NFκB–Pim-1–Eomesodermin axis is critical for maintaining CD8 T-cell memory quality

Significance

Mice and humans whose T cells are deficient in NFκB signaling lack memory T cells, but the mechanism behind this is unclear. We found that NFκB signaling is required during the resolution phase of the immune response to maintain long-term CD8 memory. NFκB signaling is necessary for preserving expression of Eomesodermin and prosurvival Bcl-2 in memory T cells, in a cell-intrinsic process where T-cell receptor (TCR) signals and Pim-1 kinase are involved. Our study defines an unexpected role of NFκB and Pim-1 signaling in the maintenance of T-cell memory quality. Furthermore, it identifies targets and specific times of intervention where protective T-cell memory could be reinforced in vaccines and cancer immunotherapies by manipulation of the NFκB–Pim-1–Eomesodermin axis.

Abstract

T-cell memory is critical for long-term immunity. However, the factors involved in maintaining the persistence, function, and phenotype of the memory pool are undefined. Eomesodermin (Eomes) is required for the establishment of the memory pool. Here, we show that in T cells transitioning to memory, the expression of high levels of Eomes is not constitutive but rather requires a continuum of cell-intrinsic NFκB signaling. Failure to maintain NFκB signals after the peak of the response led to impaired Eomes expression and a defect in the maintenance of CD8 T-cell memory. Strikingly, we found that antigen receptor [T-cell receptor (TCR)] signaling regulates this process through expression of the NFκB-dependent kinase proviral integration site for Moloney murine leukemia virus-1 (PIM-1), which in turn regulates NFκB and Eomes. T cells defective in TCR-dependent NFκB signaling were impaired in late expression of Pim-1, Eomes, and CD8 memory. These defects were rescued when TCR-dependent NFκB signaling was restored. We also found that NFκB–Pim-1 signals were required at memory to maintain memory CD8 T-cell longevity, effector function, and Eomes expression. Hence, an NFκB–Pim-1–Eomes axis regulates Eomes levels to maintain memory fitness.

Memory CD8 T cells provide long-term protection against intracellular pathogens and tumors. T-cell receptor (TCR), costimulatory, and inflammatory signals are required early during infection for efficient memory CD8 T-cell differentiation (1). In addition, IL-7 and IL-15 signals support survival and self-renewal of the memory pool when antigen and inflammatory signals have ceased (2). Interestingly, the memory T-cell pool is heterogeneous and contains distinct T-cell subsets [effector memory (T_EM), central memory (T_CM), stem cell memory (T_SCM), and resident memory (T_RM)] that differ in phenotype, longevity, location, and recall capacity (3). It is currently unknown whether functional differences in the memory pool are maintained by cell-intrinsic mechanisms or by local tissue environmental signals, as described recently for T_RM cells (4, 5). Likewise, whether the transcription factors and signaling pathways driving memory differentiation are also involved in the maintenance of memory quality remains unexplored. The T-box transcription factors T-bet and Eomesodermin (Eomes) are major regulators of effector function and memory programming. During the immune response, the Eomes/T-bet ratio changes, gradually increasing as T cells transition to memory (6). Eomes is highly expressed in memory T cells and is considered crucial for the maintenance of T_CM quality (7, 8). As such, failure to express Eomes leads to poor development of T_CM function and an inability of memory cells to survive, homeostatically proliferate, and reexpand upon rechallange (7, 9). Upon infection, high levels of inflammation repress Eomes and increase T-bet expression in a process that is dependent on the mechanistic target of rapamycin (mTOR) signaling pathway (10). On the other hand, we have described that weak TCR signals induce strong expression of Eomes and favor T_CM differentiation (11). Together, these results suggest that both TCR and inflammatory signals play an important role in regulating Eomes expression. However, how the TCR regulates Eomes to generate memory T cells and maintain the quality of the memory pool has not been addressed.

TCR stimulation leads to induction of several signaling pathways key for T-cell activation. One of these is NFκB. NFκB signaling has been previously implicated in the maintenance of naïve T cells and generation of memory T cells (12, 13), although the molecular mechanisms regarding how NFκB regulates these processes remain undefined. NFκB signaling is coupled to the TCR through PKCθ (14). PKCθ mediates the activation of the IKK complex or IKKc (composed of IKKα and IKKβ and IKKγ/NEMO), which phosphorylates IκBα, a negative regulator of NFκB signaling, and NFκB. Phosphorylated IκBα is ubiquitinated and degraded by the proteosome. This phenomenon frees NFκB subunits to translocate to the nucleus and mediate transcription of different genes, including IκBα (15).

Here, we investigated how Eomes is regulated during infection and at memory. We found that the NFκB pathway regulates the expression of Eomes and is critical for long-term maintenance of memory T cells and their responses. Our data indicate that this process is cell intrinsic and regulated by proviral integration site for Moloney murine leukemia virus-1 (PIM-1) and TCR signals.

Results

Maintenance of Eomes Expression at Memory Is Not Regulated by IL-15, IL-7, CD27, or OX40.

Memory T cells express high levels of Eomes, but it is unclear whether this elevated expression is the result of extrinsic or intrinsic cell mechanisms. To address this issue, we investigated the role of survival and homeostatic signals that could induce Eomes expression during the memory phase. Self-peptide–MHC–TCR (self-pMHC) interaction is not required for memory CD8 T-cell maintenance (16); however, it is unclear whether tonic signals via the TCR are required for Eomes expression at memory. Thus, we adoptively transferred memory T cells into MHC class I sufficient or deficient lymphopenic hosts (17) and determined Eomes levels. At 28 d posttransfer, we observed no change in Eomes expression regardless of the presence of self-pMHC (Fig. S1A).

Fig. S1.

IL-7, IL-15, self-pMHC, CD27, and OX40 signaling do not regulate Eomes expression in memory T cells. The 2 × 10⁵ naive OT-I cells were adoptively transferred into congenic hosts and challenged with 7 × 10⁶ cfu Att-LM-OVA. After ≥28 d, memory cells were used. (A) The 5 × 10⁵ memory OT-I T cells were isolated and transferred into MHC-sufficient (B6Rag^−/−) or -deficient (B6Rag^−/−b2m^−/−) hosts. After 28 d, Eomes expression was determined by flow cytometry. (B and C) Memory cells were cultured with 0.1 ng/mL IL-7 (lo IL-7) or 20 ng/mL IL-7 + 10 ng/mL IL-15 ± 5 μg/mL agonist antibodies for CD27 and/or OX40 or respective control antibody (B) or 20 ng/mL IL-7 and/or 10 ng/mL IL-15 ± 25 μg/mL IL-7Ra and/or IL-15Rb blocking antibodies or respective control antibody (C). Eomes levels were determined by flow cytometry after 24 h. All graphs show geo mean fluorescence intensity (gMFI) of memory OT-I population minus the respective isotype control gMFI (mean ± SD). All data are representative of two to three independent experiments; n = 3 mice.

