The onset of immune aging varies across individual mice

To understand aging variation across individuals, we studied C57BL/6 J mice, the most widely used inbred strain. We measured the immune aging phenotype by calculating the BG ratio in the peripheral blood (Supplementary Fig. 1A). As expected, the mice generally presented increased levels of granulocytes and decreased levels of B cells at 30 months of age (Fig. 1A). However, we found substantial variations in the BG ratios across individuals (Fig. 1B). At 30 months of age, many mice presented BG ratios lower than those observed in 4-month-old mice. However, some 30-month-old mice maintained ratios similar to those of 4-month-old mice and thus lacked the expected aging phenotype (Fig. 1B). The variation across individuals was also evident at 16 months of age, when a smaller portion of individuals presented the aging phenotype (Fig. 1B).

Fig. 1

The onset of immune aging varies across individual mice. A Granulocyte and B-cell abundances in the peripheral blood of 4-, 16- and 30-month-old mice (n = 15, 16 and 13, respectively). B Individual mice from A are ranked on the basis of their ratios of B-cell abundance to granulocyte (Gr) abundance. The dashed line marks the lower boundary of the ratios from 4-month-old mice, which is used to discriminate between early and delayed aging. C Granulocyte and B-cell abundances in the peripheral blood of early and delayed aging mice. A, C Data are presented as the means ± SEMs. *P < 0.05, **P < 0.01, ***P < 0.001 compared between the early and delayed aging groups by two-tailed Student’s t test; WBC, white blood cells. D UMAP density plot of HSCs from early  and delayed aging mice. Each dot represents one cell. E Unsupervised clustering of HSCs identified 17 clusters. The top 3 clusters most enriched by HSCs from early aging mice or those from delayed aging mice are highlighted in different colors. F Gene set enrichment analysis (GSEA) profiles showing significant enrichment of genes associated with aging and telomere organization in the clusters overrepresented by HSCs from early aging mice. NES normalized enrichment score, FDR false discovery rate

Compared with 4-month-old mice, 30-month-old mice with reduced BG ratios had significantly greater levels of granulocytes and significantly lower levels of B cells (Fig. 1C). In contrast, 30-month-old mice with BG ratios similar to those of 4-month-old mice did not show significant shifts in B-cell or granulocyte abundance (Fig. 1C). Similar results were found when Mac1+ myeloid cells and T cells were analyzed (Supplementary Fig. 2). T cells were not extensively analyzed in this study because of their distinctive maturation and expansion in the thymus, as well as their distinct clonality in comparison to that of HSCs [35]. Our data demonstrate that variation exists in the onset of the immune aging phenotype among individual mice with the same genetic background and housing conditions. For subsequent analyses, we classified individual mice into two groups, the early and delayed aging groups, on the basis of whether their BG ratios at an advanced age fell below or within the range observed in mice at 4 months of age (Fig. 1B). There was no significant sex disparity between the early aging and delayed aging groups (Supplementary Table 1). By identifying elderly individuals exhibiting or lacking the immune aging phenotype at the same chronological age, we can investigate the mechanisms triggering the onset of the immune aging phenotype.

Distinct subsets of HSCs in early and delayed aging mice differentially express aging-associated genes

To identify the gene expression signature associated with the variation in the onset of immune aging, we compared the transcriptomes of individual HSCs from 30-month-old mice exhibiting or lacking the aging phenotype (Supplementary Fig. 3A, Supplementary Table 2). While the transcriptomes of most HSCs were similar between the early and delayed aging groups (Fig. 1D), our unsupervised clustering analysis identified a few clusters that were overrepresented by HSCs from one group (Fig. 1E, Supplementary Fig. 3B to F, Supplementary Table 3). We refer to the top three clusters that were most enriched with early aging HSCs as early aging clusters and the top three clusters enriched with delayed aging HSCs as delayed aging clusters. A comparison of their gene expression profiles revealed significant overrepresentation of genes associated with aging and telomere organization in the early aging clusters (Fig. 1F). A comparison of the gene expression profiles of either early or delayed aging HSCs with those of HSCs from young mice revealed that the difference between early aging HSCs and young HSCs was substantially greater than the difference between delayed aging HSCs and young HSCs (Supplementary Fig. 4). This result indicates that among mice of the same chronological age, gene expression differences in a subset of HSCs align with the aging phenotype observed in immune cell counts in the peripheral blood.

