Age-related dysregulation of the retinal transcriptome in African turquoise killifish

Abstract

Age-related vision loss caused by retinal neurodegenerative pathologies is becoming more prevalent in our ageing society. To understand the physiological and molecular impact of ageing on retinal homeostasis, we used the short-lived African turquoise killifish, a model known to naturally develop central nervous system (CNS) ageing hallmarks and vision loss. Bulk and single-cell RNA-sequencing (scRNA-seq) of three age groups (6-, 12-, and 18-week-old) identified transcriptional ageing fingerprints in the killifish retina, unveiling pathways also identified in the aged brain, including oxidative stress, gliosis, and inflammageing. These findings were comparable to observations in ageing mouse retina. Additionally, transcriptional changes in genes related to retinal diseases, such as glaucoma and age-related macular degeneration, were observed. The cellular heterogeneity in the killifish retina was characterised, confirming the presence of all typical vertebrate retinal cell types. Data integration from age-matched samples between the bulk and scRNA-seq experiments revealed a loss of cellular specificity in gene expression upon ageing, suggesting potential disruption in transcriptional homeostasis. Differential expression analysis within the identified cell types highlighted the role of glial/immune cells as important stress regulators during ageing. Our work emphasises the value of the fast-ageing killifish in elucidating molecular signatures in age-associated retinal disease and vision decline. This study contributes to the understanding of how age-related changes in molecular pathways may impact CNS health, providing insights that may inform future therapeutic strategies for age-related pathologies.

Keywords: Ageing; Gliosis; Inflammageing; Neurodegeneration; Nothobranchius furzeri; Oxidative stress; Retina; Transcriptomics.

Figure 1.. Turquoise killifish retinas have age-related changes in gene expression.

(A) Experimental setup for the bulk RNA-seq experiments. (B) Principal component analysis of 6-, 12-, and 18-week-old individual RNA-seq samples (n=10 each) showing that age-related variance in transcriptome across samples. (C) Scaled Venn diagrams depicting up- (red) and down-regulated (blue) genes between young (6-week-old), middle-aged (12-week-old) and aged (18-week-old) retinas. There are more up-regulated genes than down-regulated genes identified in the dataset with age. Differentially expressed genes were identified using following thresholding criteria: FDR ≤ 0.05 and |log2FC| ≥ 1. (D) Boxplots representing the age-dependent fold change of transcript expression of mouse orthologs (Xu et al., 2022) to Up, Down, and Unchanged sets of differentially expressed killifish genes. The killifish-specific unchanged (grey) and up-regulated genes (red) show a similar trend of expression as in the ageing mouse retina. This is not observed for the down-regulated genes (blue) (Wilcoxon rank-sum test). FC = fold change, mo = months, ns = not significant, PC = principal component, RIN = RNA integrity number, RPE = retinal pigment epithelium, Unch = unchanged, w = weeks.

Figure 2.. Aged killifish retinas show signs of gliosis, inflammageing, and neurodegeneration detected by bulk RNA-seq.

(A) Scatter plot showing mean expression of all genes detected by bulk RNA-seq across 6w and 18w samples. The different colours denote genes that passed FDR and FC thresholds (FDR < 0,05 and|log2FC| ≥ 1). Genes that are not significantly changed are shown in grey; up-regulated genes are in red; down-regulated genes in blue. ZC3HDC1L is representing the optn gene and CU571169.1 is slc7a11. (B) Immunohistochemistry for Rlbp1 shows a visual increase in Rlbp1 protein expression in aged retinas, corresponding to increased transcript expression in RNA-seq. Additionally, images highlight Müller glia expansion and elaboration. (C) Immunostaining of glutamine synthetase similarly shows expansion of Müller glia cell morphology in 18-week-old killifish retinas, as compared to 6-week-old retinas. (D) in situ HCR for apoeb in 6 and 18-week-old killifish retinas highlight an increase in immune cells (microglia/macrophages) with age as well as changes in their morphology (insets), which are signs of inflammageing. (E) Microglia/macrophage-especific labelling by apoeb is confirmed using antibody co-labelling with the pan-leukocyte marker L-plastin. Arrowheads indicate cells that are both apoeb+ and L-plastin+. Images are acquired as mosaic Z-stack and are visualised as maximum projections. Merged images with nuclei shown in left third of image; remaining image is without nuclei. Scale bars = 50 μm. HCR = hybridisation chain reaction, FC = fold change, FDR = false discovery rate, GCL = ganglion cell layer, GS = glutamine synthetase, INL = inner nuclear layer, ONL = outer nuclear layer, unch = unchanged, w = weeks.

