Characterization of MK-4166, a Clinical Agonistic Antibody That Targets Human GITR and Inhibits the Generation and Suppressive Effects of T Regulatory Cells

The significant clinical benefit obtained from the use of immune checkpoint inhibitors in the treatment of cancer and the durability of such responses has generated tremendous interest in immunotherapy of cancers (1–4). Therapeutic antibodies designed to block the inhibitory feedback mechanisms of CTLA-4 or PD-1 have been approved for the treatment of cancer. In addition, several agonist antibodies that target costimulatory molecules on T cells like GITR, CD27, 4-1BB, CD40, and OX40 are in various stages of pre-clinical or clinical development (5, 6).

Glucocorticoid-induced tumor necrosis factor receptor-related protein (GITR), also referred to as TNFRSF18, is a type I transmembrane protein of the tumor necrosis factor receptor superfamily that is expressed on human T lymphocytes (preferentially on T_regs; refs. 7–10) and natural killer (NK) cells (11–13). Expression of GITR is significantly increased upon T-cell activation, and ligation of GITR provides a costimulatory signal that positively modulates antigen-specific T-cell responses, leading to enhanced cellular and humoral immunity (11, 12, 14). The counter structure, GITR-Ligand (GITR-L), is expressed on antigen-presenting cells (dendritic cells, B cells, and macrophages; refs. 15–18).

Engagement of murine GITR by either an agonistic anti-mouse GITR antibody (clone DTA-1) or GITR-L, in the presence of a primary TCR signal, results in enhanced T-cell proliferation and cytokine production (11, 12). In mice, DTA-1 abrogates T_reg-mediated suppression either by eliminating GITR-expressing tumor-infiltrating T_regs (19, 20) or by causing them to become unstable thereby attenuating their suppressive activity (21, 22). Agonist anti-GITR antibodies can prevent tumor growth or cure established tumors in murine tumor models (21, 23–28), but the question of whether or not human GITR biology parallels that of the mouse remains largely unanswered (29). Several human GITR agonists are currently in early clinical development for the treatment of solid tumors (NCT01239134, NCT01216436, NCT02583165, NCT02628574, NCT02697591, NCT02132754, NCT02437916).

On the basis of the potent antitumor efficacy of DTA-1 in murine models (21, 23–26), we developed the humanized IgG1 agonist anti-GITR mAb MK-4166, focusing our selection on binding to an epitope on human GITR analogous to the epitope that DTA-1 binds on mouse GITR. We show that engagement of GITR by MK-4166 is a high-affinity interaction that enhances TCR-driven in vitro proliferation of human and cynomolgus monkey naïve CD4⁺ T cells similar to the effect of DTA-1 on mouse T cells. In addition, MK-4166 enhanced the proliferation of human tumor-infiltrating T cells. To address the translatability of murine efficacy data to human cancers, we evaluated GITR expression in mouse and human. The expression of GITR on tumor-infiltrating T_regs was comparable between mouse and human despite differences in GITR expression on many peripheral blood populations and tumor-infiltrating CD4⁺ and CD8⁺ T cells. Because T_regs have been shown to play a critical role in antitumor efficacy of DTA-1, we focused on characterization of MK-4166 activity on human T_regs. MK-4166 significantly inhibited the generation of T_regs in mixed lymphocyte reactions (MLR) and decreased the suppressive effects of T_regs. Moreover, MK-4166 induced phosphorylation of NFκB, upregulated ERK pathway genes, and decreased expression of FOXP3 in tumor infiltrating T cells in ex vivo human tumor cultures. MK-4166 did not induce any adverse events when administered intravenously to cynomolgus macaques at a broad range of doses. The safety and tolerability of MK-4166 in patients with advanced solid tumors is currently being evaluated in a phase I study (NCT02132754).