Distinct regulators of hematopoiesis are associated with early and delayed immune aging

Many genes upregulated in the early aging HSC clusters are involved in regulating myeloid differentiation (Supplementary Fig. 3G and Supplementary Data 1), including those of the top 20 most upregulated genes (Fig. 2A). In particular, Itga2b (CD41), the top-ranked upregulated gene, has been previously associated with HSC aging [36]. CD41-positive HSCs are known to exhibit myeloid bias [36] and increased expression of Gata1, which is also upregulated in the early aging HSC clusters (Supplementary Fig. 3G). Moreover, another upregulated gene, TGF-β1, has been shown to regulate HSC proliferation and stimulate myeloid-biased HSCs while inhibiting the growth of lymphoid-biased HSCs [37,38,39]. Additional regulators of myelopoiesis associated with early aging clusters include Cd9 [40], Hmgb2 [41], Hmgb3 [42], and Zeb2 [43]. The latter three genes, in addition to Cdk6 [44], have also been shown to play roles in regulating HSC proliferation. Therefore, some of the genes whose expression was most upregulated in the early aging HSC clusters were associated with myeloid differentiation and HSC proliferation.

Fig. 2

Variation in the onset of immune aging is associated with distinct gene expression characteristics of HSCs. A The top 20 DEGs in the clusters enriched with HSCs from early or delayed aging mice are shown (Fig. 1E). The dot size indicates the fraction of the HSCs expressing the gene. The color indicates the fold change in gene expression between the clusters of interest and all other clusters. B Gene Ontology enrichment and transcription factor binding motif analyses of upregulated genes in clusters enriched with HSCs from early and delayed aging mice. The circle size corresponds to the term size. The number of significantly enriched terms in each category is shown in parentheses. MF, molecular function; BP, biological process; CC, cellular component; TF, transcription factor. Selected terms related to hematopoiesis are listed in the table below. C Semantic similarity REViGO scatterplot of the top 100 GO:BP terms reveals that early aging is associated primarily with increased stem cell proliferation, whereas delayed aging is associated primarily with increased stem cell regulation and response to external signals. D Primary mouse HSCs from Cas9-expressing transgenic mice were transduced with lentivirus carrying a mixture of sgRNAs targeting each gene. A non-targeting (NT) sgRNA mixture was used as a control. The ratios of B cells (CD19+ B220+ ) to granulocytes (Mac1+ Gr1+) were analyzed after 10 days of co-culturing HSCs with OP9 cells. Two or more biological replicates with a total of 14 or more replicates per gene were performed. The data are shown as the means ± SEMs. Two-tailed Student’s t test. E BG ratio of hematopoietic cells in the peripheral blood derived from ckit high and ckit low donor HSCs. One-tailed Student’s t test was used. *P < 0.05; **P < 0.01; ***P < 0.001

Among the top 20 most upregulated genes in the delayed aging clusters (Fig. 2A), many are transcriptional regulators, including Hlf, Mecom, Nfat5, and Mllt3, and long non-coding RNAs, including Malat1, Neat1 and Linc-pint [45,46,47,48,49,50,51]. In particular, the long non-coding RNA Linc-pint is associated with healthy aging in mice, and knockout mice exhibit signs of premature aging in multiple tissues [52]. Additionally, activated leukocyte cell adhesion molecule (ALCAM) has been shown to counteract myeloid skewing, and HSCs in knockout mice exhibit aging-like phenotypes [53]. Therefore, some of the most upregulated genes in the delayed aging HSC clusters have already been shown to offset aging phenotypes.

