The use of DNA methylation clock in aging research

Impact statement

The purpose of the Life Sciences is to provide for health and longevity. A significant component of this purpose is understanding the pathophysiological mechanisms of aging. DNA methylation clocks, which are based on the concept that DNA methylation status is a measure of chronological age, have been employed as biomarkers of both healthy and unhealthy aging, as well as disease risk. Herein we describe the utility of DNA methylation clocks as biomarkers of aging as well as their use during intervention therapies designed to promote healthy aging and longevity. A proper understanding of DNA methylation clocks, as aging biomarkers, will assist in the quantification of the rate of biological aging and will provide a foundation for the development of effective therapeutic aging interventions.

Introduction

Implementation of effective and reliable biomarkers for aging is essential to an understanding of the means by which to delay, halt, or even reverse the aging process. Chronological age, despite its wide use as a measure of aging, is not an effective biomarker for aging in that individuals of the same chronological age, especially the elderly, often have different physiological features. ¹ Thus, an active area of scientific pursuit has been the identification of more reliable and predictive markers of biological aging. Recently, studies involving both animals and humans have suggested that biomarkers based on DNA methylation are appropriate molecular markers of aging. DNA methylation-based biomarkers, which are also termed “epigenetic clocks” or “DNA methylation clocks,” can be extracted from different tissues and cell lines throughout the entire lifespan (antenatal to centenarian tissues).^2–5 The reversibility of DNA methylation is the most interesting feature of epigenetic clocks, suggesting their potential use for measurement of the efficacy of anti-aging interventions.^6,7 A recent systematic study compared six major biomarkers of biological aging including, telomere length, proteomic predictors, transcriptomic predictors, metabolomic predictors, composite biomarker predictors, and DNA methylation clocks. ⁸ The results of that study demonstrated DNA methylation-based biomarkers to be the most promising biomarker for aging. Herein we provide a brief overview of DNA methylation as well as a description of identified epigenetic clocks in humans and animals. Further, the utility of DNA methylation clocks as a guide for development of anti-aging therapies is discussed.

Overview of DNA methylation

DNA methylation is a form of chemical modification of genomic nucleotide bases, which can change gene expression without altering the DNA sequence. Cytosine-phosphate-guanine dinucleotides (CpGs) are prime targets for mammalian DNA methylation modification by addition of methyl (-CH3) groups to cytosine forming 5-methylcytosine (5mC). ⁹ Most CpGs are irregularly distributed and relatively rare in the mammalian genome. ¹⁰ In some regions, however, CpGs exist in groups or clusters known as CpG islands (CGIs). ¹⁰ Approximately 60%–70% of gene promoters have associated CGIs, ¹¹ with promoters classified based on their CpG content. Most CpGs are methylated in mammalian genomes. However, CGIs within transcriptional start sites (TSSs) of active genes are generally hypomethylated. ⁹

DNA methylation plays a key role in gene transcription by suppression of certain repetitive elements,^12–14 X-chromosome inactivation,^15–17 and repression of imprinted genes. ¹⁸ The process of DNA methylation-mediated gene silencing, which maintains the stability of transposable elements and other repeated DNA sequences, is believed to be driven by the binding of specific transcription factors to DNA and/or in cooperation with histone modifications.^19–21

DNA methylation is catalyzed by DNA methyltransferases (DNMTs) and Ten-eleven translocation (TET) enzymes. Mammalian DNMTs are a family of enzymes responsible for the production and maintenance of 5mC throughout the genome and include: DNMT1, DNMT3A, DNMT3B, and DNMT3L. ²² DNMT1 and the E3 ubiquitin-protein ligase, UHRF1, play a critical role in the maintenance of DNA methylation following replication. ²² UHRF1 specifically binds hemi-methylated CpGs and attracts DNMT1 via its ubiquitin-like domain.^23,24 DNMT1 then methylates the daughter DNA strand. Further, DNMT1 augments the activity of de novo DNMTs.^25–27 The DNMT3L enzyme, a non-catalytic DNMT type, binds to and activates DNMT3A and DNMT3B, ²⁵ which are both implicated in the process of de novo DNA methylation. TET1, TET2, and TET3, which are three members of the TET enzyme family, convert 5mC to 5-hydroxymethylcytosine (5hmC).^28–30 The establishment, maintenance, and removal of 5mC by DNMTs or TETs suggest that DNA methylation is a dynamic process that is balanced by the addition and removal of methyl groups. ³⁰ Even in the brain, which is composed of numerous slowly proliferating or non-proliferating cells, the methylation status of CpGs can oscillate in a circadian rhythm.^31,32 DNA methylation is also associated with the function of enhancers.^28,33 Taken together, DNA methylation is a tightly controlled and dynamic process that plays an important role in the regulation of gene transcription.

