Clonally expanded memory CD8⁺ T cells accumulate in atherosclerotic plaques and are pro-atherogenic in aged mice

Abstract

Aging is a strong risk factor for atherosclerosis and induces accumulation of memory CD8⁺ T cells in mice and humans. Biological changes that occur with aging lead to enhanced atherosclerosis yet the role of aging on CD8⁺ T cells during atherogenesis is unclear. Here we found that depletion of CD8⁺ T cells attenuated atherogenesis in aged, but not young mice. Furthermore, adoptive transfer of splenic CD8⁺ T cells from aged wild-type, but not young wild-type donor mice, significantly enhanced atherosclerosis in recipient mice lacking CD8⁺ T cells. We also characterized T cells in healthy and atherosclerotic young and aged mice by single-cell RNA-sequencing. We found that specific subsets of age-associated CD8⁺ T cells including a granzyme K⁺ effector memory subset that accumulated and was clonally expanded within atherosclerotic plaques. These had transcriptomic signatures of T cell activation, migration, cytotoxicity, and exhaustion. Overall, our study identified memory CD8⁺ T cells as therapeutic targets for atherosclerosis in aging.

Introduction

Atherosclerosis is the underlying pathology leading to myocardial infarction and stroke which account for ~31% of deaths worldwide and are the leading cause of death in the United States^{1, 2}. Atherosclerosis is a chronic inflammatory disease driven by hypercholesterolemia characterized by accumulation of leukocytes in the arterial wall, which lead to foam cell accumulation². The strongest risk factor for atherosclerosis is age, and >80% of those who die from cardiovascular diseases are >65 years of age¹. Aging also enhances atherosclerosis in a murine models³.

Lymphocytes were discovered in human atherosclerotic plaques over 30 years ago⁴ and play a key role in development of atherosclerosis⁵. Murine studies in young mice have elucidated that regulatory T cells are athero-protective⁶, CD4⁺ Th1 cells are pro-atherogenic⁷, and the role of CD4⁺ Th2 cells is context-dependent.⁸ The role of CD8⁺ T cells in murine atherosclerosis models is less clear. In one study, germline knockout of Cd8 showed no effect on atherosclerotic lesion size in Apoe^−/− mice⁹. Subsequent studies depleted CD8⁺ T cells and show they are pro-atherogenic in Ldlr^−/− and Apoe^−/− mice^{10, 11}. Hence, the role of CD8⁺ T cells in atherosclerosis pathogenesis is unclear even in young hosts and unknown in aged hosts.

T cells change substantially during aging¹². Thymic involution leads to reduction in naïve CD4⁺ and CD8⁺ T cell production which coincides with increasing memory subsets during aging^{12, 13,14}. How aging impacts the transcriptome of T cell subsets and how these contribute to atherosclerosis is unknown. Given the significant age-related changes in T cell populations and the discrepant roles CD8⁺ T cells play during atherogenesis, we aimed to clarify the role of CD8⁺ T cells in atherosclerosis in the context of aging.

Results

CD8⁺ T cells infiltrate atherosclerotic plaques in aging

We treated disease-free young (2–3 months) and aged (18–19 months) C57BL/6N wild type (WT) female mice with an adeno-associated virus to overexpress proprotein convertase subtilisin/kexin type 9 serine protease (PCSK9-AAV) and then switched the mice from low-fat diet (LFD) to a western-diet (WD) for 10-weeks (Fig. 1A) to generate a model of atherosclerosis¹⁵. The AAV treatment alone does not significantly alter T cell subsets from baseline through 3-weeks (Fig. S1). We previously demonstrated that aged mice develop enhanced atherosclerotic plaques with no significant impact on fasting total low-density lipoprotein cholesterol (LDL)³. Here, we show that aged atherosclerotic mice have a greater number of CD8⁺ T cells within plaques compared with young mice (Fig. 1B).

Fig. 1: CD8⁺ T cell depletion reduces atherosclerotic plaque size in Aged WT mice but not young WT mice.

(A) Schematic of murine PCSK9-AAV-induced hypercholesterolemia model of atherosclerosis and paradigm showing PCSK9 treatment protocol in young and aged female C57BL/6N WT mice. (B) CD8⁺ immunostaining in cross-section of one leaflet of the aortic sinus in young and aged C57BL/6N atherosclerotic mice with quantification of the number of CD8⁺ T cells normalized to total plaque area. (C) Schematic of murine PCSK9-AAV-induced hypercholesterolemia model of atherosclerosis and paradigm showing PCSK9 treatment protocol in C57BL/6N WT mice with anti-CD8 and anti-IgG antibody injection every 2 weeks during the WD feeding period. Representative flow cytometry plots demonstrate the reduction of CD8⁺ T cells in anti-CD8 treated mice compared to isotype control, and a line graph of serial flow cytometry data on the blood of mice demonstrates the mean reduction in CD8⁺ T cells in the anti-CD8 treatment group compared to the isotype control treatment group. Representative histology and atherosclerotic plaque size and necrotic core size quantification of the aortic sinus of young PCSK9-AAV (D), aged PCSK9-AAV (E), and Ldlr^−/− mice (F) treated with either anti-CD8 antibody treatment or anti-IgG treatment. Dashed lines indicate the atherosclerotic plaque outline and dotted lines indicate the acellular necrotic core. Scale bar = 100μm. In B, each point is a biological replicate and measurements were taken from distinct samples. In B and D-F, data are pooled from 3 independent experiments and all data are shown. In C, data were independently replicated in 3 experiments and one experiment is shown. In B and D-F, for young anti-IgG N=10, for young anti-CD8 N=11, for aged anti-IgG N=17, for aged anti-CD8 N=16, for young Ldlr^−/− anti-IgG N=7, for young Ldlr^−/− anti-CD8 N=6 over 3 independent experiments. In C, N=5 per group over 3 independent experiments (only 1 experiment shown). Data in this figure represents mean +/− SEM. 2-tailed Students T test was used for B. 2-Way ANOVA was used for C and D-F with Tukey’s post-hoc test. * = P<0.05, ** = P<0.01, *** = P<0.001, **** = P<0.0001. CM = central memory, EM = effector memory, PCSK9 = proprotein convertase subtilisin/kexin type 9 serine protease.

We investigated differences in CD8⁺ T cell populations by flow cytometry in the blood in young C57BL/6N WT, aged C57BL/6N WT, and young Ldlr^−/− mice, an established model for atherosclerosis (Extended Data Fig. 1 and 2). For WT mice, we employed the same PCSK9-AAV approach as previously³, and for all groups, we fed a WD for 10-weeks and performed flow cytometry at baseline, week 5, and week 10. Aged WT mice have fewer circulating naïve CD8⁺ T cells and more circulating effector memory (EM) and central memory (CM) CD8⁺ T cells than young WT mice (Extended data Fig. 1). The largest proportion of circulating CD8⁺ T cells in aged mice were CM CD8⁺ T cells (~20% of CD3⁺ T cells), followed by EM CD8⁺ T cells (~15% of CD3⁺ T cells) (Extended data Fig. 1). We used PD1 and Tox as markers of exhaustion and surrogate markers of a recently described population of age-associated Gzmk-expressing EM CD8⁺ T cells (CD8⁺ EM Gzmk)¹⁶, along with a polyclonal anti-Granzyme K antibody. This population is related to cancers^{17, 18, 19} and aging^{16, 20} but their association with atherosclerosis is unknown. We found that aged WT mice have more circulating CD8⁺ EM Gzmk cells than young WT mice; however, these were small proportions within the CD3⁺ T cell pool even in aged mice, approximately <2% (Extended data Fig. 1 and 2). Young Ldlr^−/− mice have a similar number of naïve CD8⁺ T cells as young WT mice, but young Ldlr^−/− mice have significantly greater frequency of circulating CM CD8⁺ T cells than young WT mice (Extended Data Fig. 1). This demonstrates that young Ldlr^−/− mice share phenotypic features of both young and aged WT mice.

Depleting CD8⁺ T cells attenuates atherogenesis in aging

To determine the necessity of CD8⁺ T cells for atherogenesis with aging, we used a monoclonal anti-CD8 antibody to deplete CD8⁺ T cells compared to isotype control, every 2 weeks during the 10-week WD feeding in young and aged C57BL/6N WT PCSK9-AAV mice (Fig. 1C). CD8⁺ depletion significantly reduced the number of circulating CD8⁺ T cells in the blood during atherogenesis (Fig. 1C). Young Ldlr^−/− mice had higher fasting LDL prior to PCSK9-AAV injection in WT mice and before beginning WD feeding, but we found no difference in fasting LDL cholesterol levels between either groups 2- and 10-weeks into the WD feeding protocol (Extended data Fig. 3). CD8⁺ T cell blockade significantly reduced aortic sinus atherosclerotic lesion size, necrotic core area, and necrotic core area percentage in aged PCSK9-AAV mice compared to isotype control, but it had no effect on atherogenesis in young PCSK9-AAV WT mice (Fig. 1D and E, and Fig. S2A–C). CD8⁺ T cell blockade significantly reduced the brachiocephalic artery atherosclerotic lesion size in aged PCSK9-AAV mice compared to isotype control but had no effect on young PCSK9-AAV WT mice (Extended data Fig. 4). We demonstrate that depletion of CD8⁺ T cells significantly reduced CD8⁺ T cells within the atherosclerotic aorta in aged mice by approximately 50% (Extended data Fig. 5).

