An intrinsic role of IL-33 in Treg cell-mediated tumor immunoevasion.

Regulatory T (Treg) cells accumulate into tumors, hindering the success of cancer immunotherapy. Yet, therapeutic targeting of Treg cells shows limited efficacy or leads to autoimmunity. The molecular mechanisms that guide Treg cell stability in tumors remain elusive. In the present study, we identify a cell-intrinsic role of the alarmin interleukin (IL)-33 in the functional stability of Treg cells. Specifically, IL-33-deficient Treg cells demonstrated attenuated suppressive properties in vivo and facilitated tumor regression in a suppression of tumorigenicity 2 receptor (ST2) (IL-33 receptor)-independent fashion. On activation, Il33-/- Treg cells exhibited epigenetic re-programming with increased chromatin accessibility of the Ifng locus, leading to elevated interferon (IFN)-γ production in a nuclear factor (NF)-κB-T-bet-dependent manner. IFN-γ was essential for Treg cell defective function because its ablation restored Il33-/- Treg cell-suppressive properties. Importantly, genetic ablation of Il33 potentiated the therapeutic effect of immunotherapy. Our findings reveal a new and therapeutically important intrinsic role of IL-33 in Treg cell stability in cancer.

Introduction

Foxp3+ regulatory T (Treg) cells play an instrumental role in immune homeostasis and maintenance of self-tolerance while their absence leads to fatal autoimmunity[1]. Treg cells are enriched in the circulation and tumor microenvironment of cancer patients and their presence correlates with tumor progression, invasiveness and metastasis, where they hamper the success of cancer immunotherapy [2, 3]. Treg cells represent a putative therapeutic target with checkpoint inhibitor-targeted immunotherapy against molecules mainly expressed by Treg cells to demonstrate promising results. However, still cancer immunotherapy remains ineffective in a large proportion of patients, while responses are frequently accompanied by autoimmune manifestations [4, 5]. Consequently, an urgent need exists to precisely target the tumor-specific Treg cells without affecting the peripheral Treg cell repertoire. To achieve this goal, the molecular events that dictate the suppressive program of tumor Treg cells need to be delineated.

Interleukin 33 (IL-33), an “alarmin” of the IL-1 family, has been correlated with the progression of several types of malignancies and is associated with low patient survival [6]. IL-33 is constitutively expressed by a broad range of stroma and hematopoietic cells acting as a transcription repressor released in the extracellular space upon cell death [6, 7]. Extracellular IL-33 binds to the suppression of tumorigenicity 2 receptor (ST2) and acts directly either on tumor cells enhancing their proliferation, invasion and migration or on endothelial cells promoting angiogenesis [8]. Although the impact of IL-33 in immune cell function during tumor immunosurveillance, remains unclear [8], in autoimmunity, IL-33–ST2 axis promotes Treg cell stability, expansion and conversion of CD4+Foxp3–T cells to Foxp3-expressing inducible Treg (iTreg) cells [4, 8]. Whether extracellular IL-33 supports Treg cell-mediated tumor-immune evasion and intranuclear IL-33 could shape the transcriptional landscape of Treg cells and dictate their function in an anti-tumor immune response remain unexplored.

In this report, we describe a cell-intrinsic role of IL-33 in Treg cell functional stability during tumor development. Ablation of IL-33 expression by Foxp3+ Treg cells resulted in tumor regression while IL-33-deficient Treg cells exhibited impaired suppressive properties, promoted tumor eradication and evolution of robust anti-tumor immunity. Notably, in the absence of IL-33 Treg cells maintained Foxp3 expression, consistent with a “fragile” phenotype [9, 10]. Epigenetic re-programming of tumor-exposed IL-33-deficient Treg cells resulted in the up-regulation of IFN-γ expression, which accounted for Treg cell dysfunction. Finally, genetic ablation of Il33 potentiated the therapeutic efficacy of immunotherapy. Overall the findings presented here delineate a molecular program orchestrating Treg cell stability within the tumor microenvironment.

Results

Tumor regression in IL-33-deficient mice

The precise role of IL-33 in anti-tumor immunity remains ill defined. To address IL-33 in tumors, we first performed a meta-analysis of The Cancer Genome Atlas (TCGA) Skin Cutaneous Melanoma (SCKM) dataset, which revealed a significant up-regulation of Il33 expression and correlation with metastasis (Fig. 1a). In addition, IL-33 was increased in tumors (Fig. 1b) and tumor-draining lymph nodes (tdLNs) of B16.F10 melanoma cell (B16.F10)-inoculated compared to naïve animals and correlated to tumor progression (Fig. 1c), suggesting a role for IL-33 in promoting tumor growth. In support, B16.F10-inoculated IL-33-deficient mice (Il33–/–) exhibited significantly reduced tumor growth compared to wild-type mice (Fig. 1d) accompanied by increased numbers of tumor-infiltrating CD45+ leukocytes, CD4+ and CD8+ T cells (Fig. 1e,f). Paradoxically, frequencies and numbers of intratumoral CD4+Foxp3+ Treg cells (Fig. 1e,f) were significantly elevated in Il33–/– animals. In a similar fashion, inoculation of Il33–/– mice with the poorly immunogenic Lewis lung carcinoma (LLC) cells, led to substantially diminished tumor growth but increased numbers of tumor-infiltrating CD4+Foxp3+ Treg cells (Fig. 1g,h). To exclude immune cell alterations in Il33–/– mice at steady state, naive Il33–/– mice were compared to wild type animals. No differences were found in frequencies of CD4+ T cells and CD4+Foxp3+ Treg cells and expression of Treg cell signature molecules, CD25 and glucocorticoid-induced TNFR-related protein (GITR), as well as the activation marker CD44 in the thymus and skin LNs (Supplementary Fig. 1a, b). Moreover, frequencies of CD11c+ dendritic cells (DCs) and CD11b+ monocytes in the bone marrow and spleen of the two groups did not demonstrate differences (Supplementary Fig. 1c). Overall, these findings suggested that IL-33 deficiency impaired tumor development and promoted the anti-tumor immunity with an unexpected increase in intratumoral Foxp3+ Treg cells.

Figure 1

IL-33 deficiency promotes anti-tumor immunity and inhibits tumor growth.

(a) Whiskers plot (min to max) of IL-33 expression levels (FPKM-UQ) based on TCGA datasets of primary (n=102 biologically independent human samples) and metastatic SCKM (n=367 biologically independent human samples), ***P=0.0006 (b) IL-33 levels (pg/ml) in homogenates of cultured B16.F10 melanocytes (n=8, biological independent cell cultures) and tumors from day 14 B16.F10 inoculated WT mice (n=5) determined by ELISA, **P=0.0016 (c) IL-33 levels (pg/ml) in tdLNs homogenates from B16.F10-inoculated mice at indicated time points (day 0 n=3, day 4 n=4, day 7=3, day 10=4, day 15=4), *P=0.00195, **P=0.049, ***P=0.039 (d) Tumor volume (mm3) curve of WT (n=8) and Il33–/– (n=10) mice following B16.F10 inoculation. *P=0.003, **P=0.0025, ***P=0.0027, ****P=0.0001. Gating strategy (e), percentages and numbers per 5 × 105 tumor cells (f) of intratumoral CD45+ cells, CD4+, CD8+ and CD4+Foxp3+ T cells on day 14 after B16.F10 inoculation of WT (n=8) and Il33–/– (n=10) mice. Numbers denote percentages of gated populations. *P=0.0019, **P<0.0001, ***P=0.02, ****P=0.0117, N.S.=0.9076, #P=0.0267, ##P=0.0173, ###P=0.0002 (g) Tumor volume (mm3) curve of WT (n=4) and Il33–/– (n=3) mice following LLC inoculation. *P=0.0255, **P=0.0445 (h) Numbers per 5 × 105 tumor cells of intratumoral CD45+ and CD4+Foxp3+ T cells on day 14 after LLC inoculation of WT (n=4) and Il33–/– (n=3) mice. ***P=0.0245, ****P=0.0214. Data are shown as means ± SD. Representative data from three (g, h) and four (c, e, f) independent experiments. (a, f*, f#, f##, h) unpaired two-tailed t-test, (b, f**, f***, f****, f###) Mann-Whitney two-tailed U-test, (c) one-way ANOVA with Dunnett’s multiple comparisons test, (d, g) multiple unpaired two-tailed t-tests between WT and Il33–/– at each time point. (c, d, e, f, g, h) n=biologically independent mouse samples.

