Insights & Perspectives Ideas & Speculations
Do age-associated DNA methylation changes increase the risk of malignant transformation? Wolfgang Wagner*, Carola I. Weidner and Qiong Lin Aging of the organism is associated with highly reproducible DNA methylation (DNAm) changes, which facilitate estimation of donor age. Cancer is also associated with DNAm changes, which may contribute to disease development. Here, we speculate that age-associated DNAm changes may increase the risk of tumor initiation. Notably, when using epigenetic signatures for age-estimations tumor cells are often predicted to be much older than the chronological age of the patient. We demonstrate that aberrant hypermethylation within the gene DNA methyltransferase 3A (DNMT3A) which may contribute to initiation of acute myeloid leukemia (AML) is particularly observed in AML samples that reveal significantly more age-associated DNAm changes. The functional relevance of age-associated DNAm changes remains to be elucidated, but they occur genome wide, in a highly reproducible manner, and most likely influence chromatin organization and hence may favor acquisition of aberrant DNAm patterns contributing to cancer in the elderly.
.
Keywords: aging; cancer; DNA-methylation; DNMT3A; epigenetic; epimutation; predictor
Introduction Mutations are diverse in different cancer types, and understanding of the mechanisms causing somatic mutations in most cancer classes is remarkably limited DOI 10.1002/bies.201400063 Helmholtz-Institute for Biomedical Engineering, Stem Cell Biology and Cellular Engineering, RWTH University Medical School, Aachen, Germany *Corresponding author: Wolfgang Wagner E-mail: [email protected] Abbreviations: AML, acute myeloid leukemia; DNAm, DNA methylation; DNMT, DNA methyltransferase.
[1]. The common perception is that disturbances of the DNA sequence resemble the tumor-initiating events. However, there is increasing evidence that epigenetic modifications can also play a leading role in disease development of many tumor types, such as acute myeloid leukemia (AML) [2]. It is even conceivable that aberrant DNA methylation (DNAm) changes at specific sites in the genome – so called epimutations – mimic genomic mutations to contribute to malignant transformation [3]. Aging is the biggest risk factor for cancer. Recently, it has been recognized that aging of the organism is associated with highly reproducible epigenetic modifications, particularly in the DNAm
Bioessays 36: 0000–0000, ß 2014 WILEY Periodicals, Inc.
pattern [4]. DNAm plays pivotal roles in development and cellular differentiation [5]: specific cytosine residues in cytosine-guanine dinucleotides (CpG sites) can be transformed to 5-methylcytosine by DNA methyltransferases (DNMTs), and the process is reversed by other enzymes [6]. So far little is known about non-CpG DNAm changes in the context of aging [7]. Notably, agerelated DNAm changes are particularly observed in developmental genes [4], in regions with bivalent histone modifications [8], and in target genes of the polycomb group proteins [9]. The underlying mechanism and the functional relevance of age-related DNAm, which often occurs in non-expressed genes and intergenic regions, remain unclear, but the consistency and chronological succession is remarkable. Interestingly, age-related DNAm changes are almost entirely reversed upon reprogramming into induced pluripotent stem cells (iPSCs) – this may reflect functional rejuvenation by conversion into pluripotent state [10–12]. On the other hand, DNAm changes of many CpGs are not reflected by differential expression of corresponding genes, and may hence not be relevant for the phenotypic changes per se. Several groups have shown that age-related DNAm can be used as a biomarker to track donor age with a mean deviation from the chronological age of less than five years [13– 15]. Such age-predictors are usually based on genome-wide DNAm profiles, whereas we have recently demonstrated that even with three specific CpGs the donor age of a blood sample can be
www.bioessays-journal.com
1
W. Wagner et al.
Insights & Perspectives
Box 1
Ideas & Speculations
Epigenetic age-predictors DNAm changes are rapidly acquired during childhood, whereas they often tend to correlate linearly with age in adults [16]. It is possible to filter for CpGs that reveal almost linear correlation of either hypermethylation or hypomethylation with chronological age. Such individual CpGs can be used as biomarkers [13, 17]. Furthermore, a combinatorial model of age-associated CpGs (e.g. a multivariate linear model or an elastic net model) enables an accurate prediction of donor age [14, 15]. DNAm data used in these studies often derive from Illumina Bead Chip arrays, which simultaneously measure the DNAm level on single base resolution at more than 480,000 CpGs [18]. Such models have proven high precision in independent data sets with a mean deviation between predicted and chronological age of less than five years [11, 15]. In this study, we use an age-predictor by Horvath [11] that utilizes 353 CpGs and our previously designed multivariate model of 99 age-related CpGs that was trained and validated on published DNAm profiles of blood samples [10]. Alternatively, it is possible to use our epigenetic age-predictor for blood samples based on pyrosequencing of only three CpGs – without the requirement of whole genome analysis [10]. Analysis of age-related DNAm facilitates relatively precise age-predictions (the precision is higher than analysis of telomere length), but there is evidence that clinical and lifestyle parameters affect this process [10]. Age-related DNAm may therefore help to identify relevant factors supporting healthy aging. On the other hand, the predictive value of these age-predictions with regard to survival or susceptibility for specific diseases remains to be demonstrated.
accurately estimated [10] (Box 1). In this manuscript we discuss whether epigenetic changes of aging rather reduce proliferation rate – as safeguard of tumor formation – or if they may also favor malignant transformation and thereby contribute to cancer development in the elderly.
