Fig 1.
Breakpoint junctions increase in non-dividing, ageing cells.
(A) Spontaneous rearrangements that occur in non-dividing cells should be heterogeneously distributed throughout the population and identifiable in DNA sequencing data. (B) Proportion of junctions expected to fall in coding regions vs proportion that do (based on simulated data; Materials and Methods). Points represent biological (yellow) or in-silico (grey) repeats. One-sided two sample Mann-Whitney U to test whether observed is less than expected (U = 0, p <0.001, N = 8). Note that proportion of simulated junctions in coding regions is similar to proportion of genome reported to be coding (excluding introns [59]). (C) Number of breakpoint Junctions Per Mapped Read (JPMR) passing filter (Materials and Methods) as a function of cellular age. Cells were cultured in rich medium until the optical density no longer increased, indicating that cells were no longer dividing, and used as Day zero for the chronological ageing timecourse. Samples were taken from the cultures at Days 0, 1, 2, 3, 4 and 5. JPMR are shown relative to Day zero (N = 8, confidence interval = 68%). Ochre line: total junctions, genome wide. Grey lines from top to bottom: subset of junctions formed among fragments of nuclear DNA (nDNA); between fragments of mitochondrial (mtDNA) and nDNA; among fragments of mtDNA. (D) Histograms showing number of junctions with various lengths of microhomology (right of dotted lines), blunt joints (between dotted lines), or non-homologous insertion (left of dotted lines) at the breakpoint and during cellular ageing. Histograms at different days are overlaid on top of each other, with early timepoints in blue through to late timepoints in red (see colour legend). Right: schemes for reads with each type of junction.
Fig 2.
Similar patterns of genome rearrangement in ageing human brain cells as in ageing yeast cells.
Analysis was performed on sequence data from putamen samples of neurologically normal Caucasian males [24], six of whom were young (blue; 19, 24, 28, 28, 29 and 30 year old donors) and six of whom were old (red; 67, 71, 78, 83, 85 and 89 year old donors). The 100bp reads from these samples, which were crudely enriched for mtDNA, were re-mapped to the whole human genome and filtered (Materials and Methods). (A) Histograms showing number of junctions with various lengths of microhomology (right of dotted lines), blunt joints (between dotted lines), or non-homologous insertions (left of dotted lines) at breakpoint. Histograms at different ages are overlaid on top of each other, with samples from young donors in blue and older donors in red (as in A). In these sparse data, the difference between young and old brain tissue was marginally significant at best (comparison of number of junctions per mapped read: pMann-Whitney = 0.23; correlation between sample age and JPMR: pPearson = 0.17; comparison of average microhomology length: pT-test = 0.05, suggesting slightly shorter mean microhomologies in old tissues, 8.15bp vs 8.28bp). (B) Proportion of junctions expected to fall in coding regions vs proportion that do (based on simulated data; Materials and Methods). One-sided two sample Mann-Whitney U to test whether observed is less than expected (U = 3, p <0.001). Simulated proportions approached the 1.5% proportion of the genome reported as coding [127] when more junctions were simulated.
Fig 3.
Local and global hotspots for age-associated junction formation.
(A) Illustration of complexity of junction formation over time. In this example, junctions are represented as a link between two points in any of the three S. pombe chromosomes (depicted around the edge in Mbp). The number of junctions in a region are stacked up in the outer ring. Blue: although a region may have many junctions at a late time point (bottom), these junctions may not be the result of ageing as they could have been present earlier (top). Red: true age-associated hotspots will show an age-associated increase. Some rearrangements may occur recurrently between the same two regions (green) or one region with many others (red). (B) Heatmap showing the ratio of junction counts between Days 5 and 0 (T5/T0; ‘old-young ratio’), in 20kb windows across the genome. An element in the heatmap depicting two regions with high levels of age-associated breakpoint formation will have a higher old-young ratio and appear darker. The average number of breakpoints at each window, genome-wide, is plotted in grey at the top. An intra-chromosomal region of Chromosome II is blown up at bottom right; local hotspots are exemplified by Region II:2940000–3080000, where the old-young ratio exceeds 25 in many bins along the top edge of the heatmap. Global hotspots are reflected by the dark diagonal emanating from the right end of Chromosome II and both ends of Chromosome III.
Fig 4.
Bodies of RNA associated with DNA at global hotspots.
