Physiologic Thymic Involution Underlies Age-Dependent Accumulation of Senescence-Associated CD4⁺ T Cells

The thymus shows involution with age in humans and mice, resulting in the progressive decline of T cell output with age (1). The mechanisms underlying the physiologic thymic involution remain controversial and may involve multiple factors, including age-related changes in hematopoietic cells and the thymic microenvironment (2, 3). Although the decrease in thymic organ size becomes evident after adolescence (4), quantitative and qualitative changes in thymic epithelial cells are recognizable much earlier (5, 6). We reported that thymic epithelial cell stem cell activity is rapidly diminished soon after birth, apparently at the cost of robust T cell production during the late fetal and neonatal stages, and it may contribute to the early thymic involution (7–9). The physiologic thymic involution and resulting decrease in T cell output with age should greatly influence the dynamic homeostasis of the peripheral T cell population, although the overall impacts on immune function remain elusive.

Despite the progressive decline in T cell output with age, the overall peripheral T cell population is largely maintained throughout life. However, the T cell population shows a remarkable shift from naive (CD62L^high CD44^low) to memory (CD62L^low CD44^high) phenotype with age (10–12). Because Ag-independent, homeostatic proliferation results in phenotypic transition of peripheral naive T cells to memory phenotype in mice (11, 13), the age-dependent shift in the T cell population from naive to memory phenotype may reflect, in part, increasing homeostatic proliferation of peripheral T cells to compensate for diminishing thymic T cell output (12).

A minor subpopulation expressing PD-1 and, to a lesser extent, CD153, is increased in the memory phenotype CD4⁺ T cells as mice age (10, 14). PD-1⁺ CD44^high CD4⁺ T cells, in particular those expressing CD153, exhibit features of cellular senescence, such as a marked increase in senescence-related genes (Cdkn1, Cdkn2) and senescence-associated heterochromatin foci (SAHF), with a profound defect in the Ag-driven proliferation capacity; thus, they are referred to as senescence-associated T (SA-T) cells (14). SA-T cells show additional unique functional features, including biased secretion of abundant proinflammatory cytokines, such as osteopontin (OPN) and chemokines (Ccl3, Ccl4), reminiscent of the SA-secretory phenotype (15), as well as induction of spontaneous germinal centers (GCs) (14). SA-T cells are increased robustly and prematurely in lupus-prone mice and play a crucial role in anti-nuclear autoantibody production and resulting lupus nephritis (13, 15). A recent study reported that SA-T cells also accumulate in visceral adipose tissues under a high-fat diet, causing insulin resistance (16). Thus, SA-T cells may play a role in immune aging phenotypes, such as the decline of acquired immunity, proinflammatory traits, and a higher risk for autoimmunity (17–19). However, the exact mechanism leading to CD4⁺ T cell senescence remains to be seen.

In the current study, we investigated the possible contribution of physiologic thymic involution to age-dependent accumulation of SA-T cells. We demonstrate that the environment of aged mice is highly permissive for spontaneous, Ag-independent proliferation of naive CD4⁺ T cells, which eventually lead to the development of SA-T cells via extensive cell division. Thymectomy at the early adult stage (ATX) markedly accelerated the development of SA-T cells, whereas embryonic thymic implantation at the late adult stage (ATI) delayed the accumulation of SA-T cells. Our findings suggest that the age-dependent increase in SA-T cells is an inevitable consequence of continuous homeostatic proliferation of peripheral naive CD4⁺ T cells due to the progressive decline of thymic T cell output with age.

CD45.2 C57BL/6 (B6) mice were purchased from CLEA Japan (Tokyo, Japan) and Japan SLC (Shizuoka, Japan). CD45.1 mice were purchased from Charles River Laboratories Japan (Kanagawa, Japan). CD3ε^−/− mice were kindly provided by Dr. S. Fagarasan (RIKEN Center for Integrative Medical Sciences, Kanagawa, Japan). OPN-EGFP knock-in mice were described previously (14). All mice were maintained under specific pathogen–free conditions at the Centre for Experimental Animals of Kyoto University, and the animal experiments were performed in accordance with institutional guidelines.

