CD137 Agonist Therapy Can Reprogram Regulatory T Cells into Cytotoxic CD4⁺ T Cells with Antitumor Activity Free

Despite robust anti-tumor immunity, control and clearance of tumor growth is often impaired in cancer patients, leading to uninhibited growth, metastasis, and death. Several strategies have been developed to increase the antitumor immune responses including reversion of T cell exhaustion with agonistic Abs such as PD1 or CTLA4 (1), increasing frequency of tumor-specific T cell responses by ex vivo expansion and reinfusion (2), and increasing antitumor activity by targeting costimulatory receptors (3). One approach with promising results in an animal model has been the stimulation of the TNFR superfamily member CD137 (4-1BB), which is involved in T cell activation and function (4) including expansion, survival, and cytokine production of effector T cells (3, 5). Because CD137 is more prominently expressed on CD8⁺ T cells than CD4⁺ T cells, most immunotherapy studies have focused on the impact of anti-CD137 on CD8⁺ T cell immunity (4). However, a recent study demonstrated that, in addition to improved CD8⁺ T cell responses in B16 Melanoma, anti-CD137 treatment also drove differentiation of CD4⁺ T cells into cytotoxic effectors (6). This cytotoxicity of CD4 T cells was dependent on the regulatory T-box transcription factor Eomesodermin (Eomes) and mediated by KLRG-1⁺ terminally differentiated effector T cells (6). Curran et al. (6) nicely showed that effector KLRG-1⁺Eomes⁺ CD4⁺ and CD8⁺ T cells, generated by anti−CD137 treatment, enhanced tumor-specific cytotoxicity and in turn promoted tumor immunity. Because the majority of cytotoxic antitumor immunity is mediated by CD8⁺ T cell response, it remains unknown how much cytotoxic CD4⁺ T cells actually contribute to tumor rejection in such immune studies and in particular after treatment with CD137 agonist. In addition, the significance of regulatory T cells (Tregs) after anti-CD137 therapy has not been well defined and remains contradictory. Because CD137 is not only expressed on activated effector T cells, but also on CD4⁺Foxp3⁺Tregs (7), it is unknown whether activation of Tregs through CD137 may drive increased suppressive Treg activity as previously suggested (8).

To address these questions, we made use of a murine retrovirus-induced tumor cell line of C57BL/6 (B6) origin, namely, FBL-3 cells. FBL-3 is a Friend virus (FV)–transformed tumor cell line that does not produce infectious virus, but expresses highly immunogenic FV Ags on the cell surface (9, 10). s.c. implantation of FBL-3 cells into B6 mice results in local tumor growth with subsequent regression caused by effector CD8⁺ T cells (11, 12). In a previous article, we reported that, in this FBL-3 tumor model, Tregs suppress effective antitumor CD4⁺ T cell responses and thereby contributed toward tumor progression (13). The cytotoxic CD4⁺ T cells could fully compensate for the antitumor effect of CD8⁺ T cells, when their suppression by Tregs was abrogated because of Treg depletion. In this work, we show that administration of an agonist CD137 Ab into mice lacking CD8⁺ T cells promoted effective cytotoxic antitumor CD4⁺ T cell responses and led to FBL-3 tumor elimination in the presence of Tregs. Moreover, anti-CD137 treatment converted a subpopulation of Foxp3⁺ Tregs into tumor cell–killing effector CD4⁺ T cells. Our results reveal an intriguing immunology of a unique subpopulation of Foxp3⁺CD4⁺ T cells that can adopt antitumor effector functions after potent costimulatory signaling.

Experiments were done using sex- and age-matched B6, CD45.1, MHCII^−/−, and DEREG (14) mice that were between 8 and 10 wk old at day 0 of the experiments. Mice were housed in specific pathogen-free conditions and treated in accordance with institutional guidelines.

FBL-3 is an FV-induced tumor cell line derived from a B6 mouse (15). FBL-3 cells were maintained in complete RPMI-1640 medium (Invitrogen, Gaithersburg, MD) supplemented with 10% FCS and 0.5% penicillin/streptomycin. Mouse epithelial cells (MTECs) are transformed mouse tonsil epithelial cells, which do not carry viral proteins (HPV⁻) (16). MTECs were grown in E media containing DMEM as previously described (16).

