Oncogenic BRAF^V600E governs regulatory T cell recruitment during melanoma tumorigenesis

Abstract

Regulatory T cells (Tregs) are critical mediators of immune suppression in established tumors, although little is known about their role in restraining immune surveillance during tumorigenesis. Here we employ an inducible autochthonous model of melanoma to investigate the earliest Treg and CD8 effector T cell responses during oncogene-driven tumorigenesis. Induction of oncogenic BRAF^V600E and loss of Pten in melanocytes led to localized accumulation of FoxP3⁺ Tregs, but not CD8 T cells, within 1 week of detectable increases in melanocyte differentiation antigen expression. Melanoma tumorigenesis elicited early expansion of shared tumor/self-antigen-specific, thymically derived Tregs in draining lymph nodes, and induced their subsequent recruitment to sites of tumorigenesis in the skin. Lymph node egress of tumor-activated Tregs was required for their C-C chemokine receptor 4 (Ccr4)-dependent homing to nascent tumor sites. Notably, BRAF^V600E signaling controlled expression of Ccr4-cognate chemokines and governed recruitment of Tregs to tumor-induced skin sites. BRAF^V600E expression alone in melanocytes resulted in nevus formation and associated Treg recruitment, indicating that BRAF^V600E signaling is sufficient to recruit Tregs. Treg depletion liberated immunosurveillance, evidenced by CD8 T cell responses against the tumor/self antigen gp100, which was concurrent with the formation of microscopic neoplasia. These studies establish a novel role for BRAF^V600E as a tumor cell-intrinsic mediator of immune evasion and underscore the critical early role of Treg-mediated suppression during autochthonous tumorigenesis.

INTRODUCTION

Immune cells constantly survey host tissues for aberrant cells to limit neoplastic emergence. This cancer immunosurveillance is predicated on the ability of effector CD8 T cells to adequately detect and respond to tumor-associated antigens. While effector responses spontaneously arise against highly immunogenic tumors leading to elimination or equilibrium (1), poorly immunogenic tumors can instead evade immunosurveillance by inducing tolerance (2). FoxP3⁺ regulatory T cells (Treg) mediate this tolerance by suppressing effector responses against cancer (3). Hence, the balance of Tregs and CD8 T cells in the host is critical to effective immunosurveillance (4,5). Accordingly, high intratumoral Treg:CD8 ratios portend weak anti-tumor responses in preclinical models (6) and poor outcomes in solid malignancies (7,8). While Treg:CD8 ratios have been well characterized in the microenvironments of established tumors, less is known about these T cell responses during early tumorigenesis.

Generating and maintaining functional antitumor T cell responses requires choreographed immune events, a process termed the ‘cancer immunity cycle’ (9). Tregs are known to impede this process in draining lymph nodes, as they suppress CD8 T cells following the implantation of poorly immunogenic tumors (2). In transplantable models of ovarian cancer, Tregs preferentially migrate to tumor sites, restrain anti-tumor immunity, and, in turn, allow tumor growth (10). However, the utility of transplantable models remains limited, as they fail to mimic the autochthonous aspects of human malignancies. In an autochthonous model of pancreatic ductal adenocarcinoma, Tregs have been shown to accumulate during the pre-invasive and invasive stages of the disease (11). However, the factors governing their early accumulation during tumorigenesis remain unexplored.

Oncogene-driven transgenic (Tg) mouse models offer a basis for studying host T cell responses in settings that recapitulate the autochthonous nature of human cancers. The Braf/Pten melanoma model harbors inducible, melanocyte-restricted loss of the tumor suppressor gene Pten, and expression of the oncogenic MAPK kinase variant BRAF^V600E, a common melanoma driver mutation (12,13). We previously demonstrated that inhibiting BRAF^V600E in established Braf/Pten tumors induces the selective apoptosis of intratumoral FoxP3⁺ Tregs and promotes CD8 effector T cell responses (14). Moreover, we found that inhibiting BRAF^V600E decreases infiltrating myeloid derived suppressor cells (MDSCs) and downregulates cytokines associated with MDSC recruitment (15). Other groups have similarly described a link between BRAF^V600E and extrinsic mechanisms of immune suppression (16,17). However, it is unclear whether the initial activation of BRAF^V600E signaling is sufficient to trigger Treg recruitment to the nascent tumor microenvironment.

Little is known about the antigen (Ag) specificity or origins of the earliest Treg responders during tumorigenesis. The Treg compartment in established cancers is known to include thymic Tregs (tTregs) and their peripherally-induced counterparts (pTregs) (18). Thymic Tregs arise through high-avidity interactions with self epitopes and constitutively express FoxP3 (19). In an autochthonous mouse model of prostate cancer, such tumor/self-Ag-specific Tregs were identified, and shown to require thymic Aire expression (20). Similarly, Tregs with high avidity for tumor-expressed self Ags have been identified in patients with melanoma and pancreatic cancers (21,22). On the other hand, pTregs develop from CD4⁺FoxP3^– T cells and thus predominantly recognize non-self, tumor-specific antigens. Conversion of pTregs was shown to occur in transplantable models of kidney adenocarcinoma and colon carcinoma (23,24). Moreover, Tregs recognizing tumor-derived viral epitopes have been identified in patients with human papilloma virus-associated cervical cancer (25). In contrast to virally and chemically derived tumors, autochthonous oncogene-driven tumors exhibit a low mutational burden and harbor no predicted neoepitopes (26), suggesting that tTregs may dominate the earliest responses during tumorigenesis. However, this has yet to be explored.

In these studies we employ the inducible Braf/Pten mouse melanoma model to investigate the dynamics of early Treg responses during autochthonous tumorigenesis. By employing melanoma/melanocyte Ag-specific Tregs and CD8 T cells, we investigate the competing processes of tolerance and immunosurveillance during tumor emergence. We report that tumorigenesis induces the priming and recruitment of thymic Tregs at the expense of CD8 T cells, and further elucidate a requirement for oncogenic BRAF^V600E as a driver of Treg trafficking.

