Evasion of NKG2D-mediated cytotoxic immunity by sarbecoviruses

Abstract

SARS-CoV-2 and other sarbecoviruses continue to threaten humanity, highlighting the need to characterize common mechanisms of viral immune evasion for pandemic preparedness. Cytotoxic lymphocytes are vital for antiviral immunity and express NKG2D, an activating receptor conserved among mammals that recognizes infection-induced stress ligands (e.g., MIC-A/B). We found that SARS-CoV-2 evades NKG2D recognition by surface downregulation of MIC-A/B via shedding, observed in human lung tissue and COVID-19 patient serum. Systematic testing of SARS-CoV-2 proteins revealed that ORF6, an accessory protein uniquely conserved among sarbecoviruses, was responsible for MIC-A/B downregulation via shedding. Further investigation demonstrated that natural killer (NK) cells efficiently killed SARS-CoV-2-infected cells and limited viral spread. However, inhibition of MIC-A/B shedding with a monoclonal antibody, 7C6, further enhanced NK-cell activity toward SARS-CoV-2-infected cells. Our findings unveil a strategy employed by SARS-CoV-2 to evade cytotoxic immunity, identify the culprit immunevasin shared among sarbecoviruses, and suggest a potential novel antiviral immunotherapy.

Keywords: COVID-19; MIC-A/B; NK cells; NKG2D; SARS-CoV-2; cytotoxic; lymphocytes; sarbecoviruses.

Figure 1:. SARS-CoV-2 downregulates MIC-A/B and other NKG2D ligands from the cell surface.

(A) Schematic of A549-ACE2 cells infected or non-infected with SARS-CoV-2 to generate a mix of infected and bystander cells for NKG2D ligand staining. (B) Surface binding of NKG2D-Fc and antibodies against NKG2D ligands in non-infected A549-ACE2 cells measured by flow cytometry. Representative histograms show unstained and stained A549-ACE2 cells and bar graph shows aggregate data. Relative median fluorescence intensity (RMdFI) was calculated for n = 3 technical replicates. (C) A549-ACE2 cells subjected to wild-type SARS-CoV-2 infection or no infection, stained with NKG2D-Fc or MIC-A/B, MIC-A, and ULBP-2/5/6 antibodies and intracellular SARS-CoV-2 nucleocapsid at 48 h post-infection (hpi). Representative flow plots and histograms and bar graphs of aggregate data are shown with unstained control included. Median fluorescence intensity (MdFI) is shown for n = 3 – 8 replicates pooled from two experiments. (D) MIC-A/B surface expression in A549-ACE2 cells that were not infected or were subjected to wild-type SARS-CoV-2 infection and classified as infected (nucleocapsid⁺) or bystander (nucleocapsid⁻) cells within the same well at 48 hpi. A high infection frequency (>75% nucleocapsid⁺ cells) was used. Bars indicate mean of n = 4 replicates pooled from two experiments. (E) MIC-A/B surface expression on infected and bystander cells is shown for different SARS-CoV-2 variants at 48 hpi. A low-to-moderate infection frequency (5–30% nucleocapsid⁺ cells) was achieved by infecting a 1:1 mixture of A549-ACE2 cells and infection-resistant A549 cells (no ACE2 overexpression). Relative MdFI is expressed as a percentage (%, RMdFI) for n = 3 technical replicates. (B-E) One-way ANOVA (parametric) (B-D) or repeated measures one-way ANOVA (E) was performed with correction for multiple comparisons, statistical significance is denoted as follows: **** p < 0.0001, *** p < 0.001, ** p < 0.01, and * p < 0.05. Bars and error bars indicate mean ± SD. See also Figure S1.

Figure 2:. SARS-CoV-2 downregulates MIC-A/B via shedding in vitro.

(A) Schematic of MIC-A/B life cycle processes that could be altered by viral infection. Created with BioRender.com. (B) Bulk RNA-seq of wild-type SARS-CoV-2-infected versus non-infected A549-ACE2 cells at 48 hpi. Transcript expression was normalized to transcripts per million (TPM) and log₂ fold change (log₂FC) in transcript expression was calculated for NKG2D ligand and sheddase genes. Bars and error bars indicate mean ± SEM for n = 3 technical replicates. Official gene names for ULBP4, ULBP5, and ULBP6 are RAET1E, RAET1G, and RAET1L, respectively. Multiple unpaired t test with Welch correction and false discovery rate (FDR) analysis were performed: ** q < 0.01. No data (n.d.) for undetectable transcripts. (C) Immunofluorescence staining of SARS-CoV-2 nucleocapsid, cell nuclei with DAPI, and MIC-A or MIC-B protein on non-infected and wild-type SARS-CoV-2-infected A549-ACE2 cells at 24 and 48 hpi. Scale bar (white horizontal line) indicates 50 μm. (D) MIC-A/B surface expression on wild-type SARS-CoV-2-infected A549-ACE2 cells treated with solvent (0.1% DMSO, None), a proteasome inhibitor (10 μM MG-132, Prot), a lysosome inhibitor (1 μM concanamycin A, Lyso), or a MIC-A/B shedding inhibitor (25 μg/mL 7C6 antibody, MIC Shed). Representative histograms and a bar graph are presented and include A549-ACE2 cells that are non-infected and unstained for comparison. Bar graph shows mean ± SD for n = 4 technical replicates. (E) Soluble MIC-A and MIC-B levels in the cell culture supernatants of A549-ACE2 cells either left non-infected or infected with the indicated SARS-CoV-2 variants at 48 hpi. Lines and error bars indicate mean ± SD and dotted lines represent lower limit of detection. One-way ANOVA (parametric) was performed with correction for multiple comparisons: **** p < 0.0001, *** p < 0.001, * p < 0.05. (F) A549-ACE2 cells infected with wild-type SARS-CoV-2 and treated with different concentrations of 7C6 or isotype antibody followed by multiplexed ELISA for soluble MIC-A in supernatant and flow cytometry for MIC-A/B surface expression on cells at 48 hpi. Lines represent the mean of n = 2 technical replicates. Multiple unpaired t tests (parametric) with correction for multiple comparisons was performed for 7C6 versus isotype antibody treatment at each concentration: **** p < 0.0001, ** p < 0.01, * p < 0.05. See also Figure S2.

