Loss of immune cell identity with age inferred from large atlases of single cell transcriptomes

Abstract

By analyzing two large atlases of almost 4 million cells, we show that immune-senescence involves a gradual loss of cellular identity, reflecting increased cellular heterogeneity, for effector, and cytotoxic immune cells. The effects are largely similar in both males and females and were robustly reproduced in two atlases, one assembled from 35 diverse studies including 678 adults, the other the OneK1K study of 982 adults. Since the mean transcriptional differences among cell-types remain constant across age deciles, there is little evidence for the alternative mechanism of convergence of cell-type identity. Key pathways promoting activation and stemness are down-regulated in aged T cells, while CD8 TEM and CD4 CTLs exhibited elevated inflammatory, and cytotoxicity in older individuals. Elevated inflammatory signaling pathways, such as MAPK and TNF-alpha signaling via NF-kB, also occur across all aged immune cells, particularly amongst effector immune cells. This finding of lost transcriptional identity with age carries several implications, spanning from a fundamental biological understanding of aging mechanisms to clinical perspectives on the efficacy of immunomodulation in elderly people.

Keywords: computational pipeline; differential expression; immune cell aging; peripheral blood mononuclear cells; scRNAseq.

FIGURE 1

Overview of the approach and clustering strategy used for analyzing the circulating immune cell profiling of aging and sex. (a) scRNA‐seq data of PBMCs from a discovery cohort comprising of 356 women and 322 men (18–110 years old) and the independent 1.2 millioncell OneK1K validation atlas from 982 Australian (Tasmanian) adults (18–97 years old). (b) Donor demographics and sample type displayed as color‐coded donut plots. (c) Bar plot showing the number of marker genes for each of 19 cell types from all 2.74 M cells in the discovery atlas.(d) UMAP of PBMCs across all individuals in the discovery atlas, with 19 transcriptionally distinct populations. (e) Density plots of canonical markers of immune cells in discovery atlas. Density corresponding to expression of the labelled gene. Cells are colored based on the expression of the transcript, bottom legend.

FIGURE 2

Distribution of age and sex in the (a) discovery and (b) OneK1K cohorts. (c–d) Age distribution for deciles of the PBMC data atlases. Healthy subjects were divided into 20 groups (10 age deciles for males and 10 age deciles for females) according to sex and the age distributions for Discovery Atlas and OneK1K.

FIGURE 3

Loss of effector immune cell identity with age (a) Positive correlation between age and the number of cell‐type specific markers in effector immune cells (CD8 effector T cells; MAIT cells; NK cells; NK CD56+ cells; CD14 Monocytes). Negative correlation between age and the number of cell type specific markers in CD4 regulatory T cells. DEG analysis performed with FastDE (b) DEG analysis replicated with limma. See Figure S4 for full set of cell types.

FIGURE 4

Two alternative hypotheses on the origin of age‐related loss of cell type identity. (a) The “dispersion” hypothesis postulates that the mean intrinsic dispersion for within each cell‐type increases with aging. (b) The “convergence” hypothesis proposes that immune cell types are extrinsically heterogenous in younger individuals. However, with aging, heterogeneity is lost, and cell types converge due to the reduced repertoire of cell type‐specific genes expressed. (c) Bar plot depicting the mean and standard error for the Euclidean distance from each cell to the centroid within the subset in young (<40 years) and older (79+ years) males and females. (d) Bar plot depicting the mean and standard error for the Euclidean distance from each cell to the centroid between all subsets in young (<40 years) and older (79+ years) males and females.

FIGURE 5

Changes in transcriptional profiles during aging. (a) Numbers of age‐related differentially expressed genes (DEGs) in four immune cell subsets comparing older males (OM) to younger males (YM) and older females (OF) to younger females (YF). (b) UpSet Plot showing the integrated comparative analysis of upregulated DEGs in major immune cell lineages between OF and YF Upregulated DEGs: Upregulated in OF, downregulated in YF. (c) Numbers of sex‐related DEGs.(YM: YF and OM: OF) in four immune cell subsets. (d) UpSet Plot showing the integrated comparative analysis of upregulated DEGs in four major immune cell lineages between OF and OM. Upregulated DEGs: Upregulated in OF, downregulated in OM. (e) Venn diagram showing integrated comparative analysis of upregulated age‐related DEGs in T cells between females and males. (f) Representative GO terms and pathways enriched in age‐related DEGs based on functional enrichment analysis in T cells of females and males. (g) Distribution and comparison of the aging score in T cell subpopulations. (h) Density plots of aging score and selected gene modules in T cells. (i) Distribution and comparison of the aging score in all T cells of each age group. (CD4 TCM, CD4+ Central Memory T cells; CD8 TCM, CD8+ Central Memory T cells; CD4 TEM, CD4+ Effector Memory T cells; CD8 TEM, CD8+ Effector Memory T cells; CD4 CTL, CD4+ Cytotoxic T cells).