Heterochromatin formation and senescence

Initial studies previously established that chromatin becomes reorganised with increased population doublings, with large-scale effects being evident upon senescence (Ref. Reference Macieira-Coelho28). It has now been shown that the formation of heterochromatin at specific foci plays a significant role in senescence in response to oncogenic insult and telomere attrition, and that this senescence arrest as a cellular response to stress might act as a tumour suppressor mechanism (Refs Reference Campisi2, Reference Mathon and Lloyd3). In addition, it has also been observed that many genes upregulated at senescence are actually physically clustered (Ref. Reference Zhang, Pan and Cohen29), again indicating that re-organisation of chromatin domains are important in senescence. Therefore, formation of heterochromatin may be viewed as a tumour suppressor mechanism.

Early studies showed that expression of oncogenic Ras induced the accumulation of p16^INK4a (also known as cyclin-dependent kinase inhibitor 2A) and hypophosphorylated pRB (retinoblastoma protein) with a senescence-like arrest in normal human fibroblasts (Ref. Reference Narita30). It was also noted that this senescence-like arrest was associated with the formation of distinct heterochromatic structures, which were designated senescence-associated heterochromatin foci (SAHF) (Ref. Reference Narita30). Importantly, SAHF were shown to be formed by newly deposited heterochromatin rather than from redistributed heterochromatin, and were concentrated for repressive histone modifications such as methylation of K9 H3 and heterochromatin protein 1 (HP1), whereas permissive modifications, such as methylated K4 H3 and acetylated K9 H3, were excluded (Fig. 2). It was also established that active transcription was not occurring at the sites of SAHF formation and that chromatin features indicative of SAHF formed on E2F-responsive promoters of genes known to be downregulated at senescence. E2F is a transcription factor important for regulation of genes associated with the cell cycle and was first identified as a cellular factor that had the ability to bind the promoter for the gene encoding adenovirus E2. The genes that gained heterochromatic features and became downregulated included the genes encoding cyclin A and proliferating cell nuclear antigen (PCNA), and their promoters were bound by pRB, thereby linking pRB binding to the formation of SAHF. It is known that the upregulation of p16^INK4a during senescence can inhibit the phosphorylation of pRB, causing it to bind to E2F to cause a permanent growth arrest by silencing growth regulatory genes through altering the chromatin state of the promoter sequences. pRB has been shown to interact with factors involved in the formation of heterochromatin, including histone deacetylases (HDACs) (Refs Reference Luo, Postigo and Dean31, Reference Magnaghi-Jaulin32, Reference Brehm33), the K9 methyltransferase SUV39H1 (Ref. Reference Vandel34) and HP1 proteins (Ref. Reference Nielsen35); thus, pRB binding to regulatory sequences could be sufficient to form SAHF in order to silence gene expression. Importantly, the authors showed in the same study that promoters of genes known to be upregulated at senescence gained euchromatic patterns at their promoters at the same time as SAHF were being formed on other promoters, suggesting that the formation of SAHFs might be directed towards certain genes and not an indiscriminate global effect. This indicates that modification of the chromatin environment of gene promoters is important for the stable senescence-like arrest attained. It also suggests a model in which SAHF formation accompanies the senescence process and indicates a role for pRB-mediated heterochromatin formation at growth regulatory genes during senescence, leading to the stable repression of these genes. From this it was further proposed that failure of certain processes in these cells (e.g. mutations in genes encoding p16^INK4a or pRB, or silencing events) could lead to bypass of senescence and progression to cancer as a result of the inhibition of SAHF formation (Ref. Reference Narita30).

Figure 2

The formation of senescence-associated heterochromatin foci (SAHF) in response to oncogenic insult. In this model, (a) an active gene promoter (e.g. cyclin D1) shows methylation of Lys4 (K4) H3 and acetylation of H3 and H4, and its chromatin has an open (euchromatic) structure. Under these conditions, the transcription factor E2F mediates transcription, whereas retinoblastoma protein (pRB) remains in a phosphorylated state as a result of low p16^INK4a levels and therefore cannot bind to E2F to repress transcription. (b) Upon oncogenic insult (e.g. aberrant Ras signalling), pRB becomes hypophosphorylated, mediated by increased levels of p16^INK4a. Hypophosphorylated pRB is sequestered to the promoter along with several other proteins, such as heterochromatin protein 1 (HP1), histone deacetylase (HDAC) and the Lys9 (K9) H3 methyltransferase SUV39H1, which can bind pRB. (c) This leads to the silencing of gene expression by altering the chromatin architecture towards a heterochromatic state by promoting K9 H3 methylation and removing/inhibiting euchromatic modifications, as is observed in SAHF formation. SAHF formation as a mode of gene expression regulation in response to oncogenic insult might lead to the onset of premature senescence and might therefore inhibit tumourigenic growth.

