The Antitumor Effects of Vaccine-Activated CD8⁺ T Cells Associate with Weak TCR Signaling and Induction of Stem-Like Memory T Cells

The success of using checkpoint blockade to unleash antitumor immunity (1) confirms that activation of the immune system can control cancer growth. Intensive efforts by vaccine researchers have yielded a variety of immunization approaches that can elicit tumor-specific immunity (2). Despite induction of tumor-specific T cells, the antitumor effect of cancer vaccines remains disappointing (3). Evidence from both animal models (4, 5) and human trials (6) suggests that the presence of tumor antigen–specific T cells does not necessarily translate into antitumor effects. The limited antitumor effect of tumor-specific T cells has been attributed to immune suppression (7) as well as to T-effector cell (Teff) exhaustion (8, 9). However, the lack of correlation between the frequency of antigen-specific T cells and antitumor effects (10–12) also suggests that intrinsic cell attributes may play a role in determining their antitumor outcome (13). Although high “functional avidity” of antigen-specific T cells might be a measure of antitumor efficacy (14–16), contrasting evidence shows that high avidity T cells are associated with weaker antitumor effects (17, 18). Indeed, in chronic infections (19, 20), the competitiveness of high-avidity T cells comes at a cost, i.e., overactivation, exhaustion, and apoptosis of Teff cells in the presence of persistent antigen stimulation, a scenario that also exists in tumor lesions. Thus, the attributes of CD8⁺ T cells and mechanisms underpinning the antitumor efficacy of tumor-specific immunity remain unclear and require in-depth investigation in order to improve the efficacy of current immunotherapies.

Lentivector (lv) activates CD8⁺ T-cell responses (21, 22) because of its transduction of dendritic cells (23). We found that immunization with lv-expressing epitope-optimized α-fetoprotein (opt-AFP) activated CD8⁺ cells to prevent carcinogen-induced autochthonous hepatocellular carcinoma (HCC) in mice (24). But it is not clear whether different epitope-specific CD8⁺ cells all have antitumor effect. In the current study, we investigated the antitumor effects of two AFP epitope-specific CD8⁺ cells that could be reproducibly induced by lv immunization. We found that epitope AFP₂₁₂- and AFP₄₉₉-specific CD8⁺ cells (hereafter as Tet₂₁₂ and Tet₄₉₉) generated different antitumor effects. We thus studied the molecular and cellular attributes of vaccine-activated T cells that are associated with their antitumor effects. We found that Tet₄₉₉ were insensitive to antigen stimulation and survived persistent in vitro antigen restimulation and maintained their effector functions. In contrast, Tet₂₁₂ were sensitive to antigen restimulation, resulting in more expansion at low doses of antigen, but became exhausted and apoptotic in the presence of high and persistent antigen. In vivo, Tet₄₉₉ generated more expansion than Tet₂₁₂ upon reencountering antigen. The different antigen responsiveness of Tet₂₁₂ and Tet₄₉₉ cells correlated to their different TCR signaling strength and the presence of stem-like memory T cells (Tscm), which had robust in vivo expansion. The TCR signaling on Tet₄₉₉ was weaker, associated with severe TCR downregulation. Thus, TCR downregulation and weakened TCR signaling likely prevent Tet₄₉₉ from overactivation and enhances Tscm responses in the priming phase, but also protect Teff from exhaustion and apoptosis at effector phase. Our data suggest that further investigation into the molecular mechanisms of how to control TCR downregulation and TCR signaling strength may help design cancer vaccines to elicit a diverse and healthy mix of tumor-specific CD8⁺ T cells, especially Tscm, to generate antitumor effects.

C57BL/6 (CD45.2 and CD45.1) and C3H mice were from Charles River. Nur77^GFP mice (25) were from The Jackson Laboratory. Mice were bred and maintained in specific pathogen-free facility at Augusta University. Animal protocols were approved by the Institutional Animal Care and Use Committee.

The PE-labeled H-2D^b/AFP₂₁₂ and PE- or APC-labeled H-2K^b-D^b/AFP₄₉₉ tetramers were prepared by NIH Tetramer Core Facility. In the H-2K^b-D^b/AFP499 tetramer, the α3 domain of H-2K^b is replaced with α3 domain of H-2D^b to reduce nonspecific binding. The AFP₂₁₂ and AFP₄₉₉ peptides bind to H-2D^b and H-2K^b, respectively. Wild-type AFP peptides were used to prepare tetramers.

