Tumor-Derived IL33 Promotes Tissue-Resident CD8⁺ T Cells and Is Required for Checkpoint Blockade Tumor Immunotherapy

Immune checkpoint blockade (ICB) therapy has produced unprecedented survival benefits for patients with cancer. The efficacy of ICB depends on adaptive antitumor immune responses, which are activated by a combination of tumor antigens and tumor-derived damage-associated molecular pattern (DAMP) molecules (1). High tumor mutation load increases the chance of generating immunogenic nonself-neoantigens, which can be recognized by the adaptive immune system (2). Increased tumor mutation load is associated with the improved survival provided by ICB therapy in multiple cancer types (3, 4). The “danger hypothesis” predicts that antitumor immune responses depend on immunostimulatory DAMP molecules, also called alarmins or danger signals, in addition to neoantigens (5). Alarmins stimulate dendritic cells (DC) and T cells and are involved in initiating antitumor immune responses. Yet the role of DAMP molecules in ICB tumor therapy is not well understood.

Tumor-resident T cells have been implicated in mediating tumor immune surveillance and immunotherapy (6). Ample studies have established that the number of resident CD8⁺ T cells in the tumor tissue correlates with better prognosis (7, 8). Tissue-resident T cells can be generated in the draining lymph node and migrate to the tissue. Although sharing similar T-cell receptor repertoires with effector and central memory T cells, tissue-resident T cells reside in the tissue and do not circulate into the blood. Resident T cells also express characteristic markers such as CD103, CD69, and CD49a (9, 10). In tumor tissues, it is thought that they interact intimately with epithelial tumor cells and can initiate various effector functions against target tumor cells. The tissue signals crucial for tissue residence of T cells are not well understood.

IL33 is a member of the IL1 gene family. IL33 protein is detected in the nuclei of epithelial cells in barrier tissues such as the skin, gastrointestinal tract, lungs, and endothelial cells of blood vessels (11). The nuclear localization of IL33 suggests that it has a role as an alarmin or danger signal upon damage of endothelial or epithelial cells (11). IL33 performs diverse biological functions by targeting various immune cells. The role of IL33 in type 2 immunity is established (12). IL33 enhances the function of Th1 and CD8⁺ T cells in vitro and mediates type 1 immunity during viral infection and chronic immune pathology (13–15). Strong antitumor effects can be produced when the active isoform of IL33 is expressed in tumor cells or the recombinant IL33 is administered exogenously (16, 17). The biological function of endogenous IL33 in tumorigenesis is quite complex because it can promote immune tolerance by activating regulatory T cells (Treg) and M2 while being a positive regulator of adaptive immune responses (18–21). However, the role of IL33 in ICB tumor immunotherapy has not been defined.

In this study, we set out to determine the role of IL33 in responsiveness to ICB tumor therapy. We examined IL33 expression in mouse tumor tissues after treatment with checkpoint inhibitors such as CTLA-4 and programmed death-1 (PD-1) mAbs. We also determined the significance of IL33 signaling in mediating ICB efficacy in murine tumor models. We clarified the role of tumor-derived IL33 in ICB tumor therapy by dissecting the underlying cellular mechanisms and IL33-driving immune responses, particularly involvement of the tissue adaptive immune system. Finally, we explored the feasibility of combination therapy with IL33 and ICB for the treatment of tumors.

C57BL/6J, BALB/cJ, B6.129S2(C)-Itgae^tm1Cmp/J (CD103-deficient), RAG1 KO, and B6.129S(C)-Batf3^tm1Kmm/J (Batf3-deficient) mice were purchased from The Jackson Laboratory. The ST2^−/− mice were provided by Dr. Andrew McKenzie (MRC Laboratory of Molecular Biology, Cambridge, UK). Il33^−/− mice have been described previously (22) and were obtained from RIKEN. All “knockout” mice were on the C57BL/6 background. Mice were housed in the Specific Pathogen Free facility of the University of Pittsburgh School of Medicine (Pittsburgh, Pennsylvania) or Soochow University (Suzhou, China). Experiments were done in accordance with a protocol approved by the institutional Animal Care and Use Committee and in accordance with NIH guidelines.

Tumor-infiltrating lymphocytes (TIL) were harvested from freshly resected tumor tissues according to the method we have described previously (23). In brief, the tumor tissues were dissected and transferred into RPMI culture medium. Tumor tissues were then mechanically disrupted and digested with a mixture of 0.3 mg/mL DNase I (Sigma-Aldrich) and 0.25 mg/mL Liberase TL (Roche) in the serum-free RPMI medium in a CO₂ culture incubator at 37°C for 30 minutes. The tissues were then dispersed through a 40-mm cell strainer (BD Biosciences) to remove tissue clumps. The single cells were washed and suspended in Hank's balanced salt solution (HBSS) with 1% FCS for staining and flow cytometry analysis.

