Discovery of a Metastatic Immune Escape Mechanism Initiated by the Loss of Expression of the Tumour Biomarker Interleukin-33

A new paradigm for understanding immune-surveillance and immune escape in cancer is described here. Metastatic carcinomas express reduced levels of IL-33 and diminished levels of antigen processing machinery (APM), compared to syngeneic primary tumours. Complementation of IL-33 expression in metastatic tumours upregulates APM expression and functionality of major histocompatibility complex (MHC)-molecules, resulting in reduced tumour growth rates and a lower frequency of circulating tumour cells. Parallel studies in humans demonstrate that low tumour expression of IL-33 is an immune biomarker associated with recurrent prostate and kidney renal clear cell carcinomas. Thus, IL-33 has a significant role in cancer immune-surveillance against primary tumours, which is lost during the metastatic transition that actuates immune escape in cancer.


IL-33 expression is decreased in metastatic carcinomas.
To address the metastatic gene signature network, we conducted a comparative microarray analysis on antecedent non-metastatic and metastatic cell lines of both murine lung and prostate cancers, using the Agilent Two Color Microarray Technology (a 28005 Two-Color Agilent microarray with a total of 55821 probes). Analysis of the aberrant expression of genes in the metastatic/non-metastatic cell lines provided an avenue to begin to characterize and understand the mechanism of transition from immunologically-recognized to immunoevasive tumours. To validate the mouse data, expression of selected genes within human tumours was taken into account, using the data obtained from the Gene Expression Atlas created by the European Biostatistics Institute (Gene Expression Atlas; https://www.ebi.ac.uk/ gxa/home; access date: 30/01/2013). From these, we focused on genes known to be involved in novel aspects of inflammation or immunity and those that may interact with the antigen presentation machinery.
While many genes were found to be downregulated in metastatic cancer as compared to primary cancer, one gene in particular, IL-33, was selected for further study since it is a key inducer of helper T cells and innate immune cells. According to our microarray data, IL-33 is down-regulated in the metastatic murine prostate and lung carcinomas (data not shown; unpublished). The IL-33 gene was confirmed to be down-regulated by 3-6 fold by separate qRT-PCR from two sets of primary and metastatic murine cell lines (Lung TC1/A9; Prostate PA/LMD) grown in vitro. The gene expression level in the data was in concordance with its protein expression level, indicating that it may be a good predictor for metastatic potential in carcinomas. In support of this hypothesis, it was also found to be significantly downregulated in metastatic human prostate carcinomas found in the Gene Expression Atlas (Fig. 1).

MHC-I expression is tied to the level of IL-33.
To address the role of IL-33 on the expression of MHC-I within the tumour, IL-33 gene expression was inhibited using siRNA on primary tumour cells (TC1). After siRNA treatment for 72 to 96 h, both the IL-33 and MHC-I surface expression levels were assessed. The down-regulation of IL-33 protein was confirmed by ELISA (for secreted  and Western blot analysis (for cell-associated IL-33) (Fig. 2a). Primary lung carcinoma cells, TC1, treated with IL-33-directed siRNAs had a much lower abundance of MHC-I complexes than did controls (Fig. 2c), whereas the overexpression of IL-33 in metastatic cells (A9) (Fig. 2b) had a higher level of MHC-I production compared to untransfected controls (Fig. 2d). This suggested that IL-33 may be important for the expression of the immune recognizable phenotype of the primary TC1 cell tumour.
