RESULTS

Cohort characteristics and single-cell analysis of PBMCs in young and aged adults
To generate a comprehensive immune cell atlas reflecting cellular and systemic adaptations resulting from age and/or COVID-19 infection, we integrated scRNA-seq, CyTOF, scATAC-seq and scTCR/BCR-seq of single-cell PBMC suspensions collected from 3 separate cohorts (Fig. 1A, 1B, and Table S1A–G). In cohort-1, comprising young healthy adults (YA) (20–45 years old) and aged healthy adults (AA) (≥60 years old), we combined CyTOF (n = 10) and scATAC-seq (n = 10) with scRNA-seq (n = 16) and scTCR/BCR-seq (n = 16); in cohort-2, comprising young healthy (YH) individuals (30–45 years old), aged healthy (AH) individuals (≥60 years old), young COVID-19 onset patients (YCO) (30–50 years old) and aged COVID-19 onset patients (ACO) (≥70 years old), we performed CyTOF analysis (n = 8); and in cohort-3, comprising YH individuals, AH individuals, young recovered COVID-19 patients (YCR) (30–50 years old) and aged recovered COVID-19 patients (ACR) (≥70 years old), we performed scRNA-seq (n = 22) (Fig. 1B). By combining scRNA-seq, CyTOF, scATAC-seq and scTCR/BCR-seq analysis, we created a comparative framework detailing the impact of aging on cell type distribution and immune cell functions at the transcriptional, proteomic, and chromatin accessibility levels in cohort-1. In cohort-2, we measured single-cell protein expression using a 26-marker CyTOF panel to discover early cellular changes in incipient COVID-19 patients and how those changes were affected by age. Finally, in cohort-3, we compared cellular differences between young and aged recovered COVID-19 patients by scRNA-seq analysis (Fig. 1B).
Figure 1 Schematic illustration of the collection and data processing of PBMC from young and aged group. (A) Flowchart overview of PBMC collection in young and aged adults followed by scRNA-seq, mass cytometry, scATAC-seq and scTCR/BCR-seq experiments. (B) Schematic illustration of experimental cohorts; cohort-1: young and aged adults, cohort-2: young and aged healthy individuals, young and aged adults with COVID-19 onset, cohort-3: young and aged healthy individuals, young and aged adults recovered from COVID-19, matched with analysis as indicated: single-cell proteomic data from CyTOF studies, gene expression data from scRNA-seq studies, chromosomal accessibility data from scATAC-seq, and TCR and BCR repertoire data from scTCR/BCR-seq. (C) t-SNE projections of PBMCs derived from scRNA-seq data in cohort-1. (D) Heatmaps showing scaled expression of discriminative gene sets for each cell type and cell subset. Color scheme is based on z-score distribution from −3 (purple) to 3 (yellow)
We analyzed PBMC single-cell suspensions by CyTOF for the protein expression of several lineage-, activation- and trafficking-associated markers and converted them to barcoded scRNA-seq libraries using 10x Genomics for downstream scRNA-seq, scATAC-seq and scTCR/BCR-seq analysis. CellRanger software and the Seurat package were used for initial processing of the sequencing data. Quality metrics included numbers of unique molecular identifiers (UMIs), genes detected per cell, and reads aligned that were comparable across different research subjects. We identified red blood cells (RBCs), megakaryocytes (MEGAs) and five major immune cell lineages (TCs, NKs, BCs, MCs and DCs) based on the expression of canonical lineage markers and other genes specifically upregulated in each cluster (Figs. 1C, 1D and S1A–C). In accordance with the scRNA-seq results, we identified five immune cell lineages (TCs, NKs, BCs, MCs and DCs) in CyTOF using t-distributed stochastic neighbor embedding (t-SNE), an unbiased dimensionality reduction algorithm (See Table S2 for a list of antibodies) (Fig. S2A–D). Cell-type-specific marker genes were determined by differential gene expression values between clusters positioned and visualized in a t-SNE plot (Figs. S1 and S2). The definition of cell types in clusters in the t-SNE maps was comparable between old and young individuals (Figs. S1B and 2B) both by scRNA-seq and CyTOF, indicating that the cell type identity was not altered with age.
Figure 2 Changes in cell proportions during aging. (A) Bar chart of the relative percentage of immune cell types derived from scRNA-seq data in PBMCs. (B) Bar chart of the relative percentage of immune cell subsets derived from scRNA-seq data in PBMCs. The focused cell-subsets have been marked red. (C) Pie charts showing relative cluster abundances derived from mass cytometry data in the YA and AA groups. (D) Percentage of CD4 Naive cells in PBMCs from YA (n = 8) and AA (n = 8) groups. (E) Percentage of CD8 Naive cells in PBMCs from YA (n = 8) and AA (n = 8) groups. (F) Percentage of CD4 Naive cells in CD45+ cells from YA (n = 5) and AA (n = 5) groups. (G) Percentage of CD8 Naive cells in CD45+ cells from YA (n = 5) and AA (n = 5) groups. (H) Bar chart of the relative percentage of CD4+ T cell subsets derived from scRNA-seq data in PBMCs. (I) Bar chart of the relative percentage of CD8+ T cell subsets derived from scRNA-seq data in PBMCs. (J) Bar chart of the relative percentage of CD4+ T cell subsets derived from mass cytometry data in CD45+ cells. (K) Bar chart of the relative percentage of CD8+ T cell subsets derived from mass cytometry data in CD45+ cells. (L) Percentage of CD14 monocytes in PBMCs from YA (n = 8) and AA (n = 8) groups. (M) Percentage of CD14 monocytes in CD45+ cells from YA (n = 5) and AA (n = 5) groups. (N) Bar chart of the relative percentage of DC subsets derived from scRNA-seq data in PBMCs. (O) Bar chart of the relative percentage of DC subsets derived from mass cytometry data in CD45+ cells. P values are based on two-tailed Mann-Whitney-Wilcoxon tests between groups

Dissection of immune cell subtypes in the cellular aging ecosystem
To classify each cell subpopulation in an unbiased manner, we separately reclustered the cells of each lineage. By analyzing the most significantly upregulated genes in each cluster in scRNA-seq analysis, we identified five distinct subsets of CD3+ TCs (Fig. S3A), five distinct subsets of CD4+ TCs (Fig. S3B), four distinct subsets of CD8+ TCs (Fig. S3C), three distinct subsets of NKs (Fig. S3D), four distinct subsets of BCs (Fig. S3E), three distinct subsets of MCs (Fig. S3F) and four distinct subsets of DCs (Fig. S3G, see Table S3A for the details).
