Abstract
OBJECTIVES: Progressive supranuclear palsy (PSP) is a neurodegenerative 4R tauopathy clinically presenting with atypical parkinsonism or cognitive behavioral changes and a relatively uniform neuropathology. We recently identified rare HLA haplotypes in PSP and now examine whether HLA haplotypes are associated with different cytopathological and clinical phenotypes. METHODS: Retrospective collection of clinical data and mapping of T and B cells, microglia, and phosphorylated-tau (p-Tau) cytopathologies in 32 PSP cases. Machine learning was used to analyze whether pathological variables and their ratios, or the sequence of clinical symptoms cluster or predict HLA haplotypes. RESULTS: Four groups were defined based on HLA haplotypes: i) 12 cases with the haplotype associated with narcolepsy (DRB1*15:01-DQB1*06:02); ii) 11 cases with other DQ5-DQ6 haplotypes; iii) 8 cases with various haplotypes frequent in the general population; and iv) one case with the haplotype frequent in IgLON5-disease (DRB1*10:01-DQB1*05:01). Neuropathology revealed regional differences in the severity of microglia load, density of cytotoxic T cells, and p-Tau cytopathologies between groups. HLA haplotypes were most distinguishable using machine learned features of inflammatory markers and ratios of neuropathological variables (clustering accuracy: 86.96% and 91.30%, respectively). The sequence of clinical symptoms and the ratios of neuropathological variables were the strongest predictors of HLA haplotypes (prediction accuracy=80.00% and 71.43%, respectively). INTERPRETATION: PSP pathology might be associated with various etiological-pathogenic events including targetable autoimmune mechanisms. The HLA-haplotype dependent diversity of neuroinflammatory markers should be evaluated in clinical and biomarker studies in, and beyond, PSP to understand its relevance for patient stratification in disease modifying therapy trials.