Epigenetic age oscillates during the day

Abstract Since their introduction, epigenetic clocks have been extensively used in aging, human disease, and rejuvenation studies. In this article, we report an intriguing pattern: epigenetic age predictions display a 24‐h periodicity. We tested a circadian blood sample collection using 17 epigenetic clocks addressing different aspects of aging. Thirteen clocks exhibited significant oscillations with the youngest and oldest age estimates around midnight and noon, respectively. In addition, daily oscillations were consistent with the changes of epigenetic age across different times of day observed in an independant populational dataset. While these oscillations can in part be attributed to variations in white blood cell type composition, cell count correction methods might not fully resolve the issue. Furthermore, some epigenetic clocks exhibited 24‐h periodicity even in the purified fraction of neutrophils pointing at plausible contributions of intracellular epigenomic oscillations. Evidence for circadian variation in epigenetic clocks emphasizes the importance of the time‐of‐day for obtaining accurate estimates of epigenetic age.


Figure S2
Figure S2 Cell type composition of WBC-Neu samples.Barplot of WBC subtype proportions estimated using Houseman's method (Houseman et al. 2012) with "enhanced cell deconvolution" referenced by Salas et al. (Salas et al. 2022).Colors represent different WBC subtypes.x-axis: estimated cell type proportion in percentages; y-axis: samples arranged by collection time.

Figure S3
Figure S3 Epigenetic age oscillations for PC-based clocks.(a-c) 24 hr dynamics of epigenetic age predictions in WBC-Neu dataset for PC Horvath pan-tissue (a), PC DNAmTL (b), and PC GrimAge (c) clocks.Lines depict cosinor regression fits with solid lines indicating oscillation significance (cosinor p < 0.05).x-axis: time of day; y-axis: epigenetic age.(d) Estimated oscillation parameters of the 6 tested epigenetic clocks.Bold values indicate significant oscillations (p < 0.05), and asterisks indicate significance after Bonferroni correction for multiple testing.

Figure S4
Figure S4 Time of day effects for PC-based clocks.Boxplots of epigenetic age predictions at two time points: 12:45 and 16:15 (Apsley et al. 2023) for PC Horvath pan-tissue (e), PC DNAmTL (f), and PC GrimAge (g) clocks.Solid median lines indicate statistically significant differences (p < 0.05, paired Student's t-test).x-axis: time of day; y-axis: epigenetic age.(h) Paired Student's t-test p-values and pairwise mean differences of epigenetic age between 12:45 and 16:15, matched by donor.

Figure S5
Figure S5 Predictions made by chronological epigenetic clocks for different WBC subtypes.Violin plots of epigenetic age differences between WBC subtypes and matched whole blood samples in Wang et al. (Wang et al. 2023) (left) and Reinius et al. (Reinius et al. 2012) (right) datasets.Numbers above violin plots depict p-values (bold -p < 0.05) of Student's paired t-test between each WBC subtype compared to whole blood, matched by donor.x-axis: WBC subtype; y-axis: epigenetic age deviation of WBC subtype from whole blood.

Figure S6
Figure S6 Predictions made by mitotic epigenetic clocks for different WBC subtypes.Violin plots of epigenetic age differences between WBC subtypes and matched whole blood samples in Wang et al. (Wang et al. 2023) (left) and Reinius et al. (Reinius et al. 2012) (right) datasets.Numbers above violin plots depict p-values (bold -p < 0.05) of Student's paired t-test between each WBC subtype compared to whole blood, matched by donor.x-axis: WBC subtype; y-axis: epigenetic age deviation of WBC subtype from whole blood.

Figure S7
Figure S7 Predictions of biological and mortality epigenetic clocks across different WBC subtypes.Violin plots of epigenetic age differences between WBC subtypes and matched whole blood samples in Wang et al. (Wang et al. 2023) (left) and Reinius et al. (Reinius et al. 2012) (right) datasets.Numbers above violin plots depict p-values (bold -p < 0.05) of Student's paired t-test between each WBC subtype compared to whole blood, matched by donor.x-axis: WBC subtype; y-axis: epigenetic age deviation of WBC subtype from whole blood.

Figure S8
Figure S8 Epigenetic age oscillations after adjustment for WBC subtype proportions.(a-c) 24 hr dynamics of epigenetic age predictions in WBC-Neu dataset regressed on chronological age and cell type proportions (see methods) for Horvath pan-tissue 2013 (a), Teschendorff epiTOC2 2020 (b), and Lu GrimAge2 2022 (c) clocks.Lines depict cosinor regression fits with solid lines indicating oscillation significance (p < 0.05).x-axis: time of day; y-axis: epigenetic age.(d) Estimated oscillation parameters of the 17 tested epigenetic clocks after adjustment for WBC subtypes.