Since young Ldlr^−/− mice exhibit increased circulating CM CD8⁺ T cells similar to aged WT mice and increased number of naïve CD8⁺ T cells (Extended data Fig. 1), we determined if CD8⁺ T cell depletion impacted atherosclerosis in young Ldlr^−/− mice. CD8⁺ T cell blockade significantly reduced aortic sinus atherosclerotic plaque size, necrotic core size, and necrotic core as a percentage of total plaque area in 3-mo Ldlr^−/− mice (Fig. 1F and Fig. S2). The atherosclerotic plaque size reduction in young Ldlr^−/− mice (~26% plaque reduction) was inferior to that of aged WT mice (~67%; Figure 1D to 1F). This suggests that although CD8⁺ T cells contribute to atherogenesis in young Ldlr^−/− mice, this contribution is smaller to the impact CD8⁺ T cells play in atherogenesis in aging.

Aged CD8⁺ T cells enhance atherogenesis in young CD8^−/− mice

We next determined whether CD8⁺ T cells from young or aged C57BL/6N WT donors were more pro-atherogenic. We isolated splenic CD8⁺ T cells from young or aged disease-free C57BL/6N WT donors (Fig. 2A). We examined the composition of cells after CD8⁺ enrichment and found that most cells enriched from the young spleens were naïve CD8⁺ T cells compared to more CM and EM CD8⁺ T cells from aged spleens (Fig. 2B and Extended data Fig. 6). EM and CM CD8⁺ T cells enriched from aged spleens included a significant enrichment of PD1⁺ cells (Fig. 2B). We adoptively transferred CD8⁺ T cells from either young or aged donors via intravenous injection into young Cd8^−/− mice lacking CD8⁺ T cells. After 1-month we induced atherogenesis by PCSK9-AAV injection and 10-wk WD-feeding (Fig. 2A). Fasting LDL increased significantly upon WD feeding in all 3 groups with no significant differences between groups at any timepoint (Extended data Fig. 7). We identified CD8⁺ T cells in the blood of the Cd8^−/− recipient mice from young and aged donors 6-weeks after beginning WD feeding (Fig. S3). Transfer of CD8⁺ T cells from aged donors significantly enhanced aortic sinus atherosclerotic plaque size, total necrotic core size, and brachiocephalic artery atherosclerotic plaque size by approximately 3-fold, but not necrotic core size percentage compared with both transfer of CD8⁺ T cells from a young donor and in vehicle-transferred mice (Fig. 2C and D, and Fig. S4). Immunohistochemical assessment of CD8 expression identified more CD8⁺ T cells within the plaque of aged CD8⁺ T cell recipients (Fig. 2E and at higher magnification in Fig. S5).

Fig. 2: Aged CD8⁺ T cells contain more exhausted memory cells and are more pro-atherogenic compared to young CD8⁺ T cells.

(A) schematic of CD8⁺ T cell harvest from female young (3-mo) and female aged (18-mo) C57BL/6N WT donor spleens. Single-cell suspensions from young or aged female donors were pooled, purified for CD8⁺ T cells and then 10 million young or aged donor CD8⁺ T cells were intravenously injected into young female Cd8^−/− recipients, and a “no-cell” group received sterile PBS intravenously. Mice were allowed to recover for 1-month post-adoptive cell transfer prior to PCSK9-AAV transfection and 10-week WD feeding in all 3 groups. (B) Pie chart demonstrating the composition of enriched cell suspensions determined by flow cytometry as a percentage of all cells that were then injected into Cd8^−/− recipients. (C) Representative hematoxylin and eosin histology of the aortic sinus of Cd8^−/− mice that received either vehicle (i.e., no cells), CD8⁺ T cells from young donor spleens, or CD8⁺ T cells from aged donor spleens. Scale bars = 100μm. (D) Quantification of aortic sinus plaque size and necrotic core size of the 3 groups. 30 individual sections are analyzed from each mouse. (E) Immunohistochemical staining for anti-CD8 in tissue sections of the aortic sinus with arrowheads depicting positively stained CD8⁺ T cells with quantification to the right. Scale bars = 50μm. For B, all donor cells are pooled in each group from N=10 donors/age group over 2 independent experiments. For D and E, N=8 young Cd8^−/− + aged CD8 cells, N=13 young Cd8^−/− + young CD8 cells, and N=7 young Cd8^−/−+ no CD8 cells over 2 independent experiments. In E, each point is a biological replicate and measurements were taken from distinct samples. Data in this figure represents mean +/− SEM. For D, 2-way ANOVA with Tukey’s post-hoc test. For E, 1-way ANOVA with Tukey’s post-hoc test. CM = central memory, PD1 = programmed cell death protein 1, EM = effector memory, PCSK9 = proprotein convertase subtilisin/kexin type 9 serine protease.

CD8⁺ T cells in atherosclerotic plaques change during aging

Since CD8⁺ T cells were necessary and sufficient to enhance atherosclerosis with aging, we next characterized T cells during aging in health and atherosclerosis. We performed single-cell RNA sequencing (scRNAseq) and T cell receptor (TCR)seq on enriched CD3⁺ T cells from spleens of healthy young (2–3 months) and aged (18–19 months) female mice, and from atherosclerotic lesions of young (4–5 months) and aged (21–22 months) mice (Fig. 3A). We performed dimension reduction using uniform manifold approximation projection (UMAP)²¹ (Fig. 3B). Based on transcriptomic gene expression (Fig. S6), we determined that of the 37,260 total cells, 83% are T cells. Clustering and cell surface marker expression revealed significant differences in cell types by aging and tissue distribution (Fig. 3C and 3D and Figs. S6 and S7).

Fig. 3. Aging leads to expansion of memory CD8 cells that infiltrate aortic atherosclerotic plaque.

(A) Overview of the single cell analyses of the healthy young (2–3 month) and aged (18–19 month) female C57BL/6N WT spleens enriched for CD3⁺ T cells, and young (4–5 month) and aged (21–22 month) female C57BL/6N mPCSK9-AAV aortic plaques enriched for CD3⁺ T cells. (B) UMAP (uniform manifold approximation projection) clustering of all cells from the spleen and aorta of C57BL/6N mice showing unique cellular populations. (C) UMAP clustering of all cells split by group from the young spleens (pool of N=3 biological replicates with 18,412 total cells), aged spleens (pool of N=3 biological replicates with 11,151 total cells), young aortas (pool of N=4 biological replicates with 1,999 total cells), and aged aortas (pool of N=4 biological replicates with 5,698 total cells), color-coded by cluster identification from B. (D) Proportion of all clusters (left) and CD8⁺ T cell clusters (right) represented in each group and tissue type based on single-cell RNA seq. (E) Fold changes in cluster proportions comparing aged to young (higher fold change = increased in aged versus young) for spleen and aortic plaque. Data in this figure are from 1 independent experiment. EM = effector memory, CM = central memory, GZMK = granzyme K, LDL = low-density lipoprotein, PCSK9 = proprotein convertase subtilisin/kexin type 9 serine protease, AAV = adeno-associated virus, UMAP = uniform manifold approximation projection.

We investigated how clusters differed by age and tissue (Fig. 3D), specifically within CD8⁺ T cell clusters. Most T cells from young spleens were naïve while most in the aged spleens were memory subtypes (Fig. 3D). EM CD8⁺ T cells, CM CD8⁺ T cells, and CD8⁺ EM Gzmk were significantly increased within atherosclerotic plaques in aged mice with very few naïve CD8⁺ T cells found in young (15 cells, 4.9% of aortic CD8⁺ T cells) or aged (33 cells, 2.9% of aortic CD8⁺ T cells) atherosclerotic plaques (Fig. 3E).

We identified γδ T cells (GammaDelta_1 cluster) as significantly enriched in atherosclerotic plaques from young and aged mice (Fig. 3E). When we explored fold-change differences within each age group, we found that the CD8⁺ EM Gzmk population had the second highest fold increase in aged spleen after follicular CD4⁺ T cells and highest fold change in atherosclerotic aorta with only 89 cells identified in young spleen and 176 in young aorta compared with 767 cells in aged spleen and 805 in aged aorta. This suggests that memory CD8⁺ T cells but not naïve T cells accumulate in atherosclerotic plaques, and possibly that naïve CD8⁺T cells differentiate into memory cells upon infiltration into the plaque.