A total of 1 × 10⁷ FBL-3 tumor cells were injected s.c. on the right flank in 100 μl PBS through a 27-gauge needle on day 0. To verify tumor volume by external caliper, we determined the greatest longitudinal diameter (length) and the greatest transverse diameter (width). Tumor size based on caliper measurements was calculated by formula: tumor area (cm²) = π × a × b, where a is half of length and b is half of width. After 6 d posttumor challenge (ptc), mice were sacrificed and draining lymph nodes (inguinal) were resected. A total of 1 × 10⁶ MTECs were implanted s.c. on the right flank of B6 mice. Eight days ptc, mice were sacrificed and draining lymph nodes (inguinal) were extracted for further analysis.

In brief, mice received i.p. 0.5 ml supernatant fluid obtained from hybridoma cell line 169.4 producing CD8a-specific mAb (17). Depletion was started at day 0 and carried out every other day for the tumor growth analysis until mice were killed because of the progressive tumor growth, and three times (on days 0, 2, 4) for the experiments where mice were killed at day 6 postinoculation. The treatment depleted >95% of the CD8⁺ T cells in lymph nodes, whereas only 40–50% of CD11c⁺CD8⁺ DCs were depleted after this treatment (data not shown) (18). The CD4⁺ T cells were depleted with the same efficacy by i.p. injection of 0.5 ml supernatant fluid obtained from hybridoma cell culture YTS 191.1 producing CD4-specific mAb (17). To deplete Tregs, we injected DEREG mice i.v. with 1 μg diphtheria toxin (DT) (Merck) diluted in endotoxin-free PBS. The treatment depleted >97% of the CD4⁺ T cells expressing fluorescent protein (GFP) in DEREG mice. The anti-CD137 (LOB 12.3) (19) used in vivo were produced by BioXCell. Dosing per injection was 100 μg administered i.p. every other day from day 0.

The Abs used for cell-surface staining were anti-CD4 (eFluor 605, RM4-5), anti-CD43 (PerCP, 1B11), KLRG-1 (BV421, 2F1) (Biolegend); anti-CD11b (BV650, anti–Mac-1 [WT.5]), anti-CD11a (PE, M17/4), CD45.1 (PeCy7, A20), and Fc block anti-mouse CD16/CD32 (93) (eBioscience), and Neuropilin1 (Nrp1; allophycocyanin, FAB566A; R&D Systems). Intracellular granzyme B (monoclonal anti-human granzyme B Ab allophycocyanin-conjugated; Invitrogen) was performed as described previously (20). Dead cells for cell-surface and intracellular staining were excluded by using Fixable Viability Dye (eBioscience). To determine intracellular production of IFN-γ, TNF-α, and IL-2, we stimulated cells from lymph nodes in the presence of 2 μg/ml of the CD28 Ab and 2 μg/ml brefeldin A for 5 h at 37°C. The cells were then stained for surface expression of CD4, CD8, and CD43, fixed, and permeabilized with Cytofix/Cytoperm solution (BD). The cells were then washed, permeabilized, and incubated with Fc blocking anti-mouse CD16/CD32. After that cells were labeled with mAbs specific for IL-2 (eFluor 450, JES6-5H4), IFN-γ (FITC, XMG1.2), TNF-α (BV510, MP6-XT22), and anti-CD154 (PE, CD40Ligand, gp39, MR1). Foxp3 (PE-TexRed, FJK-16s), Eomes (PerCP-eFluor710, Dan11mag), T-bet (PeCy7, eBio4B10), and Ki67 (PeCy7, SolA15) expression were detected by intracellular staining using the Foxp3 staining kit (eBioscience). Data were acquired on an LSRII flow cytometer (Becton Dickinson) from 300,000–500,000 lymphocyte-gated events per sample. The analyses were done using the FACSDiva software (Becton Dickinson) and the FlowJo software (Tree Star).

For Treg conversion experiments, CD4⁺ T cells were isolated from spleens and lymph nodes (inguinal, cervical, mesenteric, and axillary) of naive DEREG mice using CD4⁺ selection (Miltenyi Biotec) according to the manufacturer’s instructions. Cells were then sorted on a FACSAria (BD Biosciences) to >95% pure populations of CD4⁺GFP⁺ or CD4⁺GFP⁻ cells. Between 1 and 3 × 10⁶ CD4⁺GFP⁻ cells and 1–5 × 10⁵ CD4⁺GFP⁺ cells were transferred into CD45.1 recipients by i.v. injection in 0.5 ml PBS on day 0 after FBL-3 challenge and Ab administration.