Several other papers have provided more evidence of the link between chromatin remodelling and senescence. Braig et al. have demonstrated that oncogenic Ras could induce a senescence-like arrest in mouse primary lymphocytes by a SUV39H1-dependent and K9 H3 methylation-related mechanism (Ref. Reference Braig36). In the same study, the deletion of SUV39H1 in primary lymphocytes led to the improper methylation of K9 H3 and the promotion of Ras-driven lymphomagenesis. This study also correlates with other work demonstrating that loss of methylation of K9 H3 and also K20 H4 resulted in genomic instability in a variety of normal cells and was associated with increased tumour risk (Ref. Reference Peters37). Decreased expression of SUV39H1 has also been observed in the early premalignant stage of tumour development in the context of hepatocarcinogenesis induced by methyl deficiency in rats, whereas loss of tri-methylation of K20 H4 was also observed at both preneoplastic and tumour stages of liver cancer and corresponded to a decrease in the K20 H4 histone methyltransferase, SUV420H2 (Ref. Reference Pogribny38). Interestingly, loss of tri-methylation of K20 H4, as well as loss of acetylated K16 H4 and DNA hypomethylation, has been shown to be a common epigenetic change in tumourigenic cells (i.e. an ‘epigenetic signature’ of cancer) (Ref. Reference Fraga39). These studies provide evidence that improper heterochromatin maintenance might lead to genomic instability and progression to tumourigenicity.

Michaloglou et al. have also demonstrated that BRAF^E600-mediated oncogenic signalling in melanocytes induces a senescence-like phenotype with the formation of SAHF (Ref. Reference Michaloglou40). BRAF is a protein kinase and downstream regulator of Ras, and the V600E mutation in the gene encoding BRAF is an oncogenic mutation observed frequently in human naevi (moles). A further study focusing on Ras utilised a mouse model with a conditional oncogenic K-rasV12 allele activated by the Cre recombinase (Ref. Reference Collado41). Upon activation of oncogenic Ras, the mice developed multiple lung adenomas (pre-malignant) and a few lung adenocarcinomas (malignant), and markers for senescence were evident in the adenomas but not in the carcinomas. Interestingly, HP1γ was strongly positive in the adenomas, whereas the carcinomas were negative. As the HP1 proteins can bind methylated K9 H3 to mediate gene silencing, this study further reinforces the role of epigenetics in senescence. The formation of SAHF has also been shown in response to overexpression of E2F3 (Ref. Reference Lazzerini Denchi42) and repression of the E1A (adenovirus 5 early region 1A protein)-associated protein p400 (Ref. Reference Chan43). In the first study, which analysed whether deregulated E2F activity would be sufficient to mimic loss of pRB, short-term deregulation of E2F3 led to hyperplasia of quiescent melanotrophs but did not lead to tumourigenicity due to the induction of senescence with the appearance of SAHF (Ref. Reference Lazzerini Denchi42). p400, a protein belonging to the SWI2/SNF2 family of chromatin-remodelling proteins and postulated to be a transforming agent, was studied in relation to the p53/p21 senescence pathway (Ref. Reference Chan43). When p400 expression was repressed in normal human fibroblasts and in human fibroblasts transduced with human telomerase reverse transcriptase (hTERT), senescence was induced along with the appearance of SAHF, suggesting that p400 expression could repress p53 induction of p21 expression, possibly allowing the bypass of the senescence arrest.

Further studies into the formation of SAHF have also indicated roles for the chromatin regulators histone regulating A (HIRA) and anti-silencing factor 1 (ASF1), which were shown to be rate limiting for the formation of SAHF and were required for senescence (Ref. Reference Zhang44), whereas the high-mobility group A (HMGA) proteins have been shown to be major structural components of SAHFs (Ref. Reference Narita45). HMGA proteins are usually associated with the formation of chromatin environments permissive for gene transcription, but have been shown to be vital for SAHF formation and gene repression in Ras-induced senescence of normal human fibroblasts (Ref. Reference Narita45). Also, similar to the role of pRB, the growth regulatory protein prohibitin, which has previously been linked to the regulation of senescence (Ref. Reference Liu46), has also been shown to localise to SAHFs induced by DNA-damaging agents (Ref. Reference Rastogi47). It was shown that prohibitin plays an integral role in the stability of the senescent phenotype by maintaining the repression of E2F-responsive promoters through the maintenance of a repressive chromatin environment by interaction with HP1γ. Reduction in the cellular levels of prohibitin led to a reduction in the number of SAHFs and impaired the ability of the cells to undergo senescence, further indicating an integral role for epigenetics in cellular senescence.

The formation of SAHF in senescence arrest has also been studied in vivo in ageing primates (Ref. Reference Herbig48). Increases in the presence of telomere-dysfunction-induced DNA damage foci were correlated to increasing age in dermal fibroblasts and, importantly, this was further correlated to SAHF formation. This study indicates that telomere dysfunction, occurring as a result of telomere attrition with increased cellular proliferation, also induces SAHF formation. Whether this occurs in human cells with ageing is as yet unknown, but this study does indicate that SAHF formation at senescence is not only a response to oncogenic stress, but might also be a common mechanism in cellular senescence.

In summary, multiple studies have now shown the strong correlation between the formation of heterochromatin and the attainment and maintenance of the senescence phenotype. Repression of proliferation-associated genes by the formation of heterochromatin serves to block cellular division and so acts as a potent tumour suppressor mechanism. Furthermore, improper regulation of heterochromatin is associated with deregulated cellular proliferation and tumourigenesis.

