We performed FAIRE on early passage and senescent (Fig. S1
) normal human diploid fibroblasts (HDF) and interrogated the genome using Affymetrix Human Tiling arrays (FAIRE-chip). We first computed average FAIRE enrichments of all RefSeq genes in a TSS-centered window (). As expected, we found a strong peak located directly over the TSS. Interestingly, senescent cells displayed lower enrichment than early passage cells. To examine in more detail how FAIRE enrichment changes during senescence, we performed k-means clustering of the interpolated FAIRE signals of all the TSS in our dataset (). The FAIRE-enriched cluster contained notably more genes in early passage cells compared to senescent cells (56% and 33% of total genes, respectively). In addition, reinforcing the analysis of all TSS, the magnitude of the signal in FAIRE-enriched genes was decreased in senescent cells. We also examined FAIRE signal over predicted enhancers (Heintzman et al., 2009
) and found that it decreases in senescent cells (). Finally, we obtained gene expression data from corresponding early passage and senescent HDF using Affymetrix U133 microarrays, and intersected these data with the FAIRE clustering analysis. As expected, we found a positive correlation between FAIRE enrichment and gene expression (Fig. S2
Fig. 1 FAIRE analysis of TSS and enhancers. (A) FAIRE-chip: average FAIRE signal at TSS decreases in senescent cells. Enrichment (Y-axis) was calculated as the averaged log2 (FAIRE/input) signal for each probe within a 10 kbp window (X-axis) centered on all (more ...)
Since the resolution and dynamic range of tiling arrays are limited, we examined FAIRE-extracted DNA by high-throughput sequencing (FAIRE-seq). These experiments broadly confirmed all the findings from the tiling arrays, in that both the number of FAIRE-enriched genes as well as the relative magnitude of the signal were significantly decreased in senescent cells (), and that FAIRE signal correlated positively with gene expression (). We also independently verified the FAIRE enrichment over TSS and enhancers in the FAIRE-seq datasets (Fig. S3
), and used a bootstrap randomization method to confirm statistical significance.
A whole chromosome view of the FAIRE-chip data showed that while early passage cells displayed distinct banding of high and low FAIRE enrichment, these patterns were noticeably smoothened in senescent cells (Fig. S4
). This was reminiscent of the genome-wide redistribution of SIRT1 protein in response to DNA damage (Oberdoerffer et al., 2008
). When we examined the chromosomal distribution of the FAIRE signal in the FAIRE-seq data we again noted a pronounced smoothening in senescent cells. A genome browser view showed that in early passage cells the FAIRE signal was higher in gene-rich regions and showed numerous peaks, while it was lower and more uniform over gene-poor regions (, track 5). In senescent cells, the signal over gene-rich regions notably decreased, while the signal over gene-poor regions increased (, track 4). These changes were specific to senescent cells, as the FAIRE profile in quiescent cells was very similar to early passage cells (Fig. S5
Fig. 2 Large scale genome-wide distribution of FAIRE signals. (A) A genome browser view is shown for a representative 15 Mb region of the left arm of chromosome 16. FAIRE signal from early passage and senescent LF1 cells is shown in tracks 4 and 5; other tracks (more ...)
It is well documented in public databases that gene-rich regions are demarcated with numerous features of open chromatin, such as activating histone marks (H3K4me3), DNaseI-hypersensitive sites, or RNA polymerase II localization (). In contrast, gene-poor regions are enriched for features of heterochromatin, such as repressive histone marks (H3K9me3), and are known to replicate late (Hansen et al., 2010
). These features are constant in very different cell lines (HDF, HeLa, lymphoblastoid, etc.) and likely represent basic architectural features of the human genome ().
Changes in the distribution of FAIRE signal were confirmed in a genome-wide analysis. When analyzed in the context of early and late replicating regions, the FAIRE signal distributions were relatively broad in early passage cells, indicative of considerable local variation, and early replicating regions were enriched for FAIRE signal (). In senescent cells the distributions became much sharper and moved closer together. To specifically address changes taking place in the context of euchromatin and heterochromatin, FAIRE signal was additionally analyzed in regions demarcated by H3K4me3 and H3K9me3 modifications. FAIRE signal was positively enriched in H3K4me3 regions (), and this enrichment decreased in senescent cells, suggesting that these regions were becoming relatively more closed. In contrast, FAIRE signal was negatively enriched in H3K9me3 regions (), but here relative enrichment increased in senescent cells, suggesting that these regions were becoming more open. Statistical validation using bootstrap randomization showed that all changes were highly significant. It should also be mentioned that because of the manner in which FAIRE-chip and FAIRE-seq data are normalized, the resultant comparisons reflect the relative distribution of open and closed chromatin across a single epigenome (and hence indirectly between proliferating and senescent cells), rather than a direct comparison of openness between proliferating and senescent cells.