Aging CD8⁺ T cell cytotoxicity in atherosclerosis

To determine how aging impacts CD8⁺ T cell populations in disease-free spleens and within atherosclerotic plaques, we examined naïve (CD8_Naive), CM (CD8_CM), EM (CD8_EM), and EM Gzmk (CD8_EM_GZMK) CD8⁺ clusters (Fig. 4A). Most naïve cells were from young spleens followed by aged spleens, whereas most memory cells derived from aged spleens and aged atherosclerotic plaques (Fig. 4B). We performed trajectory analysis to compares dynamic changes in gene expression of neighboring cells to determine the differentiation state of cell changes over pseudotime (Fig. 4C). We identified a sub-cluster of CD8⁺ T cells with the highest pseudotime value (dashed oval) which derived primarily from aged aortic samples and were CD8⁺ CM and CD8⁺ EM Gzmk cells (Fig. 4C–E). We compared the transcriptome of these to the other CD8⁺ T cells and found that they exhibit high expression of exhaustion (Pdcd1 and Tigit) and cytotoxicity markers (Fth1) (Fig. 4D). Naïve CD8⁺ T cells had the lowest pseudotime values (Fig. 4E). Within CD8⁺ CM, CD8⁺ EM, and CD8⁺ EM Gzmk cells, a range of pseudotime values were evident, suggesting considerable transcriptomic variability. We examined pseudotime by group which demonstrated that CD8⁺ T cells from aged aortas and spleens have the highest pseudotime values of all groups (Fig. 4F). CD8⁺ T cells from atherosclerotic plaques exhibited further pseudotime differentiation compared to splenic counterparts, suggesting either a selection for more differentiated T cells that migrate to the aorta or differentiation upon migration to the atherosclerotic plaque (Fig. 4F).

Fig. 4. CD8 T cell populations in atherosclerotic plaques are more differentiated, exhausted, and cytotoxic with aging.

(A) UMAP scattergram of splenic and aortic CD8_Naive, CD8_CM, CD8_EM_Gzmk, CD8_EM clusters, color coded by cluster. (B) UMAP scattergram of the same clusters from A but color-coded by group. (C) UMAP scattergram of the same 4 clusters from A, but color coded by pseudotime trajectory analysis score with the cells with highest pseudotime score indicated by dashed line oval. (D) Volcano plot showing differential gene expression comparing CD8_CM and CD8_EM_Gzmk clusters with the highest pseudotime score (inside the dashed oval in C) to those with lower pseudotime scores (outside the dashed oval in C). Positive values in gene expression indicate increased expression in the cells with the highest pseudotime values. (E) UMAP scattergrams of pseudotime trajectory score split by cluster for CD8_Naive, CD8_CM, CD8_EM_Gzmk, CD8_EM with dashed oval indicating cells with the highest pseudotime trajectory values. (F) UMAP scattergrams of pseudotime trajectory score split by sample group for Young_Spleen, Aged_Spleen, Young_Aorta, Aged_Aorta with dashed oval indicating cells with the highest pseudotime trajectory values. (G) Signature score violin plot calculated for the 4 clusters from A, split by cluster and sample group, for exhaustion score (genes: Tim3, Lag3, Eomes, Pdcd1, Cd160, Tox), cytotoxicity score (genes: Ctla4, Prf1, Gzmb, Gzmk, Tigit, Ifng, Tnfrs1a), Senescence score (genes: Cdkna1, Cdkna2), and SenMayo score²². For A-G, clustering of all cells split by group from the young spleens (N=3 biological replicates), aged spleens (N=3 biological replicates), young aortas (N=4 biological replicates), and aged aortas (N=4 biological replicates). For D, Venice method within BBrowser3 was used to compare groups (see Methods). Data in this figure are fromn one independent experiment. AU = arbitrary units, CM = central memory, EM = effector memory, Gzmk = granzyme K, UMAP = uniform manifold approximation projection.

We next examined how aging impacted CD8⁺ T cell exhaustion, cytotoxicity and senescence, by calculating signature scores across the 4 groups and 4 CD8⁺ T cell clusters (Fig. 4G). CD8⁺ T cells from aged aortas had the highest exhaustion score, cytotoxicity score, and senescence score (Cdkna1/Cdkna2 and SenMayo scores²²) for all 4 CD8⁺ T cell clusters compared to other groups (Fig. 4G). These data demonstrate that CD8⁺ T cells from aged and young aortas are more differentiated compared to CD8⁺ T cells from young spleens. These data are similar to a previous scRNAseq study conducted in older adults that showed most T cells within human atherosclerotic plaques are memory sub-types and more cytotoxic/inflammatory in symptomatic versus asymptomatic patients²³. This study did not report a sub-population of EM CD8⁺ T cells defined by GZMK expression; however, after re-examined this dataset we found that most human plaque T cells express GZMK (Fig. S15 A-B). The T cells with high GZMK expression (Extended data Fig. 8, dashed line) compared to the rest of the T cells had genes associated with cytotoxicity and inflammation (i.e., CCL5, CCL4, S100A4; Extended data Fig. 8C). Although all participants in the study by Fernandez et. al >65 years, comparing symptomatic to asymptomatic patients revealed that GZMK expression is significantly upregulated in CD8⁺ T cells from symptomatic patients compared to asymptomatic (Extended data Fig. 8D). These data support our findings in aged murine atherosclerotic plaques.

Aging induces atherosclerotic CD8⁺ T cell clonal expansion

We investigated T cell receptor (TCR) clonality by TCRseq and identified that clonal diversity by Chao1 diversity estimation was greatest in young spleens followed by the aged spleens (Extended data Fig. 9A). Most CDR3 sequences were 13–15 amino acids long (Extended data Fig. 9C). TCR clones from young spleens were small or medium size while clones from aged spleens were small-large. TCR clones the young and aged aortas were large and hyperexpanded (Extended data Fig. 9D). The 5 most expanded clonotypes from each sample across other samples were from multiple samples with significant overlap of CDR3 sequences in all samples by Jaccard index (Extended data Fig. 9E–F). We found the canonical NKT cell 15-amino acid TCR-α CDR3 sequence, CVVGDRGSALGRLHF (Extended data Fig. 9G).^{24, 25, 26} We identified other expanded 10-amino and 5-amino acid CDR3 sequences distributed among young and aged aortic samples, and these were most made up of polar and hydrophobic residues (Extended data Fig. 9H–L). We demonstrated that many clones are shared across samples and aged samples have more clonal expansion compared to young samples (Extended data Fig. 9M–O).

Within CD8_naive, CD8_EM, CD8_CM, and CD8_EM_Gzmk clusters we report TCR clones as bar plots with each clone as a unique color where size corresponds to the percent of cells containing that clone. We show all clones present in 2 or more cells, stratified by cluster and group (Fig. 5A). We observed that CD8⁺ EM Gzmk cells has the greatest TCR clonal expansion with more clones in aged splenic CD8⁺ EM Gzmk cells (149 clones) compared to young spleens (12 clones) and in the aged plaques (87 clones) compared to young plaques (25 clones; Fig. 5A). The average size of clones in CD8⁺ EM Gzmk cells within aged plaques (5.53 cells/clone) was significantly greater than the other groups (from 2.28 to 2.92 cells/clone; Fig. 5A). We identified 2 clones within the aged plaque CD8⁺ EM Gzmk cells that were significantly expanded compared to all others, with 178 cells identified in each. These represented the largest clones in our dataset after the canonical NKT cell clone. We identified a significant number of clones within CD8 naïve cells including one with 23 cells; however, most of these included <3 cells/clone. We also found clonal expansion within CD8⁺ CM cells, especially in aged spleens (155 clones) with the 4 largest clones including >15 cells/clone (Fig. 5A). This suggests that CD8⁺ EM Gzmk and CD8⁺ CM cells exhibit clonal expansion in aging.

Fig. 5: Granzyme K⁺ CD8⁺ effector memory and CD8⁺ central memory cells from aged mice are clonally expanded compared to young mice.

Single-cell T cell receptor (TCR) sequencing of naive, central memory, effector memory, and granzyme K⁺ effector memory CD8⁺ T cells from disease-free 3-month old female C57BL/6N WT and 18-month old female C57BL/6N WT spleens and from atherosclerotic plaques from 3-month old C57BL/6N WT and 18-month old C57BL/6N WT mice after enriching for CD3⁺ T cells. (A) The percent of cells per unique clonotype and the number of unique clonotypes with at least 2 cells are shown. Each color represents a unique clonotype. Data are represented as a percentage of the total size of the sample. The 2 largest clonotypes, identified in within the aged plaque CD8_EM_GZMK cluster are shown in black color on the UMAP scattergram which depicts the 4 clusters of CD8⁺ T cells from all groups. (B) CDR3 sequence, V_gene, J_Gene, and clone size (number of cells) are shown along with a pie chart showing the group composition and cluster composition that makes up each of the top 5 largest clones within the CD8⁺ T cell clusters. N=3 biological replicates per group for spleen and N=4 biological replicates per group for aortic plaque. Data in this figure are from 1 independent experiment. CM = central memory, EM = effector memory, Gzmk = granzyme K, Y = young, A = aged, UMAP = uniform manifold approximation projection.

