TLR-Mediated Innate Production of IFN-γ by CD8⁺ T Cells Is Independent of Glycolysis

The CD8⁺ T cells are critical members for defense against intracellular pathogens. They rapidly shift from a quiescent to a highly active state upon reinfection with the pathogen, which allows them to produce massive amounts of effector molecules within a few hours (1–3). IFN-γ is a central effector molecule of CD8⁺ T cells that has a broad spectrum of activity (4, 5). IFN-γ not only dampens the growth of pathogens, it also recruits neutrophils to the site of infection and activates immune cells such as macrophages to potentiate the innate immune response (6–8). In humans, intact IFN-γ production and IFN-γR signaling is essential for antimycobacterial immunity (9).

T cells reside in lymphoid and in nonlymphoid tissues, such as skin, lung, kidney, and the female reproductive tract (2, 10–13). This tissue-specific localization ensures that T cells exert their effector function right at the entry site of the pathogen and thereby limit the spreading of the infection. CD8⁺ T cells are not only important during reinfections with the cognate pathogen. They also have the capacity to respond to unrelated pathogenic intruders (14–16). In the absence of Ag recognition through the TCR, CD8⁺ T cells can acquire innate defense functions. Inflammatory cytokines can directly activate Ag-experienced T cells and, for example, induce the production of IFN-γ or granzyme B (15, 17–20). The production of IFN-γ during this bystander activation supports the immune response against unrelated infections (14, 16). To achieve this cross-protection, innate immune cells like dendritic cells and NK cells are recruited to the site of infection (7, 21), and macrophages and neutrophils become activated by IFN-γ (8, 22, 23).

Although it has been established that Ag-experienced CD8⁺ T cells respond to inflammatory cytokines, it is not known whether also other stimuli can trigger T cells in an Ag-independent manner. For instance, ligands for TLRs signify the presence of microbial non-self and are major activators of innate immunity (24). We therefore asked whether TLR ligands would also mobilize innate immune functions by previously activated T cells. Earlier studies have shown that bacterial lipoproteins and CpG sequences provide costimulation to the TCR engagement during T cell activation (25, 26). In particular, costimulation through TLR2 triggering lowers the threshold of Ag required for an optimal activation of CD8⁺ T cells (27). In addition, TLR2-mediated costimulation promotes the development and maintenance of memory T cells (25, 28, 29). However, whether TLR ligands alone can drive the activation of Ag-experienced T cells is yet to be determined. In the current study, we show that murine Ag-experienced CD8⁺ T cells can be directly activated by TLR2 and TLR7 ligands but not by TLR3, TLR4, TLR5, or TLR9 ligands. TLR2- and TLR7-mediated activation of CD8⁺ T cells does not induce production of TNF-α or IL-2 but results in production of IFN-γ that is sufficient to alert and prime macrophages. TLR-driven IFN-γ production is of short nature because it relies on newly synthesized mRNA without increasing the rate of translation by stabilizing IFN-γ transcripts. Similar to what is found for primary T cell activation and for reactivation of memory T cells with cognate Ag (30), TLR-mediated generation of IFN-γ transcripts relies on the PI3K-Akt signaling axis. Protein production is driven by mechanistic target of rapamycin (mTOR), a downstream target of PI3K-Akt, for both TLR- and TCR-mediated activation. When we investigated the source of energy used during T cell activation, we found different requirements for TLR-mediated and Ag-driven stimulation. Although TCR triggering depends on glycolysis to reach optimal cytokine production during recall responses, external sources of glucose are not required for the production of IFN-γ upon TLR triggering. Instead, the energy generated by mitochondrial respiration is sufficient to fully support the TLR-dependent innate-like production of IFN-γ. In conclusion, we show that bystander activation of T cells through TLRs leads to a short but biologically significant burst of IFN-γ production, which relies exclusively on newly synthetized transcripts and energy generated by mitochondrial respiration.

C57BL/6J mice and C57BL/6J.OT-I TCR (OT-I) transgenic mice were housed and bred in the animal department of the Netherlands Cancer Institute. All animal experiments were performed in accordance with institutional and national guidelines and approved by the Experimental Animal Committee of the Netherlands Cancer Institute.

Mouse T cells and MEC.B7.SigOVA cells were cultured in IMDM (Life Technologies) supplemented with 8% FCS, 50 μM 2-ME, 2 mM l-glutamine, 20 U/ml penicillin G sodium, and 20 μg/ml streptomycin sulfate, unless differently specified.

