Data-driven modelling of tau pathology reveals distinct progressive supranuclear palsy subtypes

Progressive supranuclear palsy (PSP) is a heterogeneous neurodegenerative disease characterised by the accumulation of misfolded 4-repeat tau within neurones and glial cells. There is limited longitudinal data on pathologically confirmed PSP patients with phenotypes other than classical Richardson’s syndrome (RS), and the pathomechanisms responsible for the broad variability in clinical phenotype and progression are not well understood. An unresolved question in this context is whether distinct spatiotemporal patterns of tau pathology propagation exist within the clinicopathological spectrum of PSP.

We identified 241 consecutive, pathologically confirmed patients with PSP from the Queen Square Brain Bank for Neurological Disorders (2010-2022). Phenotyping was performed based on clinical features present within the first 3 years from symptom onset according to the Movement Disorder Society (MDS) criteria, and specific clinical features and disease milestones were recorded. Genotyping was performed using Illumina NeuroBooster and NeuroChip arrays and MAPT haplotype, APOE genotype, TRIM11 rs564309, and SLC2A13 rs2242367 single nucleotide polymorphism status were collated from imputed data. Tissue sections from eight brain regions, mounted on glass slides, were immunostained for hyperphosphorylated tau and digitised using whole-slide scanning. Forty-one anatomical regions of interest were manually segmented, and total tau pathology burden was quantified using an automated, machine learning-based algorithm. The associations between survival and both clinicogenetic features and regional tau pathology burden were modelled using Cox regression and generalised linear models, respectively, and the Subtype and Stage Inference (SuStaIn) algorithm was used to identify subgroups with distinct progression patterns.

We have identified: 1) several clinical predictors of survival in PSP and the relationship between regional tau pathology burden and survival; 2) novel anatomical reference standards for the expected distribution of tau pathology across MDS-defined PSP phenotypes, emphasising region-specific white matter involvement in patients with corticobasal syndrome and speech/language variants; 3) associations linking biological sex, MAPT haplotype, and TDP-43 co-pathology to clinical phenotype and regional tau pathology burden; 4) patterns of covariance in regional tau pathology implicating inter-regional connectivity in tau spreading; and 5) three distinct spatiotemporal patterns of tau pathology progression: one characterised by initial involvement of subcortical grey matter followed by rostral spread to frontal white matter and other cortical regions, and two characterised by early, simultaneous involvement of subcortical grey matter and frontal white matter.

Taken together, these results indicate that PSP clinicopathological heterogeneity is mediated by propagation of tau pathology along anatomically connected networks, and via cell- autonomous mechanisms influenced by sex, genetic factors and possibly co-pathology.