In vivo cytotoxicity assay was performed as described by Barber et al. (21) using B6 or DEREG mice. Five days ptc, all groups of mice received i.v. peptide-loaded CFSE-labeled lymphocyte targets (Vybrant CFDA SE Cell Tracer Kit; Life Technologies), as well as unloaded cells from CD45.1 mice as a control population. Naive cells from MHCII^−/− mice were either labeled with CFSE and pulsed with peptide or labeled with CellTrace Violet (Invitrogen) without peptide incubation to confirm tumor-specific lysis of targets. Twenty hours after i.v. injection of donor cells, on day 6 ptc, the mice were sacrificed and in vivo killing activity was quantified in single-cell suspensions from the draining lymph node of each tumor-bearing mouse. In addition, in some experiments, 1 μg DT was added together with the target cell suspension. Tumor target lysis assay was performed using 2 × 10⁵ FBL-3 cells/mouse labeled with 10 μM CFSE. FBL-3 cells were injected i.p. into naive CD45.1 mice. In addition, mice received sorted CD4⁺GFP⁺ cells from either anti-CD8 or anti-CD8+anti-CD137 tumor-bearing or naive mice. The mice were sacrificed 48 h later, and i.p. lavage was performed with 10 ml PBS to obtain cells. Cells were washed once, resuspended in buffer containing fixable viability dye (eBioscience) to exclude dead cells, and analyzed by flow cytometry for CFSE⁺ signal.

MHC class II (MHCII) tetramers loaded with I-A^b–restricted FV-specific CD4⁺ T cell epitope (H19-Env; EPLTSLTPRCNTAWNRLKL) were obtained from the National Institutes of Health Tetramer Facility (Atlanta). Nucleated lymph node cells were incubated with allophycocyanin-labeled I-A^b tetramers for 3 h at 37°C and later stained with anti-CD4 and anti–Mac-1 for 10 min at room temperature.

Statistical analyses and graphical presentations were computed with GraphPad Prism version 6. Statistical differences (p value) between two groups were performed using an unpaired t test. Statistical differences (p value) between the different parameters were performed testing with the Kruskal–Wallis one-way ANOVA on ranks and Newman–Keuls multiple comparison tests. In this work, all p values ≤0.05 were determined to be significant.

To address the question whether CD137 agonist treatment can alter CD4⁺ T cell effector function and restore sufficient antitumor immunity in the absence of CD8⁺ T cell mediated control, we first depleted mice of CD8⁺ T cells and determined FBL-3 tumor progression (Fig. 1A). Indeed, similar to previous studies (11, 22), we found no tumor control in mice depleted for CD8⁺ T cells, and animals had to be euthanized on average 15 d after tumor infusion. We next administered a CD137 agonist Ab every second day starting at the time point of the FBL-3 tumor infusion to CD8⁺ T cell–depleted mice. Strikingly, despite the absence of CD8⁺ T cells, the costimulation of CD4⁺ T cells with CD137 agonist restored anti–FBL-3 tumor immunity, and injected tumor cell lines were completely rejected after on average 22 d of tumor challenge (Fig. 1B). This is only a slight delay of 7 d in tumor rejection compared with only FBL-3–challenged mice (see dotted line in Fig. 1A). To validate the CD4⁺ T cell–mediated control of tumor growth in CD8⁺ T cell–depleted animals, we next depleted mice not only of CD8⁺ T cells but also CD4⁺ T cells and treated them with a CD137 agonist Ab during FBL-3 tumor challenge (Fig. 1C). As expected, depletion of CD4⁺ and CD8⁺ T cells led to a loss of tumor control, despite the presence of an anti-CD137 Ab, indicating that this is a T cell–mediated effect. Thus, our data demonstrate that, in the absence of CD8⁺ T cells, CD4⁺ T cells can restore antitumor immunity in the presence of CD137 costimulation.

