AGE

, Volume 34, Issue 1, pp 169–179

Global DNA methylation in old subjects is correlated with frailty

Authors

  • Dina Bellizzi
    • Department of Cell BiologyUniversity of Calabria
  • Patrizia D’Aquila
    • Department of Cell BiologyUniversity of Calabria
  • Alberto Montesanto
    • Department of Cell BiologyUniversity of Calabria
  • Andrea Corsonello
    • Italian National Research Center on Aging
  • Vincenzo Mari
    • Italian National Research Center on Aging
  • Bruno Mazzei
    • Italian National Research Center on Aging
  • Fabrizia Lattanzio
    • Italian National Research Center on Aging
    • Department of Cell BiologyUniversity of Calabria
Article

DOI: 10.1007/s11357-011-9216-6

Cite this article as:
Bellizzi, D., D’Aquila, P., Montesanto, A. et al. AGE (2012) 34: 169. doi:10.1007/s11357-011-9216-6

Abstract

Epigenetic variations have been widely described to occur during the aging process. To verify if these modifications are correlated with the inter-individual phenotypic variability of elderly people, we searched for a correlation between global DNA methylation levels and frailty. We found that the global DNA methylation levels were correlated to the frailty status in middle/advanced-aged subjects but not with age. A 7-year follow-up study also revealed that a worsening in the frailty status was associated to a significant decrease in the global DNA methylation levels. These results suggest that the relaxation of the epigenetic control in aging is specifically associated with the functional decline rather than with the chronological age of individuals. Thus, the modifications of DNA methylation, representing a drawbridge between the genetic and the environmental factors affecting the age-related decay of the organism, may play an important role in determining physiological changes over old age.

Keywords

FrailtyGlobal DNA methylationEpigeneticsAging

Introduction

Frailty represents a clinical syndrome which is prevalent in the elderly population. It is characterized by the decrease of total physiological reserves and a progressive deregulation in the homeostatic network underlying maintenance and repair in the aging body (Lang et al. 2009; Bandinelli et al. 2010). A wide range of geriatric phenotypes are commonly observed to be associated with frailty including weight loss and sarcopenia, muscle weakness with low grip strength, low activity level, exhaustion, low body mass index, blood pressure instability and balance and gait abnormalities (Fried et al. 2004; Bergman et al. 2007; Topinková 2008; Davis et al. 2010). In addition, in the characterization of the clinical phenotype of frailty, it has been very important the demonstration of the role in this syndrome of multi-system impairments including immune/inflammatory, multiple hormonal and neuromuscular deregulations, as well as metabolic and vascular alterations and oxidative stress that commonly occur with aging (Ferrucci et al. 1999; Leng et al. 2002; Schalk et al. 2004; Walston et al. 2006; Barzilay et al. 2007; Landi et al. 2008; Hubbard et al. 2009; Blaum et al. 2009; Serviddio et al. 2009; Desai et al. 2010; Lustosa et al. 2010; Evans et al. 2010; Hyde et al. 2010; Maggio et al. 2010; Hubbard and Woodhouse 2010). As a consequence, frailty increases the individual vulnerability to age-related disorders, such as myocardial infarction, rheumatoid arthritis, diabetes and hypertension (Hubbard et al. 2010). Moreover, an association between frailty and the risk of mild cognitive impairment and a rapid rate of cognitive decline was described (Buchman et al. 2007; Boyle et al. 2010).

In recent years, many operational definitions of frailty have been proposed. Montesanto et al. (2010), by using a hierarchical cluster analysis that availed of specific geriatric parameters, identified a specific approach to define the frailty status. The diagnostic and predictive soundness of this classification was confirmed by a detailed survival analysis which showed higher survival chance for subjects characterized by lower degree of frailty.

