Nrf2 induces IL-17D to mediate tumor and virus surveillance

Cells undergoing xenobiotic or oxidative stress activate the transcription factor Nrf2, which initiates an intrinsic “stress surveillance” pathway. We recently found that the cytokine IL-17D effects a form of extrinsic stress surveillance by inducing antitumor immunity, but how IL-17D is regulated remains unknown. Here, we show that Nrf2 induced IL-17D in cancer cell lines. Moreover, both Nrf2 and IL-17D were induced in primary tumors as well as during viral infection in vivo. Expression of IL-17D in tumors and virally infected cells is essential for optimal protection of the host as il17d −/− mice experienced a higher incidence of tumors and exacerbated viral infections compared to WT animals. Moreover, activating Nrf2 to induce IL-17D in established tumors led to natural killer cell-dependent tumor regression. These data demonstrate that Nrf2 can initiate both intrinsic and extrinsic stress surveillance pathways and highlight the use of Nrf2 agonists as immune therapies for cancer and infection. factor nuclear factor erythroid derived 2- like 2, or Nrf2, the interleukin-17D. Nrf2/IL-17D-mediated natural cell IL-17D using


INTRODUCTION
Cells undergoing malignant transformation or viral infection constitute cells in a "stressed state," characterized by altered metabolism and imbalanced reactive oxidative species (ROS) (Gorrini et al., 2013;Martindale and Holbrook, 2002;Schwarz, 1996). In order to deal with ROS, cells activate the transcription factor Nrf2 (Nuclear factor erythroid derived 2-like 2 or Nfe2l2). Nrf2 is a member of the cap 'n' collar family of bZip transcription factors and is recognized as the primary responder to cellular oxidative stress (Ma, 2013). Nrf2 induces genes involved primarily in antioxidant defense, oxidant signaling, and drug metabolism; and, secondarily, in metabolism, cell proliferation, and proteasome activity (Malhotra et al., 2010;Schafer et al., 2012;Turei et al., 2013).
The cell-protective pathways induced by Nrf2 can have opposing effects on cancer development. For example, Nrf2 protects somatic and premalignant cells from carcinogens, and in this context fulfills a tumor suppressor role Ma and He, 2012). On the other hand, it is well documented that Nrf2 can promote the growth and survival of established tumors by inducing anti-oxidative pathways that help cancers deal with chronic oxidative stress, a hallmark of cancer progression (Jaramillo and Zhang, 2013;Sporn and Liby, 2012). It is not known whether Nrf2 regulates extrinsic stress response mechanisms that operate to suppress carcinogenesis.
Recently, our group identified the cytokine IL-17D as a molecule expressed at higher levels in highly immunogenic tumor cells compared to poorly immunogenic tumor cells Saddawi-Konefka et al., 2014). When IL-17D was overexpressed in poorly immunogenic cancer cells, it led to immune rejection mediated by natural killer (NK) cells. The regulation of IL-17D is not known, but based on its tumor rejection activities, we hypothesize that IL-17D would initiate tumor surveillance and thus accompany early cellular transformation.
Here, we show direct and compelling evidence that Nrf2 induces the expression of IL-17D, therefore initiating antitumor immune responses. We also find that viral infection induces Nrf2 and IL-17D, presumably by causing local oxidative stress. The induction of IL-17D is required for effective cancer surveillance and antiviral responses, as mice deficient in IL-17D had increased formation of MCA-induced tumors and exacerbated pathology when infected with vaccinia virus (VV) or murine cytomegalovirus (MCMV). Our data document a link between cellular oxidative stress -resulting from viral infection or tumorigenesis -and the initiation of immunity. Moreover, our results define an immune activating role for the well-studied factor Nrf2, which has not been previously defined. This role for Nrf2 in inducing IL-17D has therapeutic benefit in our mouse models.

