Distinct plasma lipids predict axonal injury and multiple sclerosis activity

Background Lipids are of particular interest for the study of neuroinjury and neuroinflammation as structural lipids are major components of myelin, and a variety of lipid species modulate inflammation. In this study, we performed an in-depth lipidomics analysis to identify lipids associated with injury and disease activity. Methods Plasma samples were collected from paediatric-onset multiple sclerosis (MS) cases within 4 years of disease onset from 17 sites. The lipidome was measured using untargeted and targeted mass spectrometry. For cross-sectional analyses, the agreement between multiple machine learning models was used to predict neurofilament light chain (NfL) levels. In longitudinal analyses, the association between clinical (relapse count) and imaging (MRI count with ≥1 enhancing or new T2 lesion) outcomes with each metabolite was estimated using adjusted negative binomial regression. Results At sample collection, 68% of the 435 included individuals were treatment-naive, with a median disease duration of 0.8 years (IQR 0.3–1.7). For longitudinal analyses, 381 and 335 subjects had at least 1 year of clinical and imaging follow-up, respectively. In cross-sectional analyses, NfL chain levels identified structural lipids (phosphatidylcholines and phosphatidylethanolamines) as the highest-performing predictors, including external validation. In contrast, longitudinal analyses found polyunsaturated fatty acids (PUFAs) and their derivatives to be protective from subsequent disease activity (q<0.001, multiple outcomes). Conclusion There are two categories of lipids associated with MS processes. First, structural lipids strongly associated with NfL levels may result from cell lysis secondary to acute inflammation. In contrast, PUFAs, especially ω−3, had a protective effect on subsequent disease activity.