Pro- and anti-tumorigenic capacity of immunoproteasomes in shaping the tumor microenvironment

Abstract

Apart from the constitutive proteasome, the immunoproteasome that comprises the three proteolytic subunits LMP2, MECL-1, and LMP7 is expressed in most immune cells. In this study, we describe opposing roles for immunoproteasomes in regulating the tumor microenvironment. During chronic inflammation, immunoproteasomes modulated the expression of pro-tumorigenic cytokines and chemokines and enhanced infiltration of innate immune cells, thus triggering the onset of colitis-associated carcinogenesis (CAC) in wild-type (WT) mice. Consequently, immunoproteasome-deficient animals (LMP2/MECL-1/LMP7-null mice) were almost completely resistant to CAC development. In ulcerative colitis (UC) patients with high risk for CAC, immunoproteasome-induced pro-tumorigenic mediators were upregulated. In melanoma tumors, the role of immunoproteasomes is relatively unknown. We found that high expression of immunoproteasomes in human melanoma was associated with better prognosis. Similarly, our data revealed that the immunoproteasome has anti-tumorigenic activity in a mouse model of melanoma. The anti-tumor immunity against melanoma was compromised in immunoproteasome-deficient mice due to the impaired activity of CD8⁺ cytotoxic T lymphocytes (CTLs), CD4⁺ Th1 cells, and antigen-presenting cells (APCs). These findings show that immunoproteasomes may exert opposing roles with either pro- or anti-tumoral properties in a context-dependent manner.

Introduction

The 20S proteasome, which is a part of the proteolytic enzyme called the 26S proteasome, is a barrel-shaped complex consisting of four heptameric rings that envelop the catalytic chamber, in which unneeded or damaged proteins are degraded by proteolysis (1). The two outer rings of the 20S proteasome are structurally related α subunits (α1–α7), and the two inner rings are made of seven β subunits (2). Three of the β proteasomal proteins, namely β1, β2, and β5, are catalytically active in vertebrate proteasomes. Based on their substrate specificity, catalytic subunits of the 20S proteasome (called the constitutive or standard proteasome) exhibit caspase-like (β1), trypsin-like (β2), or chymotrypsin-like (β5) activities (3). In mammals, interferon (IFN)-inducible catalytic subunits low-molecular mass polypeptide (LMP)2, LMP7, and multicatalytic endopeptidase complex-like (MECL)-1 can incorporate into the 20S proteasome instead of the constitutive subunits, leading to the assembly of immunoproteasomes (4,5). The immunoproteasome was originally described as an enzymatic complex abundantly expressed in immune cells, with an enhanced capacity to cleave proteins after their basic and hydrophobic residues, thereby optimizing the antigen presentation of major histocompatibility (MHC) class I-restricted-epitopes to CD8⁺ T cells (6,7). Thus, the function of the immunoproteasome is to optimize CD8⁺ T-cell responses to various epitopes during viral and intracellular bacterial infections (8). Immunoproteasomes may also be important for recognition of neo-antigens, which are newly formed antigens created by mutations in tumor cells. Various antigenic peptides derived from human tumors have been described to be more efficiently produced by immunoproteasomes or intermediate proteasomes than constitutive proteasomes (9,10). Overall, the composition of proteasomes in the cell dictates the peptide repertoire displayed by MHC class I molecules, which has important consequences for elimination of viral infection and recognition of tumor antigens by cytotoxic T lymphocytes (CTLs).