EL4 and 293T cells were purchased from ATCC in 2010. Cells received from ATCC were immediately expanded and stored in liquid nitrogen. Each time, one vial of cells were thawed and used for less than 6 passages to maintain their authenticity. EL4-AFP tumor cells were established previously (24) by transducing parental EL4 cells with lv expressing mouse AFP. Cell lines were checked for mycoplasma by PCR test (Fisher Scientific). For tumor challenge, 1 × 10⁵ EL4-AFP cells were injected subcutaneously into the flank of C57BL/6 mice. The induction of autochthonous HCC with diethylnitrosamine (DEN; Sigma) was conducted as described (24). Briefly, 2-week-old F1 mice of B6XC3H cross-bred were intraperitoneally injected with 50 μg of DEN per gram of mouse weight.

The plasmid expressing epitope-optimized mouse AFP (opt-AFP) was described (24). The shorter AFP fragments of AFP142 and AFP164 were cloned into pLenti6 (Invitrogen) by PCR cloning. The lvs were prepared by transient cotransfection of 293T cells, and the vectors were concentrated and titered by measuring the p24 level as described (21).

To construct recombinant vv expressing the opt-AFP, a shuttle plasmid vector pG10 was used (26). The opt-AFP gene was cloned into the vector behind the p7.5 early gene promoter to create pG10-opt-AFP. CV-1 cells in 6-well plates were infected with a wild-type vaccinia virus of WR strain at multiplicity of infection of 0.1, and then transfected with pG10-opt-AFP by SuperFect reagent (Qiagen). The recombinant vv was selected in human TK-143 cells with addition of BrdUrd in the medium. After three rounds of plaque purification, the purity of the virus was verified by PCR assays for the presence of the transgene and deletion of the viral thymidine kinase gene, and by fluorescence of DsRED in the infected cells. The virus, designated as opt-AFP-vv, was amplified in HeLa cells and purified by a standard procedure (26).

For immunization, 2 × 10⁷ transduction units of opt-AFP-lv were injected via footpad. To boost immune responses, 1.5 × 10⁷ infectious units of recombinant opt-AFP-vv were injected intraperitoneally. For immunization in the autochthonous HCC model, 2-month-old F1 mice of B6XC3H that had been induced by carcinogen DEN were immunized with opt-AFP-lv and then boosted with opt-AFP-vv at 3 months.

The splenocytes (6 million) were restimulated for the indicated time with various concentration of wild-type AFP₂₁₂ (GSMLNEHVC) or AFP₄₉₉ (SSYSNRRL) peptide in the presence of 20 IU/mL IL-2 (ProSpec-Tany Technogene). The wild-type AFP peptides were used to make sure that the restimulated T cells would recognize and kill tumor cells expressing wild-type AFP. In some experiments, 10 ng/mL of IL-7 and IL-15 were also added. To maintain the peptide level, half of the RPMI medium containing the indicated concentrations of peptides was changed every other day.

In the second setting, the splenic CD8 T cells were purified by magnetic beads from immunized CD45.2 mice. The splenocytes of naïve CD45.1 congenic mice were pulsed with indicated concentrations of AFP₂₁₂ or AFP₄₉₉ peptides for 2 hours. Free peptides were washed away. Two millions of the purified CD8 T cells were then restimulated with four millions of peptide-loaded CD45.1 splenocytes in a 24-well plate in RPMI media containing 20 IU/mL IL-2. After 48 hours, 4 million peptide-loaded CD45.1 fresh splenocytes were added to continue stimulation for another 2 days before analysis.

Peripheral mouse blood cells and the splenocytes (fresh or stimulated) were stained with indicated combinations of antibodies plus H-2D^b/AFP₂₁₂ or H-2K^b-D^b/AFP₄₉₉ tetramers. For tetramer staining after restimulation with peptide in the media, the stimulated cells need 12 hours rest to recover the TCR. DAPI was added to exclude dead cells. Intracellular staining of IFNγ was conducted as described (24) after the cells were restimulated with peptides in the presence of GolgiStop for 3.5 hours. Events were collected on the LSR II, and the data were analyzed using FCS Express III software. The antibodies used in this study include antibodies to mouse CD8 (Clone: 53-6.7, Biolegend), CD45.1 (Clone: A20, Biolegend), CD45.2 (Clone: 104, BD Bioscience), CD44 (Clone: 1M7, Biolegend), CD62L (Clone: MEL-14, Biolegend), CD127 (IL-7Ra; Clone: A7R34, Biolegend), PD-1 (Clone: 29F.1A12, Biolegend), CCR7 (Clone: 4B12, Biolegend), CD122 (IL-2Rβ; Clone: TM-β1, Biolegend), Sca-1 (Clone: D7, Biolegend,), CD95 (Clone: SA367H8, Biolegend), IFN-γ (Clone: XMG1.2, Biolegend), Bcl-2 (Clone: BCL/10C4, Biolegend), H-2^K b/H-2D^b (Clone: 28-8-6, Biolegend), and TCRβ (Clone: H57-597, BD Bioscience). The Annexin-V and DAPI (4,6-diamidino-2 phenylinole dilactate) were from Biolegend.