MC-38 cells and Panc02 cells were kindly provided by Dr. Zongsheng Guo (University of Pittsburgh School of Medicine, Pittsburgh, Pennsylvania) in 2017. 4T1.2 mouse breast tumor cells were provided by Dr. Zhaoyang You (University of Pittsburgh School of Medicine, Pittsburgh, Pennsylvania) in 2017. The cell lines were cultured in RPMI1640 (B16F0) or DMEM (MC-38 and 4T1.2) media supplemented with 10% FBS in the presence of benzylpenicillin (100 U/mL), streptomycin (100 μg/mL), and 2 mmol/L l-glutamine. Construction of the B16-IL33 cell line has been reported previously (16). Cells were cultured for less than 2 weeks before injection into mice.

IL33^−/− MC38 cells were generated using the CRISPR/Cas9 system. Briefly, single-guide RNA (sgRNA) was designed using online CRISPR Design Tool (https://crispr.cos.uni-heidelberg.de) and cloned into plasmid lentiCRISPRv2GFP (Addgene, catalog no. 82416). The sgRNA sequences were designed to delete exon2 and exon3 of mouse IL33. They were 5′-CATTCTAGGTCTCATTTTTC-3′ for sgRNA1, and 5′-TACTGCATGAGACTCCGTTC-3′ for sgRNA2. The plasmids were transfected into MC38 cell line using Lipofectamine 2000 (Thermo Fisher Scientific, catalog no. 11668030). Transfected cells were sorted and single cell cloned, and mutant cells were identified using PCR and confirmed by Western blot analysis (Supplementary Fig. S1B–S1D). The genomic target sequences used for targeting screening were 5′-AGCCAAGGTTGCTTCTGATGA-3′ and 5′-TAGATGCCCATCAGTCTTTC-3′. IL33 expression by Western blot analysis used an IL33 antibody (R&D Systems, catalog no. AF3626).

The construction of IL33 expression plasmid has been described before (16). Briefly, the IL33 expression construct was generated by fusing the nucleotide sequence encoding the human CD8α signal sequence to the 5′ end of IL33 (S109–I266) sequence downstream the elongation factor alpha promoter. The detailed procedure of synthesis of POEG-st-Pmor polymer was described previously (24). For plasmid DNA complexation, POEG-st-Pmor was diluted to different concentrations in water and mixed with plasmid DNA solution to obtain the desired N/P ratios. Mice were treated intravenously with IL33 plasmid/POEG-st-Pmor nanoparticles at a dose of 30 μg DNA/mouse in 200 μL 5% dextrose.

For the MC-38 tumor model, MC-38 cells (1 × 10⁶) were injected intradermally into the right flank of C57BL/6 mice. CTLA-4 mAb (clone 9D9, catalog no. BE0164, BioXcell)– and PD-1 mAb (clone j43, catalog no. BP0033–2, BioXcell)–based ICB therapies were administered on the 4th day after the tumor inoculation, when tumor diameters had reached approximately 4 mm). A total of 200 μg antibodies were intraperitoneally injected four times with 4-day intervals. Hamster IgG (catalog no. BE0091) and mouse IgG2b (catalog no. BE0086, BioXcell) were used as controls. Tumor size was monitored by a caliber and recorded every 2 days after tumor inoculation.

For the B16 tumor model, the B16-Vec and B16-IL33 cells (1 × 10⁵ in 100 μL PBS) were injected intradermally into the right flank of C57BL/6j mice. CTLA-4 mAbs (200 μg) and IgG2b control (200 μg) were intraperitoneally injected starting on the 4th day after tumor inoculation for a total of four injections separated by 4-day intervals. Tumor size was recorded every 2 days.

For the 4T1.2 tumor model, 4T1.2 cells (2 × 10⁵ in 100 μL PBS) were intravenously injected via the tail vein of BALB/c mice. The CTLA-4 mAbs (200 μg) in combination with the IL33 plasmid-nanoparticles were injected intravenously on the 5th day after tumor inoculation. These injections were repeated three times at intervals of 4 days. Mouse IgG2b (200 μg) and control plasmid-nanoparticles were used as controls.

B16-IL33 cells were injected intradermally into the right flank of RAG1^−/− mice to establish the B16-IL33 tumor model. The CD8⁺ T cells from ST2^−/−, CD103^−/−, and wild-type (WT) CD57BL/6 mice were purified using magnetic bead–based methods (catalog no. 130–116–478, Miltenyi Biotec). The purified CD8⁺ T cells (around 10 million in 100 μL PBS) were intravenously injected into the retro-orbital venous sinus of RAG1^−/− mice. Tumor cells were inoculated on the same day. Tumor size was monitored and recorded every 2 days.

Expression of the IL33 mRNA in MC-38 cells under different stimulation conditions were assessed using RT-PCR. The total RNA was extracted from MC-38 cells using TRIzol reagent (Invitrogen) and reverse transcribed using Superscript II (Invitrogen). RT-PCR reactions were done using the StepOnePlus Real-Time PCR System (Applied Biosystems) according to the manufacturer's instructions. Primers were synthesized as follows: GAPDH 5′-CCTGCACCACCAACTGCTT-3′ and 5′-ATGACCTTGCCCACAGCCT-3′; IL33 5′-TATCCACGGGATTCTAGGAAGAG-3′ and 5′-CTCATAGTAGCGTAGTAGCACCT-3′. Relative amounts of mRNA were calculated using the 2^–ΔΔC_t method.