Aging affects the development and function of TCs and NKs (Pinti et al., 2016). We identified known T cell subsets, including CD4+, CD8+, CD4+CD8+, CD4−CD8− and proliferative T cells (mitotic T cells, T-mito), based on the expression of canonical lineage markers (Fig. S3H). The CD4+ T cells were subdivided into five classes: CCR7high CD69low naive CD4+ T cells (CD4 Naive); CCR7med CD69high CCR6− central memory CD4+ T cells (CD4 Tcm); CCR6+ effector memory CD4+ T cells (CD4 Tem); FOXP3+ regulatory T cells (CD4 Treg) and PDCD1+ exhausted CD4+ T cells (CD4 Tex) (Fig. S3I). The CD8+ T cells were subdivided into four classes: CCR7+ naive CD8+ T cells (CD8 Naive); GZMK+ effector memory CD8+ T cells (CD8 Tem); GZMB+ GNLY+ cytotoxic CD8+ TCs (CD8 CTL) and PDCD1+ exhausted CD8+ T cells (CD8 Tex) (Fig. S3J).
Analysis of NK cell-status identified circulating NKs with three separate immune states (Fig. S 3D): the CD16 (FCGR3A)low CD56 (NCAM1)bright NK population (NK1), the CD16high CD56dim CD57 (B3GAT1)− low-cytotoxic NK compartment (NK2) and the CD16high CD56dim CD57+ late NK population (NK3) (Fig. S3K). In addition, we identified four major peripheral B cell subsets: IL4R+ IGHD+ naive B cells (Naive BCs); CD27+ IGHG1+ memory B cells (Memory BCs); plasma cells or so-called antibody-secreting cells (ASCs), expressing high level of immunoglobulin genes MZB1; and a subset of ITGAX+ B cells defined as age-associated B cells (ABCs) (Fig. S3L).
In human peripheral blood myeloid cells (including MCs and DCs), known to promote antigen presentation and inflammatory activities, we identified seven transcriptionally distinct subsets: CD14high CD16− classical monocytes (CD14 MCs), CD14+/− CD16high nonclassical monocytes (CD16 MCs), CD14+ CD16+/− intermediate monocytes (Intermed MCs) (Fig. S3M), CLEC9A+ conventional DC1 (cDC1), CD1c+ cDC2 conventional DC2 (cDC2), CD123 (IL3RA)+ CLEC4C+ plasmacytoid DCs (pDCs) (Fig. S3N), and dendritic cell precursors (pre-DCs) expressing AXL and CD123 (Grabiec and Hussell, 2016; Ruffin et al., 2019) (Fig. S3N). Therefore, we targeted the immune cell changes based on more precise classification of each subgroup.
To further verify the aging-associated change in the cell ratio, we performed single-cell analysis at the protein level. Similar to the cell clusters and subsets in scRNA-seq results, in CyTOF analysis, we identified 21 sub-clusters with nine subsets of TCs (CD4 Naive, CD4 Tcm, CD4 Tem, CD4 Treg, CD8 Naive, CD8 Tem, CD8 CTL, CD4+ CD8+, CD4− CD8−), three subsets of NKs (CD56bright NK1, CD16+CD57− NK2 and CD16+CD57+ NK3), four subsets of BCs (Naive BC, Memory BC, ASCs, and ABCs), three subsets of MCs (CD14high MCs, CD16high MCs and intermediate MCs), and two subsets of DCs (pDCs and cDCs) (Fig. S4A–K, see Table S3B for the details).

Aging shifts the cellular composition toward extreme effector phenotypes
To delineate how cell-type composition changed with aging, we separately compared the proportions of each cell type across major cell types between the YA and AA groups. We observed changes at the single-cell transcriptional level, which were further confirmed at the protein level by CyTOF. Globally, we found that TCs and BCs, especially the former, decreased by approximately 10% in all PBMCs with scRNA-seq analysis (Fig. 2A, 2B, and S5A) and by 15% with CyTOF (Figs. 2C and S5B). In contrast, MCs increased by approximately 7% in scRNA-seq analysis (Figs. 2A, 2B, and S5A) and by 10% in CyTOF (Figs. 2C and S5B).
