An epigenetic clock is a weighted algorithmic combination of the methylation status of various specific CpG sites on the genome, where the number produced matches well with chronological or biological age. It is constructed by applying machine learning techniques to epigenetic data, a catalogue of DNA methylation patterns across the full structure of the genome, taken from blood samples of mice or people at various ages. The early epigenetic clocks use hundreds of CpG sites, and therefore one might reasonably hypothesize that they reflect the entire breadth of the processes of aging. That turns out not to be true, but it was reasonable.
As work on the production of epigenetic clocks progressed, as well as the establishment of other clocks based on transcriptomic, proteomic, and other data, many groups have found highly optimized clocks that use very few CpG sites, less than a dozen. Further, it is possible to identify sites, when using such small numbers, that work the same way in mice, humans, and other mammalian species. This is desirable in one sense, in that such clocks are less costly to implement, particularly at scale. On the other hand, it is not reasonable to think that such clocks will reflect more than a tiny fraction of the processes of aging.
If all one cares about is to measure biological age absent intervention, then it really doesn’t matter whether a clock measures only one or only a few of the underlying processes or dysfunctions of age. Absent intervention, all of the processes of aging proceed in parallel, so measuring just one or just a few is good enough. However, it is the case that every new approach to rejuvenation therapy will address only the one target mechanism, a limited portion of the contributions to degenerative aging. It is entirely plausible that an epigenetic clock will underestimate or overestimate the utility of a potential rejuvenation therapy, and the plausibility of that outcome increases as the number of CpG sites decreases. The most important future use for epigenetic and other clocks is to steer research and development towards larger effect sizes, more effective approaches to human rejuvenation. But we are not there yet.
Precise measurement of aging is a prerequisite to identify parameters that may attenuate the aging process. It is fascinating that the DNA methylation (DNAm) patterns change in a highly reproducible and seemingly organized manner during aging of the organism. This epigenetic modification at the cytosine residues of CG dinucleotides (CpGs) impacts on chromatin organization, transcription factor binding, and gene expression. It is therefore anticipated that age-associated DNAm might be of immediate functional relevance for the aging process, albeit this remains to be proven. Today, epigenetic clocks are considered to be the most accurate biomarker for age predictions and there is sound evidence that they also capture aspects of biological aging that are independent from chronological age.
In this study, we used the recently released Infinium Mouse Methylation BeadChip to compare such epigenetic modifications in C57BL/6 (B6) and DBA/2J (DBA) mice. We observed marked differences in age-associated DNA methylation in these commonly used inbred mouse strains, indicating that epigenetic clocks for one strain cannot be simply applied to other strains without further verification. Interestingly, the CpGs with highest age-correlation were still overlapping in B6 and DBA mice and included the genes Hsf4, Prima1, Aspa, and Wnt3a. Furthermore, Hsf4, Aspa, and Wnt3a revealed highly significant age-associated DNA methylation in the homologous regions in human. Subsequently, we used pyrosequencing of the four relevant regions to establish a targeted epigenetic clock that provided very high correlation with chronological age in independent cohorts of B6 and DBA.
Larger signatures that comprise hundreds of CpGs may be more robust than targeted assays that only consider one or few CpGs, since they reflect a broader epigenetic pattern. BeadChip technology makes large signatures easily applicable since all relevant CpGs are addressed in each sample. However, adaptation and integration of different microarray datasets remains a major hurdle and age-predictors may become outdated if a BeadChip release is discontinued. It may therefore be advantageous to rather focus on individual CpGs by targeted methods, such as pyrosequencing, digital droplet PCR, or barcoded amplicon sequencing. These methods give very precise and reproducible results on single CpG level and facilitate fast and more cost-effective analysis. Notably, all 21 CpGs covered by our pyrosequencing assay provided very high correlation with age in all training and validation cohorts. Our four CpG epigenetic age prediction model thus now outperforms our previously published three CpG signatures. Other methods for age prediction in mice have reported lower correlations with a higher number of CpG sites (9-582).