Cellular aging – a purposeful process? All primary cells can only reach a limited number of cell divisions during cell culture in vitro. Within 2–3 months they enter a senescent state, which is the socalled “Hayflick limit” [19]. This senescence process is not identical to complex changes in aging of the organism, but it is often used as a surrogate model that is associated with similar molecular changes. Both processes involve telomere attrition, accumulation of genomic damage, oncogenes, and tumor suppressive pathways [20]. In analogy to aging, longterm culture in vitro is also associated with highly reproducible DNAm changes, which are similar but not identical to
2
age-associated DNAm alterations. Such senescence-associated DNAm changes can – similar to the above-mentioned epigenetic age-predictors – be used to predict the number of passages or cumulative population doublings during culture expansion [21, 22]. Furthermore, most of the senescence-associated DNAm changes are also reversed by reprogramming in iPSCs [22]. However, neither age-related, nor senescenceassociated DNAm changes seem to be counteracted in malignant cells, despite immortalization and high proliferation rates of tumor cells [22]. It is still under debate whether aging and/or cellular senescence are tightly regulated processes, or whether they simply resemble the inevitable outcome of accumulation of stochastic damage – if they are controlled then they should be purposeful. The process of senescence – which on cellular level is associated with loss of proliferative potential after a limited number of cell divisions – may be rationalized as stress-induced barrier to protect from cancers [23, 24]. Thereby the organism is shielded from “selfish cells” that evade growth control due to
.....
mutations. In fact, the relevance of cellular senescence has been suggested by suppression of Pten-deficient tumorigenesis in a murine model [25]. Furthermore, human naevi with the oncogenic mutation BRAFV600E display classical hallmarks of senescence, suggesting that oncogene-induced senescence represents a genuine protective physiological process [26]. A limited number of cell divisions are compatible with life-long regeneration of the normal stem cell pool because these cells usually reside quiescently in their niche [27]. A benefit from cellular aging may therefore be protection of the organism from cancer. Alternatively, it is conceivable that the benefit of aging accounts rather for the species in terms of evolution than for the individual organism [28], and that the aging process of a given species is closely linked to different reproductive rates and reproductive timespans [28] – for example mice usually die after two years, whereas humans live for about 75 years. In fact, mathematical models support the idea that aging might be selected for in terms of evolution under the following assumptions: (1) there is competition between individuals (particularly between parents and their progeny residing in locally similar niches), (2) optimal conditions are not stationary, and (3) reproduction helps each species to keep competitive and adopt to new environments [29]. Whether aging is purposeful in terms of evolution is still controversially discussed – in fact, many researchers argue against this possibility [30]. Nevertheless, this idea opens further perspectives for functional implications of aging and of the potentially related process of senescence.
Cancer cells age prematurely at an epigenetic level The methylome of cancer cells changes drastically. This may be due to exposure to stimuli from the tumor microenvironment, genetic aberrations, or differentiation processes during tumorigenesis. Besides hypermethylation at specific CpG sites, a global hypomethylation occurs in both aging and cancer, which
Bioessays 36: 0000–0000, ß 2014 WILEY Periodicals, Inc.
.....
Insights & Perspectives
W. Wagner et al.
could contribute to malignant transformation [31]. Silenced transposons could become reactivated due to hypomethylation, and transposed to other genomic regions where they could integrate into genes, thereby deregulating them. In fact, the risk for carcinogenesis has been shown to be increased in individuals with hypomethylated transposons [32]. Specific DNAm changes have been used as biomarkers for disease development. However, whether such DNAm changes drive the disease has not yet been unequivocally demonstrated. Already two decades ago it has been demonstrated that specific age-associated methylation occurs in genes encoding the estrogen receptor (ER) and insulin-like growth factor 2, which become methylated in colon cancer [33, 34], suggesting that specific age-related methylation changes may account for occurrence of cancer in the elderly. More recently, Hannum et al. [15] used their agepredictor on DNAm data from The Cancer Genome Atlas (TCGA) portal of matched samples of cancer tissue and normal tissue. Their model indicated that tumors appeared to have aged 40% more than matched normal tissue from the same individual, and this accelerated tumor aging was apparent regardless of the primary tissue type. Maegawa et al. [35] demonstrated such accelerated epigenet-
ic aging phenotype in the myelodysplastic syndrome of specific mouse models. Age-associated CpGs were also observed to be over-methylated in a wide variety of tumor types in a recent work by Xu and coworkers [36], suggesting that cells may have a lower threshold for malignant transformation due to the acquisition of methylation at age-related sites. Similar results were also described by Horvath [11], although the statement that cancer is associated with an increased DNAm age in most cancer types was later corrected in an erratum. Notably, “age markers” are not necessarily transcribed genes with any known consequence to cancer, which may also be due to the fact that interplay of DNAm with gene expression is complex [11]. For visualization of these findings we have, as representative examples, used our signature of 99 age-related CpGs (99-CpGs model) [10] and the agepredictor from Horvath (Horvath’s model) [11] in 194 publically available DNAm profiles of patients with AML from TCGA [37]. On average, AML samples were predicted 21 years older (99-CpGs model; Fig. 1A) and 10 years older (Horvath’s model; Fig. 1B) than the chronological age of patients, respectively. It might be anticipated that the global DNAm changes in leukemia development result in non-specific
Bioessays 36: 0000–0000, ß 2014 WILEY Periodicals, Inc.
skewing of age-predictions. However, the deviation of age-predictions is similar in individual AML samples when comparing the results of the two different age-predictors (Fig. 1C): there is a clear correlation in premature aging of AML samples when comparing the 99-CpG model and Horvath’s model (R ¼ 0.7) even though it is much less pronounced than in normal blood (R ¼ 0.93). This finding supports the notion that age-associated DNAm changes are rather enhanced in AML. It is conceivable that accelerated epigenetic aging in these cells is due to increased proliferative history associated with more epigenetic drift [38]. Alternatively, tumor-initiating cells might be derived from cells that have already acquired many age-related DNAm changes prior to malignant transformation. Either way, aberrant DNAm in AML includes significant changes of ageassociated DNAm [39] – whether or not this is the case for all types of cancer deserves further analysis.
Epimutations may be more likely in the aging DNA methylome The aberrant epigenome in cancer may simply mirror the phenotypic changes during disease progression that are caused by genetic defects. There is, however, also ample evidence that disease-specific DNAm occurs independently of genomic aberrations and that it may contribute to disease progression, too [39, 40]. Aberrant DNAm changes at specific sites in the genome – so-called epimutations – may
3
Ideas & Speculations
Figure 1. Premature aging of the DNA-methylome in acute myeloid leukemia. This figure compares different epigenetic age-predictors on previously published datasets. A, B: DNAm profiles of 194 AML patients from TCGA [37] were analyzed using two age-predictors: our previously published 99-CpGs model (A) and Horvath’s model (B) [11]. AML samples (red dots) were predicted to be aged prematurely. Age-predictions for normal blood samples [15] are depicted as gray dots for comparison. C: Comparison of the two age-predictions using the two different models revealed clear correlation of premature epigenetic aging in normal blood (R ¼ 0.93) but also in AML samples (R ¼ 0.7).