Host-derived IL-33 directs tumor regression

IL-33 promotes Treg cell proliferation [8] while tumor cells have been reported to produce substantial quantities of IL-33 [11]. Thus, we postulated that increase of tumor-infiltrating Foxp3+ Treg cells in Il33–/– mice might be due to tumor-derived IL-33. To address this, we engineered B16.F10 cells that lacked expression of endogenous IL-33 through shRNAs targeting of Il33 gene. Thus, shIL-33_1 diminished IL-33 in both mRNA and protein levels compared to shIL-33_2 and scramble (Supplementary Fig. 2a), while B16.F10 transduction did not affect their viability and in vitro proliferation (Supplementary Fig. 2b). Therefore, B16.F10 cells transduced with shIL-33_1 (denoted as B16.F10) were used throughout this study. Accordingly, B16.F10-inoculated Il33–/– mice presented significantly reduced tumor volume and tumor weight as well as delayed tumor growth compared to wild type animals (Fig. 2a). This finding was also evident by the significantly decreased proliferation of tumor cells based on Ki67 expression and decreased formation of tumor vessels as determined by CD31 and vascular cell adhesion protein 1 (VCAM) immunofluorescence staining (Fig. 2b). Tumor regression in B16.F10-inoculated Il33–/– mice was accompanied by increased numbers of tumor-infiltrating CD45+ leukocytes that exhibited elevated IFN-γ expression and increased numbers of CD4+ T cells (Fig. 2c,d), while NK1.1+ cell numbers were significantly reduced (Fig. 2c, Supplementary Fig. 2c). Furthermore, tumors from Il33–/– mice were enriched in activated myeloid cells as determined by MHC class II expression on CD11c+ DCs and CD11b+ monocytes (Fig. 2c,e, Supplementary Fig. 2c). Strikingly, the robust anti-tumor immunity in B16.F10-inoculated Il33–/– mice was associated with significantly elevated numbers of intratumoral Foxp3+ Treg cells (Fig. 2c). Of note, no significant differences in Ki67 expression (proliferation) or phospho-γH2AX expression indicative of DNA damage were observed by intratumoral Foxp3+ Treg cells in melanoma-bearing wild type and Il33–/– animals (Supplementary Fig. 2d). Together, these findings underscore that host-derived IL-33 deficiency was responsible for tumor regression.

Figure 2

Host-derived IL-33 role in tumor regression.

(a) Tumor volume (mm3) and % of mice bearing tumors <400 mm3 of WT (n=9) and Il33–/– (n=10) mice 9-13 days following B16.F10 inoculation and tumor weight (g) on day 13. *P=0.0173, **P=0.0103, ***P=0.0158, ****P=0.0037, #P=0.0002, ##P=0.0076, ###P=0.0001 (b) Representative images of Ki67 and CD31 immunohistochemistry, VCAM and CD31 immunofluorescense and quantification plot of Ki67 intensity, CD31 intensity and vessel size from day 13 B16.F10 tumors from WT (n=6) and Il33–/– mice (n=6). 10 fields or more per tumor were quantified. Scale bars: Ki67=30 μm, CD31 immunohistochemistry=50 μm, CD31 and VCAM immunofluorescence=30 μm. *P<0.0001, **P=0.0004, ***P=0.0022 (c) Numbers per 5 × 105 tumor cells of CD45+, CD4+, CD8+, CD4+Foxp3+, CD11c+, CD11b+, NK1.1+ cells from day 13 B16.F10 tumors from WT (n=9) and Il33– (n=10) mice determined by FACS. *P=0.035, **P=0.0245, ***P=0.0123, ****P=0.0003, #P=0.0133, (d) Representative FACS plots and percentages of IFN-γ+ cells derived from in vitro cultures of intratumoral CD45+ cells of WT (n=9) and Il33–/–(n=10) mice 13 days following B16.F10 inoculation. Numbers denote percentages of gated populations, **P=0.0088 (e) Representative overlays and MHCII MFI of CD11c+ and CD11b+ cells infiltrating B16.F10 day 13 tumors of WT (n=9) and Il33–/– mice (n=10). *P=0.0003, **P=0.0146. Data are shown as means ± SD. Representative data from five (a, c, d, e) independent experiments. (b**, c**, c***, c****, d, e) unpaired two-tailed t-test, (a##, b***, c*, c#) Mann-Whitney two-tailed U-test, (a*, a**, a***, a****, a#) multiple unpaired two-tailed t-tests between WT and Il33–/– at each time point, (a####) two-tailed Log rank test. (a, b, c, d, e) n=biologically independent mouse samples.

Impaired suppressive function of Il33–/– Treg cells

Considering the pivotal role of Foxp3+ Treg cells in tumor development[2], we sought to investigate the contribution of the increased Foxp3+ Treg cell numbers in the anti-tumor immune response of B16.F10-inoculated Il33–/– mice. To address this, B16.F10-inoculated Il33–/– and wild type mice were treated with anti-CD25 monoclonal antibody that specifically depletes Treg cells [12]. Although depletion efficiency was similar in the two groups (Supplementary Fig. 3a), anti-CD25-treated wild type animals presented reduced tumor volume compared to non-treated control mice, while anti-CD25 treatment did not affect tumor growth in Il33–/– animals (Fig. 3a). These results suggested that depletion of Treg cells in IL-33 deficient mice did not further enhance tumor regression as was shown in wild type animals. Since IL-33 signals through its receptor ST2 on Treg cells to promote their stability and function[8], we asked whether IL-33-ST2 axis on Treg cells is implicated in tumor regression. To this end, we generated mice with ablation of MyD88, a major component of ST2-mediated signaling, in Foxp3+ Treg cells by crossing of Myd88fl/fl mice with Foxp3Cre mice. No significant changes in tumor growth and frequencies of intratumoral CD45+, CD4+ and CD4+Foxp3+ Treg cells were monitored in B16.F10-inoculated Foxp3CreMyd88fl/fl mice compared to Myd88fl/fl control animals (Supplementary Fig. 3b). Since IL-33 signaling could occur independently of MyD88[13] we sorted highly pure CD4+CD25hiGITR+ Treg cells (>90% Foxp3+) (Supplementary Fig. 3c) from either St2+/+ or St2–/– mice and adoptively transferred them into Rag1–/– recipients (that lack T and B lymphocytes) that were reconstituted with St2+/+CD4+CD25–Foxp3– and CD8+ T effector cells and further inoculated with B16.F10 melanoma cells. Notably, no significant differences in tumor volume and tumor weight were observed in Rag1–/– mice that received St2–/– Treg cells compared to St2+/+ Treg cells-transferred mice (Fig. 3b). Thus, ablation of ST2-mediated IL-33 signaling specifically in Treg cells did not alter their tumor-promoting function. These results raised the possibility of a cell intrinsic role of IL-33 in Treg cell function. To address this, we first assessed Il33 mRNA levels by in vitro activated CD4+Foxp3+ Treg cells isolated from Foxp3gfp reporter wild type or Il33–/– mice through qPCR. Indeed, Il33 was expressed by activated Treg cells from wild type mice but not from IL-33-deficient mice (Supplementary Fig. 3d). Importantly, a targeted proteomics approach for detecting IL-33 yielded a positive result in wild type Treg cells for one out of the nine proteotypic peptides in the sequence of IL-33 (DYSVELQR, aminoacids 198-205) (Fig. 3c) that was undetected in Il33–/– Treg cells (data not depicted). Based on our data showing that Treg cells produce IL-33, we further explored its functional importance in tumor growth. Specifically, Rag1–/– mice adoptively transferred with Il33–/– Foxp3+ Treg cells from Il33–/–Foxp3gfp mice in combination with Il33+/+ CD4+Foxp3– and CD8+ T cells exhibited markedly reduced tumor volume and tumor weight compared to Il33+/+ Treg cell-transferred mice (Fig. 3d) and accompanied by increased frequencies of intratumoral CD45+ leukocytes and CD4+ T cells (Fig. 3e). Importantly, we observed significant melanoma regression and delayed tumor growth in FoxpCreIl33fl/fl (specific ablation of IL-33 expression in Treg cells) mice compared to control Foxp3Cre mice (Fig. 3f), accompanied by increased numbers of tumor-infiltrating CD45+ and Foxp3+ T cells in comparison to control mice (Fig. 3g). Overall, these findings demonstrate an intrinsic role of IL-33 orchestrating the Treg cell suppressive activity in tumor microenvironment.