Tumor samples were resected and embedded in disposable vinyl specimen molds (Tissue-Tek Cryomold, Sakura) with optimal cutting temperature compound (catalog no. 23730571, Tissue-Tek, Sakura) on dry ice. For immunofluorescence staining of CD8 and CD103, 7-μm cryostat sections were prepared using a microtome (Leica CM1950), then air dried for 10 minutes and fixed in 4% paraformaldehyde fix solution for 10 minutes at room temperature. The sections were incubated in PBS with 3% FBS, then stained with Alexa Fluor 647 anti-mouse CD8α (Clone 53–6.7, Biolegend) and Alexa Fluor 594 anti-mouse CD103 (Clone 2E7, Biolegend) for 1 hour at room temperature. The sections were then washed twice with PBS and counterstained with 4′,6-diamidino-2-phenylindole for 30 seconds. For immunofluorescence staining of IL33, the 7-μm cryostat sections were air dried for 10 minutes, fixed in 4% paraformaldehyde for 10 minutes at room temperature, permeabilized using 0.5% Triton for 10 minutes at room temperature, washed twice with PBS. They were then stained with goat anti-mouse IL33 (R&D, AF3626), washed twice with PBS, and stained with Alexa Fluor 594 donkey anti-goat secondary (catalog no. A32758, Invitrogen). For the multicolor staining assay, APC-CD45 (Clone 30-F11, BD Pharmingen), FITC-CD11b (Clone M1/70, BD Pharmingen), FITC-B220 (Clone RA3–6B2, BD Pharmingen), FITC-CD3 (Clone 145–2C11, BD Pharmingen), CD31 (EPR17260–263, catalog no. ab222783, Abcam), fibroblast-specific protein 1(FSP1) (Rabbit Polyclonal IgG, Clone S100A4, catalog no. 810101, Biolegend), and Alexa Fluor 647 donkey anti-rabbit secondaries (catalog no. A-31573, Invitrogen) were used.

The densities of tumor-infiltrating CD8⁺ T cells and CD103⁺CD8⁺ T cells in tumor tissues were evaluated as we have described previously (24). In brief, pictures of five randomly chosen high-power fields (×200) were captured and collected using an OX83 Olympus microscope, after which the numbers of tumor-infiltrating CD8⁺ T cells and CD103⁺CD8⁺ T cells in each picture were calculated and recorded. For the assessment of IL33⁺MC-38 tumor cells in tumor sections, pictures of five random high-power fields (×100) were captured and collected using the OX83 Olympus microscope. The numbers of IL33⁺MC-38 cells were then calculated, recorded, and statistically analyzed.

Flow cytometry analysis was performed using a FACS flow cytometer Fortessa (BD Biosciences). The data were analyzed using Flowjo software. CD45 (Clone 30-F11), CD4 (Clone GK1.5), CD8 (Clone 53–6.7), CD103 (Clone M290), and CD69 (Clone H1.2F3) were purchased from BD Bioscience. PD-1 (Clone 29F.1A12), Granzyme B (Clone GB11), and Foxp3 (Clone MF-14) were purchased from BioLegend. Ki-67 (Clone SolA15) and IFNγ (Clone XMG1.2) were purchased from eBioscience (Thermo Fisher Scientific). For intracellular cytokine staining, harvested cells were stimulated with phorbol myristate acetate (10 ng/mL; catalog no. p1585–1G, Sigma) and ionomycin (1 μg/mL; I9657–1G, Sigma) for 4 hours and incubated for the last 3 hours with brefeldin A (10 μg/mL; catalog no. 00-4506-51, Thermo Fisher Scientific). The cells were transferred to a V-bottom plate, stained with surface marker antibodies in HBSS containing 1% FCS, fixed with 2% formaldehyde, and permeabilized with 0.5% saponin. The cells were stained with anti-IFNγ and examined by flow cytometry.

IHC staining was performed as described in our previous studies (25). Briefly, the paraffin-embedded tissue-array blocks were cut into 3-mm-thick consecutive sections, and were dewaxed in xylene, rehydrated and graded ethanol solutions. Antigen retrieval was performed by heating the tissue sections at 100°C for 30 minutes in EDTA solution (pH 9.0). The sections were incubated with mouse IL33 antibody (Abcam, ab229698) at 4°C overnight, followed by incubation with horseradish peroxidase–conjugated secondary antibody. Diaminobenzene was used as the chromogen, and hematoxylin was used as the nuclear counterstained. Finally, the sections were dehydrated, cleared, and mounted.

GraphPad Prism 6.0 software package (GraphPad Software, Inc.) was used for the analysis. Statistical analysis was done using the Student t test (two-tailed), two-way ANOVA, or the log-rank survival analysis. A P value of <0.05 was considered to be statistically significant. No statistical correction was used.

