Diagnosing enterovirus meningitis via blood transcriptomics: an alternative for lumbar puncture?

Background: Meningitis can be caused by several viruses and bacteria. Identifying the causative pathogen as quickly as possible is crucial to initiate the most optimal therapy, as acute bacterial meningitis is associated with a significant morbidity and mortality. Bacterial meningitis requires antibiotics, as opposed to enteroviral meningitis, which only requires supportive therapy. Clinical presentation is usually not sufficient to differentiate between viral and bacterial meningitis, thereby necessitating cerebrospinal fluid (CSF) analysis by PCR and/or time-consuming bacterial cultures. However, collecting CSF in children is not always feasible and a rather invasive procedure. Methods: In 12 Belgian hospitals, we obtained acute blood samples from children with signs of meningitis (49 viral and 7 bacterial cases). (aged between 3 months and 16 years). After pathogen confirmation on CSF, the patient was asked to give a convalescent sample after recovery. 3'mRNA sequencing was performed to determine differentially expressed genes (DEGs) to create a host transcriptomic profile. Results: Enteroviral meningitis cases displayed the largest upregulated fold change enrichment in type I interferon production, response and signaling pathways. Patients with bacterial meningitis showed a significant upregulation of genes related to macrophage and neutrophil activation. We found several significantly DEGs between enteroviral and bacterial meningitis. Random forest classification showed that we were able to differentiate enteroviral from bacterial meningitis with an AUC of 0.982 on held-out samples. Conclusions: Enteroviral meningitis has an innate immunity signature with type 1 interferons as key players. Our classifier, based on blood host transcriptomic profiles of different meningitis cases, is a possible strong alternative for diagnosing enteroviral meningitis. Overall design: Gene expression profiles of host blood can be used to differentiate between different types of meningitis. Libraries were sequenced in 4 lanes and each sequencing lane is represented as a column in the processed data files.