Bile acid metabolomics reveals distinct immunometabolic niches and enables accurate diagnosis of AQP4-IgG–seronegative NMOSD

Background and Objective Differentiating neuromyelitis optica spectrum disorder (NMOSD) from multiple sclerosis (MS) is clinically difficult, especially when patients are seronegative for anti–aquaporin-4 immunoglobulin G (AQP4-IgG). Bile acids (BAs) function as important immunoregulatory metabolites, yet their metabolic signatures have not been profiled across NMOSD subtypes or compared with MS. We therefore profiled BA metabolism in NMOSD and MS to determine diagnostic utility. Methods We enrolled 112 NMOSD patients (32 AQP4-IgG seronegative, 80 AQP4-IgG seropositive), 50 MS patients, and 66 healthy controls. Targeted liquid chromatography–mass spectrometry quantified 15 bile acids, from which 71 derived metabolic indices were computed. Candidate biomarkers were identified using nested cross-validation combined with stability selection and integrated into diagnostic models evaluated under a pre-specified internal multi-level validation framework. Associations between BA profiles and neurological disability were assessed using Spearman correlation. Results Serum BA signatures differed markedly between NMOSD and MS. NMOSD showed a pronounced increase in primary conjugated BAs, whereas MS displayed enhanced secondary BA metabolism. Notably, AQP4-IgG seronegative NMOSD had significantly lower secondary BA concentrations than all other groups. A BA-based diagnostic model distinguished NMOSD from MS with an AUC of 0.874. When specifically differentiating AQP4-IgG–seronegative NMOSD from MS, deoxycholic acid (DCA) demonstrated strong discriminative potential (AUC 0.965), with internally consistent performance across resampling- and cross-validation–based robustness analyses. Disease-specific correlations between BA profiles and Expanded Disability Status Scale scores were also observed. Conclusions Serum bile acid metabolomic profiling reveals disease- and subtype-specific signatures that may assist in the differential diagnosis of NMOSD and MS, particularly in AQP4-IgG–seronegative cases. Given the cross-sectional design, these findings should be interpreted as disease-associated metabolic signatures rather than causal mechanisms. DCA showed strong discriminatory performance; however, its clinical utility remains hypothesis-generating and requires replication in a fully independent external cohort.