Ideas & Speculations
W. Wagner et al.
Insights & Perspectives
Figure 2. Epimutations in DNMT3A are more frequent in patients with prematurely aged DNAmethylome. Acute myeloid leukemia is frequently associated with genomic mutations (mut) in DNMT3A [41], or with aberrant hypermethylation at an internal promoter-region of DNMT3A (epimutation; epi) [3]. The previously published AML datasets from TCGA [37] were categorized according to these DNMT3A modifications as described before [3] (numbers of samples per category: mut()epi() ¼ 84; mut()epi(þ) ¼ 61; mut(þ)epi() ¼ 38; mut(þ)epi(þ) ¼ 11) and further analyzed with our previously published age-predictor [10]. A: Epimutations in DNMT3A (red) were not significantly more frequent in AML of elderly patients. B: However, AML samples with epimutations are predicted to be of significantly higher age than samples without DNMT3A methylation modifications (99-CpGs model depicted here). C: Deviation of predicted and chronological age increases in AML samples with DNMT3A epimutation. D: The percentage of AML samples with epimutations was subsequently analyzed in samples that were estimated to be younger than chronological age (100 years; samples with genomic mutations in DNMT3A were excluded from this analysis). P values were calculated by Wilcoxon’s rank-sum test (A–C) and hypergeometric distribution (D).
even replace genomic aberrations and entail malignant transformation [3]. Horvath [11] has suggested that the accelerated epigenetic age of cancer tissue might be inversely correlated with the number of mutations. This may indicate that aberrations resulting in malignant transformation are either acquired on a genetic level or on an epigenetic level in association with age-related DNAm changes. If epigenetic drift favors epimutations then this might be the reason why some tumors, such as AML, occur particularly in the elderly. We have recently demonstrated that aberrant hypermethylation within the gene DNMT3A is observed in about 40% of AML patients and that it is associated with shorter overall survival [3]. This epimutation is rather mutually exclusive with genomic mutations in DNMT3A, which occur in about 20% of AML patients [41]. Notably, epimutations as well as genomic mutations entail variant transcripts of DNMT3A, which may contribute to the related patterns in transcriptome and DNAm profiles [42]. We have revisited age-predictions of AML
4
.....
Conclusions and prospects
they are regulated at specific sites in the genome and whether they are functionally relevant remains largely unknown. On the other hand, aging is clearly associated with certain genome-wide DNAm changes that are highly reproducible and can even be used for reliable age-predictions. It will be important to gain better insight into the underlying mechanisms: non-coding RNAs and epigenetic modifiers may play a role in this process. The fact that age-related DNAm changes can be reset by reprogramming somatic cells into iPSCs [10, 11] suggests some kind of reversible trigger. It is conceivable that the global epigenetic changes upon aging favor dysregulations contributing to disease development. This was particularly well depicted by the epimutation of DNMT3A, which is significantly enriched in AML-patients that are prematurely aged on DNAm level. However, this is only one example for possible interplay of age-related DNAm and epimutations. Further research is required to decipher how age-related changes are affected in malignant cells, and if they coincide with epigenetic modifications involved in malignant transformation. In this regard, ageassociated alterations of the DNAm pattern resemble a double-edged sword: they may provide an anti-proliferative barrier for aging cells to prevent cancer initiation, but they may also favor aberrant DNAm resulting in malignant transformation.
DNA methylation is now recognized as a highly dynamic system, which can be modulated during development and disease. Such changes may be a sequel of initial signaling events – but how
Acknowledgments This work was supported by the German Research Foundation (WA/1706/2-1), by ¨ner-Fresenius Stiftung, by a the Else Kro
samples with regard to epimutations/ mutations in DNMT3A. In fact, AML samples with epimutations in DNMT3A (mut() epi(þ) DNMT3A) revealed accelerated aging and significantly higher deviation of age-predictions (Fig. 2A–C). Concordantly, the percentage of AML patients with DNMT3A-epimutation increased significantly in patients with premature epigenetic aging (Fig. 2D). If the epimutation in DNMT3A is relevant for the development of AML then this association might be one reason why AML is more frequent in elderly people: age-related changes in DNAm profiles may favor epimutations that contribute to formation of age-associated tumors.
Bioessays 36: 0000–0000, ß 2014 WILEY Periodicals, Inc.
.....
Insights & Perspectives
RWTH Aachen University Medical School has applied for patent applications for the Epigenetic-Aging-Signature and the epimutation in DNMT3A. Wolfgang Wagner is involved in the company Cygenia, which provides services for these methods to other researchers (www.cygenia.com). Apart from this, the authors have nothing to disclose.
References 1. Alexandrov LB, Nik-Zainal S, Wedge DC, Aparicio SA, et al. 2013. Signatures of mutational processes in human cancer. Nature 500: 415–21. 2. Jeong M, Sun D, Luo M, Huang Y, et al. 2014. Large conserved domains of low DNA methylation maintained by DNMT3a. Nat Genet 46: 17–23. 3. Jost E, Lin Q, Ingrid WC, Wilop S, et al. 2013. Epimutations mimic genomic mutations of DNMT3A in acute myeloid leukemia. Leukemia 28: 1227–34. 4. Bork S, Pfister S, Witt H, Horn P, et al. 2010. DNA methylation pattern changes upon longterm culture and aging of human mesenchymal stromal cells. Aging Cell 9: 54–63. 5. Smith ZD, Meissner A. 2013. DNA methylation: roles in mammalian development. Nat Rev Genet 14: 204–20. 6. Wu H, Zhang Y. 2011. Mechanisms and functions of Tet protein-mediated 5-methylcytosine oxidation. Genes Dev 25: 2436–52. 7. Lister R, Mukamel EA, Nery JR, Urich M, et al. 2013. Global epigenomic reconfiguration during mammalian brain development. Science 341: 1237905. 8. Rakyan VK, Down TA, Maslau S, Andrew T, et al. 2010. Human aging-associated DNA hypermethylation occurs preferentially at bivalent chromatin domains. Genome Res 20: 434–9. 9. Teschendorff AE, Menon U, Gentry-Maharaj A, Ramus SJ, et al. 2010. Age-dependent DNA methylation of genes that are suppressed in stem cells is a hallmark of cancer. Genome Res 20: 440–6. 10. Weidner CI, Lin Q, Koch CM, Eisele L, et al. 2014. Aging of blood can be tracked by DNA methylation changes at just three CpG sites. Genome Biol 15: R24.