Costimulatory molecules of the TNFR family can regulate memory generation and secondary responses (18–20). One of them, CD27, has been previously linked to Eomes expression (21). However, we observed that neither activation of CD27 nor OX40 signaling with agonist antibodies (22, 23) altered Eomes expression in memory T cells (Fig. S1B). In addition, we were unable to detect expression of other TNFR family members previously associated with memory quality, such as 4-1BB and TRAIL, in memory cells (reviewed in ref. 20), suggesting they are most likely not involved in the regulation of Eomes once a T cell is at memory (Fig. S2).

Fig. S2.

Memory CD8 T cells do not express TNF receptors OX40, 41BB, or TRAIL. The 5 × 10⁴ naïve OT-I T cells were adoptively transferred into congenic hosts and challenged with 7 × 10⁶ cfu Att-LM-OVA. After ≥60 d, the expression of OX40, 41BB, and TRAIL were determined in the spleen and lymph nodes by flow cytometry.

We also evaluated the role of IL-7 and IL-15 in the induction of Eomes expression because both are key regulators of memory survival and homeostasis (2). Neither blockade (24, 25) nor addition of exogenous IL-7 and IL-15 to memory cells (26, 27) changed Eomes levels (Fig. S1C). Unexpectedly, these results indicate that the maintenance of Eomes expression in resting memory T cells is not dependent on self-pMHC or TNFR or homeostatic cytokines (9).

NFκB Signaling Is Necessary for Maintaining High Eomes and Bcl-2 Expression in Memory CD8 T Cells.

mTOR signaling regulates both Eomes and T-bet expression upon infection (10). Thus, we tested whether this pathway was involved in maintenance of Eomes levels in memory T cells. For this testing, we used rapamycin, a chemical inhibitor of mTOR signaling, because memory T cells have a slow cell division rate and are not amenable to retroviral transduction. mTOR activity was higher in memory cells than in naïve T cells (Fig. S3A). Nonetheless, treatment of memory CD8 T cells with rapamycin did not alter Eomes or T-bet expression. This finding was in stark contrast to what happened to effector T cells treated in parallel (Fig. S3B) (10). These data suggest that, whereas rapamycin-sensitive mTORC1 regulates Eomes expression to program the generation of memory T cells (28), it does not contribute to support Eomes expression at memory. Interestingly, rapamycin did not inhibit the phosphorylation of mTOR in memory cells (Fig. S3B).

Fig. S3.

mTOR signaling regulates Eomes and T-bet expression in activated but not memory CD8 T cells. The 2 × 10⁵ naïve T cells were transferred into congenic hosts and challenged with 1 × 10⁴ cfu LM-OVA. At 28 d p.i., memory cells were harvested and treated with ±20 μM mTOR inhibitor rapamycin with 0.1 ng/mL IL-7. (A) Phosphorylation of mTOR was determined by flow cytometry in ex vivo memory cells. (B) Expression of Eomes, T-bet, and CD122 and phosphorylation of mTOR were determined by flow cytometry in treated memory or naïve OT-I T cells stimulated with 20 nM OVA, 1 or 2 d posttreatment/stimulation, respectively. Graphs show fold induction over an isotype or naïve (for CD122 only) control (mean ± SD). Histograms are representative of more than two independent experiments. *P ≤ 0.05, **P ≤ 0.005, ***P ≤ 0.001.

NFκB has been shown to bind to the Eomes promoter in in vitro-generated effector T cells (29), but no studies have addressed whether this signaling pathway regulates Eomes during an immune response or at memory. We observed that NFκB activity was present in resting memory T cells as shown by the levels of phosphorylated NFκB at Ser-536 (Fig. 1A) (30, 31). When we treated memory cells with a well-characterized chemical IKKβ/NFκB inhibitor (NFκBi) (32), we found both a decreased level of phospho-NFκB and a severe loss of Eomes expression in memory T cells. Importantly, this finding was not due to overt toxicity, as cell viability was similar between control (vehicle treated) and NFκBi-treated memory cells (Fig. 1A). These results indicate that NFκB signals are necessary to maintain Eomes expression in memory cells.

Fig. 1.

NFκB signaling is required at memory to maintain Eomes and Bcl-2 expression. (A and B) The 2 × 10⁵ naive OT-I cells were transferred into congenic hosts and challenged with 1 × 10⁴ cfu LM-OVA. At ≥28 d postinfection (p.i.) memory cells were harvested and treated with 0.1 ng/mL IL-7 ± 40–100 μM NFκB inhibitor (NFκBi). Phosphorylation of p65-NFκB (S536), Eomes expression, and cell viability (A) and IL-7R, CD122, and Bcl-2 expression (B) were determined by flow cytometry after 24 h. Graphs show either fold induction over an isotype control or geo mean fluorescence intensity (gMFI) of OT-I memory population minus the respective isotype control gMFI. (C–G) OVA-stimulated OT-I T cells were transduced with Eomes RV or empty vector (EV) RV and 1 × 10⁶ GFP⁺ OT-I were transferred into naïve hosts and hosts challenged with 7 × 10⁶ cfu Att-LM-OVA. (C) Frequency of Bcl-2^hiEomes^hi OT-I GFP⁺ cells and (D) Eomes and Bcl-2 gMFI at day 25 p.i. (E) At day 42 p.i., memory T cells were harvested and incubated for 24 h ± 50 μM NFκBi. Graph shows frequency of Bcl-2^hiEomes^hi OT-I GFP⁺ cells after treatment. (F) Density plots and graphs show frequency of OT-I GFP⁺ cells with differential expression of IL-7R and Eomes at day 45 p.i. (G) Number of RV-Eomes or -EV transduced cells in the lymph nodes 25 d p.i. All graphs show mean ± SD. All data are representative of n ≥ 2 independent experiments, n = 3–6 mice per group. *P ≤ 0.05, **P ≤ 0.005, ***P ≤ 0.001, ****P ≤ 0.0001.