Variation in the onset of immune aging is associated with increased stem cell proliferation and decreased stem cell regulation through external signals

To comprehensively investigate the molecular mechanism underlying the variation in the onset of immune aging, we analyzed the biological implications of all genes that were significantly differentially expressed between the early and delayed aging clusters (Fig. 1E). We identified twice as many gene ontology (GO) terms and transcription factor binding motifs among genes expressed at significantly greater levels in the delayed aging clusters than in the early aging clusters (Fig. 2B and Supplementary Data 2). We focused on terms relevant to hematopoiesis and found that both the early and delayed aging clusters presented upregulated genes associated with GO terms related to hematopoiesis, myeloid cell differentiation, and lymphocyte differentiation, since these GO terms do not distinguish between positive and negative regulation. However, we also found substantial differences. For example, the early aging clusters were particularly enriched for genes associated with megakaryocyte differentiation and the regulation of platelet activation, whereas the delayed aging HSC clusters were uniquely enriched for genes associated with the regulation of T-cell differentiation and hematopoietic stem cell homeostasis (Fig. 2B).

Overall, genes upregulated in the early aging clusters were enriched in biological processes related to the stress response and stem cell proliferation, such as the regulation of the cell cycle (Fig. 2C). Genes that were upregulated in the delayed aging clusters were enriched in biological processes related to stem cell regulation and response to external signaling, such as cell communication and localization (Fig. 2C), both of which are fundamental for stem cell regulation. Taken together, these results suggest that the onset of immune aging is associated with increased stem cell proliferation and reduced control over stem cell homeostasis and differentiation, particularly in response to external signaling.

Functional roles of DEGs between early aging and delayed aging HSC clusters

To investigate the functional roles of genes significantly differentially expressed between the early and delayed aging clusters, we performed CRISPR knockout assays using primary mouse HSCs co-cultured with OP9 cells to facilitate both myeloid and lymphoid differentiation. This analysis was conducted on genes with undefined roles in hematopoiesis. Our results revealed that knocking out genes upregulated in early aging clusters, such as Lgals9, Nme1, and Slc25a5 (Supplementary Fig. 3H), led to an increase in the BG ratio (Fig. 2D), indicating that these genes contribute to reducing the BG ratio during aging. Conversely, knocking out genes upregulated in the delayed aging clusters, including Nedd4 and Prex2 (Supplementary Fig. 3I), resulted in a decreased BG ratio (Fig. 2D), suggesting their roles in maintaining the BG ratio in delayed aging mice. Additionally, ckit was significantly upregulated in delayed aging mice (Supplementary Fig. 3I). Therefore, we transplanted HSCs expressing ckit at high and low levels and found that the BG ratio derived from ckithigh HSCs was significantly greater than that derived from ckitlow HSCs (Fig. 2E). These in vitro and in vivo functional assays collectively demonstrated that genes differentially expressed in early and delayed aging HSC clusters play crucial roles in regulating immune cell production.

Tracking individual HSC clones in early and delayed aging mice

Since transcriptomic analysis revealed that differences in immune aging between early aging and delayed aging mice reside in a subset of HSCs (Figs. 1D–F, and 2), we further investigated how a subset of HSCs triggers the immune aging phenotype of an organism. To identify and characterize the HSC clones that drive the immune aging phenotype, we compared the immune cell production of individual HSC clones before and after the onset of immune aging. We isolated and labeled HSCs from young donor mice via unique genetic barcodes [29, 54] and transplanted them into young recipients (Fig. 3A, Supplementary Fig. 5 and Supplementary Table 1). The peripheral blood of recipient mice was analyzed starting from an initial time point 4 months post-transplantation, when hematopoiesis had stabilized at the cell population level [55, 56], to an end time point 15 months post -transplantation or 12 months post- transplantation if the mouse reached its end-of-life before the 15-month mark (Fig. 3B–D and Supplementary Fig. 5A–F). Similar to our analysis of naïve mice (Fig. 1), we classified individual mice into either the early or delayed aging group on the basis of their BG ratios at the end time point compared with the range of observed ratios for all mice at the initial time point (Fig. 3B and Supplementary Table 2). Mice in the delayed aging group may also develop the aging phenotype given additional time (Supplementary Fig. 5).

