Six-week-old to 8-week-old female C57BL/6, BALB/c, B6.Cg-Foxp3tm2Tch/J (B6 Foxp3^GFP) and C.Cg-Foxp3tm2Tch/J (BALB/c Foxp3^GFP) mice were obtained from Jackson Laboratory (Bar Harbor, Maine, USA) and housed under specific pathogen-free conditions.

Mouse pancreatic (Pan02) cells, mouse colon (CT26) cells, mouse mammary carcinoma (4T1) cells, mouse melanoma (B16-F10) cells and the human acute lymphoblastic leukemia cell line (CCRF-CEM) were purchased from the American Type Culture Collection. The MC38 cell line was a kind gift from Dr Holbrook (Department of Medicine, Stanford University). Pan02-OVA cells were engineered by stable transduction of Pan02 cells with pChac-puro plasmid expressing full-length chicken ovalbumin (OVA; LakePharma). All cells were confirmed mycoplasma negative and identity of all cell lines was validated by short-tandem repeat profiling.

The discovery of the selective and potent small molecule CCR4 inhibitor (referenced as CCR4-351 in this manuscript) is described elsewhere.22 This novel compound was designed, synthesized and characterized at RAPT Therapeutics. In several experiments, an independent CCR4 antagonist highly related to CCR4-351 was used. The potency, selectivity and pharmacokinetic properties of this antagonist are similar to CCR4-351.

Antibodies and dyes used for surface staining: unless specified, antibodies were purchased from BioLegend, San Diego. Anti-CD45 BV510 (clone 30-F11), anti-CD4 APC-Cy7 (clone RM4-5), anti-CD8 FITC (clone 53-6.7), anti-CD69 PE (clone H1.2F3), anti-PD-1 PE-Cy7 (clone 29F.1A12, BD Biosciences), anti-CCR4 APC (clone 2G12), anti-CCR4 PerCP/CY5.5 (clone L291H4), anti-CCR5 PE (clone HEK/1/85), anti-CCR6 PE (clone G034E3), anti-CCR7 PE (clone G034H7), anti-CXCR3 PE (clone G025H7), anti-CD45RA BV510 (clone HI100), anti-CD45RO Pacific Blue (clone UCHL1), anti-CD25 APC-CY7 (clone BC96), anti-CD127 APC (clone A019D5), anti-CD4 FITC (clone GK1.5) and 7AAD. Reagents and antibodies used for intracellular staining: cells were fixed and permeabilized following the protocol for the FoxP3/Transcription Factor Staining Buffer (eBiosciences). Fixable viability dye eFluor780 (eBiosciences) and anti-FoxP3 APC (clone FJK-16s, eBiosciences). Sample acquisition was performed on either a FACSCanto II or LSR Fortessa X-20 (Becton Dickinson). Data analysis was performed using FlowJo software (Becton Dickinson).

For both the Pan02 and the Pan02-OVA tumor model, 4×10⁶ tumor cells were injected subcutaneously into the lower right flank of female C57BL/6 mice. For the CT26 tumor model, either 2.5×10⁵ or 1×10⁶ cells were subcutaneously injected into female BALB/c mice in the lower right flank. For the B16F10 and MC38 tumor models, 1×10⁶ cells were injected subcutaneously into female C57BL/6 mice in the lower right flank. And for the 4T1 tumor model, 1×10⁶ cells were injected into the mammary fat pad of female BALB/c mice. Tumor volume was measured twice weekly using the formula: V=0.5(A×B²), where A and B are the long and short diameters of the tumor, respectively. Mice were randomized by tumor size into appropriate treatment groups when a tumor volume of 40–70 mm³ was reached. Daily oral dosing with CCR4-351 (once a day orally) or intraperitoneal administration of an antibody began on the day of randomization. For the Pan02-OVA immunization model, mice were immunized intradermally (ID) with 50 µg/mouse EndoFit OVA mixed in incomplete Freund’s adjuvant (InvivoGen) 14 and 7 days prior to tumor cell inoculation. Animals were dosed with either 50 mg/kg CCR4-351 (once a day orally) or intraperitoneal with antibodies: 50 µg/mouse or 100 µg/mouse anti-CTLA-4 (clone 9D9, Bio X Cell) or 50 µg/mouse anti-CD137 (clone LOB12.3, Bio X Cell) antibodies. Dosing frequency for antibodies is indicated in the Results section. Body weight was measured once a week. Endpoints are either tumor volume of 1500 mm³ for the CT26 and MC38 tumor models or 400 mm³ for the Pan02 and Pan02-OVA tumor models.