The in vitro CTL assay was performed as previously described (21, 24). The splenocytes from immunized mice were restimulated with 2 μg/mL of AFP₂₁₂ or AFP₄₉₉ peptide for indicated times in the RPMI media containing 20 units/mL of IL-2. The EL4 and EL4-AFP tumor cells were labeled with 1 μmol/L and 0.03 μmol/L of 5-(and 6)-carboxyfluorescein diacetate succinimidyl ester (CFSE; Invitrogen), respectively, for 10 minutes at 37°C. After washing, equal numbers (5 × 10⁴) of CFSE^lo-labeled target EL4-AFP cells and CFSE^hi-labeled EL4 control cells were cocultured in triplicate with the in vitro peptide-stimulated splenocytes at the indicated E/T ratios for 6 hours. The killing of target EL4-AFP cells was analyzed by flow cytometry. The specific killing activity was calculated using the formula [1-(ratio of CFSE^hi/CFSE^lo in the absence of CTL)/(ratio of CFSE^hi/CFSE^lo in the presence of CTL)] × 100, as previously described (21).

The splenic CD8 T cells were enriched by magnetic beads (Stemcell Technologies) and then stained with anti-CD8, CD44, and CD62L antibodies plus H-2D^b/mAFP₂₁₂ or H-2K^b-D^b/mAFP499 tetramer. The tetramer⁺ cells and their subsets were sorted on FACSAria (BD Biosciences). The sorted cells of indicated number were injected into CD45.1 congenic mice, which were then immunized with opt-AFP-lv. The CD45.2⁺CD8⁺Tet⁺ cells in the peripheral blood of CD45.1 mice were monitored once a week by tetramer staining.

The tetramer binding and dissociation assays were conducted as described (27). For tetramer binding assay, 1 million splenocytes of the immunized mice were stained with anti-CD8 antibody and the indicated concentrations of H-2D^b/AFP₂₁₂ or H-2K^b-D^b/AFP₄₉₉ tetramer. Between the tetramer concentrations of 6.5 and 13 μg/mL, the percentage of Tet₂₁₂ and Tet₄₉₉ cells did not increase further, suggesting that all the tetramer⁺ cells were stained. The Tet₂₁₂ and Tet₄₉₉ cells detected at lower tetramer concentration were compared with those detected at the concentration of 13 μg/mL, which was considered as 100%.

For the tetramer dissociation assay, the cells were stained with 6.5 μg/mL of the tetramer. After washing away the free tetramers, the cells were left at room temperature for the indicated times in the presence of anti-H-2K^b/H-2D^b antibody before they were analyzed for the percent and MFI of the Tet⁺ cells.

Statistical analyses were performed using t test or ANOVA (GraphPad Inc.).

To study whether different AFP epitope–specific CD8⁺ T cells could kill AFP⁺ tumor cells, we restimulated the splenocytes with AFP₂₁₂ and AFP₄₉₉ peptide in media and used them as effectors for in vitro CTL assays (Fig. 1A). After 6-day restimulation, splenocytes of naïve and wild-type-AFP-lv–immunized mice showed no CTL activity (Fig. 1B), consistent with our recent report (24). On the other hand, splenocytes from the opt-AFP-lv–immunized mice had CTL activity after AFP₄₉₉ peptide restimulation (Fig. 1B and Supplementary Fig. S1A and S1B). However, after AFP₂₁₂ peptide restimulation, splenocytes of the opt-AFP-lv–immunized mice generated no CTL activity. Kinetic study demonstrated that the CTL activity of opt-AFP-lv–immunized splenocytes was diminished with AFP₂₁₂ peptide, but was enhanced with AFP₄₉₉ peptide, restimulation (Supplementary Fig. S1C).

To study whether Tet₂₁₂ and Tet₄₉₉ generated different antitumor effects in vivo, we constructed two lvs, AFP142-lv and AFP164-lv, to activate Tet₂₁₂ and Tet₄₉₉ separately (Fig. 1C). We found that, like opt-AFP-lv, AFP164-lv immunization protected mice from AFP⁺ tumor challenge. In contrast, all mice immunized with AFP142-lv developed tumors (Fig. 1D). The lack of antitumor effect by Tet₂₁₂ was not due to its incapability of recognizing AFP⁺ tumor cells because both Tet₂₁₂ and Tet₄₉₉ cells produced IFNγ after coculture with EL4-AFP cells (Supplementary Fig. S1D). Together, these data suggest that Tet₂₁₂ and Tet₄₉₉ cells have different responses to tumor antigen in vitro and in vivo, resulting in different antitumor effects.

