain 6 sgRNAs per gene (3 sgRNAs in each library).
HeLa-Cas-ML cells were transduced with lentiviral libraries A and B separately and selected with puromycin to obtain a cell mixture called HeLa-Mix. For each library, cells in 4 subconfluent T150 flasks were transduced with approximately 2 x 10 7 lentiviral particles. At 24 h, medium was replaced by fresh medium containing 1 μg/mL puromycin. After three days the cells were harvested, pooled, and subcultured to 8 T225 flasks. Subsequently, cells were transferred to 24 T225 flasks maintaining the selective medium at all times for 10 days and constituting the pass 0.

Genetic screen
GeCKO KO cell pools in 2 T225 flasks (about 10 8 cells) were infected in each condition. Cells were maintained in EMEM-7% FBS except during virus adsorption, which was carried out in EMEM containing 2% FBS. Uninfected control cells were maintained in culture for 8 subcultures before analysis. Detached cells from the infected cultures were removed regularly. Infection was monitored by detecting infected cells by fluorescence (red fluorescence for V-ΔD4L and blue fluorescence for V-ΔA27ΔF13). To exert selective pressure in the cultures in the form of VV infection, cells were occasionally re-infected with fresh inoculum. If cultures reached confluence, they were subcultured into new T225 flasks. To avoid population bottlenecks in the sgRNA population, cell dilutions at each passage were never greater than 4-fold. The last passage was allowed to reach confluence and the cells were harvested and prepared for sgRNA quantitation in the surviving population. The conditions under which the infection experiments were performed are summarized in S1 Table. After the selective process, cellular DNA was extracted and the lentiviral region containing the sgRNA sequence was PCR amplified and sequenced. For DNA purification cell samples were suspended in 500 μL of lysis buffer (10 mM Tris (pH 8.0), 10 mM EDTA, 10% SDS and 100 μg/mL proteinase K) and incubated 3 h at 55˚C or o/n at 50˚C. Samples were then extracted with an equal volume of phenol twice, then 3-5 times with a volume of phenol: chloroform: isoamyl alcohol (25: 24: 1) and finally once with a volume of chloroform: isoamyl alcohol (24: 1). DNA was then precipitated by adding one-tenth volume of 3 M sodium acetate and two volumes of ethanol and, cooled to -20˚C for 1 h followed by 30 min centrifugation at 4˚C and 13000 rpm. The pellet was resuspended in 300 μL TE with RNAase mix and frozen at -20˚C for storage.
Amplicon for sequencing was derived from cellular DNA by three successive nested PCRs. PCR reactions were performed with Biotools thermostable DNApol enzyme. The first PCR was performed on all the purified DNA as template, by setting up multiple reactions containing each 5 μg of DNA in 100 μL of reaction volume, with 5'P1, 3'P1 oligonucleotides and 15 amplification cycles. After the first PCR, all reactions were pooled and a second PCR was performed in four-fold less reactions with a 1:20 dilution of the first as template in 100 μl of reaction with oligonucleotides 5'P2, 3'P2, and 20 amplification cycles. The third PCR was performed similarly (1:20 template, 1:4 number of reactions, 20 cycles) to incorporate the sequences and the barcodes necessary for Illumina sequencing. Each sample carries a different combination of R and F oligonucleotides in order to multiplex the Illumina sequencing run. Sample libraries were equimolarly mixed and the resulting pools quantified by dsDNA fluorescence (Quantifluor, Promega) and qPCR (KAPA Library Quantification kit, Roche). Singleend sequencing (76 bases) was performed in a NextSeq 500 (Illumina) sequencer or in a Hiseq (Illumina).