Epigenetic modifications by the senescence mediators p53 and ING

Similar to pRB, other tumour suppressor proteins have also been shown to interact with histone-modifying enzymes to regulate transcription (Ref. Reference Allison and Milner49). p53 is an important mediator of senescence and one of its main transcriptional targets is p21. It is known that p53 binds to and activates the p21 promoter and is mediated, in part, by increased promoter-proximal histone acetylation (Ref. Reference Espinosa and Emerson50) through the recruitment of histone acetyltransferases (HATs) (Ref. Reference Barlev51) and the displacement of HDAC1 (Ref. Reference Lagger52). Interactions between p53 and the HAT p300 and protein arginine methyltransferases, which can also mediate transcriptional activation by post-translational modifications of histones, have also been reported to be required for p53-dependent gene activation (Ref. Reference An, Kim and Roeder53). However, p53 also interacts with histone deacetylases to mediate gene repression as it has been shown that p53 binding to promoters was concomitant with increases in levels of K9 methylation and loss of acetylation of K9 H3 (Refs Reference Murphy54, Reference Nguyen55, Reference Cui56). Given that epigenetics plays a key role in transcriptional control by p53, it is important to note that deregulation of epigenetic mechanisms could affect p53-mediated senescence or apoptosis.

The inhibitor of growth (ING) family of proteins are also important regulators of cellular senescence through the regulation of gene expression (Ref. Reference Russell57). Like p53, HATs and HDACs have both been shown to interact with the ING proteins to mediate chromatin-remodelling activities and gene expression (Ref. Reference Feng, Hara and Riabowol58). Studies have shown a direct interaction of an ING family member with chromatin upon senescence, suggesting that ING proteins might be important mediators of senescence through epigenetic control of gene expression (Refs Reference Vieyra59, Reference Vieyra60). Furthermore, it has been shown that ING2 can bind di- and tri-methylated K4 H3 to mediate gene repression through histone deacetylation by associated factors in the mSin3a–HDAC repressor complex (Ref. Reference Shi61), allowing subsequent repressive lysine methylation required for the formation of heterochromatin. This reversal of chromatic states (from euchromatic to hetero-chromatic) is an important mechanism to repress proliferation-associated genes at senescence, and one method to realise such a reversal could be the sequestering to chromatin of ING2-mediated histone deacetylation by a modification associated with active gene expression (methylation of K4 H3).

Reversing the irreversible: a new epigenetic mechanism uncovered

Methylation of K9 H3 plays an important role in the maintenance of gene repression and in SAHF-mediated gene repression at senescence. Methylation of K9 H3 was generally understood to be a stable, long-lasting modification, allowing for effective gene silencing. However, multiple studies have now shown that methylation of K9 H3 is readily reversible (Refs Reference Yamane62, Reference Whetstine63, Reference Klose64, Reference Cloos65, Reference Metzger66) and it has been suggested that enzymes that mediate K9 H3 demethylation could also affect tumourigenesis (Ref. Reference Cloos65). GASC1 (for ‘gene amplified in squamous cell carcinoma 1’; also known as JMJD2C, for ‘jumonji domain containing 2 family member C’) is one such protein found with the ability to demethylate di- and tri-methylated K9 H3 (Ref. Reference Cloos65). This, coupled with previous findings that GASC1 was amplified in several carcinoma cell lines and that inhibition of the protein led to a decrease in cellular proliferation, led to the suggestion that GASC1 could act as an oncogene. Hypothetically, overexpression of GASC1 or other K9 H3 demethylases could lead to the dissolution of SAHF and to the bypass of senescence by the de-repression of proliferation-associated genes. In such a model, the formation of SAHF in response to a senescence-inducing signal would ‘force’ cells to undergo premature senescence and stop possible tumourigenicity. However, if SAHF formation was inhibited by the overexpression or aberrant localisation of a K9 H3 demethylase, this could allow cells to move preferentially towards tumourigenesis and bypass senescence (Fig. 3).

Figure 3

The role of senescence-associated heterochromatin foci (SAHF) in senescence and tumourigenesis. (a) Formation of SAHF after oncogenic stress in a ‘normal’ cell might promote the induction of senescence over tumourigenesis by the repression of specific genes. (b) However, if SAHF formation is inhibited, potentially by the reversal of Lys9 (K9) H3 methylation, tumourigenesis may be promoted owing to ‘blockage’ of the route towards senescence.

Polycomb group proteins, stem cells and senescence

Although it has been shown that the gain of heterochromatin by increased levels of repressive histone modifications is linked to senescence, the gain of repressive histone-modifying activity by Polycomb group (PcG) proteins and gene silencing has also been observed to allow extension of lifespan and is involved in tumourigenesis. There are two distinct PcG complexes, Polycomb repressive complex 1 and 2 (PRC1 and PRC2); these multiprotein complexes can repress target gene expression by catalysing the addition of repressive modifications of histones, such as K27 H3 methylation, and allowing the compaction and maintenance of chromatin into a heterochromatic structure.