Bone marrow (BM) cells were obtained from femurs and tibia of C57BL/6J wild-type (WT) or MyD88^−/− mice (provided by J. den Haan, VU University Medical Center, Amsterdam, the Netherlands) and of BALB/c WT or IFN-γR^−/− mice (provided by H. A. Young, National Institutes of Health–National Cancer Institute, Frederick, MD). BM-derived macrophages were generated by a 7-d culture in 100-mm non-tissue-culture–treated dish at a density of 0.2 × 10⁶ cells/ml in RPMI 1640 (GE Healthcare Life) supplemented with 10% FCS, 50 μM 2-ME, 2 mM l-glutamine, 20 U/ml penicillin G sodium, and 20 μg/ml streptomycin sulfate together with 15% L-929 conditioned medium containing recombinant M-CSF. Medium was refreshed once at day 4.

A total of 1 × 10⁶ purified CD8⁺ OT-I T cells (Miltenyi CD8 isolation kit; 80–99% purity) or 1 × 10⁶ FACS-sorted naive CD44^loCD62L^hiCD8⁺ OT-I T cells (BD FACSAria III Cell Sorter; 100% purity) were activated by a 20-h coculture with 0.1 × 10⁶ preseeded MEC.B7.SigOVA cells/well, as described previously (31). C57BL/6J total splenocytes (3 × 10⁶ cells/well) were activated for 48 h with Con A (2 μg/ml; Sigma-Aldrich) and IL-7 (1 ng/ml recombinant murine IL-7; PeproTech). Activated T cells were harvested, removed from the stimuli, and put to rest for 3–15 d in a density of 0.5 × 10⁶/ml the presence of 10 ng/ml recombinant murine IL-7. T cells were replated, and medium was refreshed every 3–4 d.

Resting OT-I T cells were stimulated for 6 h with 1 or 100 nM OVA_257–264 peptide, or with the TLR ligands (all InvivoGen) Pam₃CysSK₄ (Pam3; 1 ng/ml–10 μg/ml), polyinosinic-polycytidylic acid [poly(I:C)] (50 μg/ml), LPS (10 μg/ml), flagellin (10 μg/ml), R848 (1 ng/ml–10 μg/ml), CpGA (10 μg/ml), and CpGB (10 μg/ml), unless differently specified.

To test possible indirect effects of IL-12, 10 μg/ml IL-12 neutralizing Ab (C17.8; eBioscience) was added to resting OT-I T cells 30 min prior to TLR-dependent activation.

For inhibitor experiments, 1 μg/ml PI-103 (Tocris Bioscience), 10 μg/ml Akti1/2, 10 ng/ml to 5μg/ml rapamycin (both Sigma-Aldrich), 25–150 ng/ml cyclosporine A, 30 μM IC87114 (both Calbiochem), 10 μM ZTSK474 (ZTS) (Selleckchem), or DMSO control was added to resting OT-I T cells 30 min prior to Pam3 (5 μg/ml) or OVA_257–264 peptide (1 or 100 nM) stimulation.

A total of 2 × 10⁶ OT-I T cells (6 d of rest) were stimulated for 6 h with 5 μg/ml Pam3 or left untreated in a 24-well plate in 2 ml medium/well. T cell supernatant was harvested, and a total of 350 μl was used to activate 0.3 × 10⁶ macrophages/well for 5 h. Macrophages were washed with PBS prior to RNA isolation.

Cells were washed with FACS buffer (PBS containing 1% FCS and 2 mM EDTA) and labeled for 20 min at 4°C with the following mAbs (all from eBioscience): anti-CD8α (53-6.7), anti-CD8β (H35-17.2), anti-CD4 (GK1.5), anti-CD44 (IM7), anti–L-selectin (CD62L) (MEL-14), anti-CD11b (M1/70), anti-CD11c (N418), anti–IFN-γ (XMG1.2), anti–TNF-α (MP6-XT22), anti–IL-2 (JES6-5H4), anti–phospho-Akt (S473, SDRNR), anti–phospho-mTOR (S2448, MRRBY), and anti–phospho-S6 (S235/S236, cupk43k). When necessary, cells were incubated with anti-CD16/CD32 blocking Ab (2.4G2; a gift from L. Boon, Bioceros, Utrecht, The Netherlands). To exclude dead cells from analysis, Near-IR (Life Technology) was added to the cells. To analyze intracellular cytokine levels, cells were fixed and permeabilized with the cytofix/cytoperm kit, according to the manufacturer’s protocol (BD Biosciences). IFN-γ–, TNF-α–, and IL-2–producing cells were gated on Near-IR^negCD8⁺ T cells prior to analysis. To assess phosphorylation levels of Akt, mTOR, and S6, activated T cells were fixed in 4% paraformaldehyde and permeabilized with methanol prior to staining. Expression levels were acquired using FACS LSR Fortessa (BD Biosciences), and data were analyzed using FlowJo software (version 10; Tree Star).

