Reprogramming Tumor-Infiltrating Dendritic Cells for CD103⁺CD8⁺ Mucosal T-cell Differentiation and Breast Cancer Rejection

In recent years, we have witnessed an improved understanding of the critical roles that the tumor microenvironment plays in cancer growth, evasion from host immunity, and resistance to therapeutic agents (1). A better definition of the molecular and cellular components of the tumor microenvironment will enhance the clinical efficacy of current immunotherapy approaches and enable tailoring of specific therapeutic strategies. Breast and pancreatic cancers are characterized by infiltration of inflammatory Th2 (iTh2) cells, which coexpress interleukin (IL)-4/IL-13 and TNF-α but not IL-10 (2, 3). Clinically, the Th2 signature in breast cancer (4, 5) and the expression of the Th2 master regulator GATA-3 in pancreatic cancer (6) are associated with poor outcomes.

Experimentally, iTh2 cells accelerate tumor development in humanized mouse models of breast cancer through the activity of IL-13 (2). In genetically engineered mouse models of mammary cancer, iTh2 cells accelerate the development of pulmonary metastasis via IL-4 (7). IL-4 and IL-13 exert protumor activity through several pathways, including (i) the triggering of TGF-β secretion (8), (ii) the upregulation of antiapoptotic pathways in cancer cells (9), and (iii) the generation of type II–polarized macrophages that foster tumor growth directly via the secretion of growth factors, and indirectly via the inhibitory effects on CD8⁺ T-cell function (10). Indeed, CD8⁺ T cells are essential for tumor rejection through the generation of cytotoxic effectors. The presence of CD8⁺ T cells in primary tumors is associated with the long-term survival of patients with colorectal and breast cancer (10, 11). Thus, iTh2 cells have a broad and profound impact on the tumor microenvironment and cancer progression.

The generation of iTh2 cells in breast cancer depends on the presence of mature tumor-infiltrating OX40L⁺ dendritic cells (DC; ref. 3). In experimental models of breast cancer, this DC phenotype is driven by cancer-derived thymic stromal lymphopoietin (TSLP; refs. 3, 12). Previous studies have demonstrated that dectin-1, an innate immune receptor with activating motifs [immunoreceptor tyrosine-based activation motif (ITAM)], can reprogram DCs from inducing Th2 responses into Th1 responses (13, 14). We therefore investigated whether curdlan, a natural ligand of dectin-1 (15), could reprogram the function of breast tumor-infiltrating DCs to enable cancer rejection.

Breast cancer cell lines, Hs587T and MCF-7, were purchased from the American Type Culture Collection; MDA-MB-231 was purchased from Xenogen and cultured in nonselecting media. All lines are banked as low-passage stock from which working banks are periodically renewed. All lines were verified by gene microarrays twice in the past 7 years. Morphology, in vitro growth rate, and in vivo growth rate were the same as the original lines. The Mycoplasma test was performed regularly, and the cell lines were Mycoplasma-free for each in vitro and in vivo experiment.

Cell lines were cultured in RPMI [plus glutamine, 2 mmol/L; penicillin, 50 U/mL; streptomycin, 50 μg/mL; minimum essential medium (MEM) nonessential amino acids, 0.1 mmol/L; HEPES buffer, 10 mmol/L; and sodium pyruvate, 0.1 mmol/L] and 10% fetal calf serum in T150 flasks at a seed density of 2 × 10⁶ cells/25 mL. At 90% confluence, fresh medium was added, and cells were cultured for an additional 48 hours. Supernatant was centrifuged and stored at −80°C.

Peripheral blood mononuclear cells (PBMC) were obtained by leukapheresis from healthy donors (Institutional Review Board approved). Primary tissues from patients were obtained from the BUMC Tissue Bank and are exempt. Animal experiments were carried out with permission from the Institutional Animal Care and Use Committee.

β-Glucan, curdlan (Wako Pure Chemical Industries), was in PBS at a working concentration of 100 μg/mL. The working concentrations of the neutralization antibodies were as follows: 20 μg/mL for anti–dectin-1 (clone 259931; R&D Systems), 10 μg/mL anti–IL-12 (clone 20 C2; Thermo Scientific), 100 μg/mL anti–TGF-β (clone 1D11; R&D Systems), 50 μg/mL anti-CD103 (clone Ber-ACT8; BioLegend), and 100 μg/mL anti-β8 (clone 37E1). Curdlan was labeled with aminofluorescein (5-DTAF; Molecular Probes–Invitrogen).