At designated time points post randomization into the various treatment groups, spleen and tumor were harvested for analysis by flow cytometry. The tumor tissue was dissociated into a single cell suspension in tumor digestion media containing Collagenase, Dispase and DNAse (Sigma-Aldrich) on the gentleMACS Octo Dissociator (Miltenyi Biotec). Spleens were dissociated by pressing through a 70 µm filter (Miltenyi Biotec) and red blood cells were lysed with ACK lysis buffer (Gibco, Waltham, Massachusetts, USA). Samples were incubated with TruStain FcX (BioLegend) prior to antibody staining.

Mice were randomized when Pan02 tumor volume ranged between 40 and 70 mm³ and then dosed orally with various concentrations of CCR4-351. Three hours after compound dosing, 5×10⁶ miT_reg were injected intravenously via tail vein into tumor bearing mice. Six days later, spleen, tumor and other relevant tissues were harvested and analyzed. For the CT26 T_reg migration studies, animals were randomized when tumor volumes were approximately 150 mm³. Mice were dosed intravenously with 50 µg/mouse of anti-CTLA-4 on day of randomization and again 3 days later. CCR4 antagonist dosing began 7 days after randomization. Three hours after CCR4 antagonist dosing began, 5×10⁶ miT_reg were injected intravenously via tail vein. To enumerate the number of miT_reg within the tumor tissue, a single cell suspension from tumors were prepared and analyzed by flow cytometry.

Human PBMCs were isolated from buffy coats using Ficoll gradient centrifugation. nT_reg were enriched for CD4⁺CD25⁺CD127^dim/– cells using Miltenyi kit (130-094-775, Miltenyi Biotec) and the LD/LS columns (Miltenyi Biotec). Purity was determined by flow cytometry. nT_reg preparations were >90% pure. The nT_reg cells were pelleted and resuspended in human serum at 2×10⁶ cells/mL for use in chemotaxis assays as described below.

The chemotaxis assay is performed using the ChemoTX (Gaithersburg, Maryland, USA) migration system with a 5 µm pore size polycarbonate track-etch membrane (cat# 106-5). CCRF-CEM, iT_reg or nT_reg cells as described above were suspended at 2×10⁶ cells/mL in human or mouse serum (to match species of cell line). Recombinant CCL17 or CCL22 was diluted in Hanks’ Balanced Salt Solution (HBSS) with 0.1% BSA to cover a range of concentrations from 5 pM to 300 nM. Diluted chemokine (29 µL) was placed in the lower wells of the ChemoTX plate. The membrane was placed onto the plate, and 50 µL of the nT_reg cell mixture was transferred onto each well of the membrane. In experiments assessing inhibition with CCR4-351, cells are incubated with various concentrations compound for 30 min prior to transfer. The chemotaxis plates were incubated at 37°C, 100% humidity and 5% CO₂ for 60 min, after which the polycarbonate membranes were removed, and 15 µL of the ATP-binding agent CellTiter-Glo was added to the lower wells. The amount of luminescence, corresponding to the number of migrated cells, was measured using an EnVision plate reader (PerkinElmer, Waltham, Massachusetts, USA). EC₈₀ values (for ligand) and IC₅₀ values (for compound) were determined by non-linear regression using a four-parameter fit using GraphPad PRISM software.