The role of IL33 in tumorigenesis and antitumor immunotherapy is complex. We determined whether IL33 signaling in recipient mice was important for mediating the antitumor effect during ICB tumor therapy. The MC-38 tumor model was chosen due to responsiveness to ICB therapy (26). To this end, we inoculated WT and ST2^−/− mice with MC-38 colon tumor cells. Upon tumor formation, we started treatment with CTLA-4 mAbs. MC-38 cells formed tumors, which, given treatment with control IgG, grew at similar rates in WT and ST2^−/− mice. Administration of CTLA-4 mAbs inhibited tumor growth in WT mice (Fig. 1A; Supplementary Fig. S2A). In contrast, administration of CTLA-4 mAbs did not inhibit tumor growth in ST2^−/− mice (Fig. 1A; Supplementary Fig. S2A). This indicated that ST2 signaling in host cells was critical for the antitumor effect of CTLA-4 mAbs.

We also determined whether the IL33/ST2 axis was important for the therapeutic efficacy of the PD-1 blockade. Administration of PD-1 mAbs inhibited tumor growth in WT mice (Fig. 1B; Supplementary Fig. S2B). In contrast, PD-1 mAbs did not inhibit tumor growth in ST2^−/− mice (Fig. 1B; Supplementary Fig. S2B). Besides MC-38, we showed that IL33 signaling was required for the ICB treatment in a murine pancreatic cancer model Panc02 (Supplementary Fig. S2C) and a murine ovarian tumor model ID8 (Supplementary Fig. S2D). Consistently, ST2 was also specifically upregulated in tumor tissues from patients with melanoma who responded to PD-1 mAb ICB therapy (Supplementary Fig. S2E; ref. 27). These results indicated that ST2 signaling in host cells was critical for responsiveness to ICB therapy.

To further investigate the underlying cellular immune mechanisms, we performed multicolor flow cytometry and immunofluorescence analysis, finding that treatment with CTLA-4 mAbs resulted in increases in CD45⁺ immune cells in tumors from both WT and ST2^−/− mice (Fig. 1C; Supplementary Fig. S3A). However, the increases were much higher in WT than ST2^−/− mice. Within the immune cell compartment, we detected no differences between the experimental groups with regard to the percentages of CD4⁺ and CD8⁺ T cells (Supplementary Fig. S3A–S3C). However, we observed substantial decreases in tumoral Tregs after CTLA-4 mAb treatment of WT mice (Supplementary Fig. S3D–S3E). This is consistent with a plethora of studies demonstrating the Treg-depleting activities of CTLA-4 mAbs in both mice and humans (28, 29). It is worth noting that the percentage of tumoral Tregs was much less in ST2^−/− mice and that CTLA-4 mAb treatment did not further reduce tumoral Tregs in ST2^−/− mice (Supplementary Fig. S3D–S3E). These findings are consistent with the hypothesis that the IL33/ST2 axis also promotes Tregs in tumors. Therefore, a main effect of CTLA-4 mAbs was the elimination of IL33-induced Tregs in the tumor microenvironment (TME). As a result, the full brunt of antitumor effect of IL33 was unleashed.

We next examined IFNγ production by tumoral CD4⁺ and CD8⁺ T cells. The frequency of IFNγ⁺CD4⁺ and IFNγ⁺CD8⁺ tumoral T cells was similar in WT and ST2^−/− mice in the control IgG treatment groups, suggesting similar spontaneous antitumor immune responses in WT and ST2^−/− mice. We observed an increase in IFNγ⁺CD4⁺ and IFNγ⁺CD8⁺ tumoral T cells upon CTLA-4 mAb treatment of WT mice (Fig. 1D–F). In contrast, CTLA-4 mAb treatment did not result in any increase of IFNγ⁺CD4⁺ and IFNγ⁺CD8⁺ TIL in ST2^−/− mice (Fig. 1D–F).

CD103 is the best characterized marker for intraepithelial lymphocytes and tissue-resident T cells (30). CD103 expressed on CD8⁺ T cells is important during immune responses against cancer, particularly cancers of epithelial origin (31). Because IL33 is a tissue alarmin, we decided to examine whether the CD103⁺ subset of TIL is regulated by IL33/ST2 signaling. Flow cytometric analysis demonstrated that treatment of tumor-bearing mice with CTLA-4 mAbs increased the percentage of CD103⁺CD8⁺ TIL in WT mice, but not in ST2^−/− mice (Fig. 1D and G). This was confirmed by immunofluorescent analysis of tumor sections (Fig. 1I and J). In addition, IFNγ⁺CD103⁺CD8⁺ TIL in WT mice, but not ST2^−/− mice, were increased by CTLA-4 mAb treatment (Fig. 1D and H). This indicated that IL33, which enhances the number and function Th1, CD8⁺, and CD103⁺CD8⁺ TIL, was required for the antitumor effect of CTLA-4 mAb treatment.

