Global reorganization of the nuclear landscape in senescent cells

Abstract

Cellular senescence has been implicated in tumor suppression, development, and aging and is accompanied by large-scale chromatin rearrangements, forming senescence-associated heterochromatic foci (SAHF). However, how the chromatin is reorganized during SAHF formation is poorly understood. Furthermore, heterochromatin formation in senescence appears to contrast with loss of heterochromatin in Hutchinson-Gilford progeria. We mapped architectural changes in genome organization in cellular senescence using Hi-C. Unexpectedly, we find a dramatic sequence- and lamin-dependent loss of local interactions in heterochromatin. This change in local connectivity resolves the paradox of opposing chromatin changes in senescence and progeria. In addition, we observe a senescence-specific spatial clustering of heterochromatic regions, suggesting a unique second step required for SAHF formation. Comparison of embryonic stem cells (ESCs), somatic cells, and senescent cells shows a unidirectional loss in local chromatin connectivity, suggesting that senescence is an endpoint of the continuous nuclear remodelling process during differentiation.

Graphical abstract

Figure 1

Senescence Is Accompanied by Local and Global Changes in the Interaction Pattern of the Genome (A) WI-38 hTERT/GFP-RAF1-ER cells show the characteristic SAHF phenotype upon senescence induction. (B) Interaction heatmaps for chr18q for growing (top) and senescence (bottom). Strong interactions are shown in red whereas weak interactions are in blue. Green lines (center) represent topologically associating domains (TADs) determined in growing cells. Highlighted TAD (black rhombus and pink TAD line) suggests a loss of internal interactions. (C) Top: genome browser shot of interactions reaching into and from within the depicted TAD. Growing cells show more TAD internal interactions than senescent cells. Bottom: normalized read counts from highlighted TAD. Internal contacts confirm loss of TAD internal interactions and external interactions increase in senescence, both in cis (light gray) and in trans (dark gray). (D) DNA-FISH showing separation between probes located within and adjacent to the highlighted TAD. Separation increases in senescence p = 0.00003 (Mann-Whitney-Wilcoxon test). Shown to the right are representative confocal microscopy planes of the FISH separation experiment. (E) Top: schematic of the open chromatin index (OCI), which describes regions changing their ratio between local and distal contacts. (F) Browser shot of OCI in two biological replicates (G, growing; S, senescence) over highlighted TAD (pink) shows a loss of local interactions. (G) Scatterplots comparing genome-wide OCI in 200 kb windows between growing and senescence. Points are colored by DNase accessibility as measured in growing fibroblasts (left) and %GC content (right). A cluster of points can be seen to deviate from the diagonal, which shows a loss of local contacts in senescence. These regions are the least accessible in the genome and have a low GC content.

Figure 2

Sequence Composition Predicts Structural Chromatin Dynamics in Senescence (A) Scatterplots showing OCI calculated in 200 kb windows for growing and senescence cells, separated by overlap with isochores. The greatest changes can be seen to occur in the L1 isochores. (B) Top: schematic of TAD boundary strength calculation. Bottom: scatterplot comparing the cross-boundary ratios over TADs genome wide, colored by isochore. The L-isochores can be seen to gain cross-boundary interactions in senescence. (C) Top: selection of opening and closing TADs highlighted in scatter plot (a less-stringent cutoff was chosen for closing TADs, in order to reach a comparable number). Bottom: enrichment for overlap with genomic features (log2 obs/exp; also see Experimental Procedures) for opening (Op) and closing (Cl) TADs.

Figure 3

Combined Sequence Composition and Lamin Association Predict the Strongest OCI-Gaining Regions, which Are Changing Their Nuclear Positioning (A) Change in OCI levels in senescence for regions overlapping LADs (purple) and inter-LADs (iLADs, green). (B) Enrichment of LADs and iLADs for replication timing, LMNB1 regions, and several chromatin marks. (C) Browser shot indicating genomic location for FISH probes designed against two adjacent H2-LAD and L1-LAD regions (green vertical lines). (D) Distances of FISH signals to the nuclear periphery. (E) Left: representative confocal microscopic images of FISH-treated growing and senescent cells. Right: schematic showing the measurements made for the DNA-FISH from the central focal plane.

Figure 4

Local Changes Are Accompanied by the Formation of Specific Distal Interactions (A) Schematic showing how the change in interaction strength between features was calculated. g and s supertext denotes growing and senescence. Note the avoidance of features within 2 Mb. (B) Change in interaction strength calculated between isochore LADs comparing growing and senescence (C) Schematic depicting the chromatin organization of histone marks in SAHF as described in Chandra et al. (2012). (D) Change in interaction strength calculated between regions associated with histone marks. (E) Model depicting the change of the chromatin architecture. We propose that L-LAD regions with strong local interactions (high OCI) in growing (green) detach from the nuclear lamina and lose their internal structure (potentially forming the core of the SAHF). Other regions, such as selected H-LADs, move toward the nuclear periphery.

Figure 5

Presenescent Replication Timing Predicts Chromatin Changes (A) Change in replication timing between growing OIS ER-Ras cells and presenescent cells. Regions overlapping LADs are shown in green and iLADs in purple, divided by isochore. (B) Scatterplot showing difference between presenescent RT and growing RT versus growing RT for isochore L1; LADs (green) and iLADs (purple) are highlighted. Points above the horizontal are replicating earlier in presenescent cells. (C) Browser shot showing a L1-LAD region changing in presenescent RT.

Figure 6

Higher-Order Chromatin Dynamics in Senescence Reflect Changes in Progeria and May Represent the Endpoint of a Continuous Remodelling Process in Differentiation (A) Hierarchical clustering of OCI values in 1 Mb of genomic windows for growing, senescence, and progeria. Columns to the right show GC content per genomic window and the average %GC per cluster. (B) Enrichment of clusters over genomic features. Note that clusters 1 and 5 show similar behavior in growing and senescence. (C) Change of interactions calculated between regions associated with histone marks. Progeria shows no clustering of H3K9me3 regions or other histone marks compared to growing cells. Enrichment calculated as shown in Figure 4A. (D) Average OCI over LADs (top) and iLADs (bottom) split by isochore in ESCs, growing, and senescence. (E) Conservation of OCI domains called using a hidden Markov model (see Experimental Procedures) in ESCs. The x axis shows percentage of windows within domains classed as local (0) or distal (1). ESCs show a bimodal distribution as expected. Growing and senescent cells show a decaying conservation of interaction state within these domains.