(A) Scatter plot showing junctions at each of the global hotspots identified near telomeres (Fig 3). Coordinates of junctions in ~20kb region surrounding each hotspot are shown on x-axis, and coordinates of corresponding junctions in Chromosomes I-III are shown on y-axis. Annotations including pseudogenes (grey), tlh2 (red) and rRNA genes (black) are shown beneath x-axis. (B) smFISH experiment using probes against the housekeeping control gene rbp1 (green, left) and tlh2 (purple, middle and right), with DAPI-stained nuclei (blue, right). Non-dividing wild-type (wt) cells do not appear to show any tlh2 expression after 2 days (top row) or 4 days (second row) of chronological ageing, proliferating cells overexpressing tlh2 (tlh2OE) show strong, but highly heterogeneous expression, while proliferating sir2Δ mutants show homogeneous high expression in all cells. (C) Top: RNA mapping to an unannotated region downstream of tlh2 in young (100% cell viability) and old (50% cell viability) stationary-phase cells. Each sample’s read coverage is normalised to the total number of mapped reads for that sample, showing the mean of two replicates (reanalysed from ref. 67). Bottom: DRIP-seq in proliferating cells using the ⍺S9.6 antibody suggests that R-loops accumulate at the same region. The broad DRIP-seq signal points to the presence of multiple R-loops downstream of tlh2, reflected by the neighbouring peaks.
Fig 5.
Age-associated phenotypes in sir2 and RNase H mutants.
(A) High-throughput lifespan assay of rnh1Δ rnh201Δ (orange), sir2Δ (green) and wild-type (grey) cells. Points show the mean of three independent repeats, and bars show the 95% confidence interval. Samples for sequencing were taken at Day 0 and Day 3, because too few live cells were present for sequencing analyses at later timepoints. (B) Counts of indels (left) and junctions (right) per mapped read (PMR), relative to Day 0, in wild-type, sir2Δ and rnh1Δ rnh201Δ cells. Dots show relative counts PMR for each replicate. (C) Scatter plot as in Fig 4A showing the coordinates of junctions at the hotspot surrounding tlh2. The x-axis shows the coordinates of junctions in 20kb region surrounding the tlh2 hotspot (tlh2 gene depicted in red); the y-axis shows the coordinates of the corresponding junctions across the genome. Junctions from three repeats were pooled, per sample, for wild-type samples (left, grey) and sir2Δ samples (right, green), both at Day 0 (top) and Day 3 (bottom). (D) Normalised data from (C). All junctions within the region were collected for each replicate and normalised to the number of mapped reads in that sample. Data is shown relative to the median at Day 0.
Fig 6.
R-loops may contribute to local hotspots in absence of Scw1.
(A) Top: sliding mean (10bp windows) junction counts around coding regions (CDS) across genome. Bottom: mean ± standard error of junction counts in 250bp bins on either side of coding regions across genome. Comparisons between bins were performed with paired t-tests (** p<0.001; *** p<0.0001; n.s., not significant). (B) Scatter plot of 3’-UTR length vs number of junctions in that UTR. Each point represents a gene. Dotted lines show two standard deviations above population mean (solid line). All genes above dotted line were used for gene enrichment analysis. Scw1 targets >2x population standard deviation are in orange. (C) Western blot of Scw1-TAP at different days of cellular ageing. Coomasie blue was used as loading control. (D) Left: rapidly proliferating and early stationary phase cells (equivalent to Day 0 in C), which normally express Scw1, were analysed for R-loop formation in presence (wt) and absence (scw1Δ) of Scw1. Right: example images from chromosomal spreads; DNA stained with DAPI (blue) and R-loops by ⍺S9.6 antibody (yellow), with (+) or without (-) RNase H treatment which removes R-loops. Note that almost all of the S9.6 signal co-localizes with DAPI, as seen with increased magnification and brightness to reveal faint DAPI signals. Scale bar: 10μm. (E) Normalised read coverage at Scw1 target genes from DRIP-seq of scw1Δ and wild-type (wt) cells using the ⍺S9.6 antibody (ochre). The difference between strains is significant (one sample t-test: T = -2.52, p = 0.012). Samples after addition of RNase H are shown as control (blue). (F) Junctions per mapped read (JPMR), relative to Day 0, in wt and rnh1Δ rnh201Δ cells at tRNAs and core Scw1 target genes [85] as indicated. Bars show the mean (±95% confidence interval) JPMR at all tRNAs or Scw1 targets ±500bp. Days 0 and 3 were compared with paired t-tests for each genotype (tRNAs: wt T = 1.6 p = 0.11, rnh T = -4.8, p<0.0001; Scw1: wt T = 1.6, p = 0.12; rnh T = -6.4, p<0.0001). The wt likely showed no increase in junctions because chromosomal rearrangements normally only accumulate at later ageing timepoints (see Fig 1C).