Application of histopathologically derived 3D tau burden map as in‐vivo region of interest for biomarker analysis

Background Quantitative three‐dimensional maps of tau neurofibrillary tangles (NFT) burden derived from dense serial histology have potential application for in‐vivo biomarker studies. We constructed a group‐level NFT burden map from 15 medial temporal lobe (MTL) specimens, majority with Primary Age‐Related Tauopathy (PART) or low‐level Alzheimer’s disease neuropathologic change, and showed relatively greater NFT burden in the anterior vs. posterior MTL. We investigated whether in‐vivo MRI and PET measures in ROIs derived from this map show meaningful biological relationships. Method Multimodal in‐vivo imaging data from 292 participants in the Aging Brain Cohort were used. The group‐level NFT burden map was mapped to each participant’s MRI to define an ROI mask, further divided into anterior (aMTL) and posterior (pMTL) ROIs. Cortical thickness (N=292) and 18F‐Flortaucipir SUVR maps (N=86) were computed. Participants’ age was correlated with average thickness in the aMTL, pMTL, and anatomically defined MTL ROIs. 18F‐Flortaucipir uptake was compared between aMTL and pMTL. The analyses were repeated in subsets of cognitive normal participants, and those with negative amyloid PET scans. Result Cortical thickness in the aMTL ROI showed stronger correlation with age than pMTL and anatomically defined MTL subregional thickness. A polynomial fit provided the best age regression, with older participants showing a parabolic decline in thickness around age 60, when substantial NFT accumulation begins in MTL. Further, tau tracer uptake in aMTL was slightly but significantly higher than in pMTL (SUVR 1.19 vs. 1.18, p=0.02). Conclusion We demonstrate the potential use of ex‐vivo NFT burden maps for in‐vivo image analysis. NFT deposition is particularly prominent in the anterior MTL in cases without or with low amyloid burden and at early Braak stage. Prior work suggests that even in the absence of amyloid, this tangle pathology may drive neurodegeneration in the aging population. Here we show that cortical thickness, a biormarker for neurodegeneration, when measured within a histopathologically‐defined region enriched for PART NFTs in the aMTL, is better correlated with age than when measured within anatomically defined subregions, suggesting that the relationship may be driven by PART. Greater tau tracer uptake in the aMTL ROI further supports this notion.