DNA-Methylation Changes in Replicative Senescence and Aging: Two Sides of the Same Coin?

Keywords::

• aging
• DNA methylation
• predictors
• replicative senescence

First draft submitted: 25 September 2015; Accepted for publication: 25 September 2015; Published online: 23 December 2015

Replicative senescence is often considered to be a hallmark of aging [1] – but this assumption needs to be challenged by new insights into associated epigenetic modifications. Primary cells can undergo only a limited number of cell divisions in vitro before they enter the state of replicative senescence, which is reflected by unequivocal cell cycle arrest. It was first described half a century ago by Leonard Hayflick [2] – therefore often referred to as ‘Hayflick limit’ – and since then it has been speculated that replicative senescence is tightly associated with aging of the organism [3]. In fact, cellular aging during in vitro culture reflects various molecular features that seem to be indicative for aging, such as telomere attrition, activation of the p53/p21CIP1 and p16INK4A/pRb signaling pathways, alteration of cell morphology and metabolism, increased senescence-associated β-galactosidase activity, loss of differentiation potential, formation of senescence-associated heterochromatin foci and the senescence-associated secretory phenotype [1]. Cell samples from elderly donors recapitulate many of these parameters, including a higher number of positive staining for senescence-associated β-galactosidase [4]; slower proliferation rate and senescence-like morphological changes already at the initial cell passage [5]; reduced colony-forming unit frequency [6] and concordant gene expression changes [7]. These findings fueled the perception, that replicative senescence in vitro and aging in vivo are governed by the same conserved mechanism – although carried out at a different pace. Cellular changes in the course of culture expansion are therefore often considered as a good in vitro model to unravel the molecular mechanisms that drive the process of aging.

Replicative senescence and aging are both reflected by highly reproducible epigenetic changes – particularly in the DNA methylation (DNAm) pattern of developmental genes [8]. DNAm is nowadays the best-characterized epigenetic modification: it represents a covalent addition of methyl groups to cytosine residues in the context of CG dinucleotides, referred to as ‘CpG site.’ Senescence-associated DNAm changes are significantly enriched in genomic regions with repressive histone marks (H3K9me3 and H3K27me3) and at target sites of Polycomb group proteins [9,10]. Similar findings have also been reported for age-associated DNAm changes [11,12]. In fact, direct correlation of age-associated and senescence-associated DNAm changes in mesenchymal stromal cells (MSCs) revealed a moderate but significant association of the two epigenetic processes [8]. On the other hand, replicative senescence and aging can both be tracked by very specific epigenetic modifications: for example, an ‘epigenetic-senescence-signature’ based on DNAm at six specific CpG sites can predict passage numbers and cumulative population doublings of MSCs and fibroblasts and these predictions are independent from the chronological age of the donor [13,14]. In analogy, epigenetic age predictors facilitate estimation of chronological age even in cultured cells – albeit this has not been systematically analyzed so far and in vitro culture generally evokes DNAm changes. Hannum et al. developed a quantitative model of aging using DNA methylation values of 71 CpG sites in blood with a mean absolute deviation (MAD) of 3.9 years [15]; and Horvath elaborated a multi-tissue age predictor based on 353 CpG sites with a MAD of 3.6 years [16]. We have recently developed an ‘epigenetic-aging-signature’ based on DNAm levels at only three CpG sites. This method facilitates pyrosequencing of bisulfite-converted genomic DNA to provide cost-effective and robust age predictions in blood samples (MAD of about 5 years) [17]. Thus, the underlying molecular mechanisms that drive epigenetic changes in senescence and aging seem to be related – but they are certainly not identical as both processes can be separately tracked by epigenetic means.