We investigated the 6 largest clones identified (Fig. 5B) in CD8 clusters. The largest clones had CDR3 motifs CASSLREEQYF and CVLEASGSWQLIF with one TCRβ clone and the other TCRα clone. All cells expressing these CDR3 motifs derived from aged plaques and most were from CD8⁺ EM Gzmk cells with some from CD8⁺ naïve, CD8⁺ EM, and CD8⁺ CM cells (Fig. 5B). The 3 other largest clones identified were skewed toward TCRα chains and were identified in spleens from the aged mice; however, one clone was expanded within young spleens (Fig. 5B). Nearly all clones were identified within the CD8⁺ CM cells (Fig. 5B). The 2 aortic plaque clones contained 6-times more cells as splenic clones. This suggests that clonal expansion is enhanced with aging and increased within atherosclerotic plaque CD8⁺ T cells.

Aging alters CD8⁺ T cell transcriptome in atherosclerosis

Based on differences in CD8⁺ T cell sub-populations and γδ T cell frequencies by age in disease-free spleens and atherosclerotic plaques, we investigated age-effects on T cell transcriptomes. There were 183 differentially expressed genes in splenic CD8⁺ EM Gzmk cells (Table S1) and 162 differentially expressed genes in atherosclerotic plaques (Table S2), with 58 shared by both (Fig. S8). The top GO biological processes by aging in splenic CD8⁺ EM Gzmk cells were cytoskeletal organization, adhesion, and migration. In the aorta, top GO processes included regulation of T cell migration, adhesion, and T cell activation (Fig. S8). For CD8⁺ CM cells, we identified 191 genes that were significantly impacted by aging in spleens (Table S3), 151 genes impacted in atherosclerotic plaques (Table S4), and 68 that were shared by both (Fig. S9). The top GO processes impacted by aging in splenic CD8⁺ CM T cells were T cell activation and adhesion and cytotoxicity in atherosclerotic plaques (Fig. S9). For CD8⁺ EM T cells, there were 365 differentially expressed genes by age in spleens (Table S5), 98 differentially expressed genes in the atherosclerotic plaques (Table S6), with 79 that were shared by both (Fig. S10). The top age-impacted GO processes in splenic CD8⁺ EM T cells were cytoskeletal organization, adhesion, and migration and in atherosclerotic plaques they related to lymphocyte migration (Fig. S10). The effect of aging on CD8⁺ T cells in atherosclerotic plaques differed by cell type. CD8⁺ EM Gzmk cells were characterized by T cell migration, adhesion, and activation; CD8⁺ CM cells were characterized by T cell cytotoxicity; and CD8⁺ EM cells were characterized by T cell migration.

Comparing aged to young differential gene expression in naïve CD8⁺ T cells in spleens (Table S7), we identified 115 significantly upregulated genes and 36 significantly downregulated genes (Fig. S11). There were not enough naïve CD8⁺ T cells in atherosclerotic plaques of young (15 cells) and aged (33 cells) to perform differential gene expression analysis. The top impacted GO biological processes affected by age in splenic CD8⁺ naïve T cells were cell polarity, metabolism, migration, and activation (Fig. S11). γδ T cells represented 30–45% of all CD3⁺ T cells in the atherosclerotic plaques, so we examined γδ T cell transcriptomic changes by aging. For the GammaDelta_1 cluster there were 128 differentially expressed genes in spleens (Table S8), 320 differentially expressed genes in atherosclerotic plaques (Table S9), and 38 shared genes (Fig. S12). The top GO processes in splenic GammaDelta_1 cluster were cell recognition, regulation of telomerase, cellular adhesion, and regulation of cell death whereas top GO processes in atherosclerotic plaques included T cell adhesion, actin regulation, and T cell activation (Fig. S12). Aging impacted migration, activation, and metabolism in naïve CD8⁺ T cells. In contrast, aging impacted cell recognition, adhesion, and cell death in disease-free splenic γδ T cells compared to adhesion, actin regulation, and activation in atherosclerotic plaque γδ T cells.

Aging upregulates CCR5, CXCR6, and CXCR3 in atherosclerosis

CD8⁺ T cell migration, motility, and locomotion were impacted in CD8⁺ T cells by aging, and we found that CCR5, CXCR6, and CXCR3 were most dysregulated between the young and aged CD8⁺ T cells by scRNAseq. We investigated CD8⁺ T cell surface expression of these chemokine receptors at baseline and at 3-, 6-, and 9-weeks of atherogenesis in young and aged mice by flow cytometry (Fig. 6A). Within the 3 major populations of CD8⁺ T cells, CCR5 and CXCR6 expression were upregulated on CM CD8⁺ cells while CXCR3 expression is significantly upregulated on EM CD8⁺ cells in aged mice (Fig. 6B–D). We determined expression of these chemokine receptors increases during atherogenesis in circulating CD8⁺ T cells.

Fig. 6: Chemokine receptors are upregulated in CD8⁺ T cell subsets in aged mice compared to young mice.

(A) Flow cytometry gating strategy for T cells from the blood of 3-mo old female C57BL/6N and 18-mo old female C57BL/6N WT mice to determine chemokine receptor expression of CCR5, CXCR3, and CXCR6. (B) Representative histograms of CCR5, CXCR6, and CXCR3 expression at baseline from 3-month old C57BL/6N WT and 18-month old C57BL/6N WT mice. Quantification of CCR5 expression on CD8⁺ CM T cells (C), CXCR6 expression on CD8⁺ CM t cells (D), and CXCR3 expression on CD8⁺ EM T cells from blood of 3-mo old C57BL/6N and 18-mo C57BL/6N female mice from baseline through 9-weeks of PCSK9-AAV treatment with western diet feeding. N=7 biological replicates per group for 3-mo old C57BL/6N and N=7 biological replicates per group for 18-mo C57BL/6N. Data in this figure are from 2 independent experiments. Data in this figure represents mean +/− SEM. For C-E, 2-way ANOVA with Šídák’s post-hoc test. AAV = adeno-associated virus, CCR5 = C-C chemokine receptor type 5, CXCR3 = C-X-C motif chemokine receptor 3, CXCR6 = C-X-C motif chemokine receptor 6, CM = central memory, EM = effector memory, MFI = median fluorescence intensity, PCSK9 = proprotein convertase subtilisin/kexin type 9 serine protease, Y = young, A = aged, WD = western diet.

Based on this, we sought to identify dynamics of T cell migration into atherosclerotic plaques during atherogenesis. We induced atherosclerosis in young and aged mice and quantified aortic T cells and macrophages at 3-, 6-, and 9-weeks by flow cytometry (Extended data Fig. 10A). The number of CD3⁺ T cells, CD8⁺ T cells, and F4/80⁺ macrophages within the aortas of young and aged mice similar at each timepoint analyzed and all 3 cell types were higher in aged atherosclerotic mice (Extended data Fig. 10B–G). At 6- and 9-weeks, there were more CD8⁺ T cells in aged atherosclerotic aortas compared to young. These data support our histology analysis of CD8⁺ T cells in atherosclerotic plaques (Fig. 1B) that there are ~10-fold more macrophages within the atherosclerotic aorta compared with T cells (Extended data Fig. 10B–G).

Discussion

Our study provides a comprehensive atlas T cells changes with age in disease-free spleens and atherosclerotic plaques in mice. We report that CD8⁺ T cells that accumulate in atherosclerotic plaques in aging mice are primarily memory T cells with few naïve T cells represented. Similar findings have been described in atherosclerotic plaques from older humans with a mean age of 70 years²³, although there was no young comparison group. By re-analyzing T cell clusters from this dataset (Fernandez et al., 2019) we found that most CD8⁺ T cells in human atherosclerotic plaques have express granzyme K, which is increased in symptomatic patients compared to asymptomatic. Thus, we found that aged murine atherosclerotic plaque CD8⁺ T cells exhibit similarities to human atherosclerotic patients and suggests our findings in mice are similar to humans. We also found that CD8⁺ T cells in the atherosclerotic plaques are clonally expanded compared to T cells from disease-free spleens, and this is enhanced by aging. Our study also determined how aging alters CD8⁺ T cell gene signatures within atherosclerotic lesions compared to healthy splenic CD8⁺ T cells. We have provided a comprehensive description of CD8⁺ T cells with aging and tested the impact of CD8⁺ T cells in murine atherosclerosis model with cause and effect experiments. Our results show that CD8⁺ T cells are both necessary and sufficient for atherosclerosis in aging.

Our study differs from recently published single-cell RNA-seq resources of aging immune systems in disease-free mice^{16, 27, 28} in 2 distinct ways. First, we focus on T cells by enriching for CD3⁺ T cells prior to performing scRNAseq and TCRseq. This reduced the number of B cells which is the most abundant cell type in the spleen. Second, we examined T cells during aging from healthy spleens and compared with T cells from atherosclerotic plaques. Previous scRNAseq studies in atherosclerotic mice utilized young germline knockout models (Ldlr^−/− and Apoe^−/− mice)^{29, 30, 31}. Our study focused on aging and used the PCSK9-AAV model¹⁵, which allows mice to develop and age with low LDL cholesterol levels unlike Apoe^−/− and Ldlr^−/− mice³². This allowed us to identify changes in CD8⁺ T cell populations and transcriptomes by age including the accumulation of memory CD8⁺ T cells in age-enhanced atherosclerosis.