MATERIALS AND METHODS

Mice and tumor/nevus induction.

These studies were approved by the Institutional Animal Care and Use Committee (IACUC) at Dartmouth College. Mice were maintained in pathogen-free conditions. Tyr::CreER⁺Braf^CA/+Pten^fl/fl (Braf/Pten) mice were kindly provided by Marcus Bosenberg (Yale), and bred in-house onto a C57BL/6 background, with >98% purity confirmed by congenic testing (DartMouse™). To generate nevus-inducible mice, Braf^CA/CAPten^lox/lox were backcrossed to C57BL/6 to generate Braf^CA/CAPten^WT/WT, which were then crossed with Tyr::CreER⁺ to generate Tyr::CreER⁺Braf^CA/+ (BRAF^V600E) mice. For induction of Braf/Pten tumors and BRAF^V600E nevi, 0.5 mg of 4-hydroxy-tamoxifen (4-HT; Sigma) in DMSO was topically applied over ~1 cm² on the shaved right flanks of 3-week-old mice. C57BL/6, Foxp3^DTR, Rag1^−/−, and Cd8^−/− mice were obtained from the Jackson Laboratory (Bar Harbor, ME), or Charles River Breeding Laboratories (Wilmington, MA). Pmel-1 mice (referred to as pmel), which express a TCR specific for H-2D^b-restricted gp100_25–33, were a gift from Nicholas Restifo (NCI) and were crossed onto a Thy1.1⁺ background (Thy1.1⁺ C57BL/6 mice were from The Jackson Laboratory). Thy1.1⁺Rag1^−/−Foxp3^DTRTyrp^B-w (TRP-1) mice express a TCR specific for I-A^b-restricted TRP-1_113–125 (27), and were a gift from Paul Antony (University of Maryland). OT-II mice were obtained from the Jackson Laboratory and bred in-house, to homozygosity, with CD45.1 congenic mice (NCI B6-Ly5.1/Cr mice; Charles River). All animal studies were performed independently at least twice to generate a conclusion.

Gene expression analysis and protein detection.

Total RNA was isolated using RNeasy kit (Qiagen), and complementary DNA (cDNA) was amplified from 2μg RNA using High Capacity cDNA Reverse Transcription Kit (Applied Biosystems). Expression of melanocyte differentiation antigens was measured by real-time qRT-PCR using pre-validated gene-specific primers (Life Technologies) for Tyrp1 (Mm00453201_m1), gp100 (Mm00498996_m1), Tyr (Mm00495817_m1), Ccl17 (Mm01244826_g1), Ccl2 (Mm00441242_m1), Ccl22 (Mm00436439_m1), gapdh (Mm99999915_g1) and TaqMan master mix (Life Technologies) on a StepOne Plus Real-Time PCR System (Applied Biosystems). Gene expression was expressed as 2^–(ΔCT), where ΔCT = CT_{gene of interest} – CT_Gapdh. Fold changes were calculated as 2^{–(ΔCT reference sample – ΔCT tested sample)}.

Flow cytometry.

Skin samples were harvested, minced, and digested for 45 minutes at 37°C in 2 ml HBSS containing 7 mg/ml collagenase D (Roche) and 200 μg/ml DNase I (Roche), using magnetic bar stirring at 300 RPM. Tissue fragments were mechanically dissociated through a 40 um nylon mesh filter. Samples were washed in RPMI-1640 media containing 10% FBS and 2mM EDTA. Tumor-draining inguinal lymph nodes were mechanically dissociated. For surface molecule staining, samples were incubated with anti-CD16/32 (2.4G2; BioXcell) and stained with antibodies (Biolegend unless indicated) against CD45-APC-Cy7 (30-F11), CD11b-PCP (M1/70), CD4-APC (RM4–5), CD3-BrV420, -BrV510 (17A2), CD44-PE-Cy7 (IM7), CD62L-BrV510 (MEL-14), CD8-PE-Cy7, -BrV510 (53–6.7), Ccr4-BrV421 (2G12), Thy1.1-PE (30-H12), FoxP3-FITC (FJK-16s; eBioscience) on ice for 30 minutes in PBS with 0.5% BSA and 2mM EDTA. For FoxP3 staining, cells were fixed and permeabilized using the Transcription Factor Staining Buffer Set (eBioscience). Cells were acquired on MacsQuant 10 Analyzer (Myltenyi) and analysis was performed using FlowJo 9.8.1 (TreeStar). To determine absolute cell number per gram of tissue, the total number of cells was multiplied by a correction factor for the acquired tumor fraction and normalized for tissue weight.

Adoptive cell transfer.

Pmel cells were isolated from pooled LN and spleen of naïve pmel mice. Negative selection using anti-CD44-PE (IM7; Biolegend) and MACS anti-PE magnetic beads (Myltenyi) was performed, followed by positive selection using MACS anti-CD8 magnetics beads and validation of >90% purity. TRP-1 and OT-II T cells were isolated from pooled LNs and spleens of naïve TRP-1 and OT-II mice, respectively. When indicated, 3 doses of daily 25 μg/kg DT were administered to deplete FoxP3⁺ Tregs prior to CD4⁺ T cell isolation from TRP-1 mice. Positive selection using MACS anti-CD4 magnetic beads was performed, with >80% purity. Isolated lymphocytes were transferred retro-orbitally at a concentration of 1×10⁵ cells/mouse (pmel) or 2×10⁵ cells/mouse (TRP-1 or OT-II) on the indicated days. To assay in vivo migration, single cell suspensions of bulk lymphocytes were obtained from inguinal DLNs of induced mice (26–28 days post induction) or from naïve LNs of WT counterparts, and 3×10⁷ cells were transferred to Rag1^−/− mice bearing induced skin grafts. When indicated, lymphocytes were incubated for 1.5 hours at 37°C in the presence or absence of 20 ng/ml pertussis toxin (PTx).

Skin grafting.

