Graphical abstract

   [47]graphic file with name ga1.jpg
   [48]Open in a new tab

   Keywords: Aging, Transcriptome, Aging clock, Rejuvenation

Abstract

   Although aging is an increasingly severe healthy, economic, and social
   global problem, it is far from well-modeling aging due to the aging
   process’s complexity. To promote the aging modeling, here we did the
   quantitative measurement based on aging blood transcriptome.
   Specifically, the aging blood transcriptome landscape was constructed
   through ensemble modeling in a cohort of 505 people, and 1138
   age-related genes were identified. To assess the aging rate in the
   linear dimension of aging, we constructed a simplified linear aging
   clock, which distinguished fast-aging and slow-aging populations and
   showed the differences in the composition of immune cells. Meanwhile,
   the non-linear dimension of aging revealed the transcriptome
   fluctuations with a crest around the age of 40 and showed that this
   crest came earlier and was more vigorous in the fast-aging population.
   Moreover, the aging clock was applied to evaluate the rejuvenation
   effect of molecules in vitro, such as Nicotinamide Mononucleotide (NMN)
   and Metformin. In sum, this study developed a de novo aging clock to
   evaluate age-dependent precise medicine by revealing its fluctuation
   nature based on comprehensively mining the aging blood transcriptome,
   promoting the development of personal aging monitoring and anti-aging
   therapies.

1. Introduction

   Life expectancy has increased dramatically in the past 150 years. It is
   expected that worldwide 1.5 billion people aged 65 years or over will
   outnumber adolescents and the youth aged 15 to 24 years (1.3 billion)
   in 2050 [49][1]. People aged 65 years and older are experiencing the
   aging process, characterized by progressive impairment and loss of
   physiological integrity and function, leading to an increased
   vulnerability to death [50][2]. Therefore, the world is facing an aging
   challenge.

   Aging, a complex biological process, is far from well modeled though
   significant efforts have been put into understanding the aging process
   and revealing patterns in immune-aging [51][3] and inflammatory-aging
   [52][4] perspectives. It is commonly acknowledged that aging clock is a
   method to predict an individual's age using aging biomarkers [53][5].
   Such aging biomarkers include telomere length, genomic instability,
   epigenetic marks, biochemical compounds, gene expression levels, etc
   [54][5]. Until now, ‘Omics’ technologies (e.g., genomics, metabolomics,
   metagenomics, proteomics, and transcriptomics) have been widely applied
   to investigate and model the aging process [55][6]. Among these Omics,
   transcriptomics by RNA sequencing is a mature and relatively low-cost
   omics technology and has already been in clinical use [56][7]. In
   addition, transcriptome-based aging clocks, including the analyses of
   peripheral blood mononuclear cells (PBMCs) [57][8], muscle [58][9], and
   dermal fibroblast [59][10], are high in interpretability without
   compromising accuracy [60][11] compared with other aging clocks.
   Recently, an ultra-predictive aging clock built on 31 batches and 3060
   samples was introduced. The model was used to explore sex, health
   status, and other factors in the aging process [61][12]. Most studies
   modeled aging as a static linear process [62][8], [63][9], [64][10],
   fail to model it as a dynamic process [65][13]. Given that recent
   studies have shown the diversified early aging signs or pace [66][14]
   at middle age and the fluctuation in plasma protein level [67][13],
   examining the transcriptome changes of blood samples in midlife can
   help investigate and model the aging process.

   In the search for anti-aging interventions and drugs, a quantitative
   measurement of sample biological age before and after intervention
   cannot be achieved without accurate modeling. RNA-based aging clock has
   been applied to population-based drug screening for geroprotectors. In
   a study, an age classifier was built on the transcriptome of the young
   and old population, and a simulated drug-response transcriptome was
   passed to the classifier to rank the drug’s anti-aging effect [68][15].
   There is room for improvement in the application of transcriptome-based
   aging clock in personal geroprotectors assessment and drug screening.
   Therefore, an accurate and applicable transcriptome-based aging clock
   is required. This study aims to construct the aging trajectories using
   blood transcriptomics and successfully developed a new aging clock
   capable of reflexing the linear and dynamic changes with high accuracy
   using ensemble modeling. Moreover, we investigated the possibility of
   using the new aging clock to screen rejuvenation treatments.

2. Results

2.1. Trajectories of aging gene expression form functional modules

   To dissect the transcriptome landscape of the aging process, we did the
   HiseqX sequencing on blood samples from a cohort of 505 volunteers. All
   participants were self-assessed healthy and free of major diseases when
   the participants came for regular health checks. The cohort includes
   208 male and 297 female participants with an age range from 18 to 68,
   with a median of 36 ([69]Fig. S1-A). First, we grouped genes with
   similar trajectories by unsupervised hierarchical clustering to
   identify the changing pattern of age-related genes. Eight modules were
   identified, of which five (Clusters 1–5) showed an upward trend, and
   Clusters 6–8 had downward patterns ([70]Fig. 1-A, B). As visualized in
   trajectory bundles ([71]Fig. 1-B), some patterns were generally linear,
   but others were non-linear. In some of the modules (Clusters 5–8), gene
   expressions changed steadily, while other trajectories indicated
   dramatic changes in a specific age range. Gene Ontology [72][16] (GO)
   Enrichment analysis was then conducted to infer its related biological
   function. The dot plot showed the top enriched GO terms in each module
   ([73]Fig. 1-C, [74]Supplement table 1). The first module expression was
   enhanced at the age of 25–35, and its genes are related to ubiquitin
   activity and immune cell proliferation. The second module was
   wave-like, and the related genes in this module regulate transcription
   factor complex and interleukin-8 secretion. The age of 45 is the
   boosting point for the third module expression, of whose genes were
   associated with mitochondria activity. The expression of the fourth and
   fifth modules, including the genes enriched in neutrophil immune
   activity, was increased at the age of 35–45. The other three modules
   (Clusters 6–8) with downward trends were mainly involved in
   translation, and the top terms were protein targeting to membrane, RNA
   helicase activity, and viral translation, respectively. These
   biological processes, enriched in these modules, correspond to previous
   studies of ubiquitin [75][17], immune cell [76][18], mitochondria
   [77][19], ribosome [78][20] in aging. In sum, we mapped the
   trajectories of the expression patterns of aging-related gene
   expression.

Fig. 1.

   [79]Fig. 1
   [80]Open in a new tab

   Trajectories of gene expression throughout age form functional modules.
   (A) Hierarchical clustering of gene expression trajectories. A red box
   highlighted each cluster. (B) Eight clusters of the aging pattern (five
   up-regulated, three down-regulated, respectively). Redline was
   indicating the fitting curve created by generalized additive
   model(GAM). (C) Top GO terms in which each aging pattern is involved.
   Color showing the p-adjusted value of enrichment analysis. Dot size
   showing the number of genes hit by GO terms. (For interpretation of the
   references to color in this figure legend, the reader is referred to