Frailty is commonly considered as the resultant of a multidimensional interplay of genetic, biological, psychosocial and environmental factors (Fried et al. 2001; Rockwood et al. 2004; Rockwood and Mitnitski 2007; Bergman et al. 2007; Lang et al. 2009; Fulop et al. 2010; Landi et al. 2010). Several lines of evidence suggest that epigenetic reprogramming of gene expression occurs during aging (Fraga 2009; Kawakami et al. 2009; Gravina and Vijg 2010; Rando 2010). A correlation between epigenetic DNA modifications and quality of aging has been shown by Fraga et al. (2005), who found that global and locus-specific differences in DNA methylation in identical twins of various ages are influenced by different environmental factors and lifestyle. In vitro models and studies carried out in tissues and blood DNA samples demonstrated a decrease in global cytosine methylation during aging mostly due to the demethylation in transposable repetitive elements (Castro et al. 2003; Bjornsson et al. 2008; Bollati et al. 2009). On the other hand, additional studies reported an age-related hypermethylation in promoter regions of specific genes, with a consequent decrease of correspondent mRNA levels. In humans, this hypermethylation has been observed for genes involved in cell cycle regulation, tumor-cell invasion, DNA repair, apoptosis, metabolism and cell signaling, implicating their potential role in age-related diseases (Wilson et al. 1987; Oakes et al. 2003; Richardson 2003; Fuke et al. 2004; Fraga and Esteller 2007; Ling et al. 2008; Arai et al. 2010; Lee et al. 2010).

In the present study, we investigated whether the frailty status in elderly people was associated to global DNA methylation levels. Thus, we searched a correlation between the age-related functional decline and epigenetic modifications. In addition, a longitudinal study carried out in a subgroup of the original sample allowed us to determine whether the methylation levels were subject to changes over time in agreement with changes in the frailty status. Global DNA methylation was measured by applying the assay developed by Anisowicz et al. (2008) to peripheral blood DNAs collected from 318 subjects classified for frailty phenotype according to Montesanto et al. (2010). In the longitudinal study, both frailty evaluation and DNA methylation analysis were repeated after 7 years.

Materials and methods

Population sample

A total of 318 unrelated individuals (144 men and 174 women), 65–105 years old (median age 79.5 years), participated in the present study. All the subjects lived in Calabria (southern Italy) and their origin in the area had been ascertained up to the grandparents generation. A more detailed sample description can be found elsewhere (De Rango et al. 2010). Health status was ascertained by medical visit carried out by a geriatrician who also conducted a structured interview including questions on common diseases occurred in the past. At the same time of the visit, peripheral venous blood samples were also obtained. Before the interview, each subject consented to her/his phenotypic and genetic data to be used anonymously for genetic studies on aging (informed consent).

In a previous work, by using a hierarchical cluster analysis (HCA) which availed of specific geriatric parameters, this sample was used to identify specific aging phenotypes (Montesanto et al. 2010). In particular, the sample was analyzed considering two different age groups. The first (S1) included 217 subjects (94 males and 123 females) 65–89 years old (median age 75 years); the second (S2) included 101 subjects (50 males and 51 females) older than 90 years (median age 99 years). By using this approach in the S1 sample, three clusters were identified: nonfrail (the cluster with subjects showing the best scores for the classification variables), frail (the clusters with subjects showing the worst scores for the classification variables), and prefrail (the cluster with subjects showing intermediate scores for the classification variables). In the S2 sample, two clusters were identified. Similar to the first classification, the two clusters obtained were defined as frail (the cluster with subjects showing the best scores for the classification variables) and very frail (the cluster with subjects showing the worst scores for the same variables).

The diagnostic and predictive soundness of these classifications were confirmed by a 3-year longitudinal study. In fact, a detailed survival analysis showed higher survival chance for subjects characterized by lower frailty in these classifications.

In order to better evaluate the relationship between DNA methylation levels and degree of frailty, a randomly selected sample consisting of 98 prefrail and nonfrail subjects of S1 group were reconsidered after 7 years from the baseline visit. During the follow-up period, 55 subjects (56%) died. Of the remaining 43 alive subjects, 6 refused the visit while for the remaining others (37), a new multidimensional geriatric evaluation was performed and a new sample of blood for the evaluation of DNA methylation was provided.

In vitro DNA methylation of control lambda DNA

Five hundred nanograms of lambda DNA (Sigma) were completely methylated by using 2.5 U of M.HpaII methylase (New England Biolabs). The mix was incubated at 37°C for 3 h and successively at 65°C for 15 min to inactivate the enzyme.

