Sensitivity to Restimulation-Induced Cell Death Is Linked to Glycolytic Metabolism in Human T Cells

Dynamic changes in cellular metabolism are vital during an effective CD8⁺ T cell response. Like most somatic cells, naive and memory T cells operate in a generally quiescent metabolic state and use mitochondrial oxidative phosphorylation (OXPHOS) for ATP generation (1). Following TCR stimulation, however, responding T cells rapidly switch to using glycolysis even in the presence of oxygen (the Warburg effect) (2–4). Activated T cells proliferate and acquire potent effector functions (e.g., IFN-γ production), which have been linked to glycolytic metabolism (2, 4–8). Recent reports demonstrate that changes in cellular metabolism over the course of a T cell response profoundly influence cell survival and differentiation, including the generation of memory (2, 4, 8–13). Interestingly, it is precisely during this window of expansion and aerobic glycolysis that effector T cells become sensitive to activation- or restimulation-induced cell death (RICD).

RICD is a critical apoptotic program that ultimately sets an upper limit for effector T cell expansion during an infection. RICD sensitivity is dependent on prior activation, cell cycle induction via IL-2, and a subsequent, strong restimulation signal propagated through the TCR, which induces apoptosis in a subset of effectors (14–16). Unlike effector T cells, naive and resting memory T cells are relatively resistant to RICD. By constraining effector T cell numbers during the Ag-induced expansion phase, this self-regulatory death pathway helps to maintain immune homeostasis by precluding excessive, non-specific immunopathological damage to the host. Indeed, our laboratory previously demonstrated that a defect in RICD contributes to excessive T cell accumulation and lethal damage to host tissues, as noted in patients with X-linked lymphoproliferative disorder (17, 18).

Although RICD was first described over 25 years ago (16, 19–21), the molecular components that convert TCR signaling from pro-proliferative in naive cells to proapoptotic in restimulated, activated T cells have yet to be fully defined. Additionally, it remains unclear why RICD sensitivity varies for T cells from different normal human donors, and why only a proportion of expanded effector T cells are rendered competent to die after TCR restimulation. Although robust glycolytic metabolism overlaps closely with the window of RICD susceptibility in effector T cells, it is not known whether metabolic reprogramming influences RICD directly. We hypothesized that glycolytic metabolism promotes the sensitization of effector T cells to RICD. To our knowledge, in this study we show for the first time that active glycolysis enhances RICD in effector CD8⁺ T cells, specifically by enabling robust induction of Fas ligand after TCR restimulation. Our findings suggest that restricting glucose (Glc) availability and/or reducing glycolysis may prolong the survival of activated T cells by protecting them from RICD.

Blood from anonymous healthy donors (buffy coats) was generously provided by Dr. Michael Lenardo and the National Institutes of Health Blood Bank. PBMC were isolated using Ficoll density gradient centrifugation, and CD8⁺ T cells were purified from PBMC using the EasySep Human CD8⁺ T cell enrichment kit (Stem Cell Technologies). T cells were activated 1:1 with beads coated with anti-CD3/CD2/CD28 Abs (Human T Cell Activation/Expansion Kit; Miltenyi) in Glc-free RPMI 1640 (Life Technologies) + 10% dialyzed FCS (Life Technologies) + 1 mM sodium pyruvate (Cellgro) + 1% penicillin/streptomycin (Lonza) and either 10 mM d-galactose or d-glucose (Sigma) for 3 d. Activated T cells were washed in PBS and subsequently cultured in Glc- or galactose- (Gal) containing media with 100 U/ml rIL-2 (PeproTech) at 1 × 10⁶ cells/ml for ≥13 d, changing media every 3 d. In some experiments, cells on days 9–12 were washed 2× in PBS and swapped into media containing the opposite sugar as described in the figure legends. For conditioned media experiments, Glc and Gal T cell cultures were spun down on day 14 of culture in IL-2 and cells were resuspended in the opposite conditioned culture media with additional IL-2. These cells were incubated for 30 min and then assayed for RICD as described below. Additionally, cells grown in Gal were washed 2× in PBS and resuspended in media supplemented with 10-fold titrations of Glc prior to RICD assays as described.

