SARS-CoV-2 escapes direct NK cell killing through Nsp1-mediated downregulation of ligands for NKG2D

Summary Natural killer (NK) cells are cytotoxic effector cells that target and lyse virally infected cells; many viruses therefore encode mechanisms to escape such NK cell killing. Here, we interrogate the ability of SARS-CoV-2 to modulate NK cell recognition and lysis of infected cells. We find that NK cells exhibit poor cytotoxic responses against SARS-CoV-2-infected targets, preferentially killing uninfected bystander cells. We demonstrate that this escape is driven by downregulation of ligands for the activating receptor NKG2D (NKG2D-L). Indeed, early in viral infection, prior to NKG2D-L downregulation, NK cells are able to target and kill infected cells; however, this ability is lost as viral proteins are expressed. Finally, we find that SARS-CoV-2 non-structural protein 1 (Nsp1) mediates downregulation of NKG2D-L and that Nsp1 alone is sufficient to confer resistance to NK cell killing. Collectively, our work demonstrates that SARS-CoV-2 evades direct NK cell cytotoxicity and describes a mechanism by which this occurs.


INTRODUCTION
Natural killer (NK) cells are innate lymphocytes that play a critical role in the immune response to viral infection. [1][2][3][4] Since the advent of the COVID-19 pandemic, studies examining the immune response in COVID-19 have noted that NK cells are less abundant in the peripheral blood of severe COVID-19 patients than in healthy donors [5][6][7][8][9][10][11][12][13] ; a concurrent increase in NK cell frequency in the lungs of critically ill patients suggests that peripheral depletion of NK cells may be due to trafficking to the site of infection. 14 In addition, immune profiling has uncovered significant, severity-associated phenotypic and transcriptional changes in the peripheral NK cells that remain in the blood of COVID-19 patients. In severe COVID-19, peripheral blood NK cells become activated and exhausted. 6,7,9,11,13,[15][16][17] They also downregulate surface level expression of the activating receptors NKG2D and DNAM-1, possibly as a consequence of internalization after ligation 7,10 and exhibit defects in their ability to respond to tumor target cells and cytokine stimulation compared with NK cells from healthy donors. 11,13,15 Less is known about how NK cells respond directly to SARS-CoV-2-infected cells, although several studies have demonstrated that NK cells can suppress SARS-CoV-2 replication in vitro. 16,18,19 Moreover, a recent study found that NK cells are able to mount robust antibody-mediated responses against SARS-CoV-2-infected target cells. 20 However, the mechanisms underlying NK cell responses to SARS-CoV-2-infected cells are not understood. This is particularly important because many viruses employ mechanisms that allow them to evade recognition and killing by NK cells. For example, both HIV-1 and human cytomegalovirus downregulate the ligands for NK cell activating receptors, shielding infected cells from recognition by NK cells. [21][22][23][24][25][26][27][28][29][30][31][32] In this study, we utilized primary NK cells from healthy donors in conjunction with replication-competent SARS-CoV-2 to create an in vitro model system that dissects the NK cell response to SARS-CoV-2-infected cells. We focused on assessing the direct killing of infected target cells to better understand how the balance between SARS-CoV-2 recognition and escape contributes to disease. Our results demonstrate that SARS-CoV-2-infected cells efficiently escape killing by healthy NK cells, likely due to downregulation of ligands for the activating receptor NKG2D. Furthermore, we interrogated the mechanisms underlying this phenomenon and identified a specific SARS-CoV-2 protein, non-structural protein 1 (Nsp1), that mediates escape from NK cell recognition. Collectively, our work deeply interrogates the NK cell response to SARS-CoV-2 and provides insight into the role of NK cells in COVID- 19.