Multicolor flow cytometric analysis and cell sorting were performed using a FACSCanto and FACSAria II/III (BD Biosciences, San Jose, CA), respectively. The following Abs were used: biotin-conjugated anti-CXCR5, PE-conjugated anti-CD121b, CD95, BV421-conjugated streptavidin, and BV510-conjugated anti-CD44 (all from BD Biosciences); biotin-conjugated anti–CD153, FITC-conjugated anti-CD44, anti-Ki67, PE–conjugated streptavidin, and Alexa–eFluor 660–conjugated anti-GL7 (all from eBioscience, San Diego, CA); Alexa Fluor 647–conjugated anti–PD-1, PerCP-Cy5.5–conjugated anti-B220, anti–Mac-1, anti-TER119, allophycocyanin-Cy7–conjugated anti-CD45.1, anti-B220, Pacific Blue–conjugated anti–Mac-1, anti-TER119, and anti-B220 (all from BioLegend, San Diego, CA); PE-Cy7–conjugated anti-CD4 and anti-CD19 (both from Tonbo Biosciences, San Diego, CA); and FITC-conjugated anti-γH2AX (Cell Signaling Technology, Danvers, MA). Propidium iodide (Sigma-Aldrich, St. Louis, MO) or Ghost Dye Violet 450 (Tonbo Biosciences) was used for excluding dead cells. For BrdU staining, the cells were immunostained with PE–anti-CD3 and PerCP-Cy5–anti-CD4, fixed and permeabilized, treated with DNase, and then immunostained with FITC–anti-BrdU (BD Biosciences).

CD4⁺ T cells were purified from spleen and lymph node cells from 8-wk-old CD45.1 B6 mice by magnetic-bead depletion with biotin-conjugated CD8a (Tonbo Biosciences), CD11c, CD19, CD49b, CD105, CD117, and PD-1 (all from BD Biosciences), and CD11b, B220, and TER119 (all from BioLegend), as well as anti-biotin beads (Miltenyi Biotec, Bergisch Gladbach, Germany). FACS analysis confirmed that 93–97% of the CD4⁺ T cell population was CD44^low CD62L^high. The cells were labeled with CellTrace Violet (CTV; Thermo Fisher Scientific, Waltham, MA), according to the manufacturer’s instructions and injected i.v. at 2 × 10⁶ cells per mouse. In some experiments, young and aged mice were kept in the same cages for more than a month before cell transfer.

Spleens were snap-frozen in optimum cutting temperature compound (Sakura, Torrance, CA), and serial frozen sections were fixed with 95% ethanol and 100% acetone, followed by immunostaining. Abs included anti-B220, anti-CD4, anti–PD-1, anti-CD11c, anti-GL7 (all from BioLegend), and anti-CD3 (BD Biosciences). Immunostained sections were mounted in Mowiol (Calbiochem, La Jolla, CA) and examined under a microscope (Carl Zeiss, Oberkochen, Germany). Tile images of whole-tissue sections were generated using the MosaiX tool in Zeiss AxioVision software (Carl Zeiss). For heterochromatin staining, cells were cytospun, fixed for 10 min in 4% paraformaldehyde, and permeabilized with 0.1% Triton X-100 PBS for 10 min. After blocking, the cells were incubated with anti–HP-1β Ab (Millipore, Billerica, MA), followed by fluorophore Alexa Fluor 488–conjugated secondary Ab and DAPI.

OPN in the culture supernatants was assessed by with ELISA kits (cat. no. DY441; R&D Systems, Minneapolis, MN).

