GITR Pathway Activation Abrogates Tumor Immune Suppression through Loss of Regulatory T-cell Lineage Stability

The immune system is capable of recognizing malignant cells, but in most situations, tumors develop strategies to avoid elimination and escape immune surveillance (1). Recent advances in immunotherapy have succeeded in shifting the balance from tumor immune escape to tumor elimination. Instead of treating the tumor directly by inhibiting cell growth, immunotherapeutic approaches modulate a patient's immune system to induce tumor regression. The success of this approach is highlighted by the U.S. Food and Drug Administration approval of the CTLA-4–blocking antibody, ipilimumab, the first therapy to show enhanced overall survival for patients with melanoma (2). Serving as a proof-of-principle, CTLA-4 blockade has led to targeting of other immune checkpoints (PD-1/PD-L1) alone or in combination with CTLA-4, with very promising results in early-phase clinical trials (3–6). Although coinhibitory receptor blockade has shown durable clinical efficacy, a significant number of patients (∼50%–80%) remain refractory to these treatments and some tumor types do not respond as robustly as others (3, 7). To further potentiate antitumor immune responses and extend clinical benefit, activating costimulatory molecules, such as TNF receptor (TNFR) superfamily members GITR (glucocorticoid-induced TNF receptor-related gene), OX40, and 4-1BB, represent a logical next step (8, 9).

GITR became an attractive target for cancer immunotherapy after the agonistic anti-GITR antibody DTA-1 was shown to block the suppressive effects of regulatory T cells (Treg; ref. 10). Subsequently, DTA-1 was shown to enhance tumor immunity in a concomitant immunity model of melanoma. In addition to preventing growth of secondary tumor challenges, DTA-1 treatment also caused the regression of some of the primary tumor challenge (11). This observation has been extended into multiple tumor models and various combinatorial strategies with vaccines, adoptive T-cell transfer, and concurrent CTLA-4 blockade (12). With preclinical success of GITR tumor immunotherapy, it has been entered into early-phase clinical trials for the treatment of advanced malignancies. Despite its therapeutic potential, the mechanism of action on Tregs as opposed to effector T cells (Teff) has not been fully elucidated. Understanding its activity on Tregs is a necessary step to inform the effective use of GITR therapy in humans.

Whether or not GITR immunotherapy targets GITR solely on Teffs, or on both Teffs and Tregs, has been an area of investigation. Because GITR is constitutively expressed at high levels on Tregs, it was assumed that DTA-1 directly inhibited Treg-suppressive function in vitro (10). However, GITR is also upregulated on CD4 and CD8 Teffs following activation and acted as costimulatory receptor (13). Through the use of GITR^−/− Tregs, it was determined that the costimulatory role of GITR enabled Teffs to resist Treg suppression while having no direct effect on Tregs (14). Thus, initial reports of enhanced tumor immunity resulting from GITR ligation by agonist antibody DTA-1 were attributed to the modulation of Teffs (15, 16). Nevertheless, we and others have recently shown that direct modulation of Tregs is an important consequence of DTA-1 therapy (17, 18). DTA-1 treatment causes more than 50% reduction of intratumor Tregs and down modulation of Foxp3. In addition, the effects of DTA-1 are attenuated if either Teffs or Tregs is GITR^−/− (17). Our data suggest that the efficacy of DTA-1 comes not only from its effect on Teffs, but also from its modulation of Tregs.

Here, we show that GITR ligation by DTA-1 induces intratumor Treg lineage instability. DTA-1 causes loss of Foxp3 in a tumor-dependent manner and is preceded by the loss of the transcription factor Helios. This results in the acquisition of a Th1 effector-like profile and prevents Treg-mediated intratumor suppression of the antitumor immune response. Our results show that modulation of Tregs, along with Teffs, is important and necessary for the efficacy of GITR immunotherapy.

C57BL/6: CD45.1, Thy1.2⁺, Thy1.1⁺, and OT-1 TCR transgenic mice were obtained from Jackson Laboratory. Pmel-1 T-cell receptor transgenic mice were a gift from Dr. Nicohlas Restifo [National Cancer Institute (NCI), Bethesda, MD]. Foxp3-GFP knockin mice were a gift from Dr. A. Rudensky [Memorial Sloan-Kettering Cancer Center (MSKCC), New York, NY]. GITR^−/− and GITR^+/+ littermates (Sv129 × C57BL/6 background) were a gift from Dr. P.P. Pandolfi (MSKCC) and were backcrossed more than 10 generations and onto Pmel-1 Thy1.1+ C57BL/6 background using a speed congenic system. Mice were maintained according to NIH Animal Care guidelines, under a protocol (# 96-04-017) approved by the MSKCC Institutional Animal Care and Use Committee.