The composition of cell subsets across all cell lineages differed between the YA and AA groups. Among TCs, CD4+ TCs were increased, CD8+ TCs were decreased, and CD4+CD8+ and proliferating T cells were increased in the AA group (Fig. 2B and 2C). Moreover, naive TCs, especially CD4 Naive and CD8 Naive, showed a common distribution in the YA group but were reduced in the aged group (P = 0.0175, Fig. 2D–G). Conversely, effector, memory and exhausted cell subsets were dominant in the aged group (Fig. 2H–K). The AA group also had a diminished proportion of the CD56bright NK1 population and an expansion of the NK2 and late NK3 populations (Fig. S5C and S5D). Analysis of BC clusters revealed that Naive BCs were decreased while ABCs were mildly increased in the AA group compared to the YA group (Fig. S5E and S5F).
Our data also showed that elderly research subjects had increased MC subsets, particularly classical CD14 MCs and, to some extent, nonclassical CD16 MCs and intermediate MCs (Figs. 2L, 2M, S5G, and S5H). Overall MC growth mainly resulted from CD14 MC enrichment (P = 0.0012, Fig. 2M). However, given that CD14 MCs made up 70%–80% of the MCs population, the increase we observed in CD16 MCs was more remarkable as a change in the overall population proportion between the AA and YA groups, which was not observed for intermediate MCs between these groups (Fig. S5G–J). A similar analysis of the DC subset composition showed that the percentage of cDC2 cells increased, whereas cDC1, pDC, and pre-DC decreased with age (Figs. 2N, 2O, and S5I).
In summary, these results demonstrate that aging induces an immune dysfunction shift into effector and inflammatory cell populations.

Identification of aging-related cell-type-specific transcriptional expression changes
To identify cell-subtype-specific gene signatures associated with aging, we performed an integrated comparative analysis of differentially expressed genes (DEGs) from blood immune cells in the YA and AA groups. We found that blood immune cells showed heterogeneous transcriptional changes affected by aging based on the number of DEGs. Strikingly, BC was the cell type most affected by aging, followed by TC and MC (Figs. 3A, S6A; Table S4A–E). Specifically, we found a set of 60 genes whose expression was increased in all kinds of immune cells, indicative of general oxidative stress (e.g., DDIT4, CASP4, TSPO) and an inflammatory state (e.g., DUSP2, S100A10, COX5A, PSMB6) across cell populations (Fig. 3A). Conversely, genes with decreased expression shared across all cell populations included DDX17, RBM39, and SCAF11, which are involved in RNA splicing (Fig. S6A and S6B). Consistent with our understanding of the main immune cell lineages, we found that the myeloid and lymphocyte cell lineages were characterized by unique gene expression spectra, whereas TCs showed the highest heterogeneity in DEGs. To explore the biological implications of our data in the context of aging, we used Gene Ontology (GO) and pathway analysis for each immune cell population. Common aging-upregulated biological pathways included TNF signaling, IL-1 signaling, the apoptotic signaling pathway, and the adaptive immune response (Fig. 3B). We found that these pathways were especially enhanced in TCs. In addition, aging-upregulated biological pathways in MCs were enriched for interferon-gamma (IFN-γ) signaling and cell aging (Fig. 3B). To assess the impact of aging on circulating immune cells, we selected the top 20 genes of the 60 total genes that were upregulated in all immune cells (Fig. 3A) and calculated aging scores across all immune cell types. We found that MCs and DCs had the highest scores, suggesting that senescent cells are most likely present in these cell populations (Fig. 3C). Moreover, when calculating the scores of individual samples, we found that individuals in the AA group had consistently higher scores than individuals in the YA group (Fig. 3D), suggesting that aging-score assessments are suitable for studying aging-related immune dysfunction.
Figure 3 Changes in transcriptional profiles during aging. (A) UpSet Plot showing the integrated comparative analysis of upregulated DEGs in major immune cell lineages between YA and AA groups. Upregulated DEGs: upregulated in AA, downregulated in YA group. (B) Representative GO terms and pathways enriched in upregulated DEGs based on functional enrichment analysis in major immune cell populations. P value was derived by a hypergeometric test. (C) Distribution and comparison of the aging score in immune cell populations. (D) Distribution and comparison of the aging score in all cells of each sample. (E) UpSet plot showing the integrated comparative analysis of upregulated DEGs in CD4+ T cells between YA and AA groups. Upregulated DEGs: upregulated in AA, downregulated in YA. The count showing the number of DEGs. (F) Representative GO terms and pathways enriched in upregulated DEGs based on functional enrichment analysis in CD4+ T cells. P value was derived by a hypergeometric test. (G) Venn diagram showing integrated comparative analysis of upregulated DEGs in monocytes between YA and AA groups. Upregulated DEGs: upregulated in AA, downregulated in YA. The count showing the number of DEGs. (H) Representative GO terms and pathways enriched in upregulated DEGs based on functional enrichment analysis in monocytes. P value was derived by a hypergeometric test. (I) Violin plots showing the distribution of normalized expression levels of selected aging-associated genes in all DC cluster between YA and AA groups. (J) t-SNE plots segregated on the basis of DC subsets. (K) Representative GO terms and pathways enriched in biased DEGs of cDC2-A and cDC2-B clusters. P value was derived by a hypergeometric test. (L) CLEC12A expression in cDC2 is shown as flow cytometry histogram. (M) Percentage of CLEC12A+ cells in cDC2. P value are based on two-tailed Mann-Whitney-Wilcoxon tests between YA and AA groups (n = 3/group)
By analyzing age-associated DEGs in CD4+ TCs, we found enrichment in inflammatory and effector genes (Tables S5A and 6A–E). To determine cell-subtype-specific gene signatures within different CD4+ TC subpopulations, we generated UpSet plots of upregulated DEGs in different CD4+ TC subsets. We found a range of subtype-specific patterns, including the IL2 receptor (IL2RA) in Naive cells, CCR10 in Tem, and GZMB and TRBV11-2 in Tex (Fig. 3E). GO and pathway analysis of the DEGs demonstrated that effector and memory subsets were most affected by aging based on the number of DEGs. For example, in CD4 Tem, TNF, interleukin signaling and apoptotic pathways were enhanced, whereas mRNA processing was impaired (Figs. 3F and S6C). Analysis of CD8+ TC status indicated that the AA group had increased expression of chemokines and granzyme family members (Fig. S6D; Tables S5B and 7A–D). Moreover, aging was associated with a decreased proportion of CD8 Naive with increased apoptotic signaling pathway and lymphocyte activation and an expanded CD8 Tem compartment with increased cytokine production as well as reduced chromatin remodeling and antiviral function (Fig. S6E and S6F). In addition, T-mito in aged group was associated with the upregulated inflammatory signaling molecules HLA-DRB5, PDCD5 and PSMA2 (Fig. S6G; Table S5C) and inflammatory pathways (Fig. S6H).