To understand the underlying immunobiology of the described phenomenon, we next characterized differences in the antitumor CD4⁺ T cell responses in mice depleted for CD8⁺ T cells compared with those that also received CD137 agonist treatment during FBL-3 tumor challenge. Notably, there was no difference in total CD4⁺ T cell frequencies between anti-CD8– and anti-CD8+anti-CD137–treated mice (Supplemental Fig. 1A). To analyze the antitumor immune response after CD137 treatment, we quantified the population of tumor-specific effector CD4⁺ T cells by staining lymphocytes from tumor-draining lymph nodes with MHCII tetramers loaded with the H-2I-A^b–restricted CD4⁺ T cell epitope FV H19-Env (23). At 6 d ptc, anti-CD137–treated mice had significantly more tetramer⁺ CD4⁺ T cells than only CD8⁺-depleted mice that did not receive treatment (p = 0.0003; Fig. 2A), indicating that CD137 signaling enhances antitumor CD4⁺ T cell immunity. Using intracellular cytokine staining after stimulation with anti-CD137, we found not only more antitumor CD4⁺ T cells in CD8-depleted, anti-CD137–treated mice, but also that these cells were more polyfunctional, producing all three proinflammatory cytokines IFN-γ, TNF-α, and IL-2 (Supplemental Fig. 1B–D). Interestingly, the CD4⁺ T cells in CD8-depleted, anti-CD137–treated mice showed significantly enhanced expression of the T-box transcriptional factors T-bet and Eomes (Fig. 2B, 2C), which have been suggested to control the Th1 helper versus cytotoxic function of T cells (24). However, CD4⁺ T cells usually express only little Eomes because cytotoxicity is mostly mediated by CD8⁺ T cells, suggesting that CD4⁺ T cells may have acquired cytotoxic activity after CD137 signaling. Indeed, we found that Eomes⁺CD4⁺ T cells in the treated mice showed substantially higher expression of the effector molecule granzyme B (Fig. 2D). An in vivo cytotoxicity assay with MHCII-FV peptide-loaded target cells revealed that CD4⁺ T cell killing was at the detection limit of the assay in only CD8⁺ T cell–depleted tumor-bearing mice, whereas the additional anti-CD137 treatment resulted in an average of >30% target cell killing (Fig. 2E). When peptide-loaded targets from MHCII^−/− mice were used for the in vivo cytotoxicity assay, we found no killing activity anymore, confirming that we indeed measured MHCII–restricted CD4⁺ T cell killing in our assay. Collectively, the data indicate that, in the absence of CD8⁺ T cells, anti-CD137 therapy induces strong helper and cytotoxic CD4⁺ T cell responses that are capable of controlling FBL-3 tumor cell growth.

In a previous study we have demonstrated that antitumor effects of cytotoxic CD4⁺ T cells in CD8⁺ T cell–depleted animals are strictly regulated and suppressed by Tregs impairing antitumor immunity (13). Because we found that anti-CD137 treatment drives CD4⁺ T cell–mediated cytotoxicity and restores antitumor immunity, we next investigated whether anti-CD137 treatment influences the otherwise suppressive Treg response. We first confirmed expression of CD137 on Tregs of FBL-3–challenged mice and found CD137 expression on a subpopulation of activated (CD43⁺) Tregs (Supplemental Fig. 2A). We next investigated the impact of CD137 agonist on the phenotype and function of Tregs during FBL-3 tumor rejection. As previously described (13), we found that only 13% of the total CD4⁺ T cell population in draining lymph nodes was Tregs on day 6 ptc and that this frequency was unchanged in CD8⁺ T cell–depleted mice. In contrast, after anti-CD137 injection into CD8-depleted mice, we found a significant expansion of Foxp3⁺ Tregs comprising up to 30% of all CD4⁺ T cells in the tumor-draining lymph nodes (p = 0.0001; Fig. 3A). Furthermore, we also found a significant increase of recently proliferated Tregs identified by the expression of Ki67 (p = 0.017; Fig. 3B) and activated Tregs identified by the expression of the activation markers CD43 and CD11a (p = 0.0001; Fig. 3C). We also found increased expression of the maturation marker KLRG-1 on Tregs, which has been linked to cytotoxic activity of CD8⁺ and NK cells (Fig. 3D). We next addressed whether the expanded Tregs were natural Tregs (nTregs) or induced Tregs, and therefore analyzed the expression of Nrp1, a receptor for ligands of the vascular endothelial growth factor family expressed on nTregs (25), and the specificity of the expanding Tregs. Interestingly, no significant expansion of tumor-specific Tregs (tetramerII⁺) was found (data not shown), but the majority of Tregs expanding after CD137 treatment expressed Nrp1 (Fig. 3E), suggesting that CD137 treatment mostly affected thymic-derived nTregs. This was also supported by the observation that Foxp3⁺ Tregs expanded in the thymus of anti-CD8+anti-CD137–treated tumor-bearing mice (data not shown). CD137 expression on thymic nTregs was found on ∼7.5% of all Foxp3⁺ cells in tumor-challenged, CD8-depleted mice (Supplemental Fig. 2B) and were therefore potential targets for anti-CD137 therapy. Thus, anti-CD137 treatment of CD8-depleted, tumor-bearing mice induced strong activation, proliferation, and differentiation of a subset of nTregs.

