Secondary analysis of transcriptomes of SARS-CoV-2 infection models to characterize COVID-19

•Defined a consensus gene signature across three models of SARS-CoV-2 infection•Characterized subnetworks host proteins interacting with proteome•Integrated wide range COVID-19 and related data to build functional modules•Identified modules that can further the understanding This study is based on premise combining information from multiple layers result in new biologically interpretable associations several ways. The underlying unifying theme this integration, mining, meta-analysis for pattern detection supports knowledge discovery generation hypotheses. methods workflow used are disease agnostic be applied any or phenotype has heterogeneous elements. By integrating joint analysis types (multiple models, viral-host protein interaction data, single-cell RNA-sequencing protein-protein interactions, genome-wide association data), identified have direct bearing furthering COVID-19. Standard transcriptomic analyses alone limited power capturing molecular mechanisms driving pathophysiology outcomes. To overcome this, unsupervised network identify clusters genes associated distinct outcomes disease. In study, we developed an integrated framework integrates transcriptional signatures model systems find modules. Through different enriched features these modules, extract communities highly interconnected features. These higher-order features, working as multifeatured machine, enable collective assessment their contribution characterization. We show utility using transcriptomics infection pathways biological processes could hypothesizing inducing pathophysiological changes, risks, sequelae vitro vivo often fail completely recapitulate manifestations humans. Integrated secondary approaches disease-related by leveraging physiological functions Functional complexes arise out known represent functions.1Spirin V. Mirny L.A. Protein networks.Proc. Natl. Acad. Sci. U S A. 2003; 100: 12123-12128Crossref PubMed Scopus (1151) Google Scholar,2Barabasi A.L. Oltvai Z.N. Network biology: cell's organization.Nat. Rev. Genet. 2004; 5: 101-113Crossref (5745) Scholar Similarly, feature networks comprising processes, pathways, phenotypes, cell machines collectively toward common goal. Based premise, implemented multilayered data-mining methodology leverages built linking together such (RNA-seq) markers, phenotype-genotype associations. demonstrate approach, analyzed two (Calu-3 Vero E6 cells) one (Ad5-hACE2-sensitized mice) infection. Coronavirus 2019 (COVID-19), caused SARS-CoV-2, affected more than 75 million people 1.6 deaths worldwide including ∼17.3 confirmed infections >311,000 United States (World Health Organization, December 20, 2020). emerging stages surrounding disease, necessity effective interventions (e.g., vaccines, small molecules), provides strong rationale multilayered, existing collected studies. Some noteworthy discoveries offshoots available omics generated pre-COVID-19 times. include RNA-seq (scRNA-seq) data3Ziegler C.G.K. Allon S.J. Nyquist S.K. Mbano I.M. Miao V.N. Tzouanas C.N. Cao Y. Yousif A.S. Bals J. Hauser B.M. et al.SARS-CoV-2 receptor ACE2 interferon-stimulated human airway epithelial cells detected specific subsets tissues.Cell. 2020; 181: 1016-1035.e19Abstract Full Text PDF (1361) Scholar,4Sungnak W. Huang N. Becavin C. Berg M. Queen R. Litvinukova Talavera-Lopez Maatz H. Reichart D. Sampaziotis F. entry factors expressed nasal innate immune genes.Nat. Med. 26: 681-687Crossref (1501) Human Cell Atlas consortium eQTL variant data5Cao Li L. Feng Z. Wan S. P. Sun X. Wen Ning G. Wang Comparative genetic novel coronavirus (2019-nCoV/SARS-CoV-2) populations.Cell Discov,. 