Journal: EBioMedicine

Article Title: AI-guided discovery of the invariant host response to viral pandemics

doi: 10.1016/j.ebiom.2021.103390

Figure Lengend Snippet: Study design. ( From top to bottom ) Step 1 : A database containing > 45,000 human, mouse and rat gene–expression data was mined to identify and validate an invariant signature for host response to viral pandemic ( ViP ) infection. ACE2, the portal for SARS-CoV-2 entry/uptake, was used as a ‘seed’ gene and Boolean Equivalent Correlated Clusters (BECC) was used as the computational method to identify gene clusters that share invariant relationships with ACE2. Once defined, these gene clusters (a.k.a., ‘ ViP signature ’) were subsequently validated across multiple human and murine models of pandemic viral infection. Step 2 : A subset of 20 genes from the ViP signature was selected that was strongly associated with severity of viral infection. These genes were validated in other cohorts to establish the ‘Severe’ ViP signature. Both 166- and 20-gene ViP signatures were validated on COVID-19 datasets. Step 3 : Cross-validation studies in numerous other datasets helped- (i) define the nature (ii) and source of the cytokine storm in COVID-19, (iii) gain insights into the immunopathology of fatal disease, and (iv) set precise therapeutic goals. Step 4 : Findings in step 3 were validated in hamsters and in a cohort of COVID-19 patients. A comprehensive catalog of the datasets analyzed in this work can be found in Supplementary Table 1 .

Article Snippet: Bottom : Bar (top) and violin (bottom) plots for the levels of IL15 cytokine (score = Z score of the log reduced mesoscale concentration data).

Techniques: Gene Expression, Infection, Biomarker Discovery

Journal: EBioMedicine

Article Title: AI-guided discovery of the invariant host response to viral pandemics

doi: 10.1016/j.ebiom.2021.103390

Figure Lengend Snippet: Validation of the ViP signatures in global pandemic viral infections. (a) Heatmap of the 166-gene signature on test dataset (GSE47963, in vitro infections of human airway epithelial cells). Genes that are involved in cytokine signaling in immune system are highlighted in the left. (b) ReacFoam analysis on the 166-gene signature that visualizes genome-wide pathway analysis based on Voronoi tessellation. (c) Reactome pathway analysis of the 166 genes in the Vip signature. (d) ViP signature-based classification of CoV-infected samples (CoV) from uninfected controls (U) in diverse human and mouse datasets. (e) Time course of CoV infection shows that the ViP host-response signature is slowly induced in very late (48–72 h) in Calu-3 cells infected with SARS-CoV-1. (f) The accuracy (Y axis; ROC AUC) of the signature to classify viral infections differs between RNA viruses and DNA viruses ( X axis) in in vitro system (top). However, they are indistinguishable in in vivo system (bottom). See also Table S3 and Fig S1. (g) ViP signature-based classification of human and murine samples with fungal or bacterial infections in either in vitro or in vivo settings. (h) The signature captures host response to CoV infection in human primary lung alveolar epithelial cells (AE) and dendritic cells (DC) better than in Fibroblasts (FI) and Endothelial (ME) cells. The accuracy of classification (ROC-AUC) strongly correlates with ACE2 expression in these cells. (i) Classification of macrophage polarization states ‘reactive’ (M1 polarized), unstimulated M0 and tolerant M2-like samples using the 166-gene ViP signature across diverse datasets.

Article Snippet: Bottom : Bar (top) and violin (bottom) plots for the levels of IL15 cytokine (score = Z score of the log reduced mesoscale concentration data).