B16F10/LM3 (hereafter called B16) is derived from the B16F10 line provided by I. Fidler (MD Anderson Cancer Center, Houston, TX), and transfected with OVA to generate B16-OVA (19). Tumor cells were cultured in RPMI-1640 medium containing 7.5% FBS (for up to 2 weeks after thawing). Each mouse received 150,000 cells in 150 μL of growth factor–reduced Matrigel (BD Biosciences) injected subcutaneously. Four days after tumor challenge, mice were injected intraperitoneally with either 1 mg of affinity-purified DTA-1 or Purified Rat immunoglobulin G (IgG; Sigma-Aldrich) in 500 μL PBS.

Spleens, tumor-draining lymph nodes (TDLN), and tumors were excised on days indicated in the text. Tumors were weighed, and then tissue was homogenized through 40-μm strainers to produce single-cell suspensions. Red blood cells were lysed from spleens using an ACK lysis buffer (Lonza). Cells were washed with media, and tissue cell counts were calculated using Guava cell counter (Millipore). Cells were then either sorted for Tregs, stained immediately by fluorescence-activated cell sorting (FACS) or for cytokine recall, stimulated with phorbol 12-myristate 13-acetate (PMA) and ionomycin for 4 hours, and then treated with monensin before FACS staining.

Anti-GITR (DTA-1, S. Sakaguchi, Osaka University, Osaka, Japan) and anti-OX40 (OX86, A. Weinberg, Earle Chiles Research Institute, Portland, OR) were produced by the MSKCC Monoclonal Antibody Core Facility, and anti-4-1BB (LOB12.3) was procured from Bioxcell. Foxp3 Staining Kit (eBioscience) was used for intracellular staining. Antibodies to antigens listed in figures were from BD Biosciences except Foxp3 (eBioscience), Helios, CD45.2 (Biolegend), and Nrp1 (R&D systems). Dead cell exclusion was done using the Aqua LIVE/DEAD Fixable Dead Cell Stain Kit (Invitrogen). Samples were acquired on 12-color LSRII cytometer, and analyzed using FlowJo (Tree Star).

Tumor-experienced or naïve Foxp3-GFP Tregs were isolated from spleens and TDLNs of untreated Foxp3-GFP mice bearing B16 tumors 7 to 8 days after tumor challenge, or non–tumor-bearing. CD4⁺ GFP⁺ Tregs were isolated by enriching CD4⁺ cells by CD4-positive or -negative MACS microbead separation kits (Miltenyi) before sorting for GFP expression on a Cytomation MoFlo or BD FACS Aria cell sorter in the MSKCC Flow Cytometry Core Facility. For cotransfer experiments (Fig. 1C), naïve Tregs were isolated from Thy 1.1⁺ C57BL/6 mice using MACS microbead Treg isolation kit (Miltenyi). Tregs were then injected intravenously (5–7 × 10⁵cell/mouse for each Treg type being transferred) in 200 μL of sterile PBS.

The collagen-fibrin gel-killing assay is described in depth by Budhu and colleagues (20) and was adapted for ex vivo tumors. Briefly, B16-Ova tumors isolated on day 10 or 11 after tumor challenge were cut into small pieces, incubated for 5 minutes in 250 μg/mL collagenase in PBS containing Ca²⁺Mg²⁺, and homogenized through 100-μm cell strainers to create single-cell suspensions. Viable tumor cells and tumor-infiltrating lymphocytes were counted by Trypan blue exclusion. A total of 10⁴ viable tumor cells, together with all infiltrating cells, were coembedded into collagen-fibrin gels with or without 1 to 5 × 10⁵ CD8⁺ T cells activated in vitro by cognate peptide + interleukin (IL)-2. Duplicate gels were lysed every 24 hours for 3 days, and viable remaining tumor cells were diluted and plated in 6-well plates for colony formation. Seven days later, plates were fixed with 3.7% formaldehyde and stained with 2% methylene blue before counting as described by Budhu and colleagues (20).

Killing constant k is calculated as described in ref. (20). Briefly, k is calculated according to the following equation: b_t = b₀e^−kpt+gt, where bt = the concentration of B16 cells at time t; b0 = the concentration of B16 cells at time 0; k = the killing rate constant (or killing efficiency) for CD8 T cells; p = the concentration of in vitro activated CD8 T cells; g = the growth rate constant for B16 cells.