Figure 3.. scRNA-seq identifies neuronal and glial cell types within the killifish retina.

(A) Experimental setup for the single-cell RNA-sequencing experiments. (B) UMAP dimension reduction of the killifish retina scRNA-seq dataset with clusters coloured by annotated retinal cell type. All cell types expected of a vertebrate retina, as well as oligodendrocytes, are present in the killifish retina. (C) Dot plot showing the specificity of marker genes within individual retinal cell types. The size of the dot represents the percentage of cells within the population expressing transcripts for the gene, while the colour indicates the average expression across individual cells. (D-M) Spatial validation of cell type marker genes via in situ HCR confirms the cell type identification for every retinal cell type. Images are acquired as mosaic Z-stack and are visualised as maximum projections. Merged images with nuclei shown in left third of image. Scale bars = 50 μm. HCR = hybridisation chain reaction. UMAP = uniform manifold approximation and projection. GCL = ganglion cell layer, INL = inner nuclear layer, ONL = outer nuclear layer, RBC = red blood cell. RPE = retinal pigment epithelium, w = weeks.

Figure 4.. Validation of cell subtypes for specific killifish retinal populations.

UMAP dimension reduction shows the subclustering of the photoreceptors (A), horizontal cells (D), bipolar cells (G) and amacrine cells (J). Cell type specific markers are shown as dot plots for each cell type (B, E, H and K, respectively). Dot size shows the percentage of cells expressing the marker gene while the colour indicates the mean transcript expression. in situ HCRs of subtype markers for photoreceptors (C), horizontal cells (F), bipolar cells (I), and amacrine cells (L) distinguish specific retinal cell subtypes. A displaced starburst amacrine (chat+) is highlighted with an inset box and arrowhead in the GCL (L). Images are acquired as mosaic Z-stack and are visualised as maximum projections. Merged images with nuclei shown in left third of image. Scale bars = 50 μm. GCL = ganglion cell layer, INL = inner nuclear layer, ONL = outer nuclear layer.

Figure 5.. Killifish display age-associated transcriptional dysregulation, with many genes becoming expressed in all cell types.

(A) Heatmaps of data integrated from both the bulk and scRNA-seq. Genes that are increasing with age in our dataset are typically expressed in a variety of cell types within the retina, with roughly half of the transcripts displaying enriched expression within glial/immune cells (left). With age, the genes are increasing in expression across numerous cell types (right). Several genes of interest are highlighted. ZC2HDC1L is optn and CU571169.1 is slc7a11. (B) UMAP showing that the genes that show increased expression with age are overall most highly expressed in Müller glia, immune cells, and oligodendrocyte clusters. (C-E) Spatial validation of up-regulated genes; total merge with nuclei shown in left third of image. (C) The Müller glia-specific gene aqp1a.1 has transcript detected in many cell types at 18 weeks in in situ HCR, including retinal ganglion cells and photoreceptors. (D) Immunolabelling for Egfr and GS showing that Egfr protein is primarily in Müller glia at in young retinas – however, in old retinas there is increased expression in the ganglion cell and nerve fibre layers. Middle: Egfr + GS merge without nuclei; right: Egfr labelling only. (E) Antibody labelling for Tgfb3 and GS shows that Tgfb3 protein is restricted to Müller glia in young retinas but appears more universally expressed in old retinas, including the outer nuclear layer. Middle: Tgfb3 + GS merge without nuclei; right: Tgfb3 only. Images are acquired as mosaic Z-stack and are visualised as maximum projections. Scale bars = 50 μm. GCL = ganglion cell layer, GS = glutamine synthetase, HCR = hybridisation chain reaction, INL = inner nuclear layer, ONL = outer nuclear layer, RBCs = red blood cells, RPE = retinal pigment epithelium, w = weeks.