SARS-CoV-2-infected cells evade NK cell killing through a cell-intrinsic mechanism
We established a system to explore the NK cell response to SARS-CoV-2 infection using A549-ACE2 cells, 33 which are lysed by NK cells and are infectible with SARS-CoV-2. We infected A549-ACE2 cells with SARS-CoV-2/WA1-mNeonGreen 34 (which replaces ORF7a with mNeonGreen) at a multiplicity of infection (MOI) of 0.5 ( Figure 1A). After 24 h, approximately 6% of cells fluoresced green, increasing to 50% by 48 h ( Figure 1A). This suggests that, although SARS-CoV-2 only requires $8 h to complete its life cycle, 35,36 48 h is required for detection of robust viral protein expression in a low MOI system in which viral replication results (D and E) Background-subtracted percentage of A549-ACE2 cell death as measured by eFluor 780 viability dye staining in either infected versus exposed, uninfected cells (D) or mock-infected versus exposed, uninfected cells (E). Background cell death for each experiment and condition was calculated as the average level of death in four wells of the condition of interest. Data are shown from n = 18 unique healthy donors across 4 separate experiments. Lines connect points from individual donors. (F and G) Representative flow plots (F) and quantitations (G) of percentage of NK cells expressing CD107a and IFN-g upon culture with no targets, mock-infected targets, or SARS-CoV-2-infected targets. Lines connect points from individual donors (n = 6). Significance values for all plots in this figure were determined using a paired Wilcoxon signed-rank test with the Bonferroni correction for multiple hypothesis testing. in spreading infection. To understand how exposure to SARS-CoV-2-infected target cells impacts NK cell phenotype and function, we added NK cells from healthy donors that had been preactivated overnight with IL-2 to target cells that had been infected for 48 h ( Figure 1B). This is an important distinction from previous studies that added NK cells early after SARS-CoV-2 infection, before the virus-infected cell expresses the full complement of viral proteins. 16,18,19 We then assessed the ability of NK cells to directly lyse SARS-CoV-2-infected (mNeonGreen+) target cells compared with bystander (exposed but mNeonGreenÀ) and mock-infected cells ( Figures 1C-1E).
NK cell co-culture induced significantly more death of uninfected ''bystander'' cells than of SARS-CoV-2-infected cells in all 18 NK cell donors tested ( Figure 1C). We found no significant difference in the killing of bystander cells compared with mock-infected cells that were never exposed to SARS-CoV-2, indicating that the ability of SARS-CoV-2-infected cells to survive is a cellintrinsic effect ( Figure 1D). To ensure that these differences were not a result of rapid cell death resulting in cell loss and undercounting of killed SARS-CoV-2-infected cells, we assessed the ratio of infected (mNeonGreen+) target cells to uninfected (mNeonGreenÀ) target cells in cultures without NK cells compared with cultures with NK cells, gating only on ''live'' versus ''total'' cells. There was no difference in this ratio among all single cells (not live gated) in the presence and absence of NK cells, suggesting that if cells are disappearing from culture due to apoptosis, they are disappearing at an equal rate among infected and bystander cells ( Figure S1). Meanwhile, the ratio of mNeonGreen+ cells to mNeonGreenÀ cells was increased in live-gated cells upon addition of NK cells due to preferential killing of uninfected target cells by NK cells ( Figure S1).

SARS-CoV-2-infected cells do not actively inhibit NK cell functionality
We next interrogated changes in NK cell phenotype and function induced by co-culture with mock-or SARS-CoV-2-infected target cells. Importantly, we continued utilizing an MOI of 0.5, resulting in around 50% infection of the SARS-CoV-2-infected wells. We observed significant induction of CD107a, a marker of NK cell degranulation and surrogate for cytolytic activity, and IFN-g upon culture with either SARS-CoV-2-infected or mock-infected A549-ACE2 cells ( Figures 1F and 1G). Activation occurred primarily within the CD56 bright CD16 low subset, possibly due to IL-2 priming ( Figures 1F and S2). We also found no significant differences in the expression of other phenotypic and functional markers on NK cells co-cultured with SARS-CoV-2-infected targets compared with those cultured with mock-infected cells ( Figure S2). This suggests that, while healthy NK cells are unable to lyse SARS-CoV-2infected cells, the presence of SARS-CoV-2-infected cells does not inhibit the NK cell response to bystander cells. Collectively, these results support a model in which a factor intrinsic to SARS-CoV-2-infected cells allows escape of NK cell killing.

