Aging of Antiviral CD8⁺ Memory T Cells Fosters Increased Survival, Metabolic Adaptations, and Lymphoid Tissue Homing

The long-term preservation of antiviral T cell memory is a highly dynamic process that promotes the progressive molecular, phenotypic, and functional remodeling of its principal constituents, the populations of specific CD8⁺ memory T cells (T_M) distributed throughout and often trafficking between various anatomic compartments. We recently demonstrated that this process can culminate, paradoxically, in the acquisition of naive-like T cell traits, enhanced recall potential, and greater protective capacities of aged CD8⁺ T_M (1). The notion that aging can improve CD8⁺ T cell memory (1–5) stands in apparent contrast to much of the literature documenting numerous deleterious consequences of T cell aging (6–9). A more focused review (10), however, indicates that eventual “immunosenescence” is not necessarily a fate shared by all T cell subsets, and CD8⁺ T_M generated earlier in life to acute, nonpersisting pathogen challenges can be maintained over time without accruing obvious functional defects (9, 11, 12). In fact, the coexistence of age-associated alterations that either impair or improve CD8⁺ T cell immunity is illustrated in aged mice that exhibit a diminished capacity for generation of primary (I°) antiviral CD8⁺ effector T cell (T_E) responses yet readily support the greater secondary (II°) expansion of old as compared to young CD8⁺ T_M specific for the same viral determinants (1).

To elucidate the foundations and consequences of successful CD8⁺ T_M aging in greater detail, we previously generated a set of integrated data sets that collectively trace the evolving molecular, phenotypic, and functional properties of aging virus-specific CD8⁺ T_M (1). We further organized the patterns of gradual CD8⁺ T_M remodeling in a conceptual framework designated the “rebound model” of progressive CD8⁺ T_M “dedifferentiation” that postulates an inverse relationship between the extent of I° CD8⁺ T_E differentiation and the pace with which aging CD8⁺ T_M populations, over a period of ∼2 years, acquire a broad spectrum of distinctive and increasingly homogenous traits. In principle, progressive alterations of the CD8⁺ T_M compartment may emerge through a gradual conversion of individual CD8⁺ T_M (13) or an outgrowth of CD8⁺ T_M subsets with a competitive advantage (14). Although these processes are not mutually exclusive (15), work by us and others provides preferential support for the conversion hypothesis including direct evidence for the molecular, phenotypic, and functional evolution of the CD8⁺ central T_M (T_CM) subset (1–3, 5, 16). Important as these distinctions may be, they should not distract from the fact that regardless of conversion or outgrowth, the maturation of long-term CD8⁺ T cell memory is a population phenomenon that imparts potential biological relevance through changes pertaining to the overall composition and distribution of the specific CD8⁺ T_M compartment. Here, aging of CD8⁺ T_M populations modulates the expression of at least ∼80 cell surface receptors/ligands, produces a more diversified functional repertoire, and eventually endows old CD8⁺ T_M in a T cell–intrinsic fashion with an improved capacity for the generation of protective II° responses (1). In the present report, we have delineated the impact of aging on the cardinal components of CD8⁺ T_M homeostasis (survival, homeostatic proliferation, metabolism, tissue residence/trafficking), and our findings demonstrate that the cumulative homeostatic adaptations converge to establish a spatio-functional foundation for improved recall responses of aged CD8⁺ T_M.

All procedures involving laboratory animals were conducted in accordance with recommendations in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health, the protocols were approved by the Institutional Animal Care and Use Committees of the University of Colorado (permit numbers 70205604[05]1F, 70205607[05]4F and B-70210[05]1E), and Icahn School of Medicine at Mount Sinai (IACUC-2014-0170), and all efforts were made to minimize suffering of animals.

C57BL6/J (B6), congenic B6.CD90.1 (B6.PL-Thy1^a/CyJ), and congenic B6.CD45.1 (B6.SJL-Ptprc^a Pepc^b/BoyJ) mice were purchased from The Jackson Laboratory; p14 TCR transgenic mice were obtained on a B6.CD90.1 background from Dr. M. Oldstone (CD8⁺ T cells from these mice [p14 cells] are specific for the dominant lymphocytic choriomeningitis virus (LCMV)-GP_33–41 determinant restricted by D^b). We only used male mice in this study to avoid potential artifacts that may arise in gender mismatched adoptive transfer (AT) settings. LCMV Armstrong (clone 53b) was obtained from Dr. M. Oldstone (17), and stocks were prepared by a single passage on BHK-21 cells; plaque assays for determination of virus titers were performed as described/referenced. For I° challenges, 8- to 10-wk-old mice were infected with a single i.p. dose of 2 × 10⁵ PFU LCMV Armstrong; for II° challenges, naive recipients (aged 8–10 wk) of various CD8⁺ T_M populations were inoculated with 2 × 10⁵ PFU LCMV Arm i.p. All mice were housed under specific pathogen–free conditions and monitored for up to ∼2 y. Aging LCMV-immune mice were excluded from our study if they presented with 1) gross physical abnormalities such as lesions, emaciation, and/or weight loss, 2) lymphatic tumors as indicated by enlarged lymph nodes (LNs) at time of necropsy, or 3) T cell clonal expansions within the virus-specific CD8⁺ T_M compartment (D^bNP₃₉₆⁺, D^bGP₃₃⁺, or D^bGP₂₇₆⁺). According to these criteria, up to ∼30% of aging mice were excluded from the study.

Lymphocytes were obtained from blood, spleen, LNs, thymus, peritoneal cavity, and bone marrow (BM) according to standard procedures; for an estimate of total BM cells, the content from one femur was multiplied with a coefficient of 15.8 (18) (Fig. 7D). For isolation of lymphocytes from solid nonlymphoid tissues (NLTs; liver, lung, kidney), terminally anesthetized mice were sacrificed by total body perfusion with PBS and subsequent organ processing and gradient centrifugation as described (18). Enrichment of splenic T cells was performed with magnetic beads using variations and adaptations of established protocols. 1) For construction of p14 chimeras (1), p14 naive CD8⁺ T cells (T_N) (CD90.1⁺) were enriched from spleens of naive p14 mice by negative selection (EasySep Mouse CD8⁺ T Cell Enrichment Kit; StemCell Technologies) and transferred i.v. into B6 recipients at indicated numbers prior to LCMV infection 2–24 h later (Fig. 2J: 2 × 10²–2 × 10⁵; Fig. 3C, Supplemental Figs. 1H, 3F, 3G: 5 × 10⁴; Figs. 6E, 7A, Supplemental Fig. 3H: 2 × 10³, Fig. 7B: 1 × 10⁴). 2) Purification of p14 T_E/M for microarray analyses is described in Ref. 1. 3) Enrichment of CD8⁺ T_M from LCMV-immune B6 and B6-congenic donors was performed by depletion of B220⁺ cells (Miltenyi Biotec, Invitrogen/Dynal, or StemCell Technologies) followed by 1:1 combination at the level of D^bNP₃₉₆⁺ CD8⁺ T_M, i.v. AT of mixed populations containing 2 × 10³ D^bNP₃₉₆⁺ congenic CD8⁺ T_M each into naive congenic recipients and challenge with LCMV (Fig. 2I).