Mice were anesthetized i.p. with 90 mg/kg ketamine and 10 mg/kg xylazine. Tail skin pieces (~5×5 mm) from 3-week-old Braf/Pten or WT donor mice were dorsally grafted with surgical sutures onto syngeneic recipients, which were allowed to recover for 7 days before 4-HT induction. Tumor growth was assessed by measuring skin thickness using a dial caliper.

In vivo depletions.

For Treg depletion using diphtheria toxin (DT; Sigma), Foxp3^DTR mice bearing induced skin grafts were injected intraperitoneally (i.p.) with 25 μg/kg DT in PBS every 2 days. For antibody-mediated depletion, CD4-targeting (300 μl; clone GK.15) or CD8-targeting antibody (500 μl; clone 2.43) were administered i.p. on indicated days. Both antibodies (~1 mg/ml) were produced as bioreactor supernatant from hybridomas purchased from American Type Culture Collection, and each lot was confirmed to deplete >95% of target cells.

Drug treatments.

For FTY720 treatments, 2-amino-2-(2[4-octylphenyl]ethyl)-1,3-propanediol hydrochloride (Cayman Chemical) was dissolved in sterile saline and 1 mg/kg was injected i.p. daily between day 10 and 26. For BRAFi treatment, Braf/Pten mice were given a dose of 100 mg/kg PLX4032 (Selleckchem) by oral gavage 24 hours prior to adoptive transfer on day 19 or qRT-PCR on day 26. PLX4032 was compounded in aqueous vehicle (0.5% hydroxylpropyl cellulose; Sigma) on the day of treatment. When indicated, transferred mice were fed PLX4720-containing diet ad libitum to maintain inhibition from day 19 to 21. PLX4720 (provided by Plexxikon Inc. under a Materials Transfer Agreement) was compounded in AIN-76A rodent diet (417 mg PLX4720/kg) by Research Diets, Inc.

Statistical analyses.

Statistical differences in normally distributed datasets were analyzed using unpaired Student’s 2-tailed t-test when comparing 2 groups, or one-way ANOVA with Bonferroni post-test when comparing 3 distinct cohorts. When Shapiro-Wilk test indicated non-Gaussian distributions, statistical differences were assessed using Mann-Whitney test (2 groups), or Kruskal-Wallis with Dunn post-test (3 groups). Differences in kinetics of skin thickness were determined using two-way ANOVA with Bonferroni post-test. Mice were randomized when assigned to treatment groups. Statistical analyses were performed using Prism 5 software (GraphPad), and differences were considered significant if P ≤ 0.05.

RESULTS

Autochthonous melanoma tumorigenesis induces priming and recruitment of self-Ag-specific thymic Tregs.

To assess the earliest T cell responses during autochthonous tumorigenesis, a tamoxifen-inducible melanoma model driven by BRAF^V600E-expression and Pten loss (referred to as the Braf/Pten model) was employed. We previously showed that dermal injection of tamoxifen leads to palpable tumor formation in this model by day 28 (14). However, topical tamoxifen led to more restrained tumor growth. Three weeks following topical induction, macroscopic hyperpigmented focal lesions developed, coalescing into palpable lesions by day 31 (Fig. 1A). Microscopically, hyperplastic foci could be detected in proximity to hair follicles as early as 16 days post-induction, with larger pigmented neoplasia evident throughout the dermis by day 26 (Fig. 1B). Accordingly, significant increases in expression of melanocyte differentiation Ags tyrosinase-related protein 1 (Tyrp1), glycoprotein 100 (gp100; Pmel), and tyrosinase (Tyr) were detected in the skin as early as 21 days post-induction (Fig. 1C).

Figure 1. FoxP3⁺ Tregs preferentially accumulate in nascent Braf/Pten tumors.

(A) Representative skin from Braf/Pten mice on indicated days following 4-OHT application, with wild-type (WT) skin as a control. (B) Representative H&E staining of WT and Braf/Pten skin showing hyperplasia on d16 and invasive neoplasia (demarcated in yellow) d26 post-induction. (C) qRT-PCR of melanocyte differentiation antigens in tumor-induced (Braf/Pten) vs. WT skin over time (n=3–10 mice/group). (D-F) Induced Braf/Pten and WT skin analyzed for the frequency and absolute numbers of CD45⁺ cells, gated on live cells (D), CD3⁺ T cells gated on CD45⁺ cells (E), CD4⁺FoxP3⁺ Tregs gated on CD3⁺ cells (F), and CD8⁺ T cells gated on CD3⁺ cells (G), with representative plots showing differences 26d post-induction. (H) Ratio of intratumoral FoxP3⁺ Tregs to CD8 T cells over time. (D-H) Data in each panel were pooled from two independent experiments, each with n≥3 mice/group; symbols and error bars represent means ± SEM, with *P<0.05, **P<0.01, ***P<0.001, NS nonsignificant. Absence of error bar indicates SEM less than area represented by symbol. Analyzed by ANOVA (Bonferroni post hoc; C, E, F, H), and Kruskal-Wallis (Dunn’s post hoc; D, G).

Measurable infiltration of T cells into skin was not detected until 26 days post-induction, with both proportions and absolute numbers of CD45⁺ leukocytes increased in comparison to wild-type skin (Fig. 1D). Total CD3⁺ lymphocytes, while slightly decreased by total proportion of CD45⁺ cells, were still substantially increased by absolute number in tumor-induced skin (Fig. 1E). Within this CD3⁺ population, FoxP3⁺ Treg cells were significantly increased by proportion, resulting in a log-fold increase in absolute numbers of Tregs in the skin between days 21 and 26 (Fig. 1F). In contrast, CD8 T cell proportions were slightly decreased in tumor-induced skin (Fig. 1G). However, this did not translate to a drop in overall numbers of CD8 T cells (Fig. 1G), consistent with the quantitative increase in FoxP3⁺ Tregs. These dynamics resulted in a six-fold increase in the Treg:CD8 ratio between days 21 and 26 (Fig. 1H). Thus FoxP3⁺ Tregs, but not CD8 T cells, accumulated in nascent melanomas within days following detectable increases in tumor Ag expression.