SARS-CoV-2 infection modulates expression of ligands involved in NK cell recognition
We next investigated the mechanism by which SARS-CoV-2-infected cells were able to evade lysis by NK cells. We used flow cytometry to profile the expression of the ligands for various NK cell activating and inhibitory receptors. 3 We grouped antibodies for ligands recognized by the same receptor into a single channel to quantify total ligand density for a given receptor. While expression of CD112/CD155 (ligands for DNAM-1), CD54 (ligand for LFA-1), and HLA-A/B/C were decreased in infected cells compared with mock and bystander cells, the magnitude of these reductions was relatively small. In contrast, the ligands for NKG2D (MICA, MICB, and ULBPs 1, 2, 5, and 6; collectively referred to as NKG2D-L) were downregulated to a much greater extent in SARS-CoV-2-infected cells compared with uninfected cells and bystander cells (Figures 2A, 2B, and S4A). All of the individual ligands comprising NKG2D-L were strongly downregulated in SARS-CoV-2-infected cells compared with mock-infected controls ( Figures 2C and S4B). Notably, the downregulation of NKG2D-L and the downregulation of HLA-A/B/C (MHC class I) would be expected to have opposing effects on the NK cell response to infected cells: downregulation of MHC class I would enhance NK cell recognition of infected targets, while NKG2D-L downregulation could represent a mechanism of NK cell evasion. As we observed a decrease in the ability of NK cells to kill SARS-CoV-2-infected cells and other studies have already interrogated MHC class I downregulation by SARS-CoV-2, 37-39 we focused our attention on the loss of NKG2D-L as a potential evasion mechanism.
Downregulation of NKG2D-L is correlated with inhibition of NK cell killing of SARS-CoV-2-infected cells To evaluate the association between NKG2D-L expression and killing of SARS-CoV-2-infected cells, we assessed NKG2D-L expression on the cells that survived following co-culture with NK cells. We identified a significant decrease in the frequency of NKG2D-L-expressing target cells in wells containing NK cells at both time points and across all infection conditions, suggesting that NK cells preferentially kill NKG2D-L-expressing targets in both SARS-CoV-2-infected and mock-infected wells (Figure 2D). We also assessed the kinetics of NKG2D-L expression on infected (mNeonGreen+) A549-ACE2 and found that, while NKG2D-L were downregulated to some extent at 24 h postinfection compared with uninfected cells, it was not until 48 h post-infection that we observed almost total loss of these proteins at the surface level ( Figure 2E). We therefore hypothesized that NK cells would kill infected cells more robustly at 24 h post-infection compared with 48 h. Indeed, we observed significantly better killing of mNeonGreen+ target cells at 24 h post-infection compared with 48 h ( Figure 2F). Further supporting a model in which downregulation of NKG2D-L allows for evasion of NK cell killing, we identified a strong correlation between the expression of NKG2D-L in target cells and target cell lysis across all time points and infection conditions ( Figure 2G).

NK cells are able to efficiently kill SARS-CoV-2-infected cells immediately following infection
Other groups have reported that NK cells are able to successfully suppress viral replication in a system where the NK cells are added to a target cell culture soon after infection with SARS-CoV-2. 16 Figure 3A), we compared total killing of all target cells in SARS-CoV-2-infected wells at 0 and 48 h. We found that, as expected, NK cells were able to robustly kill cells that were freshly infected (0 h) but not those that had been infected for 48 h ( Figure 3B). Moreover, NK cells were slightly better at killing infected cells compared with mock-infected cells at the 0 h time point ( Figure 3C), providing additional evidence that NK cells can successfully target infected cells in the early stages of SARS-CoV-2 infection, as previously reported. 16,18,19 Finally, we conducted a similar analysis of total cell killing at 24 versus 48 h post-infection. In accordance with our other findings, we observed that NK cells can efficiently kill virus-exposed cells through 24 h post-infection, but not at 48 h ( Figure 3D). Thus, our data and other published works collectively suggest that NK cells are capable of suppressing viral replication, but their ability to do so is significantly hampered if the cell has been infected for at least 48 h.

SARS-CoV-2 protein Nsp1 downregulates ligands for NKG2D
Having identified changes in the protein-level expression of NKG2D-L in SARS-CoV-2-infected cells that may underlie escape from NK cell killing, we next sought to understand how the virus mediates this effect. SARS-CoV-2 encodes 29 individual proteins that are broadly classified into 3 categories: structural, non-structural, and accessory. While the roles of these proteins are still being investigated, many of the non-structural and accessory proteins are known to suppress antiviral innate immune responses. [40][41][42][43] We therefore transfected each individual SARS-CoV-2 protein, tagged with two Strep Tag domains (Strep Tag II) to allow for easy detection, into A549-ACE2 cells and assessed for their effect on NK cell receptor ligand expression by flow cytometry ( Figures 4A and 4B). We successfully transfected 25 of the 29 SARS-CoV-2 proteins into A549-ACE2s; we also transfected cells with GFP as a non-viral control ( Figures S5A-S5C). While several proteins downregulated NKG2D-L, SARS-CoV-2 non-structural protein 1 (Nsp1) had by far the strongest effect ( Figures 4C  and S5D). Several other viral proteins, primarily accessory proteins, also downregulated NKG2D-L expression, and some increased expression. However, as Nsp1 had the most impact on NKG2D-L expression, we chose to move forward with interrogation of this protein.