We show that significant changes occur within CD8⁺ T cells with aging, including significantly more EM and CM populations and declining naïve cells which is consistent with previous reports²⁸. We show that CD8⁺ EM, CD8⁺ EM Gzmk cells, to a lesser degree CD8⁺ CM T cells, and almost no naïve CD8⁺ T cells migrate into atherosclerotic plaques with aging. Aortas of disease-free WT mice have few CD8⁺ T cells thus rendering assessment of CD8⁺ T cells within the healthy aorta technically challenging. The Tabula Muris Senis dataset only had 16 cells out of 9,436 cells in the “heart_and_aorta” and “aorta” clusters that express Cd8a, suggesting that <0.2% of those cells are CD8⁺ T cells²⁸. The majority of T cells identified in atherosclerotic plaques in PCSK9-AAV atherosclerotic mice likely transmigrate into the aorta during the 10-week WD feeding period during hypercholesterolemia. This suggests that CD8⁺ CM, CD8⁺ EM, and CD8⁺ EM Gzmk T cells have greater migratory potential to infiltrate diseased tissues in aging. Determining the importance of each subpopulation on atherogenesis with aging requires development of tools to deplete specific subpopulations of CD8⁺ T cells.

Aging not only increases CD8⁺ CM, CD8⁺ EM, and CD8⁺ EM Gzmk T cell frequencies, it skews the transcriptome to more exhausted, cytotoxic, migratory, and senescent, particularly atherosclerotic T cells. Within atherosclerotic plaque CD8⁺ T cells, aging alters transcriptomes related to chemotaxis, cytotoxicity, and exhaustion compared to young mice and splenic CD8⁺ T cells. As cells become senescent, they typically secrete inflammatory factors^{33, 34, 35}. A prior study in young Ldlr^−/− mice showed that CD8⁺ T cells induce macrophages to release chemokines in vitro^{5, 10}, and we have shown that T cells and macrophages are present within the atherosclerotic plaques by 3-weeks after induction of hypercholesterolemia. It is possible that aged CD8⁺ T cells within atherosclerotic plaques may contribute to accumulation of macrophages or their functions in atherosclerotic lesions by secreting chemoattractants that alter monocyte/macrophage function. This is an area that will require future investigation.

We have also uncovered that CD8⁺ T cells that accumulate in atherosclerotic plaques in young and aged mice are clonally expanded compared with CD8⁺ T cells in disease-free spleens. This is enhanced in aged compared with young mice, and TCR diversity in atherosclerotic plaque CD8⁺ T cells are significantly reduced compared with splenic CD8⁺ T cells. Expansion of epitope-specific memory CD8⁺ T cells may be in response to atherosclerosis-related antigens or cryptic self-antigens related to circulating and accumulating oxidized lipids or cell death, leading these clones to traffic to plaques³⁶. It is possible that memory CD8⁺ T cells differentiate and expand upon infiltration in atherosclerotic plaques. CD8_EM_Gzmk cells had the greatest clonal expansion along with senescence, cytotoxicity, and exhaustion signature scores. Determining specific antigens that T cells respond to is technically challenging, and our study was not designed to uncover atherosclerosis-specific antigens in aging. Rather our goal was to understand how CD8⁺ T cells contribute to atherogenesis with aging. We attempted to cross-reference the TCR-clonotypes with databases of TCR epitopes like McPAS-TCR and VDJdb without success, likely because these focus on infection- and tumor-specific epitopes. Uncovering epitope-specific T cells in atherosclerosis is an area of intense study, and several promising targets have emerged including ApoB100³⁷, oxidized LDL³⁸, and other oxidation species³⁹. As databases of TCR CDR3 sequences are curated, it may be possible to determine the specific epitopes we have uncovered in CD8⁺ T cells in our study.

Our immunohistochemical analysis of CD8⁺ T cells within atherosclerotic plaques and prior scRNAseq studies suggest that CD8⁺ T cells are a minority of total plaque cells. Our CD8⁺ T cell blockade and adoptive transfer experiments suggest that manipulating circulating CD8⁺ T cells significantly impacts development of atherosclerosis with aging. Depletion of CD8⁺ T cells in young Ldlr^−/− and Apoe^−/− mice^{10, 11} leads to reduced atherosclerotic plaque size by altering monopoiesis¹⁰. Our results indicate that CD8⁺ T cell depletion attenuates atherosclerosis in aged PCSK9-AAV mice and young Ldlr^−/− mice to a lesser degree; however, we found that CD8⁺ T cell depletion has no impact on atherosclerotic lesion size in young PCSK9-AAV treated mice. We identified differences in CD8⁺ T cell sub-populations between young C57BL/6N WT mice and young Ldlr^−/− mice. Young WT and young Ldlr^−/− mice have more naïve CD8⁺ T cells, between 15–30% of all CD3⁺ T cells; however, young Ldlr^−/− mice have twice as many circulating CD8⁺ CM cells compared to young C57BL/6N WT mice. We believe the reduction of CD8⁺ CM cells by CD8 blockade contributed to the reduced atherosclerotic plaque size in the young Ldlr^−/− mice but not young C57BL/6N WT. Young Ldlr^−/− mice have fewer CD8⁺ EM T cells compared with aged WT mice; however, they have significantly more CD8⁺ EM T cells than young WT mice at baseline and 5-weeks of atherogenesis. In young Ldlr^−/− mice, either deficiency of Ldlr during development or exposure to more circulating LDL particles from birth may lead to more CM CD8⁺ T cells. Chronic exposure to LDL particles in the Ldlr^−/− mice may induce release of cryptic self-antigens and expansion of atherogenic CM CD8⁺ T cells. Given our findings that CD8⁺ T cell depletion reduced atherosclerosis in young Ldlr^−/− mice, and our observation that most CD8⁺ T cells in atherosclerotic plaques are memory subtypes, our study suggests that memory CD8⁺ T cells promote atherosclerosis, which is enhanced by aging.

We also demonstrate that ~1/3 of atherosclerotic plaque T cells are γδ T cells. Winkels et al. found that 11% of all T cells were γδ T cells by mass cytometry in young Apoe^−/− mice²⁹. We found that γδ T cells within atherosclerotic lesions exhibit upregulation of T cell activation and actin filament organization pathways with aging. γδ T cells can recognize lipid moieties without major histocompatibility complex (MHC) molecules⁴⁰. γδ T cells also elicit rapid effector functions similar to innate immune responses⁴⁰; however, their role in atherogenesis is not fully understood. One study found that that γδ T cells increase within young atherosclerotic Apoe^−/− spleens⁴¹. Another study found that young Apoe^−/− mice fed a LFD have more aortic γδ T cells compared with C57BL/6N mice⁴². Young Apoe^−/− mice crossed with γδ T cell-deficient mice (i.e., Tcrδ^−/− mice) and rendered atherosclerotic have no change in atherosclerotic lesion size in one study⁴¹ but significantly reduced aortic arch lipid in another⁴². Given these discordant findings and transcriptomic changes in γδ T cells with aging, future investigation will be required to understand the contribution of γδ T cells on atherogenesis in mice and human atherosclerosis.

There are limitations to our study. The CD8⁺ T cell blockade of CD8⁺ T cells by rat-anti-mouse CD8 antibodies could have induced mouse anti-rat antibodies; however, the anti-CD8 antibody clone used for depletion (2.43) is different from the clone used to detect CD8 T cells by flow cytometry (53–6.7). Despite the possible presence of these antibodies, the depletion was effective, but mouse-anti-rat antibodies could affect the results in unexpected ways although we have no evidence that this occurred in our study. The role of naïve CD8⁺ T cells on atherosclerosis is less clear compared with memory CD8⁺ T cells. Depletion of a subset of memory cells at young ages may be atheroprotective and future studies should test this. We cannot rule out the possibility that sub-populations of memory CD8⁺ T cells may have pleiotropic effects on atherosclerosis depending on age and cell type. Some CD8⁺ dendritic cells were shown to be atherogenic in young Ldlr^−/− mice, and these cells should be depleted by anti-CD8 antibody treatment⁴³. Our adoptive transfer demonstrates that young CD8⁺ T cells that were adoptively transferred were ~75% naïve CD8⁺ T cells and did not promote atherogenesis while CD8⁺ T cells from aged donors were ~50% memory CD8⁺ T cells with more PD1⁺ cells indicating CD8⁺ T cell exhaustion. We conclude that depletion of aged memory and exhausted CD8⁺ T cells is atheroprotective; however, the role of naïve CD8⁺ T cells is less clear because they are depleted in both young and aged mice. Future studies should determine specific contributions of each cell type including CD8⁺ dendritic cells on atherosclerosis with aging, and whether phenotypic changes related to aging like exhausted memory cells compared to non-exhausted memory cells will play a role on atherogenesis.

