Computational fingerprinting of immune-mediated pressure on SARS-CoV-2 viral evolution reveals preliminary evidence for immune-evasion

Abstract

The global emergence of SARS-CoV-2 variants and Omicron in particular, has highlighted the risk of novel variants that will evade adaptive immune responses following vaccination and natural infection. Determining which viral mutations escape T-cells and antibodies is crucial for designing therapeutics and vaccines and assessing the implications of newly emerging variants. We developed an automated immunoinformatic pipeline that considers both data on viral sequences and experimentally identified T-cell and antibody responses, to identify specific viral mutations that may be associated with immune escape. We used an unsupervised clustering method to cluster a set of 108 non-redundant anti-SARS-CoV-2 spike protein antibodies with solved 3D structures from the PDB database, and identified 9 unique antibody clusters. We computed the predicted changes in binding energies (ΔΔG) for each antibody and each specific mutation within its contact footprint. Using these ΔΔG scores, we computed an antibody escape score for each mutation. We also predicted the change in binding between predicted and experimentally verified T-cell epitopes and their presenting MHC molecules. We ranked frequent spike protein mutations based on their predicted effect on immune evasion, as well as individual SARS-CoV-2 variants, with a specific focus on VOCs. Our novel immune escape scores identified several key mutations in the Omicron variant, which were not present in previous VOCs, and may be used to identify key mutations that are likely associated with immune escape, which may appear in novel variants of concern. Supported by an NIH grant (75N93021C00016)

Quantification of heterogeneity in human CD8+T cell responses to vaccine antigens: an HLA-guided perspective

“…We would like to apply our methods to other pathogens of public health relevance such as Lassa virus and Crimean Congo hemorrhagic fever virus, with the viral sequences provided in Refs. [47, 48] Another avenue we have failed to explore is that of immune evasion and the role of MHC-restriction [17] in eliciting HLA-mediated selective pressure [11, 12, 13]. We plan to make use of the computational methods developed by Hertz et al .…”

“…We would like to apply our methods to other pathogens of public health relevance such as Lassa virus and Crimean Congo hemorrhagic fever virus, with the viral sequences provided in Refs. [47,48] Another avenue we have failed to explore is that of immune evasion and the role of MHC-restriction [17] in eliciting HLA-mediated selective pressure [11,12,13]. We plan to make use of the computational methods developed by Hertz et al [17] and the approaches adopted here to quantify the potential of a vaccine protein to exert immune pressure and drive viral evolution in different human populations, as well as to identify HLA generalists and specialists [29].…”

“…[47,48] Another avenue we have failed to explore is that of immune evasion and the role of MHC-restriction [17] in eliciting HLA-mediated selective pressure [11,12,13]. We plan to make use of the computational methods developed by Hertz et al [17] and the approaches adopted here to quantify the potential of a vaccine protein to exert immune pressure and drive viral evolution in different human populations, as well as to identify HLA generalists and specialists [29]. Finally, the CD8 + T cell metrics proposed here do not account for T cell function (cytokine secretion, proliferative capacity, or cytotoxic killing activity) or T cell half-life (of particular relevance for central and effector memory T cells).…”

“…This novel graph approach allows us to address the above challenges: 1) viral genetic diversity of the pathogen is represented in the set of peptides, 𝒫, so that wild type and all circulating (or predicted) variants can be analyzed, 2) HLA variability is considered with regard to geographical regions ℛ, HLA alleles 𝒜, and their frequencies within each region, and 3) TCR recognition variability is accounted for by peptide immunogenicity [16]. Finally, the entire multi-partite graph, 𝒢, straightforwardly provides a metric to quantify vaccine coverage (see Section 2.2), and the framework to characterize immuno-dominant peptides (experimentally identified) and to predict viral immune escape from CD8 + T cell recognition [17] (see Section 4). Our methods will be applied to Ebola virus, SARS-CoV-2, and Burkholderia pseudomallei vaccine proteins.…”