key: cord-0735372-crm82ds1 authors: Hunter, Ewan; Koutsothanasi, Christina; Wilson, Adam; Santos, Francisco C.; Salter, Matthew; Powell, Ryan; Dring, Ann; Brajer, Paulina; Egan, Benedict; Westra, Jurjen W.; Ramadass, Aroul; Messer, William; Brunton, Amanda; Lyski, Zoe; Vancheeswaran, Rama; Barlow, Andrew; Pchejetski, Dmitri; Robbins, Peter A.; Mellor, Jane; Akoulitchev, Alexandre title: 3D genomic capture of regulatory immuno-genetic profiles in COVID-19 patients for prognosis of severe COVID disease outcome date: 2021-03-16 journal: bioRxiv DOI: 10.1101/2021.03.14.435295 sha: c2fff2e8772a59180e396988876767a874123b21 doc_id: 735372 cord_uid: crm82ds1 Human infection with the SARS-CoV-2 virus leads to coronavirus disease (COVID-19). A striking characteristic of COVID-19 infection in humans is the highly variable host response and the diverse clinical outcomes, ranging from clinically asymptomatic to severe immune reactions leading to hospitalization and death. Here we used a 3D genomic approach to analyse blood samples at the time of COVID diagnosis, from a global cohort of 80 COVID-19 patients, with different degrees of clinical disease outcomes. Using 3D whole genome EpiSwitch® arrays to generate over 1 million data points per patient, we identified a distinct and measurable set of differences in genomic organization at immune-related loci that demonstrated prognostic power at baseline to stratify patients with mild forms of illness and those with severe forms that required hospitalization and intensive care unit (ICU) support. Further analysis revealed both well established and new COVID-related dysregulated pathways and loci, including innate and adaptive immunity; ACE2; olfactory, Gβψ, Ca2+ and nitric oxide (NO) signalling; prostaglandin E2 (PGE2), the acute inflammatory cytokine CCL3, and the T-cell derived chemotactic cytokine CCL5. We identified potential therapeutic agents for mitigation of severe disease outcome, with several already being tested independently, including mTOR inhibitors (rapamycin and tacrolimus) and general immunosuppressants (dexamethasone and hydrocortisone). Machine learning algorithms based on established EpiSwitch® methodology further identified a subset of 3D genomic changes that could be used as prognostic molecular biomarker leads for the development of a COVID-19 disease severity test. . 165 lists from Limma and RankProd analysis were compared and the intersection 219 of the two lists was selected for further processing. 220 To evaluate the biological relevance of the observed separation of Mild and 273 Severe COVID-19 outcomes by 3D genomic markers, we focused on the top 274 750 markers prognostically associated with each of the two outcomes. The 275 statistical testing employed in this study to determine statistically significant 3D 276 genomic markers benefits from using both parametric testing (Limma) and non-277 parametric testing (EpiSwitch ® RankProd), both procedures that correct for 278 multiple testing by using False Discovery Rate (FDR) corrections. The 279 RankProd approach also has a resampling step to control for random rank 280 Next, we used the network of genes under 3D genomic control for susceptibility 490 to severe COVID-19 outcome to evaluate the existing drugs with known gene 491 targets. We were trying to evaluate them as potential therapeutic tools for 492 mitigation of severe disease outcomes. Using GeneAnalytics we identified 25 493 drug candidates with potential utility for treating COVID-19 disease ( Table 3) . 494 Interestingly, the analysis based on 3D genomic profiling of severe COVID-19 495 patients identified at the second highest score Dexamethasone, which has 496 been now reported as beneficial in reducing mortality among severe patients 497 [61]. 498 499 CXCL2, FLT1, RPS6KB1, CD28, GRB7, CCR7, PIK3R1, BIRC3, IL18, NF1, FCER2, IL6ST, PRKCA, CD27, LGALS9, DPP4, VEGFC, RPS6, CCL3, DMBT1, PRKCB, STK11, REG1A, PDPK1, PIK3CB, IL7, LYN, BLNK, TLR4, MAPK3, MAPK1 26.202619 CCL3 63 7 CXCL2, IL6, CCR7, IL18, CCL3, TNF, TLR4 25.456246 Trastuzumab 68 7 PIK3R1, H4-16, VEGFC, CCL3, VEGFA, RAF1, MAPK1 25.012201 Everolimus 41 6 RPS6KB1, TSC1, PIK3R1, MYC, NF1, VEGFA 24.336196 Immunologic Factors 117 8 FLT3, TSC1, CD28, MYC, NF1, DPP4, CD22, TNF 23.525819 Sorafenib Tosylate 9 4 FLT1, VEGFC, VEGFA, RAF1 22.936924 Rosiglitazone 133 8 RPS6KB1, IL6, STAT1, IL18, DPP4, INS, TNF, MAPK1 22.849199 IL 10 53 6 IL6, IL18, CCL3, TNF, IL7, TLR4 22.776058 Imiquimod 27 5 IL6, STAT1, IL18, TNF, TLR4 22.457678 Hydrocortisone 139 8 FLT1, IL6, FCER2, CCL3, INS, TNF, IL7, KLK3 502 a Score: The binomial distribution is used to test the null hypothesis that the 503 input genes are not over-represented within any SuperPath. GO term or 504 compound. The presented score for each match is a transformation (-log2) of 505 the resulting p-value, where higher scores indicate better matches. Results with 506 p-values lower than 10-50 are assigned the maximum score. LDA and XGBoost algorithms to build classifier models was done to reduce the 532 potential of data overfitting by either model. In addition to using multiple 533 machine learning approaches to build classifying models, 10-fold cross 534 validation was utilized to ensure robust and rigorous training models were built 535 especially for XGBoost, this was carried out using the caret package in R. Both 536 models were built using the top 50 3D genomic