All reagents and materials used for analytical flow cytometry are summarized in Supplemental Table I, and our basic staining protocols are described and/or referenced in Ref. 1; in some cases, expression levels were normalized by dividing the geometric mean of fluorescence intensity (GMFI) of experimental stains by the GMFI of isotype control stains (Fig. 2I). Additional methodologies employed here include the use of various fluorescent dyes/probes [propidium iodide, 7AAD, YO-PRO-1, Zombie dyes, DiOC₆(3), dihydroethidium (HE), Alm Alx488, ThiolTracker Violet, Glut1.RBD.GFP (stained at 37°C for detection of surface Glut1), MitoTracker Green, tetramethylrhodamine (TMRE), and JC-1 dye according to manufacturer recommendations and/or published protocols (19–21) (Figs. 2, 4E, Supplemental Fig. 2F, data not shown)], and the detection of certain intracellular Ags using methanol permeabilization (pSTAT5, Glut1, Glut3) as described (22) or the eBioscience Foxp3/transcription factor (TF) buffer set (PGC-1α) (Figs. 3C, 3D, 4E, Supplemental Figs. 1F, 2H). Lipid content and lipid/glucose uptake were determined by incubation with Bodipy 493/503 (0.5 μg/ml PBS, 10 min at room temperature) or 37°C culture in complete RPMI 1640 in the presence of Bodipy FL C16 (overnight at 0.5 μg/ml), Bodipy low-density lipoprotein (LDL) (30 min at 10 μg/ml), or 2-NBDG (2 h at 100 μg/ml) prior to cell surface stains and acquisition (Fig. 4E, 4G). Intravascular staining of CD8⁺ T cells was adapted from the methodology developed by Anderson et al. (23) (i.v. injection of 4 μg anti–CD8β-PE [53-5.8] followed by euthanasia 4–5 min later, tissue harvesting/processing, and staining with anti–CD8α-BV421 or –PerCP-Cy5.5 [53-6.7], other cell surface receptor/ligand Abs, and MHC class I tetramers; Fig. 5B–E, Supplemental Fig. 3A–E). Samples were acquired on FACSCalibur, Accuri C6, FACSCanto, LSR II, or LSRFortessa X-20 flow cytometers (BD Biosciences) and analyzed with FACSDIVA (BD Biosciences) and/or FlowJo (Tree Star) software; dimensionality reduction and data display for polychromatic flow cytometry were performed using the Cytobank platform and the t-SNE algorithm viSNE (24) (input parameters: forward and side scatter properties and CD8α, CD8β, CD27, CD43 [S7], CD62L, CD127, KLRG1, CXCR3, CX3CR1 mean expression levels of young or old D^bNP₃₉₆⁺, and D^bGP₃₃⁺ CD8⁺ T_M populations).

Microarray analyses were conducted with highly purified p14 T_E and aging p14 T_M populations and are described in detail in Ref. 1; all data can be retrieved from the Gene Expression Omnibus repository (accession number GSE38462; https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE38462), and selected datasets are shown in this article in Figs. 1C, 1D and 4A, 4D, Supplemental Figs. 1A–E and 2A, 2G. Gene set enrichment analyses (GSEAs) were performed based on filtered data sets obtained for aging p14 T_M (day [d] 46, d156, d286, and d400) (1) against 186 Kyoto Encyclopedia of Genes and Genomes (KEGG) gene sets/pathways (http://software.broadinstitute.org/gsea) (Figs. 1A, 2E, 4C, Supplemental Figs. 1G, 2B). We treated time series as continuous phenotypes and used Pearson correlation to determine ranks for each gene. Enrichment scores (ES) were obtained as the maximum deviation from zero of P_hit − P_miss, where P_hit and P_miss are fractions of genes in or not in specific gene sets weighted by their correlations up to a given position in the rank; p values were estimated from random permutation tests by comparing random ES versus observed ES (25). For quantitative RT-PCR (Supplemental Fig. 1H), RNA isolation and DNAseI digestion were performed with spleen, LN, and BM cells stored in RNAlater using RNAqueous-4PCR kit per manufacturer protocol (Ambion/Life Technologies). RNA integrity was evaluated on a RNA Nano chip run on a Bioanalyzer 2100 (Agilent Technologies); RNA integrity numbers for all samples were 8.2–9.6. The cDNA first-strand transcription was performed using 370 ng of total RNA with the iScript cDNA synthesis kit following manufacturer protocol (BioRad). Il7 and Il15 primers were obtained from PrimerBank (http://pga.mgh.harvard.edu/primerbank/). Primer sequences for the SYBR Green quantitative PCR were as follows: Il7 (PrimerBank identifier 6680433a1) forward (5′-TTCCTCCACTGATCCTTGTTCT-3′) and reverse (5′-AGCAGCTTCCTTTGTATCATCAC-3′), Il15 (PrimerBank identifier 6680407a1) forward (5′-ACATCCATCTCGTGCTACTTGT-3′) and reverse (5′-GCCTCTGTTTTAGGGAGACCT-3′), Gapdh forward (5′-AATGAAGGGGTCATTGATGG-3′) and reverse (5′-AAGGTGAAGGTCGGAGTCAA-3′). Quantitative PCR was performed on a Roche LightCycler 480 II Real Time PCR instrument, using PerfeCta SYBR Green (Quanta Biosciences). PCR was carried out in a 20-μl volume and a final concentration of 1× reaction buffer, 385 nM forward and reverse primers, and 1.0 μl of cDNA reaction. Four log10 dilutions of pooled sample cDNA template were prepared and used for primer validation and standard curve reference. All sample reactions were performed in triplicate with nontemplate control reactions for all primer sets on a single plate. PCR cycling parameters were as follows: hot-start at 95°C for 2 min 30 s, 45 cycles of 95°C for 15 s, 60°C for 35 s, followed by a dissociation curve measurement from 65°C to 95°C. Relative comparison analysis with efficiency correction was performed using the LC 480 II data collection software release 1.5.0.39 SP4. Melt curve analysis for all assays verified single product amplification and absence of primer dimers. Nontemplate control reactions for all primer sets were >5 quantification cycle from all control and unknown samples.

Single-cell suspensions prepared from spleen or lympholyte-purified (Cedarlane) PBMCs were cultured for 12–48 h in RPMI 1640 supplemented with 10% FCS but in the absence of added growth/survival factors or T cell stimuli; in some cases, titrated amounts of pharmacological inhibitors ABT-737 (Abbott Laboratories), C75 (Cayman Chemical), Atglistatin (Cayman Chemical), chloroquine (Sigma-Aldrich), or vehicle were added to cultures (Figs. 2H, 4H, 4I, Supplemental Fig. 2C–E). CD8⁺ T cell survival was subsequently determined by combined CD8α, congenic marker, MHC class I tetramer, or CD44 and viability stains (Annexin V/propidium iodide or 7AAD, or Zombie dyes). Absolute numbers of viable CD8⁺ T cell subsets were calculated using Countess (Invitrogen) or Vi-Cell (Beckman Coulter) automated cell counters.

For competitive homing assays, splenic p14 T_M were enriched from young and old LCMV-immune p14 chimeras, differentially labeled with CFSE, and mixed at a ratio of 1:1, and, depending on experiments, populations containing 1.1–4.2 × 10⁵ p14 T_M each were injected i.v. into B6 recipients; 42–48 h later, transferred p14 T_M were retrieved and enumerated in LNs and other tissues (Fig. 6E, Supplemental Fig. 3G). Homing assays using endogenously generated CD8⁺ T_M populations were conducted in an analogous fashion using 5–6 × 10⁴ D^bNP₃₉₆⁺ CD8⁺ T_M each, B6.CD45.1 recipients, and retrieval of donor cells from various tissues 20 h later (Fig. 7E). In some cases, mixed donor populations were incubated for 1 h in complete RPMI 1640 (1.5 × 10⁷ cells/ml) in the presence or absence of 25 ng/ml pertussis toxin (RnD Systems) prior to washes and transfer (Supplemental Fig. 3G). For trafficking studies under conditions of CD62L blockade (Fig. 7A, Supplemental Fig. 3H), B6 mice were treated with a single i.p. injection of 200 μg anti-CD62L (MEL-14) or rIgGa control (RTK2758) 2 h before transfer of ∼5 × 10⁵ aged p14 T_M and retrieval 48 h later. Further details about all Abs are provided in Supplemental Table I.