Interestingly, the depletion of immunoproteasomes has a crucial impact on the onset of autoimmune responses and course of inflammation in mouse models of rheumatoid arthritis and colitis, suggesting that this enzymatic complex is not only involved in antigen presentation but also in the regulation of inflammatory reactions (11–13). Novel studies have revealed that selective immunoproteasome inhibitors, such as ONX 0914, might be promising therapeutic agents for treatment of autoimmune disorders, such as multiple sclerosis, but also for inhibiting chronic inflammation in the colon, which is associated with development of colorectal cancer (14–16). Thus, the immunoproteasome seems to have some pro-tumorigenic propensities due to its role in the regulation of expression of pro-inflammatory mediators, which perpetuates the chronic inflammation and leads to tumorigenesis. In this study, we reveal opposing roles for immunoproteasomes in two non-related cancer models. Although immunoproteasome deficiency resulted in almost complete inhibition of tumor growth in colitis-associated colonic cancer (CAC), the absence of the same enzymatic complex resulted in enhanced tumor load and insufficient anti-cancer immune responses in a murine melanoma model. These findings suggest that the involvement of immunoproteasomes in shaping the tumor microenvironment (TME) diverges between different cancer types. During development of chronic inflammation-associated cancer, the pro-tumoral capacity of immunoproteasomes appeared to contribute to carcinogenesis by inducing various pro-inflammatory factors such as CXCL1, CXCL2, CXCL3, IL17A, IL1β, and IL6. By contrast, in the TME of melanoma, the immunoproteasome acted as an anti-tumorigenic factor by supporting the effector function of T cells and by promoting CTL-mediated anti-cancer immunity.

Materials and Methods

Human samples

Human tissue was obtained from colonic specimens immediately after the surgery performed at the Charité University Hospital in Berlin, Germany. This study was approved by the local ethics committee of the Charité - Universitätsmedizin, Berlin, Germany. Written informed consent was obtained from all patients. Following the surgery, colonic tissue was separated from the underlying submucosa, subsequently frozen in liquid nitrogen, and stored at –196°C until use. The degree of inflammation in the surgical specimens was evaluated using a standard scoring system. A total of 10 patients with active ulcerative colitis (UC), 10 patients with active Crohn’s disease (CD), and 10 control tissues were examined by microarray analysis, as described in in the next section.

Gene array analysis

Samples from patients were used in microarrays. Total RNA from colonic tissues was extracted with TRIzol (Invitrogen). RNA quality was routinely tested using 2100 Bioanalyzer (Agilent Technologies). Labeling of RNA for hybridization was performed using the Low RNA Input Linear Amp Kit from Agilent Technologies according to the supplier’s instructions. Briefly, 500 ng total RNA was reverse transcribed with an oligo-dT-T7 promoter primer and MMLV-RT. Second, strand synthesis was carried out with random hexamers. Fluorescent antisense cRNA (aRNA) was synthesized with either cyanine 3-CTP (Cy3-CTP) or cyanine 5-CTP (Cy5-CTP) and T7 polymerase. Purified products were quantified, and 2 μg labeled aRNA for each sample was fragmented and mixed with control targets and hybridization buffer. Hybridizations were performed at 60 °C for 17 hours. Slides were washed according to manufacturer’s protocol (SSC wash protocol), and scanning of the arrays was performed at a 5 μm resolution using a DNA microarray laser scanner (Agilent Technologies). Features were extracted from raw image data using the Agilent Technologies image analysis software (G2567AA Feature Extraction Software, Version A7.5.1) and default settings. The microarray gene expression markup language (MAGE-ML) data analysis was performed with the Rosetta Resolver software package (licensed by Aventis Pharmaceuticals) comprising a pre-processing pipeline of the feature extraction. This includes error model adjustments to the raw scan data, background subtraction by ‘spatial detrend’, and array normalization using rank consistency filtering of normalization feature selection with a combination of linear and LOWESS curve fitting methods. Ratios were calculated by the most conservative estimates between universal error model and propagated error. Ratio profiles were combined in an error-weighted fashion by Resolver to create ratio experiments. To compensate for dye-specific effects, and to ensure statistically significant data, a color-swap analysis (fluorescence reversal) was performed. Expression patterns were identified using 1.5-fold expression cut-offs of the ratio experiments and anti-correlation of the dye reversal profiles with an error weighted p-value < 0.05, rendering the microarray analysis significant (P < 0.01), robust, and reproducible. Raw intensity data of the .txt files without preprocessing were analyzed using the R package limma (Bioconductor). Gene set enrichment analysis (GSEA) (https://www.gsea-msigdb.org/gsea/index.jsp) was performed using the “on” genes pre-ranked by gene expression-based t-score using standard settings with 1,000 permutations. The microarray data have been deposited in the National Center for Biotechnology Information Gene Expression Omnibus (GEO) under accession code GSE150979.

