Transcriptional Heterogeneity and Differential Response of Rod Photoreceptor Pathway Uncovered by Single-Cell RNA Sequencing of the Aging Mouse Retina

Abstract

Visual function deteriorates throughout the natural course of aging. Age-related structural and functional adaptations are observed in the retina, the light-sensitive neuronal tissue of the eye where visual perception begins. Molecular mechanisms underlying retinal aging are still poorly understood, highlighting the need to identify biomarkers for better prognosis and alleviation of aging-associated vision impairment. Here, we investigate dynamics of transcriptional dysregulation in the retina and identify affected pathways within distinct retinal cell types. Using an optimized protocol for single-cell RNA sequencing of mouse retinas at 3, 12, 18, and 24 months, we detect a progressive increase in the number of differentially expressed genes across all retinal cell types. The extent and direction of expression changes varies, with photoreceptor, bipolar, and Müller cells showing the maximum number of differentially expressed genes at all age groups. Furthermore, our analyses uncover transcriptionally distinct, heterogeneous subpopulations of rod photoreceptors and bipolar cells, distributed across distinct areas of the retina. Our findings provide a plausible molecular explanation for enhanced susceptibility of rod cells to aging and correlate with the observed loss of scotopic sensitivity in elderly individuals.

Keywords: aging; bipolar cells; molecular heterogeneity; photoreceptor; retina; single cell RNA‐seq; transcription; vision impairment.

FIGURE 1

Single‐cell transcriptome profiling of aging mouse retina. (a) Workflow illustrating the preparation process of single‐cell suspensions from whole retinas (WR), or rod‐depleted retinal cell suspensions accomplished through CD73 cell surface marker‐mediated cell depletion (CD73 negative cells; CD73N) from male mice aged 3, 12, 18, and 24 months. Subsequently, the samples underwent 10x Genomics single‐cell RNA sequencing. The bioinformatic pipeline involved analyzing CellRanger data using an optimized Seurat SCT pipeline, which included the removal of ambient RNA and heterotypic doublets using DecontX and DoubletFinder, respectively. Additionally, it encompassed the annotation of retinal cell types using scType with well‐established retinal cell type markers, see methods. (b) UMAP visualization of cells obtained from WR and CD73N retinal single‐cell suspensions. (c) UMAP visualization of cell annotations. Broad cell types are depicted in larger font size, while subtypes of cones and cone bipolar cells are displayed in smaller font size. (d) Validation of clustering and cell type annotations through UMAP feature plots of cell type‐specific markers: Rho (rods), Opn1sw and Opn1mw (blue and green cones, respectively), lsl1 and Grik1 (OFF and ON bipolar cells, respectively), Slc6a9 and Gad1 (glycinergic and GABAergic amacrine cells, respectively), Apoe (Müller glia cells), Thy1 (ganglion cells), and Calb1 (horizontal cells). (e) Expression dot plots showing the top 5 enriched genes for each annotated cell type when compared to the entire dataset. The expression values represent log‐normalized counts. AC, amacrine cell; BC, bipolar cell subtype; BP, bipolar cell; CD73N, CD73 negative cells (non‐rod cells); Endo, endothelial cell; Feat, features (genes); GABA AC, GABAergic amacrine cell; Gly AC, glycinergic amacrine cell; Hor, horizontal cell; Micro, microglia cell; mt, mitochondria reads; Peri, pericyte; RGC, retina ganglion cell; SCTv2, sctransform version 2; UMAP, uniform manifold approximation and projection; UMI, unique molecular identifier; WR, whole retina.

FIGURE 2

Differences in the retinal transcriptome during aging. (a) UMAP visualization of retinal cells from mice aged 3, 12, 18, and 24 months, with colors representing the age of the sample. (b) Age stratified histogram of cell counts in broadly annotated cell types. (c) Histograms depicting the number of genes with differential gene expression (DGE) (cut‐off details in Materials and Methods) in cells from 12‐, 18‐, and 24‐month‐old mice compared to 3‐month‐old mice, stratified by the direction (up or down) of expression change. (d) Genes upregulated at 24 months compared to 3 months across more than 4 broad cell types (left). Downregulated genes are shown in the right panel. Gray boxes indicate no significant change. (e) Functional gene enrichment results for Gene Ontology (GO) biological process ontology terms in the comparison between 24‐month‐old and 3‐month‐old samples. DGE, differential gene expression; FC, fold‐change; GO, gene ontology; Pct, percentage present.

