Emerging mutation in SARS-CoV-2 facilitates escape from NK cell recognition and associates with enhanced viral fitness

Abstract

In addition to adaptive immunity, natural killer (NK) cells of the innate immune system contribute to the control of viral infections. The HLA-E-restricted SARS-CoV-2 Nsp13232-240 epitope VMPLSAPTL renders infected cells susceptible to NK cells by preventing binding to the inhibitory receptor NKG2A. Here, we report that a recently emerged methionine to isoleucine substitution at position 2 (pM2I) of Nsp13232-240 impairs binding of the mutated epitope to HLA-E and diminishes HLA-E/peptide complex stability. Structural analyses revealed altered occupancy of the HLA-E B-pocket as the underlying cause for reduced presentation and stability of the mutated epitope. Functionally, the reduced presentation of the mutated epitope correlated with elevated binding to NKG2A as well as with increased NK cell inhibition. Moreover, the pM2I mutation associated with enhanced estimated viral fitness and was transmitted to descendants of the SARS-CoV-2 BQ.1 variant. Interestingly, the mutated epitope resembles sequences of related peptides found in endemic common cold-causing human coronaviruses. Altogether, these findings indicate compromised peptide presentation as a viral adaptation to evade NK cell-mediated immunosurveillance by enabling enhanced presentation of self-peptide and restoring NKG2A-dependent inhibition of NK cells.

Fig 1. A single amino acid substitution in the Nsp13_232-240 epitope is associated with enhanced viral fitness.

(A) Dynamics of variant frequencies detected by genomic surveillance. Combined data reported from Belgium, France, Germany, Norway, Portugal, and Sweden. (B) Newly reported infection cases (upper bars) and number of patients receiving intensive care tested positive for COVID-19 (lower bars) in Sweden. Reported infection cases are colored based on percentage of variant prevalence. (C) Phylogenic tree illustrating genomic divergence between ancestral SARS-CoV-2 and Omicron sub-lineages as determined by Clustalω. Length of branches denotes relative divergent nucleotides. (D) Genomic sequence identity between ancestral SARS-CoV-2 and Omicron sub-lineages as determined by Clustalω and Nsp13_232-240 sequences of the indicated sub-lineages.

Fig 2. The pM2I mutation diminishes presentation efficiency.

(A) Schematic illustration depicting cellular peptide binding assays based on loading of K562/HLA-E cells with different concentrations of synthetic peptides. (B) Representative cell surface HLA-E levels determined by flow cytometry upon loading with indicated concentrations of peptides. Numbers denote geometric mean fluorescence intensity (geoMFI). (C) Summary of HLA-E signals with dots denoting mean, error bars SEM, and lines dose-response curves (n = 4 independent experiments). Two-way ANOVA with Šídák’s multiple comparisons test (NS p ≥ 0.05, ** p <  0.01, *** p < 0.001, and **** p < 0.0001). (D) Concentrations required for half-maximal surface HLA-E signals (EC₅₀). Connected dots denote individual experiments and bars mean (n = 4 independent experiments). Paired t-test (** p <  0.01). (E) Schematic illustration depicting cellular HLA-E/peptide complex turnover assays based on loading of K562/HLA-E cells with peptides and subsequent removal. (F) Representative cell surface HLA-E levels determined by flow cytometry after removal of peptides since indicated times. Numbers denote geoMFI. (G) Summary of HLA-E levels relative to t = 0 min timepoint (t₀). Dots denote means, error bars SEM, and lines exponential decay curves (n = 4 independent experiments). Two-way ANOVA with Šídák’s multiple comparisons test (NS p ≥ 0.05, ** p <  0.01, and **** p < 0.0001). (H) Half-life of HLA-E/peptide complexes. Connected dots denote individual experiments and bars mean (n = 4 independent experiments). Paired t-test (* p <  0.05). (I) Schematic illustration depicting in vitro infection experiments by inoculation of Caco-2 cells with either SARS-CoV-2 Ancestral or SARS-CoV-2 BQ.1. (J) Representative cell surface HLA-E levels determined by flow cytometry following infection with MOI = 1. Numbers denote geoMFI. (K) Summary of HLA-E levels over time post-infection. Dots denote means and error bars SEM (n = 5 infections in n = 3 independent experiments). Two-way ANOVA with Šídák’s multiple comparisons test (NS p ≥ 0.05 and * p <  0.05).

Fig 3. The pM2I mutation alters HLA-E B-pocket occupancy.

(A) 2F_o-F_c electron density maps of BA.5 Nsp13_232-240 (left, yellow) and BQ.1 Nsp13_232-240 (right, violet) in complex with HLA-E*01:03 contoured at 1.0 σ, as determined by x-ray crystallography. (B) Side and top views (left and right, respectively) of the superimposed crystal structures of the two Nsp13_232-240 peptide variants. (C) Van der Waals distances between the CG2 atom of residue p2I in BQ.1 Nsp13_232-240 and the HLA-E residue Y7 as well as between Y7 and G26 in the B-pocket. Inter-atom distances are indicated.

Fig 4. Impaired presentation of the mutated epitope from SARS-CoV-2 BQ.1 allows for increased NK cell inhibition in mixed peptide repertoires.

