Transcriptome profiling of aging Drosophila photoreceptors reveals gene expression trends that correlate with visual senescence

Abstract

Background: Aging is associated with functional decline of neurons and increased incidence of both neurodegenerative and ocular disease. Photoreceptor neurons in Drosophila melanogaster provide a powerful model for studying the molecular changes involved in functional senescence of neurons since decreased visual behavior precedes retinal degeneration. Here, we sought to identify gene expression changes and the genomic features of differentially regulated genes in photoreceptors that contribute to visual senescence.

Results: To identify gene expression changes that could lead to visual senescence, we characterized the aging transcriptome of Drosophila sensory neurons highly enriched for photoreceptors. We profiled the nuclear transcriptome of genetically-labeled photoreceptors over a 40 day time course and identified increased expression of genes involved in stress and DNA damage response, and decreased expression of genes required for neuronal function. We further show that combinations of promoter motifs robustly identify age-regulated genes, suggesting that transcription factors are important in driving expression changes in aging photoreceptors. However, long, highly expressed and heavily spliced genes are also more likely to be downregulated with age, indicating that other mechanisms could contribute to expression changes at these genes. Lastly, we identify that circular RNAs (circRNAs) strongly increase during aging in photoreceptors.

Conclusions: Overall, we identified changes in gene expression in aging Drosophila photoreceptors that could account for visual senescence. Further, we show that genomic features predict these age-related changes, suggesting potential mechanisms that could be targeted to slow the rate of age-associated visual decline.

Keywords: Aging; Drosophila; Neurons; Photoreceptors; Transcriptome.

Fig. 1

Visual function declines with age independent of retinal degeneration. a Representative confocal images of adult retinas stained with phalloidin (red) and 4C5 (Rh1, green) from male Rh1-Gal4 > KASH-GFP flies 10 and 40 days post-eclosion (n = 5). Scale bars: 5 μm. b Survival curve showing the percentage of viable Rh1-Gal4 > KASH-GFP male flies at each age (n = 345). c Box plots showing the light preference indices (positive phototaxis) for Rh1-Gal4 > KASH-GFP flies at day 10, 25 and 40 (n = 13 experiments; 27 - 33 male flies/experiment). p value, normally-distributed data were analyzed using ANOVA followed by Tukey’s honest significant different (HSD) post hoc test. d Photoreceptor R1 – R6 nuclei in each ommatidium were labeled with nuclear membrane-localized GFP in Rh1-Gal4 > KASH-GFP flies. Affinity-enriched GFP-labeled nuclei bound to antibody-coated magnetic beads are shown in the two lower panels. DAPI, blue; GFP, green. Schematic of the RNA-seq experimental design is shown in the right panel. Graphic generated by authors

Fig. 2

Age-related changes in gene expression in adult photoreceptors. a Age-regulated genes identified by time-series analysis using maSigPro (555 genes, FDR < 0.05) were clustered using k-means into 11 clusters based on temporal expression pattern (relative expression). The median expression values (circles) and fitted curves with indicated r² values are shown in red on the line graphs. Age-regulated genes were designated as upregulated or downregulated, and early, middle or late based on the fitted curves for their respective cluster (see Additional file 1: Fig. S4). b Over-represented GO terms (p < 0.01, Fisher’s exact test) were identified for 288 upregulated or 267 downregulated genes relative to all 7579 expressed genes using TopGO (Additional file 5: Table S4). Similar GO terms were grouped based on intersecting gene members, and a single representative GO term is shown from each group in the bar plot. Enrichment score indicates the number of genes with the GO term in the target gene set versus the number of expected genes, with p-values shown to the right of each bar. c Representative functional categories identified using GO term analysis and DAVID for upregulated and downregulated genes. Selected age-regulated genes involved in the indicated functions are shown below each term based on published reports

Fig. 3

Combinations of promoter motifs identify age-regulated genes. a Receiver operating characteristic (ROC) curves for combinations of promoter motifs that identify age-regulated genes. Significantly-enriched promoter sequence motifs for up- or downregulated genes were identified using HOMER (40 motifs upregulated genes, 41 motifs downregulated genes; Additional file 6: Table S5). ROC curves representing the diagnostic ability of each motif to identify whether a gene would be up- or downregulated were compared, and the motif with the highest area under the curve (AUC) was iteratively combined with other motifs to identify motif combinations. The maximum AUC values obtained for combinations of motifs are shown. b The AUC values for ROC curves generated by combining increasing numbers of motifs for up- or downregulated genes as described in panel A. The addition of a single motif does not significantly improve the ROC curve (p < 0.05) after the first 14 motifs; we define the first 14 motifs as the top motifs. The maximum AUC value obtained for ROC curves based on 40 randomly-assigned motifs was 0.58 (100 random iterations, see methods)

Fig. 4

Gene length, expression and splicing correlate with age-related downregulation. ROC curves for gene length including introns, expression (RPKM), number of expressed exons and transcripts isoforms for down or upregulated genes. AUC values are indicated for each curve

Fig. 5

circRNA levels increase in aging photoreceptors. a Schematic showing how junction-spanning reads were used to detect circRNAs resulting from back-splicing events. b Volcano plot showing the fold change in circRNA abundance plotted as log₂(fold change in counts per million reads, CPM) for each circRNA relative to its p value (−log₂[p.value]). CircRNAs with significantly differential expression (p ≤ 0.05 and FC ≥ 2, dotted lines) are highlighted. Labels indicate the corresponding host gene for selected circRNAs. c Fold changes in circRNA abundance for pairwise comparisons between the indicated aging time-point. CircRNAs with significantly differential expression (p ≤ 0.05 and FC ≥ 2, blue/red) are highlighted. d Total abundance of circRNAs (CPM) identified at each age. p values, non-parametrical Kruskal-Wallis with Nemenyi post-hoc test for multiple comparisons. e Density plots comparing the log₂ fold changes in circRNA CPM with fold change in linear RNA RPKM from the corresponding gene for 10 versus 40 day sensory neurons. f Linear regression analysis of the mean CPM of the 35 significantly upregulated circRNAs versus age