11. Horvath S. 2013. DNA methylation age of human tissues and cell types. Genome Biol 14: R115. 12. Frobel J, Hemeda H, Lenz M, Abagnale G, et al. 2014. Epigenetic rejuvenation of mesenchymal stromal cells derived from induced pluripotent stem cells. Stem Cell Rep 3: 414– 22. 13. Bocklandt S, Lin W, Sehl ME, Sanchez FJ, et al. 2011. Epigenetic predictor of age. PLoS One 6: e14821. 14. Koch CM, Wagner W. 2011. Epigeneticaging-signature to determine age in different tissues. Aging (Albany NY) 3: 1018–27. 15. Hannum G, Guinney J, Zhao L, Zhang L, et al. 2013. Genome-wide methylation profiles reveal quantitative views of human aging rates. Mol Cell 49: 359–67. 16. Alisch RS, Barwick BG, Chopra P, Myrick LK, et al. 2012. Age-associated DNA methylation in pediatric populations. Genome Res 22: 623–32. 17. Garagnani P, Bacalini MG, Pirazzini C, Gori D, et al. 2012. Methylation of ELOVL2 gene as a new epigenetic marker of age. Aging Cell 11: 1132–4. 18. Bibikova M, Barnes B, Tsan C, Ho V, et al. 2011. High density DNA methylation array with single CpG site resolution. Genomics 98: 288–95. 19. Hayflick L. 1965. The limited in vitro lifetime of human diploid cell strains. Exp Cell Res 37: 614–36. 20. Campisi J. 2013. Aging, cellular senescence, and cancer. Annu Rev Physiol 75: 685–705. 21. Koch CM, Joussen S, Schellenberg A, Lin Q, et al. 2012. Monitoring of cellular senescence by DNA-methylation at specific CpG sites. Aging Cell 11: 366–9. 22. Koch CM, Reck K, Shao K, Lin Q, et al. 2013. Pluripotent stem cells escape from senescence-associated DNA methylation changes. Genome Res 23: 248–59. 23. Finkel T, Serrano M, Blasco MA. 2007. The common biology of cancer and ageing. Nature 448: 767–74. 24. Cruickshanks HA, McBryan T, Nelson DM, Vanderkraats ND, et al. 2013. Senescent cells harbour features of the cancer epigenome. Nat Cell Biol 15: 1495–506. 25. Chen Z, Trotman LC, Shaffer D, Lin HK, et al. 2005. Crucial role of p53-dependent cellular senescence in suppression of Ptendeficient tumorigenesis. Nature 436: 725–30. 26. Michaloglou C, Vredeveld LC, Soengas MS, Denoyelle C, et al. 2005. BRAFE600associated senescence-like cell cycle arrest of human naevi. Nature 436: 720–4. 27. Marciniak-Czochra A, Stiehl T, Wagner W. 2009. Modeling of replicative senescence in hematopoietic development. Aging (Albany NY) 1: 723–32.
Bioessays 36: 0000–0000, ß 2014 WILEY Periodicals, Inc.
28. Moore A. 2012. TOR: the ancient link between cancer and ageing? Could ageing and cancer be unified by TOR via the ‘compromise’ struck between soma and germline at the dawn of multicellular life? BioEssays 34: 443–4. 29. Martins AC. 2011. Change and aging senescence as an adaptation. PLoS One 6: e24328. 30. Kirkwood TB. 2008. Understanding ageing from an evolutionary perspective. J Intern Med 263: 117–27. 31. Zane L, Sharma V, Misteli T. 2014. Common features of chromatin in aging and cancer: cause or coincidence? Trends Cell Biol, DOI: 10.1016/j.tcb.2014.07.001 32. Zhu ZZ, Sparrow D, Hou L, Tarantini L, et al. 2011. Repetitive element hypomethylation in blood leukocyte DNA and cancer incidence, prevalence, and mortality in elderly individuals: the Normative Aging Study. Cancer Causes Control 22: 437–47. 33. Issa JP, Ottaviano YL, Celano P, Hamilton SR, et al. 1994. Methylation of the oestrogen receptor CpG island links ageing and neoplasia in human colon. Nat Genet 7: 536–40. 34. Issa JP, Vertino PM, Boehm CD, Newsham IF, et al. 1996. Switch from monoallelic to biallelic human IGF2 promoter methylation during aging and carcinogenesis. Proc Natl Acad Sci USA 93: 11757–62. 35. Maegawa S, Gough SM, Watanabe-Okochi N, Lu Y, et al. 2014. Age-related epigenetic drift in the pathogenesis of MDS and AML. Genome Res 24: 580–91. 36. Xu Z, Taylor JA. 2014. Genome-wide agerelated DNA methylation changes in blood and other tissues relate to histone modification, expression and cancer. Carcinogenesis 35: 356–64. 37. Cancer Genome Atlas Research Network. 2013. Genomic and epigenomic landscapes of adult de novo acute myeloid leukemia. N Engl J Med 368: 2059–74. 38. Teschendorff AE, West J, Beck S. 2013. Age-associated epigenetic drift: implications, and a case of epigenetic thrift? Hum Mol Genet 22: 7–15. 39. Schoofs T, Rohde C, Hebestreit K, Klein HU, et al. 2013. DNA methylation changes are a late event in acute promyelocytic leukemia and coincide with loss of transcription factor binding. Blood 121: 178–87. 40. Schoofs T, Berdel WE, Muller-Tidow C. 2014. Origins of aberrant DNA methylation in acute myeloid leukemia. Leukemia 28: 1–14. 41. Ley TJ, Ding L, Walter MJ, McLellan MD, et al. 2010. DNMT3A mutations in acute myeloid leukemia. N Engl J Med 363: 2424–33. 42. Weidner CI, Walenda T, Lin Q, Wo¨lfler MM, et al. 2013. Hematopoietic stem and progenitor cells acquire distinct DNA-hypermethylation during in vitro culture. Sci Rep 3: 3372.
5
Ideas & Speculations
generous donation of Vision4 LifeSiences, and by the Interdisciplinary Center for Clinical Research (IZKF) within the faculty of Medicine at the RWTH Aachen University.