Data handling, analysis, and graphic representation was performed using Prism 6.0c (GraphPad Software). All data summarized in bar and line diagrams are expressed as mean ± 1 SE (SEM), and asterisks indicate statistical differences calculated by Student t test (unpaired or paired) or one-way ANOVA with Dunnett multiple comparisons test and adopt the following convention: *p < 0.05, ** p < 0.01, and ***p < 0.001.

Challenge of mice with the natural murine pathogen LCMV induces a potent antiviral CD8⁺ T_E response that rapidly controls the infection and permits the subsequent development of specific CD8⁺ T_M that are maintained for life in the absence of residual viral Ags (17, 26). In B6 mice, the principal LCMV-specific CD8⁺ T cell populations target the nucleo and glycoprotein determinants NP_396–404 and GP_33–41; in addition, naive TCR transgenic p14 cells specific for LCMV-GP_33–41 and transferred into congenic B6 mice can be used to construct p14 chimeras for facilitated interrogation of a clonotypic CD8⁺ T_M population (p14 T_M). A combination of p14 chimera and B6 systems provided the experimental foundation for our comprehensive delineation of aging antiviral CD8⁺ T_M properties (1), and drawing on these resources, we have now revisited the foundations of long-term CD8⁺ T_M survival (27) by conducting modified GSEAs that specifically leverage the temporal aspect of our p14 T_M data sets (see 2Materials and Methods). Here, of 132 gene sets comprising 10,945 genes and exhibiting age-associated modulations, ∼20% (26/132) were enriched and ∼80% (106/132) were depleted in old p14 T_M, the latter group including the KEGG apoptosis pathway (Fig. 1A). Within this module, 14 genes belonged to the Bcl-2, baculoviral IAP repeat-containing (BIRC), or caspase gene families, and their combined temporal regulation pointed toward reduced apoptosis susceptibility of aged p14 T_M (Fig. 1A).

Members of the Bcl-2 family have long been implicated in lymphocyte survival and death, and the balanced expression of antiapoptotic Bcl-2 and proapoptotic BIM controls survival of T_N and, to a somewhat lesser extent, memory phenotype CD8⁺ T cells (T_MP) (28). Our interrogation of individual Bcl-2 family members revealed predominantly stable expression by aging p14 T_M with two notable exceptions, the modestly rising levels of Bcl2 and Bcl2l11 (Supplemental Fig. 1A–C). Importantly, the transcriptional changes were accompanied by a substantial increase of Bcl-2 protein content in aging CD8⁺ T_M and a slight, although significant, enhancement of BIM such that the resulting Bcl-2/BIM expression ratio steadily increased over time (Fig. 1B, 1C). Given a progressive enrichment for the CD62L^hi phenotype among aging antiviral CD8⁺ T_M (1, 16), our findings are also in agreement with the reported elevation of both Bcl-2 and BIM in antiviral CD8⁺ T_CM as compared with effector T_M (T_EM) populations (29). We emphasize, however, that these expression differences themselves are subject to an extended temporal modulation because the continuous rise of the Bcl-2/BIM ratios occurred in aging CD8⁺ T_CM and T_EM subsets alike (Fig. 1C). We also note the persistence of relatively stable Bcl-x_L levels (Fig. 1D), a gradual increase of several BIRC family genes that may contribute to an enhanced survival advantage for aged CD8⁺ T_M (30, 31) (Supplemental Fig. 1D), and the pronounced decline of Casp3 mRNA without evidence for caspase-3 activation (19) throughout long-term T cell memory (Fig. 1D, Supplemental Fig. 1E). Altogether, the kinetics of gene and protein expression therefore indicate that aging CD8⁺ T_M may be endowed with increasing overall fitness.

When assessed directly ex vivo, the viability of CD8⁺ T_M was not affected by age (Fig. 2A), but an in vitro culture in the absence of added growth/survival factors (“withdrawal apoptosis”) documented a gradual decline of CD8⁺ T_M death as a function of age (Fig. 2B). Increased apoptosis resistance has been associated with aging and cellular senescence (32), but the CD8⁺ T_M under investigation here lacked phenotypic and functional features of incapacitation (1), including the hallmark of murine T cell senescence, increased P-glycoprotein activity (33). Nonetheless, a distinct survival advantage of “nonsenescent” old CD8⁺ T_MP was previously observed under conditions of withdrawal apoptosis and attributed, despite an exacerbated decline of mitochondrial membrane potentials (ΔΨm), to reduced production of reactive oxygen species (ROS), elevated intracellular thiol levels (largely representing the abundance of reduced glutathione [GSH]), and increased expression of phase II antioxidant enzymes that combine to protect aged CD8⁺ T_MP against oxidative stress, mitochondrial dysfunction, and death (34, 35). In our model system, aging virus-specific CD8⁺ T_M also exhibited a modest decline of ex vivo ROS production (Fig. 2C) and a more striking loss of ΔΨm after in vitro culture (Fig. 2D). Yet, despite an enrichment of genes within the GSH metabolism pathway (Fig. 2E) that may collectively provide a metabolic advantage (36) for recall responses, we observed only a marginal rise of intracellular thiol levels in aging CD8⁺ T_M (Fig. 2F), and regardless of a 1.9-fold increase of Nfe2l2 mRNA (1) (the major TF in control of phase II enzyme regulation), no evidence for elevated induction of the respective genes could be obtained (data not shown). Instead, we found a pronounced augmentation of cell surface thiol levels by aging CD8⁺ T_M that was likely the result of changing microenvironments in older mice as demonstrated by their significantly increased serum thiol levels (Fig. 2G); this conclusion is also consistent with the notion that the immediate microenvironment rather than intracellular GSH levels preferentially determines the redox status of cell surface molecules (37).

Collectively, the above observations suggest that T cell–intrinsic mechanisms, in particular the rising Bcl-2/BIM expression ratio, may confer a survival advantage to aging CD8⁺ T_M. To directly evaluate this possibility we employed a coculture system to monitor survival of congenic old and young CD8⁺ T_M in the same in vitro environment. Addition of the Bcl-2 inhibitor ABT-737 (38) to cultures precipitated CD8⁺ T cell death in a dose-dependent fashion and, at a saturating concentration of 150 nM, reduced total CD8⁺ T cell survival to ∼10% (Fig. 2H). The relative survival advantage of old versus young D^bNP₃₉₆⁺ CD8⁺ T_M, however, was maintained at lower ABT-737 dosages and only disappeared at ∼100 nM, providing direct evidence for the exquisite dependence of CD8⁺ T_M survival on Bcl-2 and its role in promoting enhanced apoptosis resistance of aged CD8⁺ T_M populations (Fig. 2H). Although ABT-737 binds to Bcl-x_L and Bcl-w in addition to Bcl-2 (38), the low-level expression of corresponding mRNA species and, in the case of Bcl-x_L, also protein (Fig. 1D, Supplemental Fig. 1A), supported the notion of Bcl-2 as the major ABT-737 target in CD8⁺ T_M. Further evidence for the elevated Bcl-2/BIM ratio as a determinant for enhanced survival of aged CD8⁺ T_M came from a reversal of survival advantages at saturating ABT-737 concentrations (150 nM, Fig. 2H); although very few cells remained alive under conditions of complete Bcl-2 blockade, the slightly better survival of residual young CD8⁺ T_M could potentially be explained by their comparatively lower BIM expression (Fig. 1C) because death of Bcl-2–deficient T cells was shown earlier to decline as a function of BIM gene dosage (Bcl2l11^+/+ > Bcl2l11^+/− > Bcl2l11^−/−) (28).