Restriction analysis of control lambda DNAs, control human genomic DNAs and population sample DNAs

One hundred nanograms of methylated and unmethylated lambda DNA were separately incubated with 5 U of HpaII and MspI restriction endonucleases (New England Biolabs) at 37°C overnight and successively at 65°C for 20 min to inactivate the endonucleases. The samples were loaded on a 1.4% agarose gel, electrophoresed in TAE (Tris/Acetate/EDTA) buffer and stained with ethidium bromide.

One hundred nanograms of CpGenome™ Universal Methylated DNA (Chemicon), 100 nanograms of CpGenome™ Universal Unmethylated DNA (Chemicon) and 100 ng of a mixture obtained combining 50 ng of CpGenome Universal Methylated and 50 ng of Unmethylated DNA were separately incubated with 5 U HpaII and MspI restriction endonucleases at 37°C overnight and successively at 65°C for 20 min to inactivate the endonucleases. The samples were loaded on to 1.4% agarose gel, electrophoresed in TAE (Tris/Acetate/EDTA) buffer and stained with ethidium bromide.

One hundred nanograms of population sample DNAs, extracted from venous blood buffy coats following standard procedures, were digested with HpaII and MspI by using the above experimental conditions.

Measurement of global DNA methylation levels

The global DNA methylation levels of control lambda DNAs, control human DNAs, and population sample DNAs were estimated by using the CpGlobal assay, designed and developed by Anisowicz et al. (2008). Briefly, 2 μM of both Biotin-11-dCTP and Biotin-11-dGTP (Perkin Elmer) were added to digested DNA samples in an end-fill reaction of 20 μl carried out in presence of biotinylation buffer (40 mM Tris–HCl pH 7.5, 20 mM Tris–HCl, 50 mM NaCl), and 2 U of Sequenase (USB Corporation). After incubation at 37°C for 30 min, 100 μl of Reacti-BindTM DNA Coating Solution (Pierce) were added and the samples were shaken in an orbital platform at room temperature overnight. The solution was removed and the samples were washed three times with Dulbecco’s Phosphate-Buffered Saline (Sigma). Then, 200 μl of the detector block solution (KPL) were added and the mixtures were incubated at room temperature for 30 min. After the removal of the solution, 150 μl of the Detector Block Solution containing 0.5 μg/ml of HRP Streptavidin (KPL) were added and the samples were incubated at room temperature for 30 min. After the incubation, the Detector Block Solution was removed and the samples were washed four times with Biotin Wash Solution 1X (KPL). Then, 150 μl of LumiGlo Chemiluminescence substrate (KPL) were added and after 2 min the chemiluminescence emitted from each sample was quantified in a Lumat LB9507 luminometer (EG&G Bertold). Each sample was analyzed three independent times in triplicates. In order to determine the possible “background noise” and, thus, to calculate the net luminescence for each sample a control lacking of enzyme was also analyzed. The data were calculated as Global DNA Methylation Index (GDMI) by dividing the mean net luminescence values for the HpaII enzyme to the mean net luminescence values for the MspI enzyme. Thus, the GDMI values inversely correlate to the global DNA methylation levels.

Statistical analysis

Descriptive statistics for continuous and categorical variables were used to describe socio-demographic characteristics, the frailty phenotypes and the GDMI values of the study sample. For continuous variables, measures of central tendency and dispersion, including mean, median and standard deviations were reported. Categorical variables were examined by analyzing the relevant frequency distributions. ANOVA test followed by multiple comparison test (Dunnett’s test) was used to compare the values of GDMI among the frailty groups defined by the HCA approach. Linear regression analysis was performed for testing the dependence between GDMI and age.

After the follow-up period, in order to determine the actual frailty status of the 37 subjects (re-classification) a classification tree (CT) model has been used (Breiman et al. 1984). In fact, the main HCA drawback is that classification of new cases is usually not affordable, without repeating the whole HCA analysis. Thus, we used a supervised classification approach in order to obtain an easy-to-understand model that can be used for classifying the revisited subjects. In brief, we applied the CT algorithm to the baseline data by considering the classification provided by HCA analysis in S1 group as dependent variable, while the geriatric parameters (Hand Grip, Mini Mental State Examination, Activities of Daily Living and Self Reported Health Status) as independent variables. Adjusted values (i.e., with normalization) were used for all variables (Montesanto et al. 2010). The performance of the CT model was evaluated in terms of accuracy using a tenfold cross-validation strategy. The obtained model was then used to re-classify the revisited subjects.