Like replication-competent SARS-CoV-2, Nsp1 also downregulated MICA, ULBP-1, and ULBPs-2, 5, and 6. However, it had no effect on MICB ( Figure 4D). To ensure that the downregulation of NKG2D-L that we observed was not an artifact of the cell line we were using, we also transfected Nsp1 into 293T cells and K562 cells. Nsp1 downregulated NKG2D-L expression in both cell lines, which express NKG2D-L at baseline ( Figure S6A). Nsp1 also mediated downregulation of MHC class I, but not CD54 or the ligands for DNAM-1, in A549-ACE2s ( Figures 4F and S6B-S6D).
SARS-CoV-2 post-transcriptionally downregulates NKG2D-L and does not induce shedding, intracellular retention, or degradation Nsp1, also known as the SARS-CoV-2 leader protein, is the first protein translated when the virus enters a cell and serves as a global inhibitor of host translation. Nsp1 is highly conserved across coronaviruses as it plays an important role in enhancing pathogenicity by inhibiting the innate immune response. [44][45][46][47][48][49] Schubert et al. demonstrated that SARS-CoV-2 Nsp1 functions by sterically inhibiting entry of mRNA into the mRNA channel of the 40S ribosomal subunit. 49 Thus, it is likely that Nsp1 mediates a translational block to reduce surface NKG2D-L expression.
To orthogonally validate that NKG2D-L expression is reduced via translational blockade in SARS-CoV-2-infected cells, we assessed several other potential methods of downregulation. Consistent with a model of translational inhibition, we observed only a small decrease in transcripts encoding MICB, ULBP-1, and ULBP-2 in infected cells compared with mock-infected cells and no decrease in MICA transcript levels ( Figure 5A). This modest difference likely reflects the overall decrease in transcript levels in cells infected with SARS-CoV-2 and is consistent with the idea that NKG2D-L expression is reduced at the post-transcriptional level. We also assessed whether SARS-CoV-2 might induce degradation of NKG2D-L, as CMV has also been shown to downregulate NKG2D-L through targeting these proteins for proteasomal or lysosomal degradation. 31,32 We therefore treated mock or SARS-CoV-2-infected cells with a proteasomal inhibitor (MG-132) or a lysosomal inhibitor (BAF-A1) and then assessed NKG2D-L expression; we found that neither inhibitor rescued NKG2D-L expression in infected cells ( Figures 5B and S8A). Finally, we addressed the possibility of SARS-CoV-2-infected cells shedding of NKG2D-L from the cell surface, which has been reported for other viruses and in the setting of cancer, 24,50,51 by assessing NKG2D-L levels in the supernatants of mock-and SARS-CoV-2-infected cultures by ELISA. We quantified levels of soluble MICA and soluble ULBP-2 ( Figure 5C) as these were the two most highly expressed NKG2D ligands on mock-infected cells ( Figure 2C). We were unable to detect either of these proteins in the supernatants of uninfected or infected cultures, suggesting that secretion of NKG2D ligands is not a major mechanism by which NKG2D-L is downregulated by SARS-CoV-2 ( Figure 5C). Collectively, these data suggest that NKG2D-L are downregulated post-transcriptionally and are not degraded or shed in SARS-Cov-2-infected cells. While this supports the hypothesis that Nsp1 inhibits expression of these proteins by translational blockade, we were unable to definitively prove this mechanism, as expression of NKG2D-L could be suppressed by another mechanism such as intracellular retention. 23,29 NKG2D-L have a high rate of surface turnover Although Nsp1 is a global inhibitor of host translation, our data show that it does not equally downregulate all NK cell receptor ligands. We hypothesized that this might be due to differential rates of surface expression turnover across the various ligands, as these proteins are known to have varying levels of stability on the cell surface. 32,52-54 NKG2D-L in particular are rapidly turned over to allow for a high degree of control over its expression level. 32,52 To validate that non-specific inhibition of a post-transcriptional mechanism could have an outsized effect on NKG2D-L in comparison with the other ligands assayed, we treated A549-ACE2s with the protein transport inhibitor Brefeldin A and measured expression of NK cell receptor ligands after 24 or 48 h ( Figure S7). We observed that Brefeldin A, like Nsp1, had a much larger effect on NKG2D-L than on other ligands, including CD54 and DNAM-1 ligands, supporting a model in which global translation inhibition, such as that mediated by Nsp1, could much more dramatically downregulate NKG2D-L than other surface proteins.
