Circadian deep sequencing reveals stress-response genes that adopt robust rhythmic expression during aging

Abstract

Disruption of the circadian clock, which directs rhythmic expression of numerous output genes, accelerates aging. To enquire how the circadian system protects aging organisms, here we compare circadian transcriptomes in heads of young and old Drosophila melanogaster. The core clock and most output genes remained robustly rhythmic in old flies, while others lost rhythmicity with age, resulting in constitutive over- or under-expression. Unexpectedly, we identify a subset of genes that adopted increased or de novo rhythmicity during aging, enriched for stress-response functions. These genes, termed late-life cyclers, were also rhythmically induced in young flies by constant exposure to exogenous oxidative stress, and this upregulation is CLOCK-dependent. We also identify age-onset rhythmicity in several putative primary piRNA transcripts overlapping antisense transposons. Our results suggest that, as organisms age, the circadian system shifts greater regulatory priority to the mitigation of accumulating cellular stress.

Figure 1. RNA-seq reveals diverse classes of age-dependent rhythmicity changes.

(a) Heat map representing loss or gain of gene expression rhythmicity with age. Each column represents a single-time-point for a single-replicate sampled at 4 h intervals starting at Zeitgeber Time (ZT) 0. Hue represents the Z-score (FPKM minus mean over s.d.) of each gene (row) at each time point (column) for a given biological replicate. Genes in the top half are rhythmic in young and arrhythmic in old; those in the bottom are rhythmic in old and arrhythmic in young (Methods). (b–d) RNA-seq gene expression profiles; each period of ZT 0–20 represents a distinct biological replicate. (b) Coordinated, reduced amplitude in expression of circadian genes at the Cks30A gene locus. (c) Examples of genes rhythmic only in young (CG9507 and CG11425); rhythmic in both young and old, with age-dependent increases in amplitude (RpL32 and Hr38). (d,e) ImpL3 adopts de novo rhythmic expression in old flies according to RNA-seq. Browser tracks in e show RNA-seq read pileups at each time point; reads represent merged RNA-seq data from replicate cohorts of old flies with the same y-axis range for all time points.

Figure 2. Core clock gene expression in young and old flies.

(a) Age-dependent expression of three clock genes (RNA-seq). Each period of ZT 0–20 represents a distinct biological replicate. Error bars represent 95% confidence intervals reported by Cuffdiff. (b) qRT-PCR expression of per, tim and Clk in heads of females aged to 5, 55 and 75 days. Expression is reported as per cent of peak expression in day 5 flies. Data are mean (from two cycles)±s.e.m. (c) PER western blot from heads of young or old w flies collected at indicated ZT. The arrow indicates PER protein, with estimated molecular weight of 127-130 kDA, depending on the phosphorylation status. (d) PER levels (mean±s.e.m.) from western blot in c; n=3 biological replicates. (b,c) *P≤0.05; **P≤0.01; ***P≤0.001; ****P≤0.0001 (two-way ANOVA with Bonferroni correction; day 55 change is relative to day 5 values, and day 75 change is relative to day 55 values).

Figure 3. Identification of late-life cyclers (LLCs).

(a) Differential periodicity Z-score is plotted against differential robustness Z-score (Methods); top LLCs from Supplementary Table 3 are labelled. Data points are coloured by differential rhythmicity score (S_DR). (b) Superimposed min-max normalized RNA-seq profiles of known stress-responsive LLCs with peak expression >10 FPKM. Broken lines=day 5 data; solid lines=day 55 data. Each period of ZT 0–20 represents a distinct biological replicate. (c) Bars heights represent normalized counts of LLCs and genes rhythmic in young or old, which peak at indicated 4-h intervals (Methods). (d) qRT-PCR expression of LLCs in heads of females at age 5, 55 and 75 days. (e) qRT-PCR expression of LLCs in heads of young and old males. (d,e) Data are mean (from two cycles)±s.e.m. Expression is per cent of peak expression in the oldest group. *P≤0.05; **P≤0.01; ***P≤0.001; ****P≤0.0001 (two-way ANOVA with Bonferroni correction; day 55 change is relative to day 5 values, and day 75 change is relative to day 55 values).

Figure 4. Oxidative stress-induced LLC rhythms in young flies.

Heads of young females were collected in 12:12 light/dark (LD) at 4-h intervals on the 4th day in hyperoxia (HO) or normoxia (NO) according to the experiment scheme in a. (b,c) qRT-PCR results for RNA from heads of flies in HO or NO. Data (mean±s.e.m.) are reported as per cent of peak expression in LD under HO. n≥2 biological replicates. *P≤0.05; **P≤0.01; ***P≤0.001; ****P≤0.0001 (data from each genotype in HO are compared by two-way ANOVA with Bonferroni correction relative to w flies in NO). (b) w flies in NO and HO. (c) w flies in HO and NO; Clk^out flies in HO.

Figure 5. Age-induced circadian expression of novel genes.

(a) RNA-seq expression plots for the five LLC-like novel genes; each period of ZT 0–20 represents a distinct biological replicate. XLOC IDs were assigned by Cuffmerge. (b) Browser image of the crescendo transcript model and corresponding RNA-seq reads from merged replicate datasets of old flies, with the same y-axis range across time points. The dm6 phastCons conservation tracks were downloaded from the UCSC Genome Browser website. (c) crescendo expression in heads of females on day 5, 55 and 75 by qRT-PCR is reported as per cent of peak level at day 75. Data are mean (from two cycles)±s.e.m. *P≤0.05; **P≤0.01; ***P≤0.001; ****P≤0.0001 (two-way ANOVA with Bonferroni correction; day 55 change is relative to day 5 values, and day 75 change is relative to day 55 values).

Figure 6. Age-related and time-averaged differential expression.

(a) Scatterplot showing average daily gene expression in heads of old versus young flies according to RNA-seq. Data represent mean across six time points each with two biological replicates; differentially expressed genes are coloured according to q-values reported by Cuffdiff. Genes with significant (FDR 0.05) disparities in average daily expression between two biological replicates were included but not coloured. Labelled genes have an FPKM≥20 in young or old flies as well as an old/young fold change ≥5. (b) Heat map showing gene expression changes during aging according to RNA-seq. Each row represents a differentially expressed gene (FDR 0.01) with fold change ≥1.5 and a minimum FPKM of 1 in either young or old. Genes with significant (FDR 0.05) disparities in average daily expression between two biological replicates were excluded. For each sample (column), colour corresponds to fold change given by dividing the sample's FPKM by the average FPKM in young. Data from biological replicates are grouped in adjacent columns. Right: terms representing the top ten clusters of GO terms reported as enriched among up- or down-regulated genes by DAVID.