Analysis of NKs status revealed that the AA group had a smaller fraction of the CD56bright NK1 population and expanded late low-cytotoxic NK subsets than the YA group. Notably, NKs in the AA group had increased expression of DDIT4, ISG20, and CASP4 and decreased expression of DDX17, PCBP1 and TRIM56 (Figs. 3A, S6A; Table S8A–C). These genes were mostly enriched in apoptotic signaling pathways and cellular responses to lipopolysaccharide, along with decreased virus defense responses (Fig. S6I and S6J). As for BCs, we found increased expression of JUNB, IGHA1, SSR4 and CXCR4, indicative of increased memory BC signature and activity during aging (Figs. 3A, S7A; Table S9A–D). Moreover, the comparative functional analysis of aging-associated DEGs revealed that Naive BC in the AA group had increased cytokine-mediated signaling pathways (Fig. S7B). Additionally, analysis of downregulated DEGs and pathways in the AA group demonstrated that BCs were associated with reduced viral defense responses (Fig. S7C). These results indicate that NKs and BCs lose their capacity for antiviral activity with upregulated inflammatory states in aging.
We next studied aging-associated DEGs in MCs and found enrichment in inflammatory genes, such as IL1B, TNF and CXCL8, in the AA group (Fig. 3A). All MC subsets had increased expression of the chemokines, TNF, IL1B and CDKN1A and decreased expression of SIGLEC14 and CLEC12A (Figs. 3G, S6A; Table S10A–C). Analysis of aging-related DEGs demonstrated that the CD14 MC subset was most affected by aging, as reflected by the increased NOD-like receptor signaling pathway, NF-κB signaling pathway, Toll-like receptor signaling pathway, inflammasome pathway, and MAPK pathway (Fig. 3H) and the obvious decrease in RNA splicing, autophagy, and vesicle-mediated transport (Fig. S7D).
To complete our DEG and GO survey of immune lineage cells and their subtypes, we next analyzed aging-associated DEGs in DCs in the YA and AA groups (Figs. 3I, S7E and S7I; Table S11A–D). Among the upregulated DEGs, IFN-stimulated genes (IFITM2, ISG20), TNF and IL1B indicated an overactive inflammatory response in DC clusters in the AA group (Figs. 3A, 3I, S7E, and S7I). We observed that overrepresented pathways in DCs from the AA group included apoptotic, MAPK, IL-1, and IFN-γ signaling pathways (Fig. S7F). Notably, CLEC12A and TXNIP, which are critical for the antigen-presentation function of DCs; and MALAT1 and AHR, which are critical to inducing tolerogenic DCs (Son et al., 2008; Hutten et al., 2016; Wu et al., 2018), were decreased in AA DCs (Figs. 3I, S7G, and S7I), reflecting the decreased antigen-presenting ability of aged DCs (Fig. S7H). These results indicate that DCs acquire an inflammatory state with age but lose the antigen-presenting ability.
Within DC clusters, we found distinct aging manifestations in the cDC2 subsets by comparing DC clusters in the t-SNE map (Fig. 3J). Cells from the YA group grouped together in clusters 0 and 1 (named cDC2-A), whereas cells in AA group grouped distinctively in clusters 3, 4, 10 and 11 (named cDC2-B). The expression signature of cDC2-A cells included antigen presentation-related genes such as AHR, CLEC4E, and CLEC12A, whereas the expression signature of cDC2-B cells included inflammatory and aging-associated genes such as IFN-stimulated genes, IL1B, CDKN2D, DDIT4, CXCL8, and DUSP2 (Fig. S7J and S7K). Moreover, the comparative functional analysis of DEGs between the two clusters indicated that cDC2-A had intact immune regulation and antigen presentation function, while aging-related cDC2-B with high HLA-DQA2 expression exhibited increased inflammatory signaling pathways, such as the response to hypoxia and IL-1 signaling (Fig. 3K). We further confirmed that CLEC12A+ cDC2s were decreased in aging by FACS (Fig. 3L and 3M). Taken together, these findings indicate that aging curtails DC antigen presentation ability and upregulates inflammatory and aging-associated gene expression in DCs.

