Prediction, prevention, and precision treatment of immune checkpoint inhibitor neurological toxicity using autoantibodies, cytokines, and microbiota.

Cancer immunotherapy with immune checkpoint inhibitors (ICIs) has revolutionized oncology, significantly improving survival across multiple cancer types. ICIs, such as anti-PD-1 (e.g. nivolumab, pembrolizumab), anti-PD-L1 (e.g. atezolizumab, avelumab), and anti-CTLA-4 (e.g. ipilimumab), enhance T cell-mediated anti-tumor responses but can also trigger immune-related adverse events (irAEs). Neurological irAEs (n-irAEs), affecting 1-3% of patients, predominantly involve the peripheral nervous system; less commonly, n-irAEs can present as central nervous system disorders. Although irAEs suggest a possible correlation with treatment efficacy, their mechanisms remain unclear, with hypotheses ranging from antigen mimicry to cytokine dysregulation and microbiome alterations. Identifying patients at risk for n-irAEs and predicting their outcome through biomarkers would be highly desirable. For example, patients with high-risk onconeural antibodies (such as anti-Hu or Ma2), and elevated neurofilament light chain (NfL) levels often respond poorly to irAE treatment. However, interpreting neuronal antibody tests in the diagnosis of n-irAEs requires caution: positive results must align with the clinical context, as some cancer patients (e.g., SCLC) may have asymptomatic low antibody levels, and false positive results are common without tissue-based confirmation. Also, the use of biomarkers (e.g. IL-6) may lead to more targeted treatments of irAEs, minimizing adverse effects without compromising the anti-tumor efficacy of ICIs. This review provides a comprehensive overview of the latest findings on n-irAEs associated with ICIs, with a focus on their prediction, prevention, as well as precision treatment using autoantibodies, cytokines, and microbiota. The most interesting data concern neuronal antibodies, which we explore in their pathogenic roles and as biomarkers of neurotoxicity. Most of the available data on cytokines, both regarding their role as diagnostic and prognostic biomarkers and their role in supporting therapeutic decisions for toxicities, refer to non-neurological toxicities. However, in our review, we mention the potential role of CXCL10 and CXCL13 as biomarkers of n-irAEs and describe the current evidence, as well as the need for further studies, on the use of cytokines in guiding selection of second-line therapies for n-irAEs. Finally, no specific microbiome-related microbial signature has been proven to be linked to n-irAEs specifically, leading to the need of more future research on the topic.