Mouse tumor tissue samples were lysed in 1 mL lysis buffer (1% TritonX‐100, 50 mM HEPES, pH 7.4) and total protein concentration was determined using Pierce BCA Protein Assay Kit (cat# 23225, Thermo Fischer Scientific). Tissue lysates (200 µL) were assayed for cytokine levels using the Milliplex Map Kit according to protocol (MCYTOMAG-70K, Millipore). Relative protein concentration (pg cytokine/mg tumor lysate) for cytokine analysis was calculated by normalizing measured concentration to total tumor protein. Standards and samples were tested in duplicate wells. Data were analyzed using EMD Millipore’s Milliplex Analyst software (V.5.1).

Tumor expression data was obtained from The Cancer Genome Atlas (TCGA) project dataset downloaded from UCSC Xena public data hub on June 18, 2017.21 Only solid tumor data was plotted. Crossplots show the median and 25th through 75th percentile range transcript per million (TPM) expression for each tumor type on a log₁₀ scale. Statistics shown are from a Pearson correlation on the log-transformed TPM values.

T_reg accumulation in tumors has been hypothesized to result from either local proliferation of T_reg present in the TME, conversion from conventional CD4⁺ T cells or migration of T_reg from the periphery. To understand possible mechanisms of T_reg accumulation, we first searched ‘TCGA’ database to identify human tumors that have increased T_reg frequencies, as assessed by the common T_reg marker FOXP3 gene expression.24 Although activated effector CD8 T cells have also been shown to transiently express FOXP3, due to the low frequency of these cells, we expect the majority of the FOXP3 expression to be from T_reg.25–27 In addition, gene expression in cancer tissues was compared with gene expression in normal tissues from the ‘Genotype-Tissue Expression (GTEX)’ database. There is a high correlation between FOXP3 and CD8 expression across many tumor types and normal tissues, though correlation in tumor is higher than in normal (r=0.65 and r=0.47, respectively), suggesting that T_reg levels correlate with levels of effector T cells across tissue types (figure 1A). Importantly, there was a high correlation between FOXP3 expression and CCL17 and CCL22 (r=0.53 and r=0.66, respectively) (figure 1B). This correlation supports our hypothesis that accumulation of T_reg within the TME is predominantly the result of recruitment via CCR4. Next, we assessed the cell surface expression of a panel of CC-chemokine and CXC-chemokine receptors including CCR4 on natural thymic-derived T_reg (nT_reg) by flow cytometry. Human PBMCs were stained for all chemokine receptors for which antibodies were available (figure 1C). As described in previous studies, approximately 90% of CD4⁺CD25 ⁺ CD127^low nT_reg expressed surface CCR4.16 Human nT_reg cell populations also expressed chemokine receptors CCR2, CCR5, CCR6, CCR7 and CXCR3 at frequencies ranging between 20% and 70%. Given the high frequency of CCR4 positivity, the frequency of cells expressing chemokine receptors without co-expression of CCR4 was less than 10% (figure 1C). We also assessed expression of chemokine receptors on mouse nT_reg and observed a similar chemokine expression signature although the frequency of each chemokine receptor varied from human and only 40%–50% of mouse T_reg expressed CCR4 (online supplemental figure 1A).

Since human T_reg express chemokine receptors other than CCR4, recruitment could be mediated through other chemokine–chemokine receptor interactions. In addition, it has been described that expression of TGF-β in tumors enhances T_reg proliferation and can induce conversion of conventional CD4⁺ T cells into T_reg. Therefore, we evaluated whether FOXP3 expression correlates with other chemokine ligands and TGF-β. Our analysis showed only weak to modest correlation between FOXP3 and TGF-β or other chemokines, except for CCL18, a recently identified ligand for CCR8 (online supplemental figures 1B,C, respectively). There was a good correlation (r=0.52) between FOXP3 and CCL18 in tumors (online supplemental figure 1C). CCR8 is predominantly expressed on tumor-resident (intratumoral) T_reg and only on a small percentage of circulating peripheral T_reg.15 28 In fact, we performed flow cytometry staining for surface chemokine receptors on circulating thymic T_reg (nT_reg) from human PBMCs and confirmed the low frequency and low level of surface CCR8 expression on these cells (online supplemental figure 1A). Hence, we believe that it is unlikely that the CCR8/CCL18 axis plays a major role in T_reg migration into the tumor. Recent studies have shown a potential functional role for CCR8 on intratumoral T_reg outside of migration, such as retention of T_reg in the tumor, increased survival and expansion or increased suppressive function.29

In summary, the gene expression analysis from the TCGA and GTEX data sets supports the hypothesis that increased T_reg levels in immunologically ‘hot’ tumors are more likely due to high CCL17 and CCL22 expression and the recruitment via CCR4 engagement.