Next, we investigated whether NFκB signaling was involved in the regulation of molecules associated with memory survival, an aspect of memory quality that has been linked to the level of Eomes expression (2, 9). Treatment of memory CD8 T cells with the NFκB inhibitor did not affect expression of the receptors for IL-7 or IL-15, discarding a role for NFκB in regulating the input of homeostatic signals associated with memory survival and homeostasis (Fig. 1B).

The expression of the antiapoptotic molecule Bcl-2, however, was significantly reduced in NFκBi-treated cells (Fig. 1B). Interestingly, Eomes-deficient T cells are also impaired in Bcl-2 expression (9). Consistent with this fact, overexpression of Eomes led to an increase in the Bcl-2 levels of T cells that were differentiating to memory (Fig. 1 C and D). We then, addressed whether NFκB signaling regulated Bcl-2 expression independently of Eomes. We found that memory T cells that overexpressed Eomes were refractory to NFκB inhibition and maintained high levels of Bcl-2 (Fig. 1E). Forced expression of Eomes also resulted in an increased frequency of memory cells expressing high levels of IL-7R (Fig. 1F). Furthermore, overexpression of Eomes enhanced the generation of memory T cells (Fig. 1G). Because NFκB signaling regulates Eomes and both Eomes and NFκB control Bcl-2 expression, these data show that NFκB can regulate memory survival signals in an Eomes-dependent manner.

We also tested whether Eomes regulates NFκB activity. We found that T cells overexpressing Eomes strongly induce the transcriptional activity of NFκB compared with their control [empty vector (EV)] counterparts as shown by the NFκB reporter luciferase signal detected on Eomes transduced cells (Fig. S4). Thus, both Eomes and NFκB reciprocally regulate each other in T cells.

Fig. S4.

Eomes regulates NFκB transcriptional activity. NFκB transcriptional activity as measured by luciferase expression was determined by flow cytometry in Eomes-overexpressing T cells. All graphs show mean ± SD. Data are representative of three independent experiments. *P ≤ 0.05, **P ≤ 0.005, ***P ≤ 0.001.

NFκB Signaling Controls Eomes in Activated CD8 T Cells.

T-bet and Eomes work together to regulate CD8 T-cell memory (10). Thus, we examined whether NFκB signals were required to regulate Eomes and T-bet expression in activated T cells. To address this question, we altered NFκB signaling using gain and loss-of-function approaches in proliferating T cells. First, we transduced CD8 T cells with a construct that encodes constitutive active IKKβ (CA-IKKβ) to enhance NFκB signaling (22). CA-IKKβ GFP⁺-transduced cells exhibited lower levels of IκBα (as a consequence of increased proteosomal degradation) than their EV-transduced counterparts, confirming constitutive NFκB signaling (33). Importantly, enhanced IKKβ activity increased Eomes levels and the percentage of T cells expressing Eomes. By contrast, T-bet expression was not altered in cells expressing CA-IKKβ (Fig. 2 A–C).

Fig. 2.

NFκB signaling regulates Eomes expression in activated CD8 T cells. OVA-stimulated OT-I T cells were transduced with CA-IKKβ RV (A–C), DN-p65(trunc) RV (D–F), or the respective EV RV. Eomes, T-bet, and IκBα expression were determined by flow cytometry at day 1–2 posttransduction in OT-I⁺GFP⁺ cells expressing similar levels of GFP. Numbers in histograms are geo mean fluorescence intensity (gMFI) (A and D). Graphs show percentage of GFP⁺ cells with Eomes, T-bet, or IκBα expression over isotype level (B and E). All graphs show mean ± SD. (C and F) Plots show expression of Eomes, T-bet, and IκBα versus GFP level. Arrows show trend of OT-I⁺GFP⁺ population. Data are representative of n ≥ 3 independent experiments. *P ≤ 0.05, **P ≤ 0.005, ***P ≤ 0.001.

Then, we investigated the direct role of NFκB in the regulation of Eomes and T-bet by overexpressing a dominant negative (DN) truncated form of p65-NFκB [DN-p65(trunc)] that selectively inhibits p65-dependent transactivation (34). As expected, we observed lower levels of IκBα (a NFκB target) in DN-p65(trunc)-GFP⁺ transduced T cells (35), indicating NFκB activity was effectively inhibited. Transduction of activated T cells with DN-p65(trunc) also led to a reduction in Eomes expression and a significant loss in the frequency of Eomes expressors. Furthermore, there was a direct correlation between GFP expression and the levels of Eomes and IκBα (Fig. 2 D–F). Similar to the results obtained with the CA-IKKβ construct, no change in T-bet levels occurred when NFκB activity was inhibited. These data demonstrate that NFκB signaling regulates Eomes expression in both activated and memory CD8 T cells.

NFκB Signaling Is Required After the Peak of the Immune Response to Maintain CD8 T-Cell Memory.

Next, we determined whether NFκB signaling regulates Eomes expression and T-cell survival upon infection. The role of NFκB signaling in the early steps of T-cell activation and proliferation of naïve T cells is well established (36). We reasoned that inhibiting NFκB signaling too early in the response would lead to defects in the activation and proliferation of differentiating T cells. These defects would preclude the ability to discern whether NFκB has a role later in the response after T-cell priming (37). To circumvent this issue, OT-I T cells were primed in vivo upon Listeria monocytogenes expressing ovalbumin (LM-OVA) infection. We harvested OT-I CD8 T cells 4 d postinfection and transduced them with either DN-p65(trunc) or EV control without additional TCR stimulation. Then, equal numbers of EV-GFP⁺ or DN-p65(trunc)–GFP⁺ cells were cotransferred into WT-LM (not expressing OVA) infection-matched hosts (Fig. 3 A and B). We observed that both EV- and DN-p65(trunc)–GFP⁺ transduced OT-I T cells underwent contraction after adoptive transfer. However, whereas EV-transduced T cells persisted and generated a memory pool, DN-p65(trunc)-transduced cell frequencies started to decay 2 d posttransfer and were barely detectable at the memory phase (Fig. 3 C and D). This finding was not due to rejection of donor cells, as the ratio of day 8 to day 100 of EV-GFP⁺ OT-I cells (5.53:1) was similar to the ratio of OT-I GFP⁻ cells (5.24:1), but both were different from the ratio of DN-p65(trunc)–GFP⁺ OT-I cells (78.97:1). Importantly, loss of memory cells correlated with a defect in Eomes expression in those cells transduced with DN-p65(trunc) (Fig. 3E).