Humanized IgG1, anti-human GITR (MK-4166) and humanized IgG4, anti-human GITR (MK-1248, which has the same CDR regions but significantly reduced Fc effector function) were developed at Merck Research Laboratories, and for certain uses labeled with DyLight650 (Thermo Fisher). Generation of MK-4166 is described in the Supplementary Material. A murinized IgG2a version of rat anti-mouse GITR antibody DTA-1 was generated on the basis of published sequences (30). Antibodies used for flow cytometry are listed in the Supplementary Material.

C57BL/6J, BALB/cAnN, DBA/2J, or C3H/HeN mice (4 weeks old) were purchased from The Jackson Laboratory or Taconic and maintained in specific pathogen-free facilities according to approved IACUC protocols. The cell lines MC38 (NCI, 2015), MB49 (University of Iowa, 2015), B16F10 (ATCC, 2015), LL/2 (ATCC, 2015), TC-1 (Johns Hopkins University, 2015), 4T1 (ATCC, 2015), CT26 (ATCC, 2015), RENCA (ATCC, 2011), EMT6 (ATCC, 2015), CM3 (ATCC, 2016) were all authenticated by IDEXX CellCheck authentication service. The source of each cell line and year of authentication, including Mycoplasma testing, are indicated in parentheses. Tumor cell lines were implanted at passages 4–6 after thawing.

Approximately 200 μL of blood from healthy female C57BL/6J mice or C57BL/6J mice bearing MC38 syngeneic tumors of approximately 100 mm³ was collected by cardiac puncture. Blood from cynomolgus monkeys (Bioreclamation IVT) and patients with cancer (MT Group) was collected into K₂-EDTA tubes. Blood from healthy human subjects was collected as part of the voluntary blood donor program at Merck Research Laboratories. Human buffy coats from healthy volunteers were obtained from the Stanford Blood Center. All blood samples were collected from voluntary donors in accordance with Institutional Review Board (IRB) protocols after obtaining an informed consent. The IRBs for MT Group, Merck volunteer blood donor program, and Stanford Blood Center are Sterling IRB, Merck Research Laboratories Palo Alto IRB, and Panel on Medical Human Subjects, respectively. All human blood and tumor samples were collected in accordance with the Declaration of Helsinki ethical guidelines.

Mouse tumor cell lines were injected subcutaneously into the hind flanks of syngeneic female mice and tumors were harvested when they reached a volume of 100 mm³. Single-cell suspensions were obtained by fine mincing with a scalpel, followed by a 60-minute incubation at 37°C in 2.5 mL of digestion media containing DMEM, 0.208 WU/mL liberase TM (Roche), and 400 U/mL DNase I (Worthington). Human tumor specimens were obtained from commercial sources (MT Group, Cureline) or University of Rochester in accordance with IRB protocols after obtaining informed consent. The IRBs for MT Group, Cureline, and University of Rochester are Sterling IRB, Western IRB, and Research Subjects Review Board, respectively. Single-cell suspensions from tumors were obtained by fine mincing with a scalpel, followed by a 30-minute incubation at 37°C in digestion medium containing 8 mL of RPMI1640 medium, 40 μL of 100 mg/mL Collagenase I (Thermo Fisher Scientific), and 320 μL of 10,000 U/mL DNase I.

GITR variants with E-tag were serially diluted and added to ELISA plates coated with the parental mouse clone of MK-4166. HRP-conjugated anti-E-tag mAb was used as the detection reagent. The reaction was developed with 2, 2'-Azino-di(3-ethylbenzthiazoline-6-sulfonate; ABTS) and read at 405 nm.

Naïve CD4⁺ T cells were isolated from blood by negative selection using EasySep Human Naïve CD4⁺ T-cell Enrichment kits (Stemcell Technologies) per kit instructions. L cells expressing CD32, CD58, and CD80 on their surface were irradiated with 7000 rads, plated at 2.5 × 10⁴ cells per well, and then cocultured with 2 × 10⁴ naïve CD4⁺ T cells in Yssel's medium with 1% human AB serum. The T cells were stimulated with 0.3 ng/mL of anti-human CD3 mAb (clone UCHT1) in the presence or absence of MK-4166. T-cell proliferation was assessed after 4 or 5 days by incorporation of tritiated thymidine (³H-thymidine) added 18 to 24 hours before harvesting. T-cell proliferation in cynomolgus monkeys was assayed similarly, except that a NHP Naïve CD4⁺ T-cell kit (Miltenyi Biotec) was used to enrich naïve CD4⁺ T cells and clone FN-18 at 0.2 ng/mL was used to stimulate T cells. The degree of response to MK-4166 costimulation was variable; 6 of 12 human donors and 5 of 10 cynomolgus monkeys had dose–response curves sufficient to calculate an EC₅₀ value.

