Abstract

   The immune system undergoes progressive functional remodeling from
   neonatal stages to old age. Therefore, understanding how aging shapes
   immune cell function is vital for precise treatment of patients at
   different life stages. Here, we constructed the first transcriptomic
   atlas of immune cells encompassing human lifespan, ranging from
   newborns to supercentenarians, and comprehensively examined gene
   expression signatures involving cell signaling, metabolism,
   differentiation, and functions in all cell types to investigate immune
   aging changes. By comparing immune cell composition among different age
   groups, HLA highly expressing NK cells and CD83 positive B cells were
   identified with high percentages exclusively in the teenager (Tg)
   group, whereas unknown_T cells were exclusively enriched in the
   supercentenarian (Sc) group. Notably, we found that the biological age
   (BA) of pediatric COVID-19 patients with multisystem inflammatory
   syndrome accelerated aging according to their chronological age (CA).
   Besides, we proved that inflammatory shift- myeloid abundance and
   signature correlate with the progression of complications in Kawasaki
   disease (KD). The shift- myeloid signature was also found to be
   associated with KD treatment resistance, and effective therapies
   improve treatment outcomes by reducing this signaling. Finally, based
   on those age-related immune cell compositions, we developed a novel BA
   prediction model PHARE ([46]https://xiazlab.org/phare/), which can
   apply to both scRNA-seq and bulk RNA-seq data. Using this model, we
   found patients with coronary artery disease (CAD) also exhibit
   accelerated aging compared to healthy individuals. Overall, our study
   revealed changes in immune cell proportions and function associated
   with aging, both in health and disease, and provided a novel tool for
   successfully capturing features that accelerate or delay aging.

Supplementary Information

   The online version contains supplementary material available at
   10.1186/s12979-024-00479-4.

   Keywords: Immune aging, Biological age, Chronological age, Aging clock,
   Machine learning

Introduction

   Advanced age is often linked to increased morbidity and mortality from
   infectious diseases, and decreased vaccination efficacy [[47]1, [48]2].
   Aging is also a leading risk factor for the increased incidence of most
   cancer types [[49]3, [50]4]. Although most children develop mild and
   self-limiting symptoms of infectious diseases, such as COVID-19, a
   severe and delayed post-SARS-CoV-2 inflammatory response in children
   has been recognized worldwide, correlating with disease severity
   [[51]5, [52]6]. Moreover, compared with older individuals, children
   appear to have a more robust immune response during acute infections,
   whereas young cancer patients treated with immune checkpoint inhibitors
   often have poor outcomes [[53]7, [54]8]. The immune system exhibits
   highly age-specific features; however, the impact of age-related
   changes on different components of the immune system is not fully
   understood [[55]2]. Therefore, the characterization of immune
   populations and function among different age groups could provide
   valuable insights into mechanisms underlying disease development and
   inform more precise disease treatments in the future.

   Not everyone ages at the same rate. Usually, aging affects health and
   varies from person to person; therefore, it is not surprising that
   people with the same CA manifest diverse aging-related phenotypes
   [[56]9]. In this regard, a person's BA can often differ from his/her
   CA. Arrojo et al. revealed that most organs, even in one person, are
   also a mix of cells and proteins of vastly different ages, which depend
   on their rates of regeneration [[57]10]. The ability to accurately
   measure human aging from molecular profiles has practical implications
   in many fields, particularly in disease prevention and treatment
   [[58]11]. As expected, previous studies have developed many BA
   measurements that attempt to capture physiological changes during the
   aging process, such as DNA methylation, telomere length, and frailty
   [[59]9, [60]11, [61]12]. Meanwhile, there are tools have been developed
   for BA prediction based on blood transcriptome, given the ease of
   obtaining samples [[62]13, [63]14]. Considering that most immune
   diseases present strong age characteristics, the functional states of
   immune cells do not always match their CA in these diseases [[64]15].
   Therefore, effective measurements to comprehensively depict the age
   distribution of immune cells to accurately assess the impact of various
   diseases on the aging process are urgently needed.