Quantitative PCR analysis was performed using SYBR Green I Master on a LightCycler 480 Instrument (both from Roche, Basel, Switzerland). Cyclophilin was used as an internal control for mRNA. For single-cell PCR, RNA from single cells sorted with a FACSAria II/III (BD Biosciences) was isolated and amplified on a C1 Single-Cell Auto Prep System (Fluidigm, San Francisco, CA). Quantitative PCR analysis was performed with SsoFast EvaGreen Supermix with Low ROX (PN 172-5211; Bio-Rad, Hercules, CA) on a Biomark HD real-time PCR system (Fluidigm). Hierarchal clustering analysis of the gene expression in single cells was defined based on the Pearson correlation, and heat maps and cluster dendrograms were created with the group average method using R software (version 3.3.1; R Foundation for Statistical Computing, Vienna, Austria). Primer sequences are as follows: Ascl2, 5′-GAGAGCTAAGCCCGATGGA-3′ and 5′-AGGTCCACCAGGAGTCACC-3′; Bcl6, 5′-TTCCGCTACAAGGGCAAC-3′ and 5′-CAGCGATAGGGTTTCTCACC-3′; Blimp1, 5′-TGCGGAGAGGCTCCACTA-3′ and 5′-TGGGTTGCTTTCCGTTTG-3′; Ccl3, 5′-TCTGTCACCTGCTCAACATCA-3′ and 5′-CGGGGTGTCAGCTCCATA-3′; Ccl4, 5′-AGCAACACCATGAAGCTCTG-3′ and 5′-GAGGGTCAGAGCCCATTG-3′; Cd30, 5′-GTCCACGGGAACACCATTT-3′ and 5′-CCAACCAGTAGCACCACCAT-3′; Cdkn1a, 5′-AACATCTCAGGGCCGAAA-3′ and 5′-TGCGCTTGGAGTGATAGAAA-3′; Cdkn1b, 5′-GTTAGCGGAGCAGTGTCCA-3′ and 5′-TCTGTTCTGTTGGCCCTTTT-3′; Cdkn2a, 5′-GGGTTTTCTTGGTGAAGTTCG-3′ and 5′-TTGCCCATCATCATCACCT-3′; Cdkn2b, 5′-AATAACTTCCTACGCATTTTCTGC-3′ and 5′-CCCTTGGCTTCAAGGTGAG-3′; Cebpa, 5′-AAACAACGCAACGTGGAGA-3′ and 5′-GCGGTCATTGTCACTGGTC-3′; Cebpb, 5′-TGATGCAATCCGGATCAA-3′ and 5′-CACGTGTGTTGCGTCAGTC-3′; Cxcr4, 5′-TGGAACCGATCAGTGTGAGT-3′ and 5′-GGGCAGGAAGATCCTATTGA-3′; Cxcr5, 5′-GAATGACGACAGAGGTTCCTG-3′ and 5′-GCCCAGGTTGGCTTCTTAT-3′; Ifng, 5′-ATCTGGAGGAACTGGCAAAA-3′ and 5′-TTCAAGACTTCAAAGAGTCTGAGG-3′; IL10, 5′-CAGAGCCACATGCTCCTAGA-3′ and 5′-TGTCCAGCTGGTCCTTTGTT-3′; Il1r2, 5′-CCCATCCCTGTGATCATTTC-3′ and 5′-GCACGGGACTATCAGTCTTGA-3′; Il21, 5′-TCATCATTGACCTCGTGGCCC-3′ and 5′-ATCGTACTTCTCCACTTGCAATCCC-3′; Pdcd1, 5′-CTACCTCTGTGGGGCCATC-3′ and 5′-GAGGTCTCCAGGATTCTCTCTGT-3′; Ptpn11, 5′-AGCTGGCTGAGACCACAGAT-3′ and 5′-TGTTGCTGGAGCGTCTCA-3′; Sostdc1, 5′-AACAGCACCCTGAATCAAGC-3′ and 5′-CAGCCCACTTGAACTCGAC-3′; Tbx21, 5′-CAACCAGCACCAGACAGAGA-3′ and 5′-ACAAACATCCTGTAATGGCTTG-3′; and Tnfsf8, 5′-GAGGATCTCTTCTGTACCCTGAAA-3′ and 5′-TTGGTATTGTTGAGATGCTTTGA-3′.

Mice (7–8 wk of age) were anesthetized with pentobarbital, and the thymus was removed with vacuum suction (ATX group). Any mice with detectable thymic remnants at necropsy were excluded from the study. Mice that received a sternum incision under anesthesia without thymic suction (sham operated) were used as the control group. Six or seven whole thymic lobes from embryonic day 14.5 (E14.5) B6 embryos were implanted under the left kidney capsule of B6 mice, as described previously (7). After 3 mo, the mice were sacrificed for multicolor flow cytometric analysis.

BrdU (Sigma-Aldrich) was added to the drinking water (0.8 mg/ml) for 9 d before the analysis of ATX and sham-operated mice.

Statistical analyses were performed using an unpaired two-tailed Student t test and one-way ANOVA with the Bonferroni post hoc test. A p value <0.05 was considered to indicate a significant difference.