Mouse naïve CD4⁺ T cells were isolated from spleen by negative selection using an EasySep mouse naïve CD4⁺ T-cell kit and 2 × 10⁴ cells were stimulated with 100 ng/mL of anti-mouse CD3 mAb (clone 145-2C11) and cultured with or without DTA-1 in complete RPMI-1640. Irradiated mouse splenocytes (2,000 rads, 1 × 10⁵), depleted of T cells using CD90.2 microbeads, were used to provide costimulation and T-cell proliferation was quantified on day 3 using ³H-thymidine incorporation.

Single-cell suspensions of human tumors were layered on 25 mL of Histopaque and centrifuged at 2,000 RPM to remove dead cells and to enrich for TILs. A total of 0.5 × 10⁵ enriched TILs were stimulated with 5 ng/mL of soluble anti-CD3 mAb (clone OKT3) for 7 days in the presence of 10 μg/mL of MK-4166, MK-1248 or corresponding isotype matched control antibodies in complete DMEM. Samples were stimulated in triplicates and proliferation was quantified on day 7 using ³H-thymidine incorporation.

Monocytes enriched from human PBMCs (RosetteSep human monocyte enrichment kit) were cultured in complete DMEM with 10% FBS (SAFC Biosciences), 1,000 U/mL GM-CSF (PeproTech) and 400 U/mL IL4 (R&D Systems) for 7 days to generate monocyte derived DCs (mo-DCs). LPS (0.5 μg/mL) was added to the culture during the last 2 days to mature the mo-DCs. MLR was set up by culturing PBMCs (2 × 10⁶ cells/mL) with γ-irradiated (30 Gy) allogeneic mo-DCs (0.2 × 10⁶ cells/mL) in the presence of IL2 (100 U/mL) and IL15 (5 ng/mL). MK-4166, MK-1248, or isotype control mAb was added to the cultures and the relative abundance of CD4⁺FoxP3^high T_regs was evaluated at day 7 using flow cytometry.

Total CD3⁺ T cells, T_regs and HLA-DR⁺ cells were isolated from PBMCs of the same donor with EasySep Human T-cell isolation kit (Stemcell Technologies), CD4⁺CD25⁺CD127dim/⁻ Regulatory T-Cell Isolation Kit II, human (Miltenyi Biotec) and human anti–HLA-DR MicroBeads (Miltenyi Biotec), respectively. T_effs (Total CD3⁺) were labeled with 1.5 μmol/L CFSE (CellTrace Cell Proliferation Kits, CFSE, Invitrogen) and T_regs were labeled with 1.5 μmol/L CellTrace Violet (CellTrace Cell Proliferation Kits, Violet, Invitrogen). After labeling, cells were counted and cultured at indicated T_eff:nT_reg ratios without the addition of exogenous IL2. T_effs were stimulated with autologous HLA-DR⁺ cells (HLA-DR⁺:T_eff ratio of 2:1) and 2 μg/mL of anti-CD3 (clone OKT3) in the presence of indicated amounts of MK-4166 or control Ab for 4 days. Cells were labeled with anti-CD3, anti-CD4, and anti-CD8, and proliferation of specific T-cell subsets was measured as a decrease in CFSE-labeling intensity by flow cytometry.