To further assess if T_reg migration into the tumor is predominantly mediated through CCR4, we developed a potent and selective small molecule antagonist that blocks CCL17 and CCL22 binding to the CCR4 receptor (CCR4 antagonist, CCR4-35122 30). We first generated mouse and human iT_reg and determined the EC₈₀ concentrations for human and mouse CCL22 and CCL17 in an in vitro chemotaxis assay (figure 2A). Then, cells were preincubated with different concentrations of CCR4-351 in the presence of 100% serum and chemotaxis was assessed toward CCL22 and CCL17 (figure 2B and online supplemental figure 2B). In the presence of CCR4-351, we observed a dose-dependent inhibition of iT_reg chemotaxis towards CCL22 (figure 2B). The average chemotaxis IC₅₀ value across multiple experiments for human and mouse iT_reg was approximately 40 nM. CCR4-351 also inhibited CCL17-mediated chemotaxis of CCR4⁺ CCRF-CEM cells with similar potency to that seen for CCL22-mediated chemotaxis (online supplemental figure 2B). CCR4-351 is highly selective as it only inhibits migration of human iT_reg towards the two CCR4 ligands, CCL17 and CCL22, but did not affect migration towards other chemokine ligands (online supplemental figure 2C).

To establish a murine tumor model for assessing whether T_reg migration into tumors can be blocked by CCR4 antagonist, we screened multiple tumor models for their CCL17 and CCL22 expression levels in vivo. Pan02 pancreatic tumors expressed both CCL17 and CCL22 at high levels compared with other tumor types that we tested and were therefore selected as an in vivo tumor model (figure 3A). In order to track and enumerate only newly migrating T_reg into the tumor and not tumor-resident T_reg, we generated iT_reg isolated from syngeneic Foxp3^GFP mice. When fully differentiated into iT_reg, over 90% of these cells were GFP-positive and 40%–50% of them expressed surface CCR4 (figure 3B). The GFP⁺ iT_reg were adoptively transferred intravenously into Pan02 tumor-bearing animals that were either treated with vehicle or CCR4-351. Numbers of GFP⁺ T_reg per 10⁶ of total live cells in the tumor and periphery were determined by flow cytometry. Treatment with CCR4-351 significantly reduced the number of GFP⁺ iT_reg in the tumor in a dose-dependent manner (figure 3C). At the highest dose of 30 mg/kg, which results in plasma levels corresponding to the IC₉₀ of in vitro T_reg migration, CCR4-351 reduced the number of GFP⁺ T_reg in average by over 90% when compared with the vehicle group. The number of GFP⁺ iT_reg in the spleen was only modestly reduced and showed no dose-dependency, with about 26% fewer GFP⁺ iT_reg in all dose groups compared with vehicle control (figure 3C). In comparable studies, we also assessed the number of T_reg in the periphery such as blood and skin to further confirm that CCR4 inhibition selectively inhibits migration of T_reg to the tumor but not healthy tissue. There was no significant difference in T_reg number in the blood and skin in animals dosed with CCR4 inhibitor when compared with the vehicle group (figure 3D). These data suggest that T_reg migration into tumors expressing high levels of CCL17 and CCL22 occurs predominantly via CCR4-mediated chemotaxis.