In the context of an acute response, both I° and II° CD8⁺ T_E downregulate Bcl-2 expression (39), and control of CD8⁺ T_E subset survival is thought to switch to other factors, perhaps including the BIRC family member survivin/Birc5 (40) (Supplemental Fig. 1D). Work with a Bcl-2 reporter mouse, however, indicates that even at the peak of a pathogen-specific immune response, CD8⁺ T_E populations are characterized by a spread of Bcl-2 expression levels that permits the distinction of CD8⁺ T_E subsets with differential memory potential (41). In line with these observations, we found that II° CD8⁺ T_E derived from aged CD8⁺ T_M exhibited a slight yet significant elevation of Bcl-2 as compared with I° CD8⁺ T_E or II° CD8⁺ T_E generated from young CD8⁺ T_M (Fig. 2I). Coupled with the former cells’ improved survival during the ensuing contraction phase (1), our results therefore hinted at a direct role for Bcl-2 in promoting a more effective establishment of II° CD8⁺ T cell memory. Indeed, whereas young II° CD8⁺ T_M featured reduced Bcl-2 contents compared with I° CD8⁺ T_M as reported previously (39), old II° CD8⁺ T_M present within the same hosts exhibited substantially higher Bcl-2 levels (Fig. 2I). Thus, the largely Bcl-2–dependent survival advantage of old over young I° CD8⁺ T_M was re-established in the course of II° memory formation.

Overall, the dynamic regulation of Bcl-2 re-expression in the memory phase (Fig. 1B, 1C) followed a pattern similar to that of multiple other phenotypic/functional CD8⁺ T_M properties subject to age-associated expression modulation (1). Because the precise pace of these changes could be experimentally accelerated or delayed as a function of initial CD8⁺ T_N precursor frequency or infection dosage (1), we surmised that Bcl-2 expression by CD8⁺ T_M could be controlled in a comparable fashion. In this study, we constructed p14 chimeras with titrated numbers of p14 T_N (2 × 10²–2 × 10⁵) and challenged the mice with a standard dose of LCMV (2 × 10⁵ PFU) or generated p14 chimeras with a fixed p14 T_N number (1 × 10⁴) and infection with graded dosages of LCMV (2 × 10³–2 × 10⁷ PFU). Measuring Bcl-2 expression by p14 T_M 6–7 wk later, we found that an increase of p14 T_N input numbers enhanced, whereas an escalation of the virus challenge dose reduced respective Bcl-2 levels in p14 T_M (Fig. 2J). In summary, our results demonstrate that aging CD8⁺ T_M become more resistant to apoptosis, that their improved survival and that of their II° progeny is principally controlled through increased Bcl-2 expression, and that the specific conditions of CD8⁺ T_E generation determine the pace of progressive Bcl-2 upregulation by CD8⁺ T_M.

In direct relation to their longevity, regulation of CD8⁺ T_M fates under steady-state conditions also involves homeostatic proliferation, the slow and stochastic division of “resting” CD8⁺ T_M governed by the cytokines IL-7 and IL-15 (27, 42). In extension of our previous report (1), we now demonstrate that a progressive upregulation of the respective cytokine receptors (CD127 and CD122) by aging CD8⁺ T_M also pertains to the PBMC compartment and to differential CD8⁺ T_M specificities (Fig. 3A, 3B), suggesting that their homeostatic proliferation rates may be adjusted accordingly. To determine if enhanced cytokine receptor expression indeed conveyed greater responsiveness, we assessed the extent of IL-7/IL-15–induced STAT5 phosphorylation in young and old p14 chimeras. Here, aged p14 T_M not only exhibited greater reactivity, but at limiting concentrations IL-7 clearly proved to be a more effective activator of STAT5 than IL-15 (Fig. 3C). These findings extend the notion of superior IL-7 potency in the context of initial CD8⁺ T_M formation (43) to the long-term maintenance of CD8⁺ T_M and complement a recent observation about enhanced IL-15 reactivity of “late” p14 T_M or T_CM (>8 mo postinfection) as compared with “early” p14 T_M/T_CM (d30–d45) (5). We further note that the thymic stromal lymphopoietin receptor (TSLPR) is apparently the only CD8⁺ T_M-expressed cytokine receptor subject to a gradual downmodulation over time (1), a pattern that could contribute to the amplified IL-7 reactivity of aged CD8⁺ T_M as it may permit enhanced complex formation of CD127 with CD132 rather than TSLPR (44).

Although the above findings correlate increased CD127/CD122 expression with CD8⁺ T_M reactivity to IL-7/IL-15, we also noted a certain extent of constitutive STAT5 phosphorylation among p14 T_M analyzed directly ex vivo, similar to the basal STAT5 phosphorylation observed in human CD8⁺ T cell subsets of undefined specificity (45). Additional control experiments confirmed this conclusion (Supplemental Fig. 1F), but unexpectedly, the levels of constitutive STAT5 phosphorylation remained unaltered in aging antiviral CD8⁺ T_M populations (Fig. 3D). Because the level of active STAT5 appears to control homeostatic proliferation rates (46), stable pSTAT5 expression by endogenously generated CD8⁺ T_M therefore suggested that their homeostatic proliferation rates, despite enhanced sensitivity to IL-7/IL-15, might not be accelerated. This prediction was reinforced by our longitudinal p14 T_M GSEAs that demonstrated a negative (although not significant) enrichment of cell cycle–associated genes and thus also argued against an accelerated CD8⁺ T_M turnover (Supplemental Fig. 1G). Indeed, as assessed by ex vivo Ki67 expression, homeostatic proliferation of blood-borne LCMV-specific CD8⁺ T_M was unaffected by age (Fig. 3E), a contention corroborated through the comparable in vivo BrdU incorporation by young and old CD8⁺ T_M in various lymphatic tissues and NLTs (Fig. 3F). Thus, in contrast to murine CD8⁺ T_MP of undefined specificity (47), homeostatic proliferation rates of virus-specific CD8⁺ T_M were largely independent of age but remained susceptible to modulation by tissue-specific microenvironments as shown for young CD8⁺ T_M (18).

Finally, it is important to note that homeostatic proliferation rates are not simply an intrinsic property of phenotypically defined CD8⁺ T_M subsets. For example, the CD62L^hi CD8⁺ T_CM population, previously reported to exhibit higher homeostatic proliferation rates than CD8⁺ T_EM (16, 48), accumulates in the spleen over time (1, 16) without causing an overall acceleration of homeostatic turnover (Fig. 3E, 3F). And although we confirmed the differential homeostatic proliferation rates of splenic CD8⁺ T_CM versus T_EM in young LCMV-immune mice, we found no differences in other tissues, such as LNs (Fig. 3G). The absence of a simple correlation between CD8⁺ T_M subsets, rates of homeostatic proliferation, cytokine receptor (CD127/CD122), and even corresponding tissue-specific cytokine (Il7/Il15) expression levels (Supplemental Fig. 1H) constitutes an important caveat that needs to inform further investigations into the homeostasis of CD8⁺ T_M populations.