In summary, we provide evidence for an atherogenic role of CD8⁺ T cells with aging, and this is driven by effector and memory CD8⁺ T cells. Our findings demonstrate a significant population of CD8⁺ EM, CD8⁺ CM, and CD8⁺ EM Gzmk T cells in atherosclerotic plaques, particularly in aged mice. We also found γδ T cells in atherosclerotic plaques of both young and aged mice. This suggests that these sub-populations of T cells may contribute significantly to age-enhanced atherosclerosis. There have been significant advances in development of T cell therapies in treating cancer, and our study paves the way for potential therapeutics targeting T cells to treat and prevent atherosclerosis in older adults.

Methods

Study approval.

All animal experiments were carried out in accordance with the University of Michigan Institutional Animal Care and Use Committee (protocol #00010704) and the University of Alabama at Birmingham (protocol #22627).

Mice and diet.

C57BL/6N WT young (2–3 months) and aged (18–19 months) female mice were obtained from the National Institute on Aging rodent colony and Charles River Breeding Laboratories (stock# 027). Female C57BL/6N Ldlr^−/− mice were purchased from Jackson Laboratories (stock #002207). Female C57BL/6N CD8^−/− mice (Jackson Laboratories, stock # 002665) were a gift from Dr. Grace Chen, the University of Michigan. We chose to use female mice in these studies because we have previously demonstrated that the age-enhanced atherosclerosis phenotype is similar between males and females ³. All mice were maintained on a 12-h light-dark cycle with free access to food and water. All mice are on a C57BL/6N background. Sizes of experimental groups were based upon our prior studies ^{3, 44}.

To generate hyperlipidemic WT C57BL/6N mice, an i.p., injection of recombinant adeno-associated virus 8-D377Y-murine Pcsk9 (mPCSK9-AAV) was used. The mPCSK9-AAV was generated at the University of Pennsylvania Vector Core. The mPCSK9-AAV was diluted in sterile saline and mPCSK9-AAV was administered i.p. at 5.0×10⁹ vector genomes per gram body weight (i.e., 1.5×10¹¹ vector genomes/30g mouse), similar to the intermediate dose that was previously described ¹⁵. mPCSK9-AAV mice were then rested for 1 week in their home cage and then both WT C57BL/6N mPCSK9-AAV and Ldlr^−/− mice were fed a Western Diet (WD; 42% calories from fat, Teklad, catalog #88137) or control low-fat diet (LFD; 13% calories from fat, LabDiet catalog #5L0D) for 10-weeks. Similar to previous experiments ⁴⁵, a subset of young and aged C57BL/6N and young Ldlr^−/− mice were administered 200μg of anti-CD8α clone 2.43 monoclonal antibody (BioXcel, catalog #BE0061) or isotype control (Rat IgG2b, κ, catalog #BE0090) at 7 days and 3 days before beginning WD, and then every 2 weeks (i.e., day −7, day −3, day 7, day 21, day 35, day 49, and day 63) as has been used previously.⁴⁶ The CD8 antibody used for CD8 depletion studies was clone 2.43 and the anti-CD8 antibody used for flow cytometry experiments was clone 53–6.7. In a subset of experiments, cells were isolated from spleens from young (2–3 months) and aged (18–20 months) C57BL/6N mice and were enriched for CD8⁺ T cells (EasySep^™, StemCell Technologies, catalog #19853) and 10⁷ of either pooled young or pooled aged purified CD8⁺ T cells or vehicle only (no cells) were administered to young (2–3 months) CD8^−/− mice. CD8^−/− mice lack CD8⁺ T cells, thus mice were allowed to rest in their home cage for 1 month followed by mPCSK9-AAV administration and western-diet feeding for 10-weeks. Mice that did not survive the entire 10 western-diet feeding period were excluded from the primary endpoint of atherosclerotic plaque size.

Atherosclerotic lesion analysis and histology.

Mice were euthanized using isoflurane overdose. Blood was removed by right ventricular cardiac puncture and the vasculature was perfused with ice-cold PBS. The heart was harvested and the aortic root was placed in paraformaldehyde and paraffin embedded. The aortic root was serially sectioned (6μm each). Histochemical sectioning and staining was performed by the Unit for Animal Laboratory Management In vivo Animal Core Pathology laboratory at the University of Michigan and technicians were blinded to the experimental group identity. For morphometric analysis, 30 paraffin sections (each 6μm apart) per mouse were stained with hematoxylin & eosin and assessed for lesion size and composition as previously described,⁴⁷ for a total distance of 360μm of the aortic sinus. The brachiocephalic artery (BCA) was chosen as a second anatomical site to assess for atherosclerosis as recommended ⁴⁸. Serial sections for the BCA were performed in a similar paradigm as the aortic sinus but only for 32 sections (total distance of 140μm). For the BCA, 16 sections were stained with hematoxylin and eosin and analyzed per biological replicate. Total lesion size and acellular lesion area, a correlative of necrotic core size,⁴⁹ was quantified using ImageJ software (NIH, USA). Images compiled in GraphPad Prism (Version 9.4.1, San Diego, CA).

Immunohistochemistry.

Slides were deparaffinized using Diva Decloaker (Biocare), blocked (Biocare, catalog No. RBM961), and stained CD8α (Cell Signaling Technology, catalog #98941, 1μg/mL). Secondary antibodies were used (Biocare, catalog No. RMR622) followed by 3,3′-Diaminobenzidine (DAB) chromogen and hematoxylin counterstain for nuclei. Immunohistochemical staining was performed by the Unit for Animal Laboratory Management In vivo Animal Core Pathology laboratory which also stained positive control slides and secondary antibody-only stains. Immunohistochemistry and H&E images were captured with an Olympus LC30 camera mounted on Olympus CX41 microscope. Slides were scanned and processed using QuPath (version 0.3.2, The University of Edinburgh) and images compiled in GraphPad Prism (Version 9.4.1, San Diego, CA). CD8⁺ cells were visually identified and counted.

Flow cytometry.

Approximately 50–100 μl of whole blood was harvested from animals via the tail vein and the red blood cells were lysed via red blood cell lysis buffer (Biolegend, catalog #420302). Cells were then centrifuged for 5 minutes at 400xG and then the supernatant was poured off. For flow cytometry on aortas, the aortic root and arch containing atherosclerotic plaques were harvested into 1X PBS on ice. Aortas were minced and digested in 1X HBSS containing 450 U/mL collagenase I (Sigma, catalog #SCR103), 250 U/mL collagenase XI (Sigma, catalog #C7657), 120 U/mL hyaluronidase (Sigma, catalog #H3506), and 120 U/mL DNAse I (Roche, 10104159001) for 45 minutes followed by quenching with RPMI 1640 + 10% fetal bovine serum followed by passing through 70μm cell strainer. Pellets were washed again then resuspended with live/dead aqua stain (ThermoFisher Scientific, catalog #L34957) followed by addition of FCγR block (Biolegend, catalog #101320) in flow cytometry buffer (FACS buffer), 1× PBS (Ca²⁺- and Mg²⁺-free) containing 5% FBS, and 5mM EDTA. Blood cells were split to perform between 1–3 different flow cytometry staining panels during FCγR blocking step. Cells were incubated for 15 minutes followed by addition of antibody cocktail on ice for 30 minutes. Cells were fixed in 4% paraformaldehyde for 15 minutes. When necessary, cells were fixed and permeabilized (BD Bioscience, catalog #554714) to stain intracellular antigens according to manufacturer’s instructions and 4% paraformaldehyde fixation was omitted. Cells were washed with FACS buffer 2 times and then resuspended in 200μl FACS buffer. Flow cytometry was performed on a BioRad Ze5 equipped with 405nm, 488nm, 561nm, and 640nm lasers using Everest software. Data and compensation were analyzed with FlowJo software (FlowJo 10.8.1, Becton Dickinson). Antibodies used in flow cytometry were: APC-Fire750 anti-CD8a (Biolegend, catalog #100766, 4μg/mL), FITC anti-CD45 (Biolegend, catalog #103108, 10μg/mL), Alexa700 anti-CD44 (Biolegend, catalog #103026, 5μg/mL), BV785 anti-CD3 (Biolegend, catalog #100232, 4μg/mL), BV605 anti-CD62L (Biolegend, catalog #104438, 2μg/mL), PE-Cy7 anti-CD4 (Biolegend, catalog #100422, 2μg/mL), BV421 anti-PD1 (Biolegend, catalog #135221, 4μg/mL), PE anti-Tox (ThermoFisher Scientific, catalog #12–6502-82, 4μg/mL), PE anti-Gzmk (MyBioSource, catalog #2052533, 6.4μg/mL), BUV387 anti-CD4 (Biolegend, catalog #100492, 2.5 μg/mL), BUV661 anti-CD45 (BD Biosciences, catalog #612975, 2μg/mL), PerCP/Cy5.5 CD3e (Biolegend, catalog #152312, 2μg/mL).