Fig. 3

Tracking individual HSC clones in early and delayed aging mice. A Experimental design for tracking the immune cell production of individual hematopoietic stem cell (HSC) clones during aging. B Individual mice (n = 30) are ranked on the basis of their ratios of B-cell abundance to granulocyte (Gr) abundance (BG ratios) at the end time point. The dashed line marks the lower boundary of the corresponding ratios at the initial time point, which is used to discriminate between early and delayed aging. C Changes in the BG ratio at the population level over time in early (n = 13) and delayed aging (n = 12) mice. The dashed line is the same as in B. D Granulocyte and B-cell abundances in the peripheral blood over time. C, D One-way ANOVA with the Bonferroni-corrected paired t test was used to compare different time points; **P < 0.01 between the initial and end time points. Some mice died prior to 15 months post-transplantation, and their data at 12 months are shown separately to illustrate their end time points. E Kaplan‒Meier survival plots of early and delayed aging mice. C, D Data are presented as the means ± SEMs; N.S. not significant, WBC white blood cells

The immune aging phenotype of the mice in the early aging group started to develop after 9 months post-transplantation (Fig. 3C, D). Thus, we refer to 9 months post-transplantation as the pre-divergent time point. Like naïve non-transplanted mice (Fig. 1C), early aging mice had significantly higher levels of granulocytes and significantly lower levels of B cells at the end time point than at the initial time point, whereas delayed aging mice did not exhibit any substantial changes in granulocyte or B-cell abundance over time (Fig. 3D). Moreover, delayed aging mice tended to live longer than early aging mice (Fig. 3E). Compared with those in the delayed aging group (31.8%), almost twice as many mice in the early aging group (59%) became sick or died prior to the predetermined end point of the experiment. These findings indicate a correlation between mouse lifespan and the onset of immune aging, as delineated by the BG ratio in our analysis.

Delayed onset of aging is associated with reduced myeloid production in a subset of HSC clones

To identify the specific changes in HSC clones that drive the immune aging phenotype, we compared their immune cell production at the pre-divergent time point, right before the onset of the aging phenotype, and at the end time point, when approximately half of the mice displayed the aging phenotype (Fig. 3B–D). Between the pre-divergent and end time points, the immune cell production of most HSC clones remained unchanged (Fig. 4A). The HSC clones whose immune cell production changed exhibited different characteristics between early aging and delayed aging mice. For example, the number of HSC clones that reduced B-cell production was significantly greater in early aging mice than in delayed aging mice (Fig. 4B). Among the HSC clones that increased B-cell production, the overall increase in B-cell abundance was significantly greater in delayed aging mice than in early aging mice (Fig. 4C). In the myeloid lineage, the number of clones with reduced granulocyte production was significantly greater in delayed aging mice than in early aging mice (Fig. 4B). The amount of overall granulocyte reduction was significantly greater in delayed aging mice, and the amount of overall granulocyte increase was significantly greater in early aging mice (Fig. 4C). Taken together, these data demonstrate differences in immune cell production among distinct subsets of HSC clones when early and delayed aging mice are compared.

Fig. 4

The delayed onset of immune aging is associated with reduced myeloid production. A Changes in B-cell and granulocyte production of individual HSC clones between the pre-divergent (9 months post-transplantation) and the end time points. Each circle indicates one HSC clone. Both axes are shown on a “symlog” scale with a linear scale ranging between 0 and 0.1. B The number of HSC clones that changed granulocyte or B-cell production in early and delayed aging mice. C Changes in the overall immune cell production of HSC clones that exhibited distinct changes in each mouse. B, C Clones with changes of less than 0.1% white blood cells were not considered. Bonferroni adjusted Wilcoxon rank-sum test. *P < 0.05, **P < 0.01. D STEM clustering [70] of HSC clones on the basis of changes in granulocyte and B-cell production over time. Each box shows a cluster of HSC clones, depicted as red lines, that share similar temporal dynamics, depicted as a black line. The colored boxes highlight clusters consisting of significantly more HSC clones than expected. The numbers above each box indicate the percentage of all clones in the corresponding cluster followed by the P value, indicating that the cluster consists of a greater number of clones than expected. N.S. not significant where P > 0.05. B, C Data are presented as the mean ± SEM; WBC white blood cells