Figure 3

Impaired suppressive function of IL-33-deficient Treg cells in vivo.

(a) Tumor volume (mm3) and tumor weight (g) of B16.F10- inoculated control-injected or anti-CD25-depleted WT (n=3) and Il33–/– (n=5) mice. *P=0.016, **P=0.0005, ***P<0.001, #P=0.016, ##P=0.0012, ###P<0.0001, NS=0.975 (b) Tumor volume (mm3) and tumor weight (g) (day 14) of B16.F10 inoculated Rag1–/– mice reconstituted with St2+/+ Treg cells (n=5) and St2–/– Treg cells (n=7). NS=0.0996 (c) Intensity plot of the elution of y5 (m/z: 644.3726 charge: 1+) and y4 (m/z: 545.3042 charge: 1+) fragment ions of the IL-33 proteotypic peptide DYSVELQR (Aminoacids 198-205, m/z: 505.2511 charge:2+) extracted by protein extracts of in vitro activated lymph node WT CD4+Foxp3+ Treg cells, separated by high performance liquid chromatography and injected in a Q Exactive mass spectrometer operating in Parallel Reaction Monitoring (PRM) mode (d) Tumor volume (mm3) and tumor weight (g) (day 13) of B16.F10 inoculated Rag1–/– mice reconstituted with Il33+/+ Treg cells (n=7) and Il33–/– Treg cells (n=10). *P=0.0089, **P=0.0028, ***P=0.035, ****P=0.0021 (e) Numbers per 5 × 105 tumor cells of CD45+ cells and CD4+ T cells of Il33+/+ Treg cells (n=7) and Il33–/– Treg cells-transferred (n=10) Rag1–/– mice. *P=0.0137, **P=0.0431 (f) Tumor volume (mm3), tumor weight (g) and % of mice bearing tumors <400 mm3 of B16.F10- inoculated Foxp3Cre (n=5) and Foxp3CreIl33fl/fl (n=5) mice. *P=0.03, #P=0.0039, ##P=0.019, ###P=0.005, ****P=0.0101, ####P=0.0003. (g) Numbers per 5 × 105 tumor cells of CD45+, CD4+ and CD4+Foxp3+ T cells infiltrating day 16 tumors of B16.F10- inoculated Foxp3Cre (n=5) and Foxp3CreIl33fl/fl (n=5) mice *P=0.031, **P=0.02. Data are shown as means ± SD. Representative data from two independent experiments. (b, d, e, g) unpaired two-tailed t-test, (b,d,f) multiple two-sided t-tests between WT and Il33–/– at each time point, (a) one-way two-tailed ANOVA with Tukey’s multiple comparisons test, (f) two-tailed Log rank test. (a, b, d, e, f, g) n=biologically independent mouse samples.

IL-33-deficient Foxp3+ Treg cells present a “fragile” phenotype

To examine whether IL-33-deficient Treg cells adopt an ex-Treg cell phenotype, we assessed the methylation status of Foxp3 Treg cell-specific demethylated region (TSDR), which determines the stability of Foxp3 expression [14]. Il33–/– Foxp3gfp Treg cells displayed enhanced TSDR demethylation similar to Il33+/+Foxp3gfp Treg cells (wild type Treg cells), implying a stable Foxp3 expression (Fig. 4a). Interestingly, expression of cell surface molecules linked to Foxp3+ Treg suppressive activity, such as CTLA-4, LAG-3, GITR, killer cell lectin-like receptor subfamily G member 1 (KLRG-1), tumor necrosis factor receptor superfamily member 4 (OX-40) and inducible T-cell costimulator (ICOS) did not demonstrate any significant alteration in Il33–/–Foxp3gfp compared to Il33+/+Foxp3gfp Treg cells from B16.F10-inoculated animals. In contrast, we observed a significant up-regulation of PD-1 expression while expression of neuropilin (Nrp-1) was markedly reduced in Il33–/–Foxp3gfp Treg cells (Fig. 4b) consistent with a “fragile” Treg cell phenotype [9, 10]. Finally, assessment of the metabolic profile of tumor-infiltrating IL-33-deficient Treg cells and in particular the mammalian target of rapamycin (mTORC-1) signaling that has been shown to impact on Treg cell-mediated suppression [15], demonstrated increased abundance of phosphorylated mTOR, S6 and eukaryotic translation initiation factor 4E-binding protein 1 (4E-BP1) compared to wild type Treg cells (Fig. 4c). Overall these data pointed towards a cell-intrinsic role of IL-33 in shaping the function and metabolic profile of Foxp3+ Treg cells in tumor microenvironment by hampering the induction of “fragile” Foxp3-expressing Treg cells.

Figure 4

IL-33-deficient Treg cells acquire a “fragile” phenotype.

(a) Color-coded methylation status (white=0% methylation, red=100% methylation) of individual CpG motifs (1-7) in the TSDR of foxp3 gene in CD4+Foxp3– (Teffectors) and CD4+Foxp3+ cells (Treg cells) infiltrating tumors of WT (n=4) and Il33–/– (n=4) mice (b) Histogram overlays of KLRG-1, OX-40, CTLA-4, LAG-3, GITR, ICOS, NRP and PD-1 expression and plots of PD-1 (n=11) and NRP (n=5) MFI of tumor-infiltrating Treg cells (CD4+Foxp3+) from WT and Il33–/– mice. *P=0.023, **P=0.0133 (c) Overlays and MFI of pmTOR, pS6, p4E-BP1 from intratumoral Treg cells (CD4+Foxp3+) of WT (n=5) and Il33–/– (n=6) mice. *P=0.0056, **P=0.0085, ***P=0.0487. Data are shown as means ± SD. Representative data from two independent experiments. unpaired two-tailed t-test. (a, b, c) n=biologically independent mouse samples.