RICD assays were performed as previously described (22). Briefly, activated T cells (days 13–15) were treated in triplicate with anti-CD3ε mAb OKT3 (5–500 ng/ml; BioGems), and plated at 7.5 × 10⁵ cells/ml in 96-well round-bottom plates. For some assays, cells were pretreated for 30 min with 2 μM 2-deoxy-glucose (2-DG), 2.7 μM rapamycin, 1 μM oligomycin A, 5 μM rotenone, 10 ng/ml rIFN-γ, 5 μg/ml concanamycin A, 20 mM d-glucose (Sigma-Aldrich) or 1 μg/ml anti-FAS antagonistic Ab SM1/23 (Enzo) versus DMSO or ddH₂0 solvent control. At 24 h after TCR restimulation, cells were stained with 5 μg/ml propidium iodide (PI) (Sigma-Aldrich) and collected for constant time on an Accuri C6 flow cytometer (BD Biosciences). Cell death was quantified as percentage cell loss = 1 − [number of viable cells (treated)/number of viable cells (untreated)] × 100. For some assays, T cells were stained with Annexin V-FITC (BioLegend) 4 h after restimulation. Surface expression of FAS (CD95) and CD3 were assessed using anti-CD95-APC and anti-CD3-PE Abs respectively (BioLegend). Surface expression of CD107a (LAMP1) was measured ±4 h of anti-CD3 restimulation using anti-CD107a-APC (BioLegend). Intracellular flow cytometry for phospho-S6 at baseline and after 4 h of restimulation ±2-DG pretreatment was measured using anti-pS6-FITC (Cell Signaling) with Cytofix Fixation Buffer and Phosflow Perm Buffer III reagents according to product protocol (BD Biosciences). DNA content was used to evaluate cell cycle status ±4 h of anti-CD3 restimulation using methanol fixation and staining with PI and RNAse A (Sigma). Cell cycle status was also assessed using the Click-iT EdU (Thermo Fisher) assay by flow cytometry, according to product protocol. Transfections of Glc and Gal T cells were performed with 5 μg of either pEGFP-C2 or pEGFP-C2-FASL-3′UTR plasmids, a generous gift from the Gallouzi laboratory (23), using the Amaxa P3 Primary Cell 4D-Nucleofactor X Kit L (Lonza) according to the manufacturer’s protocol. GFP expression was analyzed 6 h post-transfection by flow cytometry. All flow cytometric assays were performed on an Accuri C6 flow cytometer (BD Biosciences).

Oxygen consumption and extracellular acidification rates (OCR and ECAR) were measured using a Seahorse XF24 analyzer (Seahorse Bioscience). Primary human effector T cells derived in Glc-containing media (as described above) were attached with Cell-Tak tissue adhesive (Corning) to 24-well Seahorse XF-24 assay plates at ∼500,000 cells/well in Seahorse BASE media with additives. Cells were incubated at 37°C in a non-CO₂ incubator for 45 min. All media was adjusted to pH 7.4 on the day of assay. Mitochondrial and glycolysis stress tests were performed according to the manufacturer’s protocol. OCR (an indicator of oxidative phosphorylation) and ECAR (an indicator of glycolysis) were automatically calculated and recorded by the Seahorse XF-24 software.

Activated CD8⁺ T cells (1 × 10⁶ per time point) were restimulated with 500 ng/ml OKT3 (0–4 h), washed in cold PBS, and lysed in 1% Nonidet P-40 (NP-40) lysis buffer (50 mM Tris [pH 7.4], 150 mM NaCl, 0.5 mM EDTA, 1% NP-40, 0.5% sodium deoxycholate, 1 mM Na₃VO₄, 1 mM NaF) containing complete protease inhibitors (Roche) for 30 min on ice. Cleared lysates were boiled in 2× reducing sample buffer, and resolved on Any kDa SDS-PAGE gels (Bio-Rad). Proteins were transferred to nitrocellulose on a Trans-Blot Turbo system (Bio-Rad), blocked in 2% Tropix I-Block (Applied Biosystems) in TBS/0.1% Tween, and probed with the following Abs: anti-FASL (Ab3; EMD Millipore); anti-BID; anti-cleaved caspase 8 (Cell Signaling Technologies); anti-BIM (Enzo); anti-cleaved caspase 9, anti-cleaved caspase 3, anti-NUR77 (BioLegend); anti-geminin; anti-Cdt1 (Santa Cruz Biotechnology); and anti–β-actin (Sigma-Aldrich). Bound Abs were detected using HRP-conjugated secondary Abs (Southern Biotech, eBioscience) and ECL (Thermo Scientific).