The Seahorse XF-96 metabolic extracellular flux analyzer (Seahorse Bioscience) was used to measure OCR (in picomoles/min) or extracellular acidification rate (ECAR) (in mpH/min). After 7 d of rest, CD8⁺ OT-I T cells were resuspended in nutrient-free unbuffered DMEM (Sigma-Aldrich) that was supplemented with 16 mg/l phenol red, 1,85g/l NaCl, and 2 M NaOH. The nutrient-free unbuffered DMEM was used with or without the presence of 25 mM glucose, 1 mM sodium pyruvate, and 2 mM l-glutamine (all Sigma-Aldrich). A total of 0.3 × 10⁶ cells/well were plated into Seahorse cell plates coated with poly-lysine (Sigma-Aldrich) to enhance T cell attachment. Pam3, OVA_257–264 peptide, or medium (used as negative control) were directly applied to plated cells by using the instrument’s multi-injection ports. Experiments with the Seahorse system were performed with the following assay conditions: 2-min mixture and 3-min measurement. Data were measured in 6-plo or 9-plo and analyzed using Wave software (Seahorse Bioscience).

For glucose-dependency experiments, resting OT-I T cells were washed once in serum-free glucose-free RPMI 1640 medium (Life Technologies) and then plated in the same medium with or without 25 mM glucose (Sigma-Aldrich). For inhibition of glucose usage, 10 mM 2-deoxy-d-glucose (2-DG; Sigma-Aldrich) was added to the stimulation.

For glutamine-dependency experiments, resting OT-I T cells were washed once in serum-free l-glutamine-free RPMI 1640 medium (Life Technologies) and then plated in the same medium with or without 2 mM l-glutamine (Sigma-Aldrich). For inhibition of glutamine usage, cells were cultured in the presence of 500 μM diazo-5-oxo-l-norleucine (DON; Sigma-Aldrich). For mitochondrial respiration-dependency experiments, resting OT-I T cells were activated in the presence of 15 μM carbonyl cyanide-p-trifluoromethoxyphenylhydrazone (FCCP) or 10 μM oligomycin (Sigma-Aldrich).

Total RNA was extracted using TRIzol reagent (Invitrogen). cDNA was synthesized using SuperScript III Reverse Transcriptase (Invitrogen). Quantitative real-time PCR was performed with SYBR green and a StepOne Plus RT-PCR system (both Applied Biosystems). The sequences of primers used are listed in Supplemental Table I. Reactions were performed in triplicate. Threshold cycle values were normalized to L32 levels.

When mRNA stability was studied, resting T cells were treated with 10 μg/ml actinomycin D (Sigma-Aldrich) for indicated time points upon 1 h up to 3h of activation.

Results are expressed as mean ± SD. Statistical analysis between groups was performed with GraphPad Prism 6, using paired or unpaired two-tailed Student t test when comparing two groups, or one-way ANOVA test with Dunnett’s multiple comparison when comparing more than two groups. A p value < 0.05 was considered to be statistically significant.

We first set out to determine whether Ag-experienced T cells can be bystander activated by TLR ligands. We generated Ag-experienced T cells by activating MACS-enriched CD8⁺ OT-I T cells for 20 h with MEC.B7.SigOVA cells expressing the cognate Ag OVA_257–264 and the costimulatory molecule CD80. This model system results in full T cell activation and allows for subsequent development of bona fide memory T cells in vivo (31). Upon activation, T cells were allowed to rest in the absence of Ag for several days. Within 24 h of Ag removal, T cells ceased to produce IFN-γ protein, but when activated again with cognate Ag, they generated substantial amounts of IFN-γ within a few hours (Supplemental Fig. 1A).

In line with previous studies in memory T cells (32), in vitro–generated Ag-experienced CD8⁺ T cells express transcripts for TLR2, TLR3, TLR5, TLR7, and TLR9 (Fig. 1A). Nonetheless, only stimulation with the TLR2 ligand Pam3 and the TLR7 ligand R848 led to production of IFN-γ in T cells (Fig. 1B). All other TLR ligands tested [i.e., TLR3 ligand poly(I:C), TLR4 ligand LPS, TLR5 ligand flagellin, and TLR9 ligands CpGA and CpGB] failed to do so (Fig. 1B).