To investigate why Tet₂₁₂ and Tet₄₉₉ generate such different CTL activities and antitumor effects, we studied their antigen responsiveness in vitro. To generate more Tet₂₁₂ and Tet₄₉₉ cells, we boosted the lv-immunized mice with vv. The purified CD8⁺ cells from immunized CD45.2 mice were restimulated with naïve CD45.1 splenocytes pulsed with different peptides. After 4 days, we observed a dose-dependent increase of Tet₂₁₂ cells between the pulsed peptide ranges of 0.1–2 μg/mL (Fig. 2A). However, at 10 μg/mL of AFP₂₁₂ peptide, the percentage of Tet₂₁₂ among total CD45.2 CD8⁺ cells began to decline. Furthermore, restimulation with 2 μg/mL of AFP₂₁₂ peptide in media reduced Tet₂₁₂ to ∼1% (Fig. 2A). In contrast, the percentage of Tet₄₉₉ increased only slightly between the pulsed AFP₄₉₉ peptide ranges of 0.1 to 10 μg/mL and increased more so after restimulation with AFP₄₉₉ peptide in media (Fig. 2A). In addition, we analyzed the PD1 and Annexin-V level on Tet₂₁₂ and Tet₄₉₉ after stimulation. We found that, compared with Tet₄₉₉, the percentage of PD1⁺ Tet₂₁₂ and the MFI of PD1 on Tet₂₁₂ was higher at every dose point (Fig. 2B). Similarly, the percentage and MFI of Annexin-V were also higher on Tet₂₁₂ (Fig. 2C). The production of cytokine IFNγ by Tet₂₁₂ cells was inversely related to AFP₂₁₂ peptide dose (Fig. 2D and E). In contrast, the percentage and MFI of IFNγ⁺ Tet₄₉₉ cells were directly related to the dose of AFP₄₉₉ peptide.

The exhaustion and apoptosis of Tet₂₁₂ and Tet₄₉₉ cells were also studied when peptides were added directly to the restimulation media (Supplementary Fig. S2A). The data showed ∼10% of Tet₂₁₂ and Tet₄₉₉ cells being apoptotic prior to restimulation. AFP₂₁₂ peptide restimulation caused Tet₂₁₂ undergo dose-dependent apoptosis, ∼70% of Tet₂₁₂ cell being apoptotic with 2 μg/mL of peptide stimulation (Supplementary Fig. S2B and S2C). In contrast, only ∼17% of Tet₄₉₉ cells were apoptotic after AFP₄₉₉ peptide restimulation. Consistent with the data, only 5% of Tet₂₁₂ cells, but 80% of Tet₄₉₉ cells, survived 4-day peptide restimulation (Supplementary Fig. S2D). Again, more Tet₂₁₂ expressed higher level of PD1 in a dose-dependent manner (Supplementary Fig. S2E). Furthermore, Tet₂₁₂ lost their cytokine production and CTL activity after restimulation (Supplementary Fig. S2F). In contrast, the majority of Tet₄₉₉ was not exhausted and maintained their effector function. The exhaustion and apoptosis were different from activation-induced anergy (28) as it could not be rescued by the addition of IL2 or even IL7 and IL15.

In summary, Tet₂₁₂ have a low threshold of reactivation and can be expanded by low-dose antigen. But its high sensitivity to antigen restimulation leads to progressive upregulation of PD1 and reciprocal decrease of IFNγ and apoptosis with increasing dose of antigen. In contrast, Tet₄₉₉ have a high reactivation threshold and thus is difficult to activate and expand. But the Tet₄₉₉′s insensitivity to antigen also protects them from exhaustion and activation-induced apoptosis. As a result, persistent antigen restimulation (such as peptide in media) decreased Tet₂₁₂ but increased Tet₄₉₉ cells.

We then studied in vivo reexpansion of Tet₂₁₂ and Tet₄₉₉ upon reencountering antigen. Tet₂₁₂ and Tet₄₉₉ cells were sorted from immunized CD45.2 mice and transferred into CD45.1 congenic mice followed by immunization (Fig. 3A). The data showed a significant more increase of Tet₄₉₉ than Tet₂₁₂ cells in the CD45.1 mice (Fig. 3B). In addition, the carcinogen-induced autochthonous HCC model was utilized to study in vivo antigen responsiveness of Tet₂₁₂ and Tet₄₉₉ cells to emerging AFP tumor antigen (Fig. 3C). We first found that the magnitude of Tet₄₉₉ responses after vv boost was greater than that of Tet₂₁₂ (Fig. 3D), suggesting Tet₄₉₉ cells primed by lv underwent more extensive expansion when reencountering antigen expressed from vv. Tet₄₉₉ cells were also able to detect emerging AFP antigen in autochthonous HCC and responded by expansion (the spike of Tet₄₉₉ at day 140 in Fig. 3D). In contrast, Tet₂₁₂ cells did not generate expansion by vv boost and could not generate productive response to emerging AFP antigen.