6: 11Crossref (547) Genotype Tissue Expression (GTEx) database.6Consortium G.T. genomics. Genotype-Tissue pilot analysis: multitissue regulation humans.Science. 2015; 348: 648-660Crossref (3055) Thus, repository datasets information, even if they were not designed specifically COVID-19, provide jump start discover sides Recently there been studies reporting analysis-based both COVID-19- non-COVID-19-related detect tissue-specific7Hernandez Cordero A.I. Yang C.X. Milne Bosse Joubert Timens van den Berge Nickle Hao K. al.Gene expression potential targets against SARS-CoV-2.Sci. Rep. 10: 21863Crossref (7) Scholar,8Guzzi P.H. Mercatelli Ceraolo Giorgi F.M. Master regulator SARS-CoV-2/human interactome.J. Clin. 9: 982Crossref (112) pan-tissue9Feng Q. Identifying profiles normal SARS-CoV-2-infected tissues.Front. Mol. Biosci. 7: 568954Crossref (13) infections. differ input “seed” construct networks; some focused entry-associated receptors and/or proteases7Hernandez Scholar,9Feng while others use expanded set virus-host interactants SARS-CoV-2.8Guzzi Scholar,10Nadeau Shahryari Fard Scheer Hashimoto-Roth E. Nygard Abramchuk I. Chung Y.-E. Bennett S.A.L. Lavallée-Adam computational identification sequence motifs putatively targeted networks.J. Proteome Res. 19: 4553-4566Crossref However, most do consider differentially (DEGs) following analysis. A recently published study11Ahmed network-based reveals mechanism vitamin D suppressing cytokine storm virus infection.Front. Immunol. 11: 590459Crossref (36) bronchial (NHBE) along protease TMPRSS2 regulatory network. however, single (three samples) samples. Additionally, did other host-virus virus. limitations address issues, (two vivo) infection, them jointly non-COVID-19/SARS-CoV-2 data. For latter, scRNA-seq markers lung studies, (GWAS) (Figure 1). While acknowledge complexity believe our formulation testable hypotheses pathogenesis. viral complexes, which manipulated hijack individual processes. Therefore, evaluate phenomenon, first interactome around transcriptome obtain signature, considered DEGs at least models12Blanco-Melo Nilsson-Payant B.E. Liu W.C. Uhl Hoagland Moller Jordan T.X. Oishi Panis Sachs al.Imbalanced response drives development COVID-19.Cell. 1036-1045 e9Abstract (2262) Scholar, 13Riva Yuan Yin Martin-Sancho Matsunaga Pache Burgstaller-Muehlbacher De Jesus P.D. Teriete Hull M.V. al.Discovery antiviral drugs through large-scale compound repurposing.Nature. 586: 113-119Crossref (439) 14Sun Zhuang Zheng Wong R.L. He Zhu Zhao al.Generation broadly useful pathogenesis, vaccination, treatment.Cell. 182: 734-743 e5Abstract (256) compared (i.e., lines, namely, transformed lung-derived Calu-3 VeroE6 cells, mouse model) 2A Table concordance was seen among upregulated downregulated models. total 732 (537 195 downregulated) shared between non-human primate lines 2B). found 325 369 Ad5-hACE2-sensitized mice. overall DEGs, each also had unique 2C S1). validated comparing patients (GEO: GSE152075; nasopharyngeal swabs 430 54 controls).15Lieberman N.A.P. Peddu Xie Shrestha M.L. Mears M.C. Cajimat M.N. Bente D.A. Shi P.Y. Bovier al.In load, sex, age.PLoS