Identification of aging-related cell-type-specific chromosomal accessibility changes
After quality control, a total of 74,102 cells (33,004 YA, 41,098 AA) were used to generate a PBMC chromatin-accessibility map. MEGAs, TCs, NKs, BCs and myeloid cells were identified based on the promoter sum of genes specifically upregulated in each cluster. After separately reclustering each lineage population, we identified 3 distinct subsets in CD4+ TCs, 3 distinct subsets in CD8+ TCs, 3 distinct subsets in NKs, 3 distinct subsets in BCs, 3 distinct subsets in DCs and 2 distinct subsets in MCs according to gene peaks and transcription factor (TF) activity using chromVAR (Satpathy et al., 2019) (Fig. 4A, 4B, and S8A–D, see Table S 3C for the details). Consistent with the scRNA-seq and CyTOF data, we observed a decrease in naive TCs and an increase in MCs in the elderly (Fig. 4C).
Figure 4 Changes in chromosomal accessibility during aging. (A) Heatmaps showing scaled expression of discriminative gene sets for each cell type and cell subset. Color scheme is based on z-score distribution from −1.5 (purple) to 1.5 (yellow). (B) t-SNE projections of PBMCs derived from scATAC-seq data. (C) t-SNE plots segregated by YA and AA groups. (D) UpSet plot showing the integrated comparative analysis of upregulated differentially expressed transcription factors (DETs) in major immune cell populations between YA and AA groups. Upregulated DETs: upregulated in AA, downregulated in YA. The count showing the number of DETs. (E) UpSet plot showing the integrative comparative analysis of downregulated DETs in major immune cell populations between YA and AA groups. Downregulated DETs: upregulated in YA, downregulated in AA. The count showing the number of DETs. (F) Venn diagram showing integrated comparative analysis of upregulated DETs in CD4+ T cells between YA and AA groups. Upregulated DETs: upregulated in AA, downregulated in YA. The count showing the number of DETs. (G) Venn diagram showing integrated comparative analysis of upregulated DETs in CD8+ T cells between YA and AA groups. Upregulated DETs: upregulated in AA, downregulated in YA. The count showing the number of DETs. (H) Venn diagram showing integrated comparative analysis of upregulated DETs in NK cells between YA and AA groups. Upregulated DETs: upregulated in AA, downregulated in YA. The count showing the number of DETs. (I) Venn diagram showing integrated comparative analysis of upregulated DETs in B cells (top) and monocytes (bottom) between YA and AA groups. Upregulated DETs: upregulated in AA, downregulated in YA. The count showing the number of DETs. (J) Mean scATAC-seq coverage at FOSL2 loci in CD8+ T cells. (K) Mean scATAC-seq coverage at NFATC2 loci in CD8+ T cells. (L) Mean scATAC-seq coverage at CDKN2B loci in B cells. (M) Mean scATAC-seq coverage at SIRT7 loci in NK1 cells. (N) Mean scATAC-seq coverage at GLI2 loci in CD4 Naive cells. (O) Mean scATAC-seq coverage at IFNG loci in CD8 Naive cells. (P) Mean scATAC-seq coverage at DUSP5 loci in CD8 Memory cells. (Q) Mean scATAC-seq coverage at PDCD1 loci in NK3 cells
Next, we focused on the differentially expressed transcription factors (DETs) in immune cells in the AA group compared to the YA group. At the TF level, MCs were the most affected by aging based on the numbers of upregulated and downregulated DETs (Fig. 4D and 4E). To identify aging-associated TF events, we performed an integrated comparative analysis of these DETs and found that AP-1 family TFs, including FOSL2 and JUNB, were increased in all immune cells during aging (Fig. 4D and 4E). Upregulation of AP-1 family TFs, including FOS, FOSB, FOSL1, FOSL2, JUN, JUNB, and JUND, was also observed in almost all cell subsets during aging (Fig. 4F and 4I). The AP-1 family regulates a wide range of cellular processes, including cell proliferation, death, survival, and differentiation. The effects of the activated AP-1 TFs, associated with the active inflammatory state, are primarily mediated through combinatorial regulation with the NFAT family, both of which are key regulators of TC activation and are enriched in TCs (Fig. 4D) (Shaulian and Karin, 2002). In addition, we visualized the chromosomal accessibility of FOSL2 loci and NFATC2 loci and found that the chromosomal accessibility of the FOSL2 and NFATC2 gene regions was also increased in aged TCs (Fig. 4J and 4K). CDKN2B, an aging hallmark gene, also showed an increase in accessibility with age (Fig. 4L). In parallel, we found 25 common decreased TFs, including nuclear respiratory factor 1 (NRF1) and ELK4, which are involved in antioxidant stress and negatively regulate cell differentiation and proliferation (Figs. 4E and S8E–I). Consistently, we found that chromatin accessibility also decreased at the loci of SIRT7 (Fig. 4M), which coordinates with NRF1 to maintain cellular energy metabolism and proliferation (Mohrin et al., 2015).
In TCs, a series of subset-specific TF changes were observed, such as GLI2 in naive cells, which has been associated with decreased TC function and impaired immune defenses (Fig. 4F and 4G). Consistently, increased chromatin accessibility was detected in GLI2 loci (Fig. 4N). Analysis of differentially accessible regions (DARs) demonstrated that the IFNG, DUSP5, and GZMB loci were highly accessible, which indicated activated CD8+ TC states (Figs. 4O, 4P and S8J). In our analysis of NK status, we identified the key TF changes in NK subsets during aging (Fig. 4H), and found that the chromatin accessibility of the inhibitory receptor gene increased, while that of the activating receptor decreased. These changes may weaken the ability to clear virus-infected cells. For example, the PDCD1 exhibited higher chromatin accessibility in the gene region of the elderly group, which might be part of the reason why older individuals were prone to infection (Fig. 4Q). In our analysis of BCs, we identified aging-related TF changes, such as TBX21, IRF4, which are consistent with our scRNA-seq results (Fig. 4I). Aging-associated TFs and DARs in MCs demonstrated enrichment in inflammatory-related TFs and gene loci in the AA group, such as NF-κB family (REL, RELA), IL1B, TNF and CXCL8 (Figs. 4I and S8K–M). In summary, aging-related chromosomal accessibility changes are associated with an increase in the inflammatory pathway and an impaired immune response.