To determine the degree of antitumor efficacy that can be achieved by blocking CCR4⁺ T_reg migration into tumors, we used the CCL17 ^high CCL22^high Pan02 tumors. However, Pan02 tumors are poorly immunogenic and respond poorly to treatment with immune modulatory agents. To increase the immunogenicity of this tumor model, Pan02 cells were engineered to stably express chicken OVA (Pan02-OVA). In addition, mice were immunized intradermally with OVA on days −14 and −7 prior to Pan02-OVA inoculation to increase the number of OVA-specific T cells. Animals were randomized at an average tumor volume of 60 mm³ and dosed daily with CCR4-351, weekly with an anti-CTLA-4 (figure 4A) or an agonistic CD137 (4-1BB) antibody (online supplemental figure 3), or a combination of CCR4-351 and CTLA-4 or CD137 antibodies. The anti-CTLA-4 antibody clone used here was previously reported not to deplete T_reg.31 When compared with the vehicle group, treatment with CCR4-351 alone resulted in significantly reduced tumor growth that was comparable with single-agent treatment with anti-CTLA-4 (figure 4A) or anti-CD137 (online supplemental figure 3). Tumor growth was further reduced when CCR4-351 and anti-CTLA-4 antibody were combined (figure 4A).

In a parallel study, we isolated tumors on day 24 post-randomization to assess tumor-infiltrating lymphocytes. As expected, the frequency of tumor-infiltrating Foxp3⁺ T_reg was significantly reduced in animals that received CCR4-351 either as a single agent or in combination with anti-CTLA-4 antibody (figure 4B) when compared with the vehicle group. When compared with the vehicle group, mice that were treated with CCR4-351 or anti-CTLA-4 antibody had significantly increased tumor CD8⁺ T cell:T_reg ratios (figure 4C), an important correlate of antitumor immune activity. CD8⁺ T cell:T_reg ratio was highest in the CCR4-351/anti-CTLA-4 combination group.

T_reg accumulation in tumors has been reported in patients receiving immune-based therapies.19 20 To assess whether CCR4 inhibition can potentiate efficacy of an immune-based therapy in tumors where CCL17 and CCL22 levels are low at baseline, mice bearing CT26 colon (CT26) tumor were randomized at an average tumor volume of 60 mm³ and treated with vehicle, a daily dose of CCR4 antagonist, twice a week dose of anti-CD137 (4-1BB) antibody or a combination of CCR4 antagonist and anti-CD137 antibody. As hypothesized, we did not observe significantly reduced tumor growth in CT26-bearing animals with CCR4 antagonist treatment alone since CCR4-mediated T_reg recruitment in this tumor type is most likely minimal and might not contribute to immune resistance (figure 5A,B). Treatment with anti-CD137 antibody alone resulted in intermediate antitumor activity with no complete tumor regression. Importantly, combination therapy of anti-CD137 with CCR4 antagonist significantly enhanced the antitumor effect and four out of nine mice were tumor-free at the end of the study (figure 5). Treated animals remained tumor free for at least 60 days, including the 30–35 days after treatment was stopped. We then dosed CT26 tumor-bearing mice with anti-CTLA-4 antibody and assessed tumor efficacy in combination with CCR4-351. Treatment with anti-CTLA-4 antibody alone resulted in intermediate antitumor activity and 1 out of 10 mice were tumor free at the end of the study (figure 5C,D). As expected, combination therapy of anti-CTLA-4 with CCR4-351 significantly enhanced the antitumor effect and 6 out of 10 mice were tumor-free at the end of the study (figure 5). Treated animals remained tumor free for more than 60 days, including the 30–35 days after treatment was stopped.

To assess the effector T cell:T_reg ratios, CT26 tumors were isolated from animals that were dosed with CCR4-351, anti-CTLA-4 or both. Both, the CD4⁺ T cell:T_reg and CD8⁺ T cell:T_reg ratios were significantly higher in the tumors of animals dosed with the combination of CCR4-351 and anti-CTLA-4 compared with animals dosed with anti-CTLA-4 or CCR4-351 alone or vehicle control (figure 5E). Similarly, activated PD-1⁺ CD8⁺ T cell:T_reg and CD44⁺ CD8⁺ T cell:T_reg ratios were highest in the combination group (online supplemental figure 4A).

