Modified aging of elite athletes revealed by analysis of epigenetic age markers

DNA methylation status

The study revealed significantly altered DNA methylation patterns between the whole group of athletes and controls in TRIM59 (P=1.894x10^-7) and KLF14 (P=8.974x10^-10), as well as separately for endurance and power athletes (Table 1). Although the results were significant for both endurance and power athletes, a more marked difference was noted in TRIM59 C7 for the power athletes (P=2.946x10^-7). In KLF14 C1, hypermethylation was observed for power and endurance athletes with a similar level of significance of P=4.987x10^-7 and P=1.866x10^-7, respectively (Table 1). Analyses performed separately for females and males confirmed TRIM59 and KLF14 markers to be significantly hypermethylated in athletes of both genders. (Supplementary Table 1).

Age prediction and age acceleration

Table 2 shows that our established multivariate 5-CpG ANN model predicted the whole group of elite athletes with a lower accuracy of MAE = 3.3, compared to healthy controls of similar mean age and age distribution that had a MAE = 2.4, P = 3.815x10^-4. When the two athlete groups were analysed separately, a significant decrease of age prediction accuracy from MAE=2.4 to MAE = 3.7 was only found for power athletes (P=5.727x10^-5). Analyses performed separately for females and males confirmed reduced accuracy of age prediction in the group of power athletes vs. controls in both genders (P=0.013 and P=0.002, for females and males respectively); outlined in Supplementary Table 2.

Age acceleration from chronological and methylation age divergence (determined using the 5-CpG model) was further calculated and tested for its association with intense physical exercise. This analysis showed the status of elite athletes is significantly associated with age acceleration (P=1.503x10^-7) and was significant for both groups (1.051x10^-9 and 0.004 for power and endurance athletes, respectively; Table 3). Analysis performed at the level of single CpG predictors showed that significant association of the status of elite athletes with age acceleration arose from increased DNA methylation of the CpGs at TRIM59 and KLF14. The level of significance of association between the status of power athletes and age acceleration was highest when just TRIM59 was used as a predictor of age (P=1.239x10^-10 comparing to P=0.003 obtained for endurance athletes), but at a comparable level in KLF14 in both groups (P=3.345x10^-7 and P=2.116x10^-8, for endurance and power groups, respectively). The type of sport competition performed was also found to be significantly associated with age acceleration (P=0.003) indicating the increased aging of athletes engaged in power sports compared to endurance sports (Table 3).

Predicted age versus chronological age in endurance and power athletes plus the group of controls is presented on Figure 1, which shows predictions based on the 5-CpGs model and TRIM59 and KLF14 separately. Figure 2 shows that age prediction based on CpG site C1 in KLF14 indicated athletes to be an average 5.5 years older than healthy controls (P=1.766x10^-9), endurance athletes to be 5.9 years older than controls from the general population (P=2.511x10^-7) and power athletes to be 5.1 years older than healthy controls (P=8.443x10^-7). Age prediction performed with DNA methylation values from CpG site C7 in TRIM59 showed athletes to be an average 4.5 years older than healthy controls (P=9.405x10^-8), power athletes to be 5.6 years older than healthy controls (P=2.105x10^-7) while endurance athletes were predicted to be 3.3 years older than healthy controls (P=0.003).

The impact of physical exercises on epigenetic age

Very intensive physical exercise as a routine part of elite athlete training has been shown to have favorable effects on lifespan longevity and is associated with a lowered risk of cardiovascular diseases and cancer in later life [19,20,24]. Importantly, individuals with active lifestyles but free from the stress of extreme exercise regimes benefit from regular physical activity by reducing their risk of all forms of mortality, and in proportion to the intensity of the training [25]. Better insight into the molecular processes behind the aging of elite athletes is now possible thanks to recent advances in epigenetics and the efficiency of DNAmet age prediction systems in providing reliable indicators of the aging rate. Our study indicates accelerated aging of the whole group of elite athletes, with power athletes shown to have higher levels of age acceleration compared to endurance athletes. The observed acceleration of aging may potentially have significant implications, as it has been shown that a difference between the chronological age of an individual and the age measured by DNA methylation is an indicator of a person’s overall condition and is predictive of mortality in later life.