Analysis of publicly available transcriptomes

For the association of melanoma patient survival with expression of PSMB8 (ENSG00000204264), PSMB9 (ENSG00000240065), and PSMB10 (ENSG00000205220), we interrogated The Cancer Genome Atlas (TCGA) database, accessed on March 30^th 2020, using the package TCGAbiolinks v. 2.15.3 in the R programming environment v. 3.5.1. We retrieved harmonized FPKM values from RNA-seq experiments for PSMB8, PSMB9, and PSMB10 from the SKCM (Skin Cancer Melanoma) project. FPKM were grouped into quartiles, and the upper and lower quartiles were used for correlation with patient survival over a course of 3650 days.

For UC patients, colon cancer-associated genes were extracted from the publicly available gene set (TCGA, M14524, GRADE_COLON_CANCER_UP). Heatmaps were constructed by plotting z-score-transformed array intensities with the R package gplots v. 3.0.1.1. GSEA (https://www.gsea-msigdb.org/gsea/index.jsp) was used to determine over-representation of selected genes in publicly available datasets. Genes were chosen on the basis of absolute fold-change in UC patients vs. control.

Mice

Mice were maintained under SPF conditions at the Biomedical Research Center, Philipps-University of Marburg. Female 8–12-week old wild-type (WT), PBSM8/PSMB9/PSMB10-null [triple-knockout (TKO)] mice, and LMP7^−/− mice on a C57BL/6 background were used for animal experiments. WT mice were obtained from The Jackson Laboratory. LMP7^−/− mice were provided by Antje Behling (Charité - Universitätsmedizin Berlin). TKO mice were generated by Kenneth Rock (University of Massachusetts) as reported earlier (17) and were bred in our own animal facility. The study was approved by Regierungspräsidium Gießen, Germany (Nr. 70/2014 and Nr. 24/2019) and conducted according to the German animal protection law.

Cell lines and cell line treatment

The human colorectal cancer (CRC) cell line HT-29 was purchased from ATCC (HTB-38). B16-F10 and B16-GFP melanoma cells were kindly provided by Tobias Bopp (Johannes Gutenberg-University Mainz). X6310-GM-CSF cell line used for generation of conditioned media was provided by Markus Schnare (Philipps University Marburg). HT-29 and X6310-GM-CSF cells were cultured in DMEM (Sigma-Aldrich) supplemented with 10% FCS,100 U/ml penicillin/streptomycin (AppliChem) and 2 mM glutamine (PAN-Biotech) at 37°C with 5% CO₂. Cell lines used in experiments were routinely checked for mycoplasma contamination. 1.5×10⁶ HT-29 cells per well were seeded into a 12-well plate and pretreated with IFNγ (200 U/mL, PeproTech) for 24 hours. Afterwards, the cells were treated with TNFα (10 ng/mL, PeproTech) and IL1β (10 ng/mL, PeproTech) in absence or presence of ONX 0914 (0.5 μM; an immunoproteasome inhibitor, Onyx Pharmaceuticals) for additional 24 hours and subjected to RT-qPCR analysis as described in the next section. An equal volume of captisol (Ligand Pharmaceuticals, solvent for ONX 0914) without ONX 0914 was used as control. For detection of immunoproteasome expression in melanoma cells, 2×10⁶ B16-F10 cells were cultured in a 12-well plate and treated with IFNγ (500 U/mL) for 48 hours. Subsequently, RT-qPCR was performed.

Subcutaneous melanoma tumor model

B16-F10 melanoma cells were cultured in RPMI 1640 medium (Thermo Fisher Scientific) supplemented with 100 U/ml penicillin/streptomycin, 10% FCS and 2 mM glutamine for 1 week before inoculation into mice. WT, TKO LMP7^−/− mice were subcutaneously (s.c.) injected with 1 × 10⁶ B16-F10 cells. In some experiments, the B16-GFP cell line was used as well. Tumor growth was monitored daily for 15 days by caliper measurements. To analyze T cells in tumors and tumor-draining lymph nodes (LNs), mice were sacrificed, and tumor tissues and draining LNs were harvested. The single-cell suspensions from LNs were obtained by mechanical disruption using 30 μm pre-separation filters (Miltenyi Biotec). Tumor-infiltrating lymphocytes (TILs) were isolated using the Miltenyi TIL Isolation kit (Miltenyi Biotec, 130–096-730) and the GentleMACS Octo Dissociator (Miltenyi Biotec) according to the manufacturer’s instructions. Briefly, the melanoma tumors were cut into small pieces and the tissue was transferred into GentleMACS tubes containing the enzyme mix. The tumors were dissociated into single-cell suspensions by combining enzymatic digestion with mechanical dissociation. After termination of the program, the cell suspensions were centrifuged (300xg, 7 minutes) and supernatant was completely aspirated.

