Dectin-1-activated dendritic cells trigger potent antitumour immunity through the induction of Th9 cells

Dectin-1 signalling in dendritic cells (DCs) has an important role in triggering protective antifungal Th17 responses. However, whether dectin-1 directs DCs to prime antitumour Th9 cells remains unclear. Here, we show that DCs activated by dectin-1 agonists potently promote naive CD4+ T cells to differentiate into Th9 cells. Abrogation of dectin-1 in DCs completely abolishes their Th9-polarizing capability in response to dectin-1 agonist curdlan. Notably, dectin-1 stimulation of DCs upregulates TNFSF15 and OX40L, which are essential for dectin-1-activated DC-induced Th9 cell priming. Mechanistically, dectin-1 activates Syk, Raf1 and NF-κB signalling pathways, resulting in increased p50 and RelB nuclear translocation and TNFSF15 and OX40L expression. Furthermore, immunization of tumour-bearing mice with dectin-1-activated DCs induces potent antitumour response that depends on Th9 cells and IL-9 induced by dectin-1-activated DCs in vivo. Our results identify dectin-1-activated DCs as a powerful inducer of Th9 cells and antitumour immunity and may have important clinical implications.

N aive CD4 þ T cells, on antigenic activation, differentiate into various T helper (Th) cell subsets, such as Th1, Th2, Th17 and T regulatory cells (Tregs) 1,2 . Th9 is a recently described Th cell subset characterized by the secretion of interleukin (IL)-9 (refs 3,4). Th9 cells and IL-9 (Th9/IL-9) are pro-inflammatory and appear to function in a broad spectrum of autoimmune diseases and allergic inflammation 5,6 . We and others have recently reported that adoptive transfer of Th9 cells induces potent therapeutic immunity against melanoma tumours in mice, better than other Th cells 7,8 . We have shown that Th9-derived IL-9 is critical in promoting an efficient host CD8 þ CTL-mediated antitumour immune response 8,9 . Th9-derived IL-9 was also shown to activate mast cells, which may contribute to Th9 cell-induced antitumour activities 7 . IL-9 has the potential to enhance the survival and proliferation of antitumour effector T cells 10 . These seminal findings provide an impetus for further investigation of efficient strategies to induce and expand Th9 cells for tumour immunotherapy.
Dendritic cells (DCs) are professional antigen-presenting cells (APCs) and play a crucial role in the induction of Th cells 22,23 . Dectin-1, a C-type lectin receptor, is expressed mainly by DCs, macrophages and neutrophils 24,25 . DCs sense fungal pathogens through dectin-1, which recognizes b-1-3-glucans present on the fungal cell wall, and trigger the host immune response against fungal pathogens 26 . Dectin-1 triggers Syk and Raf1 downstream signalling pathways, which subsequently regulate the activation of canonical and noncanonical NF-kB pathways 24 . Dectin-1 activation in DCs stimulates the secretion of IL-6, TNF-a and IL-12p40, which polarize naive CD4 þ T cells into Th17 and Th1 cells, the key effector cells for antifungal immunity 27,28 . However, whether dectin-1 activation in DCs favours the induction of antitumour Th9 cells remains unclear.
In this study, we found that dectin-1 activation in DCs potently promotes the induction of Th9 cells. We show that dectin-1 signalling stimulates DCs to overexpress TNFSF15 and OX40L, which are responsible for promoting Th9 cell differentiation primed by dectin-1-activated-DCs in vitro. Syk, Raf1 and NF-kB signalling pathways triggered by dectin-1 are required for dectin-1induced expression of TNFSF15 and OX40L. Furthermore, immunization of mice bearing melanoma or myeloma tumours with dectin-1-activated DCs induces potent antitumour responses that depend on Th9 cells and IL-9. Our results thus identify dectin-1-activated DCs as a powerful inducer of Th9 cells and antitumour immunity and may have important clinical implications.