Aging-associated heterogeneous changes in clonality and diversity of TCRs and BCRs
Although the antigen repertoire sensed by immunoglobulin receptors on both TCs (TCRs) and BCs (BCRs) is known to continuously evolve with age (Yuseff et al., 2013), the phenomenon of aging-associated TCR and BCR repertoire constriction has not yet been studied at the single-cell level. Here, we employed scTCR/BCR-seq to assess immune cell clonal expansion in the YA and AA groups. We found that relative to the YA group, the AA group was associated with a substantial decrease in unique clonotypes both in TCRs and BCRs (Fig. 5A and 5B), suggesting that both TCR and BCR clonality increased with age. Moreover, quantification of the most highly expanded (maximum) clone for each research subject showed that the ratios of the maximum clones were higher in the AA group than in the YA group (Fig. 5C). Although an aging-related clonal lymphocyte population may reflect an existing adaptive immunity of the elderly, the overall diversity was decreased in the AA group compared to the YA group (Fig. 5D). Analysis of TCR and BCR distributions across different TC and BC subtypes revealed that loss of repertoire diversity was pronounced in CD8+, T-mito and memory BCs of the AA group (Fig. 5E and 5F). To understand how clonally expanded TCs could be affected by aging, we performed DEG analysis of clonal cells between YA and AA groups and revealed increased expression of effector and memory TC signatures, including GZMB, GZMK, CXCR4, CCL3 and various TCR genes in the aged group (Fig. 5G; Table S12A). In addition, clonal BCs showed aging-associated changes, including increased expression of S100A family genes and decreased levels of naive signature genes such as IGHM and TCL1A in the aged BCs compared to their young counterparts (Fig. 5H; Table S 12B).
Figure 5 Abnormal TCR and BCR repertoire during aging. (A) Pie plots showing TCR clone differences across YA and AA groups. (B) Pie plots showing BCR clone differences across YA and AA groups. (C) Percentage of maximum clones between YA (n = 8) and AA groups (n = 8). (D) Diversity of TCR and BCR between YA (n = 8) and AA groups (n = 8). (E) Diversity of TCR in T cell subsets between YA (n = 8) and AA groups (n = 8). (F) Diversity of BCR in B cell subsets between YA (n = 8) and AA groups (n = 8). (G) Volcano plot showing DEGs of clonal T cells between the YA and AA groups. P values were calculated using a paired, two-sided Wilcoxon test and FDR was corrected using the Benjamini-Hochberg procedure. (H) Volcano plot showing DEGs of clonal B cells between the YA and AA groups. P values were calculated using a paired, two-sided Wilcoxon test and FDR was corrected using the Benjamini-Hochberg procedure. (I) Chord diagram showing pairing of V and J segments within the TRB subset from the AA group. Chord widths represent the proportion of sequences with a given V (colored) and J (gray) segment pairing. (J) Chord diagram showing pairing of V and J segments within the IGH subset from the AA group. Chord widths represent the proportion of sequences with a given V (colored) and J (gray) segment pairing
To further explore the aging-associated changes on V(D)J rearrangements in TC and BC, we next examined the frequency of genes (variable region) in the YA and AA groups and found that the frequency of several TRAVs, TRBVs, IGHVs, IGKVs and IGLVs changed with age (Fig. S9A–E), indicating that TCs and BCs had experienced unique clonal V(D)J rearrangements under the adaptive immune environment of the elderly. When analyzing isotype use in BCR repertoires in the YA and AA groups (Fig. S9F), we found that IGHA and IGHG were overrepresented in the AA group compared to the YA group, suggesting that aging might induce more frequent isotype switching (Fig. S9G). In addition, the chord diagrams of the V-J arrangement for each group showed that aging resulted in multiple cloning sites, suggesting increased antigen exposure with age (Figs. 5I, 5J, S9H and S9I). The enriched arrangements associated with aging were mainly TRBV6-5, TRBV20-1, and TRBV28 in the TRB subset and IGHV3-33 and IGHV5-51 in the IGH subset. Taken together, these data show that increased clonality and decreased diversity in aging immune cells, combined with a skewed use of variable region genes, underlie aging-associated abnormalities of TCR and BCR repertoires, elucidating the abnormal immune states and disease spectra during aging.