We isolated naive CD4⁺ T cells from the spleen of 2-mo-old CD45.1 B6 mice, labeled them with CTV, and transferred them into young (2-mo-old) or aged (12–20-mo-old) CD45.2 B6 mice with no manipulations. In young recipients, the donor CD4⁺ T cells showed minimal proliferation in the spleen until 21 d after transfer (Fig. 1A). In contrast, the same donor CD4⁺ T cells exhibited robust proliferation in aged mice as early as 5 d after transfer, and nearly 80% of the donor cells, on average, had undergone extensive cell divisions by day 21 (Fig. 1A). FACS analysis revealed that the proportions of PD-1⁺ CD44^high cells were progressively increased in the donor CD4⁺ T cell progenies in aged recipients, reaching >60% at day 21, whereas most donor T cells remained CD44^low, with only 15% being PD-1⁺ CD44^high in young recipients at day 21 (Fig. 1B). The expression of PD-1 in aged recipients was confined to donor T cells that underwent extensive cell divisions (Fig. 1C). A significant fraction of donor-derived PD-1⁺ CD44^high CD4⁺ T cells in aged recipients also expressed CXCR5 and, to a lesser extent, CD153 and CD121b, essentially similar to host endogenous SA-T cells, although the proportions varied markedly in individual recipients (Fig. 1D). Donor-derived and host CD4⁺ T cells in young recipients rarely expressed these markers (Fig. 1D). Because homeostatic T cell proliferation can be affected by an altered commensal microbiome in the hosts, we cohoused aged and young mice for more than a month to equalize their microbiomes before cell transfer. Although the aged mice cohoused with young mice again allowed potent naive CD4⁺ T cell proliferation, donor CD4⁺ T cells showed minimal proliferation in the young mice cohoused with aged mice, similar to those that were kept alone (Supplemental Fig. 1); this suggests that possible changes in microbiomes with age may play a minimal role, if any, in CD4⁺ T cell proliferation. The results suggest that naive CD4⁺ T cells show robust proliferation spontaneously in the aged mouse environment, resulting in phenotypic conversion resembling endogenous SA-T cells.

To investigate the role of homeostatic proliferation under T-lymphopenic conditions in the characteristic phenotypic conversion of naive CD4⁺ T cells, we transferred naive CD4⁺ T cells from CD45.1 B6 mice into young CD45.2 B6 mice irradiated with 5 Gy gamma irradiation. Donor CD4⁺ T cells exhibited an increase in CD44 expression within a week, and PD-1 expression in the CD44^high fraction was markedly increased with a week delay (Fig. 2A). Furthermore, at day 14, significant fractions of donor-derived PD-1⁺ CD4⁺ T cells also expressed CXCR5 and CD153, whereas host CD4⁺ T cells expressed minimal PD-1, with undetectable CXCR5 or CD153, as expected (Fig. 2A, left panels). The numbers of PD-1⁺ and CD153⁺ cells increased robustly during the progression of donor T cell proliferation until total CD4⁺ T cell numbers reached a plateau (Fig. 2A, right panels). A marked increase in CD44 expression, as well as PD-1 and CD153 expression, became detectable only after extensive cell divisions of donor CD4⁺ T cells (Fig. 2B). In particular, the expression of PD-1 and CD153 was seen only in minor fractions of CTV⁻ cells, suggesting that the phenotypic conversion occurred only after many rounds of cell divisions much exceeding the detection limit with the CTV dilution. It was noted that CD5 expression also tended to be increased in CTV⁻ cells (mean fluorescence intensity, 9489 versus 6734 in undivided cells). The expression induction profiles of PD-1 and CD153 were distinct from those in naive CD4⁺ T cells optimally stimulated with anti-CD3+CD28 Abs, in which PD-1 and CD153 were rapidly induced even before initial cell division and tended to decrease as cell division progressed (Supplemental Fig. 2A). Thus, it seemed unlikely that expression of PD-1 and CD153 during homeostatic proliferation merely reflects transient cell activation. Using CD3ε^−/− recipients, we confirmed similar generation of PD-1⁺ and CD153⁺ cells from transferred naive CD4⁺ T cells (Fig. 2C). The PD-1⁺ and CD153⁺ cells were developed rapidly in 10 d and were sustained significantly for >2 mo, when the total CD4⁺ T cell numbers reached a plateau (Supplemental Fig. 2B). The results suggest that robust cell divisions during acute homeostatic proliferation of naive CD4⁺ T cells result in the generation of cells resembling SA-T cells.