To examine the role of dectin-1 signalling in activating naturally occurring DCs in Th9 differentiation, mouse spleen CD11c þ cells were isolated, activated by curdlan and cocultured with T cells. Similarly, Curdlan-treated natural DCs drove Th9 differentiation by enhancing Th cell Il9 expression as compared with untreated natural DCs ( Supplementary Fig. 2).
Dectin-1 stimulates DCs to express TNFSF15 and OX40L. Th cell differentiation relies on specific profiles of cytokines and costimulatory molecules 1,2 . To explore the molecular mechanisms by which dectin-1-activated DCs drove Th9 differentiation, we performed gene expression profiling analyses in BMDCs versus CurDCs. Among 26,423 mouse genes, we identified 42 upregulated genes of cytokines, chemokines and costimulatory surface molecules that might be involved in Th cell activation, polarization and chemotaxis (Table 1).
Together, these results demonstrated that dectin-1 activation stimulated DCs to express a specific profile of cytokines, chemokines and surface costimulatory molecules, especially the TNF/receptor family members, suggesting the potential mechanisms for dectin-1-activated DCs in the induction of Th9 cells.
Dectin-1 directs Th9 priming via TNFSF15 and OX40L. Tnfsf15 was the top 4 among the 42 upregulated genes of cytokines, chemokines and surface costimulatory molecules in CurDCs (Table 1). We therefore investigated its role in mediating Th9 differentiation primed by dectin-1-activated DCs. First, we examined the function of TNFSF15 in Th9 cell differentiation. As shown in Fig. 4a,b, the addition of TNFSF15 significantly increased IL-9 mRNA and protein expressions by Th9 cells. To investigate the contribution of TNFSF15 to Th9 cell differentiation primed by dectin-1-activated DCs, a TNFSF15 neutralizing antibody (aTNFSF15) was used. The addition of aTNFSF15 significantly inhibited IL-9, Irf4 and Il13 expression in CurDC-induced Th9 cells as compared with the treatment of control IgG (Fig. 4c,d and Supplementary Fig. 7A,B); whereas the addition of aTNFSF15 did not affect IL-9 expression of BMDC-induced Th9 cells (Fig. 4c,d and Supplementary  Fig. 7A,B). These results indicated that TNFSF15 contributed to Th9 cell differentiation primed by dectin-1-activated DCs.
Recent study has proposed a role of OX40/OX40L axis in Th9 cell differentiation 11 . Thus, we next sought to determine whether OX40L expressed by dectin-1-activated DCs is required for the Th9 cell differentiation primed by dectin-1-activated DCs. Naive CD4 þ T cells were isolated from WT and OX40 À / À mice and differentiated into Th9 cells in the presence of BMDCs or CurDCs. CurDCs readily converted WT naive CD4 þ T cells into Th9 cells (Fig. 4e,f and Supplementary Fig. 7C,D). However, the   capability of CurDCs in converting OX40 À / À naive T cells into Th9 cells was largely abrogated as demonstrated by significantly lower IL-9 and Irf4 expression by OX40 À / À Th9 cells primed by CurDCs (Fig. 4e,f and Supplementary Fig. 7C,D), indicating that OX40L contributed to Th9 cell development primed by dectin-1activated DCs. Notably, CurDC-induced OX40 À / À Th9 cells also expressed less Il13 than CurDC-induced WT Th9 cells ( Supplementary Fig. 7C,D). To further explore the contribution of OX40L to dectin-1-activated DC-induced Th9 differentiation, an OX40L blocking antibody (aOX40L) was used. We found that the addition of aOX40L significantly diminished IL-9 expression by Th9 cells stimulated with CurDCs ( Fig. 4g, Dectin-1 induces TNFSF15 and OX40L via Syk and Raf1. We next examined the downstream signalling pathways of dectin-1 that are responsible for TNFSF15 and OX40L expression in DCs. Given that dectin-1 is known to regulate the transcription factor NF-kB via the activation of Syk and/or Raf1-dependent signalling 24 , we examined whether these signalling pathways mediated dectin-1-induced TNFSF15 and OX40L expression. We first compared the effects of Curdlan versus TNF-a/IL-1b on the activation of Syk and Raf1 signalling pathways in mouse iDCs. Curdlan treatment induced rapid increases of phosphorylated (p)-Syk and p-Raf1 in DCs as compared with the treatment with TNF-a/IL-1b ( Fig. 5a), indicating that Curdlan was more potent than TNF-a/IL-1b in activating Syk and Raf1 in DCs. To further confirm the involvement of dectin-1 in Syk and Raf1 activation, iDCs generated from WT or dectin-1 À / À mice were used. Dectin-1 deficiency remarkably reduced the expression of p-Syk and p-Raf1 in iDCs treated with Curdlan, whereas it did not affect the phosphorylation of Syk and Raf1 in iDCs treated with TNF-a/IL-1b ( Fig. 5b), indicating that Curdlan induced dectin-1-dependent activation of Syk and Raf1 in DCs.