Individual tumors and pooled control spleens were collected and stained with anti-CD45, anti-CD4, and 4′, 6-diamidino-2-phenylindole (DAPI) before CD45⁺ CD4⁺ GFP⁺ Treg or GFP⁻ Teff control, and were FACS sorted on BD FACS Aria directly into Trizol reagent (Invitrogen). Total RNA was prepared and reversed transcribed into cDNA using a High Capacity cDNA Reverse Transcription Kit (Applied Biosystems). The primer-probe sets were from TaqMan Gene Expression Assays (Applied Biosystems). Quantitative real-time PCR reactions were prepared with FasStart Universal Probe master (Rox) mix (Roche) according to the manufacturer's instructions and done using the ABI 7500 Real-Time PCR system (Applied Biosystems). Each gene was amplified in duplicate and repeated in two separate experiments. cDNA concentration differences were normalized to glyceraldehyde—3—phosphate dehydrogenase (GAPDH). Relative gene expression of the target genes was calculated by the formula 2 – ΔC_t [ΔC_t = C_t (target gene) – C_t (GAPDH)].

We previously showed that optimal GITR agonist antibody DTA-1 treatment of early established (day 4) B16 tumors caused intratumor Tregs to lose Foxp3 expression. By treating tumors grown in mice where Tregs express GFP fused in frame to Foxp3 (Foxp3-GFP), we were able to detect remnant GFP in former Tregs, after Foxp3 had been degraded (Fig. 1A). GFP remains, whereas Foxp3 is degraded, because it is less susceptible to proteolytic degradation (Fig. 1A; ref. 17). DTA-1–induced Foxp3 loss is not seen in peripheral tissues, suggesting that entry into the tumor microenvironment promotes Treg instability and increases susceptibility to modulation (17). In addition, adoptively transferred Tregs sorted from spleens and TDLNs of tumor-bearing Foxp3-GFP mice lose Foxp3 expression within 48 hours of infiltrating tumors in DTA-1–treated hosts (17). To determine what renders Tregs susceptible to GITR modulation, we used the adoptive transfer system to track the highly purified previously untreated Tregs and probe the specific conditions permitting DTA-1–induced Foxp3 loss.

Tregs have been described to contain a minor population that is less stable and characterized by low CD25 expression. This population is susceptible to Foxp3 loss after long-term transfer (4 weeks) into Rag^−/− hosts (21). It has recently been shown that stimulation with Fc-GITR-L can augment Foxp3 loss after CD4⁺ T-cell transfer into a RAG^−/− model of inflammatory bowel disease (22). Therefore, we first asked whether during our short-term (48 hours) adoptive transfer conditions, DTA-1 exclusively modulates only the minor CD25 low population (which at most accounts for only ∼10% of Tregs). Consistent with published reports, there is a slight loss in the percentage of Foxp3⁺ Tregs in control IgG-treated mice 2 days after transfer (∼90% pretransfer to ∼77% after transfer, Fig. 1B). In contrast, DTA-1 treatment induced a pronounced reduction in Foxp3⁺ Tregs (∼90% Foxp3⁺ pretransfer to ∼10% Foxp3⁺ after transfer, Fig. 1B). This translates to an average of a 6 (± 0.58 SEM)-fold decrease in the percentage of Foxp3⁺ transferred Tregs in DTA-1–treated hosts compared with controls. These data confirm that DTA-1 has the potential to modulate all Tregs and is not restricted to the minor, unstable CD25^low Treg fraction.