Detection of soluble/cleaved FASL in cell supernatants from T cells with or without anti-CD3 restimulation ( with or without inhibitor pretreatment as described above) was performed using the Quantikine Human Fas Ligand/TNFSF6 Immunoassay Kit (R&D Systems). l-Lactate levels were measured in cell supernatants using a glycolysis cell-based assay kit (Cayman Chemical). IFN-γ secretion was measured in cell supernatants from T cells with or without anti-CD3 restimulation using the Ready SET Go Human ELISA IFN-γ kit (eBioscience). ELISA plates were read using a Synergy H1 Hybrid Reader (BioTek); concentrations of soluble FASL (sFASL) (pg/ml), l-lactate (μM) or IFN-γ (pg/ml) were calculated using Gen5 data analysis software (BioTek).

RNA was isolated from Glc or Gal T cells from two donors at baseline or after 4 h of OKT3 restimulation ±2DG pretreatment using QIAshredder and RNeasy mini plus columns (Qiagen).

cDNA was prepared from 300 ng RNA using the MultiScribe Reverse Transcriptase Kit (ThermoFisher Scientific). Maxima SYBR Green/ROX qPCR Master Mix (ThermoFisher Scientific) was used for subsequent PCR with specific primers against FASL (for: 5′-CTTCCACCTACAGAAGGAGC-3′, rev: 5′-CCAGAAAGCAGGACAATTCC-3′) and RPL13 (for: 5′-GAATGGCATGGTCTTGAAGCC-3′, rev: 5′-GGGAATGTGCTGTTTCCATGG-3′) as a reference control, analyzed on a StepOnePlus Real-Time PCR System (Applied Biosystems).

In vitro cell death assays were evaluated using two-way ANOVA (α = 0.05) with Sidak correction for multiple comparisons or Student t test where appropriate. l-Lactate, maximum ECAR, OCR, percentage of cell loss, and sFASL values were correlated using Pearson’s comparison analysis and graphs were generated using linear regression. All statistical analyses were performed using GraphPad PRISM software. Error bars are defined in the figure legends as ± SEM or ± SD where appropriate. Asterisks denote statistical significance and p values are reported in figure legends.

To investigate whether donor-dependent variability in RICD sensitivity is associated with glycolytic metabolism, we first measured RICD and l-lactate production in CD8⁺ effector T cells derived from three human donors after restimulation with the agonistic anti-CD3 Ab OKT3. Donor T cells that displayed higher RICD sensitivity also produced more l-lactate, a secreted product of glycolysis, after 4 h of restimulation (Fig. 1A, 1B). Indeed, when data collected from 12 donors were subjected to linear regression analysis, we found a significant correlation existed between RICD sensitivity and l-lactate measured in the supernatant (Fig. 1C, R = 0.8772). Consistently, the maximum ECAR achieved after TCR restimulation also demonstrated significant correlation with RICD sensitivity in six donors with variable sensitivity (Fig. 1D–F, R = 0.9142). In contrast, OCR as a measure of OXPHOS did not significantly correlate with RICD sensitivity in the same six donors tested (Supplemental Fig. 1A, 1B). Moreover, effector T cells treated with oligomycin A (mitochondrial ATP synthase inhibitor) or rotenone (electron transport complex I inhibitor) showed only a slight, non-significant increase in RICD sensitivity (Supplemental Fig. 1C, 1D). These data suggest that RICD sensitivity directly correlates with glycolytic activity, but not OXPHOS, in human CD8⁺ T cells.

To establish a causal link between glycolytic metabolism and RICD, we next expanded effector T cells from single donors in Glc- versus Gal-containing culture medium. Substituting Gal for Glc severely restricts glycolysis and forces T cells to predominantly use OXPHOS (24). Interestingly, donor T cells cultured in Gal culture media were significantly less sensitive to RICD compared with T cells cultured in Glc-containing media (Fig. 1G). This difference could not be explained by a broader defect in programmed cell death, as both Glc- and Gal-cultured T cells were equally sensitive to direct FAS ligation (Fig. 1H). Glc- and Gal-cultured T cells also displayed equivalent cell surface expression of TCR (CD3) and FAS (CD95) (Fig. 1I). These data imply that RICD susceptibility is specifically influenced by metabolic status in effector T cells.