Induction of colitis-associated carcinogenesis (CAC)

Age- and sex-matched WT and TKO mice were injected with azoxymethane (AOM; Sigma- Aldrich) intraperitoneally (i.p.; 8 mg/kg body weight). After 5 days, mice were treated with 3.0% dextran sodium sulfate (DSS; MP Biomedicals) administered in the drinking water for 5 days, followed by 14 days of normal drinking water. Two additional cycles of DSS treatment were performed as described previously (16). The body weight of mice was monitored three times per week. After 80 days, mice were sacrificed and tumors numbers were counted. For investigation of recruitment of inflammatory cells, mice were analyzed by flow cytometric analysis and real-time quantitative RT-qPCR on day 30 after induction of CAC by AOM/DSS as described below.

Isolation of colonic lamina propria mononuclear cells (LPMCs)

Colonic tissue from WT and TKO mice was isolated and cut into pieces. The colonic tissue was homogenized using the Miltenyi lamina propria dissociation kit (Miltenyi Biotec, 130–097-410) and Miltenyi GentleMACS Octo dissociator, according to the protocol. After digestion, the cell suspensions were washed and further separated by a discontinuous density gradient centrifugation. The cells were resuspended in 40% percoll (Merck) and carefully layered on 70% percoll. After centrifugation at 625 x g for 20 minutes, the LPMCs were collected from the interphase, washed, and prepared for further analysis as described below.

Histology

Hematoxylin and eosin (H&E) staining was performed on 5 μm thick colon tissue cryosections derived from WT and TKO mice. The samples were stained for 10 minutes with hematoxylin (Carl Roth). Subsequently, cryosections were washed with water and incubated for 15 seconds with eosin (Carl Roth). Following eosin staining, incubation of samples with 70% (1 second), 80% (10 seconds), and 96 % ethanol (20 seconds), isopropanol (5 minutes), and Roti-Histol (Carl Roth, 2×3 minutes) was performed. The slides were dried and examined by bright field microscopy. Leica DFC480 camera was used to take digital images by Leica Application Suite V3.8 software.

Generation of bone marrow-derived dendritic cells (BMDCs)

Bone marrow progenitor cells were harvested from bone marrow (tibia or femur) of 8–12 week old WT and TKO mice and cultured for 1 week in petri dishes in RPMI-1640 media supplemented with 10% X6310-derived culture supernatant containing granulocyte macrophage colony-stimulating factor (GM-CSF). On day 7 of cell culture, the purity of cultured cells was tested by flow cytometry as described below. Afterwards, the cells were stimulated with LPS (100 ng/mL, Sigma-Aldrich) for 24 hours and assessed for NF-κB and ERK signaling by Western blot.

Western blots on BMDCs and thymocytes

For generating whole cell lysates from murine BMDCs and thymocytes, RIPA lysis buffer (Sigma-Alrich) supplemented with protease inhibitors (Protease Inhibitor Cocktail, Thermo Scientific). The quantification of total protein within the cell lysates was performed with Pierce BCA Protein Assays (Thermo Fisher Scientific). All samples were loaded on 12 % SDS PAGE gels with 20 μg of total protein. Following separation by electrophoresis, the samples were transferred on a PVDF membrane (Bio-Rad Laboratories). Subsequently, PVDF membranes containing transferred proteins were blocked for 1 hour with 5 % BSA, followed by incubating the membrane with primary antibodies overnight at 4°C. Following primary antibodies were used: anti-p105 (eBioscience, catalog no. 14–6732-81), anti-p-RelA (Cell Signalling, catalog no. 3033) and anti-p-ERK 1/2 (Cell Signaling, catalog no. 4370). Afterwards, the samples were incubated with secondary antibodies for 2 hours at room temperature. HRP linked anti-rabbit IgG (Cell Signalling, catalog no. 7074) and anti-mouse IgG (Cell Signalling, catalog no. 7076) were used. As a loading control, anti-β-actin (Sigma-Aldrich, catalog no. A551–2ML) and anti-α-tubulin (Sigma-Aldrich, catalog no. T5168) were used. The analyzed proteins were detected using Western Blotting Luminol Reagent (Santa Cruz Biotechnology, catalog no. sc-2048) at the MicroChemi high performance imager (DNR Bio-Imaging Systems).