Having established the effect of DTA-1 on Treg Foxp3 loss in lymphopenic conditions, we next assessed how the presence of the tumor and/or tumor infiltration affects Treg vulnerability to DTA-1. To accomplish this, we cotransferred congenically marked (with CD45 and Thy1) Tregs isolated from spleens and TDLNs of tumor-bearing (CD45.2⁺) or naïve (Thy1.1⁺CD45.2⁺) donors into lymphoreplete naïve or tumor-bearing recipients (CD45.1⁺) following the scheme in Fig. 1C. Naïve Thy 1.1 donor Tregs transferred into naïve recipients displayed negligible loss of Foxp3 in peripheral tissues 48 hours after transfer [88% Foxp3⁺, pretransfer (Fig. 1C) vs. ∼86%–91% Foxp3⁺, in IgG (Fig. 1D)]. DTA-1 treatment induced a maximum of 17% to 18% reduction in naïve Foxp3⁺ Tregs under these conditions (Fig. 1D, spleen IgG vs. DTA-1). In contrast, tumor-experienced CD45.2⁺ Tregs transferred into control IgG-treated hosts displayed a greater loss of Foxp3 expression [94% Foxp3⁺ pretransfer (Fig. 1C) vs. ∼72% and ∼48% Foxp3⁺ Tregs in the spleen and lymph nodes, respectively (Fig. 1D)]. DTA-1 treatment enhanced Foxp3 loss in tumor-experienced Tregs, which was most evident in the lymph node, where there was approximately 50% reduction in Foxp3⁺ Tregs in DTA-1–treated animals, as compared with animals treated with control IgG (Fig. 1D). In tumor-bearing recipients, DTA-1 treatment further potentiated the decrease in Foxp3 expression in tumor-experienced Tregs in the spleen {34% vs. 23%, IgG vs. DTA-1, P = 0.025 comparing %foxp3^NEG [(Post Foxp3 purity-Post Transfer Foxp3%)/Post Foxp3 purity] in Fig. 1D and E}. Naïve Tregs only displayed a significant loss of Foxp3 expression upon entering the tumor [(Fig. 1E) 27% drop IgG vs. DTA-1 compared with pretransfer]. Taken together, our data strongly suggest that preconditioning of Tregs in the presence of tumor or in the tumor microenvironment before DTA-1 treatment is important to their susceptibility to Foxp3 loss.

Transferred Tregs do not display cleaved caspase-3, and equal numbers of transferred Tregs are recovered after DTA-1 treatment for both tumor-experienced and naïve Tregs (Supplementary Fig. S1A and S1B). Foxp3 loss in tumor-experienced Tregs sorted by high CD25 expression seemed to be comparable with those sorted by Foxp3-GFP (Supplementary Fig. S1C). Moreover, to address the possibility that Foxp3^NEG Tregs result from DTA-1–induced proliferation of contaminating Teffs, we monitored the proliferation of transferred Teff (CD4⁺ GFP^NEG) sorted from Foxp3-GFP tumor-bearing mice. Supplementary Figure 1D shows that CD4⁺ GFP^NEG Teffs (sorted from Foxp3-GFP tumor-bearing mice) do not proliferate or accumulate after transfer and DTA-1 treatment. In sum, these data indicate that Foxp3^NEG Tregs come directly from the Foxp3⁺ Tregs, the frequency of which is not reduced as a consequence of cell death, depletion, or proliferation of contaminating Teffs.

In addition, we found that the DTA-1–induced Foxp3 loss occurs in a dose-dependent manner (Supplementary Fig. S1E). Interestingly, agonist antibodies to GITR-related TNFR family members, 4-1BB and OX40, did not affect the frequency of Foxp3⁺ Tregs (Supplementary Fig. S1F). Thus, this effect seems to be uniquely associated with GITR stimulation.

The data above suggest that lymphopenic conditions and the presence of tumor sensitize Tregs to the effects of DTA-1. In addition, the data imply that DTA-1 has the ability to modulate a large percentage of the Treg population, which remains viable after loss of Foxp3. Therefore, we hypothesized that in the tumor therapy setting, even the intratumor Tregs that maintain Foxp3 expression after DTA-1 treatment would be affected by GITR stimulation. In fact, we have previously shown that Foxp3 expression in the remaining Tregs is significantly lower in DTA-1– versus control IgG-treated tumors, supporting this concept (17). To better understand the outcome of DTA-1–induced Treg instability, we investigated whether there were changes in other markers associated with Treg stability, function, and/or ontogeny such as the transcription factor Helios, and expression of the cell surface VEGF coreceptor neuropilin 1 (Nrp1). Expression of Nrp1 has been reported to distinguish between thymus-derived (tTreg) and peripherally derived Tregs (pTreg) and is important for Treg trafficking to B16 tumors (23–25). Although the exact role of the Ikaros family transcription factor Helios remains unresolved, it has been described as a marker of Treg activation and identifies the most suppressive population of tumor-infiltrating Tregs (26, 27). In control animals, intratumor Tregs are uniformly Helios^HIGH with a majority being Nrp1^HIGH, at the peak of B16 immune infiltration compared with peripheral Tregs (spleen, 10–11 days after tumor challenge; ref. 28), suggesting a highly activated tTreg phenotype (Fig. 2A; refs. 25, 27). In contrast, DTA-1 treatment causes a clear loss of Helios expression in the remaining Foxp3⁺ intratumor Tregs (Fig. 2A). Nrp1 expression did not seem to be as significantly affected as Helios, which may be related to its role in Treg trafficking (24).