To explore the link between glycolysis and RICD, we conducted further apoptosis assays comparing Glc with Gal T cells in the presence of the competitive Glc analog 2-DG (7). Brief pretreatment with 2-DG substantially reduced the number of 4 h restimulated Glc T cells staining positive with annexin V, an early marker of apoptosis commitment, and significantly reduced RICD sensitivity at 24 h (Fig. 2A, 2B). Indeed, Glc T cells treated with 2-DG showed a dramatic reduction in l-lactate production after 4 h of restimulation, confirming decreased glycolysis (Fig. 2C). Although Gal T cells primarily use OXPHOS, Gal can be used for glycolysis at much lower efficiency than Glc (25). Hence 2-DG also reduced RICD sensitivity in Gal T cells (Fig. 2B), associated with a detectable decrease in lactate production in most donors tested (Fig. 2C). These results indicate that acute inhibition of glycolysis renders T cells less sensitive to RICD.

We next asked whether the RICD sensitivity of effector T cells relying primarily on glycolysis versus OXPHOS could be altered by prolonged or acute changes in Glc availability. Interestingly, we could generate a step-wise reduction in Glc T cell RICD sensitivity by swapping cells into Gal-containing media for 1–5 d of culture prior to restimulation (Fig. 2D). Conversely, supplementing normal culture media with additional Glc further increased RICD of Glc T cells slightly, reaching maximum sensitivity at 20 mM (data not shown). Because Gal T cells preferentially use pyruvate for the TCA cycle, they lack excess pyruvate for subsequent conversion to secreted l-lactate. We therefore asked whether extracellular l-lactate was sensitizing Glc T cells to RICD by swapping Glc T cells into Gal-conditioned culture media (low l-lactate) and Gal T cells into Glc-conditioned culture (high l-lactate). We saw no increase in Gal RICD sensitivity in the Glc-conditioned media compared with Gal T cells in Gal-containing media (Fig. 2E), suggesting that: 1) excess l-lactate and/or other secreted factors did not provide a sensitizing feedback signal to Glc T cells; and 2) little to no Glc remained in the Glc-conditioned media. Conversely, Gal T cells swapped into fresh 10 mM Glc-containing media for just 30 min were notably more sensitive to RICD (Fig. 2F). This enhanced RICD sensitivity was titratable and decreased with serial 10-fold dilutions of Glc in the media (Fig. 2F). The addition of 2-DG helped accentuate these differences, presumably by impeding hexokinase activity and restricting the entry of freshly added Glc into the glycolytic cycle. Collectively, these data imply that acute Glc availability helps set the threshold for RICD sensitivity in CD8⁺ effector T cells.

To understand how glycolytic metabolism drives RICD sensitivity, we examined several requirements known to render effector T cells competent to die through this pathway. Because only effector T cells that are actively cycling are sensitive to RICD (15, 26, 27), we next asked whether cell cycle progression was specifically enhanced in Glc T cells before or after TCR restimulation, relative to Gal T cells. Using PI cell cycle analysis, we found an equal percentage of cells actively dividing (S + G2/M phases) at baseline in both Glc and Gal T cells (Fig. 3A). After 4 h of restimulation, we measured a lower percentage of Glc T cells in cycle compared with Gal T cells, which was rescued by pretreatment with 2-DG (Fig. 3A). Glc T cell cultures displayed a concomitant increase in the proportion of sub-G1/apoptotic cells, consistent with RICD induction (Fig. 3A). Immunoblotting also showed no marked changes in the expression of key cell cycle checkpoint proteins Cdc10-dependent transcript 1 (Cdt-1) and geminin at baseline, after 4 h restimulation, or with 2-DG treatment between Glc and Gal T effector cells (28) (Supplemental Fig. 2). We further corroborated these findings by examining cell proliferation before and after 4 h of restimulation. Glc and Gal T cells demonstrated equivalent EdU incorporation in all conditions tested (Fig. 3B). These data suggest that Glc T cells are not more sensitive to RICD simply because a greater proportion of cells are proliferating or induced into cell cycle upon TCR restimulation.

IFN-γ has also been implicated in potentiating RICD sensitivity (29). Previous studies demonstrated that T cells cultured in Gal media proliferate but cannot mount sufficient effector functions, including IFN-γ secretion (6, 7, 24). We confirmed a similar defect in IFN-γ secretion by Gal T cells (Fig. 3C). However, the addition of exogenous IFN-γ did not boost RICD sensitivity in Gal T cells to levels measured in Glc T cells (Fig. 3D). Interestingly, Gal T cells demonstrated greater CD107a (LAMP1) staining by flow cytometry at baseline and after 4 h of restimulation than Glc T cells (Fig. 3E). Moreover, inhibition of lytic granule maturation with concanamycin A (CMA) (30) treatment did not preferentially decrease Glc T cell sensitivity (Fig. 3F). These data suggest that diminished RICD sensitivity of Gal T cells is not explained by deficiencies in IFN-γ production or perforin-mediated cytotoxicity.