We next studied the possible mechanisms underpinning the different antigen responsiveness of Tet₂₁₂ and Tet₄₉₉ cells. Memory T cells, especially Tscm and Tcm, are critical for generating antitumor effect as they can undergo expansion in response to tumor antigen (29–31). To study if the antigen responsiveness and antitumor effect of Tet₂₁₂ and Tet₄₉₉ cells were related to Tscm and Tcm, we stained Tet₂₁₂ and Tet₄₉₉ with CD44 and CD62L. As demonstrated in Fig. 4A, the Tet⁺ CD8⁺ cells could be grouped into CD44⁺CD62L⁻ Teff (or effector memory, hereafter referred to as Teff), CD44⁺CD62L⁺ Tcm, and the CD44⁻CD62L⁺ naïve-like Tscm. We found that Tet₄₉₉ contained more Tscm and Tcm than Tet₂₁₂ (Fig. 4A). Even after in vitro stimulation, the percentage of Tcm and Tscm was still higher among Tet₄₉₉ (Fig. 4B). Further analysis showed that, different from the counterpart of naïve CD8⁺ T cells, the Tet₄₉₉ Tscm expressed more of stem-like cell markers of ScaI, CD122, and CCR7 (Fig. 4C). To minimize nonspecific binding, beads enriched CD8⁺ T cells were pre-blocked by Fc blocker CD16/CD32 antibody. After taking these measures, we verified that Tet₄₉₉ contained more Tscm than Tet₂₁₂ (Supplementary Fig. S3A). The CD44⁻CD62L⁺ Tscm also expressed higher level of CD95 than naïve CD8⁺ T cells, another potential marker of Tscm (Supplementary Fig. S3B). The presence of Tscm was then confirmed in the Tet₄₉₉ cells by dual-color tetramer staining (Supplementary Fig. S3C).

The resistance to antigen-induced Tet₄₉₉ apoptosis may also be due to the Bcl-2 molecules (32). Indeed, we found that the Tscm of Tet₄₉₉ expressed the highest level of Bcl-2 compared with Tcm and Teff, nearly to the same level as naïve CD8+ cells (Supplementary Fig. S3D). In addition, the Bcl-2 level in the Tcm and Teff of Tet₄₉₉ cells was higher than the counterparts of Tet₂₁₂, suggesting that even at the Tcm and Teff stages, Tet₄₉₉ cells can better survive antigen-induced cell death.

In Fig. 3B, we showed that Tet₄₉₉ cells generated more reexpansion. To determine which subsets of Tet₄₉₉ cells could undergo in vivo expansion, we sorted the Teff, Tcm, and Tscm and transferred them into CD45.1 mice (Fig. 5A). Because the number of Tscm and Tcm cells in the Tet₄₉₉⁺ population was too low to be efficiently collected, we sorted the Teff, Tcm, and Tscm based on the total CD8⁺ cells (see sorting strategy and purity of Tscm, Tcm, and Teff in Supplementary Fig. S4A and S4B). The number of Tet₄₉₉⁺Tscm, Tet₄₉₉⁺Tcm, and Tet₄₉₉⁺Teff was calculated using the percentage of Tet₄₉₉⁺ in the sorted CD8⁺ Tscm, Tcm, and Teff population (Supplementary Fig. S4C). Equal numbers (2,000) of Tet₄₉₉⁺Tscm, Tet₄₉₉⁺Tcm, and Tet₄₉₉⁺Teff were transferred into CD45.1 mice. Teff generated no measurable expansion. In contrast, the CD45.2⁺ CD8⁺ cells were detected in the CD45.1 congenic mice receiving Tscm or Tcm cells (Fig. 5B and C). The percentage of CD45.2⁺Tet₄₉₉ cells in the mice receiving Tscm increased by more than 30-fold, from ∼1% to ∼35%, whereas the percentage of CD45.2⁺Tet₄₉₉ cells in the mice receiving Tcm increased only 5 times, from ∼1% to ∼5% (Supplementary Fig. S4C and Fig. 5B and C). These data suggest the Tet₄₉₉⁺Tscm undergo more expansion than Tet₄₉₉⁺Tcm in vivo after reencountering antigen. In addition, after adoptive transfer of Tscm, three populations of cells, Tscm, Tcm, and Teff, were found in the mice (Fig. 5D), suggesting that Tscm could differentiate into Tcm and Teff and maintained self-renewal capability.