We next compared the transcriptomes of PD-1⁺ and CD153⁺ CD44^high CD4⁺ T cells derived from naive CD4⁺ T cells during acute homeostatic proliferation with those of endogenous SA-T cells in aged mice. PD-1⁺, and to a greater extent CD153⁺, CD4⁺ T cells derived from naive CD4⁺ T cells in CD3ε^−/− recipients showed a significant increase in the expression of senescence-related (Cdkn1a, Cdkn2b) and follicular T cell–related (Bcl6, Cebpa) genes compared with donor-derived PD-1⁻ CD4⁺ T cells, as well as naive CD4⁺ T cells before transfer, similar to those of freshly isolated SA-T cells from aged mice (Fig. 3A). In addition, PD-1⁺ and CD153⁺ CD4⁺ T cells secreted abundant OPN upon TCR stimulation (Fig. 3A), again similar to SA-T cells (14). Taking advantage of the rapid development of these cells from naive CD4⁺ T cells in CD3ε^−/− mice, we investigated the possible expression coordination of diverse genes that are overexpressed in SA-T cells (14), using single-cell RT-PCR analysis. Hierarchal clustering analysis suggested that three groups of genes are induced in a coordinated manner in single cells, including genes related to negative regulation of cell proliferation (Pdcd1, Ptpn11, Cdkn1a, Cdkn2a, Cdkn2b), inflammation (Cebpb, Blimp1, Ccl3, Ccl4, Ifng), and follicular T cells (Bcl6, Il21, Cxcr5, Ascl2) (Fig. 3B). Notably, the increase in Tnfsf8 (Cd153) was highly coordinated with that of Ascl2, which is essential for follicular T cell development (20). Furthermore, these PD-1⁺ and CD153⁺ CD4⁺ T cells that developed from naive CD4⁺ T cells also showed markedly increased SAHF and γH2AX expression (Fig. 3C, 3D), again similar to SA-T cells in aged mice. Altogether, the findings strongly suggest that a portion of naive CD4⁺ T cells that have undergone extensive homeostatic proliferation are converted to SA-T cells.

Given that SA-T cells are often associated with spontaneous GCs in aged mice (14), we also examined the localization of transferred naive CD4⁺ T cells in CD3ε^−/− recipients. Three weeks after transfer of naive CD4⁺ T cells, the majority of donor T cells were detected diffusely in the follicular regions of the spleen but only minimally in vestigial T cell regions rich in CD11c⁺ dendritic cells (Fig. 4A, top panels). Among them, PD-1⁺ T cells were mostly confined to the IgD^low region (Fig. 4A, middle panels), which contained abundant GL-7⁺ cells (Fig. 4A, bottom panels). FACS analysis confirmed that the GL-7⁺ cells represent CD95⁺ B cells indicative of GC B cells (Fig. 4B), suggesting that the robust increase in donor PD-1⁺ T cells was associated with spontaneous GC formation. We previously reported that OPN derived from SA-T cells plays a crucial role in the spontaneous GC reaction (14, 21). Likewise, a minor, yet significant, fraction of PD-1⁺ CD4⁺ T cells derived from naive CD4⁺ T cells from EGFP-OPN reporter mice in CD3ε^−/− recipients showed potent GFP expression, indicative of de novo Spp1 activation (Fig. 4C). Kinetic analysis indicated that the increase in PD-1⁺ and CD153⁺ CD4⁺ T cells preceded the increase in CD95⁺ GL-7⁺ GC B cells by ∼2 wk (Fig. 4C). The results suggest that the robust increase in SA-T cells during acute homeostatic proliferation in follicular regions leads to spontaneous B cell activation and GC formation.

