Depth-dependent Spatial Clustering of Gene Expression in the Middle Temporal Gyrus is Elevated in Advanced Alzheimer's Disease Tissue

Abstract

Aims: Alzheimer's disease (AD) produces spatially heterogeneous molecular and cellular changes across the human cortex. This study tested whether spatial clustering of gene expression differs between low- and high-Braak-stage middle temporal gyrus (MTG) tissue and whether these differences are organised along normalised cortical depth. Study Design: Retrospective secondary computational analysis of publicly available, de-identified spatial transcriptomic data. Methodology: Multiplexed Error-Robust Fluorescence in Situ Hybridisation (MERFISH) data from the Seattle Alzheimer's Disease Brain Cell Atlas (SEA-AD) were analysed. MTG sections were grouped as Low Braak (I-III; n = 14 sections) or High Braak (IV-VI; n = 55 sections). Global Moran's I was computed for 140 measured genes using k-nearest-neighbour spatial weights. Group differences were evaluated with Mann-Whitney U tests and Benjamini-Hochberg false-discovery-rate correction. Cortical depth was normalised from pia to white matter, and expression-depth gradients were estimated with mixed-effects models and bootstrap resampling to reduce the effect of unequal group sizes. Results: Eight genes showed significantly higher global spatial clustering in High Braak tissue (q < .05 and Delta I > .01): NRG1, FGF13, SEMA3E, SCUBE1, PEX5L, KCNIP4, KIAA1217, and FRMPD4. None of the blank control probes met these criteria. All eight genes showed preferential expression towards deeper cortical layers, and seven genes, excluding SCUBE1, showed steeper expression-depth gradients in High Braak sections. Local Moran's I depth profiles did not show a uniform disease-associated depth pattern, indicating that increased expression magnitude and local spatial clustering are partially separable features. Conclusion: Advanced AD MTG tissue shows amplified laminar expression gradients for a limited set of genes related to synaptic regulation, neuronal structure, excitability, oxidative stress, and vascular biology. Because the MERFISH panel included 140 genes and the primary global screen used tissue sections as the analytical unit, the findings should be interpreted as exploratory and require donor-aware, cell-type-resolved, and independent validation. The results nevertheless support cortical depth as an important spatial axis of AD-related transcriptional remodelling.