To investigate the role of Syk and Raf1 signalling in dectin-1induced TNFSF15 and OX40L expression, Syk-specific inhibitor Piceatannol and Raf1-specific inhibitor GW5074 were used during DC maturation. The inhibition of Syk or Raf1 in DCs significantly decreased TNFSF15 and OX40L expression induced by Curdlan (Fig. 5c,d). In addition, by using small interfering RNAs (siRNAs) to specifically silence Syk or Raf1 in mouse DCs (Fig. 5e), we found that knockdown of either Syk or Raf1 in DCs reduced TNFSF15 and OX40L expression induced by Curdlan ( Fig. 5f,g). To examine the effects of Syk and Raf1 signalling in directing dectin-1-activated DCs for Th cell differentiation, Piceatannol-and GW5074-treated CurDCs were used to prime Th cells. Piceatannol-or GW5074-treated CurDCs were less effective in the induction of Th9, Th1 and Th17 cells than CurDCs, as demonstrated by the decreased expression of Il9, Ifrg and Il17a by Th cells primed by Piceatannol-or GW5074-treated CurDCs compared with control CurDCs, (Supplementary Fig. 9). Together, these results demonstrated that dectin-1 induced TNFSF15 and OX40L expression through Syk and Raf1 signalling pathways.  Relative expression

Relative expression
Tnfsf15 Ox40l Tnfsf15 Ox40l  To compare the effects of Curdlan versus TNF-a/IL-1b on the activation of NF-kB pathway, mouse iDCs were treated with TNF-a/IL-1b or Curdlan. Curdlan treatment increased the expression of p-IKKa/b and decreased that of IkB-a in mouse DCs (Fig. 6a), leading to a remarkable increase of c-Rel, p50 and RelB nuclear translocation and a slight increase of p65 and p52 nuclear translocation as compared with the treatment with TNF-a/IL-1b (Fig. 6b), indicating that Curdlan was more potent in activating NF-kB pathway in DCs than TNF-a/IL-1b. Bortezomib, a proteasome inhibitor, was used to inhibit NF-kB 14 . We found that mouse DCs matured by Curdlan plus bortezomib expressed less TNFSF15 and OX40L than those matured by Curdlan alone (Fig. 6c,d), indicating that dectin-1 induced TNFSF15 and OX40L expression through NF-kB signalling pathway.
To further determine the role of NF-kB pathway in dectin-1induced TNFSF15 and OX40L expression, we performed luciferase reporter assays to examine whether these NF-kB molecules could bind directly to Tnfsf15 and Ox40l promoters and affect their expression. We found that p50-RelB and p52-RelB dimmers could bind to and activate Tnfsf15 promoter (Fig. 6e); while p50, p50-RelB and p52-RelB could bind to and activate Ox40l promoter (Fig. 6f). Considering that p52 nuclear translocation was only slightly increased by Curdlan treatment, we speculated that p50-RelB was responsible for dectin-1-induced TNFSF15 expression, while p50 and p50-RelB were responsible for dectin-1-induced OX40L expression. Collectively, these results demonstrated that dectin-1-induced TNFSF15 and OX40L expression was dependent on NF-kB signalling pathway.