Consistent with the paradigm that T cells are primed in lymph nodes prior to their recruitment to peripheral sites, elevated proportions of Tregs were also detected in the draining lymph nodes (DLN) of tumor-induced mice compared to naïve counterparts (Fig. 2A). Further, Treg proportions and numbers were elevated in tumor-DLNs compared to non-draining LNs from tumor-induced mice (Supp. Fig. 1A). A significant increase in the Treg:CD8 ratio also became evident in DLN by day 26 (Fig. 2B) as a result of an eight-fold increase in Treg numbers compared to a more modest (4-fold) increase in CD8 T cells (Supp. Fig. 1B). Moreover, Treg populations in DLN were enriched for phenotypically activated/effector (CD44^hiCD62L^lo) and central memory (CD44^hiCD62L^hi) subsets (Fig. 2C), and were significantly more suppressive ex vivo relative to Tregs derived from naïve LNs (Fig. 2D). To further assess whether tumor Ag-specific Tregs were primed during tumorigenesis, CD4⁺ T cells specific for the melanocyte-expressed differentiation antigen Tyrp1 (TRP-1 cells) were harvested from naïve TCR Tg mice (on a Tyrp-1^KO background (27)) and adoptively transferred into WT and Braf/Pten mice 16 days post-induction (Fig. 2E). Ten days later, significantly larger populations of CD4⁺FoxP3⁺ TRP-1 cells were detected in DLN of tumor-induced Braf/Pten mice compared to WT counterparts (Fig. 2F), and compared to non-DLN within the same host (Supp. Fig. 2A), indicating tumor-driven Treg expansion. Whereas the transferred population exhibited a predominantly naïve phenotype, TRP-1 Tregs in induced skin-DLN acquired an overwhelmingly CD44^hiCD62L^low/hi effector/memory phenotype (Fig. 2G), as evidence of Ag-experience. A similar experiment involving tumor Ag-irrelevant OT-II Tg T cells indicated no difference between OT-II Treg accumulation in tumor-DLNs versus naïve LNs, further confirming the tumor/self-Ag specificity of the response (Supp. Fig. 2B). These data illustrate that melanoma tumorigenesis induces early priming and expansion of tumor/self-Ag-specific Tregs in draining lymph nodes.

Figure 2. Tumor/self-antigen-specific thymic Tregs proliferate and function in draining lymph nodes during tumorigenesis.

LNs draining 4-HT-induced skin were analyzed in Braf/Pten and WT mice. (A-C) Frequency of CD4⁺FoxP3⁺ Tregs gated on CD3⁺ cells (A), ratio of FoxP3⁺ Treg to CD8⁺ T cells over time, (B) and frequency of T cell subsets 26d post-induction, gated on CD4⁺FoxP3⁺ Tregs (C). (D) CD4⁺CD25⁺ Tregs from pooled Braf/Pten DLNs vs. WT naïve LNs (n=4–5 mice per group) were assayed for suppression of CellTrace-Violet-labeled CD8 T cell proliferation ex vivo (see Supplemental Materials and Methods). Representative flow plots depict 1:1 Treg:CD8 ratio. Dashed lines represents %divided cells in the absence of Tregs (top) or absence of stimulation (bottom); symbols represent mean of 3 wells ± SD. (E) Schematic for F-I depicting the adoptive transfer of Thy1.1⁺CD4⁺ TRP-1 T cells into WT vs. Braf/Pten mice on d16, and analysis 10d later. (F) Absolute numbers of TRP-1 Tregs on 26d post-induction. (G) Frequency of TRP-1 Tregs with indicated phenotypes on the day of transfer (input) and 26d post-induction. (H) Effect of DT administration on the input population of CD4⁺ TRP-1 cells. (I) Number of TRP-1 Tregs on d26 (gated on CD3⁺ cells) in induced Braf/Pten mice that received a transfer of intact or Treg-depleted TRP-1 cells. Data were pooled from (A-C, F, G, I) or are representative (D) of two independent experiments, each with n≥3 mice/group; *P<0.05, **P<0.01, ***P<0.001, analyzed by ANOVA (Tukey post hoc, A-D), and t-test (F, I). Symbols and error bars represent means ± SEM (A-D, G); symbols represent individual mice and horizontal lines depict means (F, I). Absence of error bar indicates SEM less than area represented by symbol.

To determine the extent to which thymically derived Tregs contributed to this early response during tumorigenesis, we pre-depleted FoxP3⁺ thymic Tregs in the input TRP-1 T cell population by diphtheria toxin (DT) treatment prior to adoptive transfer on day 16 (Fig. 2H; TRP-1 Tg mice also expressed the diphtheria toxin receptor under the control of the Foxp3 promoter (27)). Ten days post-transfer, mice that received FoxP3-depleted TRP-1 T cells generated an order of magnitude fewer Ag-specific Tregs in DLN, as compared to mice transferred with an intact FoxP3⁺ compartment (Fig. 2I). Thus, the priming of thymic Tregs, as opposed to the conversion of CD4⁺FoxP3^– T cells, dominates the early Ag-specific Treg response to tumorigenesis in DLN.

To determine whether thymic Tregs also constituted the earliest population recruited to nascent tumor sites, the same experiment was performed, and skin was analyzed on day 26. Following DT depletion of thymic Tregs, accumulation of FoxP3⁺ cells was drastically reduced in induced skin (Fig. 3A), indicating that nascent tumor-infiltrating Tregs also derived from the thymic subset. To determine whether these Tregs originated from populations primed in DLN, the sphingosine-1-phosphate analog FTY720 was used to block T cell egress from lymph nodes. Indeed, upon treatment with FTY720, the number of transferred TRP-1-specific Tregs was significantly decreased in skin (Fig. 3B). A similar decrease was observed in the polyclonal Treg population (Fig. 3C). Thus, optimal Treg accumulation in nascent tumors required thymic Treg egress from lymph nodes.

Figure 3. Thymic Tregs home from draining lymph nodes to nascent tumor sites in skin.