Fig. 3.

NFκB signaling regulates the maintenance of CD8 T-cell memory. (A–E) The 2 × 10⁵ OT-I T cells were transferred into congenic hosts and challenged with 7 × 10⁶ cfu Att-LM-OVA. Four days p.i., OT-I T cells were isolated and transduced with EV- or DN-p65(trunc) RV with 10 ng/mL IL-7. 1 × 10⁵ EV (CD45.2⁺CD90.1⁻) and 1 × 10⁵ DN-p65(trunc) (CD45.2⁺CD90.1⁺) OT-I⁺GFP⁺ cells were cotransferred (B) into infection-matched hosts challenged with WT LM (CD45.1⁺CD90.1⁻). (C) Kinetics of EV- and p65(trunc)-GFP⁺ OT-I T cells were determined after transfer. Percentage of GFP⁺ OT-I cells in the total OT-I⁺ CD8 T-cell pool is shown. Days represent postinfection. (D) Number of GFP⁺ OT-I T cells at day 100. (E) Expression of Eomes on day 8 in OT-I⁺GFP⁺ cells was determined by flow cytometry. Graph shows fold induction over an isotype control. All graphs show mean ± SD. Data are representative of three independent experiments; n = 3–4 mice. *P ≤ 0.05, **P ≤ 0.005, ***P ≤ 0.001.

These experiments do not discard a role of NFκB signaling in the generation of effector and memory T cells. Rather, the data presented here indicate that a failure in sustaining NFκB signaling after priming, results in impairment in maintaining the survival of CD8 T cells into memory. This conclusion is supported by the fact that control cells reached a plateau after day 30, whereas DN-p65–expressing CD8 T-cell numbers continue declining over time. Thus, altogether the data show that NFκB signaling regulates Eomes expression and maintenance of CD8 T-cell memory.

TCR-Dependent NFκB Signaling Is Required for Eomes Expression.

TNFR family members can induce NFκB signaling (22, 38). However, TNFR CD27, OX40, TRAIL, and 41BB (which have been linked to memory) were either not expressed at the contraction phase or did not have a role in regulating Eomes (Figs. S1 and S2). Our published data, on the other hand, suggest that TCR-dependent NFκB signals mediated by PKCθ are required for CD8 T-cell memory (31, 39). Thus, we hypothesized that TCR-dependent NFκB signaling had a role in the regulation of Eomes. To test this hypothesis, we transduced T cells with a retroviral vector (RV) encoding a kinase dead mutant form of PKCθ (DN-PKCθ) (40), which is one of the TCR-proximal intermediates connecting the TCR to the IKKc/NFκB signalosome (14). The frequency of cells expressing Eomes, but not T-bet, was diminished in DN-PKCθ–GFP⁺ transduced T cells (Fig. 4A), suggesting that PKCθ can regulate Eomes expression in CD8 T cells.

Fig. 4.

TCR-dependent NFκB signaling is required for expression of Eomes. (A) OVA-stimulated OT-I cells were transduced with DN-PKCθ RV or EV RV. Eomes and T-bet expression were determined in OT-I⁺GFP⁺ cells by flow cytometry at day 1 posttransduction. Graphs show percentage of GFP⁺ cells with Eomes or T-bet expression over isotype level. (B and C) The 1 × 10³ WT or MUT OT-I T cells were transferred into congenic hosts and challenged with 1 × 10³ cfu LM-OVA (B) or LM-Q4H7 (C). Kinetics of Eomes and T-bet expression were determined in the blood or spleen by flow cytometry. Graphs show geo mean fluorescence intensity (gMFI). Dashed line indicates the level of Eomes or T-bet for naïve T cells. All graphs show mean ± SD and all data are representative of two or more independent experiments; n ≥ 3 mice per group. *P ≤ 0.05, **P ≤ 0.005, ***P ≤ 0.001.

We also used an OT-I TCR mutant model impaired in triggering NFκB signals to unequivocally test the role of TCR-dependent NFκB signaling in Eomes expression. In this model, T cells carrying a point mutation in the TCRβ transmembane domain (βTMDmut or MUT) are exclusively defective in the transduction of NFκB signals and, importantly, in the generation of memory cells upon infection (39). We adoptively transferred WT or βTMDmut OT-I naïve T cells into congenic hosts and monitored Eomes and T-bet expression after infection with LM-OVA. We found no differences in T-bet expression between WT and MUT CD8 T cells. However, and in agreement with our hypothesis, Eomes expression was impaired in βTMDmut CD8 T cells. Of note, MUT cells were unable to induce Eomes over naïve levels as they matured to memory, a feature that has been associated with the competence of T cells to remain in the memory niche (9) (Fig. 4B). Conversely, in conditions where βTMDmut T cells regain NFκB signaling and memory development (weak TCR signal strength provided by infection with LM-Q4H7) (31), Eomes expression was recovered. As with LM-OVA challenge, there was no difference in T-bet expression (Fig. 4C). Hence, these data indicate that TCR-dependent NFκB signaling is required for sustained expression of Eomes in T cells that transition to memory.

NFκB and Pim-1 Regulate Each Other in CD8 T Cells.

The data in Figs. 1 and 3 suggest that NFκB signals are required after the peak of the response for persistent Eomes expression and survival of CD8 T cells in the memory pool. Antigenic and inflammatory signals quickly decline after the peak of the response (41) and other extracellular signals such as CD27, IL-7, or IL-15 did not affect Eomes expression (Fig. S1). Thus, we hypothesized that the persistence of NFκB signaling after the peak of the response and into memory is programmed by cell-intrinsic mechanisms. Pim-1, a Ser/Thr kinase, is constitutively active once expressed and has been shown to increase p65-mediated NFκB transactivation (42, 43). Most relevant to our studies, Pim-1 is expressed upon TCR stimulation. It is important for T-cell survival and highly expressed in memory T cells (32, 44, 45). Thus, we tested whether Pim-1 was involved in maintaining NFκB signaling in CD8 T cells. Upon CD3/CD28 stimulation, Pim-1 slowly accumulated with time (Fig. 5A). In line with the idea that Pim-1 maintains NFκB signals, we observed substantial late phosphorylation of p65-NFκB [p-p65(S536)] at times where TCR expression and therefore antigen input is low (39), but Pim-1 expression was maximal (Fig. 5A). In contrast, in CD8 T cells with inhibited Pim-1 kinase activity, the phosphorylation of p65-NFκB was ∼50% reduced at all time points. Furthermore, in control cells, IκBα expression returned to basal levels 3 h poststimulation, whereas in the presence of the Pim-1K inhibitor, IκBα levels continued to decrease due to a reduced input in the NFκB feedback loop that regulates IκBα expression (Fig. 5A) (35). These results were confirmed in T cells retrovirally transduced with a vector that encodes a kinase dead form of Pim-1 (DN-Pim-1) (Fig. 5B) (43). Together, these data suggest that Pim-1 regulates the amplitude of NFκB signaling in part by regulating the expression of IκBα in CD8 T cells.