IFN-γ mediates the anti-tumor activity of Il33–/– Treg cells

To investigate the transcriptional alterations that lead to the “fragile” phenotype of IL-33-deficient Foxp3+ Treg cells, highly pure CD4+CD25+GITR+Foxp3+ Treg cells isolated from tdLNs of B16.F10-inoculated Il33–/– and wild type animals were subjected to mRNA sequencing (RNAseq) revealing 585 differentially expressed genes in tumor-exposed Il33–/– Treg cells compared to Il33+/+ Treg cells (log2FC > 0.4 and adjusted p-value <0.05) (Fig. 5a). IL-33-sufficient and IL-33-deficient Treg cells transcriptomes were clearly separated by distance clustering (Supplementary Fig. 4a). In accordance with the cell surface phenotype of Il33–/– Foxp3+ Treg cells, no significant changes in Treg cell-signature genes such as Foxp3, Tgfb, Cd39, Ctla4, Granzyme, Perforin, Helios, Il7r, Gitr, Ox40, Ccdc22, Cmtm7 and Ecm1 were observed (data not shown). Interestingly, Ifng listed among the upregulated genes of the prominent cluster of differentially expressed chemokines and cytokines (Fig. 5b). Increased expression and protein abundance of IFN-γ was validated in tumor-exposed Il33–/– Foxp3+ Treg cells (Fig. 5c,d). Overall, gene expression data suggested a highly reshaped transcriptional landscape in IL-33-deficient Treg cells correlating with a “Treg cell-fragile” phenotype. Towards an in depth molecular mechanism delineation, transposase-accessible chromatin sequencing (ATAC-seq) revealed a significant altered chromatin landscape of Il33–/– Treg cells compared to Il33+/+ Treg cells. Chromatin rearrangement was depicted in 1,432 differentially accessible regions (Figure 5e), distributed among promoters, introns and distal intergenic regions while the majority of these peaks concerned proximal promoters (-1 kb to transcription start site (TSS)) (Suppl. Fig 4b). Notably, data analysis with Genomic Regions Enrichment of Annotations Tool (GREAT) showed enrichment of genes implicated in IFN-γ pathway (Suppl. Fig 4c) while Ifng was among the genes that exhibit simultaneous differential chromatin accessibility and gene expression (Fig. 5f). Specifically, conserved non-coding region +17.703 to +20.863 kb to TSS (CNS +17/+19) of Ifng locus, which acts as a transcription enhancer [16], gained chromatin accessibility in tumor-exposed Il33–/–compared to wild type Treg cells (Fig 5g).

Figure 5

Role of IFN-γ in the impaired suppressive function of Il33–/– Treg cells.

(a) Heatmap of differentially expressed genes of tdLNs Treg cells (CD4+CD25+GITR+Foxp3+) from WT (n=3) and Il33–/– (n=3) mice. (b) Heatmap of differentially expressed cytokines, chemokines and chemokine receptors genes of tdLNs Treg cells from WT (n=3) and Il33–/– (n=3) mice (c) Relative expression of IFN-γ in tdLNs Treg cells from WT (n=7) and Il33–/– (n=8), *P=0.0401 (Mann-Whitney two-tailed U-test) (d) Representative overlay and MFI of IFN-γ produced by Treg cells from tdLNs of WT (n=6) and Il33–/– (n=4) mice and percentages of IFN-γ positive Treg cells from CD45+ tumor infiltrating cells following in vitro activation *P=0.0469 (unpaired two-tailed t-test) (e) Heatmaps demonstrating normalized ATAC-seq tag counts around 1595 loci (+/- 1.25 kb window) of tdLNs Treg cells from WT (n=3) and Il33–/– (n=3) mice (f) Heatmaps of genes exhibiting simultaneous differential chromatin accessibility and gene expression of tdLNs Treg cells from WT (n=3) and Il33–/– (n=3) mice (g) ATACseq signal profiles across Ifng locus (Chromosome 10: 118.361.046-118.521.046 bp) and Ifng CNS +17/+19 (Chromosome 10: 118.457.858-118.462.499) of tdLNs Treg cells from WT (n=6) and Il33–/– mice (n=4) (h) Tumor volume (mm3) and tumor weight (g) (day 14) of B16.F10 inoculated Rag1–/– mice adoptively transferred with Treg cells from WT (n=8), Ifng–/– (n=3), Il33–/– (n=9) and Il33–/–Ifng–/– mice (n=7). *P=0.049, **P<0.0001, ***P=0.0279, #P=0.019, ##P=0.0129, ###P=0.0085, NS1=0.9745, NS2=0.9678, NS3=0.8742 (one-way ANOVA with Tukey’s multiple comparisons test). Data are shown as means ± SD. (a, b, f) Color corresponds to per gene z-score normalized FPKM counts across samples and genes are hierarchically clustered. (a, b, c, d, e, f, g, h) n=biologically independent mouse samples.

To investigate the functional importance of IFN-γ expression by Il33–/– Treg cells, we crossed Ifng–/– animals with Il33–/–Foxp3gfp mice to generate IFN-γIL-33 double knockout mice (Il33–/–Ifng–/–). Adoptive transfer of highly pure Il33–/– Ifng–/–Foxp3gfp Treg cells into T cell-reconstituted B16.F10-inoculated Rag1–/–recipients demonstrated that ablation of IFN-γ restored the suppressive capacity of Il33–/–Foxp3+ T cells and promoted tumor growth in contrast to Il33–/–Ifng+/+Foxp3gfp Treg cells (Fig. 5h). Collectively these findings revealed an essential role of IFN-γ in the functional re-programming of tumor-exposed Il33–/– Foxp3+ Treg cells.

IFN-γ production by Il33–/– Treg cells is dependent on NF-κB–T-bet axis

Next we sought to unravel the molecular events that drive IFN-γ up-regulation in IL-33-deficient Foxp3+ Treg cells. Enrichment of differentially expressed transcriptional regulators that govern Ifng expression emerged from gene expression analysis of tumor-exposed Il33–/– Treg cells (Fig. 6a) containing Tbx21, which encodes the transcription factor T-bet and is an important node of the Ifng gene regulatory network as illustrated by ingenuity pathway analysis (Fig. 6b). Consistently, T-bet expression was increased in Foxp3+ Treg cells from tumors or tdLNs of B16.F10-bearing Il33–/– mice compared to wild type mice (Fig. 6c), emulated by the ST2+ Treg cell population (Supplementary Fig. 5a). In support, T-bet expression was significantly up-regulated in Treg cells infiltrating tumors of Foxp3CreIl33fl/fl animals compared to control Foxp3Cre mice (Fig. 6c), while treatment of Il33–/– Treg cells with recombinant IL-33 did not alter T-bet expression (Supplementary Fig. 5b).

Figure 6

Increased NF-κB activation and T-bet expression promote IFN-γ production in Il33–/– Foxp3+ Treg cells.

(a) Heatmap of the differentially expressed transcription regulators of tdLNs Treg cells (CD4+CD25+GITR+Foxp3+) from WT (n=3) and Il33–/– (n=3) mice. Arrows indicate transcription regulators of IFN-γ (b) Ingenuity Pathway Analysis of IFN-γ upstream regulators. The molecules color represents the expression levels (log2): blue=underexpressed, red=overexpressed in Il33– samples compared to WT samples (c) Representative overlays and MFI of T-bet expression in tumor-infiltrating Treg cells (WT n=6, Il33–/– n=6, Foxp3Cre=3, Foxp3CreIL33fl/fl=3) and tdLNs Treg cells (WT n=10, Il33–/– n=10) of B16.F10 inoculated mice. *P=0.0454, **P=0.0201, ***P=0.019 (unpaired two-tailed t-test). Data are shown as means ± SD (d) Relative fold enrichment of T-bet binding (anti T-bet/IgG) to IFN-γ promoter and CNS17-19 by ChIP performed on chromatin derived from tdLNs Treg cells. Results from 2 independent experiments from pooled WT (n=4) and Il33–/– (n=4) mice *p=0.0178, unpaired Student two-tailed t test. Data are shown as means ± SD (e) Heat map of differentially expressed and differentially accessible NF-κB target genes of tdLNs Treg cells from WT (n=3) and Il33–/– (n=3) mice (f) Representative overlays of T-bet and IFN-γ expression and gating strategy of WT (n=6) and Il33–/– (n=5) CD4+CD8-Foxp3+ Treg cells in two days tdLN cell culture with aCD3/aCD28 beads in the presence or absence of IKK2 inhibitor. MFI of T-bet and IFN-γ is normalized to the MFI of Treg cells cultured in the absence of IKK2 inhibitor. Data are shown as means ± SD (c, d) Representative data of three independent experiments. (a, e) Color corresponds to per gene z-score normalized FPKM counts across samples and genes are hierarchically clustered. (a, c, d, e, f) n=biologically independent mouse samples.