To examine whether the observed TLR-mediated IFN-γ production by CD8⁺ T cells was a direct effect or driven indirectly by myeloid cell impurities present after the MACS-purification procedure, we compared the composition of MACS-purified T cells directly after selection and after 3 d of culture prior to the TLR activation. MACS selection of CD8⁺ T cells led to a substantial loss of CD11b⁺ and CD11c⁺ cells, that was even further decreased when T cells were rested for 3 d, resulting in a T cell purity > 99% (Fig. 1C, 1D). To exclude that TLR-induced IL-12 production by possible contaminations of dendritic cells or macrophages (33) was the driving force for IFN-γ production in T cells, resting T cells were stimulated with Pam3 and R848 in the presence of the IL-12 neutralizing Ab. In this setting, IFN-γ production by T cells was not at all affected (Supplemental Fig. 1B). By using FACS-sorted naive CD44^loCD62L^hiCD8⁺ OT-I T cells for primary T cell activation, we confirmed that possible contaminations were not the driving force of TLR-dependent IFN-γ production by T cells. This procedure reached a purity > 99.9% prior to primary T cell activation (Fig. 1E), and as expected, contamination of myeloid cells was not observed in FACS-sorted T cells (Fig. 1E). Importantly, irrespective of the method used for T cell isolation prior to primary T cell activation (i.e., MACS-selected or FACS-sorted cells), we found that TLR2- and TLR7-mediated activation of Ag-experienced CD8⁺ T cells triggered the production of IFN-γ (Fig. 1F, 1G). Therefore, we conclude that TLR triggering acted directly on the T cells.

Interestingly, CD8⁺ T cells specifically produced IFN-γ upon TLR triggering, whereas two other critical cytokines for T cell function, TNF-α and IL-2, and the degranulation marker CD107a were barely detectable (Fig. 2A, Supplemental Fig. 1C; data not shown). The TLR-mediated production of IFN-γ was robust over a longer culture period of T cells, from 3 d up to 15 d of rest (Fig. 2B, Supplemental Fig. 1D). The response to Pam3 and R848 was dose and time dependent, with a maximal IFN-γ production level within the first 6 h of stimulation (Fig. 2C, 2D). Furthermore, TLR-dependent innate activation is more pronounced in CD8⁺ T cells than in CD4⁺ T cells. Although CD8⁺ T cells from C57BL/6 mice clearly responded to stimulation with TLR2 and TLR7 ligands, only a minor fraction of the CD4⁺ T cell population produced IFN-γ within 6 h (Fig. 2E).

We next confirmed that the IFN-γ produced by T cells was a direct result of the TLR triggering and that it was occurring in an Ag-independent fashion. To this end, we activated T cells in the presence of cyclosporine A (CsA) to block TCR-mediated calcineurin activation, and consequentially, TCR-mediated IFN-γ production. Although CsA effectively inhibited the production of IFN-γ in OVA_257–264-activated OT-I T cells in a dose-dependent manner (Fig. 2F), CsA had no effect on the IFN-γ production upon TLR2 or TLR7 triggering (Fig. 2F). This finding reveals that T cells specifically responded to TLR activation and not to possible residual traces of Ag from the primary stimulation. Furthermore, TLR-activated T cells also effectively produced IFN-γ in FCS-free culture medium (Fig. 2G), demonstrating that TLR-mediated activation did not rely on additional triggers from FCS-derived cytokines or growth factors. Taken together, we show in this study that CD8⁺ T cells respond directly to pathogen-derived danger signals via TLRs by producing IFN-γ.

Previous studies have shown that IFN-γ synergizes with TLR ligands to activate macrophages (22, 23, 34). In light of our findings that TLR engagement triggered the production of IFN-γ by Ag-experienced T cells, we examined whether the amount of IFN-γ produced was sufficient to support the activation of macrophages. To this end, we harvested supernatant from resting T cells that were activated for 6 h with Pam3, which resulted in ∼10% IFN-γ–producing T cells (data not shown). When we cultured murine BM–derived macrophages from WT mice for 5 h with the T cell–derived supernatant, we found a significant induction of TNF-α transcripts, a signature for activated macrophages (Fig. 3A). Importantly, MyD88^−/− macrophages that completely fail to respond to TLR2 ligands (35) also exhibited an increase in TNF-α transcripts, albeit to a lesser extent than WT macrophages (Fig. 3A). To address whether IFN-γ produced by Pam3-activated T cells was the factor that contributed to the activation of MyD88^−/− macrophages, we exposed macrophages from IFN-γR^−/− mice to the T cell supernatant. In contrast to direct Pam3 activation of IFN-γR–deficient macrophages that induced similar levels of TNF-α transcripts when compared with WT (Supplemental Fig. 2A), TNF-α transcript levels upon exposure to the T cell supernatant were significantly reduced (Fig. 3B). These data thus demonstrate that IFN-γ produced by T cells is at least in part responsible for the induction of TNF-α transcripts.