   Single-cell RNA sequencing (scRNA-seq) is a powerful technology used
   for studying individual cells and delineating complex cell populations.
   Recently, scRNA-seq has been performed in several studies to profile
   the immune landscape of human peripheral blood mononuclear cells
   (PBMCs) at different ages [[65]16–[66]18]. However, most of these
   studies have primarily focused on certain age periods, rather than the
   entire lifespan from 0 to over 110 years old, and many studies only
   analyzed a single type of immune cell, such as T cells [[67]19,
   [68]20], B cells [[69]21], or myeloid cells [[70]22, [71]23]. Owing to
   the complexity of age spans and cell populations, the previous studies
   mentioned above limit our understanding of how immune profiles
   contribute to disease development from a systematic perspective. Thus,
   for the first time, we systematically analyzed all immune cells and
   their transcriptomic signatures in human PBMCs across different age
   groups encompassing the entire lifespan to establish an elaborate and
   aging-focused immune landscape.

   In the current study, we first integrated three public scRNA-seq
   datasets from 24 healthy individuals across different age groups.
   Through extensive analysis, we revealed dynamic changes in cell
   composition, signaling pathways, metabolism, differentiation, and
   functions of human PBMCs over the lifespan. Furthermore, we re-analyzed
   other independent PBMC data from COVID-19 and KD mapping to healthy
   reference data, and detected a functional shift: immune cells’ BA
   differed from their CA, which was associated with the progress of
   vaccine efficacy and complications, respectively. Hereafter, we
   enrolled the largest PBMCs scRNA-seq datasets from 343 healthy
   individuals (over 2 million cells), and constructed a Physiological Age
   Prediction (PHARE) model. Collectively, our study provides essential
   insights for the precise treatment of patients at different life
   stages, and for building a novel machine-learning model to predict the
   patients’ BA at both scRNA-seq and bulk RNA-seq levels.

Results

Depicting global features of human immune cell atlas over the lifespan

   To determine the effects of age on the immune system, we initially
   integrated three public scRNA-seq datasets of PBMCs from 24 healthy
   individuals across children (Cd), teenagers (Tg), adults (Ad), elders
   (Ed), and supercentenarians (Sc) (Fig. [72]1A). After quality control
   and filtering, we obtained high-quality single-cell transcriptomes from
   159,671 cells (Figures S1A and Supplementary Table 2). Then, the
   scRNA-seq data were normalized and used harmony to remove batch effects
   (Figures S1B). Based on an unbiased integrative analysis across all
   immune cells, 12 cell clusters (Figures S1C) across six main immune
   cell populations were annotated according to the most salient cell
   markers: myeloid cells (CD14^+LYZ^+VCAN^+), B cells
   (MS4A1^+CD79A^+CD79B^+), T cells (C3D^+CD3E^+CD3G^+), NK cells
   (KLRD1^+KLRB1^+KLRF1^+), platelets (PPBP^+GP9^+), and hematopoietic
   progenitor cells (Hpc, KIT^+CD34^+) (Fig. [73]1B, C, and Figures S1D).
   Immune cell compositions were compared among different age groups, and
   the results showed that all six immune cell types were present in each
   group, albeit in different proportions. To confirm this validity, the
   proportion of cells in each individual was analyzed, and no immune
   cells showed high degrees of interindividual heterogeneity (Fig.
   [74]1D). As shown in Fig. [75]1E, the composition of T cells
   significantly decreased, whereas that of NK and myeloid cells increased
   in old groups. Notably, the proportion of platelets gradually increased
   in old groups (Figures S1E), which is likely associated with an
   increased incidence of cardiovascular disease in these populations
   [[76]24]. The compositions of immune cells from the elderly displayed
   prominent differences because the diversities measured with Shannon
   equitability index were significantly higher than those in young groups
   (Fig. [77]1F). The results implied different aging processes in
   different individuals, and this disparity was amplified with aging.

Fig. 1.

