Virus-induced genetics revealed by multidimensional precision medicine transcriptional workflow applicable to COVID-19

Abstract

Precision medicine requires the translation of basic biological understanding to medical insights, mainly applied to characterization of each unique patient. In many clinical settings, this requires tools that can be broadly used to identify pathology and risks. Patients often present to the intensive care unit with broad phenotypes, including multiple organ dysfunction syndrome (MODS) resulting from infection, trauma, or other disease processes. Etiology and outcomes are unique to individuals, making it difficult to cohort patients with MODS, but presenting a prime target for testing/developing tools for precision medicine. Using multitime point whole blood (cellular/acellular) total transcriptomics in 27 patients, we highlight the promise of simultaneously mapping viral/bacterial load, cell composition, tissue damage biomarkers, balance between syndromic biology versus environmental response, and unique biological insights in each patient using a single platform measurement. Integration of a transcriptome workflow yielded unexpected insights into the complex interplay between host genetics and viral/bacterial specific mechanisms, highlighted by a unique case of virally induced genetics (VIG) within one of these 27 patients. The power of RNA-Seq to study unique patient biology while investigating environmental contributions can be a critical tool moving forward for translational sciences applied to precision medicine.

Keywords: PICU; RNAseq; multiple organ dysfunction syndrome; precision medicine; transcriptomics.

Fig. 1.

Transcriptomics workflow used to study pediatric intensive care unit (PICU) patient precision medicine. Shown in red are returnable data sets from the RNA-Seq. ECMO, extracorporeal membrane oxygenation; MODS, multiple organ dysfunction syndrome.

Fig. 2.

61 transcriptomes analyzed from 27 patients. Principal component analysis of transcriptomes annotated at the gene level for all 3 days (day 0 green, day 3 blue, day 8 magenta) (A) or on day 0 for the various groups (Sedation control blue, MODS green, MODS with ECMO red) (B). C: using the five PERSERVERE genes, we quantify the total transcripts per million (TPM, x-axis) and the percent of the five genes expressed (y-axis). Two outliers are labeled (sample:group:day:sex). Expression of various neutrophil genes (D) or cytokines (E) shown on the left as the sum of their TPM (x-axis) and the % of genes in the group expressed (y-axis). On the right is a heat map of each sample and the various genes. Clustering is based on one minus Pearson’s correlation. ECMO, extracorporeal membrane oxygenation; F, female; M, male; MODS, multiple organ dysfunction syndrome.

Fig. 3.

Blood cell composition of each sample. CIBERSORTx plots of each sample deconvoluted cell identifies shown as box and whisker plots for 16 of the cell types for day 0 (red), day 3 (orange), and day 8 (gray). Outliers are labeled (sample:group:day:sex). ECMO, extracorporeal membrane oxygenation; F, female; M, male; MODS, multiple organ dysfunction syndrome.

Fig. 4.

Bacterial/viral mapping from RNA-Seq. A: box and whisker plot for domain annotations normalized to every million reads of human for day 0 (red), day 3 (orange), and day 8 (gray). Box and whisker plots for the highest mapped bacteria (B) and samples with a significant outlier (C) reads mapped relative to the number of human mapped reads. Outliers are labeled (sample:group:day:sex). ECMO, extracorporeal membrane oxygenation; F, female; M, male; MODS, multiple organ dysfunction syndrome.

Fig. 5.

Organ-specific biomarker transcripts. A: mapping of each RNA-Seq to transcripts of ALB, shown as the summed TPM count (x-axis) and the percent of ALB transcripts identified (y-axis). B: BWA alignment of reads from patient 27 day 0 to ALB (NM_000477.7). C: correlation of ALB mapped reads to blood measured AST at day 0 (black) and day 3 (red). The black line connects the two points for patient 24 and red line for patient 27. D: Human Protein Atlas annotated bone marrow-specific genes assessed in each patient for the summed TPM and the % of genes identified above 0. E: GTEx tissue-specific genes annotated for 12 tissues. Shown for each tissue is the number of genes used for annotations next to the tissue label. The x-axis shows the summed transcript per million (TPM), and the y-axis shows the percent of tissue-specific genes with expression >0. Genes, listed in order of expression value, are shown in red for the top nonpatient 20 sample. Genes in blue are shown for small intestine, kidney, nerve, and stomach, where a small number of genes have very high expression. ECMO, extracorporeal membrane oxygenation; F, female; M, male; MODS, multiple organ dysfunction syndrome.