Fig. 5.

TCR-dependent NFκB signaling supports late Pim-1 expression. (A) Naïve OT-I T cells were pretreated with 10 μM Pim-1 kinase inhibitor (Pim-1Ki) or vehicle (DMSO) for 30 min and stimulated with OVA-tetramer⁺αCD28. Levels of proteins shown were determined by immunoblot. Densitometry intensity shown is relative to unstimulated cells and was corrected for loading. Graphs show fold induction over vehicle nonstimulated (t = 0) cells after normalization to loading control. (B) OVA-stimulated OT-I cells were transduced with DN-Pim-1 or EV RV and restimulated with PMA + ionomycin for the indicated times. Expression of IκBα in OT-I⁺GFP⁺ T cells was determined by flow cytometry. (C) Stimulated OT-I T cells were transduced with CA-IKKβ or EV RV. Pim-1 expression was determined in OT-I⁺GFP⁺ cells by flow cytometry at day 2 posttransduction. Graph shows geo mean fluorescence intensity (gMFI) fold induction over isotype level. (D) OVA-stimulated OT-I T cells were transduced with DN-PKCθ RV or control RV. Pim-1 expression was determined in OT-I⁺GFP⁺ cells by flow cytometry at day 2 posttransduction. Graph shows percentage of GFP⁺ OT-I cells with Pim-1 over isotype level. (E) Naïve OT-I and MUT CD8 T cells were stimulated with OVA- or Q4H7-tetramer⁺αCD28. Protein levels shown were determined by immunoblot. Densitometry intensity of Pim-1 relative to unstimulated cells was corrected for loading and is shown below the panel. (F) The 5 × 10³ (or 10⁴ for LM-Q4H7) OT-I and MUT T cells were transferred into congenic hosts and challenged with 7 × 10⁶ cfu Att-LM-OVA (or 10⁴ cfu LM-Q4H7). Pim-1 expression kinetics (shown as fold induction over a naïve control) were determined in the blood by flow cytometry. All graphs show mean ± SD. Dashed line indicates the level of Pim-1 for naïve T cells. Data are representative of two to three independent experiments; n = 4 mice per group. *P ≤ 0.05, **P ≤ 0.005, ***P ≤ 0.001.

Constitutive activation of NFκB signaling up-regulated Pim-1 expression, indicating that NFκB signaling also modulates Pim-1 expression in CD8 T cells (Fig. 5C). This finding agrees with other studies showing that NFκB signaling regulates Pim-1 expression in B cells (46). Overall, these data show that NFκB and Pim-1 regulate each other in CD8 T cells and may function to preserve a continuum of NFκB signaling when antigenic/inflammatory signals are low or not present.

TCR-Dependent NFκB Signaling Regulates Pim-1 Expression.

We then assessed whether TCR-dependent NFκB signaling was involved in the regulation of Pim-1. T cells, retrovirally transduced to overexpress a dead kinase form of PKCθ (DN-PKCθ), were indeed impaired in Pim-1 expression (Fig. 5D). Most importantly, βTMDmut T cells defective in TCR-dependent NFκB signaling and memory development (39) exhibited lower levels of Pim-1 than their WT counterparts from the peak of the response into memory. Conversely, when TCR-dependent NFκB signaling was restored (weak TCR stimulation), Pim-1 expression in βTMDmut T cells was similar to WT levels (Fig. 5E) (31). WT and MUT cells displayed similar levels of Pim-2, suggesting that only Pim-1 is a target of TCR-dependent NFκB regulation (Fig. S5).

Fig. S5.

WT and MUT T cells express similar levels of Pim-2. Naïve OT-I and βTMDmut CD8 T cells were stimulated with OVA- or Q4H7-tetramer and α-CD28 for the indicated times. Pim-2 and α-tubulin levels were determined by immunoblot. Densitometry intensity of Pim-2 relative to unstimulated cells was corrected for loading (α-tubulin) and is shown below the panel. Data are representative of two independent experiments.

Upon infection, Pim-1 levels in differentiating CD8 T cells at the peak of the LM-OVA response were similar to naïve cells and gradually increased as they developed into memory. This finding resembled the profile of Eomes expression at the end of the immune response (Figs. 4B and 5F). MUT T cells, however, failed to increase Pim-1 expression as they entered the memory phase upon LM-OVA but not LM-Q4H7 infection (Fig. 5F). This finding closely matched the defect and recovery of Eomes expression and memory development in MUT cells observed, depending on TCR signal strength (Fig. 4 B and C). These results indicate that TCR signals regulate the increase of Pim-1 levels at the end of the primary immune response. Furthermore, because MUT cells are specifically defective in triggering TCR-dependent NFκB signaling (39), these data suggest that antigenic signals use the NFκB pathway to regulate Pim-1 expression in CD8 T cells.

Pim-1 Kinase Activity Supports Eomes Expression and CD8 T-Cell Memory Fitness.

So far, our data show that sustained NFκB signals after the peak of the response are important for Eomes expression and memory quality. They also show that Pim-1 is an NFκB target capable of regulating NFκB signals. Thus, we reasoned that if Pim-1 supports NFκB, it would have a role in regulating Eomes expression and memory persistence. To test this hypothesis, equal numbers of OT-I T cells transduced with DN-Pim-1 or EV (control) were cotransferred into naïve hosts followed by LM-OVA infection (Fig. 6 A and B). Although T cells containing DN-Pim-1 expanded similarly to control T cells, impaired Pim-1 kinase activity gradually led to a loss of OT-I responders after the peak of the response (Fig. 6B). This loss was especially evident for KLRG1^loIL-7R^hi memory precursors (Fig. 6C). As a result, the number of DN-Pim-1–GFP⁺ memory T cells was severely reduced compared with the number of control memory T cells (Fig. 6D and Fig. S6). The defect in memory generation was especially marked for memory cells with a T_CM phenotype (Fig. 6 D and E). Importantly, DN-Pim-1–transduced OT-I T cells also exhibited impaired Eomes expression throughout the immune response (Fig. 6F).