W. Wagner et al.
Do age-associated DNA methylation changes increase the risk of malignant transformation? Wolfgang Wagner*, Carola I. Weidner and Qiong Lin Aging of the organism is associated with highly reproducible DNA methylation (DNAm) changes, which facilitate estimation of donor age. Cancer is also associated with DNAm changes, which may contribute to disease development. Here, we speculate that age-associated DNAm changes may increase the risk of tumor initiation. Notably, when using epigenetic signatures for age-estimations tumor cells are often predicted to be much older than the chronological age of the patient. We demonstrate that aberrant hypermethylation within the gene DNA methyltransferase 3A (DNMT3A) which may contribute to initiation of acute myeloid leukemia (AML) is particularly observed in AML samples that reveal significantly more age-associated DNAm changes. The functional relevance of age-associated DNAm changes remains to be elucidated, but they occur genome wide, in a highly reproducible manner, and most likely influence chromatin organization and hence may favor acquisition of aberrant DNAm patterns contributing to cancer in the elderly.
.
Keywords: aging; cancer; DNA-methylation; DNMT3A; epigenetic; epimutation; predictor
Introduction Mutations are diverse in different cancer types, and understanding of the mechanisms causing somatic mutations in most cancer classes is remarkably limited DOI 10.1002/bies.201400063 Helmholtz-Institute for Biomedical Engineering, Stem Cell Biology and Cellular Engineering, RWTH University Medical School, Aachen, Germany *Corresponding author: Wolfgang Wagner E-mail: [email protected] Abbreviations: AML, acute myeloid leukemia; DNAm, DNA methylation; DNMT, DNA methyltransferase.
[1]. The common perception is that disturbances of the DNA sequence resemble the tumor-initiating events. However, there is increasing evidence that epigenetic modifications can also play a leading role in disease development of many tumor types, such as acute myeloid leukemia (AML) [2]. It is even conceivable that aberrant DNA methylation (DNAm) changes at specific sites in the genome – so called epimutations – mimic genomic mutations to contribute to malignant transformation [3]. Aging is the biggest risk factor for cancer. Recently, it has been recognized that aging of the organism is associated with highly reproducible epigenetic modifications, particularly in the DNAm
Bioessays 36: 0000–0000, ß 2014 WILEY Periodicals, Inc.
pattern [4]. DNAm plays pivotal roles in development and cellular differentiation [5]: specific cytosine residues in cytosine-guanine dinucleotides (CpG sites) can be transformed to 5-methylcytosine by DNA methyltransferases (DNMTs), and the process is reversed by other enzymes [6]. So far little is known about non-CpG DNAm changes in the context of aging [7]. Notably, agerelated DNAm changes are particularly observed in developmental genes [4], in regions with bivalent histone modifications [8], and in target genes of the polycomb group proteins [9]. The underlying mechanism and the functional relevance of age-related DNAm, which often occurs in non-expressed genes and intergenic regions, remain unclear, but the consistency and chronological succession is remarkable. Interestingly, age-related DNAm changes are almost entirely reversed upon reprogramming into induced pluripotent stem cells (iPSCs) – this may reflect functional rejuvenation by conversion into pluripotent state [10–12]. On the other hand, DNAm changes of many CpGs are not reflected by differential expression of corresponding genes, and may hence not be relevant for the phenotypic changes per se. Several groups have shown that age-related DNAm can be used as a biomarker to track donor age with a mean deviation from the chronological age of less than five years [13– 15]. Such age-predictors are usually based on genome-wide DNAm profiles, whereas we have recently demonstrated that even with three specific CpGs the donor age of a blood sample can be
www.bioessays-journal.com
1
W. Wagner et al.
Insights & Perspectives
Box 1
Ideas & Speculations
Epigenetic age-predictors DNAm changes are rapidly acquired during childhood, whereas they often tend to correlate linearly with age in adults [16]. It is possible to filter for CpGs that reveal almost linear correlation of either hypermethylation or hypomethylation with chronological age. Such individual CpGs can be used as biomarkers [13, 17]. Furthermore, a combinatorial model of age-associated CpGs (e.g. a multivariate linear model or an elastic net model) enables an accurate prediction of donor age [14, 15]. DNAm data used in these studies often derive from Illumina Bead Chip arrays, which simultaneously measure the DNAm level on single base resolution at more than 480,000 CpGs [18]. Such models have proven high precision in independent data sets with a mean deviation between predicted and chronological age of less than five years [11, 15]. In this study, we use an age-predictor by Horvath [11] that utilizes 353 CpGs and our previously designed multivariate model of 99 age-related CpGs that was trained and validated on published DNAm profiles of blood samples [10]. Alternatively, it is possible to use our epigenetic age-predictor for blood samples based on pyrosequencing of only three CpGs – without the requirement of whole genome analysis [10]. Analysis of age-related DNAm facilitates relatively precise age-predictions (the precision is higher than analysis of telomere length), but there is evidence that clinical and lifestyle parameters affect this process [10]. Age-related DNAm may therefore help to identify relevant factors supporting healthy aging. On the other hand, the predictive value of these age-predictions with regard to survival or susceptibility for specific diseases remains to be demonstrated.
accurately estimated [10] (Box 1). In this manuscript we discuss whether epigenetic changes of aging rather reduce proliferation rate – as safeguard of tumor formation – or if they may also favor malignant transformation and thereby contribute to cancer development in the elderly.