Statistical analyses were performed using SPSS 15.0 (SPSS Inc., Chicago, IL). A significance level (α) of 0.05 was chosen in all the tests.

Results

We carried out a series of control experiments in order to validate the reproducibility of the CpGlobal assay described in the “Materials and methods” section (Anisowicz et al. 2008). To this purpose, we analyzed two samples of lambda DNAs and three human genomic DNAs as quality control.

Global DNA methylation levels in control samples

Control lambda DNA

Two samples of lambda DNA were analyzed: one was unmethylated; the second one was in vitro methylated by using the bacterial M. HpaII methylase. Firstly, to verify the restriction conditions and the fully methylation we treated the two samples with the methyl-sensitive HpaII and the methyl-insensitive MspI restriction endonucleases. As shown in Fig. 1a, in contrast to unmethylated DNA, no HpaII restriction fragments were observed in the methylated lambda DNA. In fact, HpaII activity was greatly inhibited by the methylation on the internal cytosine residue in its recognition sequence. As expected, the MspI activity was not affected by the absence/presence of methylated cytosines. Then, the CpGlobal assay was applied to determine the GDMI values in the two lambda DNA samples. As expected, the GDMI value was significantly lower (about 0.1) in the methylated DNA sample than in the unmethylated DNA (about 1.0; Fig. 1b).
https://static-content.springer.com/image/art%3A10.1007%2Fs11357-011-9216-6/MediaObjects/11357_2011_9216_Fig1_HTML.gif
Fig. 1

Control lambda DNA analysis. a Ethidium bromide-staining agarose gel showing unmethylated and methylated lambda DNAs digested with HpaII and MspI. ND: lambda DNA not digested. b GDMI values of unmethylated and methylated lambda DNAs. The values represent the main of three independent triplicate experiments with standard error mean

Control human genomic DNAs

CpGlobal assay was subsequently applied to three control samples of human genomic DNA: one fully unmethylated, one fully methylated, and another obtained by mixing an equal ratio of unmethylated and methylated DNA. We restricted the samples as described above. We observed that when fully methylated human DNA was digested with HpaII, no restriction fragments were observed (Fig. 2a). Conversely, when fully unmethylated human DNA was digested with the same enzyme, several restriction fragments were observed, indicating the complete digestion of the sample. Finally, the DNA sample obtained by mixing an equal ratio of unmethylated and methylated human DNA, showed an intermediate result. In fact, the intensity of the lane of undigested DNA approximates the 50% of that untreated with HpaII, as from densitometric tracings (data not shown). As expected, the GDMI values in the fully methylated, in the mix of methylated and unmethylated DNA, and in the fully unmethylated DNA were about 0.1, 0.5, and 1.0, respectively (Fig. 2b).
https://static-content.springer.com/image/art%3A10.1007%2Fs11357-011-9216-6/MediaObjects/11357_2011_9216_Fig2_HTML.gif
Fig. 2

Control human DNA analysis. a Ethidium bromide-staining agarose gel showing unmethylated and methylated human DNAs and a mixture of equal amount of unmethylated and methylated human DNAs digested with HpaII and MspI. ND human DNA not digested. b GDMI values of unmethylated, methylated and the mixture of unmethylated and methylated human DNAs. The values represent the main of three independent triplicate experiments with standard error mean

On the whole, the results obtained by the above control experiments demonstrated the accuracy and an overall high reproducibility of the CpGlobal assay in measuring the global DNA methylation levels of the different sample analyzed.

Global DNA methylation levels in the population sample

After the validation procedure, the CpGlobal assay was applied to measure the GDMI values of the DNA samples extracted from peripheral venous blood collected from 318 subjects (65- to 105-year-old subjects).