(A) Absolute number of Thy1.1⁺Foxp3⁺ TRP-1 Tregs in induced skin of Braf/Pten mice (d26) that received adoptive transfer of intact vs. DT-depleted TRP-1 T cells on d16. (B-C) Absolute numbers of FoxP3⁺ Tregs in Braf/Pten vs. WT skin 26d post-induction, +/− FTY720 treatment; (B) Numbers of TRP-1-specific Tregs; intact TRP-1 T cells were transferred 10d prior; (C) Endogenous Foxp3⁺ Tregs. (D) Braf/Pten tail skin was grafted onto Rag^−/− vs. WT hosts, and numbers of FoxP3⁺ Tregs were quantified in skin grafts 21d post-induction. (E) Schematic depicting Treg recruitment assay used in F-H; Rag^−/− mice bearing Braf/Pten skin grafts received adoptive transfer (A.T.) of T cells on d19, and Tregs were assessed in grafts 48h later (d21). (F) Absolute numbers of Tregs in skin grafts following A.T. of T cells derived from tumor-DLNs vs. naïve LNs. (G) Proportions (left) and absolute numbers (right) of Tregs in induced skin grafts vs. adjacent skin and DLN. (H) Treg:CD8 ratios in induced skin grafts vs. adjacent skin. Data in each panel were pooled from two independent experiments, each with n≥3 mice/group (A-G) or n≥2 (F), with *P<0.05, **P<0.01, ***P<0.001, analyzed by Mann-Whitney test (A), Kruskal-Wallis (Dunn’s post hoc; B), ANOVA (Tukey post hoc; C, H), and t-test (D, F, G). Symbols represent individual mice, horizontal lines depict means (A-G), or symbols and error bars represent means ± SEM (H).

As FTY720 treatment did not completely eliminate Treg accumulation in induced skin, we separately assessed whether skin-resident Tregs proliferated in situ during tumorigenesis. Skin from Braf/Pten mice was grafted onto T cell-deficient (Rag^−/−) mice (Supp. Fig. 3A). As grafting necessitated the use of melanocyte-rich tail skin, tumor appearance and Treg infiltration was accelerated in this model, occurring on day 21 (Supp. Fig. 3B-D). When tumors were induced in Braf/Pten skin grafts on Rag^−/− hosts, Tregs were virtually absent, indicating a lack of contribution from skin-resident Tregs (Fig. 3D). Separately, when Braf/Pten skin was grafted onto congenically distinct (Thy1.1⁺) mice, Tregs on day 21 overwhelmingly expressed the congenic marker of the host (Supp. Fig. 3E), further supporting their origination from the lymphoid compartment.

To formally demonstrate that migration mediates the accumulation of tumor-experienced Tregs in the skin, T cells from either naïve LN or tumor-DLN were transferred into Rag^−/− recipients 19 days following induction of Braf/Pten skin grafts, and infiltration was assessed 48h later (Fig. 3E). Whereas the input cell population from tumor-bearing donors had a slightly elevated Treg proportion as compared with naïve donors (Supp. Fig. 4A), this difference was amplified in vivo with tumor-experienced Tregs migrating in four-fold greater numbers to tumor-induced skin (Fig. 3F). More detailed analysis of DLN-derived Treg migration indicated preferential trafficking to induced skin grafts, as compared to adjacent host skin or LNs (Fig. 3G). This was in contrast to the CD8 T cells from the transferred, tumor-experienced DLN population, which predominantly trafficked to LNs, resulting in a high Treg:CD8 ratio in skin grafts (Fig. 3H). Taken together, these data illustrate that melanoma tumorigenesis induces the priming and recruitment of Tregs but not CD8 T cells, and suggest a dominant role for Ag-specific thymic Tregs at this early stage.

BRAF^V600E signaling governs chemokine-driven recruitment of Tregs during tumorigenesis.

The mechanisms of early thymic Treg recruitment to sites of tumorigenesis remained to be elucidated. To determine whether chemotaxis was involved, tumor-experienced T cells were treated ex vivo with pertussis toxin (PTx) to block chemokine receptor signaling, prior to adoptive transfer as in Fig. 3E. PTx had negligible effects on the viability of T cells in the input population (Supp. Fig. 4B). However, consistent with a role for chemokine receptors, PTx-pretreatment significantly decreased the number of Tregs migrating to induced skin (Fig. 4A). As reports have implicated Ccr4 in the accumulation of Tregs in established tumors (10,28), Ccr4 expression was assessed in Tregs from induced-skin DLN. Compared to Tregs from naïve LNs, approximately twice as many Tregs from DLN expressed Ccr4 (Fig. 4B). Similarly, a majority of TRP-1-specific Tregs expressed Ccr4 following tumor Ag encounter in DLN (Fig. 4C). Moreover, the expression of three Ccr4-cognate chemokines, Ccl17, Ccl2, and Ccl22, increased in Braf/Pten skin within the 21–26 day window following tumor induction (Fig. 4D, and Supp. Fig. 5A). To determine if these chemokines were required for Treg accumulation during tumorigenesis, mice were treated with a combination of Ccl17, Ccl2, and Ccl22 neutralizing mAbs throughout the day 21–26 Treg recruitment window. Chemokine neutralization significantly reduced Treg accumulation by proportion of CD3⁺ T cells, however absolute numbers of Tregs were only modestly decreased (Supp. Fig. 5B). To more definitively test whether Ccr4 was required for Treg accumulation during tumorigenesis, the Ccr4 antagonist C-021 was administered to tumor-induced mice throughout days 21–26. FoxP3⁺ Treg numbers were significantly decreased in induced skin of antagonist-treated mice (Fig. 4E), confirming a role for Ccr4 in Treg recruitment.

Figure 4. The Ccr4 chemotactic axis controls Treg homing during melanoma tumorigenesis.