Cellular aging – a purposeful process? All primary cells can only reach a limited number of cell divisions during cell culture in vitro. Within 2–3 months they enter a senescent state, which is the socalled “Hayflick limit” [19]. This senescence process is not identical to complex changes in aging of the organism, but it is often used as a surrogate model that is associated with similar molecular changes. Both processes involve telomere attrition, accumulation of genomic damage, oncogenes, and tumor suppressive pathways [20]. In analogy to aging, longterm culture in vitro is also associated with highly reproducible DNAm changes, which are similar but not identical to
2
age-associated DNAm alterations. Such senescence-associated DNAm changes can – similar to the above-mentioned epigenetic age-predictors – be used to predict the number of passages or cumulative population doublings during culture expansion [21, 22]. Furthermore, most of the senescence-associated DNAm changes are also reversed by reprogramming in iPSCs [22]. However, neither age-related, nor senescenceassociated DNAm changes seem to be counteracted in malignant cells, despite immortalization and high proliferation rates of tumor cells [22]. It is still under debate whether aging and/or cellular senescence are tightly regulated processes, or whether they simply resemble the inevitable outcome of accumulation of stochastic damage – if they are controlled then they should be purposeful. The process of senescence – which on cellular level is associated with loss of proliferative potential after a limited number of cell divisions – may be rationalized as stress-induced barrier to protect from cancers [23, 24]. Thereby the organism is shielded from “selfish cells” that evade growth control due to
.....
mutations. In fact, the relevance of cellular senescence has been suggested by suppression of Pten-deficient tumorigenesis in a murine model [25]. Furthermore, human naevi with the oncogenic mutation BRAFV600E display classical hallmarks of senescence, suggesting that oncogene-induced senescence represents a genuine protective physiological process [26]. A limited number of cell divisions are compatible with life-long regeneration of the normal stem cell pool because these cells usually reside quiescently in their niche [27]. A benefit from cellular aging may therefore be protection of the organism from cancer. Alternatively, it is conceivable that the benefit of aging accounts rather for the species in terms of evolution than for the individual organism [28], and that the aging process of a given species is closely linked to different reproductive rates and reproductive timespans [28] – for example mice usually die after two years, whereas humans live for about 75 years. In fact, mathematical models support the idea that aging might be selected for in terms of evolution under the following assumptions: (1) there is competition between individuals (particularly between parents and their progeny residing in locally similar niches), (2) optimal conditions are not stationary, and (3) reproduction helps each species to keep competitive and adopt to new environments [29]. Whether aging is purposeful in terms of evolution is still controversially discussed – in fact, many researchers argue against this possibility [30]. Nevertheless, this idea opens further perspectives for functional implications of aging and of the potentially related process of senescence.
Cancer cells age prematurely at an epigenetic level The methylome of cancer cells changes drastically. This may be due to exposure to stimuli from the tumor microenvironment, genetic aberrations, or differentiation processes during tumorigenesis. Besides hypermethylation at specific CpG sites, a global hypomethylation occurs in both aging and cancer, which
Bioessays 36: 0000–0000, ß 2014 WILEY Periodicals, Inc.
.....
Insights & Perspectives
W. Wagner et al.
could contribute to malignant transformation [31]. Silenced transposons could become reactivated due to hypomethylation, and transposed to other genomic regions where they could integrate into genes, thereby deregulating them. In fact, the risk for carcinogenesis has been shown to be increased in individuals with hypomethylated transposons [32]. Specific DNAm changes have been used as biomarkers for disease development. However, whether such DNAm changes drive the disease has not yet been unequivocally demonstrated. Already two decades ago it has been demonstrated that specific age-associated methylation occurs in genes encoding the estrogen receptor (ER) and insulin-like growth factor 2, which become methylated in colon cancer [33, 34], suggesting that specific age-related methylation changes may account for occurrence of cancer in the elderly. More recently, Hannum et al. [15] used their agepredictor on DNAm data from The Cancer Genome Atlas (TCGA) portal of matched samples of cancer tissue and normal tissue. Their model indicated that tumors appeared to have aged 40% more than matched normal tissue from the same individual, and this accelerated tumor aging was apparent regardless of the primary tissue type. Maegawa et al. [35] demonstrated such accelerated epigenet-
ic aging phenotype in the myelodysplastic syndrome of specific mouse models. Age-associated CpGs were also observed to be over-methylated in a wide variety of tumor types in a recent work by Xu and coworkers [36], suggesting that cells may have a lower threshold for malignant transformation due to the acquisition of methylation at age-related sites. Similar results were also described by Horvath [11], although the statement that cancer is associated with an increased DNAm age in most cancer types was later corrected in an erratum. Notably, “age markers” are not necessarily transcribed genes with any known consequence to cancer, which may also be due to the fact that interplay of DNAm with gene expression is complex [11]. For visualization of these findings we have, as representative examples, used our signature of 99 age-related CpGs (99-CpGs model) [10] and the agepredictor from Horvath (Horvath’s model) [11] in 194 publically available DNAm profiles of patients with AML from TCGA [37]. On average, AML samples were predicted 21 years older (99-CpGs model; Fig. 1A) and 10 years older (Horvath’s model; Fig. 1B) than the chronological age of patients, respectively. It might be anticipated that the global DNAm changes in leukemia development result in non-specific
Bioessays 36: 0000–0000, ß 2014 WILEY Periodicals, Inc.
skewing of age-predictions. However, the deviation of age-predictions is similar in individual AML samples when comparing the results of the two different age-predictors (Fig. 1C): there is a clear correlation in premature aging of AML samples when comparing the 99-CpG model and Horvath’s model (R ¼ 0.7) even though it is much less pronounced than in normal blood (R ¼ 0.93). This finding supports the notion that age-associated DNAm changes are rather enhanced in AML. It is conceivable that accelerated epigenetic aging in these cells is due to increased proliferative history associated with more epigenetic drift [38]. Alternatively, tumor-initiating cells might be derived from cells that have already acquired many age-related DNAm changes prior to malignant transformation. Either way, aberrant DNAm in AML includes significant changes of ageassociated DNAm [39] – whether or not this is the case for all types of cancer deserves further analysis.
Epimutations may be more likely in the aging DNA methylome The aberrant epigenome in cancer may simply mirror the phenotypic changes during disease progression that are caused by genetic defects. There is, however, also ample evidence that disease-specific DNAm occurs independently of genomic aberrations and that it may contribute to disease progression, too [39, 40]. Aberrant DNAm changes at specific sites in the genome – so-called epimutations – may
3
Ideas & Speculations
Figure 1. Premature aging of the DNA-methylome in acute myeloid leukemia. This figure compares different epigenetic age-predictors on previously published datasets. A, B: DNAm profiles of 194 AML patients from TCGA [37] were analyzed using two age-predictors: our previously published 99-CpGs model (A) and Horvath’s model (B) [11]. AML samples (red dots) were predicted to be aged prematurely. Age-predictions for normal blood samples [15] are depicted as gray dots for comparison. C: Comparison of the two age-predictions using the two different models revealed clear correlation of premature epigenetic aging in normal blood (R ¼ 0.93) but also in AML samples (R ¼ 0.7).
