Glia-Associated Secretome Signature Identifies Six CSF Proteins as Candidate Biomarkers for Temporal Lobe Epilepsy
Background
Despite advances in multi-omics, clinically accessible and stable cerebrospinal fluid (CSF) biomarkers for epilepsy remain scarce. Current diagnostic methods for temporal lobe epilepsy (TLE) often rely on imaging and electrophysiology, which can be inconclusive or invasive. Identifying reliable, biologically relevant biomarkers could significantly improve diagnosis, prognosis, and therapeutic monitoring by reflecting underlying molecular alterations like neuroinflammation and glial activation.
Study Design
Researchers employed a multi-layered approach, starting with a kainic acid (KA)-induced mouse model to identify epilepsy-associated secreted proteins. This was combined with single-cell RNA sequencing and public transcriptomic datasets to pinpoint candidates linked to inflammation and glial activation. Candidate biomarkers were then validated across mouse and human datasets. Finally, these proteins were quantified in the CSF of TLE patients and non-epileptic controls using enzyme-linked immunosorbent assays (ELISA). Machine learning models were developed to assess diagnostic performance and interpretability.
Results
Six specific proteins were consistently dysregulated in human CSF from TLE patients. CSF levels of C-X-C motif chemokine ligand 10 (CXCL10), apolipoprotein E (APOE), apolipoprotein D (APOD), galectin-3-binding protein (LGALS3BP), and lysozyme (LYZ) were significantly elevated, while transthyretin (TTR) was markedly reduced (all p < 0.001).
Key Findings
- Six CSF proteins (CXCL10, TTR, APOE, APOD, LGALS3BP, LYZ) were dysregulated in human TLE patients.
- CSF CXCL10, APOE, APOD, LGALS3BP, and LYZ levels were significantly elevated in TLE (all p < 0.001).
- CSF TTR levels were markedly reduced in TLE patients (p < 0.001).
- Machine learning models showed promising diagnostic performance, with TTR and CXCL10 as major features.
- CSF CXCL10 levels positively correlated with preoperative seizure frequency (Spearman's ρ = 0.698, p < 0.0001).
Why It Matters
Identifying a robust, glia-associated secretome signature in CSF offers a promising avenue for non-invasive temporal lobe epilepsy diagnosis and prognosis. These candidate biomarkers could enable earlier, more accurate diagnosis and potentially guide personalized treatment strategies for TLE patients. The strong correlation of CXCL10 with seizure frequency suggests its potential as a severity or progression marker, which could be crucial for monitoring disease activity and treatment response, moving beyond subjective patient reports or infrequent EEG monitoring.
temporal-lobe-epilepsy
biomarkers
csf
neuroinflammation
glial-activation
machine-learning