It is postulated that TCR signaling strength affects antigen-induced T-cell differentiation, including induction of Tscm and Tcm (29, 33, 34), which have a greater potential of expansion when reencountering the same antigen (35). To determine the TCR signaling strength of Tet₂₁₂ and Tet₄₉₉ cells, we utilized the Nur77^GFP mice (25), in which the TCR signaling strength is reflected by GFP intensity on T cells. We found that, fresh Tet₂₁₂ and Tet₄₉₉ cells from immunized Nur77^GFP mice were GFP^−/low prior to stimulation (Fig. 6A). After restimulation with 2 μg/mL peptide in media for 4 hours, all Tet₂₁₂ and ∼90% of Tet₄₉₉ cells became GFP⁺, with a higher GFP level on Tet₂₁₂ cells (Fig. 6A and B). The remaining GFP^−/low Tet₄₉₉ after peptide stimulation contained more Tscm and Tcm cells; ∼40% of them were Tscm and Tcm. Even in the GFP⁺ population, Tet₄₉₉ contained more Tscm and Tcm than Tet₂₁₂ (Supplementary Fig. S5A–S5C). Peptide titration study revealed that the percentage and MFI of GFP⁺ Tet₂₁₂ were higher than Tet₄₉₉ at any concentration (Fig. 6C). For example, at 0.1 ng/mL, 80% of Tet₂₁₂ cells were GFP⁺, whereas only 20% of Tet₄₉₉ cells were GFP⁺. These data suggest that TCR signaling on Tet₄₉₉ is weaker than Tet₂₁₂. Furthermore, Tet₂₁₂ GFP⁺ cells had higher CD44 expression than the Tet₄₉₉ GFP⁺ cells (Fig. 6D), indicating that the Tet₄₉₉ cells are in a lower activation state.

The TCR signaling strength of Tet₂₁₂ and Tet₄₉₉ cells was further studied by restimulating the purified CD45.2⁺CD8⁺ T cells from immunized Nur77^GFP mice with peptide-pulsed CD45.1 splenocytes. We found that, prior to restimulation, fresh Tet₂₁₂ of immunized mice express higher basal level GFP than Tet₄₉₉ cells (Fig. 6E). Tet₂₁₂ could become GFP⁺ even at low concentration of peptide (Fig. 6F). The Tet₂₁₂ cells restimulated by splenocytes pulsed with 0.1 μg/mL of AFP₂₁₂ peptide expressed higher level of GFP than Tet₄₉₉ restimulated by splenocytes pulsed with 10 μg/mL of AFP₄₉₉ peptide, suggesting that Tet₄₉₉ TCR signaling is at least 100 times weaker than Tet₂₁₂ TCR signaling.

The low reactivation threshold and stronger TCR signaling on Tet₂₁₂ cells (Figs. 2 and 6) may be caused by high T-cell avidity for cognate MHC/peptide. Thus, we conducted a tetramer binding (Fig. 7A) and dissociation (Fig. 7B) assays, which measure the TCR avidity for the MHC/peptide complex. The data showed that approximately 90% of the Tet₄₉₉ cells were stained with a low concentration (0.65 μg/mL) of H-2K^b-D^b/AFP₄₉₉ tetramer. In contrast, to achieve 90% staining of the Tet₂₁₂ cells, 10 times more (6.5 μg/mL) of the H-2D^b/AFP₂₁₂ tetramer was required. In addition, the Tet₄₉₉ TCR binding to MHC/peptide was dissociated more slowly (Fig. 7B). Thus, the Tet₄₉₉ TCRs have a higher avidity for MHC/peptide.

Another factor that can influence the TCR signaling strength is the amount of TCR on the cell surface, which is normally downregulated after activation (36). Thus, we examined the TCR levels on Tet₂₁₂ and Tet₄₉₉ cells. The splenocytes from immunized mice were stained with antibodies to CD8, CD44, CD62L, TCR Vβ (Clone 57-597), plus tetramer. The TCR Vβ levels on Tet⁺ CD8⁺ T cells were compared with Tet^–CD44^–CD62L⁺ naïve CD8⁺ T cells. Our data demonstrated that TCRs on Tet₄₉₉ were more downregulated (Fig. 7C–E). Tet₄₉₉ had only 40% of the TCRs as naïve CD8⁺ cells. In contrast, Tet₂₁₂ had ∼80% of the TCRs as naïve CD8⁺ cells. By direct comparison, the Tet₄₉₉ had approximately half the TCRs of Tet₂₁₂ cells (Fig. 7F). The kinetic analysis of TCR downregulation showed that TCR downregulation on Tet₂₁₂ was mild and transient (Fig. 7G). In contrast, TCR downregulation on Tet₄₉₉ was severe and persistent. As the antibody to TCR Vβ staining may be interfered with by tetramers that compete for the same TCR, the CD3 level on Tet₂₁₂ and Tet₄₉₉ cells was thus measured to determine TCR downregulation. The data showed that CD3 was also more downregulated on Tet₄₉₉ than on Tet₂₁₂ (Supplementary Fig. S6A–S6D). Together, our data showed that, the Tet₄₉₉ TCRs have higher avidity for MHC/peptide, but its level is more downregulated after activation, which may result in ensuing weaker TCR signaling and higher threshold of reactivation.