Ideas & Speculations
W. Wagner et al.
Insights & Perspectives
Figure 2. Epimutations in DNMT3A are more frequent in patients with prematurely aged DNAmethylome. Acute myeloid leukemia is frequently associated with genomic mutations (mut) in DNMT3A [41], or with aberrant hypermethylation at an internal promoter-region of DNMT3A (epimutation; epi) [3]. The previously published AML datasets from TCGA [37] were categorized according to these DNMT3A modifications as described before [3] (numbers of samples per category: mut()epi() ¼ 84; mut()epi(þ) ¼ 61; mut(þ)epi() ¼ 38; mut(þ)epi(þ) ¼ 11) and further analyzed with our previously published age-predictor [10]. A: Epimutations in DNMT3A (red) were not significantly more frequent in AML of elderly patients. B: However, AML samples with epimutations are predicted to be of significantly higher age than samples without DNMT3A methylation modifications (99-CpGs model depicted here). C: Deviation of predicted and chronological age increases in AML samples with DNMT3A epimutation. D: The percentage of AML samples with epimutations was subsequently analyzed in samples that were estimated to be younger than chronological age (100 years; samples with genomic mutations in DNMT3A were excluded from this analysis). P values were calculated by Wilcoxon’s rank-sum test (A–C) and hypergeometric distribution (D).
even replace genomic aberrations and entail malignant transformation [3]. Horvath [11] has suggested that the accelerated epigenetic age of cancer tissue might be inversely correlated with the number of mutations. This may indicate that aberrations resulting in malignant transformation are either acquired on a genetic level or on an epigenetic level in association with age-related DNAm changes. If epigenetic drift favors epimutations then this might be the reason why some tumors, such as AML, occur particularly in the elderly. We have recently demonstrated that aberrant hypermethylation within the gene DNMT3A is observed in about 40% of AML patients and that it is associated with shorter overall survival [3]. This epimutation is rather mutually exclusive with genomic mutations in DNMT3A, which occur in about 20% of AML patients [41]. Notably, epimutations as well as genomic mutations entail variant transcripts of DNMT3A, which may contribute to the related patterns in transcriptome and DNAm profiles [42]. We have revisited age-predictions of AML
4
.....
Conclusions and prospects
they are regulated at specific sites in the genome and whether they are functionally relevant remains largely unknown. On the other hand, aging is clearly associated with certain genome-wide DNAm changes that are highly reproducible and can even be used for reliable age-predictions. It will be important to gain better insight into the underlying mechanisms: non-coding RNAs and epigenetic modifiers may play a role in this process. The fact that age-related DNAm changes can be reset by reprogramming somatic cells into iPSCs [10, 11] suggests some kind of reversible trigger. It is conceivable that the global epigenetic changes upon aging favor dysregulations contributing to disease development. This was particularly well depicted by the epimutation of DNMT3A, which is significantly enriched in AML-patients that are prematurely aged on DNAm level. However, this is only one example for possible interplay of age-related DNAm and epimutations. Further research is required to decipher how age-related changes are affected in malignant cells, and if they coincide with epigenetic modifications involved in malignant transformation. In this regard, ageassociated alterations of the DNAm pattern resemble a double-edged sword: they may provide an anti-proliferative barrier for aging cells to prevent cancer initiation, but they may also favor aberrant DNAm resulting in malignant transformation.
DNA methylation is now recognized as a highly dynamic system, which can be modulated during development and disease. Such changes may be a sequel of initial signaling events – but how
Acknowledgments This work was supported by the German Research Foundation (WA/1706/2-1), by ¨ner-Fresenius Stiftung, by a the Else Kro
samples with regard to epimutations/ mutations in DNMT3A. In fact, AML samples with epimutations in DNMT3A (mut() epi(þ) DNMT3A) revealed accelerated aging and significantly higher deviation of age-predictions (Fig. 2A–C). Concordantly, the percentage of AML patients with DNMT3A-epimutation increased significantly in patients with premature epigenetic aging (Fig. 2D). If the epimutation in DNMT3A is relevant for the development of AML then this association might be one reason why AML is more frequent in elderly people: age-related changes in DNAm profiles may favor epimutations that contribute to formation of age-associated tumors.
Bioessays 36: 0000–0000, ß 2014 WILEY Periodicals, Inc.
.....
Insights & Perspectives
RWTH Aachen University Medical School has applied for patent applications for the Epigenetic-Aging-Signature and the epimutation in DNMT3A. Wolfgang Wagner is involved in the company Cygenia, which provides services for these methods to other researchers (www.cygenia.com). Apart from this, the authors have nothing to disclose.
References 1. Alexandrov LB, Nik-Zainal S, Wedge DC, Aparicio SA, et al. 2013. Signatures of mutational processes in human cancer. Nature 500: 415–21. 2. Jeong M, Sun D, Luo M, Huang Y, et al. 2014. Large conserved domains of low DNA methylation maintained by DNMT3a. Nat Genet 46: 17–23. 3. Jost E, Lin Q, Ingrid WC, Wilop S, et al. 2013. Epimutations mimic genomic mutations of DNMT3A in acute myeloid leukemia. Leukemia 28: 1227–34. 4. Bork S, Pfister S, Witt H, Horn P, et al. 2010. DNA methylation pattern changes upon longterm culture and aging of human mesenchymal stromal cells. Aging Cell 9: 54–63. 5. Smith ZD, Meissner A. 2013. DNA methylation: roles in mammalian development. Nat Rev Genet 14: 204–20. 6. Wu H, Zhang Y. 2011. Mechanisms and functions of Tet protein-mediated 5-methylcytosine oxidation. Genes Dev 25: 2436–52. 7. Lister R, Mukamel EA, Nery JR, Urich M, et al. 2013. Global epigenomic reconfiguration during mammalian brain development. Science 341: 1237905. 8. Rakyan VK, Down TA, Maslau S, Andrew T, et al. 2010. Human aging-associated DNA hypermethylation occurs preferentially at bivalent chromatin domains. Genome Res 20: 434–9. 9. Teschendorff AE, Menon U, Gentry-Maharaj A, Ramus SJ, et al. 2010. Age-dependent DNA methylation of genes that are suppressed in stem cells is a hallmark of cancer. Genome Res 20: 440–6. 10. Weidner CI, Lin Q, Koch CM, Eisele L, et al. 2014. Aging of blood can be tracked by DNA methylation changes at just three CpG sites. Genome Biol 15: R24.
