Case-control virtual histology elucidates cell types associated with cortical thickness differences in Alzheimer's disease

Abstract

Many neuropsychiatric disorders are characterised by altered cortical thickness, but the cell types underlying these changes remain largely unknown. Virtual histology (VH) approaches map regional patterns of gene expression with regional patterns of MRI-derived phenotypes, such as cortical thickness, to identify cell types associated with case-control differences in those MRI measures. However, this method does not incorporate valuable information of case-control differences in cell type abundance. We developed a novel method, termed case-control virtual histology (CCVH), and applied it to Alzheimer's disease (AD) and dementia cohorts. Leveraging a multi-region gene expression dataset of AD cases (n = 40) and controls (n = 20), we quantified AD case-control differential expression of cell type-specific markers across 13 brain regions. We then correlated these expression effects with MRI-derived AD case-control cortical thickness differences across the same regions. Cell types with spatially concordant AD-related effects were identified through resampling marker correlation coefficients. Among regions thinner in AD, gene expression patterns identified by CCVH suggested fewer excitatory and inhibitory neurons, and greater proportions of astrocytes, microglia, oligodendrocytes, oligodendrocyte precursor cells, and endothelial cells in AD cases vs. controls. In contrast, original VH identified expression patterns suggesting that excitatory but not inhibitory neuron abundance was associated with thinner cortex in AD, despite the fact that both types of neurons are known to be lost in the disorder. Compared to original VH, cell types identified through CCVH are more likely to directly underlie cortical thickness differences in AD. Sensitivity analyses suggest our results are largely robust to specific analysis choices, like numbers of cell type-specific marker genes used and background gene sets used to construct null models. As more multi-region brain expression datasets become available, CCVH will be useful for identifying the cellular correlates of cortical thickness across neuropsychiatric illnesses.

Keywords: Alzheimer's disease; Cerebral cortex; Cortical thickness; Imaging transcriptomics; Magnetic resonance imaging; Virtual histology.