Molecular Signatures of Resilience to Alzheimer's Disease in Neocortical Layer 4 Neurons

We used the 10x Genomics Xenium spatial platform to map annotated neuronal cell subtypes at single-cell resolution. In total, 16 human brain sections were analyzed—8 from Brodmann area 9 (BA9) and 8 from BA17—including samples from 4 donors with high Alzheimer's disease (AD) pathology and 4 control donors. Fresh-frozen brain sections were cut at a thickness of 10 µm using a cryostat and mounted within fiducial frames on four Xenium slides.

Spatial transcriptomic profiling was performed using the predesigned 266-gene Xenium Human Brain Gene Expression panel, supplemented with a custom 100-gene panel. Data preprocessing followed standard pipelines using Xenium Ranger. Cell segmentation was carried out using the multimodal segmentation algorithm, prioritizing interior RNA staining (ribosomal RNA) to define cell boundaries, followed by a 5 µm isotropic nuclear (DAPI) expansion.

Cell type and neuronal subtype annotation was performed using a combination of:

1. Manual curation based on k-nearest neighbor graphs, Leiden clustering, and canonical marker genes;

2. Heuristic classification using a custom Python script that assigned cell types based on the most highly expressed transcripts;

3. Deep neural network (DNN) classification with spatialID, trained on the SEA-AD DLPFC dataset; and

4. Ingest-based label transfer based on our snRNA seq data (BA9, BA17)