Nrf2 induces IL-17D
To explore IL-17D regulation, we performed a transcription factor binding site (TFBS) analysis of the promoter and intronic regions of the human and mouse il17d genes (Fig. 1A).
Our analysis revealed several putative Nrf2 binding sites, defined as anti-oxidant responsive elements (ARE) (Nguyen et al., 2003) (Fig. 1A, Table S1). Given the abundance of ARE in promoter and intronic regions of il17d, we hypothesized that the activation of Nrf2 would induce IL-17D. To test this, we treated murine embryonic fibroblasts (MEFs) and the 3methylcholanthrene (MCA)-induced sarcoma cell line F244 (O'Sullivan et al., 2012;Shankaran et al., 2001) with H 2 O 2 , a known activator of Nrf2 (Pergola et al., 2011;Tkachev et al., 2011) (Fig. 1B). H 2 O 2 treatment led to significant, time-responsive increases in the transcript of il17d. Similarly, activation of Nrf2 with tert-butylhydroquinone (tBHQ) resulted in the increase of il17d transcript in the murine melanoma cell line B16, the human Burkitt's lymphoma cell line Ramos and in the MCA-induced sarcoma cell line F244 (Fig. 1C, S1A).
Next, we determined whether the transcription factor Nrf2 directly binds to the TFBS we identified in our analysis of the il17d gene. We performed a ChIP-qPCR (chromatin immunoprecipitation followed by polymerase chain reaction amplification of specific sequences) in tBHQ-treated or control-treated B16 cell lines. Cells were fixed and sonicated before immunoprecipitation with Nrf2-specific antibody or control IgG. Fractionation and Western Blot analysis confirmed that Nrf2 preferentially accumulated in the nuclear fraction of treated cells (not shown). qPCR analysis of ChIP fractions revealed two sites upstream of the il17d start site where Nrf2 has significant binding following activation (Fig. 1D). These two binding sites for Nrf2 corresponded to Nrf2 target ARE elements identified at 4195, 4860 and 3730 bp upstream of the il17d start site (Fig. 1A, Table S1). qPCR analysis of the known gene target for Nrf2, Heme Oxygenase 1 (hmox1), also indicated Nrf2 binding following tBHQ treatment in the B16 cell line (Fig. S1B).
TFBS analysis with the ENCODE UCSC browser revealed that other transcription factors might bind and regulate IL-17D (data not shown), indicating that Nrf2 may not be wholly responsible for the induction of IL-17D. In order to examine the necessity of Nrf2 for the induction of IL-17D, we activated Nrf2 in the F244 or B16 cell lines in the presence of siRNA specific to nrf2 (Fig. 1E, Fig. S1C, D) and in F244 and B16 cell lines bearing a stable knockdown of nrf2 via shRNA ( Fig. S1E-J). Knockdown of Nrf2 in B16 and F244 (~80%, Fig. S1C-F) was sufficient to block the induction of il17d following activation of Nrf2 with either H 2 O 2 or tBHQ. Altogether, we found that Nrf2 not only directly bound to the il17d promoter region but also was required for efficient induction of il17d by oxidative stress.