The decrease in granulocyte production was unexpected, as granulocyte abundance generally increases with age (Fig. 1A and Supplementary Fig. 5B). However, a trend toward granulocyte reduction was detected in delayed aging mice at the population level (Figs. 1C and 3D). Moreover, the abundance of common myeloid progenitors was also significantly lower in delayed aging mice, but not early aging mice, than in young mice (Supplementary Fig. 5H). In addition, we analyzed the dynamic changes of HSC clones in granulocyte and B-cell production over time and found that the most significantly enriched temporal dynamics in granulocyte production were increasing in early aging mice and decreasing in delayed aging mice (Fig. 4D). Taken together, these findings suggest a compelling and surprising connection between reduced myeloid production and aging delay.

The onset of immune aging is associated with distinct shifts in the lineage preferences of a subset of HSC clones

Because immune aging manifests as concurrent changes in both myeloid and lymphoid lineages, we assessed the changes in lineage bias of HSC clones from the pre-divergent time point to the end time point. The aggregated data revealed that many HSC clones maintained their lineage bias during aging, forming a normal distribution centered around zero change (Fig. 5A). However, some clones exhibited changes in lineage bias to varying degrees (Fig. 5A). In all the mice, 60–70% of the HSC clones remained lineage stable (Fig. 5B, C). Among the HSC clones whose lineage preferences were altered, some shifted toward the myeloid lineage in line with the immune aging phenotype. Surprisingly, others shifted toward the lymphoid lineage and counteracted the aging process (Fig. 5B). We refer to the former HSC clones as “aging clones” and to the latter as “anti-aging clones”. Aging HSC clones were significantly more frequent in early aging mice, and anti- aging HSC clones were significantly more frequent in delayed aging mice (Fig. 5C). In addition, HSC clones with distinct lineage preferences presented different frequencies of lineage shifts. In particular, all the myeloid-biased HSC clones remained lineage stable, and all the anti-aging clones initially exhibited lineage balance (Fig. 5D).

Fig. 5

The temporal variation of immune aging is associated with distinct changes to HSC lineage biases. A Distribution of changes in the lineage bias of all HSC clones between the pre-divergent (9 months post-transplantation) and the end time points of the mice. A Gaussian kernel density estimate of the distribution is shown. The light gray area represents the distribution of all the clones. The blue curve shows the Gaussian fit for lineage stable clones. The remaining clones are highlighted by the pink curve. The black vertical dashed line shows the cutoff between lineage stable and lineage shifting clones that is set at the intersection of the black and pink curves. B Numbers of lineage stable and lineage shifting clones in the early and delayed aging groups. The dashed lines illustrate the thresholds that separate lineage stable and lineage shifting clones, as determined in A. Clone numbers are shown on a “symlog” scale with a linear scale ranging between 0 and 50. C Fractions of lineage stable, anti-aging and aging clones in early (n = 9) and delayed (n = 9) aging mice. D Heatmap showing the fraction of lymphoid-biased, myeloid-biased, and balanced clones whose lineage biases had shifted or remained stable by the end time point. The lineage bias category is determined by data from the pre-divergent time point. The median of all the mice in each group is shown. E Contribution of lineage stable and lineage  shifting clones to granulocytes and B cells in the peripheral blood at the end of the study. C, E Data are presented as the mean ± SEM; one-way ANOVA followed by Tukey’s HSD pairwise comparison within a group; Bonferroni-adjusted independent Student’s t test between groups. *P < 0.05, **P < 0.01, ***P < 0.001

To determine the functional significance of the lineage stable and lineage shifting HSC clones, we quantified their immune cell contributions. Our data revealed that 60–80% of myeloid cells were produced by lineage stable clones (Fig. 5E and Supplementary Fig. 6A), which aligns with the corresponding clone number (Fig. 5C). Surprisingly, lymphoid production depended on both lineage stable clones and lineage shifting clones to a similar degree (Fig. 5E and Supplementary Fig. 6A). These findings suggest that the small number of lineage shifting clones plays an important role in lymphoid production. In particular, anti-aging HSC clones produced substantial numbers of immune cells in both lineages in delayed aging mice, whereas their contribution to immune cell production in early aging mice was negligible (Fig. 5E and Supplementary Fig. 6A). These findings underscore the pivotal role of these clones in delaying the onset of the immune aging phenotype.