Nsp1 is not highly expressed until more than 24 h postinfection Thus far, our analyses of NK cell evasion mediated by replication-competent SARS-CoV-2 have relied on mNeonGreen as a correlate of viral protein expression. However, having determined that Nsp1 is the viral protein with the strongest effect on NKG2D-L expression, we wanted to validate (1) that mNeon-Green expression correlates with Nsp1 expression and (2) that Nsp1 expression inversely correlates with NKG2D-L expression in SARS-CoV-2-infected cells. We therefore stained SARS-CoV-2-infected or mock-infected A549-ACE2s with an anti-Nsp1 antibody and compared expression of Nsp1 to expression of mNeonGreen by flow cytometry. We found that essentially all mNeonGreen+ cells also expressed Nsp1 ( Figure 6A). In addition, we determined that, like mNeonGreen, we could not detect high levels of Nsp1 expression until >24 h post-infection (Figure 6A); this aligns with our data demonstrating that SARS-CoV-2-infected cells are not fully resistant to NK cell killing until >24 h post-infection ( Figures 2E and 2F).
While all mNeonGreen+ cells also expressed Nsp1, there was a significant population of cells ($10%) at 48 h post-infection that  Figure 6A). This can likely be explained by the fact that Nsp1 is encoded at the 5 0 -most end of the SARS-CoV-2 genome and is thus the first viral protein to be translated. [44][45][46][47][48][49] This suggests that identification of infected cells based solely on mNeonGreen expression slightly underestimates the number of infected cells and likely explains why bystander cells appear to have slightly decreased expression of NKG2D-L compared with mock-infected cells; the bystander population includes some cells that have been recently infected and express Nsp1 but not yet mNeonGreen. It also allowed us to assess expression of NKG2D-L-infected cells subsetted by their expression of mNeonGreen and Nsp1. As expected, Nsp1À mNeonGreenÀ (Q4) cells had high expression of NKG2D-L, while Nsp1+ mNeonGreen+ (Q2) cells had lost almost all expression of NKG2D-L. However, Nsp1+ mNeonGreen-(Q3) cells had an in-termediate level of NKG2D-L expression, with roughly 20% of this population expressing these ligands ( Figure 6B). We hypothesize that these cells are more recently infected and have not yet expressed the full complement of viral proteins. Therefore, these data suggest that NKG2D-L downregulation precedes expression of at least some viral proteins.
Nsp1 is sufficient to confer significant resistance to NK cell-mediated killing We hypothesized that, if Nsp1 is the key mediator of NKG2D-L downregulation in SARS-CoV-2 infection, transfection with Nsp1 should be sufficient to confer resistance to NK cell killing. To test this hypothesis, we co-cultured activated, healthy NK cells with cells that had been transfected with either Nsp1 or a control plasmid (GFP) and assessed target cell killing by flow cytometry. Indeed, we found that NK cells were significantly more effective at killing GFP-transfected targets compared with Nsp1-transfected targets in both A549-ACE2s and 293Ts ( Figures 7A, 7B, and S8). To determine whether other viral proteins might also mediate escape from NK cell killing, we compared killing of Nsp1-transfected target cells with killing of cells transfected with other SARS-CoV-2 proteins (Figures 7C  and S8). We randomly selected 10 additional SARS-CoV-2 proteins to test alongside Nsp1. Each protein was transfected into A549-ACE2s and healthy NK cell killing of transfected cells was assessed 48 h post-transfection. We distinguished transfected cells from untransfected cells within the same well by gating on Strep Tag II expression. Of the 11 proteins transfected, Nsp1-transfected cells were killed significantly less than those transfected with any other plasmid except Nsp10 (no significant difference) ( Figure 7C). Nsp1 was also the only protein that significantly protected transfected cells from NK cell killing ( Figures 7C and S8D). Moreover, 6 of the other 10 proteins tested caused a significant increase in NK cell killing of transfected cells ( Figure S8D). Collectively, these data suggest that Nsp1 is sufficient to protect cells from NK-mediated killing and that resistance to NK cell killing in infected cells overcomes the increase in susceptibility to killing caused by other SARS-CoV-2 proteins.