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

   Depicting global features of human immune cell atlas over the lifespan.
   A Schematic diagram of the scRNA-Seq data collection, processing, and
   analysis design. B Dot plots showing representative signature gene
   expression in the main immune cell types. C UMAP projection of immune
   cell profiles in PBMC of different age groups. D Overview of data
   collection datasets, clinical age groups; quantification of main cell
   types per patient and color-coded by main cell type. E Box plot showing
   distribution of main cell types across different age groups. The P
   values are calculated with kruskal.test. F Bar plots showing the main
   immune cell types from different age groups (mean ± SD). Average
   diversity measured with the Shannon equitability index for each tissue
   is shown. Point size of dot plot shows the fraction of cells with
   non-zero expression

Pluripotency functions in innate immune subsets decline with aging

   Given that innate immune cells provide the first line of defense to
   control pathogenic infections and instruct the subsequent adaptive
   immune response, we analyzed the functional alterations in innate
   immune cells associated with aging. To classify each cell subpopulation
   in an unbiased manner, we re-clustered each cell lineage. Six cell
   clusters comprising five myeloid cell types with unique transcriptional
   features were revealed in the PBMCs of all age groups (Figures S2A, and
   2B), including two types of CD14^+ monocytes (CD14_Mono1 and
   CD14_Mono2), one group of CD16^+ monocytes (PCGR3A^+), and two types of
   dendritic cells: CD1C^+ cDC (conventional dendritic cell) and JCHAIN^+
   pDC (plasmacytoid dendritic cell) (Fig. [80]2A and B). Among the
   different age groups, we found that CD14_Mono1 was mainly enriched in
   the young groups (Cd, Tg and Ad), whereas CD14_Mono2 was markedly
   enriched in the elderly (Fig. [81]2C and Figures S2C). Besides, we
   included healthy samples from an independent PBMC bulk RNA-seq dataset
   ([82]GSE180081) for analysis. We found the CD14_Mono2 score of the
   elderly group (≥ 60) was significantly higher than that of the young
   group (< 60), while the CD14_Mono1 score was significantly lower (Fig.
   [83]2D). GSEA analysis of CD14_Mono1 and CD14_Mono2 DEGs found the
   former was enriched for migration- and response- related pathways,
   while the latter was enriched for epigenetic alterations and oxidative
   phosphorylation (Fig. [84]2E and Supplementary Table 4). Consistent
   with previous findings [[85]25], both cDC and pDC decreased with
   increasing age (Fig. [86]2C and Figures S2D). We further want to
   evaluated the functional changes of DC in the different age groups.
   Briefly, we extracted the highly expressed genes in DC subsets and
   clustered them based on their expression patterns across different age
   groups. We identified a total of 12 distinct gene modules and selected
   those that showed a gradual increase or decrease with age for further
   analysis. We found the genes involved in antigen presentation function
   of cDC gradually decreased with age, while the expression of splicing
   molecules increased (Fig. [87]2F and Supplementary Table 5). With
   aging, the metabolic pattern of pDC was remodeled, while the
   maintenance of protein homeostasis was progressively lost (Fig. [88]2F
   and Supplementary Table 5). To thoroughly explore the function of all
   myeloid subsets, we evaluated the module scores associated with
   well-defined signatures of inflammation regulation and immune
   activation (Methods). The results indicated that TNF and IL6 signaling
   were highly activated in CD14_Mono1 cells, while CD16_Mono cells showed
   high IFN-induced signaling (Figures S2E). For DC populations, pDC
   displayed strong protein secretion ability, whereas cDC had high MHC II
   expression and antigen processing potency (Figures S2F), which was in
   agreement with their respective roles. In addition, unlike CD14^+
   monocytes, the proportion of CD16^+ monocytes did not change
   substantially among different age groups (Fig. [89]2C). However, CD16^+
   monocytes in the Tg group exhibited the highest levels of inflammation,
   IL6 and innate receptor signaling (Figures S2G). Taken together, the
   functions of monocytes and DCs gradually declined with aging.

Fig. 2.

   [90]Fig. 2
   [91]Open in a new tab

   Pluripotency of functions in innate subsets declines with age. A UMAP
   projection of myeloid profiles in PBMC of different age groups. B Dot
   plots showing expression profiles of marker genes in different myeloid
   subsets. C Box plot showing distribution of myeloid subsets across
   different age groups. The P values are calculated with kruskal.test. D
   Box plot showing signature scores of monocyte subsets in different age
   groups based on ssGSEA algorithm, t-tests (two-sided) were performed. E
   Clustering network of significantly enriched GO pathways in the GSEA
   analysis. The nodes representing the significant GO pathways are
   colored by normalized enrichment score (NES). F The curves represent
   the changing trends of DCs’ gene modules across different age groups
   (left), and heatmap showing gene expression of different modules
   (middle), and the represent functional pathways of enrichment analysis
   were shown within the boxes (right). G UMAP projection of NK profiles
   in PBMC of different age groups. H Dot plots showing expression
   profiles of marker genes in different NK subsets. I Box plot showing
   distribution of NK subsets across different age groups. The P values
   are calculated with kruskal.test. J Heatmap showing mean 14 PROGENy
   pathway scores of different NK subsets. K GO-based enrichment analysis
   illustrating indicated pathways upregulated in HLA_CD56^dim subset. L
   The curves represent the changing trends of CD56.^bright NK cells’ gene
   modules across different age groups (left), and heatmap showing gene
   expression of different modules (middle), and the represent functional
   pathways of enrichment analysis were shown within the boxes (right)