Loss of function of the PcG protein BMI1 (for ‘B-cell-specific Moloney murine leukaemia virus integration site 1’) was shown to impair the progression of mouse embryonic fibroblasts into S-phase; indeed, these cells underwent premature senescence, thereby linking transcriptional repression by PcG proteins with cell-cycle control and senescence (Ref. Reference Jacobs67). Additionally, BMI1 has been shown to be downregulated in senescing human fibroblasts (Ref. Reference Itahana68), whereas overexpression has been observed in tumours and tumour cell lines (Refs Reference Bea69, Reference Vonlanthen70, Reference Nowak71, Reference Sawa72) and BMI1 is required for the short-term survival of cancer cells (Ref. Reference Liu, Andrews and Tollefsbol73). Interestingly, additional studies have shown that overexpression of BMI1 allows immortalisation and transformation of some human cells (Refs Reference Jacobs67, Reference Leung74, Reference Dimri75). Importantly, the INK4a locus from which the cyclin-dependent kinase inhibitors p16^INK4a and p14^ARF are transcribed are crucial downstream targets for BMI1 with regard to its effects on cell proliferation and senescence (Ref. Reference Itahana68). Therefore, through inhibiting the expression of these cyclin-dependent kinase inhibitors, BMI1 can lead to the promotion of cell-cycle progression.

EZH2 (for ‘enhancer of zeste homolog 2’), another PcG protein, has been shown to methylate K27 H3 (Ref. Reference Cao and Zhang76) and is also overexpressed in some cancers, again linking the formation of heterochromatin to the extension of lifespan (Refs Reference Varambally77, Reference Kleer78, Reference Bracken79). EZH2 expression is also known to be repressed by pRB and activated p53, suggesting a possible mechanism for EZH2 overexpression in cancers (Ref. Reference Bracken79) and is further observed to be specifically downregulated in senescent cells (Ref. Reference Tang80). Loss of K27 H3 histone methyltransferase activity at senescence might correspond to loss of heterochromatin and activation of genes such as those encoding p16^INK4a and p14^ARF, leading to cellular senescence (Ref. Reference Villeponteau81). By contrast, overexpression of EZH2 in cancers might lead to the proper maintenance of heterochromatin at the INK4a locus, allowing the elongation of lifespan in cancer cells by bypassing senescence through the repression of expression of genes such as those encoding p16^INK4a and p14^ARF.

Complementing these reports have been recent studies on epigenetics in embryonic stem cells (ESCs) and how epigenetics affects pluripotency; these studies have suggested another route by which a normal cell might bypass senescence and gain tumourigenic properties. The chromatin environment of ESCs has been shown to be hyperdynamic and therefore in a highly permissive state for transcription. However, as a cell differentiates along various lineages, specific genes will become repressed, whereas others are further activated, leading to a less dynamic and therefore less permissive chromatin environment (Refs Reference Meshorer82, Reference Meshorer and Misteli83, Reference Szutorisz and Dillon84). PRC1 and PRC2 have now been shown to be integral in maintaining pluripotency by initiating and maintaining the silencing of genes involved in differentiation by marking such genes with K27 H3 methylation (Refs Reference Boyer85, Reference Bernstein86, Reference Azuara87, Reference Lee88). Therefore, if any of the components of PRC1 and PRC2 become overexpressed in a normal cell, genes involved in differentiation could become epigenetically silenced and perhaps allow the reversion of a normal cell towards a more pluripotent state (Fig. 4). This could allow pluripotent stem-like cells to arise in normal tissues and nucleate cancer growth. Supporting this, as mentioned previously, it has been shown that genes such as those encoding BMI1 and EZH2 are overexpressed in some cancers (Refs Reference Bea69, Reference Vonlanthen70, Reference Nowak71, Reference Sawa72, Reference Varambally77, Reference Kleer78, Reference Bracken79).

Figure 4

Attainment of pluripotency in normal cells: a possible path towards the cancer stem cell. (a) In embryonic stem cells (ESCs), Polycomb group (PcG) complexes [Polycomb repressive complex 1 and 2 (PRC1 and PRC2)] function to repress differentiation-associated genes by repressive histone modifications [Lys9 (K9) H3 and Lys27 (K27) H3 methylation], whereas pluripotency genes are expressed with the presence of permissive histone modifications [acetylation of histones H3 and H4 and methylation of Lys4 (K4) H3]. This allows for the self-renewal capacity characteristic of stem cells. (b) Loss of the PcG-mediated repressive modifications leads to the expression of lineage-specific genes upon differentiation of ESCs, leading to the cell becoming more committed to a certain lineage. Pluripotency genes also become repressed and therefore these cells lose some ESC characteristics. (c) Overexpression of PcG proteins is frequently observed in cancer cells and it is possible that overexpression of PcG proteins in normal cells could allow the silencing of lineage-specific genes involved in differentiation by repressive histone modifications. (d) This might also allow the cell to re-express pluripotency-associated genes and revert back to a pluripotent state, undergoing deregulated cell proliferation. This may be an epigenetic mechanism by which normal cells can bypass senescence and grow without limit, allowing sustained cell proliferation and self-renewal. Therefore, this could represent a new mechanism by which cancer stem cells can arise and nucleate cancer growth.