To evaluate the correlation between IL-33 and MHC/HLA expression (and consequent T cell infiltration) in human tumors, we used RNA-sequencing data collected from resected prostate tumours. In collaboration with Vancouver Prostate Centre, we assessed the expression of Human leukocyte antigen (HLA) and IL-33 in primary treated (high risk neo-hormone-treated (NHT)) and metastatic treated castration-resistant prostate cancer (CRPC) samples. We found that HLA-A, HLA-B, HLA-C genes have a similar pattern of reduction in CRPC and elevation in high risk NHT primary tumours (Fig. 2e,f). This also appears to parallel the IL-33 pattern and is a trend for TAP1, CD4 and CD8 genes as well. These data also suggested that IL-33 might be important for the expression of an immune-recognizable phenotype of primary tumours. In human prostate cancer, MHC-I/HLA and IL-33 are co-ordinately down-regulated during the metastatic re-programming of primary tumours revealing the involvement of the IL-33 gene in tumorigenesis and immune-escape. IL-33 expression reverses the lack of MHC-I, antigen processing and immune recognition of metastatic tumours. Down-regulation of the MHC-I expression has been shown to be a primary indicator for rapid tumour progression and metastases in human prostate cancer 34 stomach, kidney, lung, prostate and other cancers) has highlighted loss of antigen presentation with mutations in beta-2-microglobulin, HLA-A, HLA-B and HLA-C, supports our earlier findings in lung carcinoma [37][38][39] , as a key strategy in tumour immune evasiveness 40 . Therefore, we next investigate the effects of IL-33 on MHC-I expression and functionality. The pIRES2-EGFP-expressing vector system (both plus and minus the IL-33 gene) was stably transfected into TC1 and A9 cells, matched antecedent primary and metastatic cell lines. The EGFP was constitutively expressed and allowed green GFP-expressing tumour cells to be sorted and stable clones to be isolated. Whereas primary TC1 cells are known to express TAP-1 and MHC-I, metastatic A9 cells have greatly reduced expression of these genes 41 . Overexpression of IL-33 in transfected A9 cells induced increased expression of MHC-I (Fig. 3a,b), linking IL-33 to antigen presentation and suggesting that IL-33 acts upstream of this pathway to restore MHC-I expression in tumour cells.
To evaluate the functionality of MHC-I induced by IL-33, the ability of stably transfected tumour cells to present an MHC-I -restricted epitope complex was assessed using a B3Z T cell assay. The B3Z T cell line is a lacZ-inducible CD8 + T cell hybridoma genetically modified to show the activation through the only one signalrecognition of MHC-I in association with the specific antigen from ovalbumin (OVA 257− 264; SIINFEKL) 42 . After incubation with soluble SIINFEKL, the transfected tumour cells were cultured with B3Z T cells. An increased activation of the B3Z T cells corresponds to a higher number of MHC-I:OVA complexes presented on the tumour cell surface. A9 cells stably transfected with IL-33 gene were able to present a higher number of MHC-I:SIINFEKL complexes on their surface than did A9 cells transfected with vector alone (Fig. 3c). This result showed that IL-33expressing A9 cells had a greater capacity for B3Z T cell priming and activation. These data directly demonstrate that IL-33 expression improves immune recognition of the murine metastatic lung carcinoma in vitro.

IL-33 gene-complementation reverses metastasis in vivo.
To test whether the IL-33-induced increase in MHC-I expression in the metastatic A9 cell line could contribute to the anti-tumoural effect of IL-33-gene in vivo, a mouse study was conducted. Tumour cells were initially injected subcutaneously into the right flank and local tumour growth was monitored. Use of the stably transfected A9 and TC1 cells, which constitutively expressed EGFP, allowed the green tumour cells to be tracked for spread beyond the initial site of injection.  We found that expression of IL-33 significantly inhibited metastatic A9 tumour formation. The mean volume of tumours grown from A9+ IL-33 cells was ~30-45% lower than from A9 controls, although both were higher as compared to the mean volume of primary TC1 tumours over the duration of study (Fig. 4a). Moreover, we observed that mice injected with A9 alone also suffered from severe clinical presentation in the form of significant weight loss and tumour ulceration (data not shown), which partially determined the length of the tumour trial. Animals with TC1 (primary tumours) and A9+ IL-33 tumours better maintained or gained weight throughout the study. Tumour ulcerations (stage 4-5) with bleeding of adjacent tissues were detected for some cases in the group injected with metastatic A9 cells, but this was not seen in animals injected with primary TC1 or metastatic A9+ IL-33 tumours.  The most important and unfavourable prognostic factor for the clinical course of tumour development is the metastatic spread of the disease. Therefore, the appearance of circulating tumour cells (CTCs) in disaggregated tissues was assessed in the most common metastatic sites for lung carcinoma: brain, lungs, liver, adrenal glands, lymph nodes and blood tissue. The highest percentage of GFP-positive CTCs were detected in liver (up to 32.0%) and adrenal glands (up to 16.0%) of animals bearing A9 tumours. IL-33 expression by the A9+ IL-33 tumour cells reduced the average number of GFP-positive cells to ~0.15% in liver, and to ~2.63% in the adrenal glands ( Fig. 4b,c; Table 1). Only single tumour cells were detected from lung and blood, while all the brain samples appeared tumour-free. CTCs were not detected from any of the tested tissues and organs of animals bearing primary tumours (TC1) with local growing potential, which are not genetically programmed to disseminate to distant sites (Table 1). Collectively, these observations suggest that IL-33 gene-complementation can modify the malignant gene expression programming. As a result, IL-33 expression decreases the metastatic potential of the cancer cell population reducing the in vivo tumour growth rate, metastatic spread of the disease and its severity.