We subsequently investigated whether RICD sensitivity is influenced by mammalian target of rapamycin complex 1 (mTORC1) and c-Myc, critical signaling nodes activated during glycolysis that promote anabolic metabolism (31). Treatment with rapamycin reduced the RICD of Glc and Gal T cells slightly, but did not eliminate the death sensitivity difference between these subsets (Fig. 4A). As expected, Glc T cells displayed higher phosphorylated ribosomal protein S6 (pS6) staining, a surrogate for mTORC1 activity, at baseline and after 4 h of restimulation compared with Gal T cells. Whereas pretreatment with 2-DG partially reduced the pS6 signal, rapamycin reduced pS6 activity following restimulation in both Glc and Gal T cells to levels below baseline, indicating potent inhibition of mTORC1 activity. We also observed equivalent expression of c-Myc in Glc and Gal T cells ± restimulation, with no change in RICD observed after pretreatment with c-Myc inhibitor JQ1 (data not shown) (32). Together these data suggest that the difference in RICD sensitivity noted between Glc and Gal T cells is largely independent of mTORC1 activity and c-Myc signaling.

To determine the mechanism by which glycolysis promotes the execution of RICD, we generated cell lysates from Glc versus Gal T cells pre and post TCR restimulation to compare expression of critical proapoptotic proteins. As reported previously, immunoblotting revealed a robust de novo induction of full-length FASL in Glc T cells, much of which is rapidly cleaved at the plasma membrane to generate a prominent N-terminal fragment (Fig. 5A) (18). Strikingly, full-length and N-terminal fragment FASL expression was markedly reduced in restimulated Gal T cells after 4 h of restimulation (Fig. 5A, top panel). Consistent with a reduction in RICD sensitivity, FASL induction in Glc T cells was almost completely blocked with 2-DG treatment (Fig. 5A). We also noted a concomitant reduction in cleaved BID, active caspase-8, caspase-9, and caspase-3 in Gal T cells, or in Glc T cells with 2-DG treatment (Fig. 5A). In contrast, the induction of other proapoptotic proteins important for RICD of CD8⁺ T cells, including BIM and NUR77, was normal for all culture conditions tested (Fig. 5A) (17, 18). An ELISA confirmed the differential induction of FASL in Glc versus Gal T cells after restimulation, with more release of sFASL in Glc versus Gal T cell supernatants from each donor tested (Fig. 5B). Although the sFASL detected here may include both soluble, cleaved FASL (relatively inactive) and exosomal FASL (highly active), its detection in the supernatant provides another reliable readout of FASL protein induction, as we have shown previously (18). Consistent with immunoblotting results, TCR-triggered FASL release was substantially reduced with 2-DG pretreatment (Fig. 5B), but not rapamycin (data not shown).

We next used quantitative PCR (qPCR) to examine transcriptional control of FASL induction in Glc and Gal T cells. Glc T cells expressed significantly more FASL mRNA than Gal T cells at baseline and after restimulation ±2-DG treatment (Fig. 5C). However, FASL transcription was still robustly induced in Gal T cells after restimulation (Fig. 5C), despite a profound decrease in FASL protein (Fig. 5A). Similarly, 2-DG treatment substantially decreased FASL protein expression (Fig. 5A) without affecting mRNA levels postrestimulation (Fig. 5C). These data imply that a post-transcriptional mechanism governs the decrease in FASL protein expression observed with restriction of glycolysis in T cells. We confirmed this using a GFP reporter construct linked to the 3′ untranslated region (UTR) of FASL mRNA, which is known to regulate FASL protein translation (23). Indeed, the FASL 3′UTR reduced GFP expression significantly more in Gal T cells (Fig. 5D). Additionally, treatment of transfected Glc T cells with 2-DG also reduced FASL 3′UTR-GFP reporter expression [average mean fluorescence intensity (MFI) decrease = 42.3 ± 5.8%; data not shown]. Together these results imply that translation of FASL mRNA in T cells is suppressed under conditions of limited glycolysis.