The distribution frequency of the GDMI values in the population sample is reported in Fig. 3. The distribution of these values in the sample was quite normal with a mean value of about 0.517. In addition, they were quite similar for men and women (p value = 0.424) and were not correlated with age (r = −0.040, p value = 0.474). These results suggested that global DNA methylation levels do not correlate neither with the age nor with the gender of sample analyzed.
https://static-content.springer.com/image/art%3A10.1007%2Fs11357-011-9216-6/MediaObjects/11357_2011_9216_Fig3_HTML.gif
Fig. 3

Frequency distribution of GDMI values in the total population sample

Subsequently, we wondered whether the GDMI values were correlated to the frailty status rather than to chronologic age. To answer this question, we availed of the HCA classifications reported in Montesanto et al. (2010), that allowed to classify this sample in different aging phenotypes (see “Materials and methods”). The mean GDMI values across the S1 and S2 groups are shown in Figs. 4 and 5. We can observe that frail subjects of S1 group exhibit GDMI values significantly higher than those prefrail (0.658 ± 0.201 vs 0.508 ± 0.223, respectively, p value = 0.006) and nonfrail subjects (0.658 ± 0.201 vs 0.521 ± 0.196, respectively, p value = 0.006). In S2 group no difference in GDMI values was detected across the frailty phenotypes (0.484 ± 0.191 and 0.509 ± 0.197 for very frail and frail, respectively). Moreover, GDMI values were quite similar for men and women in both groups (in S1 sample 0.534 ± 0.220 vs 0.522 ± 0.210, respectively; p value = 0.668; in S2 sample 0.514 ± 0.217 vs 0.477 ± 0.167, respectively; p value = 0.335). These results indicated that a correlation between the global DNA methylation levels and the frailty phenotype exists in middle-aged subjects, but not in ultranonagenarians.
https://static-content.springer.com/image/art%3A10.1007%2Fs11357-011-9216-6/MediaObjects/11357_2011_9216_Fig4_HTML.gif
Fig. 4

Mean GDMI values across the groups defined by cluster analysis in S1 sample

https://static-content.springer.com/image/art%3A10.1007%2Fs11357-011-9216-6/MediaObjects/11357_2011_9216_Fig5_HTML.gif
Fig. 5

Mean GDMI values across the groups defined by cluster analysis in S2 sample

Then, in order to better evaluate the relationship between DNA methylation levels and degree of frailty, 37 prefrail and nonfrail subjects of S1 sample were revisited after 7 years from the baseline visit. Figure 6 shows the GDMI values at baseline (black bars) and after the follow-up period (gray bars) with respect to the changes in the frailty status after this period. We can observe that in subjects who, after the follow-up period, have maintained their nonfrail or prefail frailty status or have changed their frailty status from nonfrail to prefrail, mean GDMI value did not show significant changes over time (about 0.4). On the contrary, in subjects who became frail, mean GDMI value was significantly increased (about 0.6) over time compared to the first measurement.
https://static-content.springer.com/image/art%3A10.1007%2Fs11357-011-9216-6/MediaObjects/11357_2011_9216_Fig6_HTML.gif
Fig. 6

Variations of GDMI values with respect to the variations of the frailty status after the follow-up period

Discussion

Understanding the mechanisms that modulate the quality of aging remains one the most challenging research topics. Several lines of evidence have demonstrated how the characterization of frailty, that represents a state of vulnerability for adverse health outcomes, may contribute to disentangle the molecular mechanisms influencing the functional decline of elderly people and thus to characterize and to better define the aging process (Fried et al. 2004). The impact of genetic variants of both nuclear and mitochondrial DNA on the inter-individual susceptibility to functional decline and vulnerability to diseases in the elderly people has been largely demonstrated (Maggio et al. 2006; Moore et al. 2010; Matteini et al. 2010). Similarly, different reports have shown the influence of environmental and social factors on frailty. A “drawbridge” across genetic factors and environment may be represented by epigenetic variations which in turn depend on hereditary, environmental and stochastic factors and might explain the inter-individual variability in the frailty status (Sutherland and Costa 2003; Fraga 2009; Schneider et al. 2010). Indeed, most studies on DNA methylation demonstrated that aging is associated with a relaxation of epigenetic control, thus contributing to some age-related phenotypes such as functional and cognitive decline and the development of pathologies.