The meaning of epigenetic age

The epigenetic clock CpGs defined as correlated with age consistently between individuals and included in various age prediction models, reveal some degree of inter-individual variability [26]. There is an on-going debate about the real meaning of age measured by DNAmet status. The picture arising from recent research indicates that various markers included in age estimation models may be regulated in a different way. Particular age-correlated CpGs could be liable to moderation by inherited factors and the environment [2]. It has been shown that the epigenetic clock developed by Horvath can also measure age in non-proliferative tissues but at the same time there is evidence that hypermethylation of one of the most informative age predictors – ELOVL2, is associated with cell replications and thus indicates a mitotic age [21]. Therefore, it is conceivable that some DNAmet markers may be more sensitive to the environment than others and consequently they can provide a better measure of epigenetic drift. Consequently, it is unsurprising that not all age predictors are informative in terms of increased mortality risk. The study of Lin et al selected 37 CpGs indicative of life expectancy from their full set of 99 markers. Additionally, they tested alternative age prediction models and found that the marker sets proposed by Horvath and Hannum contained mortality sensitive CpG sites [12,1,2]. Two out of five markers investigated here, namely TRIM59 and KLF14, are among the suggested life expectancy-sensitive markers from the Hannum model. TRIM59 belongs to the ten most informative markers from the age predictors included in that age prediction model [12]. Notably, an independent study has demonstrated that 5-year higher DNA methylation age predicted by the Hannum model was associated with a 21% increased mortality risk [11]. Overall, multiple studies have reported an 11% to 35% increase in mortality risk for individuals with a 5-year higher DNA methylation age [11,12,27]. Recently, Zhang et al presented a list of life expectancy DNA methylation markers which are independent from the epigenetic clock, providing further evidence that aging related markers can be different from markers predicting chronological age [13].

The molecular role of individual age-correlated DNA methylation markers

More detailed analysis of the accelerated aging of elite athletes indicated by our model revealed that accelerated hypermethylation of TRIM59 and KLF14 were the principal cause of the changes. Both, are widely replicated age predictors used in various age estimation models based on DNA methylation analysis [2,22,28,29]. Most studies have indicated longer life expectancy in top athletes compared to the general population and our finding of an accelerated epigenetic age of elite athletes seems to contradict these results. However, careful inspection of the molecular nature of TRIM59 and KLF14 might reveal their specific impact on aging processes in these athletes. TRIM59 has recently emerged as a powerful oncogene involved in induction of cellular senescence. Valiyeva and colleagues have shown that TRIM59 can affect Ras and RB signal pathways and its knockdown leads to S-phase cell cycle arrest and cell growth retardation [30]. Importantly, it has been shown that interaction between Ras and pRb may launch cellular processes leading to cellular senescence and tumor suppression [31]. There is growing evidence that TRIM59 is a promoter of cell proliferation and migration. Increased expression of this gene has been reported in various tumors [32–35]. In contrast to the initial perception that cellular senescence is harmful, it has been proposed to have beneficial biological implications. Cellular senescence acts as a tumour-suppressive mechanism that irreversibly blocks cellular proliferation in response to stress and through this process, creates a barrier to the development and progression of cancer. Moreover, this apparent cancer moderation is accompanied by the acquisition of a senescence-associated secretory phenotype that induces cell plasticity which is important in tissue regeneration or wound healing [36]. Thus, hypermethylation of TRIM59 indicates accelerated aging of elite athletes, especially involved in power sports but at the same time it may have an anticancer effect. Interestingly, a study of mutants in the p53 tumor suppressor gene has shown that increased tumor protection occurs from modified p53 protein, while at the same time these induce early onset aging in mice; linking senescence, aging and cancer [37]. Our previous study revealed aberrant hypermethylation of TRIM59 in early-onset Alzheimer’s disease patients and in Graves’ disease patients [23]. This seems to further support the hypothesis of the involvement of TRIM59 in cellular senescence since chronic neuro-inflammation has been shown to be important in aging and neuro-degeneration, while autoimmunity is known to induce inflammation, tissue damage and remodelling [38,39].