Fig. 6.

Pim-1 supports Eomes expression and CD8 T-cell memory fitness. OVA-stimulated OT-I T cells were transduced with DN-Pim-1 or EV RV. The 10⁶ congenic EV-GFP⁺ and DN-Pim-1–GFP⁺ OT-I were transferred into the same (A, B, D, F, and G) or independent (C, E, and J–M) naïve hosts. All hosts were challenged with 7 × 10⁶ cfu Att-LM-OVA. (B) Kinetics of OT-I⁺GFP⁺ T cells p.i. were determined in the blood (frequency of total live cells). (C) Kinetics of the frequencies of MPECs (KLRG1^loIL-7R^hi) or SLECs (KLRG1^hiIL-7R^lo) GFP⁺ OT-I were determined p.i. (D) Numbers of GFP⁺ OT-I T cells and T_CM (CD44^hiCD62L^hi) at memory. (E) Frequencies of GFP⁺ OT-I T cells with a T_CM (Left) or a T_EM (Right) memory phenotype at memory. (F) Eomes level (fold induction over isotype control) of OT-I⁺GFP⁺ cells in the blood. Dashed line indicates the level of Eomes for naïve T cells. (G) Expression of IκBα in OT-I⁺GFP⁺ at day 16 p.i. Graph shows percentage of GFP⁺ OT-I cells expressing IκBα over isotype. (H and I) Memory cells were obtained as in Fig. 1 and treated with 10 ng/mL IL-7 ± 100 μM NFκBi (H) or ± 40 μM Pim-1Ki (I). Pim-1 (H) or phospho-NFκB and Eomes (I) levels were determined after 24–48 h. Histograms show geo mean fluorescence intensity (gMFI). All graphs show mean ± SD. (J) Kinetics of expression of Bcl-2, Bcl-xL, Bim, and phospho-Bad are shown in graphs. (K–M) The 2.5 × 10⁵ (K) or 2.5 × 10⁴ (L and M) DN-Pim1 and EV GFP⁺ OT-I memory T cells were transferred in equal numbers into independent congenic hosts and their frequencies (percentage of donor OT-I T cells (KbOVAtet⁺CD8⁺CD45.2⁺) that are GFP⁺) were determined over time (K) or upon infection with 2 × 10⁶ pfu VSV-OVA (L). (M) Graph shows frequencies of OT-I T cells that are GFP⁺ that express both granzyme B and IFN-γ on day 4 post–LM-OVA (5 × 10⁵ cfu) infection after ex vivo restimulation with OVA or VSV peptides. All graphs show mean ± SD. Data are representative of two to three independent experiments; n = 3–10 mice. *P ≤ 0.05, **P ≤ 0.005, ***P ≤ 0.001. NS, nonsignificant. (K) Curves were analyzed by nonlinear regression and compared via F test to determine significance.

Fig. S6.

Gradual loss of DN-Pim-1 cells upon LM-OVA infection is not due to rejection. OVA-stimulated OT-I T cells were transduced with DN-Pim-1 or EV RV. The 1 × 10⁶ congenic EV-GFP⁺ and 1 × 10⁶ DN-Pim-1–GFP⁺ OT-I (1:1 ratio) were cotransferred into the same naïve hosts followed by infection with 7 × 10⁶ cfu Att-LM-OVA. Frequencies of cotransferred congenically marked GFP⁻ OT-I T cells were determined in parallel to frequencies of congenically marked GFP⁺ OT-I T cells (shown in Fig. 6B) postinfection in the blood. Gating strategy is as follows: live cells, CD8+ T cells, CD8⁺Ly5.1⁻ (hosts were Ly5.1⁺), GFP⁺, or GFP⁻ congenically marked OT-I T-cell population.

Pim-1 kinase defective T cells were unable to maintain NFκB signaling in vivo, as indicated by their inability to induce the expression of the NFκB target IκBα (Fig. 6G). This finding is consistent with the idea that Pim-1 regulates NFκB late in the primary immune response. Indeed, at memory, NFκB inhibition led to a marked defect in Pim-1 expression (Fig. 6H) and, vice versa, Pim-1 inhibition resulted in memory T cells exhibiting lower levels of p-NFκB and Eomes expression (Fig. 6I) (9). T cells with impaired Pim-1 kinase activity showed a defect in the expression of prosurvival factors Bcl-2 and Bcl-xL after day 20 postinfection (Fig. 6J). This finding indicates that Pim-1, similar to NFκB (Figs. 1B and 3), plays a role in the survival of T cells differentiating into memory. We next evaluated the ability of DN-Pim-1–GFP⁺ memory T cells to persist and to respond to reinfection. Memory T cells with deficient Pim-1 activity showed a much faster decay than their EV-GFP⁺ control counterparts (Fig. 6K) and were remarkably impaired in their reexpansion and expression of IFN-γ and granzyme B upon reinfection (Fig. 6 L and M). We obtained similar results in analogous experiments performed with memory T cells treated with NFκB inhibitor for 24 h (Fig. S7). Collectively, these data indicate that NFκB and Pim-1 modulate each other in an immune response and at memory and together contribute to maintain memory fitness through the regulation of Eomes expression (Fig. S8).

Fig. S7.

Inhibition of NFκB signaling at memory severely impacts survival and function of CD8 memory T cells. Naïve congenic hosts were adoptively transferred with 5 × 10⁵ (A and C), 9 × 10⁵ (B), or 7.5 × 10⁴ (D–H) NFκBi- or vehicle-treated memory T cells. (A) Numbers of NFκBi- or vehicle-treated memory T cells 4 d posttransfer. (B) Graph shows frequencies of congenic donor OT-I memory T cells over time after transfer. Naïve hosts containing control (DMSO treated) or NFκBi memory T cells were infected with 1 × 10⁴ (C) or 4 × 10⁵ (D and E) cfu LM-OVA. (C) Graphs show total number or fold expansion of OT-I memory cells at day 4 of the secondary response. (D) Reexpansion of OT-I cells post–LM-OVA infection. LOD, level of detection. (E) At 4 d p.i., OT-I splenocytes were harvested and cultured in the presence or absence of VSV or OVA peptides with brefeldin A for 5 h. Graph shows percentage of OT-I T cells of the total CD8 that doubly express IFN-γ and granzyme B upon ex vivo restimulation. (F) LM-OVA titers at day 4 p.i. (G and H) Hosts containing control or NFκBi memory T cells were infected with 2 × 10⁶ pfu VSV-OVA. Graphs show donor OT-I memory T-cell frequency of total CD8 T cells over time p.i. (G) and frequencies of donor memory OT-I that are granzyme B and IFN-γ double positive at 4 d p.i. upon restimulation with OVA, VSV, no peptide (NS) (H). All data are representative of two to three independent experiments; n = 3 mice. *P ≤ 0.05, **P ≤ 0.005, ***P ≤ 0.001.