These studies firmly establish a link between PcG group proteins, senescence and the modulation of the chromatin environment. Supporting this hypothesis is a recent study of PcG group proteins in prostate cancer (Ref. Reference Berezovska89). This study showed that increased expression of BMI1 and EZH2, with increases in K27 H3 methylation and gene silencing, were linked to increased malignant potential and metastasis in prostate cancer and were also linked to failure in therapy of prostate cancer. Prior work by the same group also identified an 11-gene signature that was common between normal stem cells with normal self-renewal capacity and malignant tumour cells (Ref. Reference Glinsky, Berezovska and Glinskii90). These papers suggest that PcG-mediated gene silencing can mediate increased malignancy perhaps by bestowing stem cell properties to previously non-metastatic cancer cells with low malignant potential; this could occur through the repression of multiple developmental and differentiation pathways by epigenetic changes to gene expression. Thus, PcG-mediated pathways are now known to be involved in both normal stem cell function and highly malignant tumour cells. This reinforces the concept that regulation of the chromatin environment and gene regulation by the PcG pathway is very important to proper stem cell function and to prevent highly malignant metastatic tumours from arising.

Support for the ‘reversion’ theory has been shown in a study of committed progenitor cells expressing the mixed lineage leukaemia (MLL)–AF9 fusion protein known to impart leukaemia stem cell (LSC) properties (Ref. Reference Krivtsov91). These cells gained the expression of a subset of genes highly expressed in haematopoietic stem cells (HSCs) and also gained self-renewal properties. The MLL protein is a common fusion partner in leukaemias and is a trithorax group (Trx-G) member involved in the transcriptional activation of gene expression; MLL is known to mediate methylation of K4 H3 (Ref. Reference Daser and Rabbitts92). This study is interesting as it links the action of an epigenetic modifier with the reversion of a more committed cell towards a more primitive stem cell type and tumourigenesis.

These studies also raise questions against the rising popularity of the stem cell theory of cancer (Ref. 93), by suggesting that some cancers might arise from the gain of pluripotency by deregulation of the chromatin environment in normal somatic cells. However, lending support to the cancer stem cell hypothesis is the premise that epigenetic disruption of stem cells might be a common unifying theme in cancer aetiology (Ref. Reference Feinberg, Ohlsson and Henikoff27), which is an attractive proposition. This so-called ‘epigenetic progenitor model’ might help to explain many different aspects about cancer, including initiation, progression, and genetic and epigenetic instability. To understand whether one of these models is predominant or whether they work in conjunction to foster tumourigenesis is an important goal.

Histone acetylation and senescence

As previously mentioned, histone acetylation might also have an important role in mediating the senescence phenotype. The HATs p300 and CREB-binding protein (CBP) were both studied in senescing human melanocytes and, interestingly, levels of both proteins were observed to decrease as these cells underwent replicative senescence (Refs Reference Bandyopadhyay94, Reference Yao95). The decrease in the levels of these HATs led to dramatic decreases in total histone H4 acetylation levels. Specifically, the promoter for the gene encoding cyclin E was shown to lose H4 acetylation and become repressed upon senescence, whereas cyclin E was highly expressed and H4 acetylation was high in proliferating cells. This parallels previous studies of the cyclin E promoter during senescence, which showed that pRB and HDACs bound to the promoter to repress transcription, linking the senescent phenotype to the pRB/p16^INK4a pathway and the formation of SAHF (Ref. Reference Narita30). As mentioned previously, p53 also interacts with enzymes that control histone acetylation, indicating a role for two main senescence mediators in the modulation of histone acetylation.

However, promotion of histone acetylation by the use of HDAC inhibitors can induce a senescence-like phenotype in human fibroblasts (Refs Reference Ogryzko96, Reference Munro97, Reference Place, Noonan and Giardina98). Furthermore, in senescing normal human fibroblasts, HDAC1 becomes downregulated (Ref. Reference Place, Noonan and Giardina98), indicating that overall the promotion of acetylation leads to activation of the senescence phenotype. Interestingly, histone deacetylation has also been shown to have a role in the extension of cellular lifespan. The yeast protein Sir2 (for ‘silent mating type information regulation 2’), identified as a NAD ⁺-dependent HDAC involved in ribosomal DNA silencing, contributes to the replicative lifespan of yeast (Refs Reference Kaeberlein, McVey and Guarente99, Reference Imai100). Enhanced activity of the gene encoding Sir2 was also found to be associated with increased longevity in Caenorhabditis elegans (Ref. Reference Hekimi and Guarente101), Drosophila (Ref. Reference Rogina and Helfand102) and rodents (Ref. Reference Cohen103). Drosophila Sir2 has also been shown to be involved in epigenetic gene silencing by the PcG proteins and is physically associated with a complex containing the Drosophila homologue of EZH2. This further implicates Sir2 in the control of senescence (Ref. Reference Furuyama104) and suggests a role for Sir proteins in PcG-mediated gene repression that might affect cellular proliferation and lifespan as discussed above. A human homologue of yeast Sir2, SIRT1, has also been associated with longevity and is believed to act primarily by inhibiting cellular senescence (Refs Reference Yeung105, Reference Brunet106, Reference Vaziri107) and overexpression has been observed in leukaemia cells (Ref. Reference Bradbury108). SIRT1 and other Sir2 homologues are able to deacetylate transcription factors and p53, and also have K16 H4 deacetylase activity (Ref. Reference Michishita109). Recently, a SIRT1 inhibitor, Sirtinol, was shown to induce a senescence-like growth arrest in human cancer cells (Ref. Reference Ota110) in agreement with previous data indicating that the use of HDAC inhibitors could lead to induction of the senescence phenotype. Additionally, loss of histone acetylation has been linked to tumourigenesis through decreased K16 H4 acetylation, as well as decreased tri-methylation of K20 H4 and DNA hypomethylation, as demonstrated in cancer cells and tissues (Ref. Reference Fraga39).