To address the impact of thymic involution on the generation of SA-T cells, we examined the effect of ATX at 2 mo of age. ATX mice exhibited a rapid decrease in naive CD4⁺ T cells in the spleen after ATX that occurred at a greater rate than in sham–thymectomy-operated (STX) mice, whereas memory phenotype CD4⁺ T cells remained unchanged, resulting in a significant increase in the proportion of memory phenotype cells in the CD4⁺ T cell population, reminiscent of aged mice (Fig. 5A). When the mice were administered BrdU at 9 d before analysis, ATX mice showed a marked increase in BrdU⁺ CD4⁺ T cells, indicating the increased spontaneous proliferation of peripheral CD4⁺ T cells after ATX (Fig. 5B). Concordantly, the proportions and numbers of PD-1⁺ and PD-1⁺ CD153⁺ CD44^high CD4⁺ T cells were increased in ATX mice compared with STX mice, although the increase in the latter cell numbers barely reached statistical significance because of the very small cell numbers (Fig. 5C, Supplemental Fig. 3). We confirmed that the PD-1⁺ and CD153⁺ CD4⁺ T cell populations that developed in ATX mice exhibited unique gene-expression profiles essentially similar to the corresponding cell fractions in healthy aged mice (Fig. 5D). The results suggest that the decline in thymic function led to the accelerated generation of SA-T cells through increased homeostatic T cell proliferation.

Finally, we addressed whether resuming thymic function at the late adult stage could affect the increase in SA-T cells. To this end, we implanted embryonic (E14.5) thymi (ATI) under the unilateral kidney capsules of B6 mice at the ages of 11–15 mo. ATI mice were sacrificed at 3 mo after the thymic implantation, and those bearing well-developed ectopic thymi with normal T cell–differentiation profiles were subjected to analysis (Fig. 6A, Supplemental Fig. 4A). The proportions of PD-1⁺ and CD153⁺ CD44^high cells in the splenic CD4⁺ T cell population were significantly reduced in ATI mice compared with sham–thymus-implanted (STI) mice, whereas those of naive CD4⁺ T cells were increased (Fig. 6A). To examine whether the effects in ATI mice were related to homeostatic proliferation, we transferred CTV-labeled naive CD4⁺ T cells from CD45.1 B6 mice into CD45.2 B6 mice that underwent ATI 5 mo prior, at the age of 9 mo (Fig. 6B). ATI mice showed significantly increased CD4⁺ T cell numbers compared with STI mice (Supplemental Fig. 4B), indicative of resumed T cell output from the ectopic thymi. At day 9 after transfer, donor naive CD4⁺ T cells exhibited significantly reduced cell proliferation in ATI mice (Fig. 6B). FACS analysis confirmed that the vast majority of donor T cells retained the naive phenotype in ATI recipients, whereas the proportions of naive phenotype cells were variably decreased in STI mice (Fig. 6B). Concordantly, PD-1⁺ donor T cells were markedly repressed in ATI recipients compared with STI mice; CD153⁺ donor T cells were quite rare in ATI and STI mice in these experiments (Fig. 6B). These effects paralleled the changes in the CD4⁺ T cells of the hosts. Thus, host PD-1⁺ CD44^high CD4⁺ T cells were also reduced significantly, although the decrease in CD153⁺ CD44^high CD4⁺ T cells was only slight, with no statistical significance (Fig. 6B). The disproportional changes cannot be explained by a mere dilution effect with the increase in naive CD4⁺ T cells in ATI mice; rather, they may reflect the much longer stability of CD153⁺ cells than PD-1⁻ and PD-1⁺ CD153⁻ cells (14). To confirm that these effects in ATI mice are due to resumed T cell output, rather than a functional thymus per se, we continuously transferred naive CD4⁺ T cells from CD45.1 mice into aged CD45.2 mice and analyzed CD45.2⁺ CD4⁺ T cells in the hosts. We found that repetitive transfer of naive CD4⁺ T cells once a week for 3 mo significantly repressed the increase in host CD153⁺ SA-T cells (Supplemental Fig. 4C). These results suggest that resumed thymic T cell output at the late adult stage restrains the environmental permissiveness for CD4⁺ T cell homeostatic proliferation and, thereby, compromises the accumulation of SA-T cells with age.

PD-1⁺ and CD153⁺ CD44^high CD4⁺ T cells (SA-T cells), which are rarely found in young mice, gradually increase and accumulate as mice age. SA-T cells show characteristic features related to cellular senescence, and the dominance of SA-T cells in the CD4⁺ T cell population may account, at least in part, for immune aging phenotypes, including diminished acquired immunity, proinflammatory traits, and increasing autoimmunity risk (10, 14, 16, 21).