(A) According to schematic in Fig. 3E, mice were grafted with Braf/Pten skin and T cells were transferred following ex vivo treatment with vehicle (Veh) or pertussis toxin (PTx). CD4⁺FoxP3⁺ Tregs were enumerated in grafts 48h post-transfer. (B) Naïve LNs vs. LNs draining d26 induced skin from Braf/Pten mice (DLN) were analyzed for Ccr4 on endogenous CD4⁺Foxp3⁺ Tregs. (C) Thy1.1⁺ TRP-1 T cells were transferred into Braf/Pten mice as in Fig. 3E, and Ccr4 was assessed on Thy1.1⁺Foxp3⁺ TRP-1 Tregs on the day of transfer (input) vs. 10d later in draining lymph nodes (DLN). (D) Quantitative RT-PCR analysis of Ccr4 cognate ligands over time in Braf/Pten vs. WT skin. (E) Tregs were enumerated in the skin of WT vs. Braf/Pten mice 26d post-induction, with or without daily administration of Ccr4 antagonist (C-021) beginning on day 21. (A-E) Data in each panel were pooled from two independent experiments, each with n≥3 mice/group (“input” group in C represents 2 in vitro measurements), with **P<0.01, ***P<0.001, NS non-significant, analyzed by t-test (A), or ANOVA (Bonferroni post hoc; B, D, E). Symbols represent individual mice, horizontal lines depict means (A-C, E); symbols and error bars represent means ± SEM (D).

We have previously reported that established Braf/Pten tumors express Ccl17 and Ccl2, and that these transcripts are downregulated following treatment with small molecule BRAF^V600E inhibitors (BRAFi) (15). To assess whether Ccr4-cognate chemokines are also regulated during tumorigenesis, their expression was analyzed 26 days post-induction, with or without BRAFi administered 24h earlier. Indeed, BRAFi treatment rapidly decreased expression of all three chemokines in induced skin (Fig. 5A; Supp. Fig. 5C). To separately determine if BRAF^V600E signaling regulates Treg migration to induced skin, T cells from DLN were adoptively transferred into Rag^−/− mice bearing induced Braf/Pten skin grafts (see Fig. 3E), with or without concurrent BRAFi treatment. Forty-eight hours post-transfer, both the proportions and absolute numbers of skin migrating Tregs were reduced in BRAFi-treated animals (Fig. 5B). In contrast, BRAFi had no effect on Treg accumulation in lymph nodes (Fig. 5C). Taken together, these data show that BRAF^V600E signaling promotes the expression of Ccr4 cognate chemokines and the associated recruitment of Tregs to nascent tumor sites.

Figure 5. Oncogenic BRAF^V600E governs Treg recruitment during tumor emergence.

(A) qRT-PCR of Ccr4 cognate ligands in WT vs. induced Braf/Pten skin, in mice gavaged with PLX4032 (BRAFi; +) or vehicle (–) 24h prior to analysis. (B-C) According to schematic in Fig. 3E, mice were grafted with Braf/Pten tail skin and treated with vehicle or BRAFi 24h prior to adoptive transfer of total T cells derived from day 26 tumor-DLNs (see Methods). (B) Proportions and absolute number of FoxP3⁺ Tregs in induced Braf/Pten grafts. (C) Proportions of Tregs in LNs draining induced grafts (DLN). (D-F) Wild type (WT) vs. BRAF^V600E skin (nevus) was analyzed 31d post-induction. Representative plots showing (D) CD4⁺FoxP3⁺ Tregs, gated on CD3⁺ cells; and (E) CD8⁺ T cells, gated on CD3⁺ cells. (F) Ratio of FoxP3⁺ Tregs to CD8 T cells. (G) WT and nevus-induced mice received 2×10⁵ Thy1.1⁺CD4⁺ TRP-1 T cells on d21, and numbers of TRP-1 Tregs were analyzed in skin 10d later. (A-C, D-G) Data in each panel were pooled from two independent experiments, each with n≥4 mice/group (symbols represent individual mice, horizontal lines depict means), with **P<0.01, ***P<0.001, and NS nonsignificant, analyzed by ANOVA (Bonferroni post hoc; A), t-test (B, C, D-F), or Mann-Whitney test (G).

To formally determine if BRAF^V600E signaling is sufficient to recruit Tregs, we generated mice expressing melanocyte-restricted tamoxifen-inducible BRAF^V600E, but with normal Pten expression. Consistent with prior reports in similar models (29,30), topical induction of BRAF^V600E elicited hyperpigmented, benign melanocytic lesions in skin (Supp. Fig. 6A). These nevus-like lesions expressed elevated levels of melanocyte Ags (Tyrp1, gp100, and Tyr) and Ccr4-cognate chemokines (Ccl17, Ccl2, and Ccl22) by both RNA and protein (Supp. Fig. 6B-D). In nevus-induced skin on day 31, FoxP3⁺ Tregs were significantly increased by both proportion of CD3⁺ T cells, and by absolute number (Fig. 5D), indicating that BRAF^V600E expression is sufficient to promote the accumulation of Tregs. In contrast, CD8 T cell proportions decreased, while absolute numbers remained unchanged in nevi (Fig. 5E). This is consistent with an influx of CD3⁺ Tregs, and resulted in a two-fold increase in the Treg:CD8 ratio (Fig. 5F). This change was less pronounced in nevi than in Braf/Pten skin (see Fig. 1H), albeit significant. Moreover, adoptively transferred TRP-1-specific Tregs accumulated in nevus-bearing skin (Fig. 5G), demonstrating an Ag-specific Treg response. These data collectively support the conclusion that BRAF^V600E governs the recruitment of Tregs during autochthonous tumorigenesis.

Optimal tumorigenesis requires Treg suppression of CD8 T cell-mediated immune surveillance.