Tumour-infiltrating immune cells.
Tumour-infiltrating immune cells are intimately linked to the kinetics of tumour growth 43,44 . The involvement of innate immunity (e.g. ILC2s, neutrophils, macrophages, eosinophils) and adaptive immunity (CD4 + and CD8 + T cells) were assessed by flow cytometry and immunohistochemistry. Up-regulated immune recognition shown previously by cytotoxic CD8 + T lymphocytes (CTLs) in the B3Z T cell assay (Fig. 3c) was well in line with the number of CD8 + tumour-infiltrating lymphocytes (TIL) detected in resected tumours by FACS analysis and seen within tumour sections (Fig. 5). In disaggregated tumour tissues, FACS analysis indicated CD8 + T cell content was lower in metastatic A9 tumours than in either the primary (TC1) or the modified metastatic A9+ IL-33 tumours (Fig. 5a). This data was supported by immunohistochemical visualization of tumour tissue sections, which showed increased staining intensity for MHC-I-positive cells within IL-33-expressing tumours when compared to metastatic untreated tumours ( Fig. 5b-h). Moreover, the number of CD8+ T cells was indirectly related to the number of circulating tumour cells. The percentage of circulating tumour cells (CTC) that could be detected in distal organs was significantly lower in IL-33-expressing tumours. Reduced CTCs also corresponded to increased percentages of CD8+ cytotoxic T ccells found in both the tumour and in local lymph nodes, suggesting that IL-33-induced changes may provide favorable conditions for the immune system to reduce metastatic spread, possibly by means of CD8+ T cells (Table S1). Interestingly, the metastatic A9 tumour microenvironment stained intensely for suppressive T-regulatory cell marker (FoxP3 + ), while this was not seen in the A9+ IL-33 expressing tumour environment or in the negative control ( Fig. 5o-q) that are also associated with tumour immune-resistance. However, the expression of IL-33 by the tumours appeared to prevent the accumulation of these immune-suppressive cells, and promoted intensive infiltration by CD68 + tumour-associated macrophages (Fig. 5i-k) and neutrophils (Fig. 5l-n) in IL-33-expressing A9 tumours compared to unmodified A9 tumours. Eosinophils could be seen at the tumour periphery in metastatic A9 tumours, adjacent to the normal tissue, whereas the tumour expression of IL-33 appeared to remodel the microenvironment and allow eosinophils to flow into the tumour tissue and exert an anti-tumour effect, perhaps, due to IL-13, IL-5 production by ILC2s (Fig. 6). The increased frequency and number of innate and adaptive immune cells in IL-33 -expressing tumours supports the conclusion that tumour-infiltrating immune cells mediate protective anti-tumour immunity in murine lung carcinoma. The fact that T cells also require antigen stimulation to up-regulate expression of CD8 implies that IL-33 induction or over-expression can overcome MHC-I deficiency of metastatic tumours in vivo.

IL-33 as an immune biomarker of metastatic human carcinomas.