Nrf2 and IL-17D are co-expressed in primary tumors and during viral infection
To determine the relevance of the Nrf2 regulation of IL-17D in vivo, we examined the expression of IL-17D, Nrf2 and its known target genes in primary human and mouse tumors. Analyzing gene expression in primary MCA-induced tumors (from Fig. 4A) revealed that nrf2, its target hmox1 and il17d were upregulated compared to normal untreated skin samples ( Fig. 2A). Using data sourced from The Cancer Genome Atlas (TCGA), we found that il17d expression directly correlated with the expression of ARE-containing Nrf2 targets (signature of nine genes in total, see methods) across all available human cancers (n=9755) (Fig. 2B). The results are not significant (p=0.07), likely due to the fact that TCGA data includes many tumors harvested at late timepoints, when we hypothesize il17d and nrf2 expression to be uncoupled due to editing of IL-17D . Moreover, infiltrated immune cells that have a different gene expression profile can influence the results (Aran et al., 2015). We also found that a high level of IL-17D expression in 13 out of 31 human cancer types confers a survival advantage (Table S2), representatively shown for brain lower grade glioma and ovarian serous cystadenocarcinoma (Fig. S2A). Additionally, an analysis of our MCA-sarcoma tumor cell lines demonstrated that Nrf2 and il17d are coexpressed in murine tumor cell lines (Fig. 2C). Matching our previous data Saddawi-Konefka et al., 2014), cell lines expressing high levels of IL-17D tended to behave as regressors, now underlined by their co-expression of Nrf2. Together, these data suggest that Nrf2 regulates IL-17D during primary tumor formation in both human and mouse systems in order to initiate productive antitumor immune responses leading to tumor regression and prolonged survival. IL-17D expression only correlates with better survival in a fraction of human cancers ( Fig S2, Table S2), suggesting that its regulation might be context-dependent and underlining the importance of analyzing its regulation in defined in vivo mouse models. Viral infections represent another sort of cellular stress. Since IL-17D recruits NK cells that can mediate antiviral responses, we sought to examine the role of the Nrf2-IL-17D axis in antiviral immunity. First, we measured Nrf2 and IL-17D following vaccinia virus (VV) and murine cytomegalovirus (MCMV) infection. In vitro, we observed an increase in the transcript levels of il17d in both infected primary derived fibroblasts and tumor cell lines ( Fig. 3A, B). The results in the tumor cell lines are not significant, likely due to the fact that tumors are not the primary target for viruses. For in vivo analysis, we scarified WT mice with VV. Tissue harvested from these mice had increased expression of il17d and Nrf2 (Fig.   3C, D; example of VV scar Fig. S3A, B). To model the local activation of Nrf2 and IL-17D in vivo, we adapted a system in which we topically applied tBHQ onto the dorsal flanks of mice (Schafer et al., 2014). Mice treated topically with tBHQ had increased transcript and protein levels of il17d and Nrf2, respectively, commensurate with those observed following infection by scarification (Fig. 3E, F). Together, these results suggest that the Nrf2-IL-17D regulatory axis is activated during primary tumorigenesis and viral infection in order to confer protection from disease progression.

IL-17D protects the host from primary tumorigenesis and viral infection
We previously showed that IL-17D can mediate tumor rejection when overexpressed in cancer cells , but its endogenous role in cancer immunosurveillance had not been demonstrated. Therefore, we compared the development of primary tumors in WT versus il17d −/− mice, each treated with the carcinogen 3-methylcholanthrene (MCA). Since the immune cellularity of il17d −/− mice has not been studied, we first immunephenotyped these mice. We found that the major immune populations in the spleen, lymph node, bone marrow and blood were similar between WT and il17d −/− mice ( Fig. S4 and data not shown). Despite demonstrating a largely normal immune system at baseline, strikingly, il17d −/− mice were significantly more susceptible to the development of primary tumors ( Fig. 4A). At a 25μg dose of MCA, approximately twice the number of il17d −/− mice compared to WT mice developed primary tumors. At a 5μg dose of MCA, WT mice are largely tumor free, whereas approximately 40% of il17d −/− mice developed primary tumors.
These findings confirm that, similar to Nrf2, the cytokine IL-17D can protect the host from primary carcinogen-induced tumor formation.
Since we found Nrf2 and IL-17D to be induced after viral infection, we next tested the role of IL-17D in viral infection. Following infection of WT and il17d −/− animals, we observed an exacerbation of the VV scar in il17d −/− animals compared to WT animals at two doses ( Fig. 4B). Similarly, il17d −/− mice were more susceptible to another virus, MCMV, displayed as an increase in weight loss after systemic infection (Fig. 4C). To test whether IL-17D can directly inhibit viral replication or progression, we infected either parent or IL-17D-expressing tumor cell lines with VV ( Fig. S3C). Expression of IL-17D in vitro did not protect cells from infection, suggesting that IL-17D's role in protecting the host from viral infection may require the immune system. These findings imply that virus infection may activate the Nrf2-IL-17D axis to initiate surveillance and contribute to host antiviral defense.