In the present study, we looked at epigenetic changes occurring in blood collected from individuals classified by a precise frailty phenotype based on the individual degree of cognitive, functional, and psychological status (Montesanto et al. 2010). We availed of a method that, although could not provide precise quantitative measurements of the DNA methylation levels, is highly effective in comparative analyses of these levels among different samples. In addition, as discussed by Anisowicz et al. (2008), this method is an excellent alternative to the different “gold standard technologies” such as the methods based on chromatography, the high-performance capillary electrophoresis, the radio-labeling of CpG sites using M.SssI methyltransferase, methyl-C antibody, and pyrosequencing. In fact, this method requires small amount of genomic DNA, it is not radioactive, it is highly accurate and reproducible and allow to measure global DNA methylation in many samples simultaneously. It is also worth mentioning that the use of DNA from peripheral blood can be effective in global DNA methylation studies, although this analysis misses the important effects of tissue specific methylation (Thompson et al. 2010).

As the study deals with 65–85-year-old subjects, we found no correlation between global DNA methylation levels and age but frail individuals (with high vulnerability) exhibit lower global methylation levels than those that are in a better condition (prefrail and nonfrail). On the contrary, no difference was observed in these levels among ultranonagenarians, despite their differential classification in frail and very frail subjects. In this paper, we reported also the results obtained in a longitudinal study where DNA methylation levels and frailty status were reanalyzed after 7 years of follow-up. Methylation levels were unchanged in subjects who after the follow-up period did not became frail. On the contrary, in subjects who were prefrail at baseline and became frail after the follow-up period, these levels were significantly decreased and were comparable to those observed in frail subjects at baseline.

To our knowledge, this is the first study demonstrating as a correlation between global methylation levels and frailty occurs in aging. Therefore, the loss of global DNA methylation seems to be specifically associated to the functional decline and not to chronological age, at least after 65 years of age. It might be worth noticing that frailty is associated with age, and then it is possible that in some cases an association between age and methylation is observed as a consequence of the association between frailty and methylation. This result is particularly interesting because it could provide a re-interpretation of some studies in which variations in methylation levels were not observed in samples of middle age individuals (Tra et al. 2002; Siegmund et al. 2007; Schneider et al. 2010; Agrawal et al. 2010).

We did not observe a link between DNA methylation levels and gender, although several studies found higher global DNA methylation levels in males regardless to age of the subjects analyzed (Fuke et al. 2004; Shimabukuro et al. 2006). However, as in the case of the association between age and global methylation, gender-associated differences in DNA methylation were restricted to specific loci and often to mitotic cells (El-Maarri et al. 2007). It might also be worth mentioning that no correlation between the use of drugs and methylation has been observed (data not shown).

Since methylation is associated to gene silencing, it is possible that in old frail subjects, a reduction in DNA methylation could activate genes being in a repressed status. Some of these genes might be needed to maintain the reserve capacity and the efficiency in the response to adverse stressors at levels ensuring the survival (Coneyworth et al. 2009). Likely, the hypomethylation observed in frail subjects can be associated to an accentuated deficiency to achieve the remethylation after DNA replication that, as reported in Gravina and Vijg (2010), occurs during aging. In addition, the above hypomethylation could derepress silenced retrotransposons, such as L1 or IAP, thus inducing genome instability that might account for the decline typical of frailty status (Barbot et al. 2002; Menendez et al. 2004; Maslov and Vijg 2009). Moreover, longitudinal data provide an important contribution to studies which consider frailty as a dynamic process. In fact, our findings are in line with those reported by Bjornsson et al. (2008) who observed time-dependent changes in global DNA methylation within the same individual in two separate populations (one from Iceland and one from Utah) exhibiting both decrease and increase in methylation greater than 20% over an 11- to 16-year span. A progressive loss of DNA methylation over time in repetitive elements disperses through the genome, particularly in Alu elements, was also reported by Bollati et al. (2009).

On the whole, our results contribute to understand how changes in methylation levels are associated with the aging process. We are not able to know whether the changes in frailty status induce the changes in methylation levels or, vice versa, whether these levels determine the frailty status. In any case, our results suggest, in agreement with a series of previous evidences, that specific age-related phenotypes are modulated not only by genetic and environmental factors but also by the effects of environmental factors (dietary, life style) on the individual genome.

Acknowledgments

The work was supported by Fondi di Ateneo of the University of Calabria. AM was supported financially by a fellowship (assegno di ricerca) of the University of Calabria.

Copyright information

© American Aging Association 2011