Finally, we sought to compare the protection from NK cell killing mediated by Nsp1 transfection to that conferred by infection with replication-competent SARS-CoV-2. Like SARS-CoV-2, Nsp1 was able to provide significant protection to cells that received the protein versus bystander cells in the same well (Figure 7D). We then quantified protection from killing by calculating the fold change in killing of treated (Nsp1-transfected or SARS-CoV-2-infected) compared with bystander cells for each donor and found that there was no significant difference between the level of protection mediated by Nsp1 and that mediated by SARS-CoV-2 ( Figure 7E).

DISCUSSION
The role of NK cells in mediating clearance of SARS-CoV-2-infected cells in vivo remains unclear. While several studies have demonstrated that NK cells can reduce the levels of SARS-CoV-2 replication in vitro, no prior study has directly evaluated killing of SARS-CoV-2-infected cells. Here, we address this critical gap in knowledge and demonstrate that SARS-CoV-2-infected cells escape killing by healthy NK cells in a cell-intrinsic manner, while killing of uninfected bystander cells is uninhibited. The ability of infected cells to evade NK cell recognition requires infection to proceed long enough to allow an infected cell to express SARS-CoV-2-encoded proteins. We demonstrate that this escape mechanism is driven by downregulation of ligands for NKG2D, a critical activating receptor on NK cells. We further demonstrate that this ligand downregulation is driven by the SARS-CoV-2 Nsp1 protein and show that Nsp1 alone is sufficient to mediate direct NK cell evasion. While our experimental system using a cell line with high expression of NKG2D-L could enhance the degree of bystander killing, these findings have important implications for NK cell-mediated control of SARS-CoV-2, as preferential escape of infected cells and possible killing of bystander cells could contribute to SARS-CoV-2 pathogenesis.
These results illustrate the importance of examining the temporal dynamics of the NK cell response to SARS-CoV-2-infected cells. Other studies have assessed the ability of NK cells to suppress viral load by co-culturing NK cells with SARS-CoV-2infected targets early after infection; their results suggest that, under these conditions, NK cells can at least partially control viral replication. 16,18,19 It is worth noting that these other studies also varied from ours in parameters such as target cell type, cytokine treatment of NK cells, E:T ratio, and duration of co-culture. Our own observations demonstrate that NK cells are no longer able to effectively kill infected cells when added to the culture at 48 h post-infection, after the expression of viral proteins that suppress the innate immune response. The preferential killing of NKG2D-L-positive bystander cells may have important implications for lung pathology during COVID-19. NKG2D-L can be expressed by most cell types 55 and are upregulated during viral infections, including HIV 56 and RSV, 57 in response to stress. 58 Therefore, it is possible that NK cells may actually cause damage to the healthy tissue surrounding infected cells rather Article ll OPEN ACCESS than clearing the infection, although this hypothesis has not yet been directly tested in primary lung tissue. As NK cells appear to home to the lungs during COVID-19, 59-61 our findings indicate that the timing of NK cell trafficking to the site of infection may impact the efficacy of the NK cell response to SARS-CoV-2 infection, as there is a very narrow window for killing of infected cells before bystander killing could ensue. Interestingly, Witkowski et al. observed that frequency of peripheral blood NK cells in severe COVID-19 patients negatively correlated with viral load; however, this is difficult to interpret in the context of our data because it is unknown whether the increased NK cell frequencies observed resulted from decreased trafficking to the lungs, increased peripheral proliferation, or another mechanism. 19 Our novel finding that the SARS-CoV-2 protein Nsp1 mediates evasion of NK cell killing has significant implications for both the study of the immune response to coronaviruses and the development of therapeutics for COVID-19. Nsp1 is highly conserved across coronaviruses and is an essential virulence factor; it has been shown to inhibit translation of host antiviral factors across multiple beta-coronaviruses. 44-49, 62 One study found that, among nearly 50,000 SARS-CoV-2 sequences analyzed, only 2.4% had any mutations within Nsp1. 46 SARS-CoV-2 Nsp1 also shares 84.4% of its sequence identity with SARS-CoV Nsp1. Moreover, critical motifs within Nsp1 involved in the inhibition of innate immune responses are highly conserved across many beta-coronaviruses. 46 On a practical level, the high degree of conservation of Nsp1 and its importance in coronavirus virulence have already made this protein the focus of several therapeutic strategies. 44,63,64 Our work demonstrates that Nsp1 is an even more attractive target than previously thought, as inhibiting the function of this protein has the potential to fully or partially rescue the NK cell response to SARS-CoV-2-infected cells.