In the current study, we investigated the physiologic mechanism of the age-dependent increase in SA-T cells. We found that transferred naive CD4⁺ T cells robustly proliferate in healthy aged mice, but rarely in young mice, and a portion of the progenies that have undergone extensive cell division is eventually converted to cells with features similar to SA-T cells in aged mice. The findings are consistent with our previous observation that the generation of SA-T cells is concordant with an increase in spontaneous CD4⁺ T cell proliferation with age (14). A recent study indicates that the capacity of homeostatic proliferation of naive T cells following gamma irradiation causing acute lymphopenia is more potent in young mice than in aged mice (22). However, our current findings suggest that, under normal steady-state conditions, the endogenous force driving T cell homeostatic proliferation operates constitutively in the environment of aged mice, whereas it is minimal in that of young mice. We confirmed that T cells resembling SA-T cells in aged mice are robustly developed from naive T cells, even in young mice following gamma irradiation or in CD3ε^−/− mice through extensive homeostatic proliferation, irrespective of the possible direct driving factors involved. The PD-1⁺ and, to a greater extent, the CD153⁺, CD44^high CD4⁺ T cells that developed from naive CD4⁺ T cells under T lymphopenic conditions in young mice showed a marked increase in senescence-related Cdk-inhibitor genes (Cdkn1a, Cdkn2), SAHF, and γH2AX, suggestive of replicative cell senescence (23, 24). These T cells also showed increased expression of various proinflammatory genes (Spp1, Ccl3, Ccl4, Ifng), reminiscent of the SA-secretory phenotype, and follicular T cell signature genes (Bcl-6, Cebpa, Ascl2, Cxcr5, IL-21). These features were essentially identical to those of endogenous SA-T cells in aged mice. Notably, it was suggested that the expression of these diverse genes is coordinately induced as functional clusters in individual cells during homeostatic proliferation. In any case, it was suggested that SA-T cells develop as a direct consequence of extensive homeostatic proliferation of naive CD4⁺ T cells.

Although PD-1 and CD153 are rapidly and transiently expressed by CD4⁺ T cells with optimal TCR stimulation in a manner independent of the cell cycle, naive CD4⁺ T cells express PD-1 and CD153 only after extensive rounds of cell division that may far exceed the detection limit with CTV dilution (eight times) during homeostatic proliferation. In cell-transfer experiments with T-lymphopenic mouse models, such extensive cell division of donor naive CD4⁺ T cells leading to the generation of SA-T cells may be achieved in the early phase of homeostatic proliferation before T cell numbers are restored. However, donor-derived SA-T cells persisted for >2 mo in CD3ε^−/− mice, suggesting their stability. This may be consistent with much greater stability of SA-T cells compared with naive CD4⁺ T cells in culture (14). Ag-driven CD4⁺ T cell proliferation is associated with effector T cell differentiation with a metabolic switch (25). It seems possible that extensive homeostatic proliferation of CD4⁺ T cells not associated with such a metabolic switch may eventually lead to the cellular senescence. Recent studies indicate that repetitive Ag stimulation in CD8⁺ T cells during chronic infection also results in the stable expression of PD-1, whose signaling causes the functional exhaustion via metabolic alterations (26, 27). The exhausted Ag-specific CD8⁺ T cells show a characteristic transcriptome (26), which is apparently distinct from CD4⁺ SA-T cells (14), and the effects can be reversed by PD-1 signal blockade. In contrast, we reported previously that PD-1 signaling apparently plays no significant role in the generation and function of SA-T cells, because these T cells are similarly increased in Pd1^−/− mice with age, and the functional features are not reversed by PD-1 blockade (10, 14). Thus, it seems that the overall features of CD4⁺ SA-T cells reflect cellular senescence associated with extensive cell divisions rather than Ag-driven metabolic exhaustion.