Figure 3:. SARS-CoV-2-infected human lung xenografts and COVID-19 patient serum and respiratory tract samples show evidence of MIC-A/B modulation and shedding in vivo.

(A) Schematic of NOD Rag1^–/–Il2rg^null (NRG) mice xenotransplanted with human lung tissue and infected with wild-type SARS-CoV-2 via intra-xenograft injection for bulk RNA-seq and immunofluorescence (IF) histology and microscopy at 2 or 7 days post-infection (dpi) versus non-infected (0 dpi, naïve) mice. (B) Bulk RNA-seq of wild-type SARS-CoV-2-infected human lung xenografts at 2 dpi (n = 4) and 7 dpi (n = 6) versus non-infected (naïve) human lung xenografts (n = 3). Transcript expression was normalized to transcripts per million (TPM) and log₂ fold change (log₂FC) in transcript expression was calculated for each gene shown, categorized as an NKG2D ligand or a sheddase. Bars indicate mean. Multiple unpaired t tests with Welch correction and false discovery rate (FDR) analysis were performed: ** q < 0.01. (C) Immunofluorescence staining of SARS-CoV-2 nucleocapsid, cell nuclei with DAPI, and MIC-A or MIC-B protein on human lung xenograft tissue sections that were non-infected (naïve) or 2 dpi and 7 dpi with wild-type SARS-CoV-2. Scale bar (white horizontal line) indicates 50 μm. (D) Serum levels of soluble MIC-A and MIC-B in samples from the Massachusetts and São Paulo combined cohorts. Samples were classified as healthy (negative for COVID-19; n = 28) or positive for COVID-19 (n = 143), and further stratified by disease severity (11 mild, 62 moderate, 50 severe and 20 lethal). One-way ANOVA (non-parametric) was performed with correction for multiple comparisons: **** p < 0.0001, *** p < 0.001, * p < 0.05. (E) UMAP of clustered respiratory epithelial cells from scRNA-seq analysis of pooled nasopharyngeal swabs and bronchial brush samples from COVID-19 patients (n = 19) and healthy controls (n = 5). Dataset source (Chua et al. 2020 Nat Biotechol; PMID: 32591762): https://explore.data.humancellatlas.org/projects/7ac8822c-4ef0-4194-adf0-74290611b1c6 (F) Dotplot of transcript expression of NKG2D ligands and sheddases in respiratory epithelial cell clusters from healthy control and COVID-19 patient samples. See also Figure S3.

Figure 4:. MIC-A/B downregulation via shedding is mediated by SARS-CoV-2 accessory protein ORF6, which is uniquely conserved among sarbecoviruses.