We next characterized TIL in the PD-1 mAb–treated mice by multicolor flow cytometric analysis and immunofluorescence microscopy. Using flow cytometry, we showed that administration of PD-1 mAbs led to an increase in the percentage of CD45⁺ immune cells in both WT and ST2^−/− mice (Supplementary Fig. S3F). There was no difference in the frequencies of CD4⁺ and CD8⁺ TIL in various experimental groups (Supplementary Fig. S3G and S3H). In contrast, we observed a considerable decrease in tumoral Tregs in WT mice after PD-1 mAb treatment (Supplementary Fig. S3I and S3J). This was consistent with previous findings that PD-1 is involved in the generation and maintenance of induced Tregs, as well as in homeostasis of natural Tregs (32).

We then examined IFNγ production by CD4⁺ and CD8⁺ TIL. Administration of PD-1 mAbs led to large increases in the frequency of IFNγ⁺CD4⁺ TIL in WT mice but not ST2^−/− mice (Supplementary Fig. S3K and S3I). Treatment with PD-1 mAbs also resulted in large increases in the frequency of IFNγ⁺CD8⁺ TIL in WT but not ST2^−/− mice (Supplementary Fig. S3K and S3M). PD-1 mAb treatment led to an increase in the percentage of CD103⁺CD8⁺ TIL in WT mice but not ST2^−/− mice (Supplementary Fig. S3K and S3N). These findings were confirmed by immunofluorescent microscopic analysis of tumor sections (Supplementary Fig. S3P and S3Q). IFNγ⁺CD103⁺CD8⁺ TIL were increased following PD-1 mAb treatment of WT mice, but not ST2^−/− mice (Supplementary Fig. S3K and S3O). These results indicated that IL33 mediates PD-1 immunotherapy by increasing the number and function of Th1 and CD103⁺CD8⁺ TIL.

Because IL33 signaling is required for ICB tumor therapy, we determined the cellular source of IL33. IL33 is expressed in normal epithelial cells and tumor cells but its amount is reduced in tumor cells at the most advanced stage of cancer (25, 33). These clinical findings suggest that cancer cell–derived IL33 may be involved in antitumor immune responses. We examined the expression of IL33 protein in tumor tissues in the MC-38 model during CTLA-4 or PD-1 ICB treatment. Using immunofluorescence microscopy, we found that IL33 protein was expressed in small numbers of tumor cells in mice treated with control IgG (Fig. 2A–D). Upon treatment with CTLA-4 or PD-1 mAbs, the numbers of IL33⁺ cells were greatly increased in tumor tissues (Fig. 2A–D). The morphology of these cells suggested that IL33 was expressed mainly in tumor cells (Fig. 2A–D; Supplementary Fig. S4A). Additional immunostaining with CD45, CD11b, CD3, CD31, and FSP1 confirmed that the dominant population of IL33⁺ cells within tumor tissues was not immune cells or stromal cells (Supplementary Fig. S4B). Tumor cells located in the invasion border, as well as those deep inside tumor nests, expressed IL33. In contrast to these findings in tumors isolated from WT mice, IL33 protein was not induced in tumor tissues from ST2^−/− mice regardless of which checkpoint inhibitors were used (Fig. 2A–D). This suggests that IL33 induced its own expression in vivo, possibly through host immune and stromal cells. These results indicated that IL33 could be induced in mice after ICB therapy.

To substantiate the cellular sources of IL33 that mediates ICB therapy, we generated IL33-deficient cell lines MC38^IL33−/− using the Crispr/Cas9 technology and showed IL33 protein is not expressed in MC38^IL33−/− cells (Supplementary Fig. S1B–S1D). Neither PD-1 nor CTLA-4 mAb administration inhibited the growth MC38^IL33−/− tumors (Fig. 2E; Supplementary Fig. S1E). We examined whether IL33 produced by host mice was responsible for mediating the antitumor effect of PD-1 and CTLA-4 mAbs. We carried out PD-1 and CTLA-4 mAb treatment of MC-38 tumors in IL33^−/− and WT control mice. We found that administration of PD-1 or CTLA-4 mAbs inhibited tumor growth in both WT and IL33^−/− mice with equivalent efficacies (Fig. 2G; Supplementary Fig. S1F). This again confirmed that tumor-derived IL33 and ST2 signaling in host cells were critical for the antitumor effect of ICB tumor therapy.

We dissected the cellular basis of the reduced antitumor efficacy in MC38^IL33−/− tumors. We found that the percentage of tumor-infiltrating immune cells (CD45⁺) was drastically reduced in MC38^IL33−/− tumors compared with control tumors (Fig. 2F). In addition, we found that the percentage of IFNγ⁺ CD4⁺ and CD8⁺ TIL was also reduced in MC38^IL33−/− (Fig. 2H–J). We also found the percentage of tissue resident CD103⁺CD8⁺ TIL and IFNγ⁺CD103⁺CD8⁺ TIL was reduced in MC38^IL33−/− tumors (Fig. 2H, K, and L). These data suggested that tumor-derived IL33 is crucial for the tumor-resident antitumor immune responses.