The observed hypermethylation of KLF14 also indicates accelerated aging of elite athletes. Recent findings showing the role of KLF14 in chronic inflammatory responses and the pathogenesis of atherosclerosis may be particularly relevant to this finding [40,41]. Wei et al showed that KLF14 may be involved in induction of inflammation and for this reason knockdown of this gene reduces pro-inflammatory cytokines and the formation of atherosclerotic lesions. This anti-inflammatory effect was suggested to prevent atherosclerosis: a known chronic inflammatory disease [40]. It is interesting to speculate whether DNA methylation of TRIM59 is involved in the process of cellular senescence and the associated beneficial biological effects. Similarly, the hypermethylation of KLF14 could have an anti-inflammatory role preventing cardiovascular disease. These hypotheses deserve further investigation, but they may already explain the increased life expectancy of elite athletes and in particular, the lower risk they have to develop cancer and cardiovascular diseases [19].

Samples and analysis of DNA methylation

The study was approved by the Ethics Committee of the Józef Piłsudski University of Physical Education in Warsaw (SKE 01-50/2012) and the institutional ethical review board of the Ethics Committee of Investigation in Galicia, Spain (CAEI: 2013/543). Written informed consent was obtained from the elite athletes and counterpart controls from the general population forming the study subjects. Peripheral blood was collected in EDTA tubes from 176 elite athletes in the years 2010-2014. Chronological age at the time of blood collection was used for all the calculations in this study. Top-level professionals (including medallists of Olympic Games and World championships) in power and endurance sports, 100 athletics, 21 speed skaters, 28 swimmers and 27 rowers were enrolled in the study. 128 blood samples from the general population matched by age and gender, were recruited for the control group (Table 4). DNA was extracted using standard methods and quantified using Qubit dsDNA High Sensitivity Assay Kit (Thermo Fisher). From 500 ng to 2 μg of DNA was treated with sodium bisulfite using the EpiTect 96 Bisulfite Kit (Qiagen, Hilden, Germany), following manufacturer’s guidelines. Specific CpG sites were analysed using previously described primers and methods [22]. Details of the analysed sites are presented in Table 5.

Statistical analyses and prediction of DNA methylation age

Analyses were performed using PS IMAGO 4 (IBM SPSS Statistics 24). DNA methylation was measured in the two groups of 176 elite athletes plus 128 healthy counterparts of the age/gender matched control group at 5 CpGs included in the age prediction model (Table 5). Methylation data was analysed using independent sample Student’s t test and the balanced distribution of age in the athlete and control groups confirmed with nonparametric Kolmogorov-Smirnov test. Age predictions were made with the method developed by Zbieć-Piekarska et al. [22] after remodelling with the artificial neural networks (ANN) approach applied in the study of Spólnicka et al. [23]. The training set of 305 individuals from the above previous studies was used for the age predictions of the current study. The predicted age of athletes was compared with their true chronological age to calculate mean absolute error (MAE). Independent sample Student’s t test was used to compare mean predicted age and MAE. Age acceleration (AA) - a measure of the discrepancy between methylation age and chronological age - was designated in the form of residuals calculated from linear regression analysis; where predicted age was treated as the dependent variable and chronological age as the independent variable as described in [14]. Association analysis of individual levels of intense training with age acceleration was tested using linear regression, controlling for age and gender and in some analyses, for the number of years spent in training.