Abstract

   The current pandemic of coronavirus disease 19 (COVID-19) has affected
   millions of individuals and caused thousands of deaths worldwide. The
   pathophysiology of the disease is complex and mostly unknown.
   Therefore, identifying the molecular mechanisms that promote
   progression of the disease is critical to overcome this pandemic. To
   address such issues, recent studies have reported transcriptomic
   profiles of cells, tissues and fluids from COVID-19 patients that
   mainly demonstrated activation of humoral immunity, dysregulated type I
   and III interferon expression, intense innate immune responses and
   inflammatory signaling. Here, we provide novel perspectives on the
   pathophysiology of COVID-19 using robust functional approaches to
   analyze public transcriptome datasets. In addition, we compared the
   transcriptional signature of COVID-19 patients with individuals
   infected with SARS-CoV-1 and Influenza A (IAV) viruses. We identified a
   core transcriptional signature induced by the respiratory viruses in
   peripheral leukocytes, whereas the absence of significant type I
   interferon/antiviral responses characterized SARS-CoV-2 infection. We
   also identified the higher expression of genes involved in metabolic
   pathways including heme biosynthesis, oxidative phosphorylation and
   tryptophan metabolism. A BTM-driven meta-analysis of bronchoalveolar
   lavage fluid (BALF) from COVID-19 patients showed significant
   enrichment for neutrophils and chemokines, which were also significant
   in data from lung tissue of one deceased COVID-19 patient. Importantly,
   our results indicate higher expression of genes related to oxidative
   phosphorylation both in peripheral mononuclear leukocytes and BALF,
   suggesting a critical role for mitochondrial activity during SARS-CoV-2
   infection. Collectively, these data point for immunopathological
   features and targets that can be therapeutically exploited to control
   COVID-19.

   Keywords: COVID-19, transcriptomics, inflammation, metabolism,
   SARS-CoV-2, SARS-CoV, influenza, oxidative phosphorylation

Introduction

   The outbreak of coronavirus disease 19 (COVID-19), first recognized in
   Wuhan, China, rapidly became a pandemic of major impact not only on
   global public health but also on economy and social well-being ([29]1).
   SARS-CoV-2 infection results in clinical outcomes ranging from
   asymptomatic status to severe disease and ultimately, death ([30]2).
   Understanding of the molecular mechanisms underlying the pathology of
   COVID-19 is required to design effective therapies and safe vaccines.
   In this context, current investigations have been devoted to
   biochemical characterization and cellular phenotyping in patients to
   development of animal models of COVID-19 ([31]3).

   Transcriptomics of peripheral blood cells has been a powerful tool to
   characterize human immune responses to diverse pathogens, including
   respiratory viruses ([32]4–[33]6). Gene expression profiling by
   different analytical platforms and sample types revealed that COVID-19
   patients exhibit: (i) activation of humoral immunity, hypercytokinemia,
   apoptosis ([34]7), and dynamic toll like receptor (TLR) signaling
   ([35]8) in peripheral leukocytes; (ii) induction of interferon
   stimulated genes (ISGs), chemokines and inflammation in the lower
   respiratory tract ([36]7, [37]9, [38]10). Of importance, the results
   and interpretation of these data were based on single-gene-level
   analyses, in which significance of quantitative changes of each gene
   are calculated separately and they are latter submitted to pathway
   enrichment analysis. However, the statistical power and sensitivity to
   identify pathways, or gene modules (computational gene networks),
   associated with disease phenotypes can be enhanced by the use of
   non-parametric rank-based tests such as the robust positional framework
   Gene Set Enrichment Analysis (GSEA) ([39]11). Moreover, interpretation
   of transcriptional changes during COVID-19 has been primarily evaluated
   using canonical pathways that do not often reflect human responses.
   Therefore, we propose alternative strategies to analyze and interpret
   transcriptomics data, which provide novel insights into immune and
   metabolic responses during COVID-19.

Materials and Methods

Data Collection and Processing

   Datasets used in this study included public transcriptomes available at
   the Genome Sequence Archive (GSA) or human GSA in National Genomics
   Data Center, Beijing Institute of Genomics (BIG), Chinese Academy of
   Sciences for RNA-seq data related to SARS-CoV-2 infection (CRA002390
   and HRA000143); Gene Expression Omnibus (GEO) for RNA-seq data related
   to SARS-CoV-2 infection ([40]GSE147507
   ) and microarray data related to SARS-CoV-1 infection ([41]GSE1739) or
   Influenza A virus (IAV) infection ([42]GSE34205, [43]GSE6269,
   [44]GSE29366, [45]GSE38900, [46]GSE20346, [47]GSE52428, [48]GSE40012,
   [49]GSE68310, [50]GSE61754, [51]GSE90732); and ArrayExpress for
   NanoString nCounter data related to SARS-CoV-2 infection (E-MTAB-8871).
   DESeq2-normalized counts were used for the RNA-seq dataset CRA002390
   ([52]7), while raw read counts for the RNA-seq datasets [53]GSE147507
   ([54]9) or HRA000143 ([55]10) were treated and normalized to log[2]
   counts per million with EdgeR package for R ([56]12). Normalized data
   was acquired for NanoString nCounter E-MTAB-8871 ([57]8). Normalized
   microarray datasets were acquired with OMiCC platform ([58]13).
   Detailed information about the datasets used in this study are
   described in [59]Table 1.

Table 1.

   Publicly available datasets used in the study.
   Dataset ID Platform/Technology Virus infection[60]^a Sample type[61]^b
   Sample size (I/C)[62]^c Data Repository[63]^d References