Age-related imbalance in cellular composition is associated with poor outcomes in patients with COVID-19
To depict how the immune landscape changes with aging and SARS-CoV-2 infection, we enrolled young (YCO, n = 2) and aged (ACO, n = 2) patients with incipient COVID-19 (to assess the acute inflammatory state) in cohort-2 and young (YCR, n = 2) and aged (ACR, n = 2) patients who had recovered from COVID-19 (to assess the recovered state) in cohort-3. In addition, we performed CyTOF analysis of PBMCs from YH, AH, YCO and ACO individuals in cohort-2 (n = 2 for each group) (Figs. 6A, S10 and S11). Similar to our CyTOF analysis in cohort-1, we identified 21 clusters: 9 subsets of TCs, 3 subsets of NKs, 4 subsets of BCs, 3 subsets of MCs, and 2 subsets of DCs (Figs. 6B and S11). We first compared the peripheral immune cell composition between COVID-19 patients (at the onset stage, CO) and their age-matched healthy controls (HC). Between the CO and HC groups, we found a similar trend of variation to aging, reflected in a decreased percentage of TCs and increased MC and NK populations (Fig. 6C–E). This trend was also observed at the cell subtype levels, as evidenced by decreased pDC, naive and memory TCs and BCs and increased populations of effector TCs, CD16 MCs, intermediate MCs, ASCs and ABCs (Figs. 6F, 6H, and S12A–L). Importantly, the aging-associated increase in MCs and decrease in TCs were amplified by COVID-19 in aged patients compared with healthy aged controls (Fig. 6I). This trend was also observed at the cell subtype level, as reflected by decreased naive TCs and BCs and increased populations of effector TCs, CD16 MCs, ASCs and ABCs in each immune cell composition and total circulating immune cells (Figs. 6J–N and S12M).
Figure 6 Poor outcomes upon COVID-19 infection is associated with imbalanced cellular aging. (A) t-SNE projections of PBMCs derived from mass cytometry data in cohort-2. (B) Heatmap showing mean population expression levels of all markers. (C) t-SNE plots segregated by HC and CO groups. HC includes YH (n = 2) and AH (n = 2); CO includes YCO (n = 2) and ACO (n = 2). (D) Percentage of immune cell populations in PBMC between HC (n = 4) and CO (n = 4) groups. (E) Bar chart of the relative percentage of major immune cell populations derived from mass cytometry data between HC and CO groups. (F) Percentage of CD4 Naive cells in CD45+ cells between HC (n = 4) and CO (n = 4) groups. (G) Percentage of NK2 cells in CD45+ cells between HC (n = 4) and CO (n = 4) groups. (H) Percentage of CD16 monocytes in CD45+ cells between HC (n = 4) and CO (n = 4) groups. (I) Bar chart of the relative percentage of major immune cells derived from mass cytometry data from YH, AH and ACO groups. (J) Bar chart of the relative percentage of T cell subsets derived from mass cytometry data from YH, AH and ACO groups. (K) Bar chart of the relative percentage of NK cell subsets derived from mass cytometry data from YH, AH and ACO groups. (L) Bar chart of the relative percentage of B cell subsets derived from mass cytometry data from YH, AH and ACO groups. (M) Bar chart of the relative percentage of DC subsets derived from mass cytometry data from YH, AH and ACO groups. (N) Bar chart of the relative percentage of monocyte subsets derived from mass cytometry data from YH, AH and ACO groups. (O) Bar chart of the relative percentage of major immune cell populations derived from mass cytometry data between YCO and ACO groups. (P) CT photography showing the different evolution of Lung Ground-Glass Opacity in young and aged patients with COVID-19. CT, computed tomography. P values are based on two-tailed Mann-Whitney-Wilcoxon tests between groups
Notably, we found a higher ratio of MCs, especially inflammatory MCs, and a lower percentage of TCs in aged COVID-19 patients than young COVID-19 patients (Figs. 6O and S12N). Notably, comparative subgroup analysis demonstrated that naive BCs and pDCs were decreased in aged patients (Fig. S12O–S). The patients in cohort-2 were diagnosed with severe COVID-19 and presented with similar clinical symptoms and CT findings. Despite these similarities, the recovery and outcomes in the young and aged patients differed substantially. As was evident in high-resolution CT scans, ground-glass opacity in the lungs of young patients gradually dissipated after a period of treatment, but this parameter remained associated with extensive fluid buildup (exudation) and pleural effusion in aged patients (Fig. 6P). Infiltrating MCs can enter the lung and other organs and release abundant levels of inflammatory cytokines and chemokines, exacerbating the infection and leading to fatal outcomes. Aged COVID-19 patients had more MCs and fewer TCs than young patients, thus lowering the threshold of developing hyperinflammatory states that may trigger cytokine storms and lymphopenia.

Aging increases the expression of susceptibility genes for COVID-19, and COVID-19 enhances upregulation of aging-induced inflammatory genes
To determine how an increased MCs population and decreased TCs population at the onset of SARS-CoV-2 infection contribute to faster disease progression in the elderly at the cellular and molecular levels, we used scRNA-seq to investigate the association between aging and COVID-19. Specifically, we analyzed DEGs to explore whether differentially expressed SARS-CoV-2-related genes in aged patients could explain the impact that aging had on the susceptibility and recovery of COVID-19 patients in cohort-3 (Figs. S13A–C and S14A). ACE2 is not expressed by any blood immune cells, and recent studies have reported that CD147 (encoded by BSG), CD26 and ANPEP might be alternative cellular entry receptors for SARS-CoV-2, especially CD147, in TCs (Han et al., 2020; Qi et al., 2020; Ulrich and Pillat, 2020). Anti-CD147 antibody has been tested to treat COVID-19 patients with promising effects (Bian et al., 2020). We found that BSG expression in the AH group was increased in TCs, BCs and DCs, while ANPEP was only upregulated in MCs (Fig. 7A). Moreover, we found that aging increased the frequency of immune cells that expressed BSG and ANPEP (Fig. 7B). This result was validated using flow cytometry analysis, which showed increased CD147 expression in CD3+ TCs in the aged people compared with the young group (P = 0.0010, Fig. 7C). We further observed higher expression of the CD147-related genes NFATC1, ITGB1, and PPIB in CD4 Naive of the AH group (Fig. S14B–D). In addition, CD26 (encoded by DPP4), another potential SARS-CoV-2 receptor (Radzikowska et al., 2020), was only upregulated in CD4 Naive of the AH group (Fig. S14E). Altered expression of these molecules in circulating immune cells, especially in CD4 Naive, with age might contribute to increased susceptibility and severity of COVID-19 in the elderly.