Our findings associate the antitumor effect of vaccine-activated CD8⁺ T cells with their antigen sensitivity and activation/differentiation states. First, the antigen sensitivity of two AFP-specific CD8⁺ T cells, Tet₂₁₂ and Tet₄₉₉, is different. Tet₂₁₂ have a lower threshold of reactivation and are sensitive to antigen stimulation, but become exhausted and apoptotic when antigen persists and generate no antitumor effect in vivo. In contrast, Tet₄₉₉ have a higher threshold of reactivation and are more resistant to antigen stimulation, but can survive persistent tumor antigen, maintain their effector function, and generate a potent antitumor effect in vivo. Secondly, the vaccine-activated Tet₂₁₂ and Tet₄₉₉ cells are at different differentiation and activation state. Tet₄₉₉ cells are at a lower activation state and consist of diverse subsets at different differentiation stages, including Tscm and Tcm. In contrast, there are no Tscm and few Tcm in Tet₂₁₂. The Tet₄₉₉ Tscm can undergo extensive in vivo expansion when reencountering antigen. Thirdly, the TCR signaling on Tet₂₁₂ is stronger and associated with mild and transient TCR downregulation. In contrast, the Tet₄₉₉ TCR signaling is weaker, likely due to severe and persistent TCR downregulation.

TCR signaling strength correlates to the activation and magnitude of antigen-specific T cells (37–39). Indeed, data from Nur77^GFP mice showed that the TCR signaling on Tet₂₁₂ was stronger, corresponding to a higher magnitude of Tet₂₁₂ response by immunization (24). But, on the other hand, strong TCR signaling can negatively affect the induction of Tscm (29, 40). In agreement with this argument, the Tet₄₉₉, but not Tet₂₁₂, contains Tscm. Stem-like memory immune cells were proposed by Fearon and colleagues (41). The Tscm was found to cause graft versus host disease (35). Gattinoni and colleagues showed that the TCR transgenic gp100-specific Tscm generated in vitro by inhibiting Wnt signaling had a better antitumor effect (42). They further found that human Tscm mediated strong antitumor effect in mice bearing human tumors (43). In a retrospective study, Speiser and colleagues found that yellow fever vaccine-induced Tscm in human correlated to long-term protection (44). Here, we show that cancer vaccine can elicit Tscm in vivo, which is associated with TCR downregulation and weak TCR signaling, and correlates with stronger antitumor effect.

The engagement of TCR and MHC/peptide and proper TCR signaling are also needed for the reactivation of vaccine-induced T cells to recognize and kill tumor cells. T cells with higher basal level of TCR signaling are more ready for activation (45, 46). Even though the Tet₂₁₂ and Tet₄₉₉ activated by vaccine are not naïve cells, their reactivation seems also correlate to their basal TCR signaling. Tet212 have a higher TCR signaling strength and are more ready for reactivation and consequently also suffers from activation induced cell death when antigen persists. In contrast, the Tet₄₉₉ cells with weaker TCR signaling survive persistent antigen stimulation and generate a stronger antitumor effect. This is consistent with previous hypothesis that TCR downregulation may protect T cells from excessive signaling (47). Thus, TCR downregulation and weaker TCR signaling on the activated T cells in the effector phase may protect them from antigen-induced exhaustion and apoptosis, especially in the tumor lesions where the antigen is abundant and the stimulation is likely persistent. But the high concentration of antigens in tumor lesion and the prolonged engagement of T cells and tumor cells will compensate for the lower TCR level and generate sufficient accumulative signaling strength to activate Tet₄₉₉ cells to kill tumor cells.