Nine-day-old MLR cultures were frozen for future analysis. Upon thawing cells were rested for 24 hours to allow for stabilization of basal phosphoprotein levels. T_regs were then stimulated with either DyLight650-labeled MK-4166, DyLight650-labeled MK-1248 or DyLight650-labeled isotype matched control mAb at a concentration of 5 μg/mL for 5, 10, 15, or 30 minutes. After stimulation, cells were immediately chilled on ice and centrifuged. Supernatant media were aspirated from each well and cells were incubated with a viability dye and phenotypic antibodies. All samples were fixed in 100 μL of 1.5% formaldehyde solution for 15 minutes followed by incubation with 200 μL of ice-cold 100% methanol for 30 minutes. Cells were then washed twice with DPBS, blocked 10 minutes with 2% normal mouse serum, and subsequently incubated for 30 minutes on ice with mAbs specific for FoxP3 (clone PCH101), phosphorylated NFκB p65 (clone 93H1), and phosphorylated Erk1/2 (clone D13.14.4E).

Dissociated tumor cells were cultured in complete RPMI medium supplemented with 62.5 ng/mL of IL2 and 10 μg/mL of either DyLight650-labeled MK-4166 or DyLight650-labeled isotype control antibody at 37°C, with 5% CO₂ in a humidified incubator. After 1 or 7 days in culture, CD4⁺ T cells were sorted using FACSAriaII (BD Biosciences) into individual wells of a 96-well plate containing 5 μL of preamplification buffer with gene-specific primers and probes for RT-PCR as detailed in the Supplementary Material.

On the basis of several reports that indicate that GITR signaling enhances antitumor immunity in mouse models, a humanized IgG1 agonist mAb against human GITR, MK-4166, was developed to treat patients with advanced malignancies. Because the anti-mouse GITR mAb DTA-1 has shown impressive antitumor efficacy in rodent models, the drug candidate was selected to bind to an epitope that is analogous to that of DTA-1. To this end, the amino acid residues important for DTA-1 and MK-4166 binding to murine GITR and human GITR, respectively, were determined by domain swapping and site-directed mutagenesis experiments as detailed in the Supplementary Material. An alignment of the region containing the amino acids identified as important for antibody binding is shown in Fig. 1A. Six out of the 7 key residues were different in mice compared with humans, and when these residues on mouse GITR were replaced with corresponding residues from human GITR, MK-4166 was able to bind the modified mouse GITR, albeit with lower affinity (Fig. 1B). The average affinities of MK-4166 for human and cynomolgus monkey GITR at 25°C as determined by cell-based KinExA were comparable at 5.5 and 7.6 pmol/L, respectively (Supplementary Table S1). MK-4166 did not bind to mouse GITR (Fig. 1B) as determined by ELISA. The affinity of DTA-1 for mouse GITR was determined to be 26 pmol/L (Supplementary Table S1) by cell-based KinExA, and was comparable with that of MK-4166 for human and cynomolgus monkey GITR.

A dose-dependent augmentation of T-cell proliferation was observed when human or cynomolgus monkey naïve CD4⁺ T cells were costimulated with anti-CD3 and L-cells in the presence of MK-4166 (Fig. 2A and B). A dose-dependent enhancement of murine T-cell proliferation was observed in an analogous in vitro proliferation assay using splenic naïve CD4⁺ T cells costimulated with T-cell–depleted, irradiated splenocytes and anti-CD3 in the presence of DTA-1 (Fig. 2C). The bioactivity of MK-4166 was comparable in human and cynomolgus monkey T cells, with median EC₅₀ values of 7.7 and 7.9 pmol/L, respectively (Supplementary Table S2). The median EC50 value for DTA-1 in mouse T cells was 99 pmol/L (Supplementary Table S2). Considering the relative high potency (picomolar range) of both MK-4166 and DTA-1, and the variability attributable to the differences between rodent and primate assay formats, the bioactivities of both mAbs were deemed fairly comparable.

The ability of MK-4166 or MK-1248 (which has the same CDRs as MK-4166, but is an IgG4 with a different FcγR binding profile and minimal Fc effector function) to costimulate human TILs was determined in single-cell suspensions obtained from NSCLC tumor tissues. In the presence of either MK-4166 or MK-1248, increased proliferation of anti–CD3-stimulated TILs was observed (Fig. 2D).