The discovery of aging-associated lineage shifting clones challenges the conventional paradigm that HSCs retain their differentiation characteristics and that daughter HSCs maintain their parental, epigenetically-defined lineage preferences [2, 57]. In addition, we found that the relative HSC abundances of myeloid-biased, balanced, and lymphoid-biased clones were similar between young and old mice and between early and delayed aging mice (Supplementary Fig. 6B), indicating that differential growth at the HSC level is not involved in triggering the immune aging phenotype. These data contradict the hypothesis that immune aging originates from shifts in the relative proportions of HSCs that exhibit different lineage biases [15, 58].

Anti-aging clones shift their lineage bias toward the lymphoid lineage by reducing myelopoiesis

Despite contributing significantly to B-cell production (Fig. 5E and Supplementary Fig. 6A), anti-aging HSC clones, on average, did not increase their B-cell production from the pre-divergent time point to the end time point (Fig. 6A). Instead, their myeloid production was significantly reduced, resulting in an apparent lineage bias toward the lymphoid lineage (Fig. 6A). A comparison of the granulocyte and HSC clonal abundances at the end time point revealed that anti-aging clones presented significantly lower levels of myeloid differentiation compared to other clones (Fig. 6B). The HSC abundance of these clones was not significantly different from that of the other clones (Supplementary Fig. 6C). These data suggest that the reduced myeloid production of the anti-aging HSC clones primarily arises from changes in HSC differentiation rather than self-renewal. This result challenges the clonal expansion aging model that was derived primarily from studies comparing young and old mice [6, 26, 59].

Fig. 6

Age-associated imbalance of the innate and adaptive immune systems is driven primarily by changes in myelopoiesis. A Boxen plots showing changes in the immune cell production of HSC clones between the pre-divergent and end time points. B Quantification of the myeloid differentiation of lineage stable and lineage shifting HSC clones. C Boxen plots showing changes in immune cell production of lineage stable HSC clones between the pre-divergent and end time points. The lineage bias category is based on data from the end time point. A, C Each horizontal black bar denotes the mean of all the clones in each group. Independent Student’s t test was used for testing differences from 0; the Bonferroni-adjusted Wilcoxon rank-sum test was used between groups. D Quantification of the myeloid differentiation of aging clones and lineage stable myeloid-biased HSC clones. B, D Data are presented as the means ± SEMs; B, D Wilcoxon rank-sum test. Bonferroni correction was applied to D. *P < 0.05, **P < 0.01, ***P < 0.001. N.S. not significant, WBC white blood cells

While most clones decreased their lymphoid production over time to various degrees, significant increases in myeloid production were detected in aging clones and lineage stable myeloid-biased clones from early aging mice (Fig. 6A, C). In contrast, the myeloid production of these clones did not increase with time in delayed aging mice (Fig. 6A, C). Moreover, these clones had significantly greater levels of myelopoiesis in early aging mice than in delayed aging mice when we compared granulocyte and HSC clonal abundance (Fig. 6D). However, their HSC abundances were not significantly different (Supplementary Fig. 6D, E). These results indicate that increased myeloid production with age primarily arises from changes in HSC differentiation rather than self-renewal. Taken together, our findings suggest that changes in myelopoiesis underlie immune aging.

Clonal expansion, exhaustion, and activation in the aging of immune cell regeneration

Clonal expansion and exhaustion are thought to play important roles in several aging models [1, 6, 19, 25, 26]. We quantified the abundance of the most abundant clones at the end time point and found that while early and delayed aging mic