DCs were enriched from PBMCs obtained after Ficoll-Paque Plus density gradient centrifugation (Stemcell Technologies) by negative selection with monoclonal antibodies (mAb) to CD3, CD9, CD14, CD16, CD19, CD34, CD56, CD66b, and glycophorin A (Human pan-DC Pre-Enrichment Kit; Stemcell Technologies). Cells were labeled with anti-human lineage cocktail-FITC (CD3, CD14, CD16, CD19, CD20, and CD56), CD123-PE (9F 5), CD11c-APC (S-HCL-3; BD Biosciences), and HLA-DR-APC-eflour780 (LN3; Sigma-Aldrich); lin⁻CD123⁻HLA-DR⁺CD11c⁺ DCs were sorted with FACSAria (BD Bioscience). DCs were seeded at 100 × 10³ cells per well in 200 μL of RPMI with 10% human AB serum, and cultured with medium alone or in the presence of 20 ng/mL of rhTSLP (R&D Systems), or tumor-derived products. After 48 hours, DCs were harvested, washed, and analyzed or used in experiments.

Optimum cutting temperature (OCT)–embedded (Sakura Finetek USA), snap-frozen tissues were cut at 6 μm and air-dried on Superfrost slides (Cardinal Health). Frozen sections were fixed with cold acetone for 10 minutes. Dectin-1 was stained with mAbs prepared in-house (clone12.2D8.2D4) followed by Alexa Flour 488 or 568 goat anti-mouse immunoglobulin G 1 (IgG1; Invitrogen). Cytokeratin 19 was labeled with clone A53-BA2 (Abcam) followed by Alexa Fluor 568 goat anti-mouse IgG2a (Invitrogen). CD83 was stained with clone HB15a (Immunotech) followed by Alexa Fluor 568 goat anti-mouse IgG2b (Invitrogen). CD20 was stained with clone L26 (Dako) followed by Alexa Fluor 488 goat anti-mouse IgG2a (Invitrogen). Directly labeled antibodies used were fluorescein isothiocyanate (FITC) anti–HLA-DR (clone L243; BD Biosciences) and FITC anti-CD11c (clone KB 90; Dako). Finally, sections were counterstained for 2 minutes with the nuclear stain 4′,6-diamidino-2-phenylindole (DAPI; 3 μmol/L in PBS; Invitrogen–Molecular Probes).

mAbs to human OX40L-PE (clone Ik-1), HLA-DR (clone L243), lineage cocktail-FITC (CD3, CD14, CD16, CD19, CD20, and CD56), CD11c-APC (clone S-HCL-3), CD3-PerCP (clone SK7), CD4-PE-Cy7 (clone SK3), CD8-APC-Cy7 (clone SK1), CD80-PE (L307.4), CD86-FITC [clone 2331(FUN-1)], CD70-PE (Ki-24), CD83-FITC (HB15e), IL-13-PE (JES10-5A2), TNF-α-PECy7 (mAb11), IFN-γ–Alexa Flour 700 (B27), pSTAT4-FITC (38/p-stat4), pSTAT6-PE (J91-99358.11), pSTAT3-AF647 (4/pStat3), and pSTAT5-AF647 (16) were obtained from BD Biosciences. mAb to MHC class I-PE (W6/32) was from Dako. IL-10-Pacific blue (JES3-9D7) and Perforin-PE (dG9) were obtained from eBioscience. mAbs to IL-17A-PerCP Cy5.5 (BL168), CD103-Alexa Flour 647 (Ber-act8), Granzyme A-Pacific blue (GB9), and Granzyme B-Alexa Flour 700 (GB11) were obtained from BioLegend; anti-integrin β8 (14E5) was conjugated with Alexa Fluor 488 in-house.