Activating tumor-intrinsic Nrf2 delays tumor growth in vivo via induction of tumor-derived IL-17D
Having shown a requirement for IL-17D in effective tumor surveillance Saddawi-Konefka et al., 2014) (Fig. 4A), we hypothesized that activating Nrf2 in vivo would induce IL-17D in established tumors, which in turn would initiate protective immunosurveillance. To examine this, tumor bearing mice were treated topically with tBHQ or lanolin control cream beginning when tumors achieved an average diameter of 3mm. We found that treatment with tBHQ delayed tumor growth in WT mice bearing F244 or B16 tumors (Fig. 5A, B). Moreover, tBHQ led to the in vivo induction of both hmox1 and il17d in tumors derived from both transplanted cell lines (Fig. 5C). Recognizing that tBHQ could induce Nrf2 in both host and tumor cells and that Nrf2 might induce targets other than IL-17D, we wanted to examine: first, whether tBHQ is activating Nrf2 and IL-17D in tumor cells versus host cells; second whether tBHQ necessarily and specifically induces IL-17D to mediate tumor regression; and third, whether Nrf2 is required to induce IL-17D and delay tumor growth. Therefore, we transplanted nrf2 −/− and il17d −/− mice with B16 and treated with tBHQ. We observed delayed tumor growth of B16 in nrf2 −/− and il17d −/− mice (Fig. 5D, E), demonstrating that host expression of nrf2 or il17d is not required for the response to tBHQ.
To examine the role of tumor-derived IL-17D during tBHQ-induced tumor rejection, we utilized an MCA sarcoma cell line that we generated from an il17d −/− host (named F38K1; transplanted with F38K1 failed to delay tumor growth. This finding implies that the antitumor response resulting from the activation of Nrf2, via topical application of tBHQ, requires tumor-expressed IL-17D and not the many previously described targets of Nrf2 (Malhotra et al., 2010;Turei et al., 2013).
To analyze the requirement of tumor-derived Nrf2 during tBHQ-induced tumor rejection, we generated independent sarcoma and melanoma cell lines bearing a stable knockdown of nrf2 via shRNA (Fig. S1 E-J). We chose the two shRNA constructs that showed the best downregulation of nrf2 (90% and 83% for F244 sarcoma; 88% and 80% for B16 melanoma) for further experiments. tBHQ treatment had no influence on in vitro cell growth in any of the transduced cell lines (Fig. S5D, E and not shown). When transplanted into WT mice, tBHQ treatment failed to delay tumor growth of the two nrf2 knockdown cell lines (Fig. 5G, H and data not shown), in contrast to the shRNA control cell lines (Fig. S5F). Moreover, the tBHQ-mediated induction of il17d (and hmox1) was abolished in vivo when nrf2 was knocked down (Fig. 5I). These results show that tumor-derived Nrf2 is required for both the tBHQ-dependent induction of il17d as well as the rejection of established tumors. It should be noted that silencing nrf2 in these progressor tumors actually caused them to display growth delay compared to the parental cells (Fig. 5G, H and not shown). These results indicate that Nrf2 may have tumor promoting activity and thus, the tumor rejection by tBHQ-induction of IL-17D must be stronger than the tumor promoting activity of Nrf2.

Activating Nrf2 mediates tumor rejection via recruitment of NK cells
To determine whether the immune system was required for the antitumor effect of Nrf2 agonists, we transplanted B16 melanoma cells into immune-deficient mice: rag2 −/− , which lack adaptive immunity but possess intact NK cells and macrophages (Shinkai et al., 1992) and rag2 −/− x γc −/− , which lack adaptive immunity as well as NK cells (Mazurier et al., 1999). Notably, we found that topical treatment of tumors with tBHQ delayed tumor growth in rag2 −/− but not rag2 −/− x γc −/− (Fig. 6A, B). This suggests that NK cells are the immune population responsible for tBHQ-mediated tumor rejection. To identify the immune cells recruited by tBHQ treatment, we harvested B16 tumors after 7 days of treatment (representative image shown in Fig. 5A) and performed a FACS analysis of tumor infiltrating leukocytes (TILs). Our TIL analysis revealed an increase in the percentage and total number of infiltrating NK cells when tumors were treated with tBHQ ( Fig. 6C-E), a result consistent with our prior study showing that IL-17D recruited NK cells via induction of the chemokine CCL2 Saddawi-Konefka et al., 2014). No other TIL populations were found to be different in treated versus untreated tumors (Fig. 6C). We next investigated the requirement of Nrf2 and IL-17D for the recruitment of NK cells. When nrf2 was knocked down or il17d was deleted within the tumor cell (by transplanting the cell line F38K1), the tBHQ-mediated increase in NK cells was abolished (Fig. 6F). NK cells in tumors did not differ in activation or function (as assessed by expression of CD69, IFNγ and Granzyme B) (Fig. S5G). Together, our results show that NK cells are recruited into tBHQtreated tumors and that tumor-expressed Nrf2 and IL-17D are required for the tBHQmediated recruitment of NK cells. In order to assess if il17d expression correlated with NK cell infiltration in human cancer, we used the Z-scores from RNAseq of four NK-cellexpressed genes from TCGA to approximate the presence of NK cells in human skin cutaneous melanoma and sarcoma ( Fig S6A). We found no positive correlation, probably due to the fact that TCGA data represents chronic rather than acute induction of IL-17D as assessed in our mouse model. The same held true for the correlation of il17d with NK-and macrophage-recruiting gene products (CCL5 and CCL2, Fig. S6B, C) as well as NK cellexpressed genes NKG2D and NKP46 (Fig. S6D, E). This does not exclude a role for Nrf2 and IL-17D in NK cell recruitment in human cancers, but rather calls for a more detailed analysis of human cancer biopsies after acute induction of IL-17D. To date, this is not possible but might be done in the future if tBHQ or another Nrf2 agonist reaches clinical trials.

