Comparative analysis of epigenetic aging clocks from CpG characteristics to functional associations

Abstract

To date, a number of epigenetic clocks have been developed using DNA methylation data, aimed at approximating biological aging in multiple tissues/cells. However, despite the assumption that these clocks are meant to capture the same phenomenon-aging, their correlations with each other are weak, and there is a lack of consistency in their associations with outcomes of aging. Therefore, the goal of this study was to compare and contrast the molecular characteristics and functional associations of 11 existing epigenetic clocks, using data from diverse human tissue and cell types. Results suggest that the CpGs comprised in the various clocks differ in regards to the consistency of their age correlations across tissues/cells. Using microarray expression data from purified CD14+ monocytes, we found that six clocks--Yang, Hannum, Lin, Levine, Horvath1, and Horvath2--has relatively similar transcriptional profiles. Network analysis revealed nine co-expression modules, most of which display robust correlations across various clocks. One significant module--turquoise is involved in mitochondrial translation, gene expression, respiratory chain complex assembly, and oxidative phosphorylation. Finally, using data from 143B cells with chronically depleted mtDNA (rho0) and 143B controls, we found that rho0 cells have more than a three-standard deviation increase in epigenetic age for Levine (p=0.006), Lin (p=0.012), and Yang (p=0.013). In summary, these results demonstrate the shared and contrasting features of existing epigenetic clocks, in regards to the CpG characteristic, tissue specificity, and co-regulatory gene network signatures, and suggesting a link between two hallmarks of aging--epigenetic alterations and mitochondrial dysfunction.