While we have demonstrated the importance of IL-33 in immune recognition of metastatic cancers in mice, we next tested whether this correlation held true in human clinical samples. Whereas our microarray studies indicated the expression of IL-33 may be important for cancer development, the involvement of the immune response in limiting prostate cancer and potential effectiveness of immunotherapeutic approaches in decreasing cancer recurrence is still widely debated in the literature. To begin to address this question, we examined the expression of IL-33 in prostate cancer specimens obtained at the Vancouver Prostate Centre (VPC). At the mRNA level, IL-33 expression was higher in benign prostate tissue than primary tumours that included naïve tumours (untreated prior to surgery) or NHT tumours (treated prior to surgery (p < 0.0001; t-test), and further reduced in metastatic CRPC in a large cohort of RNA-sequencing data (Fig. 7a), suggesting an inverse correlation with disease progression. Next, we evaluated the protein expression of IL-33 by immunohistochemistry in 342 prostate cancer specimens  Table 1. IL-33 gene-complementation inhibits metastatic spread of tumour cells to distal organs. GFPpositive circulating tumour cells were isolated from sites that were distal from initial subcutaneous inoculation and assessed using flow cytometry. Shown here are representative results from each of the four tumour groups, where n = eight animals/group. obtained at prostatectomy (Fig. 7b). Likewise, low IL-33 expression was significantly associated with reduced time to prostate-specific antigen (PSA) recurrence (a marker of relapse) after prostatectomy (p < 0.0001; Logrank test) (Fig. 7b). Patients with low tumour IL-33 expression had a median time to recurrence of 56.7 months, compared to 97 months for patients with higher tumour IL-33 expression. This effect appeared to be further correlated with Gleason grade, since there was a significant decrease in IL-33 expression within tumours of Gleason pattern 5 when compared to those of Gleason pattern 3 (Fig. 7b). To confirm this clinical association, we further explored IL-33 expression in an independent publically-available cohort of mRNA expression data from 131 cancer specimens obtained at prostatectomy including 19 metastatic tumours 45 . Again, IL-33 showed reduced expression with disease progression, and IL-33 down-regulation was significantly associated with reduced time to PSA recurrence (p = 0.013; Logrank test) (Fig. 7c). Overall this data suggests that reduced IL-33 expression in prostate cancer cells is associated with progression to metastatic disease. We observed a similar correlation in patients with kidney renal clear cell carcinoma, using a dataset from The Cancer Genome Atlas. The report was based on analysis of 513 cases 46 . IL-33 down-regulation was significantly associated with reduced survival time (Fig. 7d). Thus, patients with low IL-33 expression had a shorter median time of survival (52.04 months), compared to patients with higher tumour IL-33 expression (80.62 months). Collectively, these data show that the reduced IL-33 expression is associated with progression to metastatic disease and reduced time to disease recurrence in patients who had resected metastatic human carcinomas.

Discussion
In this study, we make a number of novel and potentially paradigm-shifting observations. First, we identify that IL-33 expression is reduced in many carcinomas upon their transition to the metastatic form of disease. Second, the expression of IL-33 and MHC-I by the tumours appears to be directly correlated and possibly co-regulated. Third, the re-introduction of IL-33 into metastatic tumours induces a reduction of circulating tumour cells and boosts immune recognition against metastatic tumours in vivo. Fourth, the expression of IL-33 may be used as an immune biomarker for recurrence and survival in human prostate and kidney renal clear cell carcinomas. Overall the down-regulation of IL-33 takes place concomitantly with the transition from primary to metastatic tumours and represents an entirely new form of tumour immune-escape.
We initially performed microarray analysis to define the metastatic gene signature network within cancer cells. While we identified many important potential regulators of immune evasion and inflammation that were significantly increased or down-regulated in both metastatic forms of prostate and lung cancers, we specifically focused on the "alarmin" IL-33 that is associated with the inflammatory process, immune recognition and functional development of the newly discovered ILC2s. This observation is the first that suggests loss of IL-33 by primary cancer cells is a component of the immune-escape pathway leading to the progression to metastatic forms, possibly, through an immune-selection mechanism akin to antigenic drift in viruses.
We identified IL-33 to be important for the immune recognition of murine lung, human kidney renal clear cell and in human and murine prostate, carcinomas. In our previous studies, it has been shown that down-regulation of MHC1-related genes ultimately leads to a reduction of immune-surveillance in metastatic cancers 4,5,38,41,59 . The present study indicates that complementation of tumour cells with IL-33 induces MHC-1 production and contributes to a reversal of antigen presentation deficiency, consequently shifting the tumour phenotype from immune evasiveness to immune recognition. The opposite effect, however, has been obtained after inhibition of IL-33 in primary tumour cells. Not only does IL-33 appear to be important for the expression of the immune-recognizable phenotype in primary tumour cell lines, but also MHC-I and IL-33 appear to be correlated during the metastatic re-programming of primary murine lung tumour revealing the metastatic potential of IL-33 gene.