DISCUSSION
In this study, we have demonstrated an obligate role for IL-17D in effective tumor surveillance, optimal antiviral responses, and cancer immune therapy via Nrf2 agonists. It is well established that immune responses to viruses and transformed cells have overlapping features (Raulet and Guerra, 2009): both involve NK cells, Th1 immunity, and CD8 + T cells. Moreover, NKG2D ligands are induced by viral infection (Shafi et al., 2011;Vivier et al., 2011) as well as being constitutively expressed on cancer cells (Diefenbach et al., 2001;Guerra et al., 2008). As such, our finding that IL-17D is induced by viral infection and expressed constitutively by immunogenic cancer cells has precedence in principle and data.
We have definitively shown that Nrf2, an oxidative stress response factor, can function as a tumor suppressor via direct induction of IL-17D. A tumor suppressor role for Nrf2 is supported by previous studies showing that mice genetically deficient in Nrf2 are more susceptible to a wide range of carcinogen-induced cancers (see review (Ma and He, 2012)). For example, nrf2 −/− mice displayed increased incidence of forestomach cancer (Ramos-Gomez et al., 2003) and bladder cancer (Iida et al., 2004) induced by carcinogens known to induce oxidative stress. Nrf2 −/− mice also had increased skin cancer in a model of sulforaphane-mediated protection from DMBA/TPA induced carcinogenesis (Xu et al., 2006). To our knowledge, there are no studies to address whether Nrf2 participates in tumor immunosurveillance in any of these model systems. Importantly, cancer immunoediting and tumor elimination have been extensively documented in mouse models of MCA-induced sarcomas, and it was recently shown that MCA can acutely induce Nrf2 and its target genes in liver after 24 hours (Jin et al., 2013). Accordingly, we show that Nrf2, its target genes and IL-17D are induced in MCA tumors. We suggest that Nrf2 can mediate anti-cancer functions by inducing immune-dependent pathways, namely the IL-17D-dependent recruitment of NK cells.
In contrast to the tumor suppressor role of Nrf2, other studies have shown that Nrf2 expression in tumor cells can promote their survival in the face of oxidative stress, hypoxia, and/or chemotherapy (Jaramillo and Zhang, 2013;Sporn and Liby, 2012). In fact, Nrf2 blockade has become an anti-cancer approach since certain cancer cells (and model systems) report oncogene-induced, constitutive Nrf2 activity as associated with tumor growth and metastasis (DeNicola et al., 2011;Shelton and Jaiswal, 2013). These studies have prompted a re-evaluation of the role of Nrf2 in cancer and support a model whereby Nrf2 is a "doubleedged sword" that can suppress or promote cancer. Notably, a recent study found that Nrf2 acts early in tumorigenesis to suppress tumor formation and later-on to promote tumor formation (Satoh et al., 2013). We propose that Nrf2-mediated induction of IL-17D activates antitumor immunity at an early stage to eliminate the tumor before Nrf2 exerts its pro-tumor activity. This early immune pressure mediated by NK cells could lead to cancer immunoediting, resulting in the loss of IL-17D expression as an immune evasion mechanism. Therefore, the expression of IL-17D in late stage human cancers may not always correlate with good prognosis or NK cell infiltration, as shown in Table S2 and Figure S2B, C and S6. In fact, given the known tumor promoting roles of Nrf2, late stage human cancers that display chronic inflammation may express high levels of Nrf2 without incurring the antitumor responses that could be mediated by IL-17D and/or NK cells. Thus, it will be important to determine if acute induction of Nrf2 in human cancer can indeed serve as a therapeutic mechanism to induce IL-17D and/or recruit NK cells to mediate antitumor immune responses.
