Caloric restriction delays age-related methylation drift

In mammals, caloric restriction consistently results in extended lifespan. Epigenetic information encoded by DNA methylation is tightly regulated, but shows a striking drift associated with age that includes both gains and losses of DNA methylation at various sites. Here, we report that epigenetic drift is conserved across species and the rate of drift correlates with lifespan when comparing mice, rhesus monkeys, and humans. Twenty-two to 30-year-old rhesus monkeys exposed to 30% caloric restriction since 7–14 years of age showed attenuation of age-related methylation drift compared to ad libitum-fed controls such that their blood methylation age appeared 7 years younger than their chronologic age. Even more pronounced effects were seen in 2.7–3.2-year-old mice exposed to 40% caloric restriction starting at 0.3 years of age. The effects of caloric restriction on DNA methylation were detectable across different tissues and correlated with gene expression. We propose that epigenetic drift is a determinant of lifespan in mammals.


Supplementary Figure 2 | A limited effect of blood composition on age-related methylation.
Areaproportional Venn diagrams of overlapped CpG sites between sites showing age-related methylation drift in whole blood and sites identified as differentially methylated in blood cell subtypes compared to whole blood. Red number represents the number of sites overlapping. Balloon shows the percentage of overlapped sites of age-related sites in whole blood. To identify the differentially methylated sites between whole blood and each blood cell type, we used sites with sequencing depth ≥ 100 reads in DREAM data among samples. Then, we compared the average of methylation % between the whole blood (n=16; age; 0-86y) and each blood cell type (granulocytes; n=6; age; unknown, CD34+ cells; n=2; age; unknown, T-cells; n=3; age; 19-21y) in each site and defined sites with methylation differences ≥2% (FDR<0.05) as differentially methylated sites. A Chi-square test using 2X2 tables (Supplementary Table 9) was used to calculate p-values for the significance of the overlaps. WB; Whole blood, Gra; Granulocytes, Age; Age-related sites in WB (both hyper and hypo-methylation drifts), Diff; Differentially methylated sites between whole blood and each cell sub-type. 12 Supplementary Figure 6 | Age-related genes in mouse, monkey and human. Association of the percentages of methylated cytosines in the samples as obtained from bisulfite pyrosequencing (y-axis) with age (x-axis) (a; mice, b; monkeys, and c; humans). Each dot corresponds to one individual (mouse; young, n=6, middle, n=13, old, n=12, monkey; infant, n=12, middle, n=15, old, n=12, human; newborn, n=13, young, n=54, middle, n=27, old, n=45). Spearman r values and p-values (two-tailed) were listed in Supplementary Tables 14-16. Principal component analysis was performed on the DNA methylation data for CpG sites in all genomic regions (CGI + nCGI) (sequencing depth ≥100 reads in 75% of samples). Samples are plotted using the first two principal components (PC). The color codes of samples are shown on the right. Number of individuals (n) was as follows: mouse; AL-young, n=5, AL-middle, n=7, AL-old, n=7, CRold, n=5, monkey; AL-infant, n=4, AL-middle, n=7, AL-old, n=5, CR-old, n=6.

Supplementary
Supplementary Figure 10 | Clustering analyses of age-related hypermethylated and age-related hypomethylated CpG sites in mice and monkeys. Only CpG sites that showed age-related methylation were clustered. A total of top 100, 30, 100 and 500 sites were used that showed age-related hyper-or hypo-methylation in old animals compared with that of young age animals in average DNA methylation values by DREAM assay in promoter regions and non-promoter regions, respectively. The green to red scales indicate the methylation percentage. Shown in figures next to clustering are plots for the average methylation values in each animal group. The bar in the graphs represents the median. Figure 11 | Overlap between genes affected by CR and genes showing age-related methylation. Coefficient negative means that CR tends to decrease aging methylation drift and vice versa. To test if CR tends to decrease or increase aging methylation drift, binomial test with probability 0.5 was performed for each bar. The p-values are smaller than 0.01 for all the bars.