Statistical analysis and screen data analysis
Control and infected samples read counts were analyzed with MaGeCK test function [45] and ScreenBEAM algorithm [48]. MaGeCK log2-fold change (LFC) scores of each gene were calculated by comparing the endpoint of each experiment with the initial cell population, and used for data visualization using R coding language. A cutoff of FDR < 0.05 was applied for the heatmap representation to select the best hits of each screen experiment. For ScreenBEAM analysis, the 20 best hits were selected from each experiment following z-score criteria. Genes with less than 3 successful sgRNAs as well as gene knockouts enriched in control were filtered out. ReactomePA (R package) was performed using the top 1% gene list from MaGeCK output. Statistical analyses were performed using Student's t test and one-way ANOVA (when more than two group comparisons) using Bonferroni correction method (GraphPad Prism 8 and RStudio 1.2.5033).

Isolation of KO cell clones
HeLa B2M KO lines: To obtain a B2M KO cell line, pLentiCRISPRv2 was digested with BsmBI, and a pair of annealed oligonucleotides were cloned into the single guide RNA cloning site. Two different sgRNAs for human B2M gene were selected Oligonucleotides for clone 1 were: B2M_sg1f CACCGACTCACGCTGGATAGCCTCC and B2M_sg1r AAACGGAGGCT ATCCAGCGTGAGTC. Oligonucleotides for clone 2 were: B2M_sg2f CACC. GCAGTAAGT CAACTTCAATGT and B2M_sg2r AAACACATTGAAGTTGACTTACTG. Lentiviral plasmids containing B2M sgRNA were obtained by cloning the hybridized oligonucleotide pairs B2M_sg1f/B2M_sg1r and B2M_sg2f/B2M_sg2r to obtain plasmids pLentiCRISPRv2-B2M1 and LentiCRISPRv2-B2M2 respectively. HeLa cells were transduced with lentiviral preparations of LentiCRISPRv2-B2M1 or LentiCRISPRv2-B2M2 and puromycin resistant clones were selected. After western blotting analysis, clones that did not express B2M were selected, expanded and verified by sequencing the genomic target region [88].

VV entry and blocking assays
For VV entry assays, luciferase activity driven from a VV early promoter at 3 h.p.i. was measured. Cells were plated in MW96 wells and incubated with V-e.luc (m.o.i. = 0.8) for 1 h for adsorption. Cells were washed with PBS and incubated for two additional hours at 37˚C with media free of phenol red. Cells were lysed using ONE-Glo EX Luciferase Assay System (Promega) and luciferase activity was measured using EnSight Multimode Plate Reader (PerkinElmer). Blocking assays were performed as follows: for antibody blocking assay HeLa cells were incubated with two different concentrations of anti-B2M antibody (5 and 15 μg/mL) or anticaveolin antibody (15 μg/mL) -used as negative control-1 h at room temperature. Subsequently, the cells were infected with V-e.luc (m.o.i. = 0.8) for 1 h and then washed with fresh medium. After 3 h.p.i., cells were lysed and luciferase activity was measured. For B2M blocking assay, V-e.luc inoculum was incubated with growing concentrations of soluble BSA (control) or B2M (0, 5, 15, 30 and 50 μg/mL) 1 h at room temperature. Then, HeLa cells were incubated with the pretreated inoculum for 1h and finally washed with fresh medium. At 3 h.p.i. cell extracts were prepared and subsequently luciferase activity was measured.
Cell-cell fusion assays induced by VV particles (fusion from without). Cell-cell fusion assays induced by VV particles were performed as previously described [62,89]. Briefly, HeLa cells and B2M KO cells expressing red fluorescent protein (RFP) or green fluorescent protein (GFP) were mixed at a ratio of 1:1 and seeded into 96-well plates. 18 hours later, cells were pretreated with 40 μg/ml cordycepin (Sigma) for 60 min and subsequently incubated with wild-type (WT) WR at a m.o.i. of 100 PFU per cell in triplicate. Cordycepin was maintained in the medium throughout the experiment. After 50 min virus adsorption, cells were treated with PBS at either pH 7.4 or pH 5.0 at 37˚C for 3 min, washed with growth medium, further incubated for 2h at 37˚C, fixed in 4% paraformaldehyde and incubated with a solution of Hoechst 2 mg/mL. Cell fusion was quantitated by counting more than 600 nuclei from multiple images. Percentages of cell fusion were calculated as follows: (nuclei in fused cells/ total number of nuclei) × 100. Fusion assays were repeated three times and statistical analyses were performed using Student's t test in Prism (version 8) software (GraphPad).