Using changes in Helios expression as a surrogate marker to identify DTA-1–modulated Tregs in addition to Foxp3 loss, we expanded our analysis to early phases of Treg tumor infiltration to determine the kinetics of Helios loss and its possible correlation with Treg survival and function. At day 7 of tumor growth (3 days after DTA-1 treatment), there is already a significant increase in the Helios^LOW Treg cell population (∼18% compared with ∼55% in IgG vs. DTA-1 treatment, respectively, Fig. 2B, left). By day 10, there is an approximately 55% to 60% loss of Foxp3⁺ Tregs (Supplementary Fig. 2A), and the remaining Foxp3⁺ Tregs (∼45%–50%) are Helios^LOW in DTA-1–treated tumors. Taken together, these data suggest that approximately 75% to 80% of the Tregs (compared with control IgG) in the tumor have been modulated by DTA-1.

Helios^LOW Tregs in DTA-1–treated tumors express more of the prosurvival genes BCL-2 and BCL_XL than Helios^HIGH or total IgG Tregs (Fig. 2B). This phenotype extends to the peak of immune infiltration at day 10, but by day 14, even though the tumors are regressing, the majority of the remaining Tregs are Helios^HIGH (Fig. 2B). Tregs with the lowest levels of Helios at day 7 also displayed a pronounced reduction of Foxp3 and CD25 (Fig. 2C). Helios loss seems to parallel the extent to which free cell surface GITR is saturated/modulated by DTA-1, preventing further staining on Tregs (Fig. 2D). At day 14, the 1 mg/mouse dose of DTA-1 no longer saturates available GITR, and intratumor Tregs in DTA-1–treated mice display similar Helios expression compared with IgG (day 14 posttumor challenge in Fig. 2B and D). This supports the conclusion that Tregs lose Foxp3 expression after tumor infiltration, with gross changes in Helios expression being a reliable marker of GITR modulation. Increased expression of BCL-2 and BCL_XL, combined with the lack of activated caspase-3 (Supplementary Fig. S1), shows that modulated Tregs maintain a prosurvival phenotype.

Tregs naturally co-opt and express inflammatory T-cell lineage transcription factors (T-bet, RORγt) to facilitate the suppression of the corresponding Teff program (29, 30). When Foxp3 expression is ablated in Tregs, they have been shown to revert to cells with a Teff phenotype. In addition, Helios has been shown to stabilize Treg programming and suppress IL-2 expression (31, 32). Therefore, DTA-1–treated Tregs that have lost or are losing Foxp3 expression could acquire a Teff-like phenotype. In contrast with the reduced expression of Foxp3 and CD25 in DTA-1–treated mice, Helios^LOW Tregs showed increased protein expression of T-cell lineage transcription factors such as T-bet, RORγt, and Eomes, compared with Helios^HIGH Tregs (Fig. 3A). Expression in Helios^LOW Tregs was also higher than in Tregs in the IgG control groups at multiple time points (day 7 for T-bet, days 7–14 for Eomes, days 10 and 14 for RORγt; Fig. 3A).

To determine whether increased T-bet, RORγt, and Eomes protein levels in Tregs has biologic consequence, we conducted a cytokine recall assay on cells isolated from tumors 10 days after DTA-1 treatment. Foxp3-GFP mice were used for this experiment because the staining for Foxp3 and Helios is diminished and unreliable after PMA/ionomycin stimulation (Supplementary Fig. S2B). Using Foxp3-GFP mice also allowed us to circumvent this technical hurdle as low levels of Foxp3 expression correlate with loss of Helios (Fig. 2C), allowing us to subset our analysis to Foxp3-GFP^LOW and Foxp3-GFP^HIGH Tregs. GFP^LOW Tregs (Helios^LOW) in DTA-1–treated mice showed a more than 2-fold increase in IFN-γ production compared with control IgG-treated Tregs (Fig. 3B). Although there was no difference in the IFN-γ expression between GFP^HIGH and GFP^LOW cells in IgG control tumors (Fig. 3B, top), IFN-γ expression was restricted to GFP^LOW in DTA-1–treated Tregs (Fig. 3B, bottom). Despite increased RORγt expression in Helios^LOW Tregs, we did not detect any significant difference between IgG and DTA-1–treated Tregs in its related cytokine IL-17 (data not shown). To confirm this result and more closely measure the changes in Treg lineage phenotype, we sorted Foxp3-GFP Tregs from individual tumors and measured the expression of relevant Tregs and Teff genes. Using this approach, we found a maximum of 4-fold upregulation in IFN-γ expression (day 7) and approximately 2-fold decrease in IL-10 expression (day 10) in DTA-1–treated Tregs (Fig. 3C). Other markers, such as GITR, IL-2, IL-17, TNF-α, TGF-β, and SATB1, were expressed to equivalent levels in DTA-1–treated and IgG-treated Tregs (data not shown). Although Helios protein levels after DTA-1 treatment correlated with reduced Helios gene expression, there was no major difference in Foxp3 gene expression (Fig. 3D). This would indicate that GITR signaling may cause a posttranscription modification that leads to reduced Foxp3 protein expression. Regardless of the mechanism responsible for the loss of Foxp3 and Helios expression, these results suggest that DTA-1 induces Treg lineage instability and acquisition of a Teff-like profile.