Exposure of macrophages to IFN-γ was shown to suppress the production of the anti-inflammatory cytokine IL-10 (23). MyD88^−/− macrophages completely failed to produce IL-10 transcripts (Fig. 3C), indicating that IL-10 transcription is driven by TLR triggering. To determine the role of IFN-γ in the production of IL-10, we measured the responsiveness of IFN-γR^−/− macrophages. Interestingly, whereas Pam3 activation induces equal levels of IL-10 transcripts in macrophages from both genetic backgrounds (Supplemental Fig. 2B), the lack of IFN-γR signaling resulted in a 4-fold increase of IL-10 transcripts compared with WT cells (Fig. 3D), demonstrating that the T cell–derived IFN-γ helps to contain the TLR-mediated induction of IL-10 in macrophages. Taken together, we show that TLR-dependent production of IFN-γ by CD8⁺ T cells is sufficient to enhance macrophage responses to TLR ligands. It therefore shows that TLR-driven innate activation of CD8⁺ T cells is biologically meaningful.

We next determined how the production of IFN-γ was governed in Ag-experienced T cells upon TLR triggering. We found that Pam3-mediated activation resulted in substantial IFN-γ protein production (Fig. 4A), albeit to a lesser extent than Ag-specific activation with the cognate peptide (OVA_257–264; Fig. 4A). We also compared the kinetics of IFN-γ mRNA in TCR or Pam3-activated T cells. IFN-γ mRNA in Ag-reactivated T cells peaked 4 h after activation, and levels remained high for a prolonged period (Fig. 4B). Induction of IFN-γ mRNA by Pam3 was weaker and more transient, peaking already within the first 2 h after stimulation, and rapidly declining at 4 h poststimulation (Fig. 4B). This transient induction upon TLR2 stimulation was also reflected by a short half-life of the IFN-γ mRNA. When the de novo mRNA synthesis blocking reagent actinomycin D (ActD) was added 3 h after Pam3-mediated stimulation, we found that IFN-γ mRNA was rapidly degraded and that its half-life (t_1/2 = 30 min) was comparable to the steady state half-life of IFN-γ mRNA in unstimulated T cells (resting; Fig. 4C). In contrast, OVA_257-264-stimulation of resting T cells stabilized IFN-γ mRNA (t_1/2 > 2h; Fig. 4C). Interestingly, while peptide stimulation induced IFN-γ mRNA stabilization already at 1h after T cell activation, TLR triggering failed to do so at any time point measured (Fig. 4D, 4E). This finding shows that TLR activation is sufficient to promote rapid transcription and protein production, but lacks the signals to stabilize IFN-γ mRNA that is required for sustained protein production. To further examine whether TLR-dependent production of IFN-γ relied primarily on de novo mRNA transcription, we measured the IFN-γ protein levels of TLR- or TCR-triggered T cells that were activated in the presence of ActD. Blocking transcription with ActD for 1–3 h only affected the IFN-γ protein production upon Pam3 stimulation, but not upon OVA_257–264 peptide stimulation (Fig. 4F, 4G).

Our data thus show that the TLR-dependent production of IFN-γ protein primarily depends on mRNA transcription and not on mRNA stabilization. This finding suggests that the rapid degradation of newly synthetized IFN-γ mRNA upon TLR triggering ensures a rapid yet limited production of IFN-γ by T cells in response to unrelated infections.

TLR2 engagement during a primary activation of CD8⁺ T cells promotes cell survival and memory T cell formation via the PI3K-Akt pathway (29). On top of that, activation of mTOR signaling is required to drive the costimulatory effect of TLR2 triggering to increase the production of effector molecules (36). To investigate whether these signaling pathways are also engaged during TLR-mediated bystander activation of Ag-experienced T cells, we measured the phosphorylation of Akt, mTOR and the ribosomal protein S6K1 (S6), a direct target of mTORC1, by flow cytometry 30 min after stimulation with Pam3 or OVA_257–264 peptide. Both stimuli induced phosphorylation of Akt, mTOR, and S6. However, Ag-specific stimulation resulted in more pronounced activation of the PI3K-Akt and mTOR pathway than Pam3 stimulation (Fig. 5A). Of note, blocking PI3K with the chemical inhibitor ZTS prior to T cell activation significantly reduced the staining with Abs against phosphorylated Akt, mTOR and S6, documenting the specificity of the measured phosphorylation signal (Supplemental Fig. 3A).