Notably, a very recent report by HyeMee et al. showed that the dectin-1 agonist Curdlan downregulated OX40L expression on myeloid DCs 29 , which is exactly the opposite of our finding that dectin-1 upregulated OX40L expression on bone marrow-derived myeloid DCs. The reasons for this discrepancy are unclear. We suggest that the different concentrations of Curdlan used for DC activation in these two studies (5 mg ml À 1 in our study versus 10 mg ml À 1 in HyeMee's study) may contribute to the difference of OX40L expression in myeloid DCs ( Supplementary Fig. 11).
The observed important role of dectin-1-activated DCs in Th9 cell differentiation in vitro and the antitumour immunity in vivo prompted us to determine whether Th9 cells and IL-9 were involved in mediating the antitumour effects induced by dectin-1-activated DCs in vivo. To determine whether the immunization of dectin-1-activated DCs induces Th9 cells and IL-9 production, OT-II mice were immunized with OVA peptide-pulsed BMDCs or CurDCs. On day 3 after DC immunization, mice were killed, and mouse sera and total leukocytes from spleens and lymph nodes (LNs) were collected for the assessment of Th9 cell response. While no increase of serum IL-9 was observed in mice immunized with BMDCs as compared with mice treated with PBS (Fig. 7c), mice immunized with CurDCs showed significantly higher levels of serum IL-9 than mice receiving either BMDCs or PBS (Fig. 7c), indicating that dectin-1-activated DC immunization induced IL-9 production. Spleen and LN cells were restimulated with OT-II OVA peptides for 24 h before assay. Interestingly, intracellular staining detected significantly higher percentages of IL-9 þ CD4 þ (Th9) cells in the spleen and LN cells from mice immunized with CurDCs than mice receiving BMDCs or PBS (Fig. 7d,e), though there was a modest increase of Th9 cells in the spleen and LN cells from mice immunized with BMDCs as compared with PBS control mice (Fig. 7d,e). qPCR and ELISA further confirmed the increase of IL-9 production in mice immunized with CurDCs compared with mice receiving BMDCs or PBS ( Fig. 7f and Supplementary Fig. 12). These results demonstrated that dectin-1-activated DCs stimulated the production of Th9 cells and IL-9 in vivo. Dectin-1-activated DCs reportedly promoted the development of Th17 and Th1 cells 27 . In line with the published observations, we found that dectin-1-activated DC immunization upregulated the expression of Ifng and Il17a by spleen CD4 þ cells as compared with the treatment with BMDCs or PBS ( Fig. 7f and Supplementary Fig. 12).
To further confirm the role of dectin-1 signalling in activating DCs for the induction of Th9 cells and IL-9 production in vivo, we immunized OT-II mice with OVA peptide-pulsed WT CurDCs or dectin-1 À / À CurDCs. Mouse serum samples, spleen and LN cells were collected for the analysis of Th9 cell response. We found that dectin-1 À / À CurDC immunization failed to induce Th9 cell development in mice, since the serum IL-9, the number of IL-9 þ CD4 þ (Th9) cells and the expression of Il9 mRNA were completely abolished in mice immunized with dectin-1 À / À CurDCs as compared with those immunized with WT CurDCs (Fig. 7c-f).
Finally to define the functional relationship between Th9/IL-9 and dectin-1-activated DC-induced antitumour effects, OT-II mice with established B16-OVA tumours were treated with OVA peptide-pulsed CurDCs in the presence or absence of anti-IL-9. We found that blockade of IL-9 with anti-IL-9 abolished the antitumour effects induced by dectin-1-activated DCs (Fig. 7h), demonstrating the important role of Th9/IL-9 in mediating the antitumour effects induced by dectin-1-activated DCs in vivo. Collectively, these data demonstrated that dectin-1-activated DCs mediated their antitumour effects via their ability to induce Th9 cells and IL-9 production in vivo.