For surface staining, cells were incubated with antibodies for 30 minutes at 4°C in the dark, washed and fixed with 1% paraformaldehyde (PFA), events of stained cells were acquired with FACSCanto or LSR-II (BD Biosciences), and analyzed with the FlowJo software (TreeStar). For intracellular cytokines, cells were stained using the BD Cytofix/Cytoperm Fixation/Permeabilization Kit according to the manufacturer's instructions. For pSTATs staining, cells were fixed with 2% to 4% formaldehyde for 10 minutes at 37°C and permeabilized with ice-cold methanol for 30 minutes at 4°C. Cells were washed and stained with mAbs to pSTAT3, pSTAT4, pSTAT5, and pSTAT6 for 30 minutes at room temperature.

T cells from DC-T cocultures were resuspended at 10⁶ cells/mL in medium and activated for 5 hours with phorbol 12-myristate 13—acetate (PMA) and ionomycin (Iono). Brefeldin A (GolgiPlug; BD Biosciences) and monensin (GolgiStop; BD Biosciences) were added for the last 2.5 hours. The BD Cytofix/Cytoperm Fixation/Permeabilization Kit was used according to the manufacturer's instructions. Labeled samples were acquired with FACSCanto or LSR-II (BD Biosciences). Whole-tissue fragments of tumors from humanized mice (4 mm × 4 mm × 4 mm, 0.015–0.030 g, approximately) were placed in culture medium with 50 ng/mL of PMA (Sigma-Aldrich) and 1 μg/mL of ionomycin (Sigma-Aldrich) for 18 hours. Cytokine production was analyzed in the culture supernatant by Luminex.

Total T cells were enriched from apheresis using magnetic depletion of other leukocytes (EasySep Human T Cell Enrichment Kit; Stemcell Technologies). Blood DCs cultured with medium, TSLP, or tumor-derived factors were cocultured with naïve allogeneic T cells in a ratio of 1:5. For curdlan treatment, DCs were preincubated with curdlan for 3 minutes at room temperature.

NOD.Cg-Prkdc(scid)β2m(tm1Unc)/J, abbreviated NOD/scid/β2 null mice were irradiated the day before tumor implantation. Tumors were injected with 1 × 10⁶ monocyte-derived DCs, and with autologous T cells: 10 × 10⁶ CD4⁺ T cells admixed with 10 × 10⁶ CD8⁺ T cells. Monocyte-derived DCs were generated by culturing the adherent fraction of PBMCs with 100 ng/mL of granulocyte macrophage colony-stimulating factor (GM-CSF; Genzyme) and 10 ng/mL of IL-4 (R&D Systems). CD4⁺ and CD8⁺ T cells from the same donor as DCs were positively selected from thawed PBMCs according to the manufacturer's instructions (Miltenyi Biotec) to >90% purity. Tumor volume was monitored every 2 to 3 days: [(short diameter) 2 × long diameter]/2. Tumors were injected with 100 μg/mL of curdlan at days 3, 6, and 9 after implantation.

CD8⁺ T cells (20,000) sorted from DC-T cocultures and labeled with carboxyfluorescein succinimidyl ester (CFSE) were put on acetone-fixed breast tumor sections and incubated at 37°C. After 1 hour, the slides were washed to remove the unbound cells, fixed with 4% PFA for 10 minutes, treated with background buster for 30 minutes at room temperature, stained with cytokeratin, and finally counterstained for 2 minutes with the nuclear stain DAPI.

NOD/scid/β2 null mice were subcutaneously injected with 10 × 10⁶ MDA-MB231 cells. CD8⁺ T cells (500,000) sorted from DC-T cocultures and labeled with CFSE were injected into the tumors. After 3 days, the tumors were harvested and frozen with OCT or digested with collagenase (2.5 mg/mL; Roche Diagnostics), and processed to single-cell suspension. Some groups of mice were left for tumor growth monitoring.

Total RNA was purified using the mirVana miRNA Isolation Kit (Invitrogen). RNA integrity was assessed using the Bioanalyzer 2000 (Agilent). Target labeling was carried out using the TargetAmp Nano-g Biotin-aRNA Labeling Kit for the Illumina System (Epicentre). Labeled RNA was hybridized onto HumanHT-12 v4 Expression BeadChips (Illumina). Illumina GenomeStudio version 1.9.0 software was used to subtract background and scale samples to the global average signal intensity. Ingenuity pathway analysis (IPA) was applied to reveal transcriptional networks as described previously (17).

Immunofluorescence analysis of tissues from 27 primary breast cancers (Supplementary Table S1) revealed the presence of β-glucan receptor dectin-1 in all samples with CD11c⁺CD20⁻HLA-DR⁺CD83⁺ mature DCs; dectin-1–positive cells were found in the peritumoral areas (Fig. 1).

To establish whether the ligation of dectin-1 by curdlan in the tumor microenvironment might affect breast cancer progression in vivo, we used a humanized mouse model of human breast cancer that we have described earlier (2, 3). Intratumoral administration of 10 μg of curdlan prevented breast cancer progression (Hs578T breast cancer cell line) and was as effective as the neutralizing anti-TSLP receptor antibody (Fig. 2A). The antitumor effect of curdlan has been observed in five independent experiments with a total of 23 mice that had been grafted with monocyte-derived DCs and autologous T cells obtained from several donors (Fig. 2B). Breast tumor progression in this model is dependent on IL-13; as tumors do not grow in the absence of IL-13 or in the PBS control (2, 3), we analyzed IL-13 production by breast cancers that were harvested from humanized mice and activated with PMA/Iono. When compared with controls, curdlan-treated tumors produced significantly less IL-13 (DC+T: 1,038 ± 115 pg/mL; DC+T+curdlan: 361 ± 62 pg/mL; n = 23; P < 0.0001) but similar levels of IFN-γ (DC+T: 6,880 ± 1,796 pg/mL; DC+T+curdlan: 10,669 ± 2,081 pg/mL; n = 23; P = 0.17) and IL-10 (DC+T: 41 ± 7.7 pg/mL; DC+T+curdlan: 38 ± 8 pg/mL; n = 23; P = 0.83; Fig. 2C). We have shown earlier that blood DCs as well as monocyte-derived DCs exposed to breast cancer cell supernatants (BCsups), such as MDA-MB231, Hs578T, and MCF-7 (Supplementary Table S2), which express and secrete TSLP, can induce the differentiation of naïve T cells into iTh2 cells (2, 3). To determine whether curdlan prevents the breast cancer–induced polarization of DCs, purified blood Lin^negCD123^lowHLA-DR⁺CD11c⁺ DCs were exposed for 48 hours to BCsups with and without curdlan, and subsequently cocultured in vitro with naïve allogeneic CD4⁺ T cells for 7 days. Thereafter, T cells were activated for 5 hours with PMA/Iono and analyzed using intracellular cytokine staining (ICS) and flow cytometry (Fig. 2D). As expected, CD4⁺ T cells exposed to DCs that had been pretreated with BCsups alone produced both IL-13 and TNF-α (22% ± 3% of CD4⁺ T cells). In contrast, T cells exposed to DCs treated with both BCsups and curdlan produced less IL-13 (6% ± 0.3% of CD4⁺ T cells; n = 13; P < 0.0001; Fig. 2E). In both cases, CD4⁺ T cells produced IFN-γ (+BCsup-DC: 26% ± 0.5%; and +BCsup/curdlan-DC: 33% ± 1.5% of CD4⁺ T cells, respectively; n = 13; P = 0.0002; Fig. 2E). Thus, curdlan inhibits the progression of human breast cancer by preventing the generation of protumor iTh2 cells.

To determine whether curdlan can reprogram the function of tumor-conditioned DCs, we sorted OX40L⁺ and OX40L⁻ DCs that arise in response of blood DCs to BCsups. The sorted DCs were then exposed to curdlan for 24 hours, washed and cocultured with naïve allogeneic T cells. As expected, OX40L⁺ DCs induced T cells to express IL-13, whereas OX40L⁻ DCs did not. Treatment of OX40L⁺ DCs with curdlan altered their T-cell polarization capacity as no IL-13 was induced (Fig. 3A). Adding curdlan to DCs also prevented the induction of OX40L by BCsups (BCsups DCs: 25% ± 2%, n = 10; BCsups DC+curdlan: 6% ± 1.1%, n = 10; P < 0.0001; Fig. 3B). The inhibition of OX40L expression by curdlan was also observed when DCs were treated with human recombinant TSLP (Fig. 3C). Conversely, the addition of anti–dectin-1 antibodies, which block the binding of curdlan to DCs (13), before curdlan treatment, allowed OX40L expression by DCs exposed to BCsups in several independent experiments (Fig. 3D and E), demonstrating that curdlan does indeed engage dectin-1. In vivo the administration of anti–dectin-1 antibodies to developing breast cancer tumors prevented the protective effect of curdlan (Fig. 3F). These results confirm that curdlan acts through the dectin-1 expressed by tumor-infiltrating DCs in breast cancer.

We then observed that DCs treated with BCsups and curdlan showed high levels of CD83, CD80, CD86, CD70, and MHC class I, indicating that curdlan is able to induce DC maturation in the presence of breast cancer–derived factors (Fig. 4A; ref. 18). Thus, curdlan blocks specifically OX40L expression without interfering with the other components of the DC-maturation program. OX40L transcription in DCs depends upon the phosphorylation of STAT5 and STAT6 (19). As we showed earlier, STAT6 in both the leukocyte infiltrate and the cancer cells is activated in the breast cancer microenvironment (13). Exposure of BCsup-DC to curdlan led to enhanced phosphorylation of STAT4 and decreased phosphorylation of STAT6 (Fig. 4B), thereby resulting in an increase in the pSTAT4:pSTAT6 ratio. This switch in the activation pattern of STATs was associated with increased secretion of IL-12p70 by curdlan-treated DCs (Fig. 4C). Adding IL-12–neutralizing antibodies to cocultures of naïve allogeneic T cells with curdlan-treated BCsup-DC restored the generation of iTh2 cells (Fig. 4D). Thus, curdlan enables STAT4 activation in BCsup-DC, which is associated with increased IL-12 production and subsequent Th1 response.

Transcriptome analysis revealed the overexpression of 314 transcripts and the underexpression of 873 transcripts by curdlan-treated BCsup-DC (Fig. 4E). IPA of the overexpressed transcripts revealed networks centered on NF-κB, IL-6, and TNF (Fig. 4E). The underexpressed transcripts formed networks centered on several transcription factors (Fig. 4E). Curdlan-exposed DCs showed abundant transcription of DC-maturation markers, such as CD86 and TNFSF9 (4-1BBL); cytokines, such as GM-CSF, TNF, IL-6, IL-12, IL-15, and IL-23; integrins, including ITGB8 that is involved in the activation of TGF-β (20); and several molecules that might facilitate migration, including matrix metalloproteinase 7 (MMP7; Supplementary Table S3). MMP7 might facilitate DC migration to the draining lymph nodes, a feature that seems blocked in breast cancer–infiltrating DCs (21). Conversely, curdlan-exposed DCs underexpressed CD14, CD68, and CSF1R, all of which are associated with an immature DC phenotype. Consistent with DC maturation, CCR6, which contributes to immature DC retention at the tumor site by binding to MIP3-α (21), was also underexpressed. Thus, curdlan prevents the polarization of DCs induced by soluble tumor factors and TSLP.

As CD8⁺ T cells are essential effectors of antitumor immunity, naïve allogeneic CD8⁺ T cells were cocultured with BCsup-DC, exposed or not exposed to curdlan. ICS at day 7 revealed that upon PMA/Iono restimulation, CD8⁺ T cells cultured with BCsup-DC produce IL-13 (+BCsup-DC: 23% ± 1.3%; n = 9), IFN-γ, and TNF (Fig. 5A and B), indicating a partial type II polarization. However, CD8⁺ T cells cultured with curdlan-treated BCsup-DC displayed a type I phenotype with few IL-13–producing CD8⁺ T cells (+BCsup-DC: 23% ± 1.3%; +BCsup/curdlan-DC: 2% ± 1%; n = 9; P < 0.0001), and predominantly IFN-γ–producing CD8⁺ T cells (+BCsup-DC: 53% ± 1%; +BCsup/curdlan-DC: 68% ± 1.6%; n = 9; P < 0.001; Fig. 5A and B).

CD8⁺ T cells cultured with BCsup-DC expressed high levels of perforin but low levels of granzymes A and B (Fig. 5C and D). Similar to monocyte-derived DCs (13, 22), curdlan-exposed BCsup-DC allowed the generation of CD8⁺ T cells expressing high levels of granzymes A and B (Fig. 5C and D). To test their effector function, CD8⁺ T cells were labeled with CFSE and cultured with BCsup-DC with or without curdlan treatment for 6 days. Then, proliferating CFSE-negative CD8⁺ T cells were sorted and injected into breast cancer tumors established in immunodeficient mice. At day 3 after injection, CD8⁺ T cells generated with curdlan-treated BCsup-DC persisted in the breast cancer microenvironment better than CD8⁺ T cells generated by BCsup-DC (Fig. 5E). CD8⁺ T-cell persistence within the tumor was associated with tumor necrosis (Fig. 5F). Curdlan exposure of BCsup-DC resulted in the enhanced transcription of IL-15, IL-15-RA, and 4-1BBL (Supplementary Table S3), molecules that are known to play important roles in the generation of high-avidity CD8⁺ effector T cells, facilitating cancer rejection (23–25).

The accumulation and persistence of CD8⁺ T cells in cancer nests is critical for cancer rejection. CD103 integrin allows the retention of effector and memory T cells (26) via binding to E-cadherin expressed on epithelial cells (27–30). DCs exposed to curdlan showed an increased ability to induce CD103 on CD8⁺ T cells (Fig. 6A and B). To assess whether these CD103⁺CD8⁺ T cells adhered to breast cancer, we used a modified Stamper–Woodruff tissue binding assay (21). Proliferating CFSE-negative allogeneic CD8⁺ T cells were sorted from cocultures with DCs, relabeled with CFSE (green), and overlaid on frozen breast cancer tissue sections to allow adherence. After 60 minutes, tissue sections were washed and counterstained with anti-cytokeratin mAb (red) to visualize cancer cells. Numbers of bound T cells per 0.15 mm² cytokeratin⁺ areas were counted on a series of consecutive tissue sections. CD8⁺ T cells exposed to curdlan-treated DCs adhered significantly more to frozen breast cancer tissue sections (Fig. 6C; +BCsup-DC: 7 ± 1; +BCsup/curdlan-DC: 26 ± 2; n = 20; P < 0.0001) and blocking CD103 with an mAb decreased their numbers (Fig. 6C; +BCsup/curdlan-DC+isotype antibody: 22 ± 3; +BCsup/curdlan-DC+aCD103: 2 ± 0.5; n = 20; P < 0.0001). The binding of CD8⁺ T cells to breast cancer tissue sections was also decreased when BCsup-DCs were pretreated with anti–dectin-1 antibody before curdlan treatment (Fig. 6D; +BCsup/curdlan-DC: 32 ± 4; +BCsup/aDectin/curdlan-DC: 15 ± 3; n = 10; P = 0.003). Thus, curdlan exposure enables DCs to induce the differentiation of CD103⁺CD8⁺ T cells in a dectin-1–dependent manner.

Intratumoral injection of curdlan increased the frequency of CD103⁺CD8⁺ T cells in breast cancer tumors in vivo (Fig. 6E; DC+T: 9% ± 0.3% of CD8⁺ T cells, n = 3; DC+T+curdlan: 31 ± 1.2 of CD8⁺ T cells, n = 4; P < 0.0001). When sorted, these CD8⁺ T cells triggered tumor necrosis upon transfer into tumors established in immunodeficient mice (Fig. 6F). To establish whether the activated CD8⁺ T cells can inhibit the development of highly proliferative tumors, we used MDA-MB231 breast cancer cells that can grow in an immune microenvironment–independent manner. A single injection of CD8⁺ T cells elicited by BCsup-DC treated with curdlan completely inhibited breast cancer progression in a manner dependent upon the expression of CD103 (Fig. 6G). Indeed, breast cancer tumors only grew in mice that received control CD8⁺ T cells or in the presence of CD103-blocking antibody (Fig. 6G). Finally, CD8⁺ T cells elicited by curdlan-treated DCs were able to reject established breast cancers in vivo upon repeated adoptive transfer of as few as 0.5 × 10⁶ T cells (Fig. 6H).

To determine the mechanism by which DCs enabled the induction of CD103 expression in CD8⁺ T cells, we analyzed the role of TGF-β1 as it induces CD103 expression on T cells (31, 32). Using TGF-β1–neutralizing antibodies and pharmacologic blockade of TGF-β1 by the TGF-β RI kinase inhibitor II (33), the ability of curdlan-treated DCs to induce the differentiation of CD103⁺CD8⁺ T cells was substantially reduced (Fig. 7A). Transcriptional profiling revealed that curdlan exposure enables the overexpression of ITGB8 in DCs (Supplementary Table S3). The product of this gene is a cell-surface receptor for the latent domain (LAP) of TGF-β (34). The binding to the integrin αvβ8 with subsequent metalloproteolytic cleavage of LAP represents a major mechanism of TGF-β activation in vivo (35). Consistent with RNA expression, curdlan-treated BCsup-DC showed increased cell-surface expression of αvβ8 (Fig. 7B). Adding antibodies that neutralize αvβ8 to CD8⁺ T-cell cocultures with curdlan-treated BCsup-DC resulted in the complete inhibition of CD103 expression by CD8⁺ T cells triggered as the result of DC exposure to curdlan (Fig. 7C). Thus, curdlan-treated DCs activate TGF-β1 through αvβ8 to induce CD103⁺CD8⁺ T cells that reject breast cancer cells. The impact of curdlan on DCs is unique as DCs activated with the TLR8 ligand or poly I:C do not express αvβ8 (Fig. 7D).

Thus, reprogramming tumor-infiltrating DCs via dectin-1 ligation enables the simultaneous blockade of Th2 inflammation and induction of mucosal CD8⁺ T cells that are able to reject established cancers in vivo. This opens a novel avenue for immunotherapy of breast and pancreatic cancer, where links between type II inflammation and poor prognosis have been demonstrated.

Our previous studies have established the roles of tumor cells, DCs, and iTh2 cells in the progression of breast cancer (2, 3). Here, we show an immunotherapy strategy for breast cancer based on the reprogramming of tumor-infiltrating DCs in situ by targeting pattern-recognition receptor dectin-1. Indeed, the direct engagement of dectin-1 via intratumoral delivery of its ligand (β-glucan) initiated the reprogramming of DC maturation, resulting in the broad modulation of tumor-infiltrating CD4⁺ and CD8⁺ T-cell function leading to breast cancer rejection. The key principle is a simultaneous blockade of protumor iTh2 response, a switch to Th1 immunity, and an amplification of a potent antitumor CD8⁺ T-cell immunity. The direct binding of β-glucan to tumor-infiltrating DCs allows the reprogramming of their function, including the blockade of iTh2 cells secreting IL-4 and IL-13 in favor of the generation of IFN-γ–secreting CD4⁺ T cells, thus corroborating results from earlier studies (36, 37). β-Glucan–exposed DCs induced the generation of CD8⁺ T cells expressing CD103, a ligand for E-cadherin, with superior capacity to accumulate in and to reject breast cancer in vivo.

Suppression of type II responses is linked with enhanced IL-12 production by DCs. Interestingly, the blockade of IL-12 in DC-T-cell cocultures restored iTh2 differentiation, even though we have shown earlier that iTh2 differentiation is dependent on OX40L (3). A possible explanation is that when IL-12 is blocked and the CD40L signal is provided by T cells, the OX40L can be expressed and drive T-cell polarization (38). Whereas the abundance of IL-12 upon curdlan exposure was expected, genomic profiling revealed several intriguing transcripts, including Notch 2 and IL1F9 (IL-36γ). In the mouse, DC-specific deletion of the Notch2 receptor caused a reduction of DC populations in the spleen, and was associated with the loss of CD11b⁺CD103⁺ DCs in the intestinal lamina propria and a corresponding decrease of IL-17–producing CD4⁺ T cells in the intestine (39). The IL-36 receptor pathway has been suggested in the regulation of IFN-γ secretion by murine CD4⁺ T cells (40, 41). Furthermore, IL-36γ has been shown as downstream of the dectin-1/Syk signaling pathway upon exposure to Aspergillus fumigatus (42). Thus, curdlan exposure in the presence of tumor-derived factors leads to phenotype switch, and enables DC commitment to induce IFN-γ–secreting CD4⁺ T cells. Although assessment of the global IFN-γ secretion at the tumor level does not reveal a difference between curdlan-treated and –untreated tumors, we observe a clear difference at the level of CD4⁺ T cells. These results suggest that other cells might contribute to IFN-γ secretion in untreated tumors.

The impact on mucosal CD8⁺ T-cell differentiation was specific to curdlan-dectin-1 signaling and could not be induced by exposure of DCs to TLR8 ligand CL075 or to poly I:C (43). Dectin-1–dependent mucosal marker of CD8⁺ T cells is a CD103 integrin αE, which forms a heterodimer with the integrin β7 allowing peripheral CD8⁺ T cells to be retained in the epithelial compartments (44, 45). CD103 specifically binds E-cadherin that is expressed on murine and human epithelial cancer cells (27, 28). The expression of CD103 on CD8⁺ T cells seems to depend mostly upon TGF-β1 (31, 32). Studies on GVHD in mice lacking TGF-β receptor signaling demonstrated that the effector CD8⁺ T cells infiltrating the intestinal epithelium did not express CD103 and were less pathogenic (46). We have shown previously that human CD1c⁺ DCs use activated membrane-bound TGF-β1 to induce CD103 expression on proliferating CD8⁺ T cells, both in the allogeneic and autologous influenza-specific T-cell responses (47). Herein, we find that in the breast cancer environment, CD1c⁺ DCs are largely inhibited by TSLP in their capacity to generate CD103-expressing CD8⁺ T cells. However, dectin-1 engagement enables DCs to express integrin β8, thereby facilitating TGF-β1 activation. Accordingly, whereas in the influenza model CD103 expression was contact-dependent (47), here the activity could be transferred by exposure to DC supernatant.

In the context of cancer, CD103 expression by CTL mediates adherence to E-cadherin, resulting in cancer rejection (29). Indeed, mucosal homing and retention of CD8⁺ T cells is important for mucosal cancer vaccines (16). For example, only intranasal vaccination elicited mucosal-specific CD8⁺ T cells expressing the mucosal integrin CD49a (16). These results confirm the critical role of the route of immunization for the trafficking of effector T cells (48, 49) and the critical role of tissue DCs in imprinting the trafficking patterns of elicited T cells (50). Here, we provide another mechanism by which CD8⁺ T cells can be equipped with molecules allowing mucosal retention.

In summary, our studies have identified a number of targets generated by tumor-infiltrating DCs and T cells, the ligation of which results in tumor destruction in vivo by the human immune system in humanized mice. These include OX40L, IL-13, and now dectin-1; these agents act in a unique pathway that we have characterized. Whereas we need to characterize the impact of dectin-1 engagement on other cells present in the tumor microenvironment, in diseases where the iTh2 signature is associated with poor outcomes, as is the case in breast (4, 5) and pancreatic (6) cancers, the prevention of cancer-promoting effects combined with the expansion of potent CD8⁺ T-cell immunity might represent a novel option for these patients.

S. Nishimura has received a commercial research grant from MedImmune, LLC. No potential conflicts of interest were disclosed by the other authors.

Conception and design: T.-C. Wu, K. Xu, C.I. Yu, A. Pedroza-Gonzalez, J. Banchereau, S. Oh, K. Palucka

Development of methodology: T.-C. Wu, K. Xu, F. Marches, C.I. Yu, J. Martinek, A. Pedroza-Gonzalez, S. Nishimura, Y.-J. Liu, K. Palucka

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T.-C. Wu, K. Xu, F. Marches, J. Martinek, E. Anguiano, G.J. Snipes, J. O'Shaughnessy, K. Palucka

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T.-C. Wu, K. Xu, R. Banchereau, J. Martinek, E. Anguiano, J. O'Shaughnessy, V. Pascual, K. Palucka

Writing, review, and/or revision of the manuscript: T.-C. Wu, K. Xu, R. Banchereau, F. Marches, E. Anguiano, A. Pedroza-Gonzalez, J. O'Shaughnessy, V. Pascual, J. Banchereau, K. Palucka

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K. Xu, E. Anguiano

Study supervision: K. Palucka

The authors thank the patients and the volunteers; Luz S. Muniz, Joseph Fay, and the Cores at BIIR, including Clinical, Apheresis, Flow Cytometry, and Imaging Core and the Animal Facility; and Jennifer L. Smith for the help provided. K. Palucka acknowledges the support from the BIIR, Baylor University Medical Center Foundation, Cancer Prevention Research Institute of Texas, and NIH/NCI.

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.