Fig. S8.

Model. At the beginning of the immune response, TCR signaling induces NFκB. NFκB signaling is required after the peak of the response for the survival of T cells that mature to memory, for the longevity of memory T cells and for their response. Memory survival and function are supported via NFκB by the expression of Eomes, Pim-1, and Bcl-2. Reciprocal regulation of Pim-1 and NFκB maintains the level of NFκB signaling required to enable Eomes expression.

Discussion

Several reports have previously linked different members of the canonical and noncanonical NFκB pathway with T-cell memory (12, 13, 47). Our results confirmed this role and show how NFκB signaling regulates the generation of memory T cells in the context of infection. From our data, we conclude that NFκB signaling is required after the peak of the response to support memory. Because we deliberately impaired NFκB signals after T-cell priming, the loss of memory cells most likely indicates that NFκB signals are involved in preserving the survival of T cells that enter the memory pool. Because NFκB regulates Eomes and Eomes in turn controls Bcl-2 expression, we conclude that memory survival is in part due to NFκB-dependent regulation of Bcl-2 expression through Eomes and Pim-1. Importantly, our results also indicate that NFκB–Pim-1 activity contributes to maintaining high Eomes levels in resting memory cells and is required for reexpansion and rapid memory function upon antigen reencounter. This idea is in agreement with the phenotype of Eomes-deficient CD8 T cells (7, 9). Together, our data strongly suggest that NFκB signals play a critical role in maintaining the longevity and the function of memory T cells.

Signaling pathways mTOR and Wnt/TCF-1 have been shown to regulate Eomes expression and CD8 T-cell memory development in a cytokine-dependent manner (10, 48). Unexpectedly, the data presented here show that TCR signaling uses the NFκB pathway to program Eomes expression and CD8 T-cell memory fitness. It is possible that mTOR, Wnt, and NFκB pathways cooperate, such that a change in the activity of one of them affects Eomes expression. Alternatively, it may be that each of these signaling pathways regulates Eomes levels at different times in the CD8 T-cell differentiation process. For example, we found that rapamycin-sensitive mTOR signaling controls Eomes expression in activated CD8 T cells but not at memory. In contrast, NFκB signaling regulates Eomes expression in both activated and memory CD8 T cells. How these pathways cooperate to regulate T-cell memory is an interesting area of study that warrants further investigation.

Our study proposes that the reciprocal regulation of Pim-1 and NFκB maintains the level of NFκB signaling required to enable Eomes-dependent survival of CD8 T cells as they progress to memory and at memory. This is supported by the fact that inhibition of either NFκB or Pim-1 leads to defects in the expression of the other during the late phase of the primary immune response and at memory. What controls the activity of Pim-1 and NFκB late in the immune response? Pim-1 is involved in the survival of T cells via CD27 in an mTOR-independent manner (44), and CD27–CD70 signals have a role late in the primary response in the programming of CD8 T-cell memory (19). However, and without discarding the role of Pim-1 in survival upon CD27 signaling, our data strongly suggest that, at least upon Listeria infection, Pim-1 regulates Eomes in a CD27-independent manner.

Pim-1 expression can also be regulated by STAT-3 and STAT-5 signaling, which would suggest a role for cytokine signaling in the expression of Pim-1 and its downstream target Eomes (49). However, we were unable to observe any Eomes dependence on IL-7 or IL-15 signaling (STAT-5 mediated) at memory. Furthermore, MUT cells, which are impaired in Pim-1 and Eomes expression, exhibit normal levels of STAT-5 and STAT-4 activation (STAT-4 phosphorylation is enhanced in STAT-3–deficient T cells; ref. 50) (31). Rather, the data from the memory-defective βTMDmut model, where TCR-dependent NFκB signaling is impaired together with late expression of Pim-1 and Eomes, suggest that TCR signals (although maybe not exclusively) are involved in regulating Pim-1 expression late (after the peak of the response and at memory) in an NFκB-dependent manner. It is unclear how NFκB regulates Pim-1 expression. This regulation may happen at a biochemical level. Gradual increase of Pim-1 in the memory phase could be related to induced transcription, reduced degradation, or posttranscriptional mRNA stability (51). Future research beyond the scope of this study will be necessary to address these possibilities.

Eomes expression appears to have a higher impact on T_CM development, homeostasis, and function than on any other T-cell memory subset (9). Curiously, the absence of NFκB and Pim-1 signaling has a more profound effect on memory maintenance than what would be anticipated by only the specific loss of T_CM. Thus, it is likely that Eomes has a distinct role beyond the peak of the response to enable general memory fitness. Such a role is supported by the dramatic loss of memory cells observed after day 45 post-LCMV infection in Eomes-deficient mice (52) and by the fact that T cells appear to become Eomes “addicted” for their survival once at memory (7). This role of Eomes at memory does not exclude an additional role of Eomes in programming T_CM development. This role could be related to Pim-1, as both deficiency in Eomes and Pim-1 activity does not affect the generation of memory precursor effector cells (MPECs) [or skew the differentiation toward short lived effector cells (SLEC)] but rather results in a reduction in the frequency of memory precursors beyond the peak of the response. Interestingly, both Eomes and Pim-1 kinase deficiencies have a more pronounced effect in cells of T_CM phenotype (9). Nonetheless, our findings show that maintaining NFκB signals and Pim-1 expression is critical for supporting T-cell memory quality (maintenance and recall potential). These data indicate that, in addition to the local tissue environment signals, cell-intrinsic mechanisms are crucial to determine the quality of memory T cells.

In conclusion, the data presented here show that TCR/antigenic signals regulate the induction of a NFκB/Pim-1 feedback loop that controls expression of Eomes and thereby long-term maintenance of the memory pool. Our study identifies two molecular targets that can be of aid in the design of better vaccine strategies and tumor immunotherapies.

Materials and Methods

Mice and Reagents.