Dectin-1-NF-kB signalling induced DC production of a specific profile of cytokines, chemokines and costimulatory surface molecules, including TNFSF15 and OX40L, but little IL-12, which favoured Th9 cell induction in vitro and in vivo. MyD88-mediated signals also activate NF-kB pathway. However, we found that lipopolysaccharide (LPS), a potent stimulator of MyD88/NF-kB signalling, only slightly upregulated OX40L expression in LPS-activated DCs (LpsDCs) as compared with CurDCs ( Supplementary Fig. 13A). In addition, CurDCs but not LpsDCs could significantly upregulate IL-9 production and IRF4 expression in CD4 þ T cells under Th9-polarizing condition ( Supplementary Fig. 13B,C). Furthermore, CurDCs induced more potent inhibition on tumour growth than LpsDCs (Fig. 7h). One reason that might be contributed to this discrepancy is that MyD88-mediated signals potently stimulate DC expression of Th1-polarizing cytokine IL-12 (ref. 31). Taken together, these observations suggest that dectin-1-mediated signalling may be more efficient than MyD88-dependent signalling in promoting DC-induced antitumour Th9 cell differentiation in vitro and in vivo.

Discussion
Th9 cells have been shown to mediate potent antitumour effects in vivo 7,8 ; therefore, investigation of more efficient strategies to induce and expand Th9 cells in vivo may have high clinical significance in tumour immunotherapy. Dectin-1 signalling has been shown to induce Th17 and Th1 cell responses in vivo, which are essential for protecting a mammalian host from fungal infection 27,28,32 . However, whether dectin-1-activated DCs promote the development of antitumour Th9 cells remains unknown. In this study, we found that DCs activated by dectin-1 agonists potently promoted Th9 cell differentiation, and such Th9 cells expressed high levels of IL-9-and Th9-related transcription factor IRF4, but not Th1-, Th2-and Th17-related cytokines and transcription factors. This result was confirmed by using a dectin-1 blocking antibody or dectin-1 À / À DCs as they almost completely abrogated the capability of dectin-1 agonist-activated DCs in promoting Th9 cell induction. Importantly, in the functional tests, we found that dectin-1-activated DCs induced potently antitumour immunity against established tumours. Mechanistic studies revealed that immunization with dectin-1activated DCs efficiently promoted the production of Th9 cells and IL-9, which were required for the antitumour effects induced by dectin-1-activated DCs. Thus, our data established the role for dectin-1 in activating DCs for the induction of potent Th9 cell responses against cancers in vivo.
Previous studies showed that dectin-1 activation in DCs induces the expression of TNF-a, IL-6, IL-23 and IL-12, which are related to Th17 and Th1 cell differentiation 24,27,[32][33][34] . In this study, we identified other cytokines and costimulatory molecules, especially the TNF/receptor family members TNFSF15, OX40L, TNFSF8, TNFRSF26 and TNFRSF12a, that were upregulated by dectin-1 signalling and important for Th9 differentiation. Functional tests showed that dectin-1-induced overexpression of TNFSF15 and OX40L is responsible for the enhanced Th9 cell response induced by dectin-1-activated DCs. This result is On day 3 after tumour challenge, mice (n ¼ 10 per group) were given two weekly subcutaneously immunizations with 1 Â 10 6 OVA peptide-pulsed BMDCs or CurDCs generated from WT or dectin-1 À / À mice. Mice received PBS served as controls. Shown are the tumour growth curves. (b) Balb/c mice were injected subcutaneously with 1 Â 10 6 MPC-11 myeloma cells. On day 6 after tumour challenge, mice (n ¼ 10 per group) were given two weekly subcutaneously immunizations with 1 Â 10 6 KLH-Id-pulsed BMDCs or CurDCs. Mice received PBS served as controls. Shown are the tumour growth curves. (c-f) OT-II mice (n ¼ 4-5 per group) were injected subcutaneously with 1 Â 10 5 B16-OVA cells. On day 3 after tumour challenge, mice were immunized with OVA-peptide-pulsed BMDCs, CurDCs, dectin1 À / À BMDCs or dectin-1 À / À CurDCs. PBS served as control. On day 3 after the immunization, serum samples and