Initial CD8⁺ T_E differentiation and CD8⁺ T_M generation are both controlled and accompanied by varied metabolic adaptations. Activation of “quiescent” naive CD8⁺ T_N engages a “metabolic switch” that endows emerging CD8⁺ T_E with high rates of aerobic glycolysis and glutaminolysis to support an anabolic metabolism; the subsequent development of CD8⁺ T cell memory is characterized by a gradual return to metabolic quiescence and a preferential reliance on fatty acid (FA) oxidation (FAO) and oxidative phosphorylation (OxPhos) to meet the changing energy demands (49). The extent to which established CD8⁺ T_M populations may further adapt their metabolism over time, however, remains little explored (5). We previously reported that aging CD8⁺ T_M exhibit a subtle yet significant increase of cellular size and “granularity/complexity” (determined by forward and side scatter properties, respectively) (1), a process most likely controlled by mTOR activity (50). Indeed, we now find that basal mTOR protein (although not mRNA) expression by antiviral CD8⁺ T_M increased with age, as did message for ribosomal protein S6 (Rps6, a downstream target of the mTORC1 complex involved in the regulation of cell size, proliferation, and glucose homeostasis) and, importantly, the degree of Rps6 protein phosphorylation (Fig. 4A, 4B). Although the convergence of elevated mTORC1 activity, cell size, and recall capacity of aged CD8⁺ T_M (1) is consistent, these adjustments would appear to run counter to the shift toward reduced glycolysis and increased OxPhos, as observed for the earlier transition from CD8⁺ T_E to young T_M stage (49). Interestingly, however, most recent work indicates that enforcement of sustained glycolysis and suppression of OxPhos does not compromise but rather may accelerate CD8⁺ T_M formation (51). Therefore, to assess the extended evolution of metabolic CD8⁺ T_M profiles, we reviewed our temporal GSEAs and found that ∼25% of all pathways up or downregulated by p14 T_M over time could in fact be assigned to the broad KEGG category of metabolism. Here, a collective depletion of carbohydrate, energy, lipid, amino acid, and glycan pathways in aging p14 T_M suggested a continued trend toward metabolic quiescence, yet the gene sets comprising glycolysis, nucleotide, and GSH metabolism were simultaneously enriched (Fig. 4C, data not shown). In the absence of significant differences for the majority of these temporally regulated pathways (Fig. 4C), the age-associated alterations of CD8⁺ T_M metabolism are therefore expected to be subtle but nevertheless might be reflected in a distinct modulation of glucose and FA utilization.

With regard to glucose metabolism, our transcriptional p14 T_M data indicated that within the family of facilitative glucose transporters, robust gene expression was restricted to stable Slc2a1/Glut1 and progressively declining Slc2a3/Glut3 mRNA species (Fig. 4D). Yet whereas corresponding Glut3 protein expression levels mirrored the decline of Slc2a3 mRNA, total Glut1 expression was subject to distinct translational modulations: high in CD8⁺ T_E, reduced in young CD8⁺ T_M, but intermediate in aged CD8⁺ T_M (Fig. 4E). Importantly, a specific interrogation of surface Glut1 confirmed the enhanced expression by old versus young CD8⁺ T_M, and the differential Glut1 levels in CD8⁺ T_E/M populations correlated precisely with their respective glucose uptake capacities (Fig. 4E). In contrast, greater rates of glucose uptake by CD8⁺ T_N than either CD8⁺ T_E or young T_M, also observed in other reports (52, 53), did not correspond to enhanced Glut1 levels in our experiments (Fig. 4E); however, neither glucose uptake nor in vitro survival of resting T cells is affected by Glut1-deficiency and may instead rely on related transporters such as Glut3 (54). The notion of enhanced glucose utilization by aged as compared with young CD8⁺ T_M is further supported by the pattern of CD8⁺ T cell–expressed insulin receptor (Insr/CD220) that significantly increases with CD8⁺ T_M age (Fig. 4F, Supplemental Fig. 2A). Insulin not only regulates glucose uptake but also acts as a major growth factor that increases protein translation (55). In fact, of the 26 gene sets demonstrating a progressive enrichment in aging p14 T_M, nearly half are captured under the general category of “genetic information processing” that includes pathways for transcription; translation; folding, sorting and degradation; and replication and repair (Supplemental Fig. 2B).

If CD8⁺ T_M aging fosters a trend toward increased glucose utilization, a concurrent decrease of OxPhos and FA utilization would appear likely, and our GSEAs indicate that this may indeed be the case (Fig. 4C). To directly determine the amount of stored fat in CD8⁺ T cells (53), we quantified neutral lipid content in CD8⁺ T_N and virus-specific CD8⁺ T_E/M populations and found it to be regulated in highly dynamic fashion, with a pronounced growth of neutral lipid abundance in the course of CD8⁺ T_E differentiation followed by a progressive reduction during CD8⁺ T_M formation and aging (Fig. 4G). Similarly, the capacity for exogenous long-chain FA (FL C₁₆) and LDL uptake was markedly increased in CD8⁺ T_E, quickly declined in young CD8⁺ T_M, and was further diminished in aged CD8⁺ T_M (Fig. 4G). Reduced FA uptake, however, is not per se an indicator for decreased FA metabolism because CD8⁺ T_M fuel their bioenergetics needs in a “futile cycle” that utilizes extracellular glucose to support both increased FA synthesis (FAS) and FAO (53). To delineate the relative contribution of FAS and FAO to CD8⁺ T_M metabolism (56) specifically in the context of aging, we incubated the various CD8⁺ T cell populations in the presence of titrated amounts of selected pharmacological inhibitors and assessed their respective survival. Overall, both young and old CD8⁺ T_M proved more resistant to inhibition of lipogenesis or lipolysis than either CD8⁺ T_E or T_N (Fig. 4H, 4I, Supplemental Fig. 2C–E). Yet subtle differences between young and aged CD8⁺ T_M could be discerned at particular inhibitor concentrations. Here, inhibition of FA synthase by 30 μM of the compound C75 compromised survival of young versus old CD8⁺ T_M to a greater extent, suggesting that aged CD8⁺ T_M are somewhat less reliant on FAS (Fig. 4H, Supplemental Fig. 2C). Considering the lipolytic machinery of CD8⁺ T cells, the recent work by O’Sullivan et al. (53) ruled out a role for adipose triglyceride lipase in CD8⁺ T_M formation and survival; likewise, we found that both young and aged CD8⁺ T_M, in contrast to CD8⁺ T_N and T_E, were completely resistant to adipose triglyceride lipase inhibition (Supplemental Fig. 2D). Rather, hydrolysis of neutral lipids appears to preferentially rely on lysosomal acid lipase (53), and in our experiments, young and old CD8⁺ T_M demonstrated an overall similar sensitivity to chloroquine-mediated inhibition of lysosomal acidification; at very high chloroquine concentrations, however, aged CD8⁺ T_M exhibited comparatively enhanced rates of death, suggesting they may be slightly more reliant on the mobilization of FA for FAO (Fig. 4I, Supplemental Fig. 2E).

Finally, we wanted to determine how the subtle metabolic alterations in aging CD8⁺ T_M populations relate to their overall “metabolic fitness.” Here, our determination of mitochondrial mass and membrane potential failed to document consistent differences, but in aggregate, we observed a trend toward enhanced fitness by old CD8⁺ T_M (Supplemental Fig. 2F, data not shown). In support of this assessment, we also note that PGC-1α, a master regulator of mitochondrial biogenesis most recently shown to improve the bioenergetics of LCMV-specific CD8⁺ T cells in a chronic infection model (52), is comparatively elevated at both mRNA and protein levels in aged CD8⁺ T_M (Supplemental Fig. 2G, 2H). In summary, we conclude that the “mixed metabolic phenotype” of long-term CD8⁺ T_M populations emerges through a partial reversal of metabolic adaptations that control and accompany the original transition from CD8⁺ T_E to young CD8⁺ T_M stage and that the “intermediate” metabolic profile of old CD8⁺ T_M likely contributes to their greater recall capacity (1, 5). Defining a precise inflection point for this metabolic switch during CD8⁺ T_M aging will be difficult given the delicate and only partial nature of metabolic adaptations, but it is well possible that a net effect of these processes may become discernible only at later stages of the extended CD8⁺ T_M evolution (5).