To gain more insight into the consequences of the changes in heterochromatic gene-poor regions, we extended our analysis to repetitive sequences, which are typically heavily heterochromatinized and comprise over 50% of our genomes. We applied recently developed software (Day et al., 2010
) for assessing repeat sequence coverage from high throughput sequencing data. We first focused our analysis on SINE and LINE retrotransposons, and specifically the Alu and L1 families, which together comprise some 30% of the human genome, and include elements that can actively transpose (Batzer & Deininger, 2002
; Deininger & Batzer, 2002
). Random transposition is deleterious, and heterochromatinization of retrotransposons is a key mechanism for limiting their spread. As expected, all subfamilies of Alu elements were under-represented in FAIRE-enriched DNA relative to input, but became more abundant in senescent cells (, Fig. S6
). Alu elements continue to transpose and evolve in the human germline genome, and the most evolutionarily recent subfamilies showed the highest degree of depletion in FAIRE as well as the largest increase in senescence. Another class of recent and active retrotransposons, the SVA elements, showed the same pattern of changes.
Fig. 3 Analysis of repetitive elements in early passage and senescent cells. (A) Relative abundance of Alu, L1, SVA and satellite elements in FAIRE-seq datasets. The representation of RepeatMasker annotated repetitive elements was computed for FAIRE-seq datasets (more ...)
L1 elements are a large and diverse class, but most of its members have been extensively truncated in the human genome. Only the most recent primate (L1P) and human (L1H) subfamilies contain full-length elements that are believed to be capable of autonomous transposition. A large proportion (~75%) of L1 subfamilies were enriched in FAIRE relative to input in early passage cells and became depleted in senescence, the same pattern that was noted for the majority of RefSeq genes (, Fig. S7
). The remaining 25% of L1 subfamilies showed the same behavior as the Alu and SVA elements, namely, depleted in FAIRE relative to input and more abundant in senescence. Interestingly, this group was comprised almost entirely of primate and human subfamilies (, Fig. S8
), with the most recent members (such as L1HS, L1PA2, L1PA3, L1PA4) being among the most extreme in this trend. Thus, for all the major classes of retrotransposons (Alu, SVA, L1), the most recent subfamilies showed evidence of the strongest heterochromatinization in normal cells (which is reasonable given that they pose the greatest risk of transposition), and importantly, the most profound relative opening in senescent cells.
To further validate these bioinformatic inferences, we biochemically probed FAIRE-extracted DNA for representation of repetitive sequences using a dot-blotting strategy, and found that both Alu and L1 sequences were relatively enriched in senescent cells (). We then examined the expression of Alu and L1 RNA by qRT-PCR, and found that it was correspondingly increased in senescent cells (). Finally, we investigated the copy number of L1 elements, and found a statistically significant increase of 11% in late senescent cells (). These PCR experiments employed primers designed to consensus sequences but are biased to interrogate recent primate and human subfamilies. (Coufal et al., 2009
Given the apparent loosening of chromatin in late-replicating heterochromatic regions, we wished to investigate how this may affect the retrotransposons located there. We designed primer pairs specific to individual elements of the recent AluYb9, L1PA3 and L1PA4 subfamilies that are located in these regions, and performed qPCR on FAIRE samples from early passage and senescent cells. The design of these primers was based on the presence of single nucleotide polymorphisms found in individual elements, and all primers pairs were empirically verified to amplify their targets with single-copy kinetics. Six to seven individual elements of each subfamily on several different chromosomes were investigated, and in all cases we observed a marked increase of signal in FAIRE DNA extracted from senescent cells (Fig. S10, 11
). Finally, the same primer pairs were used in qRT-PCR experiments, where in most cases we observed notable increases in the RNA expressed from these elements (Fig. S13, 14
). Overall, the magnitude of the changes observed for the individual elements, both at the DNA and RNA levels, exceeded the changes seen with consensus primers or bioinformatic analysis of FAIRE-seq data, suggesting that retrotransposons located in heterochromatic late-replicating regions may be particularly prone to derepression and activation.
Centromeres are the most prominent sites of constitutive heterochromatin in the genome. Centromeres and pericentromeres are comprised mostly of satellite repeat sequences, which play important roles in their heterochromatinization and interphase nuclear location (Pezer et al., 2012
). In the FAIRE-seq datasets satellite sequences were depleted in FAIRE and became prominently enriched in senescent cells (, Fig. S9
). Satellites are known to be transcribed in many organisms (Pezer et al., 2012
), which is believed to play an important role in their heterochromatinization (Volpe et al., 2002
). We next designed PCR primers specific for individual elements of the prominent human pericentromeric satellite II (hSATII), using the strategies described above. We found a significant enrichment on several chromosomes in FAIRE DNA extracted from senescent cells (Fig. S12
). Finally, we used the same primers in qRT-PCR assays, and found remarkable increases in hSATII RNA (in some cases 100 to 1000-fold) in senescent cells (Fig. S15
A considerable fraction of heterochromatin, including some late-replicating regions and centromeres, are localized close to the nuclear envelope in the peripheral heterochromatic compartment (PHC) (Carone & Lawrence, 2012
). The PHC was first described on the basis of its dark staining in electron microscopic (EM) studies. Using EM we found significantly less staining of the PHC in senescent cell nuclei, indicative of a decrease of heterochromatin (). We therefore examined centromeric structure by FISH using a hSATII probe. We found compact centromeric signals in early passage cells, which became distinctly enlarged in senescent cells (). By this assay centromeres and pericentromeres thus appear to adopt a relatively more open conformation.
Fig. 4 EM and centromere FISH analysis of senescent cell nuclei. (A) The peripheral heterochromatic compartment decreases in senescent cells. Representative electron micrographs (left panels) show the distribution of heterochromatin (dark staining) on the inside (more ...)