In silico motif analysis demonstrated putative binding sites of T-bet in the regulatory region of Ifng locus CNS +17/+19, which presented increased accessibility in tumor-exposed IL-33-deficient Treg cells (Supplementary Fig. 5c). T-bet binding to Ifng proximal promoter and CNS +17/+19 regions was assayed by ChIP, demonstrating significantly increased engagement (50%) in both regulatory elements of tumor-exposed Il33–/– Treg cells compared to wild type (Fig. 6d). These data suggest that enhanced expression of IFN-γ in Il33–/– Treg cells is facilitated by increased T-bet expression and concomitant binding to the re-arranged CNS +17/+19 regulatory region.

We then sought to assess the mechanisms underlying enhanced T-bet expression in Il33–/– Foxp3+ Treg cells. Intracellular IL-33 has been implicated in the regulation of NF-κB signaling [17], which is required for T-bet expression [18]. Particularly, IL-33 binds to NF-κB and restricts its transcriptional activity [17]. Thus, we postulated that in the absence of intracellular IL-33, Foxp3+ Treg cells experience increased activation of NF-κB signaling resulting in upregulation of Tbx21 expression and IFN-γ production. In support, GO analysis of mRNAseq data highlighted the activation of NF-κB pathway in tumor-exposed Il33–/– Treg cells (Supplementary Fig. 5d), while known NF-κB targets such as Nfkbia, Nfkbid, Nfkbiz, Atf3, Fos, Jun, Gadd45b were highly enriched among the differentially expressed genes of IL-33-deficient Treg cells compared to wild type Treg cells (Fig. 6e). Moreover, five of those genes (Tnfaip3, Ifng, Cybb, Slc2a6 and Itgam) were also unveiled by ATAC-seq data (Fig. 6e). Furthermore, NF-κB has been shown to regulate Treg cell activation [19]. In accordance, intratumoral Il33–/– Foxp3+ Treg cells exhibited an increased activation profile based on the expression of CD62L, CD44 and CD25 compared to Treg cells from wild type mice (Supplementary Fig. 5e). In order to study the functional importance of the NF-κB pathway activation in T-bet and IFN-γ expression in tumor-exposed Il33–/– Treg cells, we blocked NF-κB activity using IKK2 inhibitor (SC-514) in vitro. NF-κB inhibition significantly downregulated both T-bet and IFN-γ expression in Foxp3+ Treg cells compared to untreated cells (Fig. 6f). Overall, these findings demonstrate that IFN-γ production in tumor-exposed Il33–/– Foxp3+ Treg cells is NF-κB-T-bet axis dependent.

IL-33 deficiency potentiates the effectiveness of immunotherapy

Elucidating the molecular events that stabilize the suppressive program of tumor-specific Foxp3+ Treg cells is of urgent need in order to advance the therapeutic potential of targeted cancer immunotherapy. Thus, B16.F10- inoculated Il33–/– and wild type mice were treated with anti-CTLA-4 or anti-PD-1. Strikingly, we observed significantly reduced tumor growth and delayed development in the Il33–/– animals treated with either anti-CTLA-4 (Fig. 7a) or anti-PD-1 (Fig. 7b) compared to non-treated littermates and to similarly treated Il33+/+ animals. Both treatments did not delete intratumoral Foxp3+ Treg cells (Fig. 7a, b) while tumor regression in anti-PD-1-treated Il33–/– and Il33+/+ animals was also accompanied by a significant expansion of CD8+ T cells in TILs (Fig. 7b). In addition, enhanced activation of CD8+ T cells in Il33–/– mice was observed (Fig. 7b). In contrast, neutralization of extracellular IL-33 (soluble ST2-sST2) together with anti-PD-1 did not show any beneficial effect on tumor regression compared to anti-PD-1 alone in wild type mice, further supporting a cell-intrinsic role of IL-33 in the effectiveness of checkpoint immunotherapy (Supplementary Fig. 6). Collectively, the impaired suppressive activity of Il33–/– Foxp3+ Treg cells could act in concert with checkpoint inhibitor targeted immunotherapy to induce a robust tumor regression.

Figure 7

IL-33 deficiency potentiated immunotherapy efficacy.

(a) Tumor volume (mm3) curve (days 11-14), % of mice bearing tumors <200 or 400 mm3 and tumor weight (g), percentages and CD44 MFI of CD8+ and percentages of Foxp3+CD4+ T cells in TILs (CD45+) (day 14) of B16.F10 inoculated control and aCTLA-4 (b) or aPD-1 (c) treated WT (n=5, 4) and Il33–/– (n=3) mice. Data are shown as means ± SD a*P=0.0421, a**P=0.0048, a***P=0.0081, a****P=0.0161, a#P=0.0517, a##P=0.0010, a###P=0.0022, a####P=0.0003, b*P=0.0484, b**P=0.0033, b***P=0.0069, b****P=0.0017, b#P=0.0159, b##P=0.0159, b###P=0.016, b####P=0.04 (one-way ANOVA with Tukey’s multiple comparisons test), a&P=0.0022, a&&P=0.0004, b&P=0.0187, b&&P=0.0158 (Log rank test). Representative data from two independent experiments. (a, b, c) n=biologically independent mouse samples.

Discussion

An arduous task in the success of cancer immunotherapy is the selective depletion of tumor-specific Foxp3+ Treg cells or their functional reprogramming. Identification of cell-intrinsic pathways that potentiate the suppressive program of Treg cells could provide unique therapeutic strategies for tumor eradication. Herein we identify cell autonomous IL-33 to be indispensable for Foxp3+ Treg cell function. Tumor-exposed IL-33-deficient Treg cells are reprogrammed to up-regulate IFN-γ expression and exhibit decreased suppressive activity in vivo thus promoting tumor regression.

Treg cell compartment is characterized by functional heterogeneity and lineage plasticity [20]. While, Treg cell conversion is often accompanied by loss of Foxp3 expression (exTreg cells), IL-33-deficient Treg cells downregulate Nrp expression, produce IFN-γ and maintain Foxp3 closely resembling a “fragile” Treg cell phenotype [9, 10]. In addition, IL-33-deficient Treg cells exhibited increased levels of mTOR activity which has been closely correlated with Treg cell instability [21, 22, 23]. Specifically, mTOR, functions through two complexes the mTOR complex 1 (mTORC1) and mTORC2 which direct suppressive function and proliferation respectively [21, 24], and we demonstrate activation of the mTORC1 complex expression in Il33–/– Treg cells associated with the “fragile” phenotype of Treg cells. In support ablation of autophagy [25] or loss of PTEN [22] expression in Treg cells increased mTORC1 activity and led to destabilization of Treg cells. Finally, enhanced mTORC1-mediated glycolytic metabolism in TRAF3IP3–/– Treg cells caused their instability and promoted anti-tumor immunity [26].

The concept of IFN-γ expression by Treg cells is not new, however its functional role remains controversial. To this end, IFN-γ-producing CD4+CD127lowCD25+ Treg cells are enriched in type 1 diabetes [27] and multiple sclerosis patients [28]. In addition, during T. gondii infection, Treg cells, express high levels of IFN-γ and become inflammatory aiding in the clearance of the parasite [29]. Finally, Nrp1-deficient “fragile” Foxp3+ Treg cells express elevated levels of IFN-γ and inhibit tumor growth [9] and Treg cells deficient in the CARMA1-BCL10-MALT1 (CBM) signalosome complex up-regulate IFN-γ expression aiding the effectiveness of immunotherapy [30]. Although the molecular programs leading to induction of “fragile” IFN-γ-expressing Treg cells remain ill defined, in steady state Treg cells with Foxo1 nuclear exclusion [31] or with E3 ubiquitin ligase VHL deficiency [32] secrete IFN-γ and induce severe autoimmune-like inflammation. Our findings extend this knowledge demonstrating a cell-intrinsic role of IL-33 in the induction of Foxp3+ Treg cell “fragile” phenotype during tumor immune surveillance.