Onto dissect the cellular immune mechanism underlying the antitumor function of tumor-derived IL33, we characterized TIL in B16 tumors, which do not express IL33, and B16-IL33 tumors (16), which express and secrete an active form of IL33 (Fig. 3A; Supplementary Fig. S1A). The percentage of CD45⁺ immune cells was increased in B16-IL33 tumors as compared with B16 tumors (Fig. 3B). In addition, we observed considerable increases in CD4⁺ TIL (Fig. 3C), Tregs, and CD103⁺ Tregs (Fig. 3D) in B16-IL33 compared with B16 tumors. Despite increases in Tregs, the percentage of tumoral Th1 cells was increased in B16-IL33 compared with B16 tumors (Fig. 3H; Supplementary Fig. S5A). Flow cytometric analysis showed no differences in the percentages of CD8⁺ TIL (Fig. 3E). In contrast, the percentages of CD103⁺CD8⁺ TIL were greatly increased in B16-IL33 tumors as compared to B16 tumors (Fig. 3F and G). In addition, we found increases in proliferation (Ki67), activation (CD69), and IFNγ production in total CD8⁺ TIL as well as CD103⁺CD8⁺ TIL in B16-IL33 tumors (Fig. 3G–K; Supplementary Fig. S5B–S5D). The increases in the numbers of CD103⁺CD8⁺ TIL were confirmed using immunofluorescent microscopy (Fig. 3L–O). Our findings indicated that tumor-derived active IL33 increased the number and function of tumoral Th1 cells, CD8⁺ T cells, and CD103⁺CD8⁺ T cells.

ST2 can be induced on CD8⁺ T cells (14), we set out to determine whether ST2 expression on CD8⁺ T cells was required for the antitumor effect of IL33 in vivo. CD8⁺ T cells were isolated from WT and ST2^−/− mice and adoptively transferred to Rag1^−/− mice. B16-IL33 cells were then inoculated intradermally into the recipient Rag1^−/− mice. Rag1^−/− infused with ST2-deficient CD8⁺ T cells had increased tumor growth (Fig. 4A), confirming that ST2 signaling on CD8⁺ T cells was required for the antitumor effect of IL33. Flow cytometric analysis showed a decrease in CD45⁺ TIL in mice infused with ST2^−/− CD8⁺ T cells (Fig. 4B). ST2 deficiency in CD8⁺ T cells resulted in a pronounced decrease in the percentage of CD103⁺CD8⁺, but not total CD8⁺ TIL (Fig. 4C–E), and the production of IFNγ was reduced in both CD103⁺ and CD103⁻CD8⁺ TIL as the result of ST2 deficiency in CD8⁺ T cells (Supplementary Fig. S6A–S6C). Using immunofluorescent microscopy, we confirmed that the densities of both CD103⁺CD8⁺ TIL and total CD8⁺ TIL in tumor tissues were much lower in mice infused with ST2^−/− CD8⁺ T cells than in those with WT CD8⁺ T cells (Fig. 4F–I). These findings indicated that ST2 expression on CD8⁺ T cells was required for the IL33-mediated antitumor effect, as well as the accumulation of CD103⁺CD8⁺ T cells in tumor tissues.

The ligand for CD103 is the epithelial cell surface molecule E-cadherin (34). Specific interactions between CD103 and E-cadherin are responsible for retention of antigen-specific lymphocytes within epithelial tissues (35). Because we showed that IL33 increased the function and number of CD103⁺CD8⁺ TIL, we determined whether CD103 expression on CD8⁺ T cells was required for the antitumor effect of IL33. CD8⁺ T cells were isolated from WT and CD103^−/− mice and adoptively transferred to Rag1^−/−mice. B16-IL33 cells were then inoculated intradermally CD103 deficiency in CD8⁺ T cells led to an increased rate of tumor growth (Fig. 5A), indicating that CD103 was required for the antitumor effect of IL33. Flow cytometric analysis showed a decrease in CD45⁺ TIL in mice infused with CD103^−/−CD8⁺ T cells as compared with those infused with the WT control (Fig. 5B). Although the percentage of CD8⁺ TIL was not changed, CD103 deficiency resulted in decreased proliferation (by Ki67 staining), activation (CD69), and IFNγ production in CD8⁺ TIL (Fig. 5C and D; Supplementary Fig. S6D–S6F). Using immunofluorescence microscopy, the density of CD103⁺CD8⁺ and total CD8⁺ TIL was lower in mice infused with CD103^−/−CD8⁺ T cells than it was in controls (Fig. 5E–H). These results indicated that CD103 expression on CD8⁺ T cells was required for the IL33-driven antitumor effect and tumoral accumulation of CD8⁺ TIL.