Techniques: Biomarker Discovery, In Vitro, Genome Wide, Infection, In Vivo, Expressing

Journal: EBioMedicine

Article Title: AI-guided discovery of the invariant host response to viral pandemics

doi: 10.1016/j.ebiom.2021.103390

Figure Lengend Snippet: Identification of a ‘severe ViP’ signature. (a) Heatmap of the 166 genes on a dataset (GSE101702) annotated with varying severity of infection (healthy controls, 52; mild, 63; severe, 44). Genes are ranked based on their strength of association with severity (T-test between mild and severe). Genes that are involved in cytokine signaling in the immune system are highlighted on the left. Heatmap of top 20 selected genes (‘ severe ViP ’ signature) is shown on the right. (b) Bar and violin plots display sample rank order (i.e., classification) of patient samples and distribution of the 20-gene ‘ severe ViP ’ signature in the test dataset (GSE101702). ROC-AUC values of mild and severe cases are shown below the bar plot. (c) Reactome pathway analysis of 20 genes. (d) Bubble plots of ROC-AUC values (radius of circles are based on the ROC-AUC) demonstrating the direction of gene regulation (Up, red; Down, blue) for the classification based on the 20-gene severe ViP signature (top) and 166-gene ViP signature (bottom) in the test dataset (GSE101702), three more human datasets (H7N9, GSE114466; H1N1, GSE21802; IAV/H3N1 and others, GSE61821) and one mouse dataset (H1N1 Inf A, GSE42641). For each gene signature, ROC-AUC of controls vs Mild and Mild vs Severe are shown in top and bottom rows, respectively. In the mouse dataset (GSE42641) host response to lethal (L) and sublethal (SL) infection with H1N1 virus were assessed in five different lung cell types: Alv Mac, Lymphocytes, Monocytes, Neutrophil, Epithelial cells. Number of controls, mild and severe cases are shown at the top. (e) Summary of the 20-gene severe ViP signature and pathway analysis by DAVID GO ( https://david.ncifcrf.gov/ ).

Article Snippet: Bottom : Bar (top) and violin (bottom) plots for the levels of IL15 cytokine (score = Z score of the log reduced mesoscale concentration data).

Techniques: Infection, Virus

Journal: EBioMedicine

Article Title: AI-guided discovery of the invariant host response to viral pandemics

doi: 10.1016/j.ebiom.2021.103390

Figure Lengend Snippet: The ViP signatures define and measure the host immune response in COVID-19. (a) Heatmap of 166 genes in COVID-19 (GSE147507) dataset ranked by genes up-regulated in COVID-19 infected samples. Genes that are involved in cytokine signaling in the immune system are highlighted on the left. (b–g) Bar and violin plots displaying sample rank order (i.e., classification) and distribution of gene signature scores of COVID-19 (GSE147507) infected (CoV) and uninfected controls (C) in A549 (13 C, 6 CoV; b, e ), normal human bronchial epithelial cells (NHBE, 7 C, 3 CoV; c, f ), and patient lung autopsies (2 Normal, 1 CoV; d, g ) based on 166-gene ( b–d ) and 20-gene ViP signatures ( e–g ). (h) Bubble plots of ROC-AUC values (radius of circles are based on the ROC-AUC) demonstrating the direction of gene regulation (Up, red; Down, blue) for the classification based on the 20 gene-severe ViP signature (top) and 166 gene ViP signature (bottom) in multiple independent datasets. (i) Bubble plots like panel H showing ROC-AUC of controls vs Mild and Mild vs Severe that are shown in the top and bottom rows, respectively, for each gene signature in the COVID-19 single-cell datasets (GSE145926). Dataset is analyzed as a ‘pseudo-bulk’ of all cells or after selecting individual cell types using marker genes specifically expressed in these cell types. (j) ViP and severe ViP (sViP) signature-based classification of blood samples (GSE155363) before and up to 17 days after COVID-19 infections in 8 monkeys using ROC-AUC measurements. (k) Interferon stimulated genes (ISGs) are annotated in the ViP and sViP gene lists using Interferome v2.01 web application and displayed as Venn diagrams showing the number of genes regulated by one or more IFN types (Type I, II or III). (l) Hospital-free days analysis (45 days followup) of COVID-19 patients (GSE157103) limited to less than 70 years old using sViP signature (low and high group) is displayed as Kaplan-Meier estimates (left) of cumulative probability of discharge and its relationship with days in hospital. Cox-proportional hazard univariate analysis (right; top ) of sViP (high vs low) is compared to ViP signature, Interferon Stimulated Gene-signatures (ISG1 6 6 and ISG2 6 5 ), age, gender, ICU admission (icu) and mechanical ventilation (mv). Multivariate Cox-proportional hazard analysis (right; bottom ) compares the variables that are significant in univariate settings, i.e., sViP , ICU admission (icu) and mechanical ventilation (mv).