DNA methylation clocks in animals and humans

There is growing evidence suggesting the presence of differentially methylated DNA regions associated with aging.^6,34–37 With aging, hypermethylated regions are preferentially found in CGIs,^6,37 bivalent promoters, ³⁸ and Polycomb target genes.^35,39 Whereas age-related hypomethylated regions are preferentially found in non-CGIs promoters, enhancers, and late-replicating domains. ⁴⁰ DNA methylation patterns and age-related change substantially vary across tissue types. ⁴¹ Many studies have shown that a group of age-related CpG methylation changes are independent of tissue type, gender, and array platform. Several age prediction models have been developed by use of genome-wide methylation data and supervised machine learning (regression modeling).^3,4 Age estimated by DNA methylation, also known as epigenetic age or more precisely DNA methylation age, reflects not only chronological age, but also biological function. Due to its accuracy, DNA methylation age is often referred to as an “epigenetic clock” or “DNAm clock.” The definition of a DNAm clock is drawn from genome-wide methylation data that are significantly correlated (correlation coefficient ≥0.8) with chronological age or time. ⁴²

In 2011, Bocklandt et al. published the first DNAm clock to predict age, which was based on DNA extracted from saliva. ⁴³ For this clock, the degree of methylation of 88 CpG sites was found to significantly correlate with age in or proximal to 80 genes. Drawing on these initial results, Hannum et al. developed a highly accurate clock (r = 0.95) based on 71 CpGs from blood-derived DNA extracted from 656 human subjects, aged 19 to 101. This DNAm clock is commonly referred to as “Hannum’s clock”. ³ Since this DNAm clock was based primarily on adult blood samples, Hannum’s clock is thought to be less accurate for children^44–46 and non-blood tissue, ³ when compared to DNAm clocks optimized for those purposes. Several DNAm clocks were developed specifically from blood or buccal epithelial cells of children and adolescents.^45,46 These clocks performed well and exhibited a high level of precision for prediction of chronological age. Several epigenetic clocks containing fewer CpGs have also been shown to exhibit a high degree of accuracy.^47–50 As an example, a DNAm clock has been reported (Garagnani et al.) that was composed of only a single CpG within the ELOVL2 gene. ⁵¹ However, DNAm clocks with a higher number of CpGs are generally more accurate and robust. ⁴

Owing to differences in DNA methylation patterns and age-related changes between different cell types and tissues, ⁵² DNAm clocks need to be cautiously used for prediction of age. As an attempt to bypass these limitations, Horvath et al. developed a multi-tissue clock composed of 353 CpGs, with 8000 publicly available microarray databases generated through analysis of over 30 different DNA sources extracted from young children and adults. ⁴ The methylation status of 193 CpGs was found to positively correlate with age, with the remaining 160 CpGs negatively correlated. In that report, ticks of this DNAm clock were faster during growth and development. It is worth noting that the methylation status of many of the individual CpGs was only weakly associated with age. However, the comprehensive effects of 353 CpGs generated a robust biomarker of biological aging that conferred a high degree of accuracy. ⁴ While many technological platforms (e.g. microarrays, pyrosequencing, and next-generation sequencing methods) are available for the measurement of DNA methylation, Horvath’s clock is compatible with different technological platforms. For example, Horvath’s clock was successfully used with Illumina’s 450k and 27k DNA methylation microarrays as well as the Illumina Infinium MethylationEPIC array platform, despite the absence of 19 of 353 CpGs. ⁵³

Although first generation DNAm clocks exhibited a statistically significant correlation with a sizable range of age-related conditions, the effects were at best generally small to moderate. ⁵⁴ In 2018, Levine et al. developed a new clock, known as DNAm PhenoAge, which was based on phenotypic age stemming from nine age-related clinical measures instead of chronological age. ⁵⁴ The DNAm PhenoAge, composed of 513 CpGs, vastly outperformed the first generation of DNAm clocks with regard to predicting a variety of aging outcomes including mortality, health span, and cardiovascular disease. Smoking, a significant risk factor for mortality and morbidity, has been shown to correlate with DNA methylation changes. ⁵⁵ Levine et al. found that age estimated by DNAm PhenoAge significantly differed between smokers and non-smokers. Smoking did not influence age estimated by Hannum’s or Horvath’s clocks, with no association between smoking pack-years and DNAm PhenoAge.^54,56 Of note, a recently developed DNAm clock, termed “DNAm GrimAge,” incorporated DNA methylation-based surrogate biomarkers for seven plasma proteins and smoking pack-years. ⁵⁷ The ability of the DNAm GrimAge to predict risk for all cause-death and age-related disease was found to be better than previously described for DNAm clocks. Using different CpGs, Hannum’s and Horvath's epigenetic clocks were developed to simply predict actual age. In contrast, DNAm PhenoAge and GrimAge used CpG methylation to predict previously proposed age-related mortality and phenotypic indicators, closely related to actual age.