Immunofluorescence microscopy
Cells were seeded in round glass coverslips, washed with PBS, fixed with ice-cold 4% paraformaldehyde for 12 min and permeabilized in PBS containing 0.1% Triton X-100 for 15 min at room-temperature. Cells were treated with PBS containing 0.1 M glycine for 5 min and incubated with primary antibodies in PBS-20% FBS for 30 min, followed by incubation with secondary antibodies diluted 1:300 in PBS-20% FBS.
To detect virus particles by antibody labeling on the cell surface, immunofluorescence was carried out on non-permeabilized cells treated as follows: cells were seeded in coverslips, washed and incubated at 12˚C for 1 h with primary antibodies in PBS-20% FBS followed by incubation at 12˚C for 1 h with secondary antibodies in PBS-20% FBS. Subsequently, cells were fixed by ice-cold 4% paraformaldehyde for 12 min and treated with PBS containing 0. For DNA staining, cells in glass coverslips were incubated with 2 mg/mL bisbenzimide (Hoechst dye. Sigma) for 30 min. Cell surface staining was performed by incubation with 0.02 mg/mL wheat germ agglutinin (WGA)-Alexa488 (ThermoFisher Scientific) incubation at concentration for 30 min at 12˚C.
For internalization assay, cells were seeded in coverslips, and V-A4-cherry (m.o.i. = 4) was incubated 1 h at 4˚C to allow viral adsorption. Cells were then incubated at 37˚C for different times in a 5% CO 2 incubator to allow internalization. Cells were washed three times with PBS and incubated with antibodies following the non-permeabilization method described above.
Binding assays were based on protocols described elsewhere [6,20,90,91]. Briefly, cells grown in coverslips were infected with V-A4-cherry for 1 h at 4˚C (m.o.i. = 30), washed with ice-cold PBS and fixated with paraformaldehyde 4% for 12 min at 4˚C. Cells were treated with PBS containing 0.1 M glycine for 5 min and incubated 30 min with 0.02 mg/mL Wheat Germ Agglutinin-Alexa488 diluted in PBS-20% FBS. Cells were washed with PBS, mounted with FluorSave reagent (Millipore), observed by fluorescence microscopy and counted manually. 20-30 cells were counted, with an average of 2-3 cells per field.

Western blotting
Whole cell lysates were prepared in denaturant buffer (80 mM Tris-HCl, pH 6.

Electron microscopy
For ultrastructural analysis, HeLa and HeLa B2M KO cells grown on 12 mm glass coverslips, and infected with VV for 18h were fixed with a mixture of 2.5% glutaraldehyde and 2% paraformaldehyde in 0.1 M phosphate buffer (pH 7.4) for 1 h at RT. Cells were washed twice at 4˚C with phosphate buffer; postfixation was performed with a mixture of 1% osmium tetroxide, 1% potassium ferricyanide in water for 1 h at 4˚C and afterwards 0.15% tannic acid for 1 min at RT. Samples were treated with 1% uranyl acetate for 1 h at RT and dehydrated in increasing concentrations of ethanol (50, 75 and 95%) for 10 min each; and 100% ethanol three times for 30 min each, at 4˚C. Infiltration in epoxy-resin was done at RT using increasing concentrations of resin (50 and 100%) and polymerization was performed at 60˚C for 3 days. Ultrathin sections of the samples were stained following standard procedures with 4% aqueous uranyl acetate and 2% lead citrate. Sections were imaged on a FEI Tecnai 12 electron microscope operated at 120 kV and equipped with a FEI Ceta CCD camera.