The IL-33/MHC-I correlation is a very important aspect of this study. IL-33 is a strong inducer of MHC-I and antigen processing in metastatic cancers. While the down-regulation of IL-33 ultimately decreases functionality of MHC-I and reduces immune-surveillance, the overexpression of IL-33 reverses the antigen presentation metastatic prostate tumours confirms an association between low IL-33 expression (z-score relative to normal benign < − 2) and time to PSA recurrence in this independent cohort of 131 primary prostate tumours; (d) association between low IL-33 expression (z-score relative to normal benign < − 1) and survival of kidney renal clear cell carcinoma patients. deficiency in murine metastatice lung carcinomas, and makes the tumour noticeable to both innate and adaptive immune pathways. The current paradigm in immunology declares that emergence of tumours can be limited by a robust adaptive immune response generated by the interaction of a CD8 + T cell with a TAA-MHC-I or a TSA-MHC-I complex. This mechanism of immune-surveillance is thought to work efficiently until tumour cells undergo chromosomal alterations that result in a loss of the expression of APM components and subsequent conversion to an immune-escape phenotype. While the most frequent immune escape mechanism is the reduced production of components from a cassette of genes involved in antigen processing and presentation 4,47,48 , downregulation of these genes ultimately decreases functionality of MHC-I, and reduces immune-surveillance. The relevance of this observations in cases of human cancer has been shown in lung 37 , melanoma 49 , renal carcinoma 50 , colorectal carcinoma 51 , head and neck squamous cell cancer 52 , cervical 53 and breast cancers 54 with a similar statistically significant relationship. Our study model addresses a key mechanism in immune responsiveness vs. immune evasiveness that underpins immune-surveillance to cancers. While the growth and expansion of primary tumour remains hampered by the immune system, the immune-evasive metastatic tumour phenotype allows it to acquire a growth advantage resulting in metastatic spread of the disease and its enhanced severity. Thus, the MHC-I and IL-33 co-relationship introduces a novel mechanism of cancer immune escape during tumour development, revealing the involvement of IL-33 into cancerogenesis.
Recent studies have provided equivocal results and shown both protective and promoting effects of IL-33 in tumorigenesis. The present study resolves this debate. Overall our data shows that the reduced IL-33 expression in tumours is associated with progression to metastatic disease. In addition, we demonstrate that IL-33 down-regulation is significantly associated with reduced survival time in human prostate cancer (97 months compared to 56.7 months). We found a similar relationship in patients with kidney renal clear cell carcinoma (80.6 months versus 52 months; TCGA, access date: March, 2015). The present human studies conclusively demonstrate that the level of IL-33 gene is downregulated in human metastatic prostate cancer compared to human benign and primary prostate tumours. The immunohistochemical staining of human prostate tumour tissue samples shows an agreement between mRNA and protein expression levels, and for the first time demonstrates a correlation to disease outcome, where low expression of IL-33 in radical prostatectomy specimens is associated with a significantly decreased time to relapse after surgery compared to specimens with high IL-33 expression. Collectively, these results indicate that the expression level of IL-33 may be a potential candidate as the first immune prognostic marker for prostate tumour transition to its metastatic form. In order to prove that, future mulitvariant analysis will be required to evaluate IL-33 mRNA/protein expression levels with clinicopathological variables which are currently used as important diagnostic features in prostate cancer, such as stage of disease, PSA level, and Gleason Score. Future studies may provide support that IL-33 down-regulation is a common thread in other carcinomas.
Only a small fraction of patients survive metastatic forms of cancer. A profound limitation in the application of general cancer treatments is the genetic and phenotypic heterogeneity in tumours between patients. Furthermore, current immunotherapy is hampered by the need to personalize treatments based on identification of tumour-specific antigens 55,56 , as well as on the individual tumour-specific T cell receptor expression in responding tumour-infiltrating lymphocytes. These parameters and others, such as responsiveness to PD-1, PD-L1 or CTLA4 treatment, impact therapeutic options going forward. We have demonstrated that IL-33 restores immune-responsiveness against metastatic tumours via induction of antigen processing pathways. These studies may lead to innovative and novel forms of cancer therapies for metastatic disease based on the utilization of IL-33 that could be applied as generalized cancer treatment that restores immune surveillance and combats immune-escape.

