Meta-analysis of host responses identifies immune gene network dysfunction during viral infection

Abstract

Virus infection induces a host response in infected cells which aim to reduce the spread of the virus into neighbouring tissues. In early stages of infection, viruses can limit the efficiency of host responses to facilitate rapid viral replication in infected cells. Highly virulent viruses often have distinct strategies in place to antagonize anti-viral activity through interference with host cell signalling processes. Such interactions can be difficult to characterize as many targets of viral antagonism are unknown. By characterizing genes and functional networks that are universally essential to a host response against viral infection, it will enable the identification of signalling processes that regulate these networks and pinpoint potential targets of viral interference.

To characterize the host response, I propose a unique approach to assess changes in gene expression by comparing the transcriptomes of infected and uninfected samples across multiple species. I demonstrate this method using data derived from in vivo and in vitro infections using two separate viruses: the influenza A virus and the SARS-CoV-2 virus. My approach combines meta-analysis and co-expression analysis to characterize recurrent differentially expressed genes present in datasets for each virus. Meta-analysis enabled the removal of non-replicable sources of study-specific and species-specific gene expression while highlighting gene expression that was most frequently identified across datasets. My analysis identified upregulated expression of a specific set of immune genes in response to each virus across a significant proportion of studies. Multi-functionality analysis revealed that many of these recurrent genes share roles in several functional networks, therefore it was important to determine how these genes function together as part of a universal host response to influenza and SARS-CoV-2 infection.

I next constructed aggregated co-expression networks from hundreds of publicly available gene expression datasets from mouse and human studies to determine how similar the functional properties are between genes in each species. Using these networks with recurrent gene sets, I identified a conserved module of immunity-associated genes involved in the host response to both viruses. This immune network was shown to form a fundamental component for anti-viral activity for a universal host response to viral infection. Using influenza as a model, I next sought to determine if a strategy of viral antagonism was in place to target the expression of this network. Emphasis was placed on the potential role of the influenza immune-antagonistic protein, non-structural protein 1 (NS1). Single-cell RNA-seq analysis identified variations in the expression of this immune module across several cell-types between samples infected with different NS1-deficient mutants or wildtype virus. On comparing the expression profile of this network in samples infected with influenza mutants, I determined that nuclear-located NS1 was essential for decreasing expression of this immune network through a re-distribution of RNA Polymerase II (PolII) at transcription start sites. Results from this study provide new insights into understanding the dynamic nature of how a host responds to infection. This two-pronged approach of meta-analysis and co-expression can be applied to many infectious diseases besides viral infection and offers a new way to explore viral-host interactions.