To determine whether the phenotypic changes described above alter Treg-suppressive function in vivo, we used an ex vivo collagen-fibrin gel matrix culture to measure CD8⁺ cytolytic T-cell (CTL) effector function against tumor cells from control IgG- or DTA-1–treated mice (20). Collagen-fibrin gels mimic a three-dimensional tissue-like environment and are more sensitive than packed cell-pellet assays at measuring CD8⁺ CTL effector function (20). Furthermore, we have found that collagen-fibrin gel cultures of explanted B16 or B16-expressing OVA (B16-OVA) tumors, which include all infiltrating cells, are resistant to killing by a 10- to 50-fold excess of in vitro cognate antigen-activated CD8⁺ CTL, recapitulating the suppression that exists in vivo (Fig. 4A; and Budhu and Schaer; unpublished data).

Consistent with prior results, control IgG-treated tumors become resistant to killing by in vitro activated CTLs and proliferate in the collagen gels after 24 hours, with the number of tumor cells increasing overtime (Fig. 4A; Budhu and Schaer; unpublished data). In contrast, DTA-1 treatment caused tumors to remain susceptible to ex vivo killing by activated CTLs, and the number of viable tumor cells continued to decrease at 48 and 72 hours (2-fold and 3-fold, respectively, vs. 0 hour; Fig. 4A). Calculation of the killing efficiency, k (as described in Materials and Methods and in ref. 20) highlights the differences between DTA-1– and control IgG-treated tumors. Killing efficiency of CTLs in DTA-1–treated tumors increases over 2-fold at 48 hours (5.3 × 10⁻¹⁰ at 24 hours to 1.3 × 10⁻⁹ at 48 hours; Fig. 4A) in contrast with that in IgG-treated mice, which maintains suppression. Ex vivo addition of DTA-1 had no effect on the killing of DTA-1–treated tumors, control IgG-treated tumors, or cultured B16 cells, and GITR^−/− CTL killed tumor cells from DTA-1–treated tumors and cultured B16 cells at the same rate as GITR^+/+ CTL (Fig. 4B, dashed lines and green lines vs. red lines). This suggests that killing is independent of GITR stimulation by DTA-1 on CTL (Fig. 4B). Combined, our data support the conclusion that GITR modulation of Tregs by DTA-1 removes their suppressive influence in the tumor microenvironment.

The overarching goal of cancer immunotherapy has been the activation of tumor-specific immunity that is able to overcome the hurdles established by tumors to evade immune destruction. GITR activation seems to reach an important balance by enhancing tumor immunity while inhibiting immune suppression in a tumor-dependent manner. The research presented here shows that in addition to its established role in modulating Teffs, DTA-1 treatment causes Tregs to lose lineage stability, reducing their suppressive influence over the tumor microenvironment.

Our data suggest that conditions present in tumor-bearing mice and the tumor microenvironment are responsible for making Tregs susceptible to GITR-induced Foxp3 loss. Reduced IL-2 levels have been shown to be important for Treg stability and homeostasis (33, 34). However, we do not believe that the lack of IL-2 accounts for Treg instability in our system because transferred Tregs lose Foxp3 in the periphery even after transfer into lymphoreplete hosts. In addition, equal numbers of cotransferred tumor-experienced and naïve Tregs are recovered from DTA-1–treated animals, despite the loss of Foxp3 expression in tumor-experienced Tregs. This suggests that DTA-1 does not simply deplete Foxp3⁺ Tregs (Fig. 1D and Supplementary Fig. S1B). Only upon tumor infiltration in DTA-1–treated animals do naïve donor Tregs manifest significant Foxp3 loss, highlighting further the role of tumor conditioning on Tregs and even at steady state. Therefore, although the detailed mechanism of GITR signaling-induced Foxp3 loss requires further investigation, it is evident that tumor preconditioning and the tumor microenvironment play a major role in permitting GITR-dependent modulation of Foxp3 expression.