The extended maturation of circulating aging CD8⁺ T_M populations (1) proceeds in the face of their continued anatomic redistribution but without apparent alteration of total CD8⁺ T_M maintained in various lymphoid organs and NLTs (15, 17, 57, 58). Although there are some exceptions to this rule (59), it has remained unclear how exactly the phenotypic conversion of aging CD8⁺ T_M may modulate their trafficking patterns (60). The gradual re-expression of CD62L in particular (1, 16) would be expected to affect the anatomical distribution of older CD8⁺ T_M. For example, young p14 T_EM and T_CM subsets, distinguished according to CD62L expression and with differential sensitivity to the chemokines CCL19 and CXCL12, preferentially localize to splenic red pulp (RP) and white pulp (WP), respectively (61). The progressive upregulation of CD62L, CCR7 (CCL19 receptor), and CXCR4 (CXCL12 receptor) by aging splenic CD8⁺ T_M (1), confirmed and extended here to blood-borne CD8⁺ T_M with different LCMV specificities (Fig. 5A, data not shown), may therefore also promote an altered positioning of these cells within the spleen. To evaluate this possibility, we employed the i.v. injection of fluorochrome-conjugated CD8β Ab that readily labels CD8⁺ T cells found in vascular contiguous compartments (including RP) but not in tissue stroma and parenchyma (including WP) (23, 62) (Supplemental Fig. 3A). Although the total number of specific CD8⁺ T_M in the spleen does not change over time (17), their differentiation according to RP/WP residence demonstrated a pronounced increase from ∼15 to ∼60% in the WP of aging mice (Fig. 5B). A concurrent phenotypic stratification of RP/WP subsets according to markers that are substantially up or downregulated by aging CD8⁺ T_M (1) further revealed striking differences in young mice: the ∼15% of young D^bNP₃₉₆⁺ CD8⁺ T_M residing in the WP, despite preserving some phenotypic heterogeneity, for the most part already adopted properties comparable to aged CD8⁺ T_M (CD27^hi, CD62L^hi, CD127^hi, CXCR3⁺, CD43^lo, KLRG1⁻, CX3CR1^lo), whereas RP cells (representing ∼85% of splenic D^bNP₃₉₆⁺ CD8⁺ T_M) exhibited a contrasting and largely “immature” phenotype (Fig. 5C–E); these differences also pertained to more subtle aspects of CD8⁺ T_M aging such as SSC properties and CD8α expression levels (albeit not cellular size) (Fig. 5E). In aged LCMV-immune mice, and in agreement with the observation that phenotypic maturation affects both splenic and blood-borne CD8⁺ T_M [(1), Figs. 3A, 3B, 5A], the dissimilarity of WP and RP D^bNP₃₉₆⁺ CD8⁺ T_M mostly disappeared and both populations presented with an aged phenotype (although the RP subset retained somewhat elevated CD43, KLRG1, and CX3CR1 expression) (Fig. 5C–E). Nearly identical results were also obtained for young and old D^bGP₃₃⁺ CD8⁺ T_M in splenic RP/WP compartments (Supplemental Fig. 3B–D). Last, a direct comparison of young and old CD8⁺ T_M in the RP confirmed their marked phenotypic differences, but the WP subsets, to a lesser degree, also demonstrated evidence for further age-associated phenotype maturation (Supplemental Fig. 3E). Altogether, these observations reveal the gradual emergence of coregulated complex CD8⁺ T_M phenotypes as well as their distinct spatiotemporal segregation that accompanies the more global architectural changes recently reported for the aging murine spleen (63).

Another potential consequence of increasing CD62L, CCR7, and/or CXCR4 expression by aging CD8⁺ T_M (Fig. 5A) is the gradual acquisition of an enhanced LN tropism (64, 65), especially because earlier trafficking studies have demonstrated the unequivocal requirement for virus-specific CD8⁺ T_M-expressed CD62L (66) and chemokine receptors (58) to enter LNs under steady-state conditions. Indeed, a first suggestion in support of this conjecture has come from a recent study that reported a greater proportion of late p14 T_M as compared with early p14 T_M in inguinal LNs (5). To examine if the LN-homing phenotype of aged CD8⁺ T_M confers a preferential redistribution to peripheral LNs at large, we enumerated specific CD8⁺ T_M in young and old LCMV-immune mice. In the absence of age-associated changes in LN cellularity, we observed an up to 10-fold increase of specific CD8⁺ T_M frequencies and numbers in aged mice (Fig. 6A–C), and a longitudinal analysis of mesenteric LNs (MLN) revealed a slow and continuous accumulation of CD8⁺ T_M with an estimated population doubling time of ∼190 d (Fig. 6D). We next assessed the capacity of aging CD8⁺ T_M to enter peripheral LNs by performing a competitive homing experiment (Fig. 6E). In brief, p14 T_M were enriched from young and old LCMV-immune p14 chimeras, differentially labeled with CFSE, combined at a ratio of 1:1, and transferred into naive B6 recipients. Upon retrieval 48 h later, this ratio was skewed to >10:1 in favor of old p14 T_M in peripheral LNs but not blood or spleen, demonstrating that aging CD8⁺ T_M in fact acquire a capacity for facilitated LN access (Fig. 6E).

Similar to polyclonal CD8⁺ T_M [(1), Fig. 5A), old p14 T_M exhibited higher expression levels of CCR7, CXCR4, and in particular, CD62L (Fig. 7A, Supplemental Fig. 3F). To determine if CD62L contributed directly to the facilitated LN access of aged CD8⁺ T_M, we conducted an in vivo homing assay with old p14 T_M under conditions of CD62L blockade and observed an 82–93% reduction of p14 T_M accumulation in peripheral LNs (Fig. 7A). Similar experiments designed to evaluate the role of chemokine receptors by pretreatment of young and old donor p14 T_M with pertussis toxin revealed, as expected (58), a profound inhibition of p14 T_M trafficking to LNs (Supplemental Fig. 3G). The relative reduction, however, appeared especially pronounced for young p14 T_M, indicating a slight advantage for aged p14 T_M to use pertussis toxin–insensitive pathways for residual LN access (Supplemental Fig. 3G). The importance of CD62L in conveying an enhanced LN tropism to CD8⁺ T_M populations was further illustrated by use of the “virus titration chimeras” discussed above. Following infection of p14 chimeras with escalating titers of LCMV and generation of T cell memory 7 wk later, p14 T_M expression of CD62L but not CCR7 or CXCR4 significantly declined as a function of increasing viral challenge dosage (Fig. 7B, data not shown), and reduced CD62L expression correlated with an impaired accumulation of p14 T_M in peripheral LNs (Fig. 7B). Thus, the LN tropism of CD8⁺ T_M, in addition to their survival/Bcl-2 expression (Fig. 2J), multiple phenotypic and functional properties, and II° reactivity (1), can be experimentally controlled in a fashion that accelerates or delays the CD8⁺ T_M maturation process at large.