Article Snippet: Bottom : Bar (top) and violin (bottom) plots for the levels of IL15 cytokine (score = Z score of the log reduced mesoscale concentration data).

Techniques: Infection, Marker

Journal: EBioMedicine

Article Title: AI-guided discovery of the invariant host response to viral pandemics

doi: 10.1016/j.ebiom.2021.103390

Figure Lengend Snippet: ViP signatures reveal an interplay between IL15-storm and NK cell dysfunction in fatal COVID-19. (a) Bubble plots of ROC-AUC values (radius of circles are based on the ROC-AUC) demonstrating the direction of gene regulation (Up, red; Down, blue) for the classification based on the 20 gene severe ViP signature (top) and 166 gene ViP signature (bottom) in following datasets. RNASeq data (GSE115203) from PBMCs and sorted NK cells from PBMCs incubated with uninfected A549 cells for 12 hrs compared to infected A549 cells. PBMCs treated with IL15 compared to IL2 (GSE77601). RNASeq analysis of NK cells (GSE89484) treated with GSK-J4 compared to DMSO. Skin tissue in mice (GSE45551) is treated with anti-IL15RB antibody compared to PBS. (b) RNASeq data of NK cells isolated from two donors prior to vaccination compared (left) to days 1, 3, and 7 post-TIV vaccination like panel A. RNASeq data of NK enriched and NK depleted PBMCs from healthy donors compared to 30 day post-vaccination like panel A. (c, d) Heatmap of 20-gene (panel c ) and 166-gene (panel d ) ViP signatures in tissues collected during rapid autopsies on patients who succumbed to COVID-19. Genes are ranked according to the strength of differential expression (T-test) in lung tissue between normal and infected tissue. (e) Box plots of IL15 and IL15RA in samples from varying severity of COVID-19. (f-h) Violin plots show levels of plasma IL15 in COVID-19 patients stratified by disease acquity (F), by clinical severity (G) and by gender and age (H). Welch's two sample unpaired t -test is performed to compute the p values. See also Table S5 for patient metadata.

Article Snippet: Bottom : Bar (top) and violin (bottom) plots for the levels of IL15 cytokine (score = Z score of the log reduced mesoscale concentration data).

Techniques: Incubation, Infection, Isolation, Quantitative Proteomics, Clinical Proteomics

Journal: EBioMedicine

Article Title: AI-guided discovery of the invariant host response to viral pandemics