   NK cells are innate immune cells that play critical roles in
   coordinating tumor immunosurveillance and viral infection [[92]26].
   Five clusters of NK cells were identified in the PBMCs of all the age
   groups (Figures S3A), and identified four NK cell types with unique
   transcriptional features (Figures S3B), including three types of
   CD56^dim NK (classical_CD56^dim, inflamed_CD56^dim, and HLA_CD56^dim)
   and CD56^bright NK (Fig. [93]2G and H). We found that
   Classical_CD56^dim NK cells increased, whereas Inflamed_CD56^dim NK and
   CD56^bright NK cells decreased with aging (Fig. [94]2I). Notably,
   HLA_CD56^dim NK cells were specifically enriched in the Tg group
   (Figures S3C), although there was variation between healthy individuals
   (Figures S3D). Next, we used module scoring to evaluate the functional
   pathways related to NK based on the IOBR package, which provides a
   comprehensive investigation of the estimation of reported or user-built
   signatures [[95]27]. Classical_CD56^dim NK cells had the lowest
   inflammatory signaling and CD56^bright NK cells had strong cytokine and
   chemokine secretion abilities (Figures S3E). Furthermore, we took
   advantage of PROGENy, which addressed both limitations by utilizing an
   extensive collection of publicly available perturbation experiments, to
   evaluate 14 cell global functional pathways of NK subsets [[96]28]. The
   result showed HLA_CD56^dim NK had strong inflammatory signatures and
   the p53 pathway was activated (Fig. [97]2J). Considering that
   HLA_CD56^dim NK specifically presented in the Tg group, we wanted to
   further explore its features. Pathway enrichment analysis of especially
   upregulated in HLA_CD56^dim NK cells revealed enrichment of terms
   associated with “response to IL-12”, “response to IL-15” and “NF-kB
   signaling”, which were consistent with the inflammatory phenotype (Fig.
   [98]2K). Consistent with the literature [[99]29], the CD56^bright NK
   population responsible for cytokine production also decreased with
   aging in our data (Fig. [100]2I). Additionally, we extracted the highly
   expressed genes in CD56^bright NK cells and performed gene pattern
   analysis across different age groups. We found the decrease in cytokine
   and chemokine secretion ability of this type of NK cell with aging was
   not due to cytokine regulation dysfunction but the decrease in golgi
   function and epigenetic alterations (Fig. [101]2L and Supplementary
   Table 6). Similar to CD56^bright NK cells, inflammatory CD56^dim NK
   cells also decreased with age, which may relate to cell cycle arrest
   (Figures S3F). Collectively, we revealed aging-related changes in NK
   cell composition and function.