Even though we found that CD137 agonist treatment significantly augmented Treg responses in tumor-bearing mice, we made the contradictory observation of an increase in the antitumor CD4⁺ effector T cell response. To address this apparent disconnect in the Treg CD4⁺ effector T cell balance, we asked whether CD137 treatment may have altered or impaired the function of Tregs. Using intracellular cytokine staining, we found that a subset of Tregs from tumor-draining lymph nodes of anti-CD137–treated, CD8-depleted treated mice acquired the expression of CD154 (CD40L), which is critical for T cell–mediated help (Fig. 4A). Moreover, we found that anti-CD137 treatment induced significant expression of TNF-α in Foxp3⁺CD154⁺CD4⁺ T cells, but no expression of other proinflammatory cytokines (IL-2 or IFN-γ), suggesting a major alteration of the functional Treg phenotype (Fig. 4B, 4C). We next addressed whether anti-CD137 treatment also reprograms Tregs to a cytotoxic phenotype. We therefore quantified the expression of the T-box transcriptional factor Eomes in these cells and found a significant increase in the expression of Eomes in Foxp3⁺CD4⁺ T cells in anti-CD137–treated mice compared with animals depleted only for CD8⁺ T cells and challenged with FBL-3 tumor cells (Fig. 4D). Indeed, up to 14% of the Foxp3⁺ CD4⁺ T cells coexpressed Eomes after anti-CD137 injection, whereas almost none of these double-positive cells were found before Ab treatment (Fig. 4E). Most of the Foxp3⁺ Eomes⁺CD4⁺ T cells in the tumor-draining lymph node also expressed the Treg activation marker Helios (data not shown). Remarkably, anti-CD137 treatment also had an effect on thymic nTregs. It induced a significant expansion of Foxp3⁺ Eomes⁺ CD4⁺ T cells in the thymus of FBL-3–challenged, CD8-depleted mice (Supplemental Fig. 2C). Within the tumor-draining lymph nodes, Foxp3⁺ CD4⁺ T cells started to produce granzyme B after anti-CD137 therapy, which was measured in direct ex vivo staining experiments without in vitro restimulation. The number of granzyme B–expressing cells and the expression levels significantly increased, whereas no granzyme B production was found in the cells of mice from the control groups (Fig. 4F; Supplemental Fig. 2D). Because no granzyme production in Tregs was found in tumor-challenged, CD8-depleted mice that did not receive anti-CD137 Ab treatment, we conclude that only the CD137 costimulatory signal was able reprogram a subpopulation of the Tregs to become cytotoxic effector cells. To investigate whether the effect of the anti-CD137 Ab on Tregs is intrinsic or requires other cell populations, we performed an adoptive transfer of purified Foxp3⁺ cells from DEREG mice into CD137-treated RAG mice. A subpopulation of the transferred Foxp3⁺ T cells differentiated into Eomes-expressing cells (Supplemental Fig. 4B), suggesting that the effect of the anti-CD137 Ab on Tregs was induced by intrinsic signaling.

Interestingly, similar results were obtained in the MTEC tumor model. In this model, the tumor cells are not controlled by tumor immunity, and mice implanted with MTECs develop tumors and die (16). The administration of CD137 agonist therapy led to the delay in tumor growth compared with nontreated animals (data not shown). The Ab treatment of MTEC-challenged mice significantly increased the percentages of total Foxp3⁺ T cells, as well as the frequency of Eomes⁺Foxp3⁺ T cells among the CD4⁺ T cells compared with only MTEC-challenged mice (Fig. 4G). The percentage of Eomes⁺ cells among the Foxp3⁺ T cells (around 3%) was similar to what we found in the FBL-3 model after anti-CD137 injection, when no CD8⁺ T cells were depleted (Supplemental Fig. 3C). However, as long as potent CD8⁺ T cells were present, the subpopulation of Eomes⁺ Foxp3⁺ cells did not seem to contribute to FBL-3 tumor rejection (data not shown). Taken together, these results demonstrate that CD137 signaling reprograms subsets of Tregs into cytokine-expressing Th cells or even cytotoxic killer cells. This effect of CD137 stimulation could be demonstrated in different tumor models and was most efficient in the absence of CD8⁺ T cells.