Translational approaches in pursuing GITR as a target for immunotherapy require knowledge of GITR expression in healthy and tumor-bearing mice, non-human primates (NHP), healthy human donors, and cancer patients. GITR expression was observed on blood CD4⁺ T cells, NK cells, and NKT cells in healthy human donors and cancer patients, but was very low to undetectable on CD8⁺ T cells, B cells, monocytes, and granulocytes (Fig. 3A and B). Among the CD4⁺ T-cell subsets evaluated, T_regs, TH₁₇ and effector memory T cells (T_EM) had the highest frequencies of GITR⁺ cells (Fig. 3C; gating scheme depicted in Supplementary Fig. S1). In addition, we observed that the upregulation of GITR upon activation of sorted CD4⁺ T-cell subsets was highest on T_EM, followed by central memory (T_CM) and then naïve T cells (Fig. 3C and D). In cynomolgus monkeys, GITR expression was similar with the exception of NK cells, which did not express GITR (Fig. 3A). The discrepancy in expression of GITR on human and cynomolgus monkey NK cells was further confirmed by gene expression analysis of sorted NK cells (Supplementary Fig. S2A and S2B). In marked contrast, the vast majority of murine CD4⁺ T cells, CD8⁺ T cells, NK cells, and NKT cells expressed GITR (Fig. 3A). Moreover, a large proportion of B cells and a subset of monocytes and granulocytes also expressed GITR in mice (Fig. 3B).

Expression of GITR on CD4⁺ and CD8⁺ TILs obtained from non–small cell lung carcinoma (NSCLC), melanoma, and renal cell carcinoma (RCC) tumor tissues was evaluated by flow cytometry and compared with that on mouse TILs from 10 syngeneic mouse tumor models (MC38, MB49, LL/2, B16F10, TC1, RENCA, 4T1, CT26, EMT6, and CM3). Similar to the profile in peripheral blood, a significant difference in the frequency of GITR⁺ TILs was observed between human and mouse tumors (Fig. 4A and B). In the 3 human tumor types analyzed, GITR was expressed on 22% to 42% of CD4⁺ TILs, whereas it was typically expressed on less than 11% of the CD8⁺ TILs (Fig. 4A). In contrast, GITR was expressed on nearly 100% of CD4⁺ and CD8⁺ TILs in all analyzed syngeneic mouse tumor models (Fig. 4B). Both, frequency (Fig. 4C; Supplementary Fig. S3A and S3B) and intensity (Fig. 4D) of GITR expression were considerably higher on T_regs compared with non-T_reg CD4⁺ TILs from NSCLCs and were similar to GITR expression on T_regs infiltrating mouse tumors (Fig. 4E and F). Furthermore, almost all of the tumor-infiltrating CD4⁺CD25⁺ cells expressed high levels of FoxP3 protein (Supplementary Fig. S4A), high mRNA levels of T_reg markers such as Helios and Eos (Supplementary Fig. S4B) and were able to suppress proliferation of CD8⁺ TILs isolated from NSCLC tissues by cell sorting (FACS; Supplementary Fig. S4C), thus confirming their identity as T_regs.

Since T_regs have been shown to play a central role in antitumor efficacy of DTA-1 and the expression of GITR was the highest on human intratumoral T_regs as compared with the other TIL populations, we focused on characterization of the ability of MK-4166 to affect the induction of human T_regs and their suppressive effects in vitro.