To determine the relative contribution of PI3K-Akt and mTOR signaling in controlling the production of IFN-γ, we pretreated T cells with specific inhibitors prior to activation with Pam3 or OVA_257–264 peptide. When we inhibited the PI3K-Akt axis with PI-103 (a multitargeted PI3K inhibitor), ZTS (a pan3-PI3K inhibitor), IC87114 (IC, a selective inhibitor of p110δ), or with the Akti1/2 inhibitor, IFN-γ production was effectively blocked, both at protein level and mRNA level (Fig. 5B, 5C and Fig. 5D, 5E, respectively), regardless of the stimulus used.

We next examined the role of mTOR in promoting IFN-γ production. Although microgram ranges of rapamycin (1–5 μg/ml) were required to limit the production of IFN-γ upon OVA_257–264 peptide stimulation, the effect of rapamycin on TLR2 activation was already measurable at ng ranges (10–50 ng/ml) (Fig. 5F, 5G). To determine whether the different responsiveness to rapamycin was a result of different stimuli used or because of the strength of signal provided, we activated Ag-experienced T cells with low concentrations of OVA_257–264 peptide that induced similar levels of IFN-γ as Pam3 stimulation (Supplemental Fig. 3B). Also, this suboptimal Ag stimulation required high concentrations of rapamycin to inhibit IFN-γ production, and the phosphorylation levels of Akt, mTOR, and S6 were higher when compared with Pam3 stimulation (Supplemental Fig. 3C, 3D). Taken together, these results indicate that rapamycin sensitivity was not merely correlated with the magnitude of IFN-γ production.

Intriguingly, even though rapamycin effectively blocked protein production of IFN-γ upon both TLR and TCR triggering, the initial transcription rate was not affected. Independently of the stimulus, or the concentration of rapamycin used, no differences in IFN-γ mRNA were observed (Fig. 5H). Furthermore, mTOR inhibition with 20 ng/ml or 5 μg/ml rapamycin did not change the mRNA half-life of IFN-γ in T cells activated with Pam3 or OVA_257–264 peptide (Fig. 5I, 5J). Thus, these data imply that mTOR activity in T cells primarily promotes the speed and/or the efficiency of translation.

Taken together, our results show that the PI3K-Akt pathway regulates IFN-γ mRNA transcription, whereas mTOR is an important modulator of IFN-γ protein translation in both TCR- and TLR-dependent activation of Ag-experienced T cells. Because reduced mTOR and S6 phosphorylation levels correspond to a lower rate of IFN-γ production upon TLR2 triggering, mTOR may be a limiting factor for protein synthesis.

T cell activation requires a high production of energy to meet the metabolic demands of full effector functions (37–40). This can be generated through oxidative phosphorylation (OXPHOS) and/or aerobic glycolysis (41, 42). Memory T cells have an increased glycolytic capacity compared with naive T cells (42, 43), and the PI3K-Akt and mTORC2 pathway was found to drive the immediate glycolytic switch upon TCR-mediated activation of memory T cells to produce the required energy (43). We therefore questioned whether TLR triggering of Ag-experienced T cells also resulted in an immediate switch toward aerobic glycolysis. We measured the ECAR (an indicator of aerobic glycolysis) and the oxygen consumption rate (OCR, an indicator of OXPHOS) of Pam3 or OVA_257–264 peptide–activated T cells with the in-Seahorse activation assay. Both stimuli evoked a significant and sustained increase of ECAR in T cells that lasted during the entire duration of the essay (Fig. 6A–C). In addition, upon stimulation the OCR/ECAR ratio drops (Fig. 6D), demonstrating that T cells undergo a glycolytic switch irrespective of the stimulus used.

We next investigated which substrate was preferentially used to generate energy during the activation of Ag-experienced T cells. To this end, we compared the TLR- and TCR-mediated activation when the in-Seahorse activation assay was performed in fully supplemented medium or in medium that contained glucose only or l-glutamine only or that was not supplemented at all. In line with earlier reports (43, 44), ECAR levels increased only in the presence of glucose, demonstrating that glucose is an important energy source for T cell function, independently of the stimulus (Fig. 6E). When resting T cells were activated in medium lacking glucose, we found that the oxygen consumption rate increased, indicating that respiration can substitute glycolysis as a mechanism to produce energy (Fig. 6F). Interestingly, the empty medium showed similar increased OCR levels as the medium containing l-glutamine only (Fig. 6F), suggesting that the switch to respiration in the absence of extracellular glucose did not rely on the presence of extracellular l-glutamine. Taken together, these data indicate that T cells can choose which metabolic process (aerobic glycolysis versus mitochondrial respiration) to use to produce the required energy, depending on the availability of nutrients.