Age-group-specific adaptive subsets match distinct system immune function

   Although the steps underlying the activation and differentiation of
   Age-engaged B cells have been extensively characterised [[102]30,
   [103]31], the cellular and molecular mechanisms underlying this process
   remain unclear. Therefore, we extracted B cells to further reveal seven
   cell clusters in PBMCs of all age groups (Figures S4A). Analysis of
   differentially expressed genes across these clusters revealed six B
   cell types with unique transcriptional features (Fig. [104]3A and
   Figures S4B): Naïve_B (TCL1A^+), CD83_B (CD83^+), Memory_B
   (TNFRSF13B^+), Activated_memory_B (CD86^+), Plasma_cell (XBP1^+MZB1^+),
   and Transitional_B cells (CD5^+) (Fig. [105]3B). Unlike innate immune
   cells, B-cell subsets varied erratically across age groups except
   Plasma_cell (Fig. [106]3C). However, although the number of plasma
   cells gradually increased with age, the function of protein secretio
   declined (Fig. [107]3D). Notably, CD83_B cells were specifically
   enriched in the Tg group, and each patient in this group had a higher
   proportion of these type B cells (Figures S4C and 4D). To reveal the
   features of each B cell subset, we evaluated the module scores of
   functional pathways based on the signatures collected from previous
   studies. As depicted in Figures S4E, plasma cells showed higher protein
   secretion ability and lower antigen processing features among the six
   cell types, whereas Activated_memory B cells had the highest
   inflammatory score. Besides, CD83_B cells exhibited the second highest
   inflammatory features, also characterized by high expression of MHC
   class II molecules. PROGENy analysis illustrated that CD83_B cells had
   high TGF-β and MAPK pathway scores (Figures S4F). Thereafter, we
   compared CD83_B cells (Tg-enriched) with naïve B cells and found the
   former upregulated activation-related molecules (IGFBP4 and CD69) and
   inflammatory molecules (CCL4 and CCL4L2) (Fig. [108]3E and
   Supplementary Table 7). Furthermore, gene set enrichment analysis
   (GSEA) similarly indicated that CD83_B cells were transcriptionally
   poised to IL-2 and IL-15 stimulated B cells in an inflammatory state
   (Fig. [109]3F). Aging is associated with decreased efficacy of
   vaccination in both humans and mice, with reduced B cell memory
   formation [[110]32]. In this regard, we found that BCR signaling was
   significantly impaired in the elderly Memory_B cells, which was also
   related to the reduced efficacy of vaccination and increased
   susceptibility to infection in the elderly (Figures S4G). Activated
   memory B cells were enriched in the Ad group. To reveal this type of
   cell function, we compared activated memory B cells with memory B cells
   and found that the former upregulated many cytotoxic molecules, such as
   NKG7, GNLY, CCL5, and GZMB (Figures S4H). Taken together, our data
   suggest that B cells did not adhere strictly to age variation, and
   group-specific enriched subsets played special roles with distinct
   functional features.

Fig. 3.

   [111]Fig. 3
   [112]Open in a new tab

   Age-group-specific adaptive subsets match distinct system immune
   function. A UMAP projection of B cell profiles in PBMC of different age
   groups. B Dot plots showing expression profiles of marker genes in
   different B cell subsets. C Box plot showing distribution of B cell
   subsets across different age groups. The P values are calculated with
   kruskal.test. D Densityheatmap showing indicates pathway score of
   plasma cells based on AUCELL algorithm (left), and representative genes
   dot plot (right). E Differential gene expression analysis using the
   log-fold change expression versus the difference in the percentage of
   cells expressing the gene comparing CD83_B cells versus Naïve_B cells
   (Δ Percentage Difference). F GSEA plots depict the enriched gene sets
   identified between the CD83_B cells and naïve_B cell subsets associated
   with B cell activation. G UMAP projection of T cell profiles in PBMC of
   different age groups. H Dot plots showing expression profiles of marker
   genes in different T cell subsets. I Box plot showing distribution of T
   cell subsets across different age groups. The P values are calculated
   with kruskal.test. J Densityheatmap showing indicates pathway score of
   CD8_CTL based on AUCELL algorithm (left), and representative genes dot
   plot (right). K Densityheatmap showing indicate pathway score of Treg
   based on AUCELL algorithm. L Heatmap showing transcriptomic similarity
   of T cell subsets. M Spearman correlation between percentage of CD4_CTL
   and unknown_T cells. N The distribution of T cell subtypes during the
   transition, along with the pseudo-time (upper). Subtypes are labeled by
   colors (lower). O Two-dimensional plots showing the dynamic expression
   of feature genes. P Heatmap showing the dynamic changes in gene
   expression along the different branches (left) and pathway enrichment
   results in each gene module (right). In figure D, J and K, the upper
   and lower curves of heatmap represent 75% and 25% of density,
   respectively. Point size of dot plot shows the fraction of cells with
   non-zero expression. The P values in Figure D, J and K are calculated
   with kruskal.test