The association of weak TCR signaling with stronger antitumor effect of T cells seems counterintuitive. Our finding of Tet₄₉₉ TCR downregulation associated with a strong antitumor effect is also in contrast to a report showing that TCR downregulation limits their efficacy of controlling microbial infection (48). We reason that this is likely due to the difference of antigen level and antigen stimulation duration between microbial infections and tumor lesion. In acute infections, the microbial antigen is transient, and thus T cells with higher level or sensitive TCR should not be exhausted and are more effective to eliminate the infection because of the stronger engagement of TCR and MHC/peptide. On the other hand, in tumor lesion, the T cells are immersed in a pool of tumor antigens for a prolonged period of time; thus, T cells with sensitive threshold of reactivation will likely be driven to exhaustion. Only the T cells with weaker TCR signaling may survive persistent antigen stimulation and generate antitumor effects. In agreement with this theory, in chronic persistent viral infection, T cells are exhausted (49). The report that T cells activated with intermediate affinity peptides (thus intermediate strength of TCR signaling) generate a better antitumor effect (50) is also in agreement with our data.

Based on these findings, we propose a model (Supplementary Fig. S7), in which the TCR signaling strength and the threshold of activation controls the state of T-cell activation, the generation of Tscm, antigen responsiveness, and the antitumor effect of the vaccine-activated CD8⁺ T cells. In this model, initial high-affinity TCR–MHC/peptide complex engagement results in T-cell activation and concurrently generates feedback to downregulate TCR on activated T cells, weakening the TCR signaling to avoid overactivation of T cells in the priming phase. The weakened TCR signaling will allow some activated T cells to halt differentiation at Tscm stage to reserve their potential to generate better responsiveness and stronger antitumor effect when they reencounter cognate antigen in tumor lesions. Thirdly, the weakened TCR signaling on Teff will protect them from exhaustion and apoptosis upon reencountering persistent antigen in tumor lesion. In contrast, if the TCR downregulation is mild and transient, TCR signaling will likely drive all activated T cells to become terminal Teff, which not only decrease Tscm, but also predispose the Teff to exhaustion and apoptosis when they reencounter antigen in tumor lesion.

In this model, we attribute the different TCR signaling strength and antigen sensitivity of Tet₂₁₂ and Tet₄₉₉ cells to their different TCR downregulation. However, these differences may also be due to the inherent difference of internal signaling pathways inside the Tet₂₁₂ and Tet₄₉₉ cells. Whether the high-affinity TCR–MHC/peptide engagement also reduces the internal signaling pathway is not clear, nor is the inherent signaling difference between naïve Tet₂₁₂ and Tet₄₉₉ cells. Even though our data support that the different antigen responsiveness of vaccine-activated Tet₂₁₂ and Tet₄₉₉ cells relates to their different TCR downregulation and signaling strength, further studies should be done to investigate whether the naïve Tet₂₁₂ and Tet₄₉₉ cells are different in their antigen responsiveness. If naïve Tet₄₉₉ and Tet₂₁₂ cells indeed have inherent signaling difference that is beyond the TCR level, the T cells with higher threshold of activation and capability of resisting antigen-induced cell death may be the ideal host cells for generating TCR and chimeric antigen receptor (CAR) gene–engineered T cells for immunotherapy of solid tumors, where the TCR or CAR gene–engineered T cells need to survive persistent antigen stimulation and generate antitumor effect.

This current study also raises questions in cancer vaccines and immunotherapy. The first question is whether TCR signaling strength on any antigen-specific T cells can be manipulated to enhance Tscm generation and to prevent exhaustion in tumor lesion. Secondly, how should cancer vaccine and immunization strategies be devised to find the right balance between eliciting sufficient number of tumor-specific T cells while inducing Tscm to achieve the maximal antitumor effect? Investigation into these questions should help develop more effective cancer vaccines and enhance the antitumor efficacy of current immunotherapies.

No potential conflicts of interest were disclosed.

Conception and design: S. Wu, W. Zhu, L. Huang, D. Bartlett, N. Fu, Y. He

Development of methodology: S. Wu, W. Zhu, Y. Hong, L. Huang, Z.S. Guo, Y. He

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Wu, W. Zhu, Y. Peng, L. Wang, Y. Hong, D. Dong, E. Kruse, Y. He

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Wu, W. Zhu, Y. Peng, Y. Hong, L. Huang, D. Dong, J. Xie, N. Fu, Y. He

Writing, review, and/or revision of the manuscript: S. Wu, L. Huang, D. Dong, T. Merchen, E. Kruse, D. Bartlett, Y. He

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): W. Zhu, Y. Peng, L. Huang, D. Dong, D. Bartlett, Y. He

Study supervision: Y. He

We acknowledge the stimulating discussion with Dr. Bjoern Peters of La Jolla Institute for Allergy and Immunology and Drs. Rafal Pacholczyk, David Munn, and Esteban Celis at Augusta University. The excellent service of NIH Tetramer Core Facility is greatly appreciated. We thank Dr. Rhea-Beth Markowitz of Augusta University for editing the manuscript.

This work is funded by NCI grant 1R01CA168912 and Georgia Cancer Center Grant to Y. He.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.