It is yet unknown how senescence- and/or age-associated DNAm changes are regulated and if they are at all functionally relevant. A plausible explanation for epigenetic modifications might be that the DNAm profile is not accurately maintained during cell divisions, resulting in a stochastic loss of the methylation pattern over the lifespan of an organism. This process has been referred to as ‘epigenetic drift’ [15,18]. However, there is also evidence that these modifications are directly regulated at certain CpG sites: it has been shown that age-associated as well as senescence-associated DNAm changes can be reversed by reprogramming into induced pluripotent stem cells (iPSCs) [10,16–17]. This observation fits to the general assumption that iPSCs are fully rejuvenated – telomeres are elongated during reprogramming, other age-associated molecular parameters are reset and iPSCs can ultimately give rise to new embryos [19]. Interestingly, iPSCs that are redifferentiated toward MSCs in vitro remain epigenetically rejuvenated, while the senescence-associated DNAm changes are gradually reacquired upon exit from the pluripotent state [20]. If senescence- and age-associated DNAm changes occur in such a highly reproducible manner, particularly at developmental genes, and if these changes can be reset by reprogramming into iPSCs, then the two processes should be somehow controllable. It is conceivable that this process is directly mediated by transcription factors, other DNA-binding proteins or long noncoding RNAs [21]. Furthermore, changes in the histone code or chromatin structure might favor epigenetic drift at specific CpG sites in the genome. In this case we would merely look at the molecular sequel of another process that still needs to be tightly regulated and may hence be functionally relevant.

It has been suggested that the proliferation arrest that is associated with senescence acts as a safeguard for malignant transformation [1]. Other authors demonstrated partial but highly significant overlaps between altered DNAm patterns in cancer and senescence [22]. Thus, the DNA methylome of senescent cells might even promote malignancy, if these cells escape the proliferative barrier [22]. In cancer, the epigenome is generally drastically changed and several studies demonstrated that epigenetic age predictions do not correlate with chronological age anymore [15–16]. We have recently analyzed age-associated DNAm changes in 5621 DNAm profiles of 25 cancer types from The Cancer Genome Atlas [23]. Depending on the cancer type, the epigenetic age was often over- or underestimated, whereas there was hardly any correlation with chronological age in tumor tissue. This might be attributed to the fact that cancer is a clonal disease and the tumor initiating cell only captures one specific epigenetic state of cellular aging. In contrast, conventional epigenetic profiles resemble a cross-section through many subpopulations – each of which may have different stages of cellular aging. Notably, particularly age-associated DNAm patterns at hypermethylated CpG sites seem to be coherently modified in cancer, whereas hypomethylated CpGs may rather resemble less coordinated ‘epigenetic drift’ [23]. Furthermore, epigenetic age predictions were associated with specific mutations, various clinical parameters and even overall survival in many types of cancer. Depending on the tumor type, accelerated or decelerated epigenetic age is associated with better or worse prognosis. It may hence be speculated that chromatin reorganization associated with epigenetic aging supports specific mutations in the elderly. Age-associated DNAm changes would then rather resemble a trigger than a safeguard for malignant transformation [24].

Taken together, replicative senescence and aging seem to be related processes and this is reflected by similar molecular changes. However, particularly in the light of epigenetics there are significant differences and therefore replicative senescence may not be the ideal model system for aging. Either way, both ‘aging’ processes are relevant for quality control of cell preparations or for estimation of biological age, respectively – and DNAm patterns provide suitable biomarkers to track this. There is a growing perception, that senescence and aging are not purely caused by accumulation of cellular defects, but rather orchestrated by a molecular process. A better understanding of how senescence- and age-associated epigenetic modifications are regulated will provide insight into how the two sides of the coin are related.

Financial & competing interests disclosure

This work was supported by the Else Kröner-Fresenius Stiftung (2014_A193) and by the German Ministry of Education and Research (BMBF; OBELICS). RWTH Aachen University Medical School has patents pending for the ‘Epigenetic Senescence Signature’ and the ‘Epigenetic Aging Signature’. W Wagner is one of the founders of Cygenia GmbH that may provide service for these methods (www.cygenia.com).

No writing assistance was utilized in the productionof this manuscript.

Additional information

Funding

This work was supported by the Else Kröner-Fresenius Stiftung (2014_A193) and by the German Ministry of Education and Research (BMBF; OBELICS). RWTH Aachen University Medical School has patents pending for the ‘Epigenetic Senescence Signature’ and the ‘Epigenetic Aging Signature’. W Wagner is one of the founders of Cygenia GmbH that may provide service for these methods (www.cygenia.com). No writing assistance was utilized in the productionof this manuscript.