   In contrast to B cells, there is a clear and well-accepted
   understanding that age-related aberrant T cell-driven cytokine and
   cytotoxic responses lead to the failure of immune tolerance and
   sensitivity to infectious diseases in older people [[113]33].
   Therefore, we extracted T cells to explore the timing and mechanisms
   behind the decline in T cell function. First, all T cells were
   clustered into 18 subsets (Figures S5A). We then identified four CD4^+
   T subsets (CD4_TNa, CD4_TEM, CD4_CTL, and Treg), three CD8^+ T subsets
   (CD8_TNa, CD8_TEM, and CD8_CTL), NKT, proliferating_T, and unknown_T
   cells (Fig. [114]3G). Subsequent marker gene expression analysis
   confirmed the accuracy of these annotations (Fig. [115]3H), and the
   most up- and down-regulated genes were also calculated (Figures S5B).
   Among the T cell subsets, we found unprecedented percentage variation
   across age groups (Fig. [116]3I). Notably, CD4_CTL and unknown_T cells
   were especially enriched in the Sc group, although there was variation
   between patients (Figures S5C and 5D). To characterize various T cell
   subsets, we evaluated the functional pathways of all T cell subsets
   using module scores. Our results showed that, except CD4_CTL, CD4^+ T
   cell subsets exhibited lower activities of HLA_signature and chemokine
   pathways than CD8^+ T cell subsets (Figures S5E). Interestingly,
   CD4_CTL exhibited comparable cytotoxic and inflamed scores, yet they
   did not express the same exhausted features as CD8_CTL (Figures S5F).
   Furthermore, the PROGENy score also distinctively clustered CD4^+ and
   CD8^+ T cells (Figures S5G). CD8_CTL cells, which serve as the body’s
   robust defenders, were found to be most affected by aging [[117]33].
   We, therefore, examined the antimicrobial function of this T cell type
   across different age groups and found a decline in the elderly (Fig.
   [118]3J). Dot plots of related genes revealed that CD8_CTL cells in old
   people gradually lost the CD8 molecule (CD8A and CD8B) expression,
   whereas nature killer receptors (KLRB1 and KLRD1) were upregulated. We
   also assessed the suppressive function of Tregs by analyzing the
   IL2-STAT5 and TGF-β pathways across different age groups. Results
   showed that although children had a higher percentage of Tregs, the
   Tregs in age group showed lower IL2 and TGF-β signaling, possibly due
   to limited antigenic acclimation (Fig. [119]3K). Effector memory T
   cells (TEM), which indicate the body’s response to secondary immunity.
   As shown in Figures S5H, the TCR signaling in CD4_TEM and CD8_TEM were
   found to be lower in children, peaked in adults, and declined in the
   elderly. A similar pattern was observed in TCR and inflammatory
   signaling in NKT cells, while chemokine capacity was affected only by
   advanced age (Figures S5I). Recent research has highlighted the
   significant role of CD4_CTL cells in autoimmune diseases and tumor
   immunity, and an increase in these cells has been observed in
   supercentenarians [[120]17, [121]34, [122]35]. However, the precursor
   cells for CD4 CTL remain unclear. As shown in Fig. [123]3I, CD4_CTL and
   unknown_T cells were both enriched in the Sc group, prompting us to
   investigate a potential correlation between these two T cell types.
   Correlation analysis suggested that other T cells were most similar to
   CD4_CTL at the transcriptome level (Fig. [124]3L). Further correlation
   analysis revealed a significant strong correlation between the
   proportion of CD4_CTL and unknown_T cells (Pearson correlation value
   0.88, Fig. [125]3M). According to previous study [[126]36], we
   discovered that unknown_T cells highly expressed CD4_CTL precursors’
   molecules (CD27, TCF7, and LTB) compared to CD4_CTL cells (Figures S5J
   and Supplementary Table 8). Therefore, we extracted CD4_CTL and
   unknown_T cells for pseudotime analysis to construct a developmental
   trajectory (Fig. [127]3N). We found that the expression levels of naive
   and stemness features were down-regulated, while killing molecules were
   gradually up-regulated along with pseudotime (Fig. [128]3O). We next
   investigated the transcriptional changes associated with two branches,
   and explored the functional features of two distinct CD4_CTL subsets.
   We identified one subset characterized by a response to chemokines, and
   another expressing adhesion molecules indicative of direct contact with
   target cell; both subsets acquired cell-killing abilities during
   differentiation (Fig. [129]3P and Supplementary Table 9). Together, we
   identified a new CD4 T cell subset that was enriched in Sc group, may
   related to a higher percentage of CD4_CTL cells in the blood
   supercentenarians.