Due to the success of DNAm clocks in predicting chronological age in humans, DNAm clocks were developed for other species. Within a 5-year span, several DNAm clocks were developed for mice^2,58–60 and dogs/wolves. ⁶¹ Compared with human clocks, which use DNA methylation microarrays to identify subsets of CpGs, animal clocks were developed using Reduced Representation Bisulfite Sequencing, an efficient high-throughput sequencing technique for the analysis of methylation levels at the genomic mononucleotide level. As mice are commonly used for experimentation, DNAm clocks for mice were a major endeavor for use in geroscience. Such clocks have the potential to provide critical information about the underlying mechanisms of aging related to DNA methylation. Murine DNAm clocks have been used successfully to evaluate several interventions that resulted in extended lifespan and delayed aging in mice.^6,62–65 In vivo, the DNAm clocks described above have been shown to be highly accurate and robust for prediction of the biological age of tissues and organs. However, the in vitro efficiency and accuracy of these epigenetic clocks for fibroblasts or other types of cells in culture has been less accurate and robust. Hence, another DNAm clock, referred to as the skin & blood clock, was developed, which is composed of 391 CpGs from a variety of human cell sources: fibroblasts, keratinocytes, buccal cells, skin, blood, and saliva for in vivo and ex vivo studies. ⁵ This DNAm clock was shown to accurately track the dynamic aging process of cultured cells ex vivo.

The usefulness of DNA methylation clocks for aging research

Anti-aging interventions are critical methods to identify the underlying biological mechanisms or pathways of aging. An effective, accurate, and widely applicable biomarker for aging is of major importance to the evaluation of anti-aging interventions. ⁵⁶ Use of DNAm clocks in the general population can lead to identification of outliers characterized by a difference between DNA methylation age and chronological age. “DNAm age acceleration” is a term used to define those individuals with a high DNA methylation age compared to chronological age.

Internal and external DNAm age acceleration

Change in the proportion of naïve and memory T cells is a primary feature of aging with respect to blood cell composition. The age estimated by Hannum’s clock, DNAm PhenoAge, and the GrimAge clock is influenced by variations in cell type composition. In contrast, age predicted by Horvath’s clock is primarily undisturbed by such variations.^3,4,54,57 A recent study found that the accuracy of DNAm clocks increases with an increase in training sample size, although prediction of mortality was lower. ⁶⁶ One explanation is that a near-perfect age predictor estimates internal cell aging rather than disease-based external aging dynamics. Thus, Horvath’s clock is regarded as an estimator of internal cell aging, whereas Hannum’s clock, DNAm PhenoAge, and DNAm GrimAge clocks measure external aging. The discovery of the predictive utility of intrinsic aging acceleration is biologically compelling and points to a new frontier in aging research. Internal aging may reflect an inherent epigenetic clock that is related to mortality that is independent of actual age, changes in blood cell composition, and traditional mortality risk factors. ⁵⁶

DNAm age acceleration, mortality, and age-related diseases

Age is the biggest risk factor for mortality. However, the biological responses to aging differ among individuals. The difference between DNA methylation age and chronological age (DNAm age acceleration) has been proposed as an indicator of “biological” aging associated with all cause-mortality and age-related diseases.^55,67,68

DNAm age acceleration and health risk factors

If the acceleration of DNA methylation age reflects biological age, it is reasonable to assume that known risk factors for poor health will influence acceleration. An increasing body of evidence suggests that health risk factors such as obesity, environmental exposure, lifetime stress, and/or genetic factors may significantly and positively correlate with DNAm age acceleration.