The reduction of CD25 expression and the production of IFN-γ observed in intratumor Tregs during DTA-1 therapy (Figs. 2 and 3) are similar to what has been reported when Foxp3 is deleted in mature Tregs (35). There has been evidence suggesting that inflammatory environments cause Tregs to lose stability and convert to a Teff-like phenotype (29); however, recent research has brought these findings into question. Results from Miyao and colleagues and Zhou and colleagues suggest that the conversion of Tregs into Teffs is actually due to a transient expression of Foxp3 in non-Tregs (36, 37). It is unlikely that the DTA-1–induced Treg lineage conversion we observe here is an artifact of lineage marking. The Treg transfer and gene expression analysis experiments (Figs. 1 and 3) rely on sorting an entire Foxp3-GFP–positive Treg population and do not use a lineage marking Cre recombinase system. In fact, we were unable to use Foxp3-Cre mice due to the “leaky” lineage marking seen during backcrossing to the C57BL/6 background (data not shown). Thus, we believe the results presented here illustrate that DTA-1–mediated GITR stimulation causes tumor-specific reprogramming of Tregs into a Teff-like phenotype. As we were unable to isolate or phenotype repolarized Foxp3⁻ Tregs using the Foxp3-Cre lineage marking mice, it remains to be established whether the conversion of Tregs to a Teff-like profile is necessary or secondary to the loss of Foxp3/suppressive function. Development of complex genetic models would be needed to answer this question and determine whether former DTA-1–modulated Tregs work to potentiate antitumor immunity after losing suppressive capacity.

How DTA-1–induced GITR signaling leads to Foxp3 degradation is an important question. Expression levels of Foxp3 mRNA were comparable between control IgG- and DTA-1–treated mice, but there is a marked reduction in Foxp3 protein levels (Figs. 3B and 2C). This would suggest that downstream signaling from GITR imposes posttranscriptional or posttranslational control of Foxp3 protein expression. Although downstream signaling from GITR induced by GITR-L was recently shown to alter Treg-suppressive function through the activation of c-jun-NH₂-kinase (JNK), it is unclear whether DTA-1 causes a similar effect (38). JNK activation after long-term GITR-L stimulus resulted in reduced Foxp3 mRNA expression to a level that we did not observe with DTA-1 treatment. GITR and TNFR family members use TNFR-associated factor (TRAF) proteins to transmit downstream signals (8, 39). Because many TRAF proteins function as E3 ubiquitin ligases, one hypothesis could be that overstimulation of GITR by DTA-1 could cause an intersection of this cascade with Foxp3 protein and targeting it for degradation. Because intratumor Tregs express less Foxp3 mRNA than peripheral Tregs (Fig. 3B), this may make them uniquely sensitive to GITR-induced degradation of Foxp3.

A propensity to modulate pTregs over tTregs would be a logical assumption considering their unstable nature (29). However, in the case of B16 melanoma, it seems that the majority of intratumor Tregs have a tTreg-like phenotype, as has been seen in 4T1 tumors, and without a minor pTreg population as seen in other tumors (25). In fact, transfer experiments into Rag^−/− mice established that a majority of Tregs can be rendered susceptible to GITR-induced loss of Foxp3. We found a similar result, with 75% to 80% of Tregs modulated in the tumor microenvironment during DTA-1 therapy in wild-type mice (% of intratumor Treg Foxp3 loss + % Foxp3⁺Helios^LOW Tregs; Supplementary Fig. S2A and S2B). This suggests that the effects of DTA-1 are not limited to a minor subset of Tregs, such as pTregs. Regardless, DTA-1 treatment caused Tregs to lose Helios protein and gene expression, corresponding with increased levels of inflammatory T-cell transcription factors, T-bet, RORγt, and Eomes. Treg expression of T-bet or RORγt is not unprecedented, and the expression of these transcription factors is important for the Treg-suppressive function (29). Surprisingly, Eomes, traditionally thought of as a CD8⁺ CTL transcription factor, is highly upregulated in the DTA-1–treated Tregs. We have reported recently that simulation of the closely related TNFR family member OX40 has the ability to induce Eomes in CD4 Teffs (40). Even though there has been evidence that Tregs could control immunity through granzyme-dependent killing of B cells, to date no role for Eomes in Treg function has been described (41). The significance of Eomes expression in DTA-1 modulation of Tregs will require further investigation; however, it exemplifies the level to which overstimulation of GITR on susceptible Tregs can alter their lineage program