doi: 10.1016/j.ebiom.2021.103390

Figure Lengend Snippet: Lung alveolar cells contribute to the IL15 storm in fatal COVID-19. ( a ) Normal lung tissue obtained during surgical resection ( left ) or lung tissue obtained during autopsy studies on COVID-19 patients ( right ) were stained for IL15 and IL15RA. Representative images are shown. Mag = 10X. ( b ) Violin plots display the intensity of staining for IL15RA (top) and IL15 (bottom), as determined by IHC profiler. ( c ) Hospital-free days analysis (45 days followup) of COVID-19 patients (GSE157103) limited to males less than 70 years old using the abundance of IL15 transcripts (intermediate and high groups) is displayed as Kaplan-Meier estimates (left) of cumulative probability of discharge and its relationship with days in hospital. ( d ) Cox-proportional hazard univariate analysis (right; top ) of sViP (high vs low) is compared to ViP signature, Interferon Stimulated Gene-signatures (ISG1, PMID: 15619625 ; ISG2, PMID: 21478870 ), age, gender, ICU admission (icu) and mechanical ventilation (mv). Multivariate Cox-proportional hazard analysis (right; bottom ) compares the variables that are significant in univariate settings, i.e., sViP , ICU admission (icu) and mechanical ventilation (mv). ( e ) Top : Schematic displays the workflow for patient blood collection and assessment of IL15 levels by mesoscale. Bottom : Bar (top) and violin (bottom) plots for the levels of IL15 cytokine (score = Z score of the log reduced mesoscale concentration data). ROC AUC numbers indicate the strength of classification between patients with critical/fatal disease course vs. those with non-critical infection. ( f ) Summary of IL15 signaling and the hypothetical role of NK cells in the severity of COVID-19 infections.

Article Snippet: Bottom : Bar (top) and violin (bottom) plots for the levels of IL15 cytokine (score = Z score of the log reduced mesoscale concentration data).

Techniques: Staining, Concentration Assay, Infection

Journal: EBioMedicine

Article Title: AI-guided discovery of the invariant host response to viral pandemics

doi: 10.1016/j.ebiom.2021.103390

Figure Lengend Snippet: Validation of ViP signature-guided therapeutic goals. (a-c) The 166-gene ViP signature-was used to classify liver biopsies from HCV-infected patients treated or not with directly acting anti-viral agents. ROC-AUC values are shown below each bar plot unless otherwise stated. (d) 166-gene ViP signature-based classification of blood samples from HIV-infected patients treated with anti-retroviral therapy (ART). (e) The compound EIDD-2801 (MK-4482; 500 mg/kg) or vehicle ( Veh ) was administered at indicated doses to Golden Syrian hamsters 4 h prior to intranasal infection with SARS-CoV-2. Hamsters were sacrificed on day 5 and lungs we analyzed by RNA sequencing. (f) Bar (top) and violin (bottom) plots using the ViP (left) or sViP (right) signature-based classification of lung samples from hamsters in E and uninfected controls. (g) Schematic showing the experimental design for validating the ViP signatures as useful tools to assess therapeutic efficacy. Uninf , uninfected; Den3 and Anti-CoV-2 indicate SARS-CoV-2 challenged groups that received either a control mAb or the clone CC12.2 of anti-CoV-2 IgG, respectively. ( h ) Bar ( top ) and violin ( bottom ) plots display the 166- and 20-gene ViP signatures in the uninfected and the SARS-CoV-2 challenged groups, treated with control or anti-CoV-2 IgG. ( i-k ) Lungs harvested from the 3 groups of hamsters were analyzed by H&E and IHC. Representative images are shown in I. Mag = 10X. Bar graphs in J display the abundance of cellularity and infiltrates in the lungs of the 3 groups, as determined by ImageJ. Violin plots in K display the intensity of staining for IL15RA ( top ) and IL15 ( bottom ), as determined by IHC profiler.

Article Snippet: Bottom : Bar (top) and violin (bottom) plots for the levels of IL15 cytokine (score = Z score of the log reduced mesoscale concentration data).

Techniques: Biomarker Discovery, Infection, Retroviral, RNA Sequencing, Drug discovery, Control, Staining

Journal: EBioMedicine

Article Title: AI-guided discovery of the invariant host response to viral pandemics

doi: 10.1016/j.ebiom.2021.103390

Figure Lengend Snippet:

Article Snippet: Bottom : Bar (top) and violin (bottom) plots for the levels of IL15 cytokine (score = Z score of the log reduced mesoscale concentration data).

Techniques: Immunocytochemistry, Plasmid Preparation, Inverted Microscopy, Software, Enzyme-linked Immunosorbent Assay, Electron Microscopy, Modification