CD103 is also a marker for the type-1 DC, which can cross present antigens to CD8⁺ T cells in the TME (36). The percentage of CD103⁺ DC was increased in B16-IL33 tumors compared with B16 tumors and decreased in MC38^IL33−/− compared with MC38 tumors (Fig. 6A and B; Supplementary Fig. S7A–S7C). To determine the role of CD103⁺ DC in mediating the antitumor function of IL33, we used Batf3^−/− mice, which had a selected deficiency in generating CD103⁺ DC (36, 37). Our data showed that the antitumor effect of IL33 was abrogated in Batf3^−/− mice (Fig. 6C). At the cellular level, the immune infiltration was severely reduced in B16-IL33 tumor from Batf3^−/− mice compared with control mice (Fig. 6D). Consistent with the role of CD103⁺ DC in driving CD8⁺ T-cell immune responses, the percentage of CD8⁺ TIL and IFNγ⁺CD8⁺ TIL was also reduced in B16-IL33 tumors from Batf3^−/− mice (Fig. 6E and F; Supplementary Fig. S7D). Although there was an increase in the percentage of CD4⁺ TIL, a profound decrease in Th1 cells was found in B16-IL33 tumors from Batf3^−/− mice when compared with WT control mice (Fig. 6G and H; Supplementary Fig. S7D). These data demonstrated that CD103⁺ DC were crucial for mediating the antitumor effect of IL33.

Although potent antitumor efficacy can be achieved by tumoral expression of the secretory IL33, the number of tumor-infiltrating Foxp3⁺Treg cells was also increased. Because one main mechanism of action for CTLA-4 mAbs is deletion of tumoral Tregs (28, 29), we set out to determine whether IL33 combined with CTLA-4 mAbs can act as a novel tumor immunotherapy. Mice were inoculated with B16 or B16-IL33 cells, then treated with control IgG or CTLA-4 mAbs (Fig. 7A). As expected, B16-IL33 tumors grew much more slowly than control B16 tumors. Administration of CTLA-4 mAbs did not affect the progression of B16 tumors, but greatly inhibited the growth rate of B16-IL33 tumors and prolonged survival of B16-IL33 tumor-bearing mice (Fig. 7B and C). Thus, IL33 and CTLA-4 mAbs have synergistic antitumor effects.

On the basis of the rationale that there was synergy between depletion of Treg cells and IL33, we further investigated the antitumor effect of combination of CTLA-4 mAbs and IL33 in a therapeutic setting. To this end, we delivered the IL33-expressing plasmid using a nanoparticle-based approach. Plasmid-loaded POEG-st-Pmor nanoparticles are stable in the blood and are highly effective in selective delivery of gene-expressing constructs to the lung tumor tissues (38). Mice were inoculated intravenously with 4T1.2 tumor cells, then treated with control IgG, the IL33 nanoparticle, CTLA-4 mAbs, and the IL33 nanoparticle plus CTLA-4 mAbs (Fig. 7D). All mice treated with the control plasmid nanoparticles plus control IgG died within 22 days after tumor cell injection (Fig. 7E). Mice treated with CTLA-4 mAbs alone or the IL33 nanoparticle alone did not show improvement in survival compared with the control group. However, the long-term survival rate of mice treated with IL33 nanoparticles plus CTLA-4 mAbs was significantly higher than that of the control group or mice treated with either IL33 nanoparticles or CTLA-4 mAbs (Fig. 7E). Together, these results demonstrated that a combination of targeted expression of IL33 protein in the tumor site and CTLA-4 mAbs was an effective immunotherapy for tumors. The underlying mechanism is likely due to depletion of Treg cells by CTLA-4 mAbs (28, 29).

As many T cells in B16-IL33 tumors expressed PD-1 (Supplementary Fig. S8), we then tried to determine whether combining IL33 and PD-1 blockade could further increase antitumor efficacy (Fig. 7F). Indeed, administration of PD-1 mAbs significantly inhibited growth of B16-IL33 tumors (Fig. 7G). Accordingly, PD-1 mAbs also prolonged the survival of B16-IL33 tumor-bearing mice (Fig. 7H). These data indicated that IL33 and PD-1 blockade could be combined to synergistically enhance antitumor efficacy.

Although ICB has improved the survival of countless patients with cancer, the majority of patients with cancer do not respond to this therapy. Identification of the specific molecular mechanisms that mediate responsiveness to ICB treatment will help improve responses to ICB therapies and facilitate the development of new cancer therapies. In this study, we demonstrated that IL33 was expressed in immunogenic murine tumor cells and could be further induced during immune checkpoint tumor therapy. ST2 signaling was required for both CTLA-4– and/or PD-1–based tumor immunotherapy and the tumor, but not host, cell-derived IL33 was responsible for the antitumor effect. Mechanistically, IL33 induced IFNγ production by Th1, CD8⁺, CD103⁺CD8⁺ TIL, and CD103⁺ DC in the TME. Tumor-resident CD103⁺CD8⁺ T cells and CD103⁺ DC were critical for the antitumor efficacy of IL33 and demonstrated that IL33 synergized with CTLA-4 or PD-1 mAbs to increase antitumor efficacy. Our study establishes a critical role of the “danger signal” IL33 in mediating responsiveness to ICB therapy through promoting tumor-resident T cells, thus shedding light on how to harness IL33 to improve ICB tumor immunotherapy. IL33 promotes tissue resident T cells in vitro (39), and our study confirms this in vivo.