The end result of Treg lineage instability caused by GITR immunotherapy is the removal of intratumor suppression mediated by Tregs, as shown by the collagen-fibrin gel killing assay (Fig. 4). Using the same approach, we recently determined that intratumor immune suppression in B16 tumors is Treg dependent, as specific in vivo depletion of Tregs restores killing of explanted tumors (Budhu and Schaer; unpublished data). Whether or not the DTA-1 effect is due to reduced intratumor Treg numbers, Treg lineage instability, or a combination of both remains to be determined. Interestingly, even though GITR treatment removes Treg suppression and DTA-1–treated tumors are regressing in vivo, tumor cells cocultured with total infiltrates continue to grow ex vivo (Fig. 4). We interpret the need for additional input of Teffs to continue killing as evidence that for optimal in vivo therapy, GITR's ability to enhance CD8⁺ T-cell numbers and persistence also plays an important role (42). Consequently, targeting Tregs seems to be a major mechanism for DTA-1 treatment along with its intrinsic effects on CD8⁺ T cells. This conclusion is in agreement with our prior results showing that both Tregs and Teffs must express GITR for the optimal effects of DTA-1 (17).

Development of new immunotherapies that accelerate antitumor immunity is important, as checkpoint blockade does not benefit all patients (2, 3). Our data show that ligation of GITR can accomplish both goals. By inducing Treg lineage instability, DTA-1 releases an important source of suppression of tumor immunity. At the same time, we and others have shown that GITR ligation by DTA-1 accelerates antitumor immunity to take advantage of the now permissive tumor microenvironment (12, 17). The unique ability of GITR ligation to target both axes, modulating Tregs primarily in the tumor microenvironment, supports the continued clinical development of GITR agonist agents. Accordingly, in collaborations with GITR Inc., we are currently investigating the agonist anti-human GITR antibody, TRX-518, in a phase I first-in-human trial (GITR Inc., Clinical trials.gov: NCT01239134). We believe that the knowledge gained from our study in understanding GITR mechanism of action will help facilitate the development of appropriate biomarkers and inform rational design of future clinical trials.

No potential conflicts of interest were disclosed.

Conception and design: D.A. Schaer, C. Liu, A.D. Cohen, A.N. Houghton, T. Merghoub, J.D. Wolchok

Development of methodology: D.A. Schaer, C. Liu, A.D. Cohen, T. Merghoub

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D.A. Schaer, S. Budhu, C. Liu, C.F. Bryson, N.M. Malandro, A.D. Cohen, H. Zhong, X. Yang, T. Merghoub

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D.A. Schaer, S. Budhu, C. Liu, C.F. Bryson, N.M. Malandro, A.D. Cohen, T. Merghoub, J.D. Wolchok

Writing, review, and/or revision of the manuscript: D.A. Schaer, S. Budhu, T. Merghoub, J.D. Wolchok

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): C. Liu, H. Zhong

Study supervision: A.N. Houghton, T. Merghoub, J.D. Wolchok

The authors thank current and former Wolchok Lab members: Dr. Stephanie Terzulli, Andre Burey, Judith Murphy, Kelly Crowley, Rodger Pellegrini, Drs. Arvin Yang, and Francesca Avogadri for their support on the GITR project; Rudensky lab members: Drs. Steve Josefowicz, Rachel Niec, Ashutosh Chaudhry, and Robert Samstein for generous sharing of reagents and always being available for advice and thoughtful discussion about Treg lineage stability; Dr. Joe Ponte for valuable shared insight on the mechanism of GITR immunotherapy; Dr. Michael Curran for assistance with experimental design; members of the MSKCC flow cytometry, and molecular cytology core facilities; and Dr. Roberta Zappasodi for her very helpful comments, critical reading, and editing of this manuscript.

This work was supported by NIH grants R01CA56821, P01CA33049, and P01CA59350 (to A.N. Houghton and J.D. Wolchok), D.A. Schaer was supported by the NIH Clinical Training for Scholar Grant K12 CA120121-01, and received support from the NIH/NCI Immunology Training Grant T32 CA09149-30 and John D. Proctor Foundation: Margaret A. Cunningham Immune Mechanisms in Cancer Research Fellowship Award; Swim Across America; the Mr. William H. Goodwin and Mrs. Alice Goodwin and the Commonwealth Cancer Foundation for Research and the Experimental Therapeutics Center of MSKCC (to J.D. Wolchok).

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.