Bridging brain structural/functional connectivity and tau load in PSP: A [18F]PI-2620 PET/MRI study

Objectives: Progressive supranuclear palsy (PSP) is a neurodegenerative disorder associated with the accumulation of the tau 4R isoform in the microtubule-binding domain, predominantly, in the substantia nigra (ST), subthalamic nucleus (STN), and globus pallidum. The evidence supports an ongoing reorganization of the structural and functional connectivities in PSP that might be influenced by tau pathology. We aimed to assess the network properties derived from graph theoretical analysis of the functional and structural connectivities along the deep gray matter pathways and their relation to tau burden.

Methods: Ten PSP patients with predominant PSP-Richardson/parkinsonian/cortico-basal phenotype (age 72±7 years, 6 female) underwent 60 minutes dynamic PET imaging following 300 MBq bolus injection of [18F]PI-2620 on a hybrid 3T PET/MRI. Ten healthy controls (age 65±8 years, 4 female) were included for graph theory-based comparison of functional (resting-state fMRI) and structural (probabilistic tractography in network mode on diffusion-tensor imaging data) connectivity properties. A deep gray matter network was generated consisting of thirteen PSP-target regions (nodes): putamen, globus palidus externus (GPe), STN, ST, periaqueductal gray matter, red nucleus (RN) and dentate nucleus (DN). For each node, binding potential (BP) was estimated using non-invasive pharmacokinetic modeling (MRTM2 in PMOD). Structural and functional 13x13 connectivity matrices were generated, respectively, in R using brainGraph and Matlab using CONN packages. Graph theoretical analysis was used to investigate the global and local characteristics of the networks. Between-group vertex-wise analysis of graph metrics were assessed with General Linear Model (GLM) and FDR-corrected p < 0.05 was considered as significant. Student's t-test and Spearman's rank correlation coefficient were used, respectively, for between-group differences of the global network attributes and their association with BP.

Results: For functional connectivity, at the global network level, the global and local efficiencies, clustering coefficient and betweenness centrality were significantly lower, while average path length and degree were higher in PSP compared to controls. At the vertex level, global efficiency was lower for putamen/ST (left) and STN/RN (right); average path length and degree were higher for putamen/ST (left). The BP showed strong relationship (r=-0.7, p<0.05) with local efficiency and clustering coefficient for all nodes; additionally, the STN (left) showed an association with global efficiency, degree, average path length and betweenness centrality. For structural connectivity, at the global network level, graph vulnerability showed a trend (p = 0.06) towards a decrease, and asymmetry a trend (p = 0.07) towards an increase in PSP compared to controls. At the vertex level, local efficiency was lower in SN/RN (left) and bilateral DN; average path length was decreased in STN (left) and putamen (right), while increased in ST/RN (left) and GPe (right); vertex strength was higher in the bilateral ST; nodal vulnerability was increased in STN/DN (left) and putamen (right), while decreased in all other regions. The BP showed an inverse relationship with nodal assortativity for bilateral putamen/RN/STN and with vulnerability for STN (left).

Conclusions: Graph theory analysis of DTI tractography and resting-state fMRI showed diffuse differences in structural and functional brain network properties, indicating lower network integration and higher segregation in PSP that is related to local tau burden. It will be interesting to study how these findings translate to longitudinal PET/MRI data to assess the potential of this multi-modality approach for improved disease progression monitoring.