Fig. 6.

Transcript biomarker outliers. Far right panel: the number of transcripts within a transcriptome that are more than two standard deviations from the mean with outlier samples identified. The remaining panels show biotype annotations of elevated transcripts for each patient with the red bar represented the expected frequency of the biotype based on total presence in the transcriptome. ECMO, extracorporeal membrane oxygenation; F, female; M, male; MODS, multiple organ dysfunction syndrome.

Fig. 7.

Patient unique transcripts for day 0–8 and patient to all genes. Left: genes up (gray) or down (red) in patients at day 0 RNA-Seq relative to day 8. Right: genes up (gray) or down (red) in all 3 days RNA-Seq of patient relative to all other patients. Shown below is the plot of both day 0/8 (x-axis) and patient to all (y-axis). Labeled for each patient are the primary diagnoses and the GO enriched terms that overlap to phenotype (red and blue). The most surprising observation was the HHV-7 detected infection in patient 20, who had sarcoma, that aligns to genes decreased at day 0 that correlate to herpes-associated sarcoma. ECMO, extracorporeal membrane oxygenation; EBV, Epstein-Barr virus; FDR, false discovery rate; HLH, hemophagocytic lymphohistiocytosis; PPI, protein-protein interaction; SRP, signal-recognition particle. On the bar graph, A, B, and C are shown below for each patient.

Fig. 8.

Patient with active Epstein-Barr virus (EBV) and RNASEH2B splice variant. A: patient timeline: * marks multi-omic data collection points, red marks interventions, black marks major events, and blue marks phenotypes observed. B: clustering of control (red) and patient samples (green), including a 22 mo follow-up. C: second viral read caller to confirm EBV infection with EBER1 (gray) and EBER2 (red) reads from EBV detected at day 0 shown below. D: known structure of RNASEH2A (gray), RNASEH2C (magenta), and RNASEH2B (cyan). E: zoom-in view of V20 not falling at a critical site suggesting L would be functional. F: percent of reads from the RNASEH2B alleles at multiple RNA-Seq measurements suggesting allelic bias to the allele inherited from the father. G: reads aligned near the potential splice site variant inherited from mom that confirms a frameshift and early truncation. H: total RNASEH2B levels over the multiple days suggesting reduced levels relative to controls (red). I: pathways activated in the patient’s 2 yr follow-up, suggesting immune system dysfunction. J: immune system repertoire analysis performed on the patient follow-up for seven chains (TRB, T cell receptor beta; TRA, T cell receptor alpha; IGH, immunoglobulin heavy chain; TRD, T cell receptor delta; TRG, T cell receptor gamma; IGK, immunoglobulin light chain kappa; IGL, immunoglobulin light chain lambda). Shown is the percent of CDR3 identified in each group with a callout for further dissection of the IGH subgroups. K: diversity index (DI) for each of the sequenced chains. DI corresponds to elevated recombination sites, with lower levels suggesting an enrichment of certain sequences within the chain. The IGL is shown for diversity, suggesting the elevation of epitope recognition. This epitope is seen in 27 of the top 35 CDR3 sequences for IGL with a significant enrichment (8.9e-82) and composing 21.7% of all the CDR3 IGL sequences. CRP, C-reactive protein; ECMO, extracorporeal membrane oxygenation; ER, emergency room; HDVCH, Helen DeVos Children’s Hospital; HLH, hemophagocytic lymphohistiocytosis; PICU, pediatric intensive care unit; WES, whole exome sequencing.