Thymocyte and peripheral CD8⁺ T-cell assessment

For ex vivo analysis of thymocytes and CD8⁺ T-cells derived from spleens and LNs of female 8–12-week old WT, LMP7-deficient and TKO mice were used. The single-cell suspensions were generated by mechanical disruption of thymi, spleens, and LNs using 30 μm pre-separation filters (Miltenyi Biotec). Afterwards, single-cell suspensions were counted before proceeding with the flow cytometry.

Antibodies and flow cytometry

Single-cell suspensions derived from thymocytes, spleens, LNs, TILs from melanoma tumors and LPMCs were used for flow cytometry. In some experiments, cells were incubated with Fc block (Miltenyi Biotec, catalog no. 130–092-575) for 15 minutes prior to cell surface staining according to the manufacturer’s protocol. After performing surface staining with various antibodies for 15 min at 4°C, TILs and LPMCs were restimulated for 4 hours with PMA (50 ng/mL) and ionomycin (750 ng/mL) in the presence of Brefeldin A (10 mg/mL; all reagents, Sigma-Aldrich). Following fixation with 2% formaldehyde and permeabilization with 0.3% saponine buffer, the intracellular cytokine staining was performed with antibodies at recommended dilutions for 15 min at 4°C. Flow cytometry was performed using ARIA III and FACSCalibur flow cytometers (both BD Biosciences), followed by analysis using FlowJ_V10 software (TreeStar). The following antibodies were used for staining: anti-CD3 (145–2C11), anti-CD4 (RM4–5), anti-CD8 (53–6.7), anti-IFNγ (XMG1.2), anti-IL17A (eBio17B7), anti-CD11b (clone M1/70), anti-Ly6G (1A8-Ly6g), anti-MHCI (AF6–88.5), and anti-MHCII (TIB 120). All antibodies were purchased from eBioscience, BD Biosciences or BioLegend. A sample gating strategy for murine T cells can be found in Supplementary Fig. S1.

Quantitative real-time PCR (RT-qPCR)