There is an antitumor role of IL33 (33). IL33 potently increases the effector function of CD8⁺ T cells (14). Including IL33 in a cancer vaccine regimen boosts its potency (40, 41). The tumoral expression of an active form of IL33 leads to striking tumor inhibition in multiple murine tumor models (16). Administration of a recombinant IL33 protein inhibits the growth of large, established tumors (17). The antitumor function of IL33 in these models is dependent on CD8⁺ T cells. In addition to CD8⁺ T cells, the antitumor function of IL33 can also be mediated by eosinophils (42). Silencing ST2 in CT26, a colon cancer cell line, results in increased tumor growth in vivo, implying that IL33 signaling has an inhibitory role in tumor cells (43). In one colorectal cancer model, IL33 was shown to decrease dextran sulfate sodium–induced colon tumors. The antitumor function of IL33 was attributed to its ability to restrain IL1α-dependent colitis (44).This study not only reinforces the idea that the IL33/ST2/CD8⁺ T-cell axis mediates antitumor immune responses, but also further illustrates the significance of this pathway in ICB therapy.

In addition to its antitumor activity, IL33 promotes oncogenesis in some experimental systems. Administration of low amounts of IL33 promotes the accumulation of immune suppressive cells such as myeloid-derived suppressor cell (MDSC) and Tregs, thereby suppressing active antitumor immune responses (33). In the APC^Min/+ mouse intestine tumor model, a modest but statistically significant reduction in the number and size of polyps in small intestines was observed in IL33^−/− APC^Min/+ mice as compared with WT APC^Min/+ mice (45–50). The protumor effect of IL33 can be attributed to several mechanisms, as follows. First, the APC mutation leads to activation of the WNT pathway, which inhibits CD8⁺ T-cell recruitment into the TME (51). Thus, the APC mutation inhibits the antitumor function of IL33 and tips the balance toward a protumor function. Second, IL33 promotes the function of mast cells, which are crucial for polyp formation in this model. Third, IL33 can also promote Tregs and M2 in the APC^Min/+ model, thereby promoting tumorigenesis (50). Fourth, full-length IL33, which is expressed in cell nuclei, has potential oncogenic functions, as has been demonstrated with transplant tumor models (33). Collectively, in the absence of tumor antigen–specific type 1 T cells, the protumor role of IL33 becomes predominant and is mediated by Tregs, M2, mast cells, MDSC, and nuclear IL33.

Tumor-resident CD103⁺CD8⁺ T cells might play an important role in antitumor immunity, in particular, against cancers of epithelial origin (31, 52–55). Local signals that favor the recruitment and activation of resident T cells are important. This concept has been proven in vivo by the “prime and pull” strategy (56). Nonetheless, understanding of tissue “pull” signals is lacking. Although IL33 promotes tissue-resident T cells in vitro (39), it has not been demonstrated in vivo. In this study, we demonstrated that IL33 can serve as this local signal. The effect of IL33 is two-fold. First, IL33 can directly activate CD103⁺CD8⁺ T cells. This idea is consistent with our data and other published studies (14). Second, IL33 can enhance the accumulation of tumoral CD103⁺ DC. Our data showed that IL33 increased the number of tumor-associated CD103⁺ DC. The generation of tissue-resident CD8⁺ T cells is dependent on CD103⁺ DC in nonlymphoid tissues and CD8α⁺ DC in lymphoid organs (57). Here, CD103⁺ DC were crucial to the antitumor efficacy of IL33. Our study indicates that IL33 is alocal signal for the establishment of resident CD8⁺ T-cell immune responses against tumor.

IL33 is a potent cytokine that induces both strong CD8⁺ T-cell immune responses and self-limiting Treg-mediated immune regulation. On the basis of IL33 biology, an effective immune therapy will entail tumoral delivery of an active form of IL33 and simultaneous depletion of protumor cells. We have demonstrated that such a strategy is clinically feasible by combining CTLA-4 mAbs, which eliminate tumoral Tregs, and gene delivery of IL33 by nanoparticles. These studies warrant clinical trials of this novel combination therapy for cancer treatment.

B. Lu reports grants from NIH during the conduct of the study, as well as other from Anwita Biosciences Inc. (member of board of directors and scientific advisory board) outside the submitted work. No potential conflicts of interest were disclosed by the other authors.

L. Chen: Conceptualization, resources, formal analysis, methodology, writing–original draft. R. Sun: Formal analysis, methodology, writing–original draft. J. Xu: Data curation, software, methodology. W. Zhai: Formal analysis, methodology. D. Zhang: Data curation, software, visualization, methodology. M. Yang: Methodology. C. Yue: Validation, methodology. Y. Chen: Investigation, methodology. S. Li: Resources, validation. H. Turnquist: Resources, investigation. J. Jiang: Resources, investigation. B. Lu: Conceptualization, resources, formal analysis, supervision, investigation, methodology, writing–review and editing.

This work was supported in part by R21CA2165743 to B. Lu and R01CA239716 to B. Lu and S. Li. L. Chen was supported by the AAI (American Association of Immunologists) Careers in Immunology Fellowship.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.