Single Cell RNA sequencing.

Single cell suspensions from aortic plaques of young (4–5 months) and aged (21–22 months) C57BL/6N WT PCSK9-AAV mice were generated by isolating the aortic root and arch, mincing tissue, and digesting the aorta in 1X HBSS containing 450 U/mL collagenase I (Sigma, catalog #SCR103), 250 U/mL collagenase XI (Sigma, catalog #C7657), 120 U/mL hyaluronidase (Sigma, catalog #H3506), and 120 U/mL DNAse I (Roche, 10104159001) for 45 minutes. Suspension was filtered through 45μm cell strainer and washed with RPMI 1640 + 10% fetal bovine serum. Single cell suspensions of splenocytes were obtained by mechanical digestion of spleen by smushing through 70μm cell strainer and washing with RPMI 1640 + 10% fetal bovine serum. Single cell suspensions of aortic plaques and splenocytes were enriched for CD3⁺ cells using CD3ε positive selection microbead kit (Miltenyi Biotec, catalog #130–094-973) with Miltenyi MS columns (Miltenyi Biotec, catalog #130–042-201) and OctoMACS^™ separator (Miltenyi Biotec, catalog #130–042-108) according to manufacturer’s instructions. After CD3ε enrichment, cells were washed with warm RPMI 1640 + 10% fetal bovine serum. Cells were kept in RPMI 1640 + 10% fetal bovine serum on ice for 10–30 minutes until delivery to the Advanced Genomics Core laboratory at the University of Michigan for single cell RNA sequencing.

Samples with greater than 80% live cell count were bioanalyzed on D1000 ScreenTape with TapeStation Analysis Software 3.1.1 for quality control and were prepared for single-cell 5′ Gene expression and VDJ repertoire immune profiling, using the Chromium Single-Cell Controller (10X Genomics, CA). A part of the isolated cDNA was set aside for BCR target enrichment and VDJ repertoire library preparation, while another of the cDNA was processed for 5′ gene expression (GEX) library. Each library was sequenced on an Illumina NovaSeq-6000 platform to generate 151×151 paired-end reads. Processing of raw sequencing reads for each library was carried out using the Cell Ranger pipeline (v 6.0.1, 10x Genomics, CA). Reads were aligned to the pre-built mm10 genome references (versions: ‘mm10–2020-A’ or ‘vdj_GRCm38_alts_ensembl-5.0.0’) and the “filtered” Cell Ranger output, containing the expression profile or immune profile of cells with a correctly detected cellular barcode, was used for downstream analysis. Similar numbers of cells were recovered for both the expression profiling and immune profiling. Analysis was performed in R (v 4.0.3). For 5’ GEX data, analysis was primarily performed using the Seurat package (v 4.0.5) ⁵⁰. Doublets in the splenic and aortic single-cell RNA sequencing datasets were identified by the “scDblFinder()” function in the “scDblFinder” package⁵¹. 7% of the cells in the splenic dataset and 2.8% of cells in the aortic dataset were identified as doublets. We added these to a cluster “doublets” within the dataset to exclude from downstream analyses. Splenic and aortic single-cell RNA sequencing datasets were integrated using the “FindIntegrationAnchors()” and “IntegrateData()” functions developed by Stuart et al to identify anchor points (nfeatures = 3000).⁵⁰ Gene-barcode matrices and metadata for each sample were loaded and further filtering and clustering analyses were performed as described in the Seurat tutorials (http://satijalab.org/seurat/). The following parameter(s) were used to exclude cells with extreme values, which indicates low complexity, doublets, or apoptotic cells: less than 5% of UMIs mapping to mitochondrial genes and greater than 200 unique RNAs/cell. Counts were normalized using the default normalization approach and variable features were identified. Principal component analysis (PCA) was then performed, and 16 significant PCA components were used for finding nearest neighbors followed by graph-based, semi-unsupervised clustering into distinct populations (resolution = 0.5). To identify marker genes, the clusters were compared pairwise for differential gene expression using the Wilcoxon rank-sum test for single-cell gene expression (FindAllMarkers function; default settings). Cell-type predictions were also generated with scCATCH⁵², specifying mouse as species, using the CellMatch function in scCATCH to reduce memory requirements with heart and blood options included as tissues of origin. Thresholds of 0.10 were set for minimum numbers of cells for gene detection and minimum log-fold change. Cluster markers genes and scCATCH cell-type predictions were then manually reviewed. Further analysis was performed using BBrowser software (BioTuring, v3, San Diego, CA) ⁵³. Feature-barcode matrix files were imported into BBrowser. All UMAP plots used default settings. Cluster marker genes and scCATCH cell-type predictions were combined with BBrowser marker feature analysis to identify and combine clusters of the same cell types into a total of 37 clusters. Final cluster identifications were manually validated by incorporating marker genes, scCATCH predictions, and known cellular markers (i.e., Cd3 for T cells, Cd8a and Cd8b for cytotoxic T cells, Ccr7, Sell, and Cd44 to identify memory and naïve T cells). Marker gene expression on t-SNE plots were and significant differentially expressed genes were plotted. Differential gene expression was performed using BBrowser via the “Venice” method using logFC < −0.25 and logFC > 0.25; adjusted p-value < 0.05 ⁵⁴. Outputs were then used to create violin plots using GraphPad Prism software. The same differential gene expression output data were used to generate gene ontology biological process analysis using ShinyGO (v 0.76). Signature score analysis was performed using BBrowser and data were exported and graphs were created using Graphpad Prism software. Trajectory analysis (pseudotime analysis) was performed using BBrowser software.

For the VDJ data, initial analysis was performed using the scRepertoire package (v1.1, release 3.12) ⁵⁵. Filtered contig annotations were loaded and combined with sample metadata and assay information (combine TCR function; cells = T-AB). Clonotype abundances were defined as TcrIg identities to be consistent with the previous SC1 analysis. VDJ data was then integrated with the Seurat RNA-seq object (combineExpression function; cloneCall = “gene”). VDJ data were imported into BBrowser and figures were generated using Graphpad Prism software using the same thresholds used above (−0.25 > logFC > 0.25; adjusted p-value < 0.05). Additional VDJ analysis was performed in R (version 4.2.1) using Immunarch package (version 0.9.0).

Statistics and Reproducibility.

Sample size was determined based on power calculations performed in our previous publication.³ No mice were excluded from these studies unless they died prior to completing the 10-week western diet feeding period for atherosclerosis studies. Young and aged mice cannot be randomized to the age group; however, where possible, mice were randomly assigned to groups. For example, within the aged group for the anti-CD8 or anti-IgG treatment, some mice within the same cage were either randomized to anti-CD8 treatment or anti-IgG treatment. All histology experiments (cutting sections, staining, imaging, and quantifying) were performed in a blinded manner by technician who did not know the group assignment. For other experiments, investigators were not blinded. All results are presented as mean ±standard error of the mean (SEM). Normality was determined using Shapiro-Wilk test. Nonparametric tests were used for data that are not normally distributed or when groups contained <8 biological replicates. Data with 1 independent variable (i.e., age) were analyzed using 2-tailed Student’s t test or 2-tailed Mann-Whitney U-test (non-parametric). When more than 1 independent variable were analyzed (i.e., age and treatment or repeated measures), one-way ANOVA (or Kruskal-Wallis test) or 2-way ANOVA with multiple comparisons were used. Specific statistical tests and P values are denoted in figure legends. Two-sided P values were used and values <0.05 were considered significant. Graphpad Prism (v9.3.1, San Diego, CA), BBrowser (v3, BioTuring) was used to generate figures and for statistics.

Extended Data

Extended Data Fig. 1. Aging leads to expansion of memory CD8+ T cells in circulation.

Schematic of experimental procedure to isolate blood and perform flow cytometry at 3 timepoints during atherogenesis in young wild-type, aged wild-type, and young Ldlr−/− mice. (b) Blood-cell flow cytometry gating strategy to identify CD8+ T cells. Live CD45+ lymphocytes were gated on CD3e+ cells, CD8+ cells, and subdivided into naive (CD62L+CD44−), central memory (CD62L+CD44+), and effector memory (CD44+CD62L−) then PD1+ and either granzyme K+ or Tox+ cells. (c-l) Flow cytometry quantification of T cell populations as a frequency of CD3+ T cells for young C57BL/6, aged C57BL/6, and young Ldlr−/−, mice at baseline, midway through the western diet feeding period (5-weeks), and just prior to sacrifice (10-weeks). N = 6 biological replicates per group over 1 independent experiment. Measurements were taken from distinct samples. Data in this figure represents mean ± SEM. 2-Way ANOVA with Tukey’s post-hoc test. PCSK9 = proprotein convertase subtilisin/kexin type 9 serine protease, AAV = adeno-associated virus, CM = central memory, EM = effector memory.