Previous studies have found that effector CD8⁺ T cells rely on glucose metabolism for optimal production of IFN-γ (45–47). In this study, we asked whether TLR-mediated IFN-γ production also depends on glucose availability. To this end, we activated Ag-experienced T cells with Pam3 or with OVA_257–264 peptide in FCS-free, glucose-free medium, and we compared their capacity to produce IFN-γ protein with that in the presence of glucose. Intriguingly, although the IFN-γ signal was significantly increased upon Ag-specific stimulation when glucose was available, we found that Pam3-mediated IFN-γ production was insensitive to changes in glucose levels in the culture medium, both in terms of percentage of IFN-γ⁺ T cells and of levels of IFN-γ production as determined by the geometric mean fluorescence intensity (Geo-MFI) (Fig. 7A, 7B). Furthermore, addition of glucose plus its antimetabolite 2-DG significantly reduced the production of IFN-γ in OVA_257–264 peptide stimulated T cells, both at high (100 nM) and at low (1 nM) concentrations of Ag (Fig. 7B, Supplemental Fig. 3E). In contrast, this treatment left IFN-γ levels unchanged in Pam3-activated T cells (Fig. 7A). Therefore, we conclude that TLR-mediated T cell activation is independent of the availability of glucose levels.

Because glucose was redundant for the production of IFN-γ upon TLR2 triggering, we next examined whether l-glutamine was a critical energy source. We activated Ag-experienced T cells with Pam3 or with OVA_257–264 peptide in FCS-free, l-glutamine-free medium, and we compared their responsiveness to activation in l-glutamine–supplemented medium. Interestingly, regardless of the stimulus, l-glutamine significantly increased the ability of T cells to produce IFN-γ. Again, addition of 6-diazo-5-oxo-l-norleucine (DON), a glutamine antagonist that inhibits both glutamine uptake and glutaminolysis (48), reduced the responsiveness of T cells back to basal levels of the empty medium (Fig. 7C). Because TLR- and TCR-activated T cells already produced significant levels of IFN-γ in glucose/glutamine-free medium, we hypothesized that the immediate production of IFN-γ may be supported by intracellular storage of nutrients that T cells acquired during the culture period.

In light of the findings that resting T cells use oxidative phosphorylation to generate ATP necessary for homeostatic functions (38, 47), we examined whether this mitochondrial basal ATP production was also sufficient to fuel the initial production of IFN-γ in response to immunogenic signals. To test this, T cells were activated in the presence of oligomycin to block ATP synthesis or of FCCP, which uncouples ATP synthesis from the electron transport chain. Both drugs substantially affected the IFN-γ production induced by OVA_257–264 peptide and Pam3, demonstrating that mitochondrial respiration is required for the initial production of IFN-γ (Fig. 7D). Of note, T cells displayed the same requirements for nutrients when stimulated between day 3 and 6 or after 15 d of rest, thereby revealing the robustness of our findings (Fig. 7, Supplemental Fig. 3F, 3G, respectively).

Together these data show that TLR-mediated triggering of Ag-experienced T cells does not require external substrates for metabolic demands, but that internal stores of energy substrates are sufficient to promote innate production of IFN-γ in response to Ag-independent triggers.

In this study, we demonstrate that CD8⁺ T cells can be directly activated by the TLR2-ligand Pam₃CysSK₄ and the TLR7-ligand R848 in an Ag-independent manner, allowing CD8⁺ T cells to respond to unrelated infections. The innate-like effector function of T cells primarily triggers the production of IFN-γ but not of cytotoxic cytokines like TNF-α or the release of cytotoxic granules. Rather than killing infected cells, TLR-activated CD8⁺ T cells appear to exert a helper function. In an uninfected host macrophages and CD8⁺ T cells are already located in close proximity within the lymph nodes (49), which should allow for a rapid communication between the two cell types. Furthermore, a cross-talk between myeloid and lymphoid cells was shown to augment host defense in the early phase of infection (7, 8, 49). Therefore, it is tempting to speculate that TLR-mediated IFN-γ production supports this communication with myeloid cells, not only by providing costimulatory signals to effector T cells but also by bystander-activating resting T cells. In line with these findings, the fast TLR-driven release of IFN-γ by CD8⁺ T cells is sufficient to enhance the TLR-mediated activation of macrophages.