FIGURE 3

Transcriptional heterogeneity in rod photoreceptors. (a) UMAP projection of the rod photoreceptor cell sub‐clusters in WR/CD73N dataset determined by the Louvain algorithm in Seurat (left) and alternative clustering and UMAP using Monocle3 of the rod clusters identified in Seurat (middle). Cell population in each rod cluster for each age (right). (b) Dot plots highlighting differential expression of genes identified in PER versus SER comparison and enriched in selected Gene Ontology terms (biological process). Left column shows terms enriched in PER and right column shows terms enriched in SER. Average expression values are scaled. A vertical dashed line separates PER clusters (4, 8, 14, 20) from SER clusters (0, 1, 5, 3, 33). (c) UMAP projection of rod clusters identified in the second independent scRNA aging retina dataset obtained from superior and inferior (Sup/Inf) retinal punches. Expression dot plots of the enriched ontology terms identified as significant in (b) are populated with data from superior and inferior retina aging data to highlight the presence of rod heterogeneity in the Sup/Inf dataset. Average expression values are scaled. (d) UMAP projections showing the presence of marker genes for PER (Rcvrn) and SER (Rpgrip1 and Syt1) rod photoreceptors in our two current retina aging datasets (WR/CD73N and Sup/Inf), as well as publicly available datasets including adult mouse (Heng et al. 2019), human (Lukowski et al. 2019), and macaque (Yi et al. 2021) retinas. Expression values (color) represent normalized counts. CD73N, CD73 negative cells (non‐rod cells); GO‐BP, gene ontology—biological process; PER, phototransduction‐efficient rods; SER, synaptic transmission‐efficient rods; Sup/Inf, superior/inferior retina; WR, whole retina.

FIGURE 4

Molecular signatures in aging rod photoreceptors. (a) DGE results in PER. UMAP projection highlighting PER cell population in rods (top left). Venn diagram displaying the number of significant DGEs in PER cells for each age comparison: 12, 18, and 24 months versus 3 months (bottom left). Progression of scaled gene expression over age of 245 DGE genes clustered for similar patterns. The red line indicates the median expression of the cluster (middle). Enrichment analysis of GO biological process enrichment for the DGE gene clusters (right). (b) DGE results in SER. UMAP projection highlighting SER cell population in rods (top left). Venn diagram displaying the number of significant DGEs in SER cells for each age comparison: 12, 18, and 24 months versus 3 months (bottom left). Progression of scaled gene expression over age of the 192 DGE genes clustered for similar patterns. The red line indicates the median expression of the cluster (middle). Enrichment analysis of GO biological process for the DGE gene clusters (right). (c) Comparative analysis of PER (left) and SER (right) 24‐month versus 3‐month DGE results alongside flow‐sorted rod aging samples from Corso‐Diaz et al. . (d) Selected gene expression plots showing cell expression distribution from PER and SER over aging. Images show the validation of selected genes through in situ hybridization for both 3‐ and 24‐months retinas. (e) UMAP projection of Dpp10 normalized expression counts for the entire rod population (left) and the percentage of rods expressing Dpp10 at each age (right). (f) Percentage of rod cells expressing Dpp10 across age groups and retinal regions in the Sup/Inf dataset. DGE, differential gene expression; FC, fold‐change; GCL, ganglion cell layer; GO, gene ontology; Inf, inferior retina; INL, inner nuclear layer; ONL, outer nuclear layer; PER, phototransduction‐efficient rods; SER, synaptic transmission‐efficient rods; Sup, superior retina.

FIGURE 5

Gene expression differences in aging bipolar cells. (a) UMAP projection highlighting rod bipolar cell clusters (top left) and percentage of cells per cluster at each age (top right). Venn diagram displaying the number of significant DGEs for each age comparison: 12, 18, and 24 months versus 3 months (bottom). (b) Progression of scaled gene expression over age of the 628 DEG genes clustered for similar patterns. The red line indicates the median expression of the cluster (left). Enrichment analysis of GO biological process enrichment for the DGE gene clusters (right). (c) UMAP visualization of ON cone bipolar cells colored by age of sample and labeled by sub‐cell type (top). Number of significant DGEs in each sub‐cell type in the 24‐month versus 3‐month comparison (bottom). (d) GO biological process enrichment of the 24‐month versus 3‐month DGEs for each sub‐cell type (left). Expression dot plot of the DGE genes for the selected GO term shown at each age in each sub‐cell type. Scaled average expression is shown (right). (e) UMAP visualization of OFF cone bipolar cells colored by age of sample and labeled by sub‐cell type (top). Number of significant DGEs in each sub‐cell type in the 24‐month versus 3‐month comparison (bottom). (f) GO biological process enrichment of the 24‐month versus 3‐month DGEs for each sub‐cell type (left). Expression plot of the DGE genes for the selected GO term shown at each age in each sub‐cell type. Scaled average expression is shown (right). DGE, differential gene expression; GO, gene ontology; GO‐BP, gene ontology—biological process.