total leukocytes from LNs and spleen cells were collected and restimulated with 5 mg ml À 1 OVA 323-339 peptides for 24 h. consistent with previous observations that ligation of OX40 inhibits the production of regulatory T cells and Th17 cells and diverts CD4 þ T cells to Th9 cells 11 . Interestingly, a recent study reported that TNFSF15 potently promotes Th9 differentiation and the pathogenicity of IL-9-dependent allergic diseases 35 . Thus, we identify TNF family members TNFSF15 and OX40L as key mediators for promoting Th9 cell differentiation primed by dectin-1-activated DCs. OX40L/OX40 were suggested to be unique in the TNF/receptor family in promoting Th9 cell induction 11 . In contrast, we and others found that TNFSF15 is also a powerful inducer of Th9 cells. Thus, other TNF family members may also be involved in the development of antitumour Th9 cells. Further studies will be necessary to investigate the function of other TNF/receptor family members expressed by dectin-1-activated DCs in the induction of Th9 cells and antitumour immunity. Engagement of dectin-1 by b-glucan was shown to activate the transcription factor NF-kB through Syk-and Raf1-dependent signalling pathways, leading to the production of IL-6, IL-1b, IL-23 et al. 24,33,34 . However, the dectin-1 downstream signalling pathways responsible for TNFSF15 and OX40L expression were not defined. In this study, we found that dectin-1 agonists potently activated Syk, Raf1 and NF-kB signalling pathways in DCs, more powerful than regular DC maturation reagents TNF-a and IL-1b. Thus, we speculated that Syk, Raf1 and NF-kB signalling pathways triggered by dectin-1 might be responsible for TNFSF15 and OX40L expression. Indeed, we found that blocking Syk, Raf1 and NF-kB signalling by their specific inhibitors or siRNAs inhibited the expression of TNFSF15 and OX40L in dectin-1-activated DCs. Furthermore, we found that p52-RelB and p50-RelB dimmers directly bound to and activated Tnfsf15 promoter, while p52-RelB, p50-RelB and p50 directly bound to and activated Ox40l promoter, providing direct evidence for NF-kB signalling in promoting TNFSF15 and OX40L expression. Thus, we identify Syk, Raf1 and NF-kB signalling pathways as key downstream pathways responsible for TNFSF15 and OX40L expression induced by dentin-1 in DCs. Nevertheless, we cannot exclude the possibility that some other factors that regulate TNFSF15 and OX40L expression might be induced through dectin-1-triggered Syk, Raf1 and NF-kB signalling pathways.
In this study, we found that dectin-1-activated DCs induced potent antitumour effects in vivo. However, cellular mechanisms underlying the antitumour effects were not determined. Dectin-1 agonists have been shown to induce Th17 and Th1 cell responses in vivo. In this study, we found that dectin-1-activated DCs induced potent Th9 cell responses in vivo. Th1 cells are a traditional antitumour Th subset 36 , while Th17 cells may also play a role in antitumour immunity 37 . Th9 cells were found to induce potent antitumour immunity in mouse models, better than Th1 and Th17 cells 7,8 . Based on these observations, we speculated that the antitumour effects of dectin-1-activated DCs in vivo mainly relied on Th9 cells and IL-9 induced by dectin-1-activated DCs. Indeed, we found that blockade of IL-9 with anti-IL-9 antibody abolished the antitumour effects induced by dectin-1-activated DCs. Thus, our data demonstrated that Th9 cells are the main effector cells in mediating the antitumour effects induced by dectin-1-activated DCs.
IL-9 expressed by dectin-1-activated DC-induced Th9 cells could have the potential to improve Th1/17/9 cell survival and proliferation in vivo and in vitro. In addition, curdlan may improve DC survival as compared with LPS, which exhibits detrimental effects on DC survival 38 . As our results showed that the improved efficacy of CurDCs to control the established tumour is dependent on IL-9, we believe that even if CurDC improved the T-cell activation, proliferation or survival or DC survival, Th9 induction by CurDCs played the most important role for the observed antitumour responses.