The tumor-promoting role of Tregs has previously been demonstrated in established cancers (3). However, it has remained unknown whether Tregs prevent early T cell-mediated immunosurveillance against autochthonous, poorly immunogenic cancers. To assess this, Braf/Pten skin was grafted onto Foxp3^DTR mice, and Tregs were depleted by DT treatment beginning 17 days post-induction. The depletion of Tregs resulted in a forty-fold increase in CD8 T cells in the skin 26 days post-induction (Fig. 6A), indicating that CD8 T cells could readily infiltrate early tumor lesions if Tregs were absent. To separately address this question in a more high-throughput (i.e. non skin-grafted) setting, we selected anti-CD4 mAb as an established therapy to deplete Tregs without impairing CD25⁺ activated effector T cells (4). Consistent with data from Foxp3^DTR mice, anti-CD4 treatment also resulted in robust CD8 T cell accumulation in tumor-induced skin (Fig. 6B). These data collectively indicate that Tregs suppress CD8 T cell accumulation, and further suggest that CD4 T cell help is not absolutely required for CD8 T cell infiltration during tumorigenesis.

Figure 6. Optimal Braf/Pten tumorigenesis requires FoxP3⁺ Tregs.

(A) Foxp3^DTR mice bearing d17 induced Braf/Pten skin grafts received diphtheria toxin (DT) every 2 days, or were untreated, and absolute numbers of CD8 T cells were assessed in grafts 26d post-induction. (B) Absolute numbers of CD8 T cells in Braf/Pten skin 26d post-induction, +/− anti-CD4 depleting antibody (clone GK1.5) on d16 and 22. (C) Braf/Pten vs. WT mice (treated as in B) received A.T. of 1×10⁵ Thy1.1⁺CD8⁺CD44^- naïve pmel cells on d12; proportions of pmel cells in lymph nodes draining induced skin (DLN; left) and absolute numbers in induced skin (right) were assessed 26d post-induction. (D-E) WT vs. Braf/Pten mice received naïve pmel cells 14d prior to analysis on indicated days (post-induction); transfer was followed by CD4 depletion on d4 and 10; pmel cells were then enumerated in DLN (D) or induced skin (E). (F) Proportions of pmel cells in DLN (left) and absolute numbers in induced skin (right) 31d (WT and nevus BRAF^V600E mice) post induction (Braf/Pten mice were analyzed on day 26 as a positive control); mice were treated with anti-CD4 as in D. (G) Thickness of induced Braf/Pten skin grafts on WT vs. Foxp3^DTR mice that received DT (white arrows) alone or concurrent with anti-CD8 mAb (grew arrows). Data in each panel were pooled from two independent experiments each with n≥3 mice/group (B-G) or n≥2 (F), with *P<0.05, **P<0.01, ***P<0.001, NS nonsignificant, analyzed by t-test (A, B, D), Mann-Whitney test (E), ANOVA (Bonferroni post hoc; C left, F left, G), and Kruskal-Wallis (Dunn post hoc; C right, F right). Symbols represent individual mice (A-F), or symbols and error bars represent means ± SEM (G).

To determine if early, Treg-restrained CD8 T cell responses are tumor/self-Ag specific, mice received sentinel populations of congenically marked naïve CD8 T cells specific for gp100_25–33 (pmel cells). Despite local tumorigenesis, pmel cells remained virtually undetectable in DLN or early tumors on day 26 (Fig. 6C). However, Treg depletion with anti-CD4 promoted a significant accumulation of pmel cells in both locations in tumor-induced mice, but not in WT mice (Fig. 6C). To determine if Tregs functionally prevent CD8 T cell responses during incipient stages of tumorigenesis, earlier time points were analyzed. Indeed, in DLNs of Treg-depleted mice, pmel cells accumulated as early as 16 days post-induction (Fig. 6D), which was concurrent with the earliest microscopically detected melanocytic lesions (see Fig. 1B). However, pmel cells were not detected in induced skin at these earlier time points (Fig. 6E), suggesting that recruitment of CD8 T cells, like Tregs, is limited by local tumor-derived factors. Interestingly, Treg depletion did not elicit pmel cell priming or migration in response to benign melanocytic nevi (Fig. 6F), indicating that malignant transformation is required for CD8 T cell responsiveness. Finally, to determine if Treg-restrained immunosurveillance can suppress tumorigenesis, tumor formation was assessed in Foxp3^DTR mice bearing Braf/Pten skin grafts. Indeed, the depletion of Tregs with DT resulted in marked control of induced skin thickening, an effect that was lost upon co-depletion of CD8 T cells (Fig. 6G). Therefore, in this early autochthonous malignancy, Treg suppression potently restrains CD8 T cell-mediated immune surveillance.

DISCUSSION

This work has uncovered the origins of Treg responses during autochthonous tumorigenesis, and unveiled oncogenic BRAF^V600E as a key driver of early Treg recruitment. We show that Treg recruitment requires the pre-activation of Tregs in draining lymph nodes, and that self-Ag-specific thymic Tregs are involved in the process. This Treg response leads to the suppression of CD8 T cell-mediated immunosurveillance as early as microscopic neoplastic lesions develop. Thus, Treg-mediated immune suppression potently counteracts cancer immunosurveillance even in settings of early autochthonous tumorigenesis.

Previous studies by ourselves (14) and others (16,17,31) have defined T cell infiltrates in established BRAF^V600E-driven melanoma tumors, yet little was known about T cell responses during tumorigenesis. The current studies now identify Tregs as the dominant T cell responders to early tumors. This observation is consistent with prior studies showing early preferential accumulation of Tregs in pancreatic ductal adenocarcinoma in patients (32) and in a Kras-driven pancreatic mouse model (11). The present finding that Ag-specific Tregs, but not CD8 T cells, are activated in early tumor-draining lymph nodes is consistent with previous findings that Tregs outpace CD8 T cells during early tumor growth (2). In accordance with the poor immunogenicity of autochthonous tumors and their lack of strong neoantigens, we observe a lack of endogenous CD8 T cell responses during tumorigenesis, which is due to the presence of Tregs. In conjunction with published studies (2), our work supports the conclusion that Tregs are the earliest T cell responders to neoplastic transformation.