Based on the above evidence, and in the absence of locally increased homeostatic proliferation (Fig. 3E, 3F), the progressive accumulation of aging CD8⁺ T_M in II° lymphoid tissues (Figs. 5B, 6, Supplemental Fig. 3B) most likely emerged through the redistribution of CD8⁺ T_M from other anatomic reservoirs. We estimated, according to the numbers of specific CD8⁺ T_M in the LNs of young and aged LCMV-immune mice (Fig. 6A–C), as well as the number and variable size of murine LNs (67), that over a period of ∼17 mo, up to 1 × 10⁶ NP₃₉₆- and 1.9 × 10⁶ GP₃₃-specific CD8⁺ T_M were added to the entire LN pool. Given the stable CD8⁺ T_M numbers in the spleen (17), the potential sources for the new LN CD8⁺ T_M are therefore the blood and marginated pool (BMP), as well as NLTs. In a recent and comprehensive accounting of organism-wide CD8⁺ T_M distribution, based on an evaluation of LCMV-immune p14 chimeras, Steinert et al. (62) demonstrated that NLTs and BMP (excluding splenic RP) together contain ∼6 × 10⁶ p14 T_M. The p14 model used in that study and our B6 system are roughly comparable because flow cytometry-based calculations revealed the presence of ∼2.9 × 10⁶ splenic p14 T_M while we documented a total of ∼2.0 × 10⁶ endogenously generated NP₃₉₆/GP₃₃-specific CD8⁺ T_M in the spleen (Fig. 6B, 6C, data not shown). In regards to LN-residing CD8⁺ T_M, however, the models are expected to differ because of increased p14 T_N numbers used for chimera construction (62), correspondingly accelerated upregulation of CD62L by p14 T_M (1), and an experimental evaluation at a somewhat later time points (4–5 mo after challenge) (62) that together should result in apparently enhanced LN accumulation. Indeed, the reported grand total of ∼2.3 × 10⁶ p14T_M in peripheral LNs (62) clearly exceeded the ∼4.3 × 10⁵ NP₃₉₆/GP₃₃-specific CD8⁺ T_M we found in the LN compartment of B6 mice at ∼2 mo following LCMV infection (a 5.4-fold difference). With these caveats in mind, we calculated that in the time of ∼2–19 mo postinfection, a <2-fold loss from BMP and NLTs could account for the corresponding gain of NP₃₉₆/GP₃₃-specific CD8⁺ T_M in the LNs of old LCMV-immune B6 mice.

To test this prediction, we first evaluated the preservation of D^bNP₃₉₆ CD8⁺ T_M in the blood by combining data obtained in numerous experiments performed over a period of several years. Interestingly, the aggregate data uncovered an unexpected biphasic loss of blood-borne CD8⁺ T_M (Fig. 7C). In the period of ∼7–14 wk after virus challenge, and thus well after completion of the “contraction phase” in the spleen (17), specific CD8⁺ T_M numbers continued to decline in the blood before attaining seemingly stable levels around day 100 postinfection. A careful inspection of subsequent time points, however, revealed a subtle decrease of blood-borne CD8⁺ T_M with a theoretical population t_1/2 of ∼3 y (Fig. 7C). This finding is noteworthy because it evokes, even under experimental conditions that optimize CD8⁺ T_M preservation, the natural decline of blood-borne virus-specific CD8⁺ T_M in humans (12). Inasmuch as the cumulative ∼60% loss (between weeks 7 and 86) of specific CD8⁺ T_M from peripheral blood also reflects a changing CD8⁺ T_M abundance in the larger BMP, these cells could provide a relevant contribution to the growing CD8⁺ T_M LN pool. The biphasic erosion of blood-borne CD8⁺ T_M (Fig. 7C), however, would seem at odds with the dynamics of CD8⁺ T_M accumulation in the LNs (Fig. 6D). We therefore proceeded with an enumeration of young and old CD8⁺ T_M in NLTs (peritoneal cavity, liver, lung, kidney) and observed a 1.4- to 2.7-fold relative reduction of aged CD8⁺ T_M numbers (Fig. 7D). Thus, both theoretical considerations and experimental results support the notion that a loss of aging CD8⁺ T_M is not restricted to the lung (5, 59) but involves the BMP and especially NLTs in general.

How can the above conclusions be reconciled with the notion that NLTs are preferentially populated by nonrecirculating tissue-resident T_M (T_RM) (68)? According to Steinert et al. (62), ∼9% of CD8⁺ T_M found in NLTs can in fact recirculate, a fraction that is lower in some (e.g., lung) but higher in other (e.g., liver) compartments. These calculations are based on parabiosis experiments that were conducted, similar to multiple other studies, over a period of just ∼1 mo (62). A notably longer observation period was employed by Jiang et al. (69), who found that the frequencies of skin CD8⁺ T_RM in the donor parabiont declined by ∼2-fold between 8 and 24 wk after surgery, suggesting limits to CD8⁺ T_RM longevity and/or mobilization of the CD8⁺ T_RM compartment. The latter observation is not only in agreement with a classic study that reported a trend toward continued equilibration of CD8⁺ T_RM within intestinal lamina propria and epithelium for at least 8 wk (58) but also consistent with our experiments that compare CD8⁺ T_M populations recovered from NLTs at time points separated by ∼18 mo and thus may offer sufficient time for some CD8⁺ T_RM to re-enter the circulation. Of further importance is the recent observation that traditional flow cytometry–based methods of CD8⁺ T_M quantification in NLTs markedly underestimate the true number of CD8⁺ T_M found in these tissues (62). Although our quantification of young and old CD8⁺ T_M in liver, lung, and kidney therefore cannot accurately account for absolute CD8⁺ T_M numbers, it is the relative reduction of CD8⁺ T_M recovered from the NLTs of aged LCMV-immune animals, readily revealed even by use of flow cytometry, that is important for the present context. Consistent with this interpretation, we also observed an age-associated decrease of CD8⁺ T_M numbers in the peritoneal cavity (Fig. 7D), an organ that is not subject to the inefficiency of CD8⁺ T_M recovery from solid NLTs. Finally, in considering the role of CD8⁺ T_RM as highly effective first responders to infections re-encountered at body surfaces (70) and the established role of LN-residing CD8⁺ T_M as direct precursors for II° CD8⁺ T_E expansions, it is worth noting that LN CD8⁺ T_M themselves also act as gate-keepers and immediate effectors capable of curtailing peripheral infections and preventing systemic viral spread (71). In fact, following a footpad LCMV challenge of mice that received limiting numbers of young versus old CD8⁺ T_M, we found that only the latter population prevented systemic dissemination of the virus (data not shown). Therefore, the gradual accumulation of aging CD8⁺ T_M in peripheral LNs, even at the expense of CD8⁺ T_M in NLTs, may represent a progressively enhanced “strategic positioning” in anatomic locations that constitute a critical site for both local pathogen control and the coordination of effective recall expansions (72).

Considering the tissue redistribution of aged CD8⁺ T_M in their overall numerical context (Figs.5B, 6, 7C, 7D), it appears that the relative loss from NLTs and blood might even exceed the corresponding gain in peripheral LNs. A clue to another anatomic site for potential CD8⁺ T_M accrual comes from the increased CXCR4 expression by old antiviral CD8⁺ T_M [(1), Fig. 5A, Supplemental Fig. 3F]. CXCR4 is held to be a “BM homing receptor,” and consistent with this notion, recent work demonstrated that conditional CXCR4 deletion in LCMV-specific T cells resulted in a reduced abundance of CD8⁺ T_M populations especially in the BM (73). Thus, it is conceivable that greater CXCR4 expression levels by aged CD8⁺ T_M preferentially promote increased BM access, an important anatomic niche for CD8⁺ T_M (74). Indeed, the frequencies and numbers of D^bNP₃₉₆⁺ and D^bGP₃₃⁺ CD8⁺ T_M retrieved from the BM of aging LCMV-immune mice roughly doubled over a period of ∼1.5 y (Fig. 7D), although in contrast to LNs, accumulation of aging CD8⁺ T_M in the BM was independent of CD62L (Supplemental Fig. 3H). In competitive homing experiments similar to those shown in Fig. 6E but conducted here with endogenously generated CD8⁺ T_M, aged CD8⁺ T_M also displayed a slightly enhanced BM tropism; at the same time, their facilitated LN access was expectedly more pronounced (Fig. 7E).