Combination treatment with CCR4 antagonist and anti-CTLA-4 in CT26 tumor-bearing mice led to increased CD8⁺ T cell numbers, specific to the CT26 tumor antigen AH-1, when compared with anti-CTLA-4 treatment alone (data not shown). In addition, the generation of a memory T cell population was tested by challenging tumor-free animals with either CT26 cells or 4T1 cells, a mouse breast cancer cell line that expresses relatively low levels of AH-1 antigen and grows in Balb/c animals at a similar rate to CT26. All tumor-free animals inoculated with 4T1 cells demonstrated tumor growth, although tumor growth was slightly slower than in Balb/c mice not previously treated (online supplemental figure 4B and data not shown). In contrast, all tumor-free mice inoculated with CT26 tumor cells remained tumor-free. Taken together, in tumors with low levels of CCL17 and CCL22, CCR4 inhibition enhanced the antitumor effect of anti-CTLA-4 and resulted in long-lasting immunity.

Our observation that an enhanced antitumor efficacy was achieved in CCL17^low CCL22^low CT26 tumors when combined with agonistic antibody or checkpoint blockade raised the question of whether treatment with these antibodies upregulate CCL17 and CCL22 expression and increase CCR4-mediated T_reg migration into tumors. To address this question, mice bearing CT26 tumors received treatment with either vehicle (PBS), agonistic anti-CD137 (4-1BB) or anti-CTLA-4 blocking antibodies and CCR4 ligand levels were assessed in tumor lysates 3 days after antibody treatment. In the vehicle group, CCL22 protein levels in tumors were low while both anti-CD137 and anti-CTLA-4 treatment significantly increased CCL22 levels in tumors (figure 6A). Similar to CT26 tumors, treatment with anti-CTLA-4 antibody also induced upregulation of CCR4 ligand levels in the CCL17^low CCL22^low MC38 tumors (online supplemental figure 5A). These observations demonstrate that immunomodulatory agents that generate a robust antitumor immune response induce increased expression of CCL17 and CCL22 in the TME.

Proinflammatory cytokines such as interferon gamma (IFN-γ), tumor necrosis factor alpha (TNF-α) and interleukin 1 beta (IL-1β) have been reported to induce CCR4 ligands in dendritic cells and macrophages.32 33 We hypothesized that treatment of tumor-bearing mice with anti-CTLA-4 results in the induction of proinflammatory cytokines and consequently increased levels of the CCR4 ligands. Indeed, protein levels of IFN-γ, TNF-α and IL-1β increased in the tumors of anti-CTLA-4-dosed animals compared with the control group (figure 6B). The increase appears to be selective for the CCR4 ligands since other chemokine ligands for receptors that are expressed on T_reg were not elevated (online supplemental figure 5B).

Since CCL17 and CCL22 levels in CT26 tumors increased after treatment with checkpoint inhibitors, we assessed whether we could detect increased recruitment of CCR4⁺ T_reg into tumors following anti-CTLA-4 antibody treatment. CT26 tumor-bearing animals were randomized when tumor volume was between 40 and 70 mm³ and dosed with vehicle, CCR4 antagonist or anti-CTLA-4 antibody alone or in combination to induce CCL17 and CCL22 expression. Seven days after the first antibody treatment, syngeneic GFP⁺ iT_reg generated from BALB/c Foxp3^GFP were transferred into the mice followed by 6 days of treatment with either vehicle or CCR4 antagonist. Very few GFP⁺ T_reg migrated into tumors of control mice (figure 6C). As hypothesized, anti-CTLA-4 treatment prior to T_reg transfer significantly increased the number of tumor-GFP⁺ T_reg consistent with the observed upregulation of CCR4 ligands. Importantly, this effect was almost completely abrogated in animals that were treated with CCR4 antagonist (figure 6C). Anti-CTLA-4 treatment-induced upregulation of CCR4 ligands and concomitant GFP⁺ T_reg increase were limited to the tumor, with no significant change observed in the spleen (figure 6C). The data suggests that CCR4 inhibition has the potential to increase the effectiveness of checkpoint inhibitor and other immunotherapies.