Our work also underscores the participation of thymically derived, self-Ag-specific Tregs during melanoma emergence. Our demonstration that self-Ag-specific tTregs characterize the early response during tumorigenesis is consistent with reports showing that the TCR repertoires of tumor-infiltrating CD4⁺FoxP3⁺ and CD4⁺FoxP3^– cells are largely distinct, and instead resembled their draining lymph node counterparts (33). Our findings are also consistent with a previous report that Tregs in autochthonous prostate tumors are Aire dependent and recognize a tumor-expressed self Ag (20). As autochthonous oncogene-driven tumors were shown to lack predicted neoepitopes (26), our findings lend credence to the notion that self-Ag-specific thymic Tregs dominate the T cell responses elicited during early tumorigenesis.

The present in vivo migration studies illustrate that BRAF^V600E is a critical tumor cell-intrinsic driver of Treg recruitment. This observation is consistent with reports by us and others showing that treatment with BRAF^V600E inhibitors for >7 days decreases Treg prevalence in established Braf/Pten tumors (14–16). The present work underscores the rapidity with which Treg migration is inhibited in response to BRAFi. Importantly, this work also provides a new example in a growing body of research linking oncogenic signaling and immune regulation (34,35). Previous reports have linked c-KIT signaling to IDO expression in gastrointestinal stromal cell tumors (36), Pten loss to innate anti-PD1 resistance in melanomas (37), Myc to CD47 and PD-L1 expression in lymphomas (38), and β-catenin to reduced effector T cell infiltration in melanomas (31). As BRAF^V600E was previously shown to govern the accumulation of MDSCs in established tumors (14,15), future studies should address whether BRAF^V600E induction also promotes MDSC accumulation during tumorigenesis.

Our studies underscore the role of oncogene-driven chemotactic cues in Treg recruitment. The finding that FTY720 decreases Treg recruitment during tumorigenesis corroborates a previous report that S1PR1 deficiency reduced Treg accumulation in implanted tumors (39). We demonstrate that Treg recruitment to early melanoma sites requires Ccr4 signaling, consistent with the role of this axis in Treg trafficking to transplantable ovarian and melanoma tumors (10,28). Our studies also indicate that tumor induction significantly increases expression of multiple Ccr4 cognate chemokines in the skin. This is consistent with previous findings that BRAFi results in the downregulation of Ccl2 and Ccl17 in established Braf/Pten tumors (15) and Ccl2 in transplantable tumors derived from Braf/Pten mice (40). While it remains unclear which chemokine(s) mediate Treg migration, this relationship may be complex due to the functional redundancy of Ccl2, Ccl22, and Ccl17 (41). Additionally, the inability of the Ccr4 antagonist to completely abrogate Treg migration suggests that additional chemokine receptors may be involved.

BRAF^V600E mutations exist in 78% of human nevi (42), and persist in the course of melanoma’s genetic evolution from precursor lesions (43). Yet, T cell responses have remained poorly understood in relation to pre-neoplastic lesions. Our studies reveal that Tregs also preferentially accumulate in response to BRAF^V600E-driven nevus formation. This recruitment was less pronounced than in induced Braf/Pten skin, which may relate to the modest induction of chemokines and melanocyte Ags, and underscores the role of additional oncogenic pathways (e.g. Pten) in the optimal recruitment of Tregs. Prior studies indicate that BRAF^V600E nevus-like lesions are largely benign (29,30) and exhibit senescence-like cell cycle arrest typical of human nevi (44). However, these lesions progress to melanoma with a median latency of 12 months (29). Interestingly, FoxP3⁺ Tregs are prevalent in human atypical nevi and radial growth phase melanomas (45), suggesting a link between Tregs and the transition of melanocytic lesions to malignancy. One may speculate that nevus-associated Tregs could facilitate progression to melanoma through local immune suppression.

While our studies do not detect Treg accumulation in tumor-induced skin prior to day 26, Tregs play a critical role in suppressing immunosurveillance prior to this day. Indeed, when Tregs were depleted, tumor/self-Ag-specific CD8 T cells accumulated in draining lymph nodes very early. Our analyses show that T cell-mediated immunosurveillance is compromised even before increased tumor/self Ag expression is detectable in induced skin. Our failure to detect CD8 T cell responses in Treg-intact mice is in contrast to a recent report that SV40 neoantigen-specific CD8 T cells proliferate in response to pre-malignant lesions of SV40-driven autochthonous liver cancer, and only subsequently do these antitumor T cells become dysfunctional (46). In our model, CD8 T cell ignorance is Treg mediated from the outset, and only following Treg depletion are CD8 T cells capable of controlling tumor outgrowth. Interestingly, pmel cells remained ignorant in nevus-bearing animals even after Treg depletion, indicating that neoplastic transformation is required to break tolerance. The expansion of self-Ag-specific CD8 T cells in induced Braf/Pten mice corroborates our previous reports that pmel cells are primed following Treg depletion in mice bearing established B16 and Braf/Pten tumors (14,47), while further demonstrating the highly sensitive nature of CD8 T cells during early microscopic tumor formation.

The present work is not without limitations. Analysis of the causal connection between Tregs and CD8 T cells required the use of skin grafts on Foxp3^DTR mice, so it remains uncertain if Tregs are the predominant tumor escape mechanism during autochthonous tumorigenesis. These studies are also limited to a single model, and it is not yet clear if implications extend to all melanomas or other types of cancer. Finally, the mechanistic underpinnings of oncogene-driven chemokine production remain unclear. Oncogenic BRAF^V600E increases the activity of transcriptional factors including NFκB and AP-1 (48,49), which can promote chemokine expression (50). However, pathways linking BRAF^V600E signaling and Ccr4-cognate chemokine expression will require further investigation.

In summary, this work illustrates an absence of effective cancer immunosurveillance in the presence of dominant early Treg suppression. By establishing a link between BRAF signaling and Treg responses, these studies highlight the inherently immunosuppressive nature of oncogene-driven cancers. Tregs remain a formidable obstacle to generating effective immunity against cancer. Their key requirement during early tumorigenesis further underscores the integral role of Tregs throughout cancer progression.

STATEMENT OF SIGNIFICANCE.

This work provides new insights into the mechanisms by which oncogenic pathways impact immune regulation in the nascent tumor microenvironment.