The effect of MK-4166 on the induction of T_regs (iT_regs) in MLR cultures was assessed. The iT_regs were identified as CD4⁺ CD25⁺ FoxP3^high and the expression of GITR in this population was confirmed by flow cytometry (Supplementary Fig. S5A and S5B). Because activated T cells can express low levels of FoxP3, we confirmed that the iT_regs were functionally suppressive (Supplementary Fig. S5C and S5D). The addition of MK-4166 to MLR cultures on day 0 resulted in a dose-dependent decrease in abundance of iT_regs after 7 days, measured either as relative abundance (Fig. 5A) or as absolute numbers (Supplementary Fig. S6A). This decrease in iT_regs was observed only when MK-4166 was present from the start of culture. When it was added on day 7 after iT_regs were already established, the effect was not observed, indicating that the decrease in the number of iT_regs is due to lack of induction as opposed to loss of established iT_regs (Supplementary Fig. S6B and S6C). A similar dose-dependent decrease in the induction of iT_regs was observed with MK-1248 (Fig. 5B), indicating that Fc effector functions do not likely play a role in this assay, even though MK-4166 does demonstrate the potential to induce ADCC in human MLR-derived iT_regs using a Jurkat reporter cell line (Supplementary Fig. S7A and S7B).

To determine whether MK-4166 could reduce the suppressive effects of human natural T_regs (nT_regs), a nTreg suppression assay was used in which donor-matched CD4+CD25+CD127^low nT_regs and CD4⁺ and CD8⁺ T_effs isolated from blood were stimulated with anti-CD3 and autologous HLA-DR⁺ cells and proliferation measured as dilution of CFSE (Fig. 5C and D; Supplementary Fig. S8A and S8B). This was further confirmed in independent experiments using a MLR suppression assay where T cells were stimulated with allogeneic-DCs and T-cell proliferation was tracked using ³H-thymidine incorporation (Supplementary Fig. S8C). The purity of nT_regs (CD4+FoxP3+CD127^low) used in the nT_reg suppression assay was >85% (Supplementary Fig. S9A) and in the MLR suppression assay was 40 to 70% (Supplementary Fig. S9B). Addition of MK-4166 increased T-cell proliferation compared with isotype-matched controls at several T_eff:nT_reg ratios tested (Fig. 5C and D), indicating that MK-4166 partially attenuates the suppressive effects of T_regs. MK-4166 did not enhance proliferation of T_effs alone (T_eff:nT_reg ratio1:0; Fig. 5C and D; Supplementary Fig. S8A–S8C), indicating that there was no costimulatory effect of GITR agonism in this assay. In addition, MK-4166 did not enhance the proliferation of nT_regs in an independent experiment where nT_regs were stimulated by anti-CD3/anti-CD28–coated beads (Supplementary Fig. S10A and S10B).

Published reports indicate that GITR signaling is mediated by NFκB and MAP kinases p38, JNK, and ERK in mice (31–35). To determine whether binding of the agonist mAb, MK-4166, to GITR elicits similar early signaling events, phosphorylation of MAPK/Erk and NFκB was evaluated by flow cytometry. MK-4166 and MK-1248 bound at similar levels to T_regs (CD4⁺FoxP3⁺) or to T_effs (CD4⁺FoxP3⁻) as determined by flow cytometry (Fig. 6A). Following stimulation with MK-4166, a rapid induction of phospho-NFκB p65 was observed in MLR-derived GITR⁺ CD4⁺ T cells but not in GITR⁻ CD4⁺ T cells (Fig. 6B). Maximal phosphorylation of NFκB p65 was observed at 5 minutes following engagement of GITR by MK-4166, and this time point was chosen for further analysis. Increases in phosphorylation of NFκB p65 were observed in both MLR-derived and intratumoral T_regs and T_effs (Fig. 6C) following stimulation with MK-4166 but not with isotype control mAb. DyLight650-labeled MK-4166 was used in these experiments to identify T cells that bound to MK-4166 (GITR expressing) and T cells that did not (GITR non-expressing). Within each sample stimulated with DyLight650-labeled MK-4166, T cells that expressed GITR had significantly higher phospho-NFκB levels compared with T cells that did not (Fig. 6D). Furthermore, all of these results were reproduced with MK-1248, indicating that GITR signaling is independent of isotype in this assay (Fig. 6C and D). Phosphorylation of Erk1/2 in CD4⁺ T cells was not detected in this assay upon addition of MK-4166 (data not shown).