Material and Methods
Cell lines. Murine prostate cancer model. The primary PA and metastatic LMD prostate tumours were used as models of non-metastatic and metastatic prostate cancer, respectively. PA is a primary murine prostate cancer cell line derived from a 129/Sv mouse using a mouse prostate reconstitution model system that displays high expression of MHC-I. LMD is a metastatic TAP-and MHC-I deficient derivative of PA which emerged as a metastatic daughter after escaping and metastasising from the kidney capsule during serial transfer of the PA cells 57 .
Murine lung tumour model. The TC1 cell line is a murine lung tumour model derived from primary lung epithelial cells of C57BL/6 mice immortalized using the amphotropic retrovirus vector LXSN16 carrying human papillomavirus genes E6/E7, and subsequently transformed with pVEJB plasmid expressing the activated human c-Ha-ras oncogene. TC1 cells display high expression of TAP-1 and MHC-I. The cell line A9 was derived from the TC1 tumour cell line and display spontaneous down-regulation of MHC-I (or MHC-I) by immunoselection in vivo during immunization/challenge experiment 58 . All the above cell lines were grown in Dulbecco's modified Eagle medium, supplemented with 10% heat-inactivated fetal bovine serum, 2 mM l-glutamine, 100 U/ml penicillin, 100 μ g/ml streptomycin, and 10 mM HEPES.
Microarray analysis of both cell systems. Purified mRNA samples from both cancer cell model systems were sent to the Microarray Centre at the University Health Centre (now called The Princess Margaret Genomics Centre) in Toronto, Canada, where they were hybridized to a 28005 Two-Color Agilent microarray with a total of 55821 probes. Data analysis was performed using GeneSpring GX software (Agilent Technologies).
Gene expression constructs and transfection. Note: untransfected tumour cells were used for microarray studies and for siRNA studies.
Scientific RepoRts | 6:30555 | DOI: 10.1038/srep30555 Transient clones. A gene expression construct with full-length cDNA for the selected mouse IL-33 gene (NM_001164724.1) was produced using the pIRES2-EGFP vector (Clontech Lab). Vector constructs, i.e. either vector alone or vector plus IL-33 gene, were transfected into both the primary TC1 cell line and the antecedent immune evasive, MHC-I deficient cell line, A9, using a FuGene 6 transfection reagent (Promega).

Stable clones.
A gene expression construct with full-length cDNA for the selected mouse IL-33 gene (NM_001164724.1) was produced using the pIRES2-EGFP vector (Clontech Lab). TC1 and A9 cells were transfected with the pIRES2-EGFP-gene constructs using FuGene 6 transfection reagent (Promega) in vitro. Forty-eight hours after transfection GFP-positive cells were sorted by FACS (BD Aria II cell sorter) in order to obtain stable transfectants. Selection and expansion in culture was repeated twice before the cells were finally sorted into single-cell clones. Stably transfected clones were further isolated by flow cytometry from a population of GFP-positive cells. These stable GFP-expressing cells were used for all mouse studies.
siRNA study. TC1 cells were transfected with siRNAs targeted against IL-33 (GS 77125), against MHC-I (GS 1027416), untargeted siRNA (1022076) using HiPerFect Transfection reagent (301704) or left untreated. All reagents were purchased from Qiagen. The expression level of IL-33 and MHC-I was assessed in 72 to 96 hours.
Reverse Transcription. Total cellular RNA was extracted using Illustra RNAspin Mini Kit (GE Healthcare Life Science). Reverse transcription of 1 μ g of total cellular RNA was performed using the reverse transcription kit (SSII RT) from Invitrogen.

Real-time quantitative PCR analysis (qRT-PCR).
Purified genomic DNA was used as a template for amplifications using 200 to 500 nM of each primer and 1 μ l SYBR Green Taq ReadyMix (Roche) in a total 10 μ l reaction mixture. Thirty-five to forty cycles of denaturation (5 s, 95 °C), annealing (5 s, 61 to 63 °C), and elongation (20 s, 72 °C) were carried out using a Roche LightCycler 480 instrument.