Treg cells adopt the transcriptional program of lineage specific T effector cells under inflammatory conditions [33]. In this regard, the transcription factor T-bet (Tbx21), which drives expression of IFN-γ [34], is expressed in a subpopulation of Foxp3+ Treg cells however its role remains controversial [33]. Thus, ablation of T-bet+ Treg cells results in the development of aggressive Th1-mediated autoimmunity [35] while genetic deletion of Tbx21 in Foxp3+ Treg cells does not cause immunopathology [36]. Noticeable, T-bet expression is not always accompanied by IFN-γ production by Foxp3+ Treg cells [35] while in Nrp1–/– “fragile” Treg cells both Ifng and Tbx21 were highly expressed [9]. Our findings demonstrate major differences in chromatin accessibility of tumor exposed IL-33-deficient compared to wild type Treg cells characterized by an enhanced open chromatin conformation in the CNS +17-19 flanking the Ifng locus of Il33–/–Foxp3+ Treg cells, which is a known cis-regulatory element driving IFN-γ transcription [16] and permissive for T-bet binding [37]. In addition expression of Bhlhe40, which serves as a cofactor to regulate IFN-γ expression was significantly up-regulated in Il33–/– Foxp3+Treg cells. Overall, IL-33 regulates the transcriptional landscape of Foxp3+ Treg cells in tumors through the T-bet-IFN-γ axis.

Despite its role as an alarmin, IL-33 has been shown to have a prominent role in regulating chromatin structure [38], through binding to nucleosome acidic patches composed of H2A and H2B histones [39] and also to exert transcriptional regulatory properties [40]. Of interest, IL-33 was demonstrated to interfere with the DNA-binding activity of NF-κB and to significantly impair its transcriptional activation capacity [17]. Although this finding was challenged in subsequent studies [41, 42] nevertheless a context- and cell-dependent function of IL-33 cannot be excluded. In support to data presented by Ali et al [17], we demonstrate that intracellular IL-33 acts as a “rheostat” for NF-κB activity in Treg cells orchestrating their suppressive program and lineage stability. In line with this, increased NF-κB activation in Treg cells was accompanied by increased IFN-γ secretion and impaired suppressive activity [43]. Although a specialized role of NF-κB subunits in Treg cell stability and function through ablation of NF-κB activity in Foxp3+ Treg cells has been reported [19, 44], it is possible that the fine-tuning of NF-κB activity rather than its complete absence dictates the reprogramming of activated Treg cells.

Unleashing the Treg cell-mediated breaks without affecting immune homeostasis remains an unmet need in cancer immunotherapy. The efforts have been hampered by the lack of knowledge on the precise molecular programs that underscore Treg cell function due to increased plasticity and heterogeneity of the Treg cell compartment, which is also shaped by the distinct anatomic locations and inflammatory environments [20]. In the context of tumors, a major achievement would be to specifically re-program the intratumoral Treg cells while their peripheral counterparts remain unaffected in order to avoid the undesired and some times life-threating immune related adverse events described during Treg cell systemic ablation [45, 46]. In this regard, we reveal intracellular IL-33 as a potential “rheostat” in the lineage stability of effector Treg cells in tumor-bearing animals. Importantly, IL-33 deficient mice develop a prominent response to checkpoint inhibitor immunotherapy, which is moderately effective in wild type animals [47, 48] and cancer patients [49]. In particular, consistent with previous reports [50], anti-PD-1 therapy promotes the expansion of CD8+ T cell compartment in wild type and Il33–/– mice. Furthermore, it is possible that the increased activation observed in CD8+ T cells may act synergistically with “fragile” Treg cells to promote tumor regression enhancing the effectiveness of immunotherapy in Il33–/– mice. Collectively these findings underscore the importance of delineating the molecular programs of Treg cell stability in order to identify therapeutic targets to destabilize Treg cells favoring anti-tumor immunity without ensuing autoimmunity.

Methods

Animals

C57BL/6J and Rag1–/– (C57BL/6J background) were purchased from Jackson Laboratory, Il33–/– mice (C57BL/6J background) were obtained from Amgen Inc and previously described by Martin et al. [51], Ifng–/– were purchased from Jackson Laboratory and kindly provided by E. Andreakos (Biomedical Research Foundation Academy of Athens-BRFAA), B6.129P2(SJL)-Myd88tm1Defx/J (Myd88fl/fl) were purchased by Jackson Laboratory and kindly provided by V. Panoutsakopoulou (BRFAA), Foxp3Cre and Foxp3gfp.KI mice (C57BL/6 background) were kindly provided by A. Rudensky (Memorial Sloan-Kettering Cancer Center, New York, USA), St2–/– (Il1rl1–/–) mice were kindly provided by AN. McKenzie (Medical Research Council Laboratory of Molecular Biology, Cambridge, UK) and Il33fl/fl mice were purchased from Jackson Laboratory. Il33–/–Foxp3gfp mice were generated by crossing Il33–/– mice with Foxp3gfp.KI mice, Il33–/–Ifng–/– mice by crossing Il33–/– with Ifng–/–, Foxp3CreMyd88fl/fl mice by crossing Foxp3Cre with Myd88fl/fl and Foxp3CreIl33fl/fl by crossing Foxp3Cre with Il33fl/fl. All mice were maintained in the animal facility of BRFAA. All procedures were in accordance to institutional guidelines and were approved by the Institutional Committee of Protocol Evaluation in conjunction with the Directorate of Agriculture and Veterinary Policy, Region of Attika, Greece (1202/19.03.2018). Unless otherwise indicated, experiments used sex- and age-matched mice between 6 and 12 weeks of age.

Cell lines

The cancer cell lines B16.F10 and LLC were kindly provided by A. Eliopoulos (School of Medicine, University of Crete, Greece) and were negative for mycoplasma, tested by PCR.

Transplant tumor models

Transplant tumor models were performed as described previously [52]. Mice were implanted subcutaneously (s.c.) on the back with 3 × 105 B16.F10 melanoma cells or shIL-33 transduced B16.F10 cells (B16.F10) or LLC cells. For the application of immunotherapy or the depletion of Treg cells every three days following the implantation, mice were treated with anti-CTLA-4 (clone 4F10, Bioceros LB) (100 μg/ 100 μl/ intraperitoneally-i.p. in each mouse) or anti-PD-1 (clone RMP1-14 Bioceros LB) (200 μg/ 100 μl i.p./ mouse) or anti-CD25 (clone PC61, Bioceros LB) (100 μg/ 100 μl i.p. / mouse). For the adoptive transfer experiments 3 × 105 CD4+CD25+GITR+Foxp3+ cells sorted highly pure (98%) from magnetically-CD4-enriched LN and spleen cells of naïve C57BL/6 or Il33–/– or Ifng–/– or Il33–/–Ifng–/– or St2–/– mice were intravenously transferred to Rag1–/– hosts with 8 × 105 CD4+CD25–GITR– cells and 7 × 105 CD8+ cells sorted highly pure (96%) from LNs and spleen of naïve C57BL/6 mice. Three days later Rag1–/– mice were implanted with B16.F10 cells. Tumors were measured every day by caliper, and tumor volume was calculated as 0.5 × length × width × width. Mice were euthanized when tumors grew larger than 1,100 mm3. At the endpoint of each experiment tumor weight was also determined.

Lentivirus production and B16.F10 transduction

To knockdown the expression of Il33 we designed 2 shRNA sequences (shIL-33_1: 5′-GGTGCTACTACGCTACTATGA-3′ and shIL-33_2: 5′-GAACATGAGTCCCATCAAAGA-3′) that specifically target the transcripts of mouse Il33. The shRNA sequences were cloned into the pLKO.1 lentiviral vector. HEK 293T (kindly provided by A. Klinakis) were transiently co-transfected with plasmids containing the viral packaging genes (pMDG and R891) and with pLKO.1 carrying either the shIL-33 or an shRNA scramble sequence. Lentivirus-containing medium was collected at 24 post-transfection and used to transduce B16.F10 cells for 16 h. Two days later selection of transduced cells was started with puromycin (5 μg/ml) or blasticidin (5 μg/ml), for 5 and 12 days, respectively. A pLKO vector carrying the GFP gene was used to assess transduction efficiency and the titer according to the manufacturer’s instructions.