Figure 7 Aging and SARS-CoV-2 infection are characterized by similar hyper-inflammatory states. (A) Dot plot showing increased BSG and ANPEP expression in major immune cell populations in the AH group compared to YH group. P values are based on two-tailed Mann-Whitney-Wilcoxon tests between groups. (B) Expression levels of BSG and ANPEP in specific cell types in YH and AH groups. (C) Recapitulative graph of the MFI of CD147 expression in CD3+ T cells. MFI, mean fluorescence intensity. (D) Bar charts of the relative percentage of major immune cell populations derived from scRNA-seq data in YH, AH and ACR group (left), YCR and ACR group (right). (E) Venn diagram showing the integrated comparative analysis of upregulated DEGs in T cells between YH and AH group, YCR and YH group, ACR and AH group. The count shows the number of DEGs. (F) Venn diagram showing the integrated comparative analysis of upregulated DEGs in monocytes between AH and YH groups, YCR and YH groups, ACR and AH groups. The count shows the number of DEGs. (G) Volcano plot showing DEGs in T cells between YCR and ACR groups. P values were calculated using a paired, two-sided Wilcoxon test and FDR was corrected using the Benjamini-Hochberg procedure. (H) Volcano plot showing DEGs in monocytes between YCR and ACR groups. P values were calculated using a paired, two-sided Wilcoxon test and FDR was corrected using the Benjamini-Hochberg procedure. (I) Dot plot showing expression levels of the top 20 aging-induced and disease-associated genes in T cells per group in cohort-3. (J) Dot plot showing expression levels of the top 20 aging-induced and disease-associated genes in monocytes per group in cohort-3. (K) CT photography showing the different manifestation of evolution of Lung Ground-Glass Opacity in young and aged patients with COVID-19
Upregulation of SARS-CoV-2-related genes in aged individuals indicates that aging increases susceptibility to this infection. Our data demonstrated that the inflammatory response was sustained in the blood environment of COVID-19 patients recovering from SARS-CoV-2 infection (Wen et al., 2020). In the recovery stage, the aging-associated increase in MCs and decrease in TCs were amplified by COVID-19 in aged covered patients compared with healthy aged controls. Importantly, the ACR group still had more MCs and fewer TCs than the YCR group (Figs. 7D and S14F). To compare the effects of age on disease recovery, we next analyzed the upregulated DEGs between the YCR and YH groups and between the ACR and AH groups, along with combined analysis of the upregulated aging-related DEGs (Table S 13-14). We identified aging-induced and disease-associated genes in TCs, including CD69, JUNB, CDKN2A, and IFN-related genes, including IRF1 and ISG15 (Fig. 7E). In MCs, we identified several aging-induced genes involved in disease development, such as TNF, IL1B, JUNB, DUSP2, OSM, the CDKN family, IFN-related genes, and chemokine family members (Fig. 7F). Analysis of the DEGs between TCs of the YCR and ACR groups demonstrated that several granzyme genes and inflammatory genes were increased in aged recovered patients (Fig. 7G). Moreover, we found that MCs in the ACR group had increased expression of inflammatory genes such as FOS, DUSP1, IL1B, and JUN and chemokines including CXCL8 and CCL3 compared to those of the YCR group (Fig. 7H). We finally compared the expression of the top 20 specific aging-induced and disease-associated genes among the 4 groups in MCs and TCs, respectively (Fig. 7I and 7J). The results showed that COVID-19 amplifies aging-induced upregulation of inflammatory genes and senescence hallmark genes (CDKN family) (López-Otín et al., 2013). As expected, although the initial clinical manifestations and diagnosis were similar, lung ground-glass opacity in young patients had been dissipated and absorbed completely, but in aged patients, it was not absorbed completely at one week after a negative nucleic acid test (Fig. 7K). These findings indicate that aged people have a slower recovery from COVID-19 than young people.
We next predicted cell-to-cell interactions that might contribute to the distinct functional status of circulating TCs and BCs of the YCR and ACR patients (Fig. S15). In ACR patients, we discovered that TCs expressed high levels of IFNG, the ligands for IFNGR1, which was expressed on MCs (Fig. S15A). Other TC-MC interactions involved the inflammatory response, cell-cell signaling and cell adhesion. Notably, TCs might activate MCs through the expression of CCL5 ligands that bind to CCR1 and contribute to inflammatory activation. Interestingly, TCs in the ACR patients expressed high levels of IL-4, which was predicted to bind IL-4R and IL-2R in TC-MC interactions and was reported to enhance viral infection (Rogers et al., 2019). In the ACR group, BCs expressed increased levels of genes encoding ligands of IL1R and TNFRSF1B (Fig. S15B). The expression of these molecules in MCs suggests that BCs may contribute to the activation of IL1B and TNF signaling in circulating MCs. Compared with the ACR group, the YCR group was characterized by the presence of signals that negatively regulate the inflammatory response molecules IL10-IL10RA in TC or BC interactions with MCs. Downregulation of negative regulatory signals may also contribute to the slow dissipation of inflammation in the elderly. Overall, enhanced inflammatory signals and impaired regulatory signals between TCs and MCs, or, BCs and MCs, slow recovery in elderly patients.