11. Horvath S. 2013. DNA methylation age of human tissues and cell types. Genome Biol 14: R115. 12. Frobel J, Hemeda H, Lenz M, Abagnale G, et al. 2014. Epigenetic rejuvenation of mesenchymal stromal cells derived from induced pluripotent stem cells. Stem Cell Rep 3: 414– 22. 13. Bocklandt S, Lin W, Sehl ME, Sanchez FJ, et al. 2011. Epigenetic predictor of age. PLoS One 6: e14821. 14. Koch CM, Wagner W. 2011. Epigeneticaging-signature to determine age in different tissues. Aging (Albany NY) 3: 1018–27. 15. Hannum G, Guinney J, Zhao L, Zhang L, et al. 2013. Genome-wide methylation profiles reveal quantitative views of human aging rates. Mol Cell 49: 359–67. 16. Alisch RS, Barwick BG, Chopra P, Myrick LK, et al. 2012. Age-associated DNA methylation in pediatric populations. Genome Res 22: 623–32. 17. Garagnani P, Bacalini MG, Pirazzini C, Gori D, et al. 2012. Methylation of ELOVL2 gene as a new epigenetic marker of age. Aging Cell 11: 1132–4. 18. Bibikova M, Barnes B, Tsan C, Ho V, et al. 2011. High density DNA methylation array with single CpG site resolution. Genomics 98: 288–95. 19. Hayflick L. 1965. The limited in vitro lifetime of human diploid cell strains. Exp Cell Res 37: 614–36. 20. Campisi J. 2013. Aging, cellular senescence, and cancer. Annu Rev Physiol 75: 685–705. 21. Koch CM, Joussen S, Schellenberg A, Lin Q, et al. 2012. Monitoring of cellular senescence by DNA-methylation at specific CpG sites. Aging Cell 11: 366–9. 22. Koch CM, Reck K, Shao K, Lin Q, et al. 2013. Pluripotent stem cells escape from senescence-associated DNA methylation changes. Genome Res 23: 248–59. 23. Finkel T, Serrano M, Blasco MA. 2007. The common biology of cancer and ageing. Nature 448: 767–74. 24. Cruickshanks HA, McBryan T, Nelson DM, Vanderkraats ND, et al. 2013. Senescent cells harbour features of the cancer epigenome. Nat Cell Biol 15: 1495–506. 25. Chen Z, Trotman LC, Shaffer D, Lin HK, et al. 2005. Crucial role of p53-dependent cellular senescence in suppression of Ptendeficient tumorigenesis. Nature 436: 725–30. 26. Michaloglou C, Vredeveld LC, Soengas MS, Denoyelle C, et al. 2005. BRAFE600associated senescence-like cell cycle arrest of human naevi. Nature 436: 720–4. 27. Marciniak-Czochra A, Stiehl T, Wagner W. 2009. Modeling of replicative senescence in hematopoietic development. Aging (Albany NY) 1: 723–32.
Bioessays 36: 0000–0000, ß 2014 WILEY Periodicals, Inc.
28. Moore A. 2012. TOR: the ancient link between cancer and ageing? Could ageing and cancer be unified by TOR via the ‘compromise’ struck between soma and germline at the dawn of multicellular life? BioEssays 34: 443–4. 29. Martins AC. 2011. Change and aging senescence as an adaptation. PLoS One 6: e24328. 30. Kirkwood TB. 2008. Understanding ageing from an evolutionary perspective. J Intern Med 263: 117–27. 31. Zane L, Sharma V, Misteli T. 2014. Common features of chromatin in aging and cancer: cause or coincidence? Trends Cell Biol, DOI: 10.1016/j.tcb.2014.07.001 32. Zhu ZZ, Sparrow D, Hou L, Tarantini L, et al. 2011. Repetitive element hypomethylation in blood leukocyte DNA and cancer incidence, prevalence, and mortality in elderly individuals: the Normative Aging Study. Cancer Causes Control 22: 437–47. 33. Issa JP, Ottaviano YL, Celano P, Hamilton SR, et al. 1994. Methylation of the oestrogen receptor CpG island links ageing and neoplasia in human colon. Nat Genet 7: 536–40. 34. Issa JP, Vertino PM, Boehm CD, Newsham IF, et al. 1996. Switch from monoallelic to biallelic human IGF2 promoter methylation during aging and carcinogenesis. Proc Natl Acad Sci USA 93: 11757–62. 35. Maegawa S, Gough SM, Watanabe-Okochi N, Lu Y, et al. 2014. Age-related epigenetic drift in the pathogenesis of MDS and AML. Genome Res 24: 580–91. 36. Xu Z, Taylor JA. 2014. Genome-wide agerelated DNA methylation changes in blood and other tissues relate to histone modification, expression and cancer. Carcinogenesis 35: 356–64. 37. Cancer Genome Atlas Research Network. 2013. Genomic and epigenomic landscapes of adult de novo acute myeloid leukemia. N Engl J Med 368: 2059–74. 38. Teschendorff AE, West J, Beck S. 2013. Age-associated epigenetic drift: implications, and a case of epigenetic thrift? Hum Mol Genet 22: 7–15. 39. Schoofs T, Rohde C, Hebestreit K, Klein HU, et al. 2013. DNA methylation changes are a late event in acute promyelocytic leukemia and coincide with loss of transcription factor binding. Blood 121: 178–87. 40. Schoofs T, Berdel WE, Muller-Tidow C. 2014. Origins of aberrant DNA methylation in acute myeloid leukemia. Leukemia 28: 1–14. 41. Ley TJ, Ding L, Walter MJ, McLellan MD, et al. 2010. DNMT3A mutations in acute myeloid leukemia. N Engl J Med 363: 2424–33. 42. Weidner CI, Walenda T, Lin Q, Wo¨lfler MM, et al. 2013. Hematopoietic stem and progenitor cells acquire distinct DNA-hypermethylation during in vitro culture. Sci Rep 3: 3372.
5
Ideas & Speculations
generous donation of Vision4 LifeSiences, and by the Interdisciplinary Center for Clinical Research (IZKF) within the faculty of Medicine at the RWTH Aachen University.
W. Wagner et al.