Cell isolation from tumors and lymphoid organs

Single cell suspensions from LNs, spleens and thymus were generated by passing them through a 40 μm cell strainer. BM cells were isolated by flushing out femur and tibia. TILs were isolated by dissociating tumor tissue in the presence of collagenase D (1 mg/ml, Roche) and DNAse I (0.25 mg/ml, Sigma) for 45 min before passing through a 40 μm cell strainer. For LN cells activation, 4 × 105 total tdLN cells were cultured with Dynabeads™ mouse T-activator CD3/CD28 (Gibco) in a ratio 1:1 for 48 h. Recombinant murine IL-2 (1000 U/ml, Peprotech) and recombinant murine IL-33 (20 ng/ml, Immunotools) were added where indicated.

Flow cytometry

For staining of extracellular markers, cell suspensions were incubated with antibodies for 20 min at 4°C. The following antibodies were used (all antibodies were purchased from BioLegend): CD45 (clone 30-F11), CD4 (RM4-4), CD8 (53-6.7), CD11c (N418), CD11b (M1/70), NK1.1 (PK136), IAb (AF6-120.1), CD25 (PC61), GITR (DTA-1), KLRG-1 (2F1/KLRG1), OX-40 (OX-86), CTLA-4 (UC10-4B9), LAG-3 (C9B7W), ICOS (7E.17G9), PD-1, CD44 (IM7), CD62L (MEL-14), NRP (3E12), ST2 (DIH9), Ki67 (16A8), pγH2AX (2F3). For Foxp3 (150D), phospho-mTOR (Ser2448) (MRRBY), phospho-S6 (Ser235, Ser236) (cupk43k), phospho-4E-BP1(T36/T45) (V3NTY24) and T-bet (4B10) intracellular staining, cells were stained for the extracellular markers and then fixed and stained using the Foxp3 Staining Set (eBioscience) according to the vendor instructions. For IFN-γ (XMG1.2) intracellular staining, cells were incubated with 50 ng/ml PMA (Sigma-Aldrich), 2 μg/ml Ionomycin (Sigma-Aldrich) and Brefeldin (1/1000) (BD) for 4 h at 37°C and 5% CO2, stained for extracellular markers, fixed and stained using the Foxp3 Staining Set (eBioscience). All samples were analyzed with ARIA III (BD). Flow cytometry data were analyzed with FlowJo 8.7 and vX0.7 software.

Quantitative PCR

Cells RNA was extracted using Macherey Nagel Nucleospin RNA kit and Nucleospin RNA XS kit according to manufacturer guidelines. First-strand cDNA synthesis was performed using PrimeScript™ RT reagent kit (Takara). qPCR was carried out in 20 μl reactions using the KAPA SYBR@ FAST (KapaBiosystems). The program used was: 95°C for 3 min, 95°C for 10 sec and 62°C for 30 sec for 40 cycles. Relative expression of target genes was calculated by comparing them to the expression of the house-keeping genes Hprt. The following primers were used:

Hprt: Forward 5′-GTGAAACTGGAAAAGCCAAA-3′, Reverse 5′-GGACGCAGCAACTGACAT-3′, Il33: Forward 5′-GCTGCGTCTGTTGACACATT-3′, Reverse 5′-CACCTGGTCTTGCTCTTGGT-3′, Ifng: Forward 5′-ATGAACGCTACACACTGCATC-3’, Reverse 5’-CCATCCTTTTGCCAGTTCCTC-3′ from Invitrogen.

ELISA

Tumor and LN homogenates were generated in phosphate buffered saline (PBS) containing a cocktail of protease inhibitors (Roche) using a pestle. B16.F10 transduced with scramble, B16.F10 transduced with shIL-33_1 and B16.F10 transduced with shil33_2 lysates were generated by 5 freeze-thaw cycles of cultured cells. The homogenates and lysates were centrifuged and used for IL-33 ELISA (R&D) according to manufacturer’s guidelines.

Immunohistochemistry

Tumor samples were fixed overnight in 4% PFA/PBS, washed with PBS and a small part was immersed in 30% sucrose/PBS, embedded in OCT (VWR Chemicals) and frozen, while the rest was embedded in paraffin. 4-μm-thick sections of FFPE (Formalin-Fixed, Paraffin-Embedded) tumors were blocked using 1% BSA in TBS containing 0.05% Tween 20 (Sigma) (TTBS) and incubated with anti-Ki67 (Thermofisher Scientific, MA5-14520) or anti-CD31 (Abcam, ab28364) primary antibodies. A biotinylated secondary anti-rabbit antibody (Vector Laboratories) and the Vectastain ABC kit (Vector Laboratories) were used for signal detection and amplification, respectively. Signal development was performed with the Vectastain DAB (3,3-diamino-benzidine) kit (Vector Laboratories) and hematoxylin was used as a counterstain. Images were acquired with an Eclipse E800 microscope (Nikon) equipped with a QImaging Digital Camera. Cryosections were incubated with the anti-CD31 and anti-VCAM1 (Abcam, ab19569) antibodies, followed by the secondary A568-conjugated anti–rabbit antibody (Invitrogen, A110011) and the A488-conjugated anti-rat antibody (Invitrogen, A11006), respectively. Mounting medium containing DAPI (Sigma-Aldrich) was used to stain the nuclei and images were acquired using a Leica TCS SP8X White Light Laser confocal system. Quantifications were performed using ImageJ software analysis

mRNAseq and Microarray data

Microarray data generated by TCGA were downloaded from TCGA data portal (https://tcga-data.nci.nih.gov/tcga/).

The library preparation for mRNAseq was carried out in the Greek Genome Center (GGC) of BRFAA. RNA was collected from CD4+CD25+GITR+Foxp3+ sorted cells from tdLNs. RNA-seq libraries (GSE138871) were prepared using the TruSeq RNA kit using 500 ng of total RNA. The libraries were constructed according to Illumina's protocols and then were mixed in equal amounts. Paired-end 38 bp reads for 6 samples were generated with NextSeq500 in the GGC. RNA-seq analysis was done using in-house developed sub-pipeline [53]. Specifically, sequencing reads of sgRNA knockdown samples and their control samples aligned to reference genome mm10 using STAR-aligner version 2.4.2 [54] with parameters suggested by TCGA expression mRNA-seq pipeline(https://docs.gdc.cancer.gov/Data/Bioinformatics_Pipelines/Expression_mRNA_Pipeline/), and the raw read counts were generated. Then DESeq2 [55] was used to perform differential expression analysis.

Assay for Transposase Accessible Chromatin (ATAC)-Sequencing

ATAC-seq was performed as described previously [56]. In brief, 2 × 104 – 6 × 104 Treg cells from tdLNs were sorted as CD4+CD25+GITR+Foxp3+ using the BD AriaII. Cells were resuspended in 50 μl cold lysis buffer (10 mM Tris-HCl, pH 7.4, 10 mM NaCl, 3 mM MgCl2, 0,1% IGEPAL CA-630) and centrifuged at 500 g for 10 min. After centrifugation supernatant was aspirated and transposition was performed with in-house Tn5 Transposase in 20 μl tagmentation buffer (10 mM TAPS-NaOH, pH8.5, 5 mM MgCl2, 10% DMF) for 30 min at 37°C. DNA was purified using the Qiagen MinElute Kit (Qiagen) and amplified using the NEBNext High-Fidelitiy PCR Master Mix (New England BioLabs) and customized Nextera PCR primers:

Libraries were purified using the Qiagen MinElute Kit (Qiagen) and quality checked using the Agilent Tapestation 2200. Libraries were sequenced on an Illumina HiSeq1500 with a minimum of 10 × 106 reads/sample (GSE138872). ATAC-seq analysis was done using in-house developed sub-pipeline [53]. Specifically, sequencing reads were aligned to reference genome mm10 using sing Botwie2 [57] with default parameter. MACS2 [58] is used for peak calling with broad peak calling mode. DiffBind is used to perform differential binding analysis with EDGER normalization method, and q-value of 0.05 was used as the cutoff for calling differential binding peaks.