(A) Schematic illustrates generation of dual mRNA lipid nanoparticles containing in-vitro-transcribed mRNA encoding an individual SARS-CoV-2 protein and mRNA encoding eGFP for transfection of A549 cells. Surface ligand staining via flow cytometry was performed 36 – 48 h post-transfection. (B) Representative flow cytometry plots of MIC-A/B surface expression and eGFP expression in A549 cells after transfection with the indicated mRNAs. (C) MIC-A/B surface expression in A549 cells transfected with mRNAs encoding for all known individual SARS-CoV-2 proteins and eGFP measured via flow cytometry. Expression is represented relative to baseline expression in untransfected/bystander cells from the same well (depicted by dotted line at 100%). Line represents mean of n = 3 technical replicates. One-way ANOVA (parametric) was performed between transfected and untransfected/bystander cells from the same well with correction for multiple comparisons: *** p < 0.001. (D) Levels of soluble MIC-A and MIC-B in the supernatant of A549 cells transfected with eGFP alone, N, Nsp1, or ORF6 measured via multiplexed ELISA, normalized to the number of cells counted via flow cytometry to adjust for number of producing cells. Scatter plots show aggregate data of n = 5 technical replicates with lines and error bars indicating mean ± SD. One-way ANOVA (parametric) was performed with correction for multiple comparisons: **** p < 0.0001, *** p < 0.001. (E) Cladogram of genera within coronaviruses and subgenera within the genus Betacoronavirus, including Sarbecovirus, which uniquely contains the accessory protein ORF6 encoded in their genome between structural protein M and N, which may also contain other accessory proteins (i.e., ORF7, ORF8) between them. (F) Sequences of ORF6 protein from SARS-CoV-2, SARS-CoV-2-related viruses, SARS-CoV, and SARS-CoV-2-related viruses were aligned and show high sequence conservation. GenBank and GISAID accessions are provided. For SARS-CoV-2, the ORF6 sequence shown is from the original Wuhan-Hu-1 virus (YP_009724394) but represents virtually all variants that infect humans except BA.2 and BA.4, which contain a single D61L mutation. ORF6 sequences from SARS-CoV-2-related viruses included are RaTG13 (QHR63304), ZXC21 (AVP78046), and RmYN01 (GISAID; EPI_ISL_412977) isolated from bats, as well as MP789 (QIG55949) and GX-P5L (QIA48636) isolated from pangolins. For SARS-CoV ORF6 sequences, two strains that have been shown to infected humans and civets were included: Tor2 (YP_009825056) and BJ01 (AAP30035), which have been shown to infected humans and civets. ORF6 sequences from SARS-CoV-related viruses included are WIV-1 (AGZ48837), Rm1 (ABD75327), and HKU3–1 (AAY88870), which have been isolated from bats. ORF6 from SARS-CoV-2 and related viruses is 61 a.a. in length while ORF6 from SARS-CoV and related viruses is 63 a.a. Mammals from which these viruses were isolated are indicated (left-to-right): pangolin, bat, civet, human. See also Figure S4.

Figure 5:. Primary human NK cells efficiently recognize and kill SARS-CoV-2 infected cells, which can be enhanced by inhibiting virus-induced MIC-A/B shedding.

(A) Schematic of experimental set-up of NK-cell co-culture with SARS-CoV-2-infected A549-ACE2 cells (with or without antibody treatments) to measure infection frequency, target-cell survival, and NK-cell degranulation via flow cytometry and live cell counting. (B) A549-ACE2 infection frequency (% nucleocapsid⁺) in non-infected cells and wild-type SARS-CoV-2 infected cells cultured either alone (SARS2 inf.) or co-cultured with increasing quantities of cytokine-expanded NK cells (+ NK cells/well). Representative flow cytometry plot shows infection frequency. Bar graphs show aggregate data of target-cell counts of bystander (nucleocapsid⁻) and infected (nucleocapsid⁺) A549-ACE2 target cells. Bars show mean and connecting lines indicate paired data for n = 6 healthy donors. (C) A549-ACE2 cells were left non-infected or infected with wild-type, Delta, or Omicron/BA.1 variants of SARS-CoV-2 for 24 h before co-culturing with cytokine-expanded NK cells and assessing target-cell survival and NK-cell degranulation. In this assay, cytokine-expanded NK cells were added at 1.2 × 10⁴ cells/well for co-culture with target cells. Bars show mean and connecting lines indicate paired data for n = 6 healthy donors. (D) Representative flow plots of target-cell survival and NK-cell degranulation after co-culture of non-infected and SARS-CoV-2-infected A549-ACE2 cells with primary human NK cells. The target-cell marker used in this assay was PVR, which is highly expressed in A549-ACE2 cells, while the NK-cell markers used in this assay were CD56 and CD16. (E) Target-cell survival and NK-cell degranulation after co-culture of primary human NK cells (5 × 10⁴ cells/well) with A549-ACE2 cells that were non-infected, infected with wild-type SARS-CoV-2, or infected with wild-type SARS-CoV-2 and treated with 7C6 (20 μg/mL) in the presence or absence of an NKG2D blocking antibody (20 μg/mL). NK cells alone were also assessed to measure background NK-cell degranulation. Bars show mean and connecting lines indicate paired data for n = 7 healthy donors. (F) Primary human NK cells from the same assay in E phenotyped for their expression of various NK cell markers via Boolean gating to assess subset-specific degranulation. Only subsets with a frequency of >0.1% for at least one donor were included in analysis. Bar graph shows mean frequency of each NK-cell subset among bulk NK cells in n = 1 – 7 healthy donors. Heat map depicts degranulation frequency (CD107a⁺, %) of each NK-cell subset when cultured alone (No targets) or co-cultured with A549-ACE2 cells that were non-infected (A549-ACE2); infected with wild-type SARS-CoV-2 (SARS2 inf.); infected with wild-type SARS-CoV-2 and treated with 7C6 (SARS2 + 7C6); or infected with wild-type SARS-CoV-2 and treated with 7C6 and NKG2D blocking antibody (SARS2 + 7C6 + α-NKG2D). CD56⁺ denotes CD56^bright and CD56⁻ denotes CD56^dim. (C,E) Repeated measures one-way ANOVA was performed with correction for multiple comparisons: *** p < 0.001, ** p < 0.01, * p < 0.05, ns = not significant. See also Figure S5.