Epigenomic signature of accelerated ageing in progeroid Cockayne syndrome

Abstract

Cockayne syndrome (CS) and UV-sensitive syndrome (UVSS) are rare genetic disorders caused by mutation of the DNA repair and multifunctional CSA or CSB protein, but only CS patients display a progeroid and neurodegenerative phenotype, providing a unique conceptual and experimental paradigm. As DNA methylation (DNAm) remodelling is a major ageing marker, we performed genome-wide analysis of DNAm of fibroblasts from healthy, UVSS and CS individuals. Differential analysis highlighted a CS-specific epigenomic signature (progeroid-related; not present in UVSS) enriched in three categories: developmental transcription factors, ion/neurotransmitter membrane transporters and synaptic neuro-developmental genes. A large fraction of CS-specific DNAm changes were associated with expression changes in CS samples, including in previously reported post-mortem cerebella. The progeroid phenotype of CS was further supported by epigenomic hallmarks of ageing: the prediction of DNAm of repetitive elements suggested an hypomethylation of Alu sequences in CS, and the epigenetic clock returned a marked increase in CS biological age respect to healthy and UVSS cells. The epigenomic remodelling of accelerated ageing in CS displayed both commonalities and differences with other progeroid diseases and regular ageing. CS shared DNAm changes with normal ageing more than other progeroid diseases do, and included genes functionally validated for regular ageing. Collectively, our results support the existence of an epigenomic basis of accelerated ageing in CS and unveil new genes and pathways that are potentially associated with the progeroid/degenerative phenotype.

Keywords: Cockayne syndrome; DNA methylation; UV-sensitive syndrome; ageing; epigenetic clock; progeroid diseases.

FIGURE 1

Global features of DNA methylation in CS skin fibroblasts and profile of the differentially methylated positions in progeroid versus non‐progeroid groups. (a) Characteristics of primary skin fibroblasts derived from three healthy donors (WT), two UVSS patients and seven CS patients. A plus indicates the platform used for the DNAm analysis (seven samples analysed with both platforms), and the presence of a progeroid phenotype in individuals. For each cell type is shown the age at skin biopsy, age at death/age at latest follow up. The star (*) indicates that the individual was alive at the indicated age (A. Sarasin, personal communication). All age columns are expressed in years. F, female; M, male; N, no; S, suspected; Y, yes. More information on these patients (except UVSTA24VI) is available in Laugel et al. (2010). (b) Principal component analysis (PCA) of DNAm levels of probes assessed in the Infinium450k and InfiniumEPIC beadchips (merged dataset), after batch correction; dots are coloured according to the pathological (CS, UVSS) or control (WT) condition. (c) Boxplot of global percentage of DNAm in the merged dataset after batch correction. (d) Graphical representation of the approach used to identify CS‐specific (progeroid‐related) epigenetic signature. ‘CS‐specific’ refers to epigenetic changes present in CS but not in UVSS samples that, although harbouring the mutation, do not show the progeroid phenotype. Progeroid‐associated epigenetic changes result from comparing DNAm between CS (‘Progeroid’ group) versus WT and UVSS cells, merged in the same ‘Non‐Progeroid’ group. A CS‐specific CpG site will tend to have similar DNAm values in CS samples and differ from those in WT and UVSS cells (top panel). On the contrary, a CpG site whose DNAm is associated with the phenotype shared between CS and UVSS, like photosensitivity, will have similar values in CS and UVSS cells and differ from WT cells (bottom panel). Analysis of variance between the progeroid and the non‐progeroid group will tend to return significant p‐values in the first but not in the second case. (e) Volcano plot showing nominal p‐values vs the difference in mean percentage of methylation between the progeroid and the non‐progeroid group. Green: hypomethylation; gold: hypermethylation. (f) Enrichment of DMPs in different genomic regions. Odds ratio values are reported on the y‐axis. Green: hypomethylation; gold: hypermethylation. The position of the various regions is schematized in the bottom. (g) Heatmap and hierarchical clustering on DMPs methylation values in the various patient and donor‐derived cells.

FIGURE 2

Filtering and functional GSEA of methylated regions in progeroid versus non‐progeroid groups. Plots reporting the correspondence between the enriched GO terms identified in the progeroid vs non‐progeroid GSEA analysis (right y‐axis) and the genes associated with DMRs, x‐axis. This representation displays genes in the x‐axis (listed in Table S7, sheet 3), and their corresponding GO term annotation on the y‐axis. GO terms on the y‐axis are grouped according to GO categories (Biological process, Molecular function, Cellular component). Genes may belong to more than one GO category as well as more than one GO term. Indeed, the same gene can be annotated with in multiple semantically similar terms, and the three GO categories provide complementary biological information. Numbered red frames indicate identified clusters of genes (related GO terms/GO categories); #1: transcription factors implicated in embryogenesis/development, #2: ion/neurotransmitter membrane transporters, #3: synaptic neuro‐developmental genes. These genes are more precisely identified because they belong to at least two independent GO categories (the colour code of the corresponding GO category/ies is/are reported below the x‐axis, and their name is reported on the right).