An increasing endogenous force driving peripheral T cell homeostatic proliferation with age may be attributable, at least in part, to the physiologic thymic involution resulting in the age-dependent decline in T cell output (12, 28, 29). In humans, the proportions of proliferating naive phenotype T cells increase progressively with age (30), and this is accelerated in those who underwent thymectomy in early childhood (31). In contrast, in mice, the proliferating T cells with naive phenotype increase slightly with age; thus, it was claimed that peripheral T cell proliferation contributes minimally to the maintenance of naive phenotype T cells (30). However, a remarkable increase in proliferating cells is detected in memory phenotype T cells as mice age (30); thus, the apparent difference may be due, in part, to the rapid conversion of naive T cells to the memory phenotype following homeostatic proliferation in mice but not in humans (32). We confirmed that ATX caused a marked increase in the spontaneous proliferation of CD4⁺ T cells compared with STX mice. Although naive phenotype CD4⁺ T cells declined in ATX mice much faster than in STX mice, memory phenotype CD4⁺ T cell numbers remained the same as those in STX mice for ≥6 mo. Furthermore, PD-1⁺ and CD153⁺ T cells in the memory phenotype CD4⁺ T cells were significantly increased in ATX mice, similar to aged mice. It was confirmed that these T cells show the cell senescence features indistinguishable from SA-T cells in aged mice. The results suggest that thymic involution plays a crucial role in the age-dependent increase in SA-T cells.

An intriguing question is whether the accumulation of endogenous SA-T cells with age can be suppressed by resuming thymic T cell output after the thymic involution. To address the question, we implanted the embryonic thymi at the late adult stage. We found that the environment of aged ATI mice allowed significantly limited homeostatic proliferation of transferred naive CD4⁺ T cells compared with STI mice; accordingly, we recorded significantly reduced proportions of SA-T cells in ATI mice. Although it is possible that the presence of functional thymus per se limited peripheral T cell proliferation (e.g., via soluble factors), we also found that continuous administration of CD45.1 naive CD4⁺ T cells to CD45.2 mice suppressed the age-dependent increase in CD45.2 SA-T cells. Together, these results strongly suggest that diminished thymic T cell output due to physiologic thymic involution is a major cause underlying the increase in SA-T cells with age. Our current results indicate that the endogenous force driving the homeostatic proliferation of transferred naive CD4⁺ T cells is unexpectedly potent in healthy aged mice, and it is conceivable that the sustained homeostatic proliferation in the aged environment leads to the age-dependent increase in SA-T cells.

Homeostatic CD4⁺ T cell proliferation is driven by tonic TCR signaling via self–MHC class II and homeostatic cytokines (33–36), and this is consistent with the finding that the generation of SA-T cells is dependent on the presence of B cells (14). A recent report indicated that persistent self-antigens play an important role in the selective maintenance of long-lived CD4⁺ T cells with age (37). As such, sustained homeostatic proliferation may result in the selection of T cells with higher intrinsic TCR affinity for self-antigens (38–42) and may underlie certain overt autoimmune diseases (37, 38). Consistently, T cell populations that underwent homeostatic proliferation showed higher CD5 expression. We reported previously that SA-T cells are robustly increased in the splenic follicular regions of lupus-prone mice, where they secrete abundant OPN in response to autologous B cells and induce spontaneous GC reactions with anti-nuclear Ab production (14, 21). A more recent study indicated that SA-T cells accumulate in large numbers in visceral fat tissues of mice fed a high-fat diet and cause visceral adiposity with increased insulin resistance (16). In humans, it was reported that the mTOR inhibitor, which delays the onset of various age-related disorders (43), significantly improves influenza vaccination efficacy in the elderly, and the effect is associated with a decrease in peripheral T cells expressing PD-1 (44), although the exact features of the PD-1⁺ T cells remain to be determined. Thus, it is an intriguing possibility that the increasing dominance of SA-T cells in the peripheral T cell population with age plays a part in the development of various age-related disorders, including diminished acquired immunity, proinflammatory traits, and increased autoimmunity risk.

Although physiologic thymic involution that begins early in life strongly affects the homeostatic dynamics of the peripheral T cell population throughout the lifespan, its possible impact on overall immune function has remained controversial. Our current results indicate that thymic involution underlies the age-dependent increase in SA-T cells. The extent of immune aging may vary significantly elderly individuals, and SA-T cells may serve as a suitable biomarker for immune aging and a potential target controlling the disorders related to immune aging in humans.

We thank Dr. Sidonia Fagarasan for providing CD3ε^−/− mice.

The authors have no financial conflicts of interest.