The surprising alteration of the effector cell characteristics of CD4⁺Foxp3⁺ T cells after CD137 treatment prompted us to ask whether reprogrammed Tregs are involved in the elimination of FBL-3 tumor cells and restore tumor immunity as shown in Fig. 1. We performed a series of in vivo killing experiments in DEREG mice, which express a DT receptor under the control of the Foxp3 promoter. i.v. administration of DT allowed to rapidly (within hours) and selectively delete >95% of Foxp3⁺CD4⁺ T cells (data not shown). This was important because we wanted to study the immediate effect of the Foxp3⁺ cells on target cells and not the secondary effect that a more sustained Treg depletion has on effector T cell responses. DEREG mice were challenged with tumors, depleted for CD8⁺ T cells, and treated with anti-CD137 to induce the population of Foxp3⁺ CD4⁺ T cells with the cytotoxic effector phenotype. Cell tracer-labeled peptide-loaded (FV CD4⁺ T cell epitope peptides that are expressed on FBL-3 tumor cells) target cells were injected i.v. simultaneously with DT or PBS on day 5 ptc, and the in vivo killing of the targets was determined 20 h later. In PBS control mice, the total population of anti-CD137–stimulated CD4⁺ T cells, most likely including conventional T cells and the Foxp3⁺ CD4⁺ T cells that expressed Eomes, mediated a potent in vivo killing activity (Fig. 5A). Strikingly, this activity was significantly decreased after specific Foxp3⁺ cell ablation because of i.v. DT administration (Fig. 5A), suggesting a potent contribution of the converted Tregs to the total MHCII-restricted killing of peptide-loaded targets (Figs. 2D, 5A). This killing activity of CD137-stimulated Foxp3⁺ CD4⁺ T cells was confirmed in another in vivo cytotoxicity assay, in which we directly used FBL-3 tumor cells as targets. The FBL-3 cells were transferred into the peritoneal cavity of naive mice together with enriched donor Foxp3⁺ CD4⁺ T cells from tumor-challenged mice either left untreated, treated with anti-CD8 alone, or treated with anti-CD8+anti-CD137. Forty-eight hours later, we reisolated and quantified the remaining FBL-3 cells by a peritoneal lavage. In mice that received Foxp3⁺ CD4⁺ T cells from anti-CD8+anti-CD137–treated animals, up to 90% of the tumor cells were eliminated (Fig. 5B). This FBL-3 cell killing was significantly higher than in groups that received Foxp3⁺ cells either from nontreated or only anti-CD8–treated tumor-bearing mice (Fig. 5B). Thus, responding to anti-CD137 therapy, a subset of Tregs obtained an effector phenotype of cytotoxic CD4⁺ T cells and was able to eliminate FBL-3 tumor cells in vivo.

The appearance of Foxp3⁺CD4⁺ T cells with cytotoxic activity against tumor cells after anti-CD137 therapy may be caused by Tregs converting into cytotoxic effectors or rather, cytotoxic CD4⁺ T cells starting to express Foxp3 after CD137 signaling. To test both possibilities, we performed an adoptive transfer experiment where Foxp3⁺ or Foxp3⁻ donor cells (CD45.2⁺ cells were from DEREG mice and distinguished based on their GFP expression) were injected into CD45.1 mice that were challenged with FBL-3 cells and treated with anti-CD8⁺ and anti-CD137 Abs. After 6 d the donor cells were isolated from the tumor-draining lymph node and analyzed for Foxp3 and Eomes expression. Interestingly, none of the conventional CD4⁺ T cells (Foxp3⁻ donor cells) that expressed Eomes also expressed Foxp3. However, 14% of the transferred Tregs (Foxp3⁺ donor cells) additionally induced Eomes expression in the recipient mice (Fig. 5C), indicating that indeed Tregs can be reprogrammed into effector CD4⁺ T cells after CD137 signaling (Fig. 5C).

Taken together, these data suggest that anti-CD137 treatment drives a subpopulation of nTregs into cytotoxic effector T cells that significantly contribute to FBL-3 tumor cell elimination.