Characterizing developmental hierarchies and quiescent immune cells using
CytoTRACE

   Considering that not all immune processes are uniformly sensitive to
   aging, various immune cell populations are heterogeneously influenced
   by the increase in age [[130]37]. To reveal cellular states with
   intrinsic differences in differentiation state across different age
   groups, we used CytoTRACE to construct a cell differentiation atlas
   among PBMCs (Methods). As depicted in Figures S6A, all myeloid cells
   were divided into different cell clusters, which were highly consistent
   with the age groups according to their CytoTRACE scores. Additionally,
   the box plot showed that myeloid cells in the Cd, Ad, and Tg groups
   were less differentiated, while those in the elderly showed higher
   levels of differentiation. Surprisingly, myeloid cells in the Sc group
   exhibited a higher potential for differentiation capacity. The
   conclusion that aging markedly affects NK function remains
   controversial and is related to the diverse health statuses of the
   study subjects [[131]38]. Our results demonstrated that NK cells in the
   Ad group had the highest differentiation capacity, suggesting that
   their differentiation function was well maintained into adult stage
   (Figures S6B). In contrast to innate immune cells, previous studies
   have indicated that adaptive immune function declined with age
   [[132]30, [133]39]. However, our findings illustrated that the
   differentiation state of B/T cells did not decline in a straight line
   but gradually improved before the adulthood, peaked in adults, and then
   declined (Figures S6C and 6D). Unlike other immune cells, T cells in
   the Sc group were the most differentiated, in contrast to Ed, which
   strictly followed a trend of gradual decline with increasing age.
   Conversely, these results revealed that lymphoid and myeloid cells had
   different differentiation trajectories; the former had a
   well-established differentiation potential early on, while the latter
   developed this ability until adulthood.

Analysis of the metabolic pathways reveal a shift in energy supply during
immune cell aging

   Metabolism, being central to all biological processes, is crucial for
   the proper regulation of immune cells. Perturbations in metabolism can
   cause immune dysfunction and disease progression [[134]40]. Therefore,
   we aimed to determine the metabolic activity of all immune cells across
   different age groups at single cell resolution. Following the
   instructions recommended by KEGG's official website, we divided the
   extracted specific metabolic pathways into six categories (Methods).
   Lipid and carbohydrate metabolism declined the most dramatically with
   increasing age (Figures S7A). Notably, some lipid-related metabolic
   pathways showed an inverse trend, such as arachidonic acid metabolism
   in myeloid and NK cells, alpha-linolenic acid metabolism, steroid
   hormones, and fatty acid biosynthesis in B and T cells. As shown in
   Figures S7B, oxidative phosphorylation activity was heightened in
   elderly groups across all immune cells, with other energy metabolism
   pathways also displaying increased activity in advanced age, excluding
   myeloid cells. Amino acid metabolism is essential for metabolic
   rewiring, supporting various immune cell functions beyond increased ATP
   generation, nucleotide synthesis, and redox balance [[135]41]. Notable
   amino acid metabolism-specific enhancements were observed in aged T
   cells. In particular, taurine and hypotaurine metabolism pathways were
   detected exclusively in T cells and were more active in the elderly.
   Intrigued by these findings, we further explored this metabolic pathway
   to differentiate between CD4^+ and CD8^+ T cell subsets. Surprisingly,
   it was not the CD8^+ T cells (Figures S7C), but the CD4^+ T cells
   (Figures S7D) that showed an increasing trend with age. Collectively,
   these results highlighted that besides the uniformly active metabolites
   found in various immune cells of the elderly, there were also specific
   metabolites that were elevated exclusively in certain immune cells,
   which deserves further study.

Decline of infection-fighting immune function with aging

   To delineate the characteristics of entire immune cell subsets, we
   integrated all immune cells (Panage_data) and calculated the
   percentages of each cell type within all PBMCs across samples
   (Supplementary Table 3). Unsupervised hierarchical clustering, based on
   cellular composition, showed that the patients formed a remarkable
   divergence in different age groups (Fig. [136]4A). Odds ratio (OR)
   analysis revealed the cell distribution preferences of each age group,