Colon tissues were homogenized using TRI Reagent (Sigma-Aldrich). For HT-29 and B16-F10 cells, the RNA isolation was performed using EXTRACTME TOTAL RNA KIT (blirt, catalog no. EM09.1–250). Total RNA was extracted according to manufacturer’s instructions. RevertAid First Strand cDNASynthesis Kit (Thermo Scientific) was used to generate complementary DNA (cDNA) according to the manufacturer’s instructions. RT-qPCR was conducted using 500 ng template on a StepOne Plus device (Applied Biosystems). For RT-qPCR analysis, the Takyon ROX SYBR® Master Mix blue dTTP Kit (Eurogentec) was used. Quantification of cDNA was carried out by normalization to expression of the housekeeping gene Hprt1 using the 2^-ΔΔCt method. Following murine primers were used: Cxcl1 fwd GCT TGA AGG TGT CCT CAG, rv AAG CCT CGC GAC CAT TCT TG; Cxcl2 fwd GCG GTC TCA ATG CCT GAA GA, rv TTT GAC CGC CCT TGA GAG TG; Cxcl3 fwd CAT CCA GAG CTT GAC GGT GAC, rv CTT GCC GCT CTT CAG TAT CTT CTT; Cox2 fwd ATT CTT TGC CCA GCA CTT CA, rev GGG ATA CAC CTC TCC ACC AA, Il1b fwd TGG GCC TCA AAG GAA AGA AT, rv CAG GCT TGT GCT CTG CTT GT; Il6 fwd GGA TAC CAC TCC CAA CAG ACC, rv TTC TCA TTT CCA CGA TTT CCC A, Il12p40 fwd ATG TGT CCT CAG AAG CTA ACC ATC, Il12p40 rv CGT GTC ACA GGT GAG GTT CAC T, Il12p35 fwd TAC TAG AGA GAC TTC TTC CAC AAC AAG AG, Il12p40 rv TCT GGT ACA TCT TCA AGT CCT CAT AGA, Tnfα fwd AAA ATT CGA GTG ACA AGC CTG TAG, Tnfα rv CCC TTG AAG AGA ACC TGG GAG TAG, Hprt1 fwd CTG GTG AAA AGG ACC TCT CG, rv TGA AGT ACT CAT TAT AGT CAA GGG CA. Following human primers were used: Cxcl1 fwd AGT GTG AAC GTG AAG TCC CC, rv GAT GCA GGA TTG AGG CAA GC; Cxcl2 fwd TGC AGG GAA TTC ACC TCA AGA, rv TGA GAC AAG CTT TCT GCC CA; Cxcl3 fwd GCG CCC AAA CCG AAG TCA, rv GGT GCT CCC CTT GTT CAG TAT C; Psmb8 fwd TGC TCG AGA TGT GAT GAA GG, rv TGT AAT CCA GCA GGT CAG CA; Hprt1 fwd TGC TCG AGA TGT GAT GAA GG, rv TGT AAT CCA GCA GGT CAG CA.

Statistical analysis

For all experiments with two groups, mean values were compared by using an unpaired Student’s t-test (GraphPad Prism 8). For the comparison of the multiple groups, the statistical significance was determined by the one-way analysis of variance (ANOVA) test (GraphPad Prism 8). P values of p<0.05 were considered as significant. Following p-values were used: *P=0.01–0.05; **P=0.001–0.01; ***P<0.001. Data are presented as mean ± s.e.m.

Results

The immunoproteasome is essential for effective anti-tumor immunity against melanoma

The immunoproteasome is involved in modulation of chronic inflammatory environments, as well as in optimal antigen presentation of tumor epitopes to tumor-infiltrating CD8⁺ T cells (10,18). This dichotomy emphasizes the importance of this enzymatic complex in the tumor microenvironment (TME). The role of immunoproteasomes in the onset of melanoma tumorigenesis and in melanoma-specific anti-tumor immunity is largely unknown. To assess the impact of immunoproteasomes on the survival of melanoma patients, we examined the expression of immunoproteasome subunits LMP2 (PSMB9), MECL-1 (PSMB10), and LMP7 (PSMB8) in TCGA datasets from 470 patients with skin cutaneous melanoma. We observed that high expression of all three immunosubunits significantly associated with better overall survival (OS), suggesting that the immunoproteasome might be an important prognostic biomarker for melanoma patients (Fig. 1A).

Figure 1. Immunoproteasome expression in the TME is crucial for optimal CTL-mediated anti-tumor responses.

(A) Kaplan-Meier plots of melanoma patient survival depending on expression (FPKM, lower and upper quartile) of the respective immunoproteasome subunits (TCGA datasets). Number of patients per group indicated in the panel. (B-G) B16-F10 melanoma cells were transplanted subcutaneously into WT and immunoproteasome-deficient TKO mice (n=9 mice/group) for 15 days. (B-D) Tumor volume and tumor mass were analyzed. (E) Representative contour plots of CD4⁺ and CD8⁺ T cells in mice. Bar graphs of (F) frequencies and (G) cell numbers of CD8⁺ and IFNγ⁺CD8⁺ T cells in tumors. One out of three independent experiments with n=3 mice/condition. (B-G). Data were analyzed by two-tailed unpaired Student’s t-test (*P=0.01–0.05, **P=0.001–0.01, mean±s.e.m).