Even though bystander-activated CD8⁺ T cells are beneficial to mount an inflammatory response, they need to contain their cytotoxic capacity upon an unrelated infection to avoid immunopathology (50). We noticed that the levels of IFN-γ produced by TLR-triggered T cells are lower than upon Ag-specific activation. This restricted IFN-γ production is provided by lower production of IFN-γ transcripts and the failure to stabilize IFN-γ mRNA. This may ensure that T cell responses to unrelated infections are of short nature but sufficient to restrain pathogenic spread early on during infection, as was demonstrated in several infectious mouse models (7, 15, 51).

Although the production of IFN-γ transcripts upon activation of resting CD8⁺ T cells by TCR or TLR engagement depended on the activation of PI3K-Akt, only triggering with cognate Ag allowed for mRNA stabilization. Whether the difference in mRNA stabilization is determined by the strength of signaling or a result of a qualitative difference in stimulation remains to be determined. Interestingly, mTOR is the limiting factor for protein synthesis for both stimuli, even though TLR-driven IFN-γ production was more sensitive to rapamycin. Again, this difference may be due to the strength of Ag-specific stimulation that requires high levels of rapamycin to achieve a block in protein production. Alternatively, previous studies have indicated that naogram levels of rapamycin inhibit the mTORC1 complex, whereas high (i.e., microgram) levels of rapamycin also affect the mTORC2 complex (52). Hess and colleagues (43) recently found that the immediate glycolytic switch upon TCR-mediated restimulation relies on mTORC2. Our data show that in contrast to triggering with cognate Ag, bystander activation by TLR ligands is also sensitive to low nanogram levels of rapamycin. Therefore, it is tempting to speculate that TLR-mediated activation may be more dependent on the mTORC1 complex, whereas Ag-specific stimulation primarily requires mTORC2.

A critical feature of T cells is to exert immunosurveillance so that reinfection is prevented (53–56). When T cells are in a resting state, they proliferate only at a low rate to replenish their population (57, 58). They have a limited need for energy, which is also reflected in their reduced capacity to take up nutrients (37). To maintain their housekeeping functions T cells produce ATP mainly through catabolic processes that breakdown amino acids, glucose and lipids via OXPHOS (37, 38). Upon Ag-specific activation, however, T cells undergo an immediate glycolytic switch to meet the energy demands for proliferation and for optimal effector function (46, 47, 59). Amino acids and lipids are then used for the production of molecules required to respond to infectious agents, and the generation of ATP is progressively more dependent on aerobic glycolysis (60, 61). In this study, we observe that the production of IFN-γ protein can also occur when the aerobic glycolysis is limited. We show that TLR-mediated activation of resting T cells mainly drives on ATP generated by mitochondrial respiration, independently from the delivery of exogenous nutrients. In fact, supplementing T cell cultures with glucose upon Pam3 stimulation does not increase the production of IFN-γ, and the addition of l-glutamine only has slight effects. We therefore suggest that intracellular sources of nutrients, possibly also of other amino acids such as arginine and leucine (62–64), are sufficient to fuel the ATP production in the mitochondria, which is required for bystander activation-induced IFN-γ production by T cells. In fact, T cells take up and store high levels of metabolites during T cell priming, such as amino acids, lipids, and nucleotides (65), and these may serve as an immediate source of energy upon T cell activation. When in a resting state, IL-7R signaling mediates the synthesis and storage of fatty acids in memory T cells, which promote T cell survival and a fast response to secondary infections (66, 67). Memory T cells also display a substantial mitochondrial mass (42), thereby exhibiting a greater capacity to use both OXPHOS and glycolysis to generate ATP. We therefore suggest that these features may also already be manifested in resting T cells.

In conclusion, we show that bystander activation of resting T cells through TLRs leads to significant but restrained production of IFN-γ that relies exclusively on newly synthetized transcripts and on energy generated by mitochondrial respiration. Ag-experienced T cells are able to recognize pathogen-derived danger signals and alert innate immune cells, and this helper function does not require a glycolytic switch. The limited activation presumably acts only at short range and in the vicinity of the infection and at the same time prevents the mobile T cells to spill out IFN-γ at distant location where no infectious pathogens are. This mechanism allows T cells to act as sensors of unrelated infections without the risk of inducing immunopathology.

We thank the animal caretakers from the NKI for excellent assistance. We also thank J. den Haan for providing BM from MyD88-deficient mice, H.A. Young for BM from IFN-γR–deficient mice, R.H. Houtkooper (Academic Medical Center, University of Amsterdam) for help with the Seahorse assay, and A. Zaal and S. Engels for technical support. We thank F. Vieira Braga and S. Libregts for helpful discussions and H.A. Young and D. Amsen for critical reading of the manuscript.

The authors have no financial conflicts of interest.