Importantly, linear regression analysis of Glc T cells from nine separate donors revealed a strong positive correlation between l-lactate and sFASL protein concentrations in cell supernatants (Fig. 5E, R = 0.8033), suggesting a direct association between glycolysis and TCR-induced FASL upregulation in CD8⁺ effector T cells. To test whether differences in FASL protein induction specifically explained relative RICD sensitivity in Glc versus Gal T cells, we blocked the death receptor FAS during TCR restimulation. Pretreatment with the antagonistic FAS blocking Ab SM1/23 reduced RICD sensitivity of Glc T cells to a much greater extent than Gal T cells (Fig. 5F). More importantly, FAS blockade completely abolished any difference in RICD sensitivity between Glc and Gal T cells (Fig. 5F). These data suggest that glycolytic metabolism specifically promotes RICD sensitivity in effector CD8⁺ T cells by permitting robust FASL induction and FAS-mediated apoptosis.

This study highlights glycolytic metabolism as a novel requirement for licensing the FASL-dependent component of RICD sensitivity in effector CD8⁺ T cells. Interestingly, we detected no substantial differences in other known requirements for RICD between Glc and Gal T cells, including TCR expression and IL-2–dependent cell cycling. For proliferating T cells, glycolysis is not only employed for macromolecule synthesis, but is also critical for acquiring full effector functions (3, 6, 24, 25). However, our results suggest differences in IFN-γ or cytolytic granule release cannot account for differential RICD between Glc and Gal T cells. We also noted comparable upregulation of TCR-induced proapoptotic molecules BIM and NUR77 following restimulation of Glc versus Gal T cells ± 2-DG treatment. Unlike SAP-deficient T cells, these data imply that decreased RICD is not caused by a global attenuation of downstream TCR signal strength when glycolysis is restricted. Moreover, TCR-induced de novo translation of these proteins is still intact under these conditions, explaining why mTORC1 inhibition had a small, comparable effect on RICD of Glc and Gal T cells.

Our data demonstrate that Glc availability and glycolytic activity enhances RICD specifically through the induction of FASL after TCR restimulation. FASL is one of several proapoptotic proteins that contribute to RICD of effector T cells (18, 33). Previous work showed that FASL is a TCR-responsive transcriptional target of c-Myc, a crucial driver of glycolytic reprogramming in T cells (31, 34). However, we observed no differences in c-Myc expression for Gal and Glc T cells, and no effect on RICD upon treatment with the c-Myc inhibitor JQ-1 (data not shown). Instead, our data suggest glycolysis enables FASL expression by releasing FASL mRNA from posttranscriptional repression. GAPDH, which normally catalyzes the sixth step of glycolysis, can bind 3′UTR sequences of cytokine mRNAs (e.g., IFN-γ) to limit their translation, especially when glycolysis is limited (24). However, partial knockdown of GAPDH (75%) failed to rescue FASL expression and boost RICD in Gal T cells (data not shown). It will be of interest to survey other glycolytic intermediates and enzymes for potential moonlighting activities in the control of FASL expression (35).

Linking RICD sensitivity to glycolysis provides an elegant control mechanism for maintaining immune homeostasis by precluding the excessive expansion of terminally differentiated, highly glycolytic effector T cells. It will be important to determine whether certain effector T cells that modulate glycolysis and switch to OXPHOS, fatty acid oxidation, and/or autophagy can preferentially enter the memory pool in part by escaping RICD (5, 31, 36). Recently identified effector T cell subsets such as short-lived effectors and memory precursor T cells differ in their survival and potential for memory formation (37, 38), which may be influenced by a divergence in glycolysis-dependent RICD sensitivity. For example, exposure to cytokines like IL-15 might protect selected effectors from RICD by facilitating a switch from aerobic glycolysis to OXPHOS (4, 38–40).

In linking glycolysis with susceptibility to a specific self-regulatory apoptosis program, our study complements many recent reports tying T cell survival and memory generation to catabolic metabolism (5, 31, 36). We posit that pharmacological interventions that promote memory T cell formation via metabolic reprogramming may work in part by allowing greater numbers of T cells to escape RICD (12, 41). Therefore, agents that alter metabolic programming in T cells may prove useful in regulating the magnitude of any given T cell response specifically by tuning RICD sensitivity, an emerging therapeutic concept for correcting dysregulated immune homeostasis (17).

We thank the laboratory of Dr. Chou-Zen Giam for advice on interrogating cell cycle proteins, Dr. Michael Lenardo and the National Institutes of Health Blood Bank for providing access to anonymous blood donations, Dr. Barrington Burnett’s laboratory for assistance with qPCR, Dr. Cara Olsen for statistics evaluation, Dr. Imed Gallouzi’s laboratory for GFP reporter plasmids, and Drs. Edward Mitre, Brian Schaefer, Joseph Mattapallil, and Pamela Schwartzberg for helpful discussions.

The authors have no financial conflicts of interest.