Abstract
Purpose :
Dysregulated immune response is a hallmark of normal aging and a prominent feature in many neurodegenerative diseases including AMD. However, the mechanisms underlying immune dysregulation-mediated neurodegeneration are multifaceted and have not been completely resolved in the context of AMD. Here we analyze the transcriptome data (RNAseq) from 453 post-mortem donor retinas to uncover the cellular and molecular signatures underlying AMD.
Methods :
The bulk RNA-seq data, comprising 453 samples categorized as 105 controls, 175 early AMD, 112 intermediate AMD, and 61 advanced AMD, underwent machine learning analysis to unveil transcriptional signatures associated with AMD. Next, we built a reference for the average expression of retinal cell types using cell-type specific markers across different cell types in single cell data from six human samples. Employing three distinct deconvolution methods—CIBERSORTx, dTangle, and BayesPrism—we estimated the proportions of cell types in both normal and AMD stages. To validate the results of the deconvolution, we analyzed single-nuclei data from 13 controls and 17 late AMD patients from two published studies.
Results :
We identified a set of 81 genes associated with AMD. Subsequent, integrative analysis of single cell data highlighted the enrichment of these signature genes in retinal glial cells, astrocytes and microglia. Cellular deconvolution of normal and AMD cohorts identified distinct differences in the cellular composition of microglia, astrocytes and müller glia between the normal and advanced AMD. Notably, alterations in microglia were significant at early, intermediate, and advanced stages of AMD. We found 20/81 to be differentially expressed in retinal glial cells. Specifically, within the microglia, a subset of 10 genes demonstrated differential expression, with 8 upregulated and 2 downregulated.
Conclusions :
Our study underscores the significance of microglia at both the molecular and cellular level in a large cohort. These results reveal gene signatures that modulate the retinal glial function, acting as the driving force in disease progression and photoreceptor degeneration in AMD. Additionally, it highlights the merits of machine learning and data analytics in gaining mechanistic insights into AMD. Finally, these approaches offer an opportunity to regain the holistic view of the AMD that is lost in experimentally tested reductionist approaches.
This abstract was presented at the 2024 ARVO Annual Meeting, held in Seattle, WA, May 5-9, 2024.