Western blots. RIPA Lysis Buffer System (sc-24948, Santa Cruz Biotechnology) was used for protein isolation from cells and tissues. To shear genomic DNA, lysed samples were passed ten times through a 21-gauge needle then incubated on ice for 30 minutes. The homogenate was centrifuged at 4 °C at 14000 × g for 10 minutes. Protein concentrations from the supernatants were determined by BCA assay (Pierce) and samples were adjusted to final concentration of 50 μ g per lane. Proteins were separated with 10% sodium dodecyl sulfate-polyacrylamide gel electrophoresis and transferred to nitrocellulose membranes (Bio-Rad). Blots were blocked with 5% skim milk in phosphate-buffered saline and incubated overnight at 4 °C with the anti-IL-33 [Nessy-1] (ab54385, Abcam) mouse monoclonal antibody, mouse antiserum directed toward the region of MHC-I encoded by exon 8 (kindly provided by Dr. David Williams, University of Toronto), rabbit anti-mouse TAP-1 polyclonal antibodies (made in-house by W. Jefferies' lab 59 ), followed by the secondary antibodies, which were complimentary to the species of the primary conjugated with Alexa Fluor 680 (Invitrogen). CD4, CD8 staining. APC-conjugated antibody against CD4 (553051, RM4-5, BD Bioscience) and PE-Cy7-conjugated antibody against CD8a (25-0081-82, 53-6.7, Affymetrix eBioscience).

MHC-I expression. PE-conjugated anti-Kb mouse monoclonal antibody (553570, BD Pharmingen).
Flow Cytometry. BD FACS Aria II was used for cell sorting and phenotypic analysis. The program FlowJo ver-sion10.1 was used for data analysis.
Mice. C57BL/6 mice were purchased from the Jackson Laboratories and maintained in the British Columbia Cancer Research Centre (BCCRC) pathogen-free animal facility. Mice were used at 4-8 weeks of age. These experiments were approved by the Animal Care Committee of the University of British Columbia. Animals were maintained and euthanized under humane conditions in accordance with the guidelines of the Canadian Council on Animal Care.
Tumour establishment. Genetically modified tumour cells (50 μ l of 5 × 10 5 ) were injected into chimera animals subcutaneously into the right flank. Tumour growth was monitored by measuring tumour dimensions with calipers. Tumour length and width measurements were obtained three times weekly. Tumour volumes were calculated according to the equation tumour volume= length x width x height x π /6 with the length (mm) being the longer axis of the tumour. Animals were weighed at the time of tumour measurement.
Human Prostate Samples. mRNA Sequencing Cohort. RNA-sequencing data from the Vancouver Prostate Centre was obtained and analyzed exactly as previously described 61 . Public data from Taylor et al. 45 was explored using the cBioPortal 62 .
Immunohistochemistry. Immunohistochemical stains were conducted at the Vancouver Prostate Centre using a Ventana autostainer model Discover XT (Ventana Medical System) with enzyme labeled biotin streptavidin system and solvent resistant DAB Map kit using 1/50 concentration of rabbit polyclonal IL-33 antibody (HPA024426; Sigma-Aldrich). Staining was performed on 342 prostate cancer specimens obtained from the Vancouver Prostate Centre. The H&E stained slides were reviewed and the desired areas were marked on them and their correspondent paraffin blocks. Five tissue microarrays (TMAs) were manually constructed (Beecher Instruments) by punching duplicate cores of 1mm for each sample. Stained slides were digitalized with the SL801 autoloader and Leica SCN400 scanning system (Leica Microsystems) at magnification equivalent to × 20. Representative cores (clearly positive, clearly negative and mixed positive/negative) were manually identified by an experienced pathologist (L Fazli) and a four-point scale was assigned as follows: 0 represents no staining in any tumour cells, 1 represents a faint or focal, or questionably present stain, 2 and 3 represents a stain of convincing intensity in a majority of cells. For comparisons, a score of 0 or 1 was considered low IL-33 expression.
Statistics. Data were analyzed with Excel. A Student's t test was used for determining statistical significance between groups; p ≤ 0.05 was considered significant. The statistical analysis of microarray results was carried out with FlexArray (Genome Quebec).