Although Nsp1 is a global inhibitor of host translation, our data demonstrate that it has an outsized effect on NKG2D-L and MHC class I surface expression compared with that of other ligands for NK cell receptors. This appears to be due to the varying stabilities of the different ligands on the cell surface, rather than explicit specificity of Nsp1 for NKG2D-L or MHC class I. It has been established that NKG2D-L are rapidly turned over on the cell surface and are quickly lost upon treatment with a protein transport inhibitor such as Brefeldin A. 32,52 MHC class I is similarly transient on the cell surface in the presence of translation inhibition, although its stability varies with haplotype and peptide binding. 53 CD54, which was not affected by Nsp1, is highly stable for at least 48 h, even after treatment with similar inhibitors. 54 Thus, the differential effects of Nsp1 on various ligands for NK cell receptors are likely explained by the varying kinetics of surface turnover.
One of our findings that has been demonstrated by multiple groups is the downregulation of MHC class I upon SARS-CoV-2 infection. The mechanism of this downregulation remains unclear; while our data suggest that Nsp1 is responsible for this loss, ORF3a, 37 ORF7a, 37 ORF6, 38 and ORF8 39 have also been implicated. This could be due to differential downregulation of various HLA molecules by different SARS-CoV-2 proteins. In our study, we grouped together HLAs A, B, and C as there are no commercially available antibody clones that can robustly differentiate HLAs A and B; this is an important limitation of our work. According to the well-established ''missing self'' model of NK cell activation, 65,66 the downregulation of self-MHC can induce NK cell activation through subsequent lack of inhibitory signaling through the killer cell immunoglobulin-like receptors. Therefore, it might be expected that the downregulation of MHC by SARS-CoV-2 would enhance the ability of NK cells to lyse infected cells-precisely the opposite of what we observed in our study. We hypothesize that this can be explained by (1) the relative magnitudes of MHC class I and NKG2D-L downregulation on infected cells and (2) the accepted dogma in the field that missing self alone is not sufficient to cause robust NK cell activation. 67,68 As a result, we propose that the loss of NKG2D-L is the dominant factor in the NK cell response (or lack thereof) to SARS-CoV-2.
While our study focuses on direct lysis of target cells, NK cells can also kill through antibody-dependent cellular cytotoxicity. A recent study by Fielding et al. found that antibody-dependent NK cell activation can overcome SARS-CoV-2's inhibition of direct cytotoxicity, allowing healthy NK cells to mount stronger responses to infected targets. Thus, prior vaccination or infection that results in pre-existing antibodies to SARS-CoV-2 could tip the balance in favor of killing SARS-CoV-2-infected cells. This study also identified downregulation of NKG2D-L on SARS-CoV-2-infected cells through an orthogonal method. 20 This work has significant implications for the ongoing study of COVID-19. Our results deeply interrogate a potential flaw in the ability of the immune system to mount a comprehensive immune response to COVID-19. We demonstrate that the timing of the NK cell response to SARS-CoV-2-infected target cells is critical, with NK cells being able to control viral replication early in infection, but not after expression of viral proteins has begun. This should be further interrogated in vivo to explore whether the kinetics of NK cell trafficking during COVID-19 affects disease outcome. Finally, we reveal that SARS-CoV-2 protein Nsp1 is a major factor in mediating evasion of NK cell killing. This finding reinforces the attractiveness of Nsp1 as a therapeutic target.
Limitations of the study Our study has several limitations. To focus on NK cell responses in the respiratory tract, we used A549-ACE2 cells, which are an immortalized, malignant cell line. This could therefore have enhanced NK cell targeting of bystander cells. In addition, while we demonstrated that Nsp1 was sufficient to confer NK cell escape, we were unable to test whether the absence of Nsp1 rescues NK cell killing because knockout of Nsp1 is lethal to the virus. We also did not fully evaluate why Nsp1 blocks NKG2D-L more effectively than other proteins, but we hypothesize that these proteins are downregulated first as part of the global translation block because they are turned over on the cell surface more quickly and cannot be replaced. Finally, we did not interrogate the ability of every individual SARS-CoV-2 protein to mediate escape from NK cell killing.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following: