Thymic stromal lymphopoietin fosters human breast tumor growth by promoting type 2 inflammation

Our earlier study using a pilot cohort of 19 samples of primary breast cancer tumors revealed the secretion, upon activation with PMA and ionomycin, of both type 1 (IFN-γ) and type 2 (IL-4 and IL-13) cytokines (Aspord et al., 2007). The current study extends the analysis to a total of 99 consecutive samples (Table S1). Supernatants of activated tumor fragments display high levels of IFN-γ, IL-2, IL-4, IL-13, and TNF (Fig. 1 A and Table S1). Supernatants from tumor sites contained significantly higher levels of IL-2, type 2 (IL-4 and IL-13), and inflammatory (TNF) cytokines than those from macroscopically uninvolved surrounding tissue (Fig. 1 B and Table S1). Whereas a significant correlation can be observed between IL-2 levels in tumor and adjacent tissue (P = 0.02), IL-4 and IL-13 levels are not correlated (P = 0.5), further suggesting polarization of cytokine environment in breast tumors (Fig. 1 B). IFN-γ did not correlate with other cytokine levels (Table S1). However, levels of TNF were correlated with those of IL-13 (P < 0.0001; r = 0.62; n = 98) and IL-4 (P = 0.0175; r = 0.31; n = 59; Fig. 1 C). Thus, this survey of cytokine expression suggested Th2 polarization in the breast tumor microenvironment.

To identify the cells producing these cytokines, single-cell suspensions were prepared from tumors; activated for 5 h with PMA and ionomycin; stained with antibodies against T cells and cytokines; and analyzed by flow cytometry. Gated viable CD4⁺CD3⁺ T cells expressed IL-13 (3.67%), most of them coexpressing IFN-γ and TNF (Fig. 1 D and Fig. S1 A). A small fraction of IL-13⁺CD4⁺ T cells coexpressed IL-4, but none expressed IL-10 (Fig. 1 D). Such T cells have been referred to as inflammatory Th2 cells that are involved in allergic inflammatory diseases (Liu et al., 2007). Flow cytometry analysis of consecutive tumor infiltrates (n = 22) shows the overall increased percentages of IL-4– and IL-13–secreting CD4⁺CD3⁺ T cells (P = 0.0313 and P = 0.0156, respectively) when compared with adjacent tissue samples (Fig. 1 D and Fig. S1 B). Thus, the difference between tumor tissue and adjacent tissue appears to be caused by both the increased numbers of infiltrating T cells and enhanced polarization. The analysis of frozen tissue sections further demonstrated that infiltrating T cells in primary breast cancer tumors express IL-13 (Fig. 1 E). Thus, breast cancer tumors are infiltrated with inflammatory Th2 cells.

Because OX40 ligation drives the differentiation of CD4⁺ T cells into inflammatory Th2 (Ito et al., 2005), we analyzed the presence of OX40L in primary breast cancer tumors. Immunofluorescence staining of frozen tissue sections of primary breast cancer tumors showed the expression, in 57 out of 60 analyzed tumors, of OX40L by a majority of HLA-DR^high cells (Fig. 2 A). These OX40L⁺ cells are located in peritumoral areas (Fig. 2 A). Flow cytometry analysis of single-cell suspensions further confirmed the expression of OX40L by a fraction of HLA-DR^high CD14^neg CD11c^high mDCs (Fig. 2 B). Paired analysis demonstrated that the tumor beds express higher percentages of OX40L⁺ mDCs than the surrounding tissue (P = 0.0156; n = 7 paired samples; mean ± SE for surrounding tissue = 1.5 ± 0.8% [n = 7] and for breast cancer tumors 11 ± 1.67% [n = 12], respectively; Fig. 2 B). Thus, breast cancer tumors are infiltrated with OX40L⁺ mDCs.

To identify the breast cancer tumor factors that induce OX40L on mDCs, LIN^negHLA-DR⁺ CD123⁻ CD11c⁺ mDCs were sorted from blood of healthy volunteers and exposed to breast cancer supernatants. These were generated from established breast cancer cell lines expanded in vitro (Hs578T, MDA-MB-231, MDA-MB-468, MCF7, HCC-1806, and T47D; Table S2) and breast cancer tumors established in vivo by implanting breast cancer cell lines in immunodeficient mice (Aspord et al., 2007). As illustrated in Fig. 2 C, mDCs exposed for 48 h to Hs578T and HCC-1806 supernatants expressed OX40L. Four of the five breast cancer cell lines, with the notable exception of T47D, induced OX40L expression on mDCs (Fig. S1 C).

To test whether primary breast cancer tumors could also regulate OX40L expression, fragments of primary tumors were sonicated, centrifuged, filtered, and used in cultures with blood mDCs. As illustrated in Fig. 2 D, mDCs acquired both CD83 (a DC maturation marker) and OX40L. To determine the impact of OX40L on the generation of inflammatory Th2 responses in breast cancer, blood mDCs were first exposed for 48 h to either TSLP or breast tumor–soluble fractions. Exposed mDCs were then used to stimulate naive allogeneic CD4⁺ T cells with either the anti-OX40L antibody or a relevant isotype control. Blocking OX40L prevented the expansion of IL13⁺CD4⁺ or TNF⁺CD4⁺ T cells by TSLP-primed mDCs (>50% inhibition; Fig. S2 A), mDCs exposed to Hs578T breast cancer cells (n = 4; median inhibition of IL13⁺CD4⁺ cells = 74%; range = 67–80%; Fig. 2, E and F), and mDCs exposed to sonicates of randomly selected primary breast cancer tumors (Fig. 2 F and Fig. S2 B). Thus, breast cancer cells produce factors that activate mDCs and induce them to express OX40L and to elicit inflammatory Th2 cells.

OX40L can be induced on mDCs by TSLP, an IL-7–like cytokine produced by epithelial cells (Liu et al., 2007; Ziegler and Artis, 2010). All tested breast cancer cell lines expressed and secreted TSLP (Fig. 3 A). Supernatants of some primary breast cancer tumors stimulated with PMA and ionomycin displayed up to 300 pg/ml TSLP (Fig. 3 B). The expression of TSLP by cancer cells was further analyzed using an anti-TSLP antibody and immunofluorescence of frozen breast cancer tumors generated in the xenograft model (Aspord et al., 2007). There, subcutaneous MDA-MB-231 tumors transplanted in mice expressed TSLP (Fig. 3 C). The specificity of the staining is demonstrated by pretreatment of the antibody with recombinant TSLP (Fig. S3).

Importantly, TSLP is expressed in 35 out of 38 analyzed primary breast cancer tumors obtained from patients regardless of grade, histology, or stage of analyzed tumors. Fig. 3 D illustrates the pattern of TSLP staining and coexpression with cytokeratin 19–positive cells. It demonstrates that TSLP is expressed in the cytoplasm and the nucleus of breast cancer cells that display IL-13 on their surface (Fig. 3 E). Importantly, TSLP is also expressed in lung and kidney metastasis of MDA-MB-231 tumors in humanized mice (Fig. 3 F) and in breast cancer tumor metastasis from patients (Fig. S4 A). TSLP expression is specific to epithelial cells as no staining can be found in tumor-infiltrating fibroblasts (Fig. S4 B). Furthermore, nonmalignant breast epithelia can also express TSLP (Fig. S4 C).

Thus, similar to normal skin or lung epithelium, breast epithelial cells and breast cancer cells have the capacity to express, produce, and secrete TSLP.

To determine the impact of blocking TSLP on the generation of inflammatory Th2 responses in breast cancer, blood mDCs were first exposed for 48 h to either TSLP or breast tumor–soluble fractions. Addition of anti-TSLP–neutralizing antibodies to breast cancer tumor supernatants inhibited their ability to induce OX40L on mDCs (Fig. 4 A). Such mDCs displayed a diminished capacity to expand IL13⁺CD4⁺ or TNF⁺CD4⁺ T cells (n = 3; median inhibition = 73%, range = 72–77%; Fig. 4, B and E). Likewise, adding anti-TSLP–neutralizing antibody to sonicate of randomly selected primary breast cancer tumors led to down-regulation of OX40L expression by mDCs (Fig. 4 C) and decreased expansion of IL13⁺TNF⁺CD4⁺ T cells (Fig. 4, D and E). Finally, when anti-TSLP receptor β chain (TSLPR) antibody was added to mDCs during their exposure to the supernatant of three different breast cancer cell lines (Hs578T, MDA-MB-231, and MCF7; Fig. 5 A), resulting mDCs showed a much diminished expansion of IL13⁺ CD4⁺ T cells (Fig. 5 B). Thus, TSLP is the factor secreted by breast cancer cells that contributes to generation of inflammatory Th2 responses.

Our results thus far suggest a role for the TSLP–OX40L axis in generation of IL13⁺TNF⁺CD4⁺ T cells, but do not establish whether this axis might actually contribute to breast cancer tumor development. To address this question, humanized mice were reconstituted with both Hs578T cells and T cells with or without anti-OX40L–, anti-TSLPR–, and anti-TSLP–neutralizing antibodies. As shown in Fig. 6 A, the administration of neutralizing anti-OX40L antibodies leads to significant inhibition of tumor development.

The administration of a neutralizing anti-TSLP antibody also results in the inhibition of tumor development (Fig. 6 B). TSLP blockade also leads to decreased secretion of IL-4 and IL-13 by tumor infiltrating T cells upon PMA and ionomycin activation (Fig. 6 C). Finally, the administration of antibody blocking TSLPR nearly completely blocks tumor development driven by CD4⁺ T cells (Fig. 6 D). These results indicate that the TSLP contributes to breast cancer pathogenesis.

Here, we show that breast cancer is infiltrated with inflammatory Th2 cells and that such T cells are driven by OX40L on DCs. Blocking OX40L in vitro prevents generation of these CD4⁺ T cells without impact on IL-10–producing CD4⁺ T cells. Blocking OX40L in vivo partially prevents T cell–dependent acceleration of breast cancer tumor development. OX40L is not constitutively expressed, but can be induced on DCs, macrophages, and B cells; e.g., upon CD40 engagement or cytokine signals such as TSLP or IL-18, as well as upon TLR stimulation (Ito et al., 2005; Croft et al., 2009). Thus, the presence of OX40L⁺ mDCs in breast tumors indicate sustained activation of DCs in tumor environment. Indeed, OX40L expression by DCs is driven by TSLP secreted from breast cancer cells. Accordingly, TSLP expression can be found in primary and metastatic tumors. Blocking TSLP reduces inflammation and partially inhibits tumor development.

Based on our results presented herein and in our earlier studies, we propose a vicious circle of smoldering type 2 inflammation that perpetuates breast cancer and which is maintained by TSLP. There, breast cancer attracts DCs, possibly through macrophage inflammatory protein 3 α (MIP3-α; Bell et al., 1999). Tumor-infiltrating DCs are then exposed to TSLP secreted by breast cancer cells, which triggers their maturation and OX40L expression. This might explain the aseptic mDC maturation that we found in breast cancer (Bell et al., 1999; Aspord et al., 2007). OX40L⁺ mDCs induce CD4⁺ T cells to secrete IL-13, as well as TNF. These inflammatory CD4⁺ T cells contribute to tumor development in an IL-13–dependent pathway (Aspord et al., 2007). Thus far, TSLP represents the only factor that activates mDCs without inducing them to produce Th1-polarizing cytokines (Liu et al., 2007). Under normal physiological conditions, TSLP appears to play a critical role in CD4⁺ T cell homeostasis in the peripheral mucosa-associated lymphoid tissues and in the positive selection and/or expansion of regulatory T (T reg) cells in the thymus (Watanabe et al., 2005a,b). In inflammatory conditions such as atopic dermatitis and asthma, epithelial cells markedly increase TSLP expression (Liu et al., 2007). The TSLP-activated DCs migrate to the draining lymph nodes, prime CD4⁺ T cells via OX40L to differentiate into inflammatory Th2 effector and memory cells, and thus initiate the adaptive phase of allergic immune responses. Interestingly, in breast cancer, OX40L⁺ mDCs are present in the tumor. It remains to be determined whether this reflects their inability to migrate from the tumor to draining lymph nodes. Indeed, such an evasion mechanism has been documented recently in human and mouse tumors showing an inhibition of DC migration from tumors to tumor-draining lymph nodes (Villablanca et al., 2010). This effect depends on tumor-derived ligands of the liver X receptors (Villablanca et al., 2010). It also remains to be determined whether these DCs are able to prime Th2 immunity in situ in tertiary lymphoid structures or whether their main role is to maintain the activation and survival of Th2 cells at the tumor site. Their ability to maintain Th2 cell phenotype and effector function is supported by our earlier studies showing that T cells isolated from experimental breast tumors and transferred to naive tumor-bearing humanized mice can promote tumor development even at low numbers and upon single injection (Aspord et al., 2007).

Our results add another feature to the role of OX40L in tumors. Indeed, several studies in mouse models of transplantable tumors suggested that engaging OX40 via an agonist antibody, OX40L.Fc, or transfected tumor cells and DCs appears to promote antitumor effects (Weinberg et al., 2000; Morris et al., 2001; Ali et al., 2004; Piconese et al., 2008). However, Tnfsf4 (gene coding OX40L) is regulated by the mRNA MIRN125B (Smirnov and Cheung, 2008), whose expression is down-regulated in breast cancer (Iorio et al., 2005). Tnfsf4 is up-regulated in ataxia telangiectasia carriers and patients (Smirnov and Cheung, 2008), who have been shown to have an increased risk of breast cancer (Swift et al., 1987). Furthermore, a recent pilot study described the expression of OX40 and OX40L in >100 tissue samples from invasive ductal carcinomas, and suggested a possible association of OX40 expression and lymph node metastatic status (Xie et al., 2010). OX40L signaling has several important features that might help explain the results observed in our and other studies. Thus, OX40L triggers Th2 polarization independent of IL-4, promotes TNF production, and inhibits IL-10 production by the developing Th2 cells, but only in the absence of IL-12. In the presence of IL-12, OX40L signaling instead promotes the development of Th1 cells that, like inflammatory Th2 cells, produce TNF but not IL-10 (Liu et al., 2007).

Interestingly, inflammatory Th2 cells coexist within the tumor immune environment alongside IL-10–secreting CD4⁺ T cells. Recent studies demonstrate a colocalization of cells with the phenotype of T reg cells) with mature DCs in lymphoid infiltrates in breast cancer (Gobert et al., 2009). A high number of FoxP3⁺ T reg cells was associated with disease relapse (Gobert et al., 2009). Thus, the niche finding and metastasis formation might be facilitated by inflammatory Th2 response, whereas in the established tumor a T reg cell response might prevail. Interestingly, CCL22 (a macrophage-derived chemokine) appears involved in the attraction of T reg cells (Gobert et al., 2009), but also, in cooperation with CCL17 (TARC), of inflammatory Th2 cells (Liu et al., 2007). How this fine balance between T reg cells and Th2 responses is regulated will require further study.

Two key questions arise from our work: (1) what are the mechanisms allowing TSLP release from cancer cells; and (2) what is the impact of IL-13 (and IL-4) on cancer cells, on the stroma, and on the immune infiltrate? Our data show that nonmalignant breast epithelia can express TSLP. This further demonstrates that cancer cells can exploit pathways that are present in the normal tissue to establish a microenvironment facilitating tumor development. The mechanisms regulating TSLP expression and secretion from cancer cells, including a potential link with oncogenic events, remain to be established. It will also be important to determine the impact of TSLP and inflammatory Th2 environment on the stromal fibroblasts. Indeed, recent studies point to the critical role of cross talk between cancer cells and fibroblasts in determining the type of microenvironment established by cancer cells originating from different types of breast cancers (Camp et al., 2011).

Interestingly, studies in mice suggest the role of Th2 cytokines in mammary gland development. Indeed, the differentiation of the luminal epithelial lineage appears to require autocrine signaling by Th2 cytokines, as shown by reduced differentiation and alveolar morphogenesis in both Stat6 and IL-4–/IL-13–deficient mice during pregnancy (Khaled et al., 2007). Yet, in the murine model of endogenous breast cancer type 2, polarized CD4⁺ T cells are essential to the metastatic process via secretion of IL-4, which induces macrophages to secrete epidermal growth factor (DeNardo et al., 2009). Similar to endogenous mouse model, in our model of transplanted metastatic human breast cancer IL-13 is derived from microenvironment. IL-13 can exert pro-cancer activity in several ways, including the triggering of TGF-β secretion (Terabe et al., 2000, 2003; Park et al., 2005; Shimamura et al., 2010). Furthermore, IL-4 exposure of cancer cells leads to the up-regulation of antiapoptotic pathways via mobilization of STAT6 (Zhang et al., 2008). We have shown that STAT6 is phosphorylated in primary breast cancer tumors (Aspord et al., 2007). All these antiapoptotic pathways are likely to synergize to promote the survival of cancer cells and facilitate metastasis. Importantly, such a protective effect on cancer cell susceptibility to apoptosis might increase their resistance to chemotherapy (Todaro et al., 2008) and immune-mediated cytotoxicity driven by Granzyme B (Sarin et al., 1997; Heibein et al., 2000). Thus, the TSLP–OX40L–IL13 axis might offer a novel therapeutic target.

DCs were purified from buffy coat of blood from healthy donors. In brief, DCs were enriched from mononuclear cells by negative selection using a mixture of antibodies against linage markers for CD3, CD14, CD16, CD19, CD56, and glycophorin A (Dynabeads Human DC Enrichment kit; Invitrogen). Cells from negative fraction were immunolabeled with anti–human FITC-labeled lienage cocktail (CD3, CD14, CD16, CD19, CD20, and CD56; BD); PE-labeled CD123 (mIgG1; clone 9F5; BD), QR-labeled HLA-DR (mIgG2a; clone HK14; Sigma-Aldrich) and APC-labeled CD11c (mIgG2b; clone S-HCL-3; BD). DCs (lin⁻, CD123⁻, HLA-DR⁺, CD11c⁺) were sorted in a FACSAria cytometer (BD). DCs were seeded at 100 × 10³ cells/well in 200 µl of medium (RPMI supplemented with 2 mM glutamine, 50 U/ml penicillin, 50 µg/ml streptomycin, 0.1 mM MEM nonessential amino acids, 10 mM Hepes buffer, 0.1 mM sodium pyruvate, and 10% of human AB serum). DCs were cultured with medium alone or in the presence of 20 ng/ml of TSLP, or different tumor derived products. After 48 h, DCs were harvested and washed. The stimulated cells were stained for phenotype analysis or co-cultured with allogeneic naive CD4⁺ T cells.

6-µm-frozen sections from tissues were fixed with cold acetone for 5 min. The sections were labeled with 5 µg/ml of anti-OX40L antibody (mouse IgG1, 8F4), followed by anti–mouse IgG conjugated to Texas red (Jackson ImmunoResearch Laboratories). For IL-13, tissue was labeled with 10 µg/ml of anti–IL-13 (polyclonal goat IgG; AF-123-NA; R&D System) followed with Texas red anti–goat IgG (Jackson ImmunoResearch Laboratories). TSLP was detected with 10 µg/ml of mouse anti-TSLP antibody prepared in-house (mIgG1; clone 14C3.2E11). Cytokeratin 19 was labeled with monoclonal antibody clone A53-BA2 (IgG2a; Abcam), followed by Alexa Fluor 568 goat anti–mouse IgG2a (Invitrogen). The following direct-labeled antibodies used were as follows: FITC anti–HLA-DR (mouse IgG2a; L243; BD), FITC anti-CD11c (clone; KB 90; Dako), FITC anti-Ki67 (clone ki-67; Dako), and Alexa Fluor 488 anti-CD3 (mIgG1; UCHT1; BD). Finally, sections were counterstained for 2 min with 3 µM of the nuclear stain DAPI (in PBS; Invitrogen). To confirm specificity of TSLP staining, primary anti-TSLP antibody was preincubated with 100 µg of recombinant human TSLP (R&D Systems) for 30 min at room temperature before staining of tissue sections that previously showed to be TSLP positive.

Cell suspensions from human breast carcinoma tissue and tumor or draining lymph nodes from humanized mice were used for phenotypic characterization of leukocytes. Cell suspensions were obtained by digestion with 2.5 mg/ml of collagenase D (Roche), and 200 U/ml of DNase I (Sigma-Aldrich) for 30–60 min at 37°C. The anti–human antibodies used were as follows: labeled FITC-labeled lineage cocktail (CD3, CD14, CD16, CD19, CD20, and CD56; BD); PE-labeled OX40L (mIgG1; clone Ik-1; BD); QR-labeled HLA-DR (mIgG2a; clone HK14; Sigma-Aldrich); APC-labeled CD11c (mIgG2b; clone S-HCL-3; BD); PerCP-labeled CD3 (mIgG1; clone SK7; BD); PECy7-labeled CD4 (mIgG1; clone SK3; BD); APCCy7-labeled CD8 (mIgG1; clone SK1; BD); CFS-labeled IL-4 (mIgG1; clone 3007; R&D Systems); Pacific blue–labeled IL-10 (rat IgG1; clone JES3-9D7; eBioscience); PE-labeled IL-13 (rat IgG1; clone JES10-5A2 BD); APC-labeled TNF (mIgG1; clone 6401.1111; BD); Alexa Fluor 700–labeled IFN-γ (mIgG1; clone B27; BD). Cells were incubated with the antibodies for 30 min at 4°C in the dark, and then washed three times and fixed with 1% paraformaldehyde to be analyzed in a FACSCalibur or LSR-II cytometer (BD). For intracellular cytokines, cells were stained using BD Cytofix/Cytoperm fixation/permeabilization kit according to the manufacturer’s directions.

Tumor factors were obtained from supernatant of established breast cancer cell lines cultured in vitro (Table S2; Soule et al., 1973; Hackett et al., 1977; Lacroix and Leclercq, 2004; Neve et al., 2006) or by sonication from tumor cell lines, human breast tumor tissue, or tumors from humanized mice. In brief, cell lines were culture in medium (RPMI supplemented with 2 mM glutamine, 50 U/ml penicillin, 50 µg/ml streptomycin, 0.1 mM MEM nonessential amino acids, 10 mM Hepes buffer, 0.1 mM sodium pyruvate, and 10% fetal calf serum), and when the cells reached 90% of confluence fresh medium was added and the cells were left in culture for an additional 48 h. For sonication, cells or tissues were placed in PBS and disrupted for 30 s at 4°C, with the power output adjusted at 4.5 level of the 60 sonic dismembrator (Thermo Fisher Scientific). Cellular debris was removed by centrifugation, and the supernatant was collected and stored at −80°C.

Tumor samples from patients diagnosed with breast carcinoma (in situ, invasive duct, and/or mucinous carcinoma of the breast, as well as lobular carcinoma) were obtained from the Baylor University Medical Center Tissue Bank (Institutional Review Board no. 005–145). Tumors and draining lymph nodes from mice implanted with breast cancer cell lines H578T and MDA-MB-231 were also analyzed. Whole-tissue fragments (4 × 4 × 4 mm, 0.02 g, approximately) were placed in culture medium with 50 ng/ml of PMA (Sigma-Aldrich) and 1 µg/ml of ionomycin (Sigma-Aldrich) for 16 h. Cytokine production was analyzed in the culture supernatant by Luminex. For intracellular staining, cells were resuspended at a concentration of 10⁶ cells/ml in medium and activated for 5 h with PMA and ionomycin; brefeldin A (GolgiPlug; BD) and monensin (GolgiStop; BD) were added for the last 2.5 h.

Total CD4⁺ T cells were enriched from PBMCs of healthy donors using magnetic depletion of other leukocytes (EasySep Human CD4⁺ T Cell Enrichment kit; STEMCELL Technologies, Inc.). Naive CD4⁺ T cells were sorted based on the expression of CD4⁺ CD27⁺ and CD45RA⁺. Activated mDCs with medium, TSLP, or tumor-derived factors were co-cultured with naive CD4⁺ T cells at a ratio of 1:5 for 7 d. After that, the cells were washed and restimulated for 5 h with 50 ng/ml PMA and 1 µl/ml ionomycin. Brefeldin A and monensin were added for the last 2.5 h, followed by surface and intracellular staining. For blocking OX40L, tumor-activated mDCs were co-cultured with naive CD4 T cells in the presence of 50 µg of anti-OX40L (Ik-5 clone) or control IgG2a isotype antibody. For blocking TSLP, tumor-derived factors were preincubated with 20 µg/ml of anti-TSLP antibody (rabbit; AB 19024) or normal rabbit IgG (R&D Systems) at room temperature for 30 min after DC activation. DCs were activated with the neutralized tumor-derived factors and finally co-cultured with naive CD4 T cells for 7 d. For TSLPR blocking, DCs were preincubated with anti-TSLP receptor antibody (clone AB81_85.1F11; mouse IgG1) for 3 min at room temperature.

Mice were humanized either by CD34⁺ hematopoietic progenitor cell (HPC) transplant or by co-administration of DCs and T cells as described previously (Aspord et al., 2007; Institutional Animal Care and Use Committee no. A01-005). CD34⁺HPCs were obtained from apheresis of adult healthy volunteers mobilized with G-CSF and purified as described previously (Aspord et al., 2007). The CD34⁻ fraction of apheresis was Ficoll purified, and the PBMCs obtained were stored frozen and used as a source of autologous T cells. 3 million CD34⁺ HPCs were transplanted intravenously into sublethally irradiated (12 cGy/g body weight of 137Cs γ irradiation) nonobese diabetic/LtSz-scid/scid β2 microglobulin–deficient (NOD/SCID/β2m^−/−) mice (Jackson ImmunoResearch Laboratories). After 4 wk of engraftment, 10 million Hs578T breast cancer cells were harvested from cultures and injected subcutaneously into the flanks of the mice. Mice were reconstituted with 10 million CD4⁺ T cells and 10 million CD8⁺ T cells autologous to the grafted CD34⁺ HPCs. CD4⁺ and CD8⁺ T cells were positively selected from thawed PBMCs using magnetic selection according to the manufacturer’s instructions (Miltenyi Biotec). The purity was routinely >90%. T cells were transferred at days 3, 6, and 9 after tumor implantation. For experiments with NOD/SCID/β2m^−/− mice, they were sublethally irradiated the day before tumor implantation. Mice were then reconstituted with 1 million monocyte-derived DCs (MDDCs) and autologous T cells as described in the previous paragraph. MDDCs were generated from the adherent fraction of PBMCs by culturing with 100 ng/ml GM-CSF (Immunex) and 10 ng/ml IL-4 (R&D Systems). Tumor size was monitored every 2–3 d. Tumor volume (ellipsoid) was calculated as follows: ([short diameter]² × long diameter)/2.

For different experimental purposes, tumor-bearing mice transferred with autologous T cells were injected intratumorally with 200 µg of anti-OX40L (clone IK-5; mIgG2a) blocking antibody, 100 µg of anti-TSLP antibody (clone AB19024; rabbit IgG), 100 µg of anti-IL-13 neutralizing antibody prepared in-house (clone 13G1.B2; mIgG1), 200 µg of anti-TSLPR neutralizing antibody (clone AB81_85.1F11; mouse IgG1), or isotype control, respectively at days 3, 6, and 9 after tumor implantation. For blocking TSLP, anti-TSLP antibody injection was also given at day 0 while the tumor was implanted.

Fig. S1 illustrates gating strategy for flow cytometry analysis of tumor-infiltrating lymphocytes and the accumulation of lymphocytes in tumor beds as compared with surrounding tissue. Fig. S2 shows that mDCs exposed to TSLP or to soluble tumor factors in the presence of anti-OX40L are unable to drive the differentiation of inflammatory Th2 cells. Fig. S3 shows the specificity of anti-TSLP antibody staining in frozen tissue sections. Fig. S4 shows the expression of TSLP in breast cancer cells and in nonmalignant breast tissue. Table S1 provides the pathological diagnosis information and Luminex data for cytokine levels of all 99 breast cancer patients used for cytokine analysis. Table S2 describes the characteristics of breast cancer cell lines studied.

We are grateful to Albert Barnes, Sebastien Coquery, Elizabeth Kraus, Mark Michnevitz, Gina Stella Garcia-Romo, Jenny Smith, and Lynnette Walters for help; Dr. Joseph Fay for help with healthy volunteers; Cindy Samuelsen for continuous support; and Drs. Sally M. Knox and Michael Grant at the Department of Surgery and Ms. Dan Su at the Department of Pathology at Baylor University Medical Center. We thank Dr. Michael Ramsay for continuous support.

This work was supported by the Baylor Health Care Systems Foundation, and the National Institutes of Health (grants R0-1 CA89440 and R21 AI056001 to A. Karolina Palucka; grants U19 AI057234, R0-1 CA78846, and CA85540 to J. Banchereau). J. Banchereau holds the Caruth Chair for Transplantation Immunology Research. A. Karolina Palucka holds the Ramsay Chair for Cancer Immunology Research.

The authors have no conflicting financial interests.