These studies suggest that the increases in global histone acetylation can lead to senescence, perhaps mediated by the re-expression of genes previously silenced by heterochromatin that, when activated, promote senescence. By contrast, de-acetylation might serve to repress genes that contribute to the senescence phenotype. Again, this shows that regulation of the chromatin environment is crucial for proper cellular control.

DNA methylation and senescence

Generally, histone modifications and DNA methylation are often discussed separately in relation to their effects on gene expression, although it is now known that they are in fact intimately linked (Refs Reference Stancheva111, Reference Lachner and Jenuwein112, Reference Vire113, Reference Esteve114), with multiple interactions now known between proteins associated with histone modifications and those associated with DNA methylation. Generally, DNA methylation of CpG islands (regions of adjacent cytosine and guanine residues linked by phosphodiester bonds) at gene promoters is correlated with gene repression, whereas expression is linked to the lack of methylation. Global DNA methylation has been shown to decrease with age (Ref. Reference Wilson and Jones115) and this loss had been hypothesised to function as a counting mechanism for senescence, such that senescence would occur when a cell loses a certain degree of DNA methylation (Refs Reference Wilson and Jones115, Reference Hoal-van Helden and van Helden116). Treatment with 5-azadeoxycytidine (5-azadC), a DNA demethylating agent, has been shown to reduce the replicative lifespan of normal human fibroblasts, supporting the DNA methylation counting mechanism hypothesis (Refs Reference Holliday117, Reference Fairweather, Fox and Margison118). Also in agreement with this hypothesis are the findings that DNA methyltransferase activity is elevated in cancer cells in vitro, in tumours in vivo (Ref. Reference Belinsky119) and in transformed cells (Refs Reference Slack120, Reference Rouleau, MacLeod and Szyf121), whereas DNA methyltransferase activity decreases in serially passaged normal human fibroblasts (Ref. Reference Issa122). Therefore, the loss of DNA methylation with age might eventually have deleterious consequences, such as deregulation of gene expression. Consistent with this view, it has been hypothesised that DNA methylation might repress a set of growth inhibitory genes (Refs Reference Jones and Baylin24, Reference Young and Smith123) and, as the number of cell population doublings increases and DNA methylation decreases, the repression of such growth inhibitory genes might become less stringent, leading to their eventual expression and contributing to senescence. However, it has also been shown that a striking feature of some tumours (both benign and malignant) is that global DNA methylation is reduced (Refs Reference Feinberg and Vogelstein124, Reference Goelz125). This loss of DNA methylation is also linked to chromosomal instability, which can contribute to tumourigenesis (Refs Reference Eden126, Reference Gaudet127). Thus, paradoxically, it has been shown that the loss of DNA methylation can contribute to senescence and tumourigenicity. However, it is acknowledged that gene-specific methylation is variable in cancer and that many tumour suppressor genes show promoter-specific DNA hypermethylation leading to gene silencing, including the genes encoding pRB (Ref. Reference Sakai128) and p16^INK4a (Ref. Reference Gonzalez-Zulueta129), whereas the genome is hypomethylated overall. Interestingly, recent data have shown that p16^INK4a is crucially involved in DNA methylation, leading to gene silencing and breast tumourigenesis (Ref. Reference Reynolds130). Loss of p16^INK4a leads to the increased expression of key PcG proteins, resulting in a specific pattern of DNA hypermethylation and importantly the silencing of the gene encoding HOXA9, a transcription factor that is epigenetically silenced in breast tumourigenesis. Therefore, it is likely that global loss of DNA methylation leading to chromosomal instability could trigger cellular senescence and that promoter-specific alterations in key tumour suppressor genes might allow the bypass of senescence and lead to tumourigenicity.