Neutralization of IL-9 was shown to promote melanoma tumour growth in C57BL/6 mice 7,8 . However, in this study, we observed no significant difference in B16-OVA tumour growth between mice received IL-9 neutralization antibodies and control IgG in either PBS-, BMDC-or LpsDC-treated groups. Reasons for this discrepancy could be that (i) we used OT-II mice in this study which largely lack CD8 þ T cells, the important antitumour effector cells and (ii) the beginning of administration of anti-IL-9 antibodies was on day 4 after tumour challenge in this study versus day 0 in previous studies.
In summary, our study demonstrates dectin-1-activated DCs as a powerful inducer of antitumour Th9 cells in vitro and in vivo and identifies TNFSF15 and OX40L as key factors in mediating Th9 cell differentiation primed by dectin-1-activated DCs. Syk, Raf1 and NF-kB signalling pathways triggered by dectin-1 were required for dectin-1-induced OX40L and TNFSF15 expression. Finally, dectin-1-activated DCs induce potent therapeutic antitumour effects in mouse models, and the antitumour effects depended on induced Th9/IL-9. Our findings may have important clinical implications.
In vitro Th9 cell differentiation. Naive CD4 þ T cells (CD4 þ CD25 À CD62L hi ) were obtained from mouse spleen cells by fluorescence activated cell sorter (FACS) as previously described 8,10 . Naive CD4 þ T cells (1 Â 10 5 per well) were cultured with BMDCs or dectin-1-activated DCs (1 Â 10 5 per well) in the presence of plate-bound anti-CD3 (2 mg ml À 1 ) plus TGF-b (3 ng ml À 1 ) and IL-4 (10 ng ml À 1 ). Th0 cells were generated without addition of TGF-b and IL-4 in the culture medium. In some experiment a blocking anti-OX40L mAb (5 mg ml À 1 ) or a neutralization anti-TNFSF15 mAb (5 mg ml À 1 ) were added. After 3 days of culture, cells were collected and analysed for gene expression at the mRNA or protein levels by flow cytometry, qPCR and ELISA.
To examine the Th9-polarizing function of TNFSF15, CD4 þ naive T cells (1 Â 10 5 per well) were cultured under Th0 or Th9 polarizing conditions in the presence of anti-CD3 and anti-CD28 (2 mg ml À 1 ) with or without addition of TNFSF15 (50 ng ml À 1 ) for 3 days. Cells and culture supernatants were collected and analysed for IL-9 expression by qPCR and ELISA.
To examine the function of OX40L-OX40 interaction in Th9 cell differentiation primed by dectin-1-activated DCs, CD4 þ naive T cells (1 Â 10 5 per well) were isolated from WT or OX40 À / À C57BL/6 mice and cocultured with BMDC or CurDC (1 Â 10 5 per well) in the presence of plate-bound anti-CD3 with or without addition of Th9 polarizing cytokines TGF-b and IL-4 for 3 days. Cells and supernatants were analysed by qPCR and ELISA.
qPCR and western blot analyses. qPCR was performed as previously described 10 . Total RNA was extracted from cells by using an RNeasy Mini kit (Qiagen) according to the manufacturer's instructions. The mRNA levels of Dectin1, Il9, Ifng, Il4, Il5, Il13, Il17, Tnfsf15, Ox40l, Spi1, Irf4, Tbx21, Gata3 and Rorc by Th cells or DCs were analysed. Expression was normalized to the expression of the house-keeping gene Gapdh. Primer sets used for these analyses are listed in Supplementary Table 1.
Gene-expression profiling. Immature DCs were generated from Balb/c mice and matured with TNF-a/IL-1b or Curdlan for 48 h. mDCs were collected and stored in Trizol reagent (Invitrogen) at À 80°C. All samples were sent to OneArray (http://www.OneArray.com.cn/, Beijing, China) for transcription profiling via genome-wide microarrays, and the subsequent data analysis was also performed by OneArray.