markers associated with Severe 537 We used whole 3D genome arrays to analyse molecular profiles across 1.1 603 million sites per patient from 80 COVID-infected patients from around the world. In agreement with earlier observations, the regulatory 3D genome contained a 605 stable network of 3D genomic alterations tightly associated with distinct clinical 606 outcomes. The 3D genomic changes identified here provide further insight into 607 functional networks of genome regulation after SARS-CoV-2 infection, as many 608 were associated with genomic loci implicated in immune regulation. By 609 biological network analysis, we identified several immune-related pathways that An interactive web-based dashboard to track COVID-19 in real time. The Lancet Infectious Diseases Severe Covid-19 The mystery of the pandemic's 'happy hypoxia Do I need to Take Extra Precautions Against COVID-19 | CDC How COVID-19 induces cytokine storm with high mortality Clinical features of 800 patients infected with 2019 novel coronavirus in Wuhan Cytokine release syndrome in 803 severe COVID-19: interleukin-6 receptor antagonist tocilizumab may be the 804 key to reduce mortality COVACTA trial raises questions about tocilizumab's benefit in COVID-19 Regeneron and Sanofi Provide Update on Kevzara® (sarilumab) Phase 3 Trial in COVID-19 Patients Distinct inflammatory profiles distinguish COVID-19 from influenza with 811 limited contributions from cytokine storm Is a "cytokine Storm" Relevant to 813 COVID-19? Tocilizumab for Covid-19 -The Ongoing Search 815 for Effective Therapies Association between Early Treatment with Tocilizumab and Mortality among Critically Ill Patients with COVID-19 The genome 820 conformation as an integrator of multi-omic data: The example of damage 821 spreading in cancer Transcription factors and 3D genome 825 conformation in cell-fate decisions Dissecting super-827 enhancer hierarchy based on chromatin interactions The 829 human noncoding genome defined by genetic diversity Methods for mapping 3D chromosome architecture Rac activation by the T-cell receptor inhibits T cell migration Dexamethasone in Hospitalized Patients with Covid-19 -Preliminary COVID-19 and the human innate immune 950 system Chromatin 952 conformation signatures: ideal human disease biomarkers The genome 955 conformation as an integrator of multi-omic data: The example of damage 956 spreading in cancer Pathway: Potential Regulator of Inflammation and Innate Immunity. Frontiers 959 in Immunology PI3K signalling in inflammation PI3Kδ and PI3Kγ: Partners in crime in 963 inflammation in rheumatoid arthritis and beyond? Nitric oxide synthase in innate and adaptive immunity: An 966 update The role of nitric oxide in innate 968 immunity What We Know So Far SARS-CoV-2 infection: The role of cytokines in COVID-19 Cytokine and Growth Factor Reviews Immunoregulation with mTOR inhibitors to prevent COVID-19 severity: A novel intervention strategy beyond vaccines and 986 specific antiviral medicines Dexamethasone treatment for the acute respiratory distress syndrome: a 989 multicentre, randomised controlled trial Clinical Trial to Evaluate Methylprednisolone Pulses Tacrolimus in Patients With COVID-19 Lung Injury. Open Randomized Single Cent Clin Trial to Eval Methylprednisolone Pulses Tacrolimus Patients With Sev Lung Inj Second to COVID-19 Association between Administration of Systemic Corticosteroids and Mortality 996 among Critically Ill Patients with COVID-19: A Meta-analysis Molecular mechanisms 999 underlying prostaglandin E2-exacerbated inflammation and immune diseases Prostaglandin E2 inhibits production of the 1002 inflammatory chemokines CCL3 and CCL4 in dendritic cells Prostaglandin E2 suppresses chemokine production in human 1006 macrophages through the EP4 receptor Non-steroidal anti-inflammatory 1008 drugs, prostaglandins, and COVID-19 Celebrex adjuvant 1010 therapy on COVID-19: An experimental study. medRxiv Hypothesis: mPGES-1-derived prostaglandin E2, a so far missing link in 1013 COVID-19 pathophysiology? Preprints Rapid review: Nonsteroidal anti-inflammatory 1015 agents and aminosalicylates in COVID-19 infections Estimating real-world COVID-19 vaccine effectiveness in Israel Patterns of 1020 COVID-19 pandemic dynamics following deployment of a broad national 1021 immunization program. medRxiv There is a high need for high content technologies that could be reduced to 720 practice for assessment of cellular, systemic host profiles to identify potential 721 biomarkers and underlying molecular network mechanisms addressing 722 outcome prediction. Taking an array-based EpiSwitch ® approach, we identified 723 blood-based differences in 3D immuno-genetic molecular profiles of patients 724 with mild forms of COVID-19 compared to those with more severe clinical 725 outcomes. Our results suggest that 3D immuno-genetic profiles in multi-cohort 726 study could be used as a biomarker modality for COVID specific disease 727 classification and outcome prediction. 728 reduction in PCR format for the initial findings and biomarker leads described 731here, use of an extended patient cohort and an independent cohort validation 732 will serve to further validate the biological relevance of the 3D genomic changes 733 and could help to establish the minimal set of prognostic markers that could 734 robustly stratify patients. Ultimately, as a next step, a reduced set of markers in 735