Much research has focused on the inhibition of proteasome activity in tumor cells to promote cell death (19,20). However, the effect of immunoproteasomes and constitutive proteasomes on the TME are poorly characterized. To test our hypothesis that the immunoproteasome might play a crucial role in CD8⁺ T cell-mediated anti-cancer immunity, we analyzed WT mice and animals lacking all three immunoproteasome proteins (TKO mice) implanted with B16-F10 melanoma tumors. In TKO mice, we observed a significantly higher increase in tumor volume and tumor weight compared to WT counterparts, suggesting that the immunoproteasome exhibits anti-tumor effects in this model (Fig. 1B–D). TKO mice had significantly fewer CD8⁺ T cells in tumor-draining lymph nodes (LNs; Supplementary Fig. S2). Less cytotoxic T lymphocytes (CTLs) were recruited into the tumors at day 15 after injection of tumor cells, which led to reduced cell numbers of IFNγ-producing CD8⁺ T lymphocytes in the TME (Fig. 1E–G).

Absence of LMP7 in the TME results in impaired anti-tumor immunity against melanoma

A new catalytic subunit of the proteasome, β5t, which is expressed primarily in cortical thymic epithelial cells, has been described (21). Together with LMP2 and MECL-1, β5t forms so-called thymoproteasomes that are crucial for positive selection of immature thymocytes (22). We observed that both the frequencies and cell numbers of CD8 single-positive thymocytes and CD8⁺ T cells in the periphery were reduced in naïve TKO mice as compared to WT animals (Supplementary Fig. S3). These results confirm previous findings (17), and suggest that the lack of LMP2 and MECL-1, as a part of the thymoproteasome, is responsible for reduced numbers of CD8⁺ T cells in TKO mice because the LMP7-deficient animals displayed normal CD8⁺ T-cell numbers and frequencies at steady state (Supplementary Fig. S4).

To assess the impact of TME-derived LMP7 on melanoma growth, we subcutaneously injected B16-F10 cells into WT and LMP7-deficient animals and monitored the tumor growth for two weeks. We observed progressively increased growth of melanomas in mice lacking LMP7 compared to WT mice, as shown by accelerated tumor size growth and weight (Fig. 2A–B). Although the frequency of CD4⁺ and CD8⁺ T lymphocytes were not reduced in tumors and draining LNs in mice lacking LMP7, the cell number of IFNγ⁺ T lymphocytes was impaired in LMP7-deficient mice (Fig. 2C–D, Supplementary Fig. S5). Apart from CD8⁺ CTLs, Th1 cells are also crucially involved in promoting anti-cancer immunity (23). In the absence of LMP7, reduced expression of IFNγ and IL17A was detected in CD4⁺ T cells isolated from tumors (Fig. 2E). Similarly, compared to WT CD8⁺ T cells, LMP7-deficient CTLs derived from tumors produced less IFNγ (Fig. 2F). The defective anti-tumor response in LMP7-deficient mice was also reflected in reduced TNFα and IL12 production in the TME (Fig. 2G), indicating that antigen-presenting cell (APC)-derived cytokines that induce T-cell IFNγ secretion are also affected by the absence of LMP7. In contrast to WT control cells, in vitro-generated BMDCs from TKO and LMP7-deficient mice were not able to upregulate IL12 expression following stimulation with LPS (Supplementary Fig. S6). Collectively, these results highlight the important role of immunoproteasomes in shaping the TME and promoting anti-tumor activity of CTLs and Th1 cells.

Figure 2. LMP7 deficient mice display reduced anti-tumor immunity against melanoma.

(A, B) Tumor volume and tumor mass were analyzed after subcutaneous inoculation of B16-F10 melanoma cells into WT and LMP7-deficient mice (n=6 mice/group). Tumor weight was analyzed at day 15 post tumor inoculation. (C, D) The percentage of (C) CD4⁺ and CD8⁺ T cells and (D) total cell number of IFNγ-producing CD4⁺ and CD8⁺ T lymphocytes in tumors were analyzed by flow cytometry at day 15 after tumor inoculation. (E, F) Bar graphs and representative contour plots show percentages of (E) IFNγ⁺CD4⁺ T cells and IL17A⁺CD4⁺ T cells; and (F) IFN-γ⁺CD8⁺ T cells on day 15 post tumor inoculation. (G) Relative mRNA expression for Tnfa and Il12p35 was measured in the TME by RT-qPCR on day 15 after B16 