To determine the downstream effects of GITR signaling, human tumor single-cell suspensions were cultured either with DyLight650-labeled MK-4166 or DyLight650-labeled isotype matched control mAb. CD4⁺ T cells were sorted after 24 hours or 7 days, and gene expression was analyzed by RTqPCR. Even though phospho-Erk1/2 protein was not detected in TILs from NSCLCs, an increase in the expression of DUSP6, a gene induced by the Erk signaling pathway (36), was observed in CD4⁺ T cells in RCC and colorectal carcinoma after 24 hours of incubation with DyLight650-labeled MK-4166 (Fig. 7A). Cells from NSCLC were not available for this experiment. Some reports indicate that GITR signaling among murine intratumoral T_regs decreases expression of FoxP3 and causes instability of T_reg lineage commitment (21, 22). We also observed a decrease in FOXP3 mRNA in CD4⁺ T cells sorted from human RCC and NSCLC tumor cultures treated with MK-4166 for 7 days (Fig. 7B).

Potential toxicity of MK-4166 was characterized in a 1-month study in cynomolgus monkeys with a 2-month post-dosing monitoring period, where a broad range of doses (0.03, 1, 30, or 200 mg/kg/dose) was administered once weekly by intravenous infusion. MK-4166 was well tolerated at all doses and no treatment-related toxicity was detected. Details can be found in the Supplementary Materials.

Increased understanding of the immunosuppressive mechanisms in cancers has identified several molecular pathways, including members of the TNF/TNFR family, as potential targets for anti-cancer therapies (6). In particular, the agonist anti-GITR antibody DTA-1 has impressive antitumor efficacy in murine syngeneic tumor models (21, 23–26). On the basis of these observations, we developed MK-4166, a humanized agonist mAb against human GITR for the treatment of solid tumors. To take advantage of the robust efficacy seen with DTA-1, we ensured that MK-4166 and DTA-1 bound to highly analogous epitopes in the respective species. This innovative approach of ensuring that the clinical candidate binds to an analogous epitope as that of its mouse surrogate enabled us to make translational comparisons between the observations in the two species.

Although the role of GITR is studied extensively in murine models, there is a dearth of information on the role GITR plays in humans (29), and the experiments reported here not only provide insights on the translatability of observations made in murine models, but extend our understanding of GITR biology in humans. DTA-1 has been shown to deplete intratumoral T_regs in vivo (20) and to decrease the induction of T_regs in murine splenocytes treated in vitro with TGFβ (21). In contrast, GITR agonism (with DTA-1 or mouse GITRL) has been reported to enhance proliferation of T_regin vitro and in vivo (37, 38). This suggests that the effects of GITR agonism can be context dependent (12). It is unclear whether our observation that MK-4166 did not enhance nT_reg proliferation in vitro is due to conditions of testing or to species-specific differences. We found that the expression of human GITR is comparable to that of mouse GITR in tumor infiltrating T_regs, despite being drastically lower in other TIL populations and in peripheral blood. MK-4166 decreased the number of T_regs induced in a MLR culture. This is likely not due to ADCC because the decrease in the number of T_regs was also seen with MK-1248 (an IgG4, which has the same CDRs but possess minimal Fc effector functions) and only when MK-4166 was added at the beginning of the coculture. Furthermore, the NK cell-to-target cell ratio in MLR cultures is insufficient for significant ADCC. However, in an assay optimized to detect ADCC potential (Supplementary Fig. S7A and B), we show that T_reg-bound MK-4166 has the potential to induce ADCC of human T_regs. Taken together these data suggest that MK-4166 may deplete intratumoral T_regs as well as potentially inhibiting their de novo generation in the tumor.