We have implicated that Nrf2 agonists, some of which are currently in clinical trials, would be highly efficacious inducers of cancer immunosurveillance and immune therapy by acutely inducing IL-17D. Further studies are needed to determine whether Nrf2 agonists can be used to treat a broad range of established human tumors and/or prevent the development of cancer. It will be important to cast a broad and deep net in these studies, as it is likely that the role of the Nrf2-IL-17D pathway in tumor progression is context-dependent. Notably, our data suggest that the clinical use of antioxidants to prevent cancer and promote overall health may inadvertently limit Nrf2 induction and attenuate an endogenous tumor surveillance pathway. On the other hand, judicious use of oxidative species may find a niche in immunotherapy. For example, drugs such as tBHQ, which do not induce ROS but can directly induce the Nrf2-IL-17D pathway may have even higher efficacy as they would activate an endogenous tumor surveillance pathway without producing genotoxic ROS.
The role of other IL-17 family members in cancer is still controversial. For the most studied member, IL-17A, tumor-promoting roles (Charles et al., 2009;Nam et al., 2008;Numasaki et al., 2003), antitumor functions (Benatar et al., 2008;Benchetrit et al., 2002;Muranski et al., 2008), and immune cell recruiting roles (Martin-Orozco et al., 2009) have all been documented. Specifically, an antitumor role in humans was shown whereby IL-17A production correlated with CD8 + T cell and CD57 + NK cell presence in esophageal squamous cell carcinomas (Lv et al., 2011) and was associated with the induction of T cell-, NK cell-and DC-attracting chemokines (Lu et al., 2013). On the other hand, a pro-tumor role for IL-17A was shown in breast cancer whereby neutralizing IL-17A with antibodies reduced chemokine expression, and thereby breast cancer cell migration and metastasis (Roy et al., 2014). Although we have only defined antitumor activities for IL-17D, the fact that Nrf2 can induce IL-17D might suggest that IL-17D could also have tumor promoting activities, similar to IL-17A. This could certainly limit therapies based on IL-17D.
We also found that IL-17D is induced by viral infection and is required for optimal antiviral responses. The IL-17 family of cytokines has been characterized as essential to antimicrobial host defense (see reviews (Gaffen, 2011;Gu et al., 2013;Iwakura et al., 2011;Jin and Dong, 2013)). Specifically, IL-17C and IL-17A/F are thought to mediate anti-bacterial and antifungal responses via recruitment of neutrophils. IL-17E contributes to anti-helminth responses via recruitment of eosinophils. Our finding that IL-17D is induced during viral infection, recruits NK cells, and is required for optimal responses to VV and MCMV infection, suggests that the IL-17 family may have evolved to mediate distinct and specific anti-pathogen responses. Given the ancient origin of the IL-17 family, it is tempting to speculate that IL-17D and IL-17C were the first family members to evolve to mediate local control of pathogen infection prior to the evolution of adaptive immunity.
To date, the signals that induce and maintain Nrf2 activity during carcinogenesis are still unclear. We speculate that during tumor formation, Nrf2 activation occurs acutely due to carcinogen exposure, is maintained subacutely by oncogenes, and can be detected constitutively in advanced cancers by mutation or inflammatory cells providing an oxidative burst. Indeed, previous studies have found that MCA induced Nrf2 and its target genes acutely (Jin et al., 2014), in line with our finding that MCA induced Nrf2 and IL-17D in our tumor model. Moreover, oncogenic alleles have been shown to induce Nrf2 (DeNicola et al., 2011), and mutations in KEAP1 can be detected in cancer cells, leading to constitutive Nrf2 activation (Padmanabhan et al., 2006;Shibata et al., 2008;Sjoblom et al., 2006). Future studies will clarify the kinetics of Nrf2 and IL-17D induction during tumor formation.
Finally, it is clear that cancers display metabolic and oxidative stress while also demonstrating an inflammatory component. In fact, these characteristics of cancer have been touted as "next generation hallmarks" (Hanahan and Weinberg, 2011). Based on our findings, it is tempting to conclude that the Nrf2/IL-17D pathway represents an important molecular bridge that connects two hallmarks of cancer -inflammation and oxidative stress.