Despite advances in the development of immune therapeutic strategies to combat cancer, the underlying immunobiology and the changes occurring under immune-modulating therapies are not well understood. In this study, we used an FV-induced FBL-3 tumor model to study the mechanisms of immunological tumor control in the course of CD137 costimulatory immunotherapy and lack of proper CD8⁺ T cell responses. We demonstrate that, in the absence of CD8⁺ T cells, cytotoxic CD4⁺ T cells can restore antitumor immunity after systemic CD137 activation. Interestingly, this tumor clearance was partially due to a subset of CD4⁺Foxp3⁺ T cells that were reprogrammed to eliminate immunogenic virus-induced tumor cells in response to CD137 agonist treatment. Thus, for the first time, to our knowledge, we were able to demonstrate that under certain circumstances Tregs can turn on the cytolytic program and gain cytotoxic effector functions. The antitumor effect of these CD4⁺Foxp3⁺ Eomes⁺ T cells was demonstrated in the highly immunogenic FBL-3 tumor model, but whether this also applies to less immunogenic tumors is an area of future investigation. The cytolytic program might be directly induced by CD137 signaling in Tregs, because a subpopulation of Foxp3⁺ cells expresses CD137. This is supported by our findings that a subpopulation of Foxp3⁺ T cells transferred into CD137-treated RAG mice turned on the expression of Eomes (Supplemental Fig. 4B), indicating that no other lymphocyte population is needed for the effect of anti-CD137 Ab on Tregs.

The TNFR family members have been found to play major roles as costimulatory receptors for CD4⁺ and CD8⁺ T cells. Among various TNFR family members including CD134 (OX-40), CD27, CD357 (GITR), and others, CD137 was the first identified as a possible immunotherapy target (26). The efficacy of its ligation was shown in different tumor studies, demonstrating that treatment of tumor-bearing mice with CD137 agonist therapy promoted effective antitumor immune responses (27). In this work, an anti-CD137 Ab that provides additional costimulatory signal to cells was used. It remains unknown whether this Ab blocks natural ligand binding. Part of the biological effect of the Ab could therefore also be induced by competing with CD137 ligand binding. However, this is very unlikely for a gain-of-function effect of the Ab, as the one that we describe in this article.

Previous studies have demonstrated that in the FBL-3 tumor model CD8⁺ T cells are essential to control tumor growth. However, in many tumor entities and tumor models, CD8⁺ T cells are exhausted and impaired to control tumor spread, which likely contributes to the often-observed limited efficacy of immunotherapy (28). Previously we demonstrated that tumor-specific CD4⁺ T cells apart from being helper cells can substitute for the function of CD8⁺ T cells and efficiently control tumor growth (13). A targeted activation of tumor-specific CD4⁺ T cells by CD137 costimulatory Ab might therefore substantially contribute to antitumor immunity and immune control of tumor progression.

It has previously been reported that the cytotoxic potential of tumor-specific CD4⁺ killer T cells depends on transcription factor Eomes expression (6). In addition, it was demonstrated that Eomes plays an important role in initiating granzyme production in T cells (24). In agreement with this study, we show in this article that an increased tumor-specific killing by CD4⁺ T cells after anti-CD137 therapy correlates with levels of Eomes and granzyme B in the CD4⁺ T cell compartment. Another study by Qui et al. (24) described in vitro differentiation of CD4⁺ T cells into cytotoxic Th cells, in which Eomes, similar to CD8⁺ T cells, initiates the cytotoxic function. Thus, CD137 signaling can obviously induce a transcriptional program in CD4⁺ T cells that results in potent cytotoxic activity of these cells.

CD137 can be constitutively expressed on CD4⁺Foxp3⁺ Tregs (7, 29), but whether signaling through CD137 treatment increases suppressive Treg activity, and therefore counteracting antitumor CD4⁺ T cell responses, is currently unknown (5, 29–31). Interestingly, we found that CD137 ligation on Tregs increases not only the size of the Treg population but also drives alterations of Treg functions to effector cells in the absence of CD8⁺ T cells. Induction of CD4⁺ T cells that express Foxp3 and at the same time TNF-α and granzyme B by CD137 Ab was also observed in the presence of CD8⁺ T cells, although the size of the cytotoxic CD4⁺ T cell population was smaller than in CD8-depleted mice (Supplemental Fig. 3). In contrast with our study, Choi et al. (29) showed that CD137 ligation on Tregs inhibited their suppressive function in vitro without inducing proliferation, whereas stimulation of Tregs with a CD137-Fc fusion protein expanded Tregs but did not manipulate their functions (31). However, the identification of Tregs in Choi et al.’s study (29) was performed by staining for CD25 surface expression (29). The use of a nonspecific Treg marker such as CD25, which is also expressed on recently activated CD25⁺ effector T cells, makes the interpretation of the data somewhat complicated.