Finally, the apparently generalized pattern of age-associated increasing CD8⁺ T_M abundance in both II^o (splenic WP, LN) and I° (BM) lymphatic tissues also warranted an analysis of the thymus. Interestingly, we observed an almost 2-fold relative increase of old over young CD8⁺ T_M populations for this I° lymphatic organ (counts normalized to 1 × 10⁶ cells); due to thymic involution, however, absolute numbers of aged CD8⁺ T_M were expectedly reduced, here by a factor of ∼2.5 (Fig. 7F).

As detailed in our recent work on CD8⁺ T cell memory (1), aging of established antiviral CD8⁺ T_M populations introduces a series of cumulative molecular, phenotypic, and functional changes that collectively confer naive-like T cell traits, greater proliferative potential, and protective capacities onto old CD8⁺ T_M populations. To account for these sweeping processes in a simple fashion, we have introduced the rebound model of CD8⁺ T_M maturation, according to which the extent of initial CD8⁺ T_E differentiation directly determines the kinetics of protracted CD8⁺ T_M dedifferentiation (1). We now demonstrate that this remodeling process also impinges on the homeostasis of CD8⁺ T_M as evidenced by their evolving survival capacity, metabolic adaptations, and microanatomic redistribution. Here, both the Bcl-2–dependent enhancement of apoptosis resistance and the accumulation of old CD8⁺ T_M in lymphoid tissues (including the CD62L-guided peripheral LN access/residence) as a likely consequence of a redistribution from NLTs and blood are consistent with the progressive modulation of aging CD8⁺ T_M phenotypes, in particular at the level of increasing CD62L, CD122, CD127, CCR7, and CXCR4 expression (1–3, 5, 16). We further document that the gradual acquisition of mature phenotypes by aging CD8⁺ T_M populations proceeds through coregulated modulation of receptor/ligand expression and at a pace that is contingent on the specific microenvironment (i.e., accelerated in splenic WP, delayed in RP). In addition, all of these dynamics are readily captured by the basic tenet of the rebound model that posits a broad harmonization of CD8⁺ T_M and T_N traits while simultaneously reinforcing the development of a simple CD8⁺ T_M core signature (1).

The imperviousness of aging CD8⁺ T_M to changes of their basal homeostatic proliferation rates, however, was unexpected. Our results document a simple association between cytokine receptor (CD122/CD127) expression levels and functionality, and the importance of CD127 abundance as well as the intermittent rather than continuous IL-7 signaling for the homeostasis of naive CD8⁺ T cell populations has been illustrated by the work of A. Singer’s group (75). Yet we previously noted a lack of association between CD127/CD122 expression levels on CD8⁺T_M and their tissue-specific pace of homeostatic turnover (18), and the heightened responsiveness of aged CD8⁺ T_M to IL-7 and IL-15 as shown here failed to confer increased homeostatic proliferation rates. Old CD8⁺ T_M may therefore have adopted an exquisite balance with age-associated changes in various tissue microenvironments; the homeostasis of CD8⁺ T_N and pathogen-specific CD8⁺ T_M, although reliant on the same cytokines (IL-7, IL-15), may be regulated in a differential manner, or other factors contributing to the regulation of CD8⁺ T_M homeostasis may become more dominant over time. We also note that changing levels of CD8⁺ T_M–expressed CXCR4, recently proposed to control the homeostatic turnover of CD8⁺ T_M (73), had no apparent impact on their homeostatic self-renewal over time. A recent analysis of human CD45RO⁺CD8⁺ T_MP populations also found no differences in homeostatic turnover rates between young and healthy elderly individuals (76).

Although subtle, the metabolic adaptations of aging CD8⁺ T_M would appear to contradict the rebound model because they are characterized by a partial reacquisition of CD8⁺ T_E–like profiles, in particular an increase of glucose utilization (49). Yet the shift toward enhanced glucose uptake and decreased neutral lipid content as well as reduced FA and LDL uptake also indicates a gradual return, albeit incomplete, toward respective CD8⁺ T_N capacities. Nevertheless, CD8⁺ T_N consistently displayed greater sensitivity to in vitro FAS and FAO inhibition than either young or old CD8⁺ T_M suggesting that the latter cells’ distinctive and evolving metabolic profiles should be considered part of the memory “core signature” that distinguishes CD8⁺ T_M from T_N.

Three aspects of CD8⁺ T_M homeostasis will require further clarification to define relevant age-associated adaptations and their potential impact on II° reactivity and immune protection in more detail. 1) The progressive conversion of aging CD8⁺ T_M documented primarily for spleen and blood (1–3, 5) will have to be considered for other anatomic compartments, including various NLTs (68), in the context of continued CD8⁺ T_M subset migration versus extended tissue residence (including the precise developmental relations and potential phenotypic/functional modulation of CD8⁺ T_M populations as they enter and exit various tissues) (18, 77, 78) and for human CD8⁺ T_M (79). 2) The transcriptional control of CD8⁺ T_M aging is another topic of broad interest. For example, among the major transcriptional regulators of CD8⁺ T_E/M differentiation predicted on the basis of coregulated gene expression in activated CD8⁺ T cells (80) are several TFs (Tcf4, Zeb2, Rora, Hif1a, Arntl) that also demonstrate progressive downmodulation in aging CD8⁺ T_M (1). In agreement with this observation, enhanced activity of hypoxia-inducible factors was recently shown to sustain a CD8⁺ T_E–like state (81) whereas Zeb2-deficiency accelerated CD8⁺ T_CM formation (82); the extent to which the evolution of complex TF expression profiles in aging CD8⁺ T_M supports a return to a CD8⁺ T_N–like transcriptional program while simultaneously reinforcing the emergence of a highly focused CD8⁺ T_M core signature is currently under investigation. 3) In conjunction with transcriptional regulation, epigenetic DNA and chromatin modifications provide irreducible contributions to the specification of CD8⁺ T_M fates (83). Although it remains unclear if established CD8⁺ T_M are subject to epigenetic modulations under steady-state conditions, it is conceivable that exposure to or withdrawal from different microenvironmental cues may alter the epigenetic landscape of aging CD8⁺ T_M.

In summary, the present work confirms and expands the central tenets of the rebound model (1) by documenting the fundamentally temporal nature of CD8⁺ T_M homeostasis and identifying associated determinants for improved CD8⁺ T_M survival, metabolic alterations, and lymphoid tissue homing that collectively may brace aged CD8⁺ T_M for enhanced II° expansion (1–5) and associated immune protection (1, 5). The dynamic adaptations of long-term CD8⁺ T cell memory and the possibility to accelerate or delay these processes at large (1) provides an experimental framework for the focused interrogation of suitable targets that may be exploited for the prophylactic or therapeutic modulation of specific CD8⁺ T_M responses. To this end, we have explored elsewhere the specific contribution of 16 molecular pathways to the improved II° reactivity of aged CD8⁺ T_M populations (B. Davenport, J. Eberlein, Tom T. Nguyen, F. Victorino, K. Jhun, V. van der Heide, and D. Homann manuscripts in preparation).

We thank the National Institutes of Health Tetramer Core Facility for provision of biotinylated MHC:peptide monomers, Dr. D. Hildeman for advice about the use of ABT-737, Dr. C. Buettner for provision of atglistatin, R. Wong for help with quantitative RT-PCR analyses, and Dr. P. Marrack for the gift of the BIM antibody. Due to the wide-ranging nature of topics discussed herein we have frequently relied on the citation of review rather than original research articles and wish to apologize to the authors whose work is not explicitly mentioned.

The authors have no financial conflicts of interest.