Biol. 18: e3000849Crossref There stronger line Finally, 1,467 (833 634 S2) included 106 41 all (Figures 2D), representing “core” dysregulated Both sets terms (Tables S3, S4, S5) cell-type (Table S6 Figure S2). pathological traits (from Phenotype-Genotype Integrator [PheGenI]16Ramos E.M. Hoffman Junkins H.A. Maglott Phan Sherry S.T. Feolo Hindorff (PheGenI): synthesizing genomic resources.Eur. Hum. 2014; 22: 144-147Crossref (130) GWAS catalog17Buniello MacArthur J.A.L. Cerezo Harris L.W. Hayhurst Malangone McMahon Morales Mountjoy Sollis al.The NHGRI-EBI catalog arrays summary statistics 2019.Nucleic Acids 2019; 47: D1005-D1012Crossref (1601) databases) S7 S8; S3).Table 1List (0.6 logFC; FDR p ≤ 0.05) modelsDifferentially list nameNo. DEGsGEO IDReferenceCalu3 SARS-CoV-2: downregulated2,272GSE147507Blanco-Melo al.12Blanco-Melo ScholarCalu3 upregulated2,509Ad5-hACE2-sensitized downregulated2,109GSE150847Riva al.13Riva ScholarAd5-hACE2-sensitized upregulated1,217Vero downregulated953GSE153940Sun al.14Sun ScholarVero upregulated1,369Overall number DEGs: 8,286 Open table tab interactome, SARS-CoV-2-human (PPI) dataset 332 involved assembly trafficking RNA.18Gordon D.E. Jang G.M. Bouhaddou Xu Obernier White K.M. O'Meara M.J. Rezelj V.V. Guo J.Z. Swaney D.L. al.A map drug 583: 459-468Crossref (2261) addition ACE2, proteases, TMPRSS2, CTSB, CTSL. More half (151 genes) 336 3A). Of these, 29 (16 13 part signature. Using SARS-CoV-2-proteome input, queried STRING (v11) database19Szklarczyk Gable Lyon Junge Wyder Huerta-Cepas Simonovic Doncheva N.T. Morris J.H. Bork al.STRING v11: increased coverage, supporting experimental datasets.Nucleic D607-D613Crossref (7428) DEG-PPI Only interactions highest confidence score (0.9) 0.7 used. observed enrichment PPIs (p < 1.0 × 10−16) combined 3B). words, significantly themselves would expected random similar size drawn genome. next Markov clustering (MCL) algorithm. brief, MCL determine intramodular (within module) intermodular (with modules) interactions. Each only assigned module method. inflation factor parameter determines granularity (or “tightness”) thereby cluster size. experiments default (2.5). With clustering, 153 varying counts S9). selected 35 candidate having five genes. made up 797 627 (see Figures 3C–3H six example S9 details). clusters, gene-encoding interacts proteome. hypothesize SARS-CoV-2-targeted informative deciphering inferring function members clusters. step approach S10). Cluster C-1 (190 (48 type I interferon signaling (26 C-2 (92 transport (31 tube genes). abnormal cardiovascular (21 C-2. Clusters C-7 (20 C-8 interleukin secretion phenotypes. C-12 (14 genes), C-28 (6 C-23 (8 mitochondrion translation, organization, transport. regulating circadian rhythm mammals (NFIL3, PER1, PER2, PER3, SIK1) C-25 (7 evaluated performing marker compiled studies.20Habermann A.C. Gutierrez A.J. Bui L.T. Yahn S.L. Winters N.I. Calvi C.L. Peter M.-I. Taylor C.J. Jetter al.Single-cell RNA sequencing profibrotic roles mesenchymal lineages pulmonary fibrosis.Sci. Adv. eaba1972Crossref (241) 21Adams T.S. Schupp J.C. Poli Ayaub E.A. Neumark Ahangari Chu S.G. Raby B.A. DeIuliis Januszyk ectopic aberrant lung-resident populations idiopathic eaba1983Crossref (285) 22Travaglini K.J. Nabhan A.N. Penland Sinha Gillich Sit R.V. Chang Conley S.D. Mori Seita atlas sequencing.Nature. 