Extended Data Fig. 2. Aging leads to expansion of memory CD8+ T cells in circulation.

Representative flow cytometry plots of T cells for young WT, aged WT, and young Ldlr−/− mice after PCSK9-AAV and 10-weeks of western diet feeding. Representative flow cytometry plots demonstrating increased frequency of central memory CD8+ T cells in aged C57BL/6 WT mice and young Ldlr−/− mice compared to young C57BL/6 WT mice, increased frequency of effector memory CD8+ T cells in aged C57BL/6 WT mice versus young C57BL/6 WT and Ldlr−/− mice, and greater frequency of naive CD8+ T cells in young C57BL/6 WT mice and young Ldlr−/− mice compared with aged C57BL/6 WT mice. Aged WT mice also demonstrate greater frequency of PD1+ EM CD8+ T cells and granzyme K+ PD1+ EM CD8+ T cells compared to young WT and Ldlr−/− mice. PCSK9 = proprotein convertase subtilisin/kexin type 9 serine protease, AAV = adeno-associated virus, CM = central memory, EM = effector memory.

Extended Data Fig. 3. Similar fasting total cholesterol level in young and aged C57BL/6 WT mice treated with anti-IgG or anti-CD8 treatment over 10-weeks of WD feeding.

Fasting total cholesterol was quantified from plasma via colorimetric assay. 2-way ANOVA with Tukey’s post-hoc test. N = 11 young WT anti-CD8a mice, N = 10 young WT anti-IgG mice, N = 16 aged WT anti-CD8 mice, N = 17 aged WT anti-IgG mice, N = 6 young Ldlr−/− anti-IgG mice, and N = 7 young Ldlr−/− anti-CD8 mice over 3 biological replicates. Data in this figure represents mean ± SEM. Measurements were taken from distinct samples. Ldlr = low-density lipoprotein receptor. WD = western diet, WT = wild-type.

Extended Data Fig. 4. CD8+ T cell depletion reduces brachiocephalic artery atherosclerotic plaque size in Aged WT mice but not young WT mice.

Murine PCSK9-AAV-induced hypercholesterolemia model of atherosclerosis with anti-CD8 or anti-IgG antibody injection every 2 weeks during the WD feeding period. Representative histology and atherosclerotic plaque size quantification of the brachiocephalic artery of young PCSK9-AAV (a), aged PCSK9-AAV (b), and Ldlr−/− mice (c) treated with either anti-CD8 antibody treatment or anti-IgG treatment. Scale bar = 100 μm. In A-C, data are pooled from 3 independent experiments and all data are shown. N = 10 for young anti-IgG, N = 11 for young anti-CD8, N = 17 for aged anti-IgG, N = 16 for aged anti-CD8, N = 7 for young Ldlr−/− anti-IgG, and N = 6 for young Ldlr−/− anti-CD8 over 3 independent experiments. Measurements were taken from distinct samples. Data in this figure represents mean ± SEM. 2-Way ANOVA with Tukey’s post-hoc test. BCA = brachiocephalic artery, CM = central memory, EM = effector memory, PCSK9 = proprotein convertase subtilisin/kexin type 9 serine protease.

Extended Data Fig. 5. Anti-CD8 treatment significantly reduces the number of CD8+ T cells in atherosclerotic aortas.

Aortas from aged mice transfected with PCSK9-AAV and subjected to 10-weeks of western diet feeding and either treated with anti-CD8 or anti-IgG isotype control antibody during western diet feeding were digested and analyzed by flow cytometry. The flow cytometry gating strategy is shown demonstrating CD4+ and CD8+ T cells in anti-CD8 and anti-IgG treated atherosclerotic mice. Quantification demonstrates a significant reduction of CD8+ T cells in anti-CD8 treated mice compared to isotype control. N = 5 anti-IgG treated mice and N = 5 anti-CD8 treated mice over 1 independent experiment. Data in this figure represents mean ± SEM. Each point is a biological replicate and measurements were taken from distinct samples. Two-tailed Mann-Whitney U test.

Extended Data Fig. 6. CD8+ T cell enrichment from young and aged C57BL/6 WT donor spleens.

(a) Flow cytometry gating strategy for young and aged WT mice before and after CD8+ T cell enrichment prior to CD8+ T cell adoptive transfer into Cd8−/− mice showing gating and frequency of previous gate. (b) Representative flow cytometry plots demonstrating CD4+ and CD8+ T cells, memory and naive CD8+ T cells, PD1+ effector memory CD8+ T cells, and PD1+ central memory CD8+ T cells showing gating and frequency of parent gate.

Extended Data Fig. 7. Similar fasting total cholesterol level in young Cd8−/− mice adoptively transferred with 10 million CD8+ T cells or vehicle from either young C57BL/6 WT or aged C57BL/6 WT mice over 10-weeks of WD feeding.

All mice were adoptively transferred and allowed to rest for 4-weeks before PCSK9-AAV injection and 10-week WD feeding. Fasting total cholesterol was quantified from plasma via colorimetric assay. 2-way ANOVA with Tukey’s post-hoc test. N = 13 young Cd8−/− mice + young CD8+ T cells, N = 8 young Cd8−/− mice + aged CD8+ T cells, and N = 7 young Cd8−/− mice + no CD8+ T cells over 2 independent experiments. Data in this figure represents mean ± SEM. Measurements were taken from distinct samples. WD = western diet, WT = wild-type.

Extended Data Fig. 8. Human atherosclerotic plaque CD8 T cells express GZMK and associate with symptomatic atherosclerotic disease.

(a) UMAP plot of T cells from human atherosclerotic plaques from public study deposited to Zenodo: 3361716. (b) GZMK expression overlayed on UMAP plot of T cells in human atherosclerotic plaques with dashed line indicating cells with the highest GZMK expression. (c) top 15 differentially expressed genes in the CD8+ T cell cluster. (d) Violin plot of GZMK expression on 3 T cell clusters, stratified by symptomatic group. For D, two-tailed Venice non-parametric benchmarking method within BBrowser3 was used to compare groups (see Methods) and data represents mean with lines above and below representing 1st and 3rd quartiles. Raw data for this figure is from a publicly available dataset archived on Zenodo: 3361716. UMAP = uniform manifold approximation projection.

Extended Data Fig. 9. Atherosclerotic plaque T cells are more clonally expanded than splenic T cells and aging enhances clonal expansion.

(a) Abundance of unique clonotypes. (b) Clonotype diversity estimation by Chao1 index. (c) Length of CDR3 sequences by group. (d) Size of clonotypes by group. (e) Clonotype tracking of the top 5 clonotypes from each sample across all other samples. (f) Overlap of CDR3 sequences by sample in circos plot. Top 100 expanded CDR3 sequences, colored by group, of 15 amino acids (g), 10 amino acids (h), and 5 amino acids (i). Composition of amino acid sequences of different lengths including 15 amino acids (j), 10 amino acids (k), and 5 amino acids (l). Depiction of which samples had the greatest clonal expansion shared across the number of cells (x-axis) and number of samples (y-axis) and stratified by group (m), age (n), and tissue-type (o). Data includes young spleens (N = 3 biological replicates and 18,412 total cells), aged spleen (N = 3 biological replicates and 11,151 total cells), young aorta (N = 4 biological replicates and 1,999 total cells), and aged aorta (N = 4 biological replicates and 5,698 total cells) over 1 independent experiment.

Extended Data Fig. 10. Aged atherosclerotic mice have more CD3+, CD8+, and F4/80+ cells in the aorta compared to young mice.

Aortas from 3-mo old and 18-mo old mice transfected with PCSK9-AAV and subjected to 10-weeks of western diet feeding were digested and analyzed by flow cytometry. (a) The flow cytometry gating strategy is shown CD3+, CD8+, and CD11b+F4/80+ cells from the aortas of atherosclerotic mice. Quantification demonstrates a significant increase of CD3+ and CD8+ T cells as well as CD11b+F4/80+ macrophages in aged atherosclerotic mice compared with young atherosclerotic mice normalized to total aorta weight (b-d) or by mg of aorta weight (e-g). In B-G, each point is a biological replicate and measurements were taken from distinct samples. N = 6 aged and N = 7 young mice at each timepoint for a total of N = 18 aged mice and N = 21 young mice over 1 independent experiment. Data in this figure represents mean ± SEM. 2-Way ANOVA with Šídák’s post-hoc test. PCSK9 = proprotein convertase subtilisin/kexin type 9 serine protease, AAV = adeno-associated virus, CM = central memory, EM = effector memory, WD = western diet.

Data availability

Raw and processed mouse single-cell RNA sequencing and TCR sequencing data are deposited to GEO repository (GSE210719). Genome builds used are found here: https://support.10xgenomics.com/single-cell-gene-expression/software/release-notes/build. Human atherosclerotic plaque data were reanalyzed from publicly available data archived on Zenodo: 3361716. The source data fore figures in this manuscript are published along with the manuscript. Raw data are available by the corresponding author upon request.