FIGURE 3

Epigenetic age acceleration in CS and commonalities with regular ageing. Density of DNAm of each sample for the repetitive Alu (a) and LINE‐1 (b) elements. (c) Scatter plots of epigenetic age, calculated with the Skin&Blood clock, vs chronological age. The black line represents the regression of epigenetic age on chronological age. The age of samples (only CS) with a filled symbol was calculated on the last report/death since the age at the time of biopsy was not known, thereby their chronological age (x‐axis) is likely overestimated. (d) Boxplots of epigenetic age acceleration values according to the Skin&Blood clock in WT, UVSS and CS groups.

FIGURE 4

Correlation between DNA methylation and gene expression in skin fibroblasts and from different CS transcriptomic datasets. (a) For each cell type of our dataset, the expression of selected genes was plotted against the percentage of their DNAm. Out of 31 genes tested, 11 genes with a significant correlation between DNA methylation and gene expression are shown, eight of which were inversely correlated (EPB41, EPB49, ASAH1, HOXA11, VARS, SLC1A5, PRDM16, ZIC1), and three were directly correlated (SLC7A1, NDUFC2, CLDND1). The linear regression curve is indicated in red. The Pearson's correlation coefficient was used to assess the correlation between methylation and transcription levels. The R squared (R²) and p‐value (p) are indicated in each graph. The correlation was considered significant when p < 0.05 (gene name underscored in yellow). The characteristics of selected genes are indicated below. The direction of methylation changes (hypermethylation and hypomethylation in the progeroid vs non‐progeroid group) is indicated with gold and green highlight of the x‐axis label (% Methylation), respectively. Genes that showed differential methylation also in normal ageing and/or other progeroid diseases (10/11) are framed in black (rectangles). Genes present in the top list of DMRs and/or StringentDMRs (6/11) are framed in dark blue (rounded rectangles). Genes selected specifically for their function (HOXA11, CLDND1) are framed in grey (ovals). dTFs are identified with a red square on the left upper corner. CLDND1 was selected for being specifically differentially methylated in CS. These characteristics/criteria are not mutually exclusive. The remaining 20 genes are shown in Figure S7. (b) Heatmap of odds ratios representing the strength of association between genes associated with DMPs, DMRs or StringentDMRs and genes differentially expressed in different CS transcriptomic datasets (from iPSC‐derived Neurons (Vessoni et al., 2016), iPSC‐derived MSCs (Wang et al., 2020) and post‐mortem Cerebellum [_1 (Wang et al., 2014) and _2 (Okur et al., 2020)]. The statistical significance of overlaps was tested with Fisher's exact tests, and the corresponding p‐values are indicated in orange. (c) Venn diagram showing the total number of common genes between genes associated with StringentDMRs and genes differentially expressed in Wang et al. (2014) (Post‐mortem Cerebellum_1) (upper panel). Venn diagrams displaying among the Hypo‐(blue) or Hypermethylated (magenta) StringentDMRs, number of genes in common with genes found Down‐ (lower‐left panel) or Up‐regulated (lower‐right panel) in the same dataset. (d) Clustering of semantically related GO terms (left panel; molecular function right panel; biological process) identified in GSEA (green circles) and in the enrichment analysis of Wang et al. (2014) (purple circles) using REVIGO. Common GO terms are represented by the orange circles. Semantically similar terms are grouped in clusters identified by a ‘meta‐term’ (black). The significativity of GO term overlap and odds ratios are displayed in red. (e) Heatmap representing the direction of expression changes (green; up‐regulated, red; down‐regulated) of the 88 unique genes in common between genes associated with StringentDMRs and genes differentially expressed in the four previously mentioned CS transcriptomic datasets. Genes already tested for expression in Figure 4a and Figure S7 are underlined in orange. (f) Quantitative RT‐qPCR of ZIC4, PDE4B and IRX3 transcripts in the progeroid vs non‐progeroid groups. ZIC4 is common to StringentDMRs and three out of four transcriptomic datasets (TDs); PDE4B is common to StringentDMRs and 2 out 4 CS TDs, and IRX3 is common to DMRs and the 4 CS TDs. n = 3 independent experiments; mean ± SD; n.s, non‐significant; p, p‐value; unpaired t‐test vs non‐progeroid.