Identification of Conserved Detrimental Host Immune Response Predicts Severity of Bacterial and Viral Infections

Introduction: Host immune response has been repeatedly shown to diagnose the presence and type of infection. Recently, we described a 42-gene blood-based signature, conserved across several viruses, including influenza, Ebola, SARS-CoV-2, chikungunya, that is associated with and predicts the severity of viral infection, irrespective of age, sex, and host or pathogen genetics. Importantly, we showed the 42-gene signature is composed of 4 modules (2 protective, 2 detrimental). We investigated whether these modules, individually or collectively, are also associated with severity in patients with a bacterial infection. Methods: We analyzed 29 publicly available datasets comprised of blood transcriptome profiles from 1,806 patients (637 healthy patients, 1169 patients with bacterial infection) from 10 countries. We co-normalized these datasets using COCONUT. We also included 3,183 blood samples across an additional 29 datasets from 18 countries from patients with viral infection (1,663 healthy patients, 1,520 patients with viral infection) from our previous study. Severity of disease was stratified into healthy controls, asymptomatic infection, mild, moderate, serious, critical and fatal illness. We assessed the performance of our previously described four module scores and a composite severe-or-mild “SoM” score in these samples. We then applied our previously described 7 gene signature (7GS) that distinguishes viral from bacterial infections to both the bacterial and viral samples. Results: Similar to viral infections, the two detrimental module scores were positively correlated with severity of bacterial infections (module 1: r=0.64, module 2: r=0.53), and one of two protective modules was inversely correlated (module 4: r=-0.59). Module 3, initially protective in viral infections, was minimally positively correlated with severity of bacterial illness (module 3: r=0.20). The SoM score, which integrates the four module scores, was positively correlated with severity (r=0.63) and distinguish mild/moderate bacterial infections from severe (serious/critical/fatal) bacterial infection with 71% sensitivity and 73% specificity (Figure 1A, AUROC=0.77, 95% CI:0.73-0.80). Interestingly, the SoM score was not different between patients with severe bacterial or viral infection, but was significantly higher in mild/moderate bacterial infections compared to mild/moderate viral infections. Conclusion: The SoM score can accurately distinguish patients with severe infection, irrespective of bacterial or viral infection. When used in conjunction with our previously described 7-gene signature, it may help decide whether a patient should be (1) treated with an antibiotic and (2) discharged or admitted to hospital upon presentation to an emergency department.