21
Supplementary Figure 12 | CR effects on aging drift detected by bisulfite pyrosequencing assay. DNA methylation profiles in AL and CR animals (a; mice, and b; monkeys). The averaged data were derived from the methylation data of hypermethylated genes (n=24 in mice, n=24 in monkeys) and hypomethylated genes (n=10 in mice, n=12 in monkeys) listed in Supplementary Tables 14 and 15. Each dot corresponds to one individual (mouse; AL-young, n=6, AL-middle, n=13, AL-old, n=12, CR-old, n=12, monkey; AL-infant, n=12, AL-middle, n=15, AL-old, n=12, CR-old, n=18). The bar in the graphs represents the median. p-values were obtained using the unpaired t-test with Welch's correction. Figure 13 | Correlation between changes in methylation by CR and changes in methylation by age. The x-axis is methylation changes per year in AL fed animals. Positive/negative value means methylation increases/decreases with age, respectively. The y-axis is the differences of methylation percentage between CR-old and AL-old animals. Each dot represents an averaged value of DNA methylation status detected by each bisulfite pyrosequencing assay. Spearman's rank correlation coefficient (r) and corresponding two-tailed p-value were calculated by GraphPad.

Supplementary Figure 14 | CR effects on aging drift detected by bisulfite pyrosequencing assay in multiple tissues.
Unsupervised hierarchical clustering analysis of the 12 loci assayed. The green to red color scale indicates the methylation percentage. The color codes of caloric status and age are shown on the bottom-right. Correlation between changes in methylation by CR and changes in methylation by age are shown in the lower-left of the clustering. The x-axis is methylation changes per year in AL fed animals. Positive/negative value means methylation increases/decreases with age, respectively. The yaxis is the differences of methylation percentage between CR-old and AL-old animals. Each dot represents an average value of DNA methylation status detected by each bisulfite pyrosequencing assay. Spearman's rank correlation coefficient (r) and corresponding two-tailed p-value were calculated by GraphPad. Differences between the predicted ages and chronological ages in CR-old animals are shown in the lower-right of the clustering. Each pair of dots connected by a line represents the difference between predicted age and chronological age in each individual. Paired t-test was used for p-value calculation.
Supplementary Figure 15 | Correlation between telomere length and age. Relative telomere length was shown in mice, monkeys and humans. Relative telomere length declined with age in AL animals. There was no statistical evidence of a CR effect in telomere shortening. Two primer sets were used to amplify telomeres by quantitative PCR. The values are presented as means of triplicate determinations (PCR reactions). Bars represent standard error. p-values were obtained using the unpaired t-test with Welch's correction.

Supplementary Figure 16 | Correlation between DNA methylation and BMI in monkey.
The association between Body Mass Index (BMI) (x-axis) and average methylation percentage of 24 hypermethylated ARM genes (y-axis). Each dot corresponds to one individual. Black and red dots represent AL and CR animals, respectively. Correlation between methylation and BMI was calculated using Spearman's correlation (r). r-values and the corresponding two-tailed p-values were calculated by GraphPad.
Supplementary Figure 17 | Seven technical replicates of human whole blood sample in DREAM analysis. Correlation matrix of Pearson r and scatter plots of DNA methylation levels in all genomic regions (sequencing depth ≥100 reads, 25,780 sites). The lowess regression line is shown in red. Correlations are significant at p<0.001 (two-tailed).   Coefficient calculated by the linear model shows the effect of CR on aging methylation drift t-test for coefficient; p<0.05 promoter (-1kb<TSS<+500bp) average of methylation≥1% Negative (positive) coefficient means that CR tends to decrease (increase) aging methylation drift. All genes, except 7 genes, with negative (positive) coefficient showed age-related hypermethylation (hypomethylation). *; gene with negative coefficient showed age-related hypomethylation. **; gene with positive coefficient showed age-related hypermethylation. ***; gene with positive coefficient showed both age-related hypermethylation and hypomethylation. ****; non age-related gene Binomial pvalue; p<0.001 (in each species)  TCCACCCACACATTCCTA  AAA  109  60  GGTAGTTAAAGGGAATA  AAT  2  GAYGGTTTTGTTYGGGGTAGTAGGGA  GTGG   H-RILP-py1-F  TGGAGTTATGGGGTTAATT  TGGTA *   TCCAACCCAACCCTTTAA  ACTTT  159  60  GGGTTAATTTGGTAGTG  GT  2  TYGTTTTAGGYGAGAGGTGATAGGTAG  GTAG We used UNIV-reverse primer and Biotin-UNIV primer for 2nd PCR. Annealing temperature is same as 1st PCR. *; forward and UNIV-reverse primer for 1st PCR, forward and Biotin-UNIV primer for 2nd PCR. YG; CpG site for pyrosequencing analysis.