Senescence and DNA damage detection by histone modification

Histone modifications can also play a role in the detection and repair of DNA damage that, as previously mentioned, can lead to premature senescence. DNA damage can be incurred through the actions of various types of radiation (Ref. Reference Herskind and Rodemann131), DNA-damaging drugs (many of which are used as anti-tumourigenic therapies) (Ref. Reference Robles and Adami132) and oxidative stress (Ref. Reference von Zglinicki133), and histone phosphorylation is now know to play a role in mediating cellular responses to this DNA damage. Signalling kinases, notably ATM (for ‘ataxia telangiectasia mutated’) and ATR (for ‘ataxia telangiectasia and Rad3 related’), are recruited to the site of damage and are activated, leading to phosphorylation of histone H2AX molecules (γ-H2AX) (Ref. Reference Shroff134). Histone H2AX is a variant of the H2A histone that makes up part of the histone octamer. The phosphorylation of histone H2AX facilitates the recruitment of checkpoint and DNA repair factors including p53-binding protein 1 (153BP1), and promotes the activation of checkpoint kinase 1 (Chk1) and Chk2, which can activate p53 and thereby mediate cellular senescence. Furthermore, it is known that the uncapping of telomeres can lead to senescence through a DNA damage response mediated by a similar mechanism (Refs Reference Karlseder, Smogorzewska and de Lange135, Reference d'Adda di Fagagna136). Phosphorylated histone H2AX also serves to recruit histone modifiers and ATP-dependent chromatin remodellers to these sites of DNA damage and aid the repair of these sites (Ref. Reference van Attikum and Gasser137).

Two recent studies have uncovered a link between histone lysine methylation and the DNA damage checkpoint and double-strand break (DSB) repair proteins 53BP1 in mammals and Schizosaccharomyces pombe putative homolog Crb2 (for ‘Crumbs-like protein 2’): the relocalisation of 53BP1 and Crb2 to DNA DSBs has been found to require histone methylation (Refs Reference Huyen138, Reference Sanders139). It was proposed that 53BP1 senses DSBs through changes in higher-order chromatin structure that occur upon exposure of the 53BP1 binding site [methylated Lys79 H3 (K79 H3), which lies in the globular domain of histone H3] (Ref. Reference Huyen138). In S. pombe, a similar mechanism occurs with Crb2 binding to methylated K20 (Ref. Reference Sanders139). However, recent studies have shown that in fact 53BP1 and Crb2 bind to the same modified lysine residue, di-methylated K20 H4 (Ref. Reference Botuyan140).

Telomerase and telomeres in senescence

In adult humans, most normal cells have low or no detectable telomerase activity (Ref. Reference Shay and Bacchetti141). The lack of telomerase activity, or other telomere maintenance mechanisms, leads to the loss of around 50–100 bp of telomeric sequence at each cell division (Refs Reference Harley, Futcher and Greider142, Reference Hastie143). The loss of telomeric tract during cell division presents a limit to the maximum number of divisions a cell can undertake before entering senescence and therefore provides a limit to its lifespan (Ref. Reference Allsopp144). Cellular senescence as a response to telomere attrition is thought to have evolved as a tumour suppressive barrier against excessive clonal expansion. However, the detection of telomerase activity in an estimated 85% of malignant tumours, together with the strong links between telomerase activity and malignant cells (Refs Reference Shay and Bacchetti141, Reference Kim145), indicates the importance of telomerase activity in tumourigenesis and also suggests that telomerase activation is the most common mechanism for telomere maintenance in cancer cells. However, several immortal cell lines and tumours lack telomerase activity, yet have the ability to elongate their telomeres. Such non-telomerase-dependent mechanisms of telomere maintenance include the alternative lengthening of telomeres (ALT) pathway (Ref. Reference Reddel146) (Fig. 1).

In cells where ALT is utilised, telomere length is maintained by a recombination-based DNA replication mechanism (Ref. Reference Dunham147) and leads to heterogeneity in telomere lengths (Ref. Reference Bryan and Reddel148). Also associated with use of the ALT mechanism is the presence of ALT-associated promyelocytic leukemia protein (PML) bodies, called APBs (Ref. Reference Yeager149), which contain the protein PML, telomeric DNA, the telomeric-binding proteins telomere repeat factor 1 (TRF1) and TRF2, and proteins involved in DNA recombination and replication. The utilisation of ALT in tumour cells is much rarer than the utilisation of telomerase activity (Ref. Reference Bryan and Reddel148) and it has been hypothesised that ALT might normally occur under selective pressure when expression of telomerase is prevented. ALT is commonly activated in tumours of mesenchymal origin, including osteosarcomas, soft tissue sarcomas and glioblastoma multiforme (Refs Reference Henson150, Reference Ulaner151, Reference Hakin-Smith152, Reference Johnson153). Furthermore, in normal mesenchymal tissues, telomerase activity is more tightly repressed that in epithelial cells, possibly leading to the selective pressure for ALT activation over telomerase activity in these cell types.

Epigenetic regulation of telomere maintenance mechanisms

Expression of the human telomerase RNA component (hTR) and hTERT are understood to be the minimal requirements for telomerase activity. Studies have shown that specific epigenetic modifications at the hTR and hTERT promoters correlate with expression or repression of hTR and hTERT (Refs Reference Atkinson154, Reference Serakinci155). Lack of expression of hTR in ALT cell lines, and hTERT in ALT and normal cell lines, is associated with a repressive chromatin environment at the appropriate gene promoter. Conversely, expression of hTR in ALT and telomerase-positive cell lines, and hTERT expression in telomerase-positive cell lines, is associated with a more permissive chromatin environment at the appropriate gene promoter. Moreover, in the pre-senescent normal fibroblast cell line WI38, which has no detectable hTERT expression and low hTR expression, the telomerase gene promoters contain no great quantities of either positive or negative chromatin modifications, and therefore the chromatin environment of the promoter remains in a neutral chromatin configuration (Fig. 5a).