C57BL/6, C57BL/6 Rag^−/−, B6.SJL, OT-I, and βTMDmut (MUT) TCR transgenic mice (39), and B6Rag^−/−β2m^−/− were bred and maintained in accordance with University of Missouri Office of Animal Resources Animal Care and Use Committee. Experimental procedures were approved by the University of Missouri Institutional Animal Care and Use Committee. OVA, Q4H7, and VSV peptides were from New England Peptides. rmIL-7 and rmIL-15 were from Peprotech. IL-2 (X63-IL-2 hybridoma) was used at 50 units/mL Wedelolactone (NFκB inhibitor); rapamycin, Pim-1 kinase inhibitor IV, and Pim-1/2 kinase inhibitor V were from EMD/Calbiochem.

Antibodies and Flow Cytometry.

Anti-CD8α (53-6.7), -CD90.1 (OX-7), –IL-7Rα (A7R34), -CD62L (MEL-14), -CD28 (37.51), -CD122 (TMβ1), IFN-γ (XMG1.2) and -Bcl-2 (3F11) were from BD Pharmingen. Anti-CD45.1 (A20), -CD45.2 (I04), -CD27 (LG.7F9 or LG.3A10), -OX40 (OX-86), KLRG1 (2F1), TRAIL and -Eomes (Dan11Mag) were from eBioscience. Anti–T-bet (4B10) and –Pim-1 (12H8 and C20) were from Santa Cruz Biotechnology. Anti-IκBα (5A5 and L35A5), –p65-NFκB (D14E12), –phospho-p65-NFκB (S536), -Bim (C34C5), –phospho-BAD (40A9) and –phospho-mTOR (S2481) were from Cell Signaling. Anti–α-tubulin was from Sigma. Secondary antibodies and anti-granzyme B (GB11) were from Invitrogen. Anti–Bcl-XL (7B2.5) was from Southern Biotech. Flow cytometry was performed on a Coulter Cyan ADP or a FACSCalibur or LSRII Fortessa flow cytometer (Becton Dickinson) and analyzed with FlowJo FACS Analysis Software (Tree Star).

Adoptive Transfer.

Naïve (1 × 10³–5 × 10⁴) or memory CD8 T cells were purified from the lymph nodes of OT-I, MUT, or host mice and transferred i.v. into congenic mice.

Bacteria and Viral Infections.

Listeria strains were generously provided by M. Bevan, University of Washington, Seattle (LM-OVA and LM-Q4H7) and Sing Sing Way, University of Minnesota, Minneapolis (Att-LM-OVA). Att-LM-OVA strains were grown to an OD₆₀₀ of 0.1 as in ref. 39, whereas LM-OVA and LM-Q4H7 strains were grown as described in ref. 11. All infections were performed i.v. at least 1 d after adoptive transfer of naïve transgenic T cells.

OVA-expressing vesicular stomatitis virus (VSV) was provided by L. Lefrançois, University of Connecticut Health Center, Farmington, CT and used as described previously (53).

Ex Vivo T-Cell Function.

T cells at different times of the secondary immune response were harvested from spleens of host mice and stimulated with cognate (OVA) or control (VSV) null peptides in the presence of Brefeldin A (BD Biosciences) as described previously (11, 54). Responding T cells were identified with a gating strategy that included Kb-OVA tetramers, CD8, and congenic markers.

Listeria Titer Assays.

Listeria numbers in the spleens of infected mice were determined as described previously (31).

Retroviral Transduction.

Retroviral particles were generated using the Platinum-E Retroviral Packaging Cell Line (Cell Biolabs) and Genejammer (Agilent Technologies) according to the manufacturer’s instructions. Spin transduction was performed two ways. For transduction of in vitro-activated CD8 T cells, 24 h after stimulation with 20 nM OVA, OT-I T cells were plated in 96-well plates (5 × 10⁵ cells per well). A total of 100–200 μL viral supernatant with 8 μg/mL Polybrene (Millipore) and IL-2 was added, and plates were spun at 575 × g for 75 min at 32 °C. After a 4-h incubation at 32 °C, cells were resuspended in media with 20 nM peptide and IL-2 and incubated at 37 °C overnight. The second spin transduction was performed 24 h later and cells were resuspended in IL-2–containing media. For transduction of in vivo-primed CD8 T cells, OT-I T cells were harvested 4–5 d after challenge. OT-I T cells were plated in 96-well plates (5 × 10⁵ per well). Two spin transductions were performed by spinning plates at 1,800 rpm for 75 min at 32 °C after adding 100–200 μL viral supernatant with 8 μg/mL Polybrene (Invitrogen) and 10 ng/mL IL-7. After a 4-h incubation at 32 °C, cells were resuspended in media with 10 ng/mL IL-7 and incubated at 37 °C overnight. In both conditions, flow cytometry was used to calculate transduction efficiency to ensure that equal number of congenically marked OT-I GFP⁺ cells transduced with the retroviral construct or empty vector were adoptively transferred 24 h after the last transduction.

For retroviral transduction of Jurkat T cells, the Platinum GP Packaging cell line was used, and Eomes was cotransduced together with a construct that encodes VSV-G (55).

NFκB Activity Luciferase Assay.

Retroviral particles containing Eomes or EV RV control and VSV-G were produced using Plat-GP cells (Cell Biolabs) and Lipofectamine 3000. Jurkat T cells were retrovirally transduced and RV-Eomes GFP⁺ or control RV-EV GFP⁺ sorted in a BC MoFlow XDP. Next, RV-Eomes or EV control expressing T cells were transfected with pGL3-NFκB luciferase reporter (56) construct using Trans-IT Jurkat (Mirus) reagent. Luciferase expression was determined by flow cytometry with a biotin-conjugated anti-luciferase antibody (Rockland) followed by streptavidin-PE staining.

Enrichment of CD8 T Cells.

Before transfer into recipient mice, OT-I CD8 T-cell donors were enriched by positive or negative selection using magnetic beads (Miltenyi Biotec).

Immunoblot Analysis.

The 2–5 × 10⁶ T cells were stimulated with OVA- or Q4H7-tetramer and anti-CD28 subjected to immunoblot as described in ref. 17. Densitometry was determined using Photoshop CC2015.

Statistical Analysis.

For statistical analysis, two-tailed unpaired Student’s t test was applied using GraphPad Prism software. Significance was set at P < 0.05. *P ≤ 0.05, **P ≤ 0.005, ***P ≤ 0.001.