Objective:

To identify cerebrospinal fluid (CSF) gene expression signatures that differentiate infectious encephalitis (IE) from autoimmune encephalitis (AE) using metagenomic next-generation sequencing (mNGS).

Background:

Rapid and accurate distinction between IE and AE is critical given earlier treatment improves outcomes, yet immunosuppression poses significant risks in cases of undiagnosed infection. Current diagnostic tools – such as targeted pathogen-specific testing for IE and antibody panels for AE – are limited by long turnaround times and risks of false negatives. CSF host transcriptomics analysis via mNGS provides a novel approach by directly measuring the host immune response through changes in gene expression and has the potential to identify biomarkers and immune mechanisms to more rapidly and accurately differentiate IE from AE.

Design/Methods:

CSF bulk RNA-sequencing was performed in a cohort of patients with antibody-positive AE and confirmed IE. Differential gene expression and gene ontology (GO) pathway analyses were conducted to identify immune signatures distinguishing the two cohorts.

Results:

Forty-seven patients were included for analysis: 11 with IE (all viral) and 36 with AE (21 anti-NMDAr, 9 GAD65, 3 LGI1, 3 GABA-B). Median age was 10 (45% male) and 22.5 years (36% male) for IE and AE, respectively. A total of 1,310 genes were differentially expressed between IE and AE (adjusted p<0.05). Of these, 805 genes were upregulated in IE, including key antiviral and type 1 interferon-related genes. Top enriched GO pathways included “defense response to virus”, “regulation of the innate immune response”, and “interferon-beta production”.

Conclusions:

IE and AE demonstrate distinct CSF gene expression profiles. IE is characterized by robust activation of antiviral and interferon-mediated pathways. CSF host transcriptomics may serve as a powerful diagnostic tool to differentiate encephalitis etiologies and guide early treatment. Ongoing work is the application of machine learning approaches to develop a predictive classifier for IE versus AE.