587: 619-625Crossref (356) 17 (633 4 2). proliferating basal), lymphoid T natural killer cells), myeloid macrophages) types. showing epithelial, mesenchyme, vascular endothelial, lymphoid, C-9 (18 showed fibroblasts, myofibroblasts, smooth muscle enrichments C-1, C-2, C-3. certain Ionocyte marker22Travaglini genes, instance, C-5 (40 genes; 12 markers); C-7, C-11, C-13 S11).Table 2Candidate DEG typesClusterEnriched markersC-1 genes)proliferating killer/T basal, macrophage, adventitial alveolar 1C-2 (91 ciliated, classical monocytes, 1, fibroblastsC-3 (81 genes)adventitial lipofibroblasts, vessel 2, mast cellsC-4 (73 genes)ciliated, capillary endothelial cellsC-5 genes)ionocytes, proximal ciliatedC-6 (34 genes)bronchial mesothelialC-7 genes)dendritic monocytesC-9 genes)alveolar cellsC-10 (17 genes)lymphatic, peribronchial, arterialC-11 (15 genes)dendritic, cellsC-13 genes)classical monocytesC-18 (10 cellsC-22 macrophages, cellsC-30 cellsC-31 genes)ArteriesC-34 (5 genes)plasma cellsC-35 cellsClusters ≥5 shown. complete see S11, respectively. phenotypic insights into pathogenesis S12 S13). Among respiratory system (clusters C-8), asthma (C-7), autoimmune C-29), allergic rhinitis (cluster diabetes (C-15). risk inflammatory disorders bowel Crohn's ulcerative colitis (C-8), rheumatoid arthritis ankylosing spondylitis (C-8). Apart elucidating potentially help researchers understand formulate long-hauler survivors. neurodegenerative disease? plausible overactivated system.23Lehnardt Massillon Follett Jensen F.E. Ratan Rosenberg P.A. Volpe J.J. Vartanian T. Activation immunity CNS triggers neurodegeneration Toll-like 4-dependent pathway.Proc. 8514-8519Crossref (841) 24Godbout J.P. Chen Abraham Richwine A.F. Kelley K.W. Johnson R.W. Exaggerated neuroinflammation sickness behavior aged mice activation peripheral system.FASEB 2005; 1329-1331Crossref (623) 25Elson C.O. Cong McCracken V.J. Dimmitt R.A. Lorenz R.G. Weaver C.T. Experimental reveal innate, adaptive, dialogue microbiota.Immunol. 206: 260-276Crossref (408) acute delayed neurological neuropsychiatric effects previous pandemics.26Fazzini Fleming Fahn Cerebrospinal fluid antibodies Parkinson's disease.Mov Disord. 1992; 153-158Crossref (136) Scholar,27Troyer Kohn J.N. Hong Are facing crashing wave COVID-19? Neuropsychiatric symptoms immunologic mechanisms.Brain Behav. Immun. 87: 34-39Crossref (512) proposed same another corona SARS-CoV-1. extracted SARS-CoV-1 models28Sims Tilton S.C. Menachery V.D. Gralinski L.E. Schafer Matzke M.M. Webb-Robertson B.J. Luna Long C.E. al.Release severe syndrome nuclear import block enhances transcription cells.J. Virol. 2013; 3885-3902Crossref (84) 29Regla-Nava J.A. Nieto-Torres J.L. Jimenez-Guardeno J.M. Fernandez-Delgado Fett Castano-Rodriguez Perlman Enjuanes DeDiego Severe coronaviruses mutations E attenuated promising vaccine candidates.J. 89: 3870-3887Crossref (89) 30Totura Whitmore Agnihothram Katze M.G. Heise M.T. Baric R.S. 3 via TRIF contributes protective infection.mBio. e00638-15Crossref (303) (Calu3 models), DEGs. 699 1,385 systems. generate SARS-CoV-1-targeted 366 host-SARS-CoV-1 basis localization cells.31Gordon Hiatt Ulferts Braberg Jureka Batra al.Comparative host-coronavirus pan-viral mechanisms.Science. 370: eabe9403Crossref Comparing over 300 (196 119 135 shared, large them. steps described previously SARS-CoV-2. 68 S14). computed analyzing compositions (SARS-CoV-2 SARS-CoV-1), viruses. contained 90% its (37 40) Interestingly, ionocytes ciliated pathways. 9 11 C-15