METHODS AND MATERIALS Transcription Factor Binding Analysis
Sequences for mouse and human il17d genes were analyzed for the presence of antioxidant responsive elements (ARE) [TGAcTCAGCa], a sequence to which the Nrf2-sMAF heterodimer is known to bind (Nguyen et al., 2003). ARE sequences identified in mouse and human il17d genes are listed in Table S1.

Cell Lines
MCA-induced sarcoma cell lines were generated from primary tumors and expanded in vitro until at least the second passage before freezing. For experimentation, tumor cell lines were thawed from early passages and grown in RPMI 1640 (Gibco) supplemented with 10% FCS (Atlanta Biologics) (as previously described (O'Sullivan et al., 2012). Other tumor cell lines used -Ramos, B16, LLC -were cultured similarly. Primary derived mouse embryonic fibroblasts were derived from fetuses 12.5-13.5 days p.c. (as described (Conner, 2001a, b).

Nrf2 Activation and Knockdown
Nrf2 was activated in cell lines in vitro with either tert-butylhydroquinone (Spectrum) or H 2 O 2 (Fisher). tBHQ was used at 50μM in DMSO, and H 2 O 2 was used at 10μM for 0.5-1 hours before being washed out. Treated cells were harvested at timepoints between 6 and 12 hours for analysis. Activation of Nrf2 in vivo was adapted from methods described by (Schafer et al., 2014). For in vivo Nrf2 activation, a cream containing 50mM tBHQ solubilized in DMSO was mixed 1:1 (v/v) with Lanolin cream (Sigma) (controls were a 1:1 mixture of DMSO:Lanolin) and heated gently in a water bath to allow for mixing. The mixture was allowed to cool overnight before use. Hair along the flank was removed one day before the initiation of topical treatments. To knockdown Nrf2, a mixture of three siRNAs to nrf2 or control siRNA were used (Invitrogen) (Fujita et al., 2011). siRNA was transfected into cells with Lipofectamine 2000 (Thermo Fisher) as recommended by the manufacturer. For shRNA knockdown, five different shRNAs to nrf2 or control shRNA (Sigma) were cotransfected with lentiviral plasmids into 293T cells using Lipofectamine 2000. Viruscontaining supernatant was used to transduce sarcoma and melanoma tumor cell lines. Cells were grown in the presence of 10-40 μg/ml puromycin, and nrf2 knockdown was confirmed with qPCR.

Tumorigenesis and Tumor Transplantations
Primary tumorigenesis was performed as previously described (O'Sullivan et al., 2012;Shankaran et al., 2001). MCA was dissolved in corn oil (Sigma) prior to instillation. To induce primary tumors, 5μg or 25μg doses of MCA were injected subcutaneously along a single flank. Tumor development was monitored and measured weekly between two and six months post MCA instillation. Tumors were harvested for cell line generation and RNA when tumors achieved an average diameter of 25mm. For transplantation studies, tumor cell lines were trypsinized, washed with cold PBS three times, and injected subcutaneously along the flanks of mice (previously described (O'Sullivan et al., 2012)). Hair was removed from the flanks of mice at least one day prior to transplantation. Tumor progression was assessed by averaging the greatest two diameter measurements of the tumor.