(A) Schematic illustration depicting loading of K562/HLA-E cells with mixed peptide repertoires and subsequent NKG2A binding assay. Peptide concentrations within mixes were used as in S4G Fig and described in Methods. (B) Representative cell surface HLA-E levels (left) and binding of recombinant NKG2A/CD94 (right) determined by flow cytometry. Numbers denote geoMFI. (C-D) Summary of (C) HLA-E signals and (D) binding of recombinant NKG2A/CD94 upon loading with the indicated peptide mixes. Connected dots denote independent experiments (n = 4) and bars display mean. Repeated-measures one-way ANOVA with Šídák’s multiple comparisons test (NS p ≥ 0.05 and *** p < 0.001). (E) Schematic illustration depicting the NKG2A⁺ NK cell inhibition assay based on co-culture of peptide-loaded K562/HLA-E cells with purified NK cells. (F) Representative NKG2A⁺ NK cell activation measured by CD107a surface mobilization and expression of TNF-α as well as IFN-γ upon co-culture with K562/HLA-E cells loaded with solvent control (no peptide) or the indicated peptide mixes. Gated on viable CD3^- CD56^dim NKG2A⁺ NKG2C^- NK cells (gating strategy in S4D Fig). (G) Summary of HLA-E signals detected in parallel to functional assays. Dots denote independent experiments (n = 3) and bars display mean. Repeated-measures one-way ANOVA with Šídák’s multiple comparisons test (NS p ≥ 0.05). (H-J) Summary of NKG2A⁺ NK cell activation measured by frequency of (H) CD107a, (I) TNF-α, and (J) IFN-γ. Dots denote individual donors and bars mean (n = 5 donors in 3 independent experiments). Repeated-measures one-way ANOVA with Šídák’s multiple comparisons test (NS p ≥ 0.05, *** p < 0.001, and **** p < 0.0001). (K) Distributions of polyfunctional responses within NKG2A⁺ NK cells. Arcs denote mean frequency of activation marker expression and segments denote mean of cells positive for the corresponding number of markers (n = 5 donors in 3 independent experiments). (L) Summary of polyfunctional NKG2A⁺ NK cells co-expressing CD107a, TNF-α, and IFN-γ in the indicated conditions (n = 5 donors in 3 independent experiments). Repeated-measures one-way ANOVA with Šídák’s multiple comparisons test (NS p ≥ 0.05, *** p < 0.001, and **** p < 0.0001). (M-N) Peptides were mixed at a 1:9 ratio (self:viral) and K562/HLA-E cells were either treated with solvent control or loaded with 0.125, 0.5, 1.0, 100, 150, and 200 μM of peptide mixes, followed by co-culture with NK cells as in (E). Surface HLA-E levels on K562/HLA-E cells were determined in parallel to frequency of CD107a on NKG2A⁺ NK cells (see S4I Fig). (M) Response curve of HLA-E surface levels and normalized NKG2A⁺ NK cell inhibition. K562/HLA-E cells without peptide were set to 0% inhibition and NK cells cultured without target cells were set to 100% inhibition. Dots and error bars denote mean±SD of individual donors (n = 4 in 2 independent experiments) and lines display curve fit. (N) EC₅₀ values of HLA-E levels at half-maximal inhibition. Connected dots denote individual donors and bars mean (n = 4 donors in 2 independent experiments). Paired t-test (** p < 0.01).

Fig 5. The pM2I mutation is associated with enhanced viral fitness and resembles peptides found in common cold-causing human coronaviruses.

(A) Viral fitness estimated by comparing the observed occurrence of each mutation with its anticipated frequency based on the neutral mutation rate of SARS-CoV-2. Data is based on 6.5 million publicly available SARS-CoV-2 genomes [26, 27]. Every observed amino acid substitution within the Nsp13_232-240 epitope except Stop codons are displayed. Grey tiles indicate not detected (N.D.) amino acids. (B) Phylogenic tree illustrating genomic divergence between SARS-CoV-2 variants as determined by Clustalω. Tabular inlets display the sequence of the corresponding Nsp13_232-240 epitopes in (left) variants and (right) descendants of BQ.1. (C) Tabular summary of peptides related to the Nsp13_232-240 epitope as found in sarbecoviruses from wildlife, SARS-CoV-1, SARS-CoV-2 variants, and common cold-causing human coronaviruses. Amino acid alterations are highlighted in red compared to SARS-CoV-2 BA.5 Nsp13_232-240 and the presence of methionine in position two of the peptide (p2M) is indicated in the last column. See S1 Table for a complete list. (D) Representative cell surface HLA-E levels determined by flow cytometry following loading with viral peptides. (E) Summary of HLA-E signals upon loading of the indicated peptides. Dots denote individual experiments and bars mean (n = 4 independent experiments). Repeated-measures one-way ANOVA with Šídák’s multiple comparisons test (**** p < 0.0001). (F) Hamming distance of the peptide-encoding genomic sequence relative to SARS-CoV-2 BA.5. See S5 Fig. (E) Correlation between Hamming distance as in (F) and HLA-E levels as in (E). Dots and error bars denote mean±SD (n = 4 independent experiments). Colored areas indicate 95% confidence bands of linear regression between BA.5 and BQ.1 in red (best fit slope±SE –192±32) as well as BA.5 and HCoV in green (best fit slope±SE –59±6.9).