In recent years, reports of non-linear regulations in age-and longevity-associated biological processes have been accumulating. Inspired by methodological advances in precision medicine involving the integrative analysis of multi-omics data, we sought to investigate the potential of multi-omics integration to identify distinct stages in the aging progression from ex vivo human skin tissue. For this we generated transcriptome and methylome profiling data from suction blister lesions of female subjects between 21 and 76 years, which were integrated using a network fusion approach. Unsupervised cluster analysis on the combined network identified four distinct subgroupings exhibiting a significant age-association. As indicated by DNAm age analysis and Hallmark of Aging enrichment signals, the stages captured the biological aging state more clearly than a mere grouping by chronological age and could further be recovered in a longitudinal validation cohort with high stability. Characterization of the biological processes driving the phases using machine learning enabled a datadriven reconstruction of the order of Hallmark of Aging manifestation. Finally, we investigated non-linearities in the mid-life aging progression captured by the aging phases and identified a far-reaching non-linear increase in transcriptional noise in the pathway landscape in the transition from mid-to late-life.

The development of ‘age clocks’, machine learning models predicting age from biological data, has been a major milestone in the search for reliable markers of biological age and has since become an invaluable tool in aging research. However, beyond their unquestionable utility, current clocks offer little insight into the molecular biological processes driving aging, and their inner workings often remain non-transparent. Here we propose a new type of age clock, one that couples predictivity with interpretability of the underlying biology, achieved through the incorporation of prior knowledge into the model design. The clock, an artificial neural network constructed according to well-described biological pathways, allows the prediction of age from gene expression data of skin tissue with high accuracy, while at the same time capturing and revealing aging states of the pathways driving the prediction. The model recapitulates known associations of aging gene knockdowns in simulation experiments and demonstrates its utility in deciphering the main pathways by which accelerated aging conditions such as Hutchinson–Gilford progeria syndrome, as well as pro-longevity interventions like caloric restriction, exert their effects.

The dermal sheath (DS) is a population of mesenchyme-derived skin cells with emerging importance for skin homeostasis. The DS includes hair follicle dermal stem cells, which exhibit self-renewal and serve as bipotent progenitors of dermal papilla (DP) cells and DS cells. Upon aging, stem cells exhibit deficiencies in self-renewal and their number is reduced. While the DS of mice has been examined in considerable detail, our knowledge of the human DS, the pathways contributing to its self-renewal and differentiation capacity and potential paracrine effects important for tissue regeneration and aging is very limited. Using single-cell RNA sequencing of human skin biopsies from donors of different ages we have now analyzed the transcriptome of 72,048 cells, including 50,149 fibroblasts. Our results show that DS cells that exhibit stem cell characteristics were lost upon aging. We further show that HES1, COL11A1, MYL4 and CTNNB1 regulate DS stem cell characteristics. Finally, the DS secreted protein Activin A showed paracrine effects on keratinocytes and dermal fibroblasts, promoting proliferation, epidermal thickness and pro-collagen production. Our work provides a detailed description of human DS identity on the single-cell level, its loss upon aging, its stem cell characteristics and its contribution to a juvenile skin phenotype.