Viral Infections
VV Western Reserve was kindly donated by the Dr. Ananda Goldrath (UCSD) and MCMV Smith Strain by Dr. Elina Zuniga (UCSD). Primary fibroblasts or tumor cell lines were incubated with 1×10 5 pfu of MCMV or VV, respectively, per 4×10 5 cells for 2h, washed and harvested after 24h. Viral titers following VV infection were determined by plaque assays on Vero cells (kindly donated by Dr. Elina Zuniga). In vivo, age and sex-matched eight to twelve week old C57Bl/6 WT or il17d −/− mice were infected with VV by scarification at 1×10 5 or 1×10 6 pfu. VV scars were monitored daily and expressed as the average of the two maximum scar diameters. For MCMV, mice were infected with 3×10 5 pfu/mouse i.p. and weighed daily for disease progression.

Antibodies and FACS Analysis
Tumor tissues were digested both mechanically by chopping with razor blades and chemically with 1mg/mL type IA collagenase (Sigma-Aldrich) for 30 minutes at 37°C. Following digestion, cell suspensions were washed, filtered and stained as previously described (O'Sullivan et al., 2012

Western Blot
Cells were lysed in 4x sample buffer containing SDS (Bio-Rad) and β-mercaptoethanol (Sigma) before boiling at 95°C for 5 min. Samples were run on SDS-PAGE gels (Biorad), and expression of Nrf2 was analyzed by Western Blotting using anti-Nrf2 (C-20, Santa Cruz Biotech). β-actin (A5441, Sigma-Aldrich) was used as loading control. Blotted bands were quantified with CS6 Photoshop imaging software.

Immunohistochemistry
Tissues were fixed in 10% Formalin (Sigma-Aldrich) for 24 hours before embedding in paraffin. Sections were stained with α-Nrf2 (C-20) by the UCSD Core Histology and Immunohistochemistry service using the Ventana Discovery Ultra (Roche). Slides were imaged on a Leica DM 2500 microscope and photographed with a Leica DFC 420 digital camera.

Chromatin Immunoprecipitation
ChIP for Nrf2 was performed as described previously (DeNicola et al., 2011). Following activation of Nrf2 with tBHQ, cells were fixed in 1% formaldehyde for 10 minutes at room temperature, quenched with 0.125M glycine for 5 minutes at room temperature, washed with cold PBS and resuspended in lysis buffer [1% SDS, 10mM EDTA pH 8, 50mM Tris-HCL pH 8], fresh protease inhibitor cocktail for 5 minutes on ice. To generate chromatin fragments of around 200bp, cell lysates were sonicated on ice for 15 cycles [15 seconds on, 45 seconds off] and pelleted by centrifugation at 13,000 rpm for 5 minutes. Protein A dynabeads (Life Technologies), pre-blocked in 0.5% BSA in PBS (w/v), were incubated with 8μg α-Nrf2 (C-20) or normal rabbit IgG (sc-2027, Santa Cruz Biotech) overnight at 4°C and then washed with additional blocking buffer and RIPA. To immunoprecipitate Nrf2chromatin complexes, conjugated dynabeads were mixed with sonicated lysate -diluted 1:9 in dilution buffer with protease inhibitors -and allowed to rotate overnight at 4°C. Immunoprecipitates were processed as suggested by the dynabeads manufacturer (Life Technologies) and purified using the QiaQuick PCR DNA kit (Qiagen). qPCR sequences used for ChIP samples appear in Table S3.

•
The transcription factor Nrf2 induces the cytokine IL-17D • IL-17D is required for effective antitumor and antiviral immune responses  Experiments repeated at least twice. Error bars represent ± SEM. Supported by Fig. S1 and Tables S1 and S3.

Fig. 5. Activating Nrf2 induces IL-17D and delays tumor growth in vivo
Tumor cells were transplanted subcutaneously and allowed to reach an established size (~3×3mm) before the initiation of topical treatments with tBHQ once daily for seven days. (C) -(E) Topical treatment of B16 melanomas with tBHQ increases the percentage (C-E) and total number (E) of NK cells present in tumors, while other immune populations remain unchanged (C).
(F) The tBHQ-induced increase in NK cell recruitment is prevented when nrf2 is knocked down or il17d is deleted (F38K1) in tumors.
Experiments repeated at least twice. Error bars represent ± SEM. Supported by Fig. S4, S5 and S6.