RNA interference. RNA interference was performed as previously described 10 . In brief, day 6 iDCs were transfected with 50 nM siRNA with transfection reagent DF4 (Dharmacon) according to the manufacturer's protocol. Silencing was confirmed at the mRNA levels by qPCR. On day 8, iDCs were matured with TNF-a/IL-1b or Curdlan for 48 h. On day 10, mDCs were collected and analysed by qPCR and flow cytometry staining. siRNAs used are listed in Supplementary Table 2.
In vivo functional tests for dectin-1-activated DCs. iDCs generated from WT or dectin-1 À / À C57BL/6 mice were matured with TNF-a (10 ng ml À 1 ) plus IL-1b (10 ng ml À 1 ) or Curdlan (5 mg ml À 1 ) for 48 h, and pulsed with OT-II OVA peptide (OVA 323-339 ; 10 mg ml À 1 ) for additional 6 h. Treated DCs were collected for mouse immunization. In the therapeutic model, 1 Â 10 5 B16-OVA cells were injected subcutaneously into OT-II mice. On day 3 after tumour challenge, mice were randomly divided into groups and immunized subcutaneously with 1 Â 10 6 treated DCs. Mice were immunized twice (1 week apart). Mice injected with PBS served as controls. In some experiments, mice were given control IgG or neutralization anti-IL-9 antibody (aIL-9, 100 mg per mouse) every 3 days beginning 1 day after the first DC immunization. Tumour development was monitored over time. Tumour volume was calculated by the formula: 3.14 Â (mean diameter) 3 /6. Mice were killed when the tumour diameter reached to the range between 1.5 and 2 cm.
OT-II mice were challenged with B16-OVA tumour cells and immunized with OVA-peptide-pulsed DCs as described above. On day 3 after DC immunizations, serum samples were collected and total leukocytes from tumour-derived LNs and spleens were restimulated with 5 mg ml À 1 OVA323-339 peptide for 24 h. Cytokine expression was analysed by qPCR, ELISA and intracellular staining.
To test the capacity of dectin-1-activated DCs in priming therapeutic immunity against native tumour-specific antigens, a myeloma tumour model MPC-11 was used. MPC-11 cells secrete an Id protein (IgG2b) which is a native tumour-specific antigen. To enhance its immunogenicity, Id protein was conjugated to keyhole limpet hemocyanin (KLH, EMD Biosciences) as described previously 39 . iDCs generated from Balb/c mice were pulsed with Id-KLH (50 mg ml À 1 ) for 6 h and matured with TNF-a/IL-1b or Curdlan for additional 48 h. MPC-11 cells (1 Â 10 6 ) were injected subcutaneously into Balb/c mice. On day 6 after tumour challenge, mice were grouped and immunized subcutaneously with 1 Â 10 6 Id-KLH-pulsed DCs. Mice were immunized twice (1 week apart) and tumour development were monitored over time. On day 3 after the second immunization, spleen cells were collected and re-stimulated with Id-KLH-pulsed BMDCs or CurDCs with addition of control IgG or aIL-9 for 48 h.
Cytotoxicity assay. The cytotoxicity assay was performed as previously described 10 . In brief, Id protein secreted by MPC-11 cells was used as tumour antigen. Balb/c mice (n ¼ 3 per group) were immunized twice (1 week apart) with 1 Â 10 6 Id-KLH-pulsed BMDCs or CurDCs. Mice treated with PBS served as control. Mice were given control IgG or an IL-9 neutralizing antibody (aIL-9,100 mg per mouse) every 3 days beginning 1 day after the first DC immunization. On day 2 after the second immunization, spleen cells from the mice were re-stimulated with Id-KLHpulsed BMDCs for 48 h and used as effector cells. MPC-11 cells labelled with 5 mM CFSE were used as target cells, whereas MOPC-315 labelled with 0.5 mM CFSE were used as non-target cells.
Statistical analysis. The Student's t-test was used to compare various experimental groups. A P value of o0.05 was considered significant.
Data availability. The microarray data of BMDCs and CurDCs are stored in the GEO repository and is accessible under the accession number GSE81111. All data are available within the article (as figure source data or Supplementary Information Files) and/or from the authors on request.