Figure 5

Epigenetic regulation of telomere maintenance mechanisms. (a) Histograms showing a simplistic view of what we know about the relationship between the relative levels of repressive and permissive chromatin modifications at the telomerase gene promoters in telomerase-expressing cancer cells, pre-senescent normal cells and cells that maintain their telomeres by the alternative lengthening of telomeres (ALT) mechanism. The presence of permissive histone modifications at the telomerase gene promoters is linked to telomerase activity, whereas the presence of repressive histone modifications is linked to the activation of ALT. Both of these scenarios are in contrast to the chromatin environment of pre-senescent normal cells, in which permissive and repressive modifications are at a minimum. This suggests a model by which epigenetic deregulation of the telomerase gene promoters could lead to aberrant telomere maintenance. (b) Model depicting how chromatin modifications might influence telomere maintenance mechanisms. In a ‘normal’ cell that does not have telomerase activity and has no or low levels of expression of the telomerase genes, few chromatin modifications are present at the gene promoters. The choice of telomere maintenance mechanism may be influenced through epigenetic mechanisms and changes in histone modifications, possibly during tumourigenesis. In the left-hand pathway shown, telomerase gene expression is linked to the accumulation of permissive histone modifications [acetylation of histones H3 and H4 and methylation of Lys4 (K4) H3], which leads to telomerase activity and the elongation/maintenance of telomere length. However, in the right-hand pathway shown, heavy repression of the telomerase gene promoters by the accumulation of repressive histone modifications [methylation of Lys20 (K20) H4 and/or methylation of Lys9 (K9) H3] could lead to utilisation of the ALT mechanism rather than telomerase activity for telomere elongation/maintenance owing to the repression of telomerase gene expression.

Furthermore, upon comparison of the hTERT promoter chromatin environment in ALT cell lines (including a cell transformed from the parental WI38 cell line WI38-SV40) with the normal WI38 cell line, an increase in the amounts of repressive modifications is evident and links the heavy repression of hTERT expression by the chromatin environment to the ALT phenotype. In the telomerase-positive tumour cell lines, permissive modifications are gained at the hTERT promoter relative to the WI38 cell line. Similarly, heavy repression of hTR is evident in some ALT cell lines with no hTR expression, whereas high levels of permissive chromatin modifications are linked to hTR expression. Additionally, treatment of normal and ALT cells with 5-azadC in combination with the HDAC inhibitor trichostatin A (TSA) caused chromatin remodelling of both promoters and led to the formation of a more permissive chromatin environment, as well as reactivation or increased expression of hTR and hTERT expression. This suggests a model by which gain of permissive modifications at the telomerase gene promoters leads to telomerase activity by the expression of hTR and hTERT, whereas the gain of repressive modifications leads to activation of the ALT mechanism (Fig. 5b). Therefore, epigenetic deregulation of the telomerase gene promoters might allow the activation of a telomere maintenance mechanism, allowing bypass of senescence.

Epigenetic deregulation and tumourigenicity in adult stem cells

Further studies into the epigenetic regulation of the telomerase gene promoters utilised human mesenchymal stem cells (hMSCs) (Ref. Reference Serakinci156). These cells do not show telomerase activity but have been shown to be the target for neoplastic transformation after transduction with hTERT (Ref. Reference Serakinci155). However, by supplying hTERT, the molecular events required to upregulate hTERT expression in cancer stem cell development are missed. Therefore, hMSCs are ideal for the identification of molecular mechanisms regulating telomerase gene expression in stem cells.

It has now been shown that repression of hTERT expression in hMSCs is a result of promoter-specific histone hypo-acetylation coupled with low RNA polymerase II (RNAPII) and transcription factor IIB (TFIIB) tracking. This repression can be overcome by treatment of hMSCs with TSA, concomitant with increases in promoter-specific histone acetylation and increases in RNAPII and TFIIB tracking. hTR expression is also increased in TSA-treated hMSCs, concomitant with changes in RNAPII and TFIIB dynamics. These data suggest that deregulation of chromatin modifications at the hTERT and hTR promoter might facilitate tumourigenesis in hMSCs and supports the hypothesis that tumourigenesis can arise in epigenetically deregulated stem cell populations.

Epigenetic regulation of telomere length

It has been shown that epigenetic modifications can also regulate telomere length through the modulation of the chromatin environment of the telomeres themselves. Telomeres are rich in heterochromatin features (Ref. Reference Blasco157) and disruption of this heterochromatin structure has been shown to lead to altered telomere length in mice, established by knockout of the SUV39H1 and SUV39H2 histone methyl transferases in mice (Ref. Reference Garcia-Cao158) and by the knockout of either DNA methyltransferase 1 (DNMT1) or both DNMT-3a and -3b in mouse ESCs (Ref. Reference Gonzalo159). In the latter study, the lack of DNMTs led to increased intra-telomeric recombination and increased the levels of other markers of the ALT telomere maintenance mechanism, further suggesting the role for epigenetics in the choice of telomere maintenance mechanism used in cancer cells. This also shows that epigenetic deregulation can lead to immortalisation events and might facilitate tumourigenesis.