2026 American Academy of Neurology Abstract Website

Greer Waldrop¹, Sumanth Reddy², Miranda Sullivan¹, Arshdeep Kaur², Colette Caspar², ShiLu Vanasupa¹, Kelsey Zorn¹, Michelle Hughes¹, Chloe Gerungan², Susana Dominguez Penuela³, David Acero-Garces³, Munther Queisi³, Darlisha Williams⁴, Biyue Dai⁴, Fiona Cresswell⁵, David Meya⁶, David Boulware⁴, Barney Stern³, Jeffrey Gelfand¹, Carlos Pardo-Villamizar³, Michael Wilson², Ahmed Abdelhak¹
¹University of California, San Francisco, ²University of California San Francisco, ³The Johns Hopkins University, ⁴University of Minnesota, ⁵Brighton and Sussex Medical School, ⁶Makerere University

Objective:

Identify key protein markers and pathways distinguishing CNS sarcoidosis (CNS-S) from Tuberculosis (TB) Meningitis using unbiased CSF proteomics.

Background:

Differential diagnosis for CNS sarcoidosis is broad, with an overlapping clinical and imaging pattern with TB meningitis. Studying distinct CSF protein changes associated with CNS sarcoidosis can provide unique insights into the pathophysiology of CSN sarcoidosis, a disease characterized by high morbidity and limited molecular understanding.

Design/Methods:

Clinically collected, CSF samples were selected from two cohorts: 1) biopsy-supported CNS sarcoidosis in the United States and 2) tuberculosis meningitis in Uganda. Over 5000 proteins were measured using the Olink Explore HT, proximity extension assay platform. Multivariate linear regression models of protein abundance values determined differentially expressed proteins in CNS sarcoidosis compared to TB meningitis. Multiple comparisons adjusted for using the Benjamini–Hochberg method. Gene Set Enrichment Analysis identified enriched Gene Ontology pathways. Least absolute shrinkage and selection operator (LASSO) regression models were fitted to determine proteins that best discriminate between CNS sarcoidosis and TB meningitis, with performance and feature stability estimated via 1,000 bootstrap iterations.

Results:

5,398 proteins passed QC criteria and were analyzed in 58 CSF samples (CNS-S, n=34, TB Meningitis, n=24). Overall, 128 proteins (Increased: 89, Decreased: 39) demonstrated differences in CNS sarcoidosis compared to TB meningitis after adjustment for age, sex, and CSF pleocytosis, which explained up to 52% of variance (median 36% [IQR: 31-41%] of the variance). Gene enrichment analysis, did not illustrate difference in immune pathways, however synaptic signaling and neuronal development pathways were among the top enriched pathways. LASSO regressions demonstrated a median AUC of 0.89 for diagnostic classification.

Conclusions:

CSF proteomics revealed distinct protein abundance in CSF that could discriminate between CNS sarcoidosis and TB meningitis in training models. More detailed analyses including validation datasets are essential as well as identification of protein relationships.