Pyrosequencing

1 × 105-1 × 106 T effectors (CD4+CD25–GITR–Foxp3–) and Treg cells (CD4+CD25+GITR+Foxp3+) were sorted from tdLNs and their DNA was isolated using Nucleospin Tissue (Macherey Nagel) according to manufacturer’s guidelines. Sodium bisulfite treatment of genomic DNA was carried out using the EZ DNA Methylation Gold Kit (Zymo research) according to the manufacturer´s instructions. Fragments for pyrosequencing were generated by PCR using the PyroMark PCR Kit (Qiagen) with the following protocol denaturation at 95°C for 15 min, followed by 44 cycles at 94°C for 30 s, 56°C for 30 s, and 72°C for 30 s. PCR products were visualized by gel electrophoresis. The primers used were: SET1: Forward 5′-GGAGGAGGAAGAGGAGGT-3′, Reverse 5′-AAATATAAACACCAAATAAAACCC-3′, Sequencing 5′-GGAGGAAGAGGAGGTT-3′ and SET2: Forward 5′-GAAGAGGTTTTGGGGTTGT-3′, Reverse 5′-CCATCCTCTTCCTCATCAAC-3′ and Sequencing 5′-AGGTTTTGGGGTTGTTT-3′. For purification of biotinylated fragment, 20 μl of PCR products were added to a mix consisting of 3 μl Streptavidin Sepharose HP™ Beads (GE Healthcare) and 40 μl binding buffer (QIAGEN). Sepharose beads with the single-stranded templates attached were added to a 24 well Plate (QIAGEN) containing a mix of 25 μl annealing buffer (QIAGEN) and 0.3 μM of the corresponding sequencing primers (Invitrogen). Pyrosequencing was performed in a PyroMark Q24 Instrument with the PyroMark Gold Q24 Reagent Kit (QIAGEN). For pyrogram exposure including CpG-site methylation calculation, the PyroMark Q24 Method 010 Software Version 2.0.7 (QIAGEN) was applied. Only pyrograms including sharp peaks with sufficient height for each injected nucleotide of interest and without peaks for unsuccessful bisulfite treatment or background controls were considered.

Chromatin immunoprecipitation (ChIP)

2 × 106 cells were cross-linked with 1% (vol/vol) formaldehyde for 10 min at 25 °C (followed by extensive wash with PBS) and lysed with 120 μl lysis buffer (1% (wt/vol) SDS, 10 mM EDTA and 50 mM Tris-HCl, pH 8.1, 1 × protease inhibitor ‘cocktail’ (Roche), 1 mM PMSF). Chromatin is sheared by Covaris Sonicator System to 200-400 bp fragments. Supernatants are collected after centrifugation and diluted at least 5 volumes in Dilution Buffer (1% (vol/vol) Triton X-100, 2 mM EDTA, 150 mM NaCl and 20 mM Tris-HCl, pH 8.1). Antibody (2 μg) is incubated with diluted chromatin overnight, after preclearing. Immunoprecipitation continues by incubation with protein G Dynal magnetic beads (Invitrogen) for at least 3 h at 4°C. Magnetic bead– immunoprecipitated chromatin complexes are then washed 8 times with High Salt Wash Buffer (2 times), Low Salt Wash Buffer (2 times), LiCl Wash Buffer (2 times) and TE Buffer (2 times). Immunoprecipitated chromatin is then eluted from Magnetic beads with Proteinase K Digestion Buffer and heated at 65 °C for at least 6 h for reversal of the formaldehyde crosslinking. DNA fragments are purified with NucleoMag beads kit (MN) and analyzed by SYBR Green Quantitative Real-time PCR. The following antibody was used for ChIP: anti-mouse T-bet (4B10). The following primer pair was used for ChIP: Forward 5′-GCTGTCTCATCGTCAGAGAG-3′, Reverse 5′-GATGGTGACAGATAGGTGGG-3′.

Targeted proteomics

The Skyline software was used for selecting nine proteotypic peptides for IL-33 along with the m/z values of the precursor ions and specific fragment ions. These nine precursor ions were selected for detection by PRM. The samples were processed with the GeLC-MS sample preparation protocol [59]. Specifically, cell pellets were homogenized in 100 μL of lysis buffer (7 M Urea, 2 M Thiourea, 4% w/v CHAPS, 1% w/v DTE) and the protein concentration was determined by the Bradford assay. Subsequently, 10 μg of protein per sample were loaded on SDS-PAGE wells and the electrophoresis was stopped when the samples entered the resolving gel. Gels were stained with Coomassie colloidal blue and a single band containing all proteins was excised. In gel reduction and alkylation was followed by trypsin digestion in 0.01 M ammonium bicarbonate pH 8.5. Peptides were extracted from the gel bands, dried in a vacuum centrifuge and re-suspended in 10 μL 0.1% v/v formic acid, pH 3.5. Analysis of the extracted peptides was performed with an UltiMate 3000 Nano HPLC Dionex Ultimate® 3000 RSLS system (Dionex™; Thermo Fisher Scientific, Inc.) coupled to a Q Exactive (Thermo Fisher Scientific, Inc) mass spectrometer operating in PRM mode. A volume of 5 μL was injected into the chromatography system and peptides were fractionated in a C18 column during a 70 min run (0.3 μL/min flow rate) using a gradient of up to 80% v/v Acetonitrile for elution. The MS2 parameters were: resolution 35,000 and injection time (IT) 128 ms. The PRM data files were analyzed using the Skyline software. All the acquired data were inspected and processed manually.

Data analysis and statistics

Results are presented as mean ± s.d. Data were analyzed by two-tailed Student’s t-test or two-tailed Mann–Whitney U-test, as appropriate (after testing for normality with the F-test). Multiple-group comparisons were performed using one-way analysis of variance (ANOVA) and the Dunnett’s or Tukey’s multiple comparison tests. Kaplan Meier statistics were evaluated by Log rank test. All statistical analyses were performed on GraphPad Prism 5 and 8.2.1 software. P values < 0.05 were considered to be statistically significant.

Ad1_noMX	5′AATGATACGGCGACCACCGAGATCTACACTCGTCGGCAGCGTCAGATGTG′3
Ad2.1_TAAGGCGA	5′CAAGCAGAAGACGGCATACGAGATTCGCCTTAGTCTCGTGGGCTCGGAGATGT′3
Ad2.3_AGGCAGAA	5′CAAGCAGAAGACGGCATACGAGATTTCTGCCTGTCTCGTGGGCTCGGAGATGT′3
Ad2.4_TCCTGAGC	5′CAAGCAGAAGACGGCATACGAGATGCTCAGGAGTCTCGTGGGCTCGGAGATGT′3
Ad2.8_CAGAGAGG	5′CAAGCAGAAGACGGCATACGAGATCCTCTCTGGTCTCGTGGGCTCGGAGATGT′3
Ad2.9_GCTACGCT	5′CAAGCAGAAGACGGCATACGAGATAGCGTAGCGTCTCGTGGGCTCGGAGATGT′3
Ad2.10_CGAGGCTG	5′CAAGCAGAAGACGGCATACGAGATCAGCCTCGGTCTCGTGGGCTCGGAGATGT′3
Ad2.13_GTCGTGAT	5′CAAGCAGAAGACGGCATACGAGATATCACGACGTCTCGTGGGCTCGGAGATGT′3
Ad2.14_ACCACTGT	5′CAAGCAGAAGACGGCATACGAGATACAGTGGTGTCTCGTGGGCTCGGAGATGT′3
Ad2.15_TGGATCTG	5′CAAGCAGAAGACGGCATACGAGATCAGATCCAGTCTCGTGGGCTCGGAGATGT′3