Generally the role of Tregs is an immunosuppressive regulation of activated CD8⁺ T cells and CD4⁺ T cells. However, there is a certain degree of plasticity that Tregs retain, and under specific conditions they may adopt a proinflammatory phenotype (32). This phenotype includes a helper-like role and has been observed in different models, including vaccination (33), tumors (34), or graft rejection (35). In particular, Addey et al. (34) demonstrated that, in the murine urothelial carcinoma model, Tregs upregulated genes characteristic for a Th cell phenotype, such as proinflammatory cytokines. However, in this study, Foxp3 expression by the converted Tregs was lost, suggesting a complete reversion of Tregs into Th cells. In our present work, Tregs rather maintained Foxp3 expression but additionally increased expression of markers of Th cells (CD154). CD154 is an important functional mediator of T cell help (36). It has been shown that in draining lymph nodes of vaccinated mice, Tregs can undergo reprogramming and upregulate CD154 (37). In line with this study, we showed that upon anti-CD137 therapy, CD154 was expressed on a subset of Foxp3⁺CD4⁺ T cells. Moreover, the CD154⁺Foxp3⁺ CD4⁺ T cell subset expressed the proinflammatory cytokine TNF-α, supporting the functional change of the Foxp3⁺CD4⁺ T cells. In a recent study by Sharma et al. (38) it has also been demonstrated that in response to specific inflammatory signals, Foxp3⁺CD4⁺ T cells were transformed into biologically important helper cells, without loss of Foxp3. However, it has not been addressed so far whether Foxp3⁺CD4⁺ T cells can undergo conversion that will change their functional capability and turn them into cytotoxic CD4⁺ T cells. Interestingly, Foxp3⁺CD4⁺ T cells also showed increased expression of markers delineating a cytotoxic program (Eomes/granzyme B). We found the emergence of a Treg subpopulation that expresses the cytotoxic transcription factor Eomes while maintaining Foxp3 identity after anti-CD137 treatment. Interestingly, this novel population of CD4⁺Foxp3⁺ T cells coexpressing Eomes could also be found in another tumor model (Fig. 4G) and in chronically FV-infected mice after anti-CD137 therapy (data not shown). Moreover, the reprograming of Tregs into cytotoxic Eomes⁺CD4⁺ T cells was accompanied by granzyme B production (Fig. 2D). Loebbermann et al. has shown that granzyme B–expressing Tregs in the lungs of RSV-infected animals play a critical role in dampening inflammation (39). Their main function was clearly immune suppressive because they killed inflammatory effector T cells (39). However, in that study, the authors did not perform costimulatory treatment. After anti-CD137 stimulation, the subpopulation of granzyme B⁺ Foxp3⁺ cells seems to acquire an effector rather than regulatory phenotype. We show in this study that they kill tumor cells but do not seem to affect T cell responses. Their suppressive effect on conventional T cells cannot be directly tested because their phenotypic characterization is based on intracellular markers, which do not allow isolation as life cells. However, a population of total Tregs that contained ∼15% of Foxp3⁺ Eomes⁺ cells was slightly less suppressive for the proliferation of effector CD8⁺ T cells than a Treg population without this subpopulation (Supplemental Fig. 4A). These results provide indirect evidence that the Foxp3⁺Eomes⁺CD4⁺ T cells may not mediate immunosuppressive effects on conventional T cells and are therefore distinct from the Treg population described by Loebbermann et al. (39). It has recently been shown that granzyme B–producing Tregs can kill tumor target cells (40). However, these cells were redirected by bispecific Abs and classical TCR involvement in specific tumor cell recognition was absent.

Another member of the TNFR superfamily that is transiently expressed on T cells, including Tregs, is CD134. It has been shown that triggering the CD134 pathway impairs the immunosuppressive function of Tregs and results in increased lethality in mice with graft-versus-host disease (41). Another study from the same group demonstrated that in vivo administration of anti-CD134 Abs protects mice from tumor progression by inhibiting the suppressive function of Tregs (42). This is indeed very interesting data for developing new concepts in tumor immune therapy, but the phenotype of Tregs has not been investigated in detail in these studies. Thus, the plasticity of Tregs during treatment with anti-CD134 Abs or other costimulatory molecules has to be determined in future studies. In this work, we provide evidence that CD137 stimulation can induce the production of cytolytic molecules in Tregs and converts these cells into Foxp3⁺ cytotoxic killer cells that contribute to Ag-specific tumor rejection in vivo. Because agonistic Abs to CD137 are in clinical trials for immunotherapy of cancer (43), our study provides new important information about their possible mode of action.

The authors have no financial conflicts of interest.