DNAm age acceleration and genetic factors

In addition to the above factors, genetics impacts DNA methylation age with high heritability (h²) estimates of approximately 40%.^56,121 This is consistent with the observation that ticks of DNA methylation clocks are highly stabile throughout life. ¹²² The importance of the cell-intrinsic factors controlling DNA methylation age has been assessed in a recent study of hematopoietic stem cell transplantation (HSCT). ¹²³ DNA methylation age of the transplanted donor’s blood cells progressed without detectable interference from the recipient's age, even 17 years after HSCT. ¹²³ Therefore, identifying genes that affect the rate of the DNA methylation clock may provide insight into the underlying mechanistic basis for the aging process. Deep-Bis DNA methylation analysis has identified several loci associated with DNAm age acceleration. These include MLST8, EFCAB5, and TERT genes in the cerebellum, multiple brain regions, and the blood, respectively.^124–126 The specific functions of these genes are described in a recent review. ⁵⁶ Other studies have identified several genes that are highly correlated with DNA methylation age including ELOVL fatty acid elongase 2 (ELOVL2) and H3K36 methyltransferase (NSD1).^127,128 ELOVL2 is the master controller of poly-unsaturated fatty acid synthesis and was most significantly related to DNA methylation age. Although the relationship of ELOVL2 to DNA methylation clocks has been well documented, the mechanism underlying how age-related ELOVL2 methylation contributes to aging is unknown. A recent report documented a dramatic increase in ELOVL2 methylation that was accompanied by downregulation of ELOVL2 expression levels in aged human fibroblasts. ¹²⁸ Moreover, dysfunction of the Elovl2 protein interferes with lipid metabolism and synthesis, increasing endoplasmic reticulum stress and mitochondrial dysfunction, which are associated with accelerated aging. A recent gene screening study found that DNA methylation clocks tick more quickly in the blood of patients with Sotos syndrome, a familial developmental disorder caused by inactivation mutations of NSD1. ¹²⁷ NSD1 is a histone methyltransferase containing a SET catalytic domain, which regulates the di-methylation of histone H3 lysine 36. ¹²⁷ Further, the aging process and the Sotos syndrome share methylation changes as shown by genome-wide methylation analysis, which suggest that developmental diseases may serve as models for the study of the mechanistic basis for aging. P53 is considered to be one of the most important cellular tumor suppressors, with numerous studies demonstrating P53 mutations to strongly correlate with a slower acceleration of epigenetic age in several types of malignancies. ⁴ P53 has been hypothesized to be part of an epigenetic maintenance system. This hypothesis has been supported by a recently published study showing that the P53 inhibitor, SV40LT, maintained or reversed DNAm age acceleration in different protoculture fibroblast lines. ¹¹⁸

DNA methylation clocks and the anti-aging effect of caloric restriction

A significant number of interventions have been reported to prolong the lifespan of mice. Currently, caloric restriction (CR) and CR mimetics, such as rapamycin, are the most studied anti-aging interventions. ¹²⁹ Numerous reports have shown CR and CR mimetics to extend lifespan and delay the aging rate of rodents and non-human primates, with a current focus on the identification of an accurate age-related biomarker such as DNA methylation clocks. Petkovich et al. ⁵⁸ demonstrated CR to delay DNAm age acceleration in mice. Rapamycin, an inhibitor of mTOR activity, recapitulates the effect of calorie-restriction. Based on the skin & blood clocks noted in the Section “DNA methylation clocks in animals and humans,” epigenetic aging of primary human keratinocytes was independent of replicative senescence, somatic cell differentiation, and cellular proliferation rate, which are delayed by rapamycin. ¹³⁰ The effects of mTOR on DNA methylation age are consistent with the findings of a study that showed the MLST8 protein, a subunit of the mTOR complex, positively correlated with DNA methylation aging of the human cerebellum. ¹²⁴ Hence, these experiments not only demonstrate the power of DNAm clocks to reveal biological features at the cellular and organismal levels, but also support the possibility that mTOR pathway signaling may be involved in epigenetic aging.

Future prospects and recommendations

DNA methylation clocks, novel biomarkers of aging, are considered useful for measuring chronological age, as well as reflecting some characteristics of biological aging. As noted above, these clocks have a unique function in capturing differences in biological aging, and they also have great potential in predicting changes in aging, cognitive and physical function, survival rates in association with age-related diseases, as well as identification and/or validation of effective anti-aging interventions. Most aging biomarkers are based on molecules, cell or animal systems, and require strict verification in humans. An epigenetic clock suitable for in vitro studies could be used to test the effects of potential anti-aging drugs on lifespan and aging in a relatively short period of time with a low cost, not easily achieved in human cohort studies. ⁵ However, certain issues deserve attention with regard to the application of DNAm clocks. Such clocks are based on differing statistical methods, specific sample sizes, and tissue types. Each of these must be considered for individual strengths and weaknesses in selection of specific DNAm clocks for evaluation. Moreover, there is no definitive evidence supporting a cause-and-effect relationship between DNA methylation age and aging. Hence, data derived from DNAm clocks should be interpreted prudently with a thorough and comparative analysis with other measures of aging. In this manner, DNAm clocks can be effectively used to improve animal and human health.