Computational analysis for infectious diseases surveillance and host-pathogen interactions (in the context of Influenza A viruses)

• Main Author: Beyene, Biruhalem Taye
• Other Authors: School of Biological Sciences
• Format: Thesis-Doctor of Philosophy
• Language: English
• Published: Nanyang Technological University 2017
• Subjects: DRNTU::Science::Biological sciences
• Online Access: http://hdl.handle.net/10356/72814
• Institution: Nanyang Technological University

Influenza A virus (IAV) is a major public health problem responsible for the death of half a million people every year worldwide. Zoonotic transmissions of the virus from swine and avian origin have occurred and can in the worst-case result in pandemics. Prediction of the next pandemic strain is still a major challenge as mechanisms of antigenic shift and the zoonotic nature of influenza viruses are still poorly understood. On the other hand, the current vaccination strategy treating seasonal influenza viruses, given its own problems in efficacy, is not well suited to mitigate an impending pandemic. While there are few clinically approved anti-IAV drugs for therapeutic and prophylactic use, these drugs have also been challenged by the emergence of drug resistance, toxicity, adverse effects and low efficacy. Hence, development of a broad range anti-IAV drug that could target all of the IAV strains irrespective of their source is critically important. To discover effective IAV vaccines and alternative anti-IAV therapeutics, it is crucial to understand how IAVs adapt to different species’ host ranges and vice versa, how different species hosts respond to different IAV subtypes. Additionally, it is crucial to understand IAVhost interactions and identifying critical viral or host factors that could support the replication of the virus or induce pathologic conditions. Recently, high throughput technologies such as mRNA microarray-based gene expression, genome-wide siRNA screening, and proteomic analysis are providing in-depth insights into host-pathogen interactions of IAVs. Hence, this study aimed to computationally investigate IAV-host interactions using three data set (transcriptome, genomewide siRNA screens and interactome) and identify 1) virus and host specific responses, 2) host determinants in host adaptations and 3) host targets for IAV therapeutics. This thesis contains four main projects. In the first project (chapter 3), we investigated the host gene expression changes of eight IAVs (H1N1/WSN, pH1N1, H5N2/F118, H5N2/F59, H5N2/F189, H5N3, H7N9 and H9N2 viruses) in A549 cells at different time points of infections. Then we integrated the differentially expressed genes (DEGs) in at least 3 viruses, with 1713 and 1780 host factors comprehensively curated from 11 siRNA and 13 interactome studies respectively. The integration of the three data set highlighted plausible influenza A virus required host factors (IHFs) that could be targeted against IAVs. The up-regulated IHFs (e.g. TRIM21, TRIM26, IRF2, and SAMHD1) might support the replication of IAV through suppression of the innate and adaptive antiviral immune responses. The other up-regulated IHFs could also enhance the replication of IAV at different stages of the virus lifecycle: endocytosis (BTC), prevention of apoptosis (TIMM17A), nuclear import (JAK2), and translation elongation of viral proteins (HEXIM1). Although several of these IHFs have been implicated in other viruses, the detailed mechanisms of how several of these up-regulated IHFs could support IAV replication require further investigation. The second project (chapter 4), explored a comprehensive analysis of host gene expression changes in different IAV infections in different host species. We used host gene expression signatures of cell lines from three species (A549 (human), CEF (chicken), and MDCK (canine) in response to six IAVs (H1N1/WSN, H5N2/F59, H5N2/F118, H5N2/F189, H5N3 and H9N2 viruses). To compare the expression changes between the three species, we performed comprehensive probe set re-annotation and human ortholog mapping. The result showed that the expression signatures of different IAV isolates in a single species cell type are more similar to each other compared to the expression signatures of a single isolate in different species’ cell types. The functional annotations (pathways) and the highly expressed cell-specific signatures indicated that IAVs up-regulated host factors could induce virus infectivity (e.g. OSBPL1A and ARHGAP21), reduce apoptosis (e.g. MRPS27) and increase cell proliferation (e.g. COPS2) in CEF cells. Conversely, increased antiviral, pro-apoptotic and inflammatory signatures have been identified in A549 cells. Except in H5N3 virus infections, generally IAV infections down-regulates genes associated with cell cycle and metabolic pathways with the strongest effect in MDCK cells, followed by A549 cells in a strain dependent manner, but not in CEF cells. Previously our group demonstrated that the replication of the examined viruses was significantly higher in CEF cells than the other cell types. Thus, we hypothesise that this could partially explain the mechanism how infectious LPAI viruses shed by chickens lack the inflammatory response and cellular disruption that may lead to disease conditions. In the third project (chapter 5), we used a systems-based approach to investigate changes to the transcriptome of primary murine lung macrophages (PMФ) in response to infection with the mouse-adapted H1N1/WSN virus and low pathogenic avian influenza (LPAI) viruses H5N2 and H5N3. The results showed that while all viruses induced antiviral responses, the H5N3 infection resulted in higher expression levels of cytokines and chemokines associated with inflammatory responses. Previously, our group showed that the LPAI H5N2 and H5N3 were able to infect murine lung macrophages and together with increased expression of inflammatory mediators particularly in H5N3 virus could impose threats to human health in the future. The IHFs identified by the IAV genome-wide siRNA screening studies could be used as potential anti-IV targets. However, these studies were not consistent (reproducible) mainly due to false negative results. Hence, in the fourth project (chapter 6), we applied computational gene network growing for discovering gene network links that could have been missed by these experimental investigations. Using the known IHFs we compared the network growing function of two free tools GeneMANIA and STRING and the commercial IPA for their performance of recovering other known IHFs previously identified from siRNA screens. The result showed that, given small (~30 genes) or medium (~150 genes) input sets, all three network growing tools detected significantly more known host factors than random human genes with STRING overall performing strongest. Extending the networks with all the three tools significantly improved the detection of GO biological processes of known host factors compared to not growing networks. Notably, the rate of identification of true host factors using computational network growing is equal or better to doing another experimental siRNA screening study which could also be true and applied to other biological pathways/processes. Finally, using the known IHFs and the "new IHF candidates" (genes connected to the IHFs from the network growing analysis), we predicted computational drug-target interactions using MetaCore. We identified 343 US Food and Drug Administration (FDA) approved drugs that had an inhibitory effect on either the known or new candidate IHFs, of which 258 were new predictions. Furthermore, using different criteria, we computationally ranked the 343 FDA approved drugs for further experimental validation.