Furthermore, using an in vitro nT_reg suppression assay, we show for the first time that an agonist anti-human GITR antibody (MK-4166) attenuates the suppressive effects of human nT_regs on T_eff proliferation. Unlike in assays where TCR signaling is suboptimal (Fig. 2A–D), a direct proliferative effect of MK-4166 on T_effs was not observed in these assays where a very strong primary TCR signal is present (T_eff:nT_reg ratio 1:0; Fig. 5C and D; Supplementary Fig. S8A-C). However, we cannot rule out the possibility that MK-4166 might act directly on T_effs in these cultures to increase their refractoriness to T_reg suppression as has been suggested for mouse GITR agonism (21, 23, 39, 40). In either case, MK-4166 decreases the suppressive effects of nT_regs. Analysis of proximal signaling events showed that NFκB p65 is phosphorylated immediately upon engagement of GITR on intratumoral T_regs and T_effs (Fig. 6B–D) by MK-4166, suggesting that direct GITR signaling may contribute to the effects of MK-4166 on both populations.

Some studies indicate that DTA-1 causes lineage instability and dedifferentiation of T_regs, and this instability is mechanistically tied to its anti-tumor and pro-inflammatory activity (21, 22), though other studies do not (19, 20, 41). We observed a MK-4166-mediated decrease in the expression of FOXP3, a transcription factor linked to T_regs function, suggesting that MK-4166 may function by attenuating the suppressive activity of T_regs.

MK-4166 has been shown to inhibit growth of established SK-MEL-5 tumors in humanized mice (19). In this model, MK-4166 reduced the number of T_regs in the spleen and to a lesser extent in tumor. In both tissues, a decrease in the activation marker ICOS was observed on T_regs, indicating that MK-4166 can attenuate human T_reg function in vivo.

The data presented here elucidate the effects of GITR agonism on human T cells and provide a strong scientific and translational rationale for the clinical development of MK-4166 to treat advanced malignancies. These are very early insights into the effects of agonist mAbs on human GITR using ex vivo models and suggest that MK-4166 offers an alternative immune-modulatory mechanism to the reversal of checkpoint inhibition in the treatment of cancer. TNFR-agonist antibodies elicit a higher concern for safety, however MK-4166 was well tolerated in non-human primates at all doses administered and no treatment-related toxicity was detected. The safety and tolerability of MK-4166 antibody either as a monotherapy or in combination with pembrolizumab (anti–PD-1 mAb) in patients with advanced solid tumors is currently being evaluated in a phase I study (NCT02132754).

S. Sukumar has ownership interest (including patents) in Merck & Co. G. Ermakov has ownership interest (including patents) in WO 2011/028683 A1 and 14/261,152. No potential conflicts of interest were disclosed by the other authors.

Conception and design: S. Sukumar, D.C. Wilson, Y. Yu, B. Bhagwat, P. Georgiev, V. Sriram, A. Willingham, A.M. Beebe, S. Sadekova

Development of methodology: S. Sukumar, D.C. Wilson, Y. Yu, S. Naravula, G. Ermakov, R. Riener, B. Bhagwat, A.S. Necheva, T. Churakova, R. Mangadu, P. Georgiev, D. Manfra, E.M. Pinheiro, S. Sadekova

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Sukumar, D.C. Wilson, Y. Yu, J. Wong, S. Naravula, G. Ermakov, R. Riener, B. Bhagwat, A.S. Necheva, J. Grein, T. Churakova, R. Mangadu, P. Georgiev, V. Sriram

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Sukumar, D.C. Wilson, Y. Yu, S. Naravula, G. Ermakov, R. Riener, B. Bhagwat, A.S. Necheva, J. Grein, T. Churakova, R. Mangadu, P. Georgiev, W.J. Bailey, A. Willingham, S. Sadekova

Writing, review, and/or revision of the manuscript: S. Sukumar, D.C. Wilson, Y. Yu, S. Naravula, G. Ermakov, R. Riener, A.S. Necheva, P. Georgiev, W.J. Bailey, D. Herzyk, A. Willingham, A.M. Beebe, S. Sadekova

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): W.J. Bailey, T.K. McClanahan, S. Sadekova

Study supervision: D. Manfra, E.M. Pinheiro, W.J. Bailey, T.K. McClanahan, S. Sadekova

The cell lines MB49 and TC-1 were a generous gift from Dr. Michael O'Donnell (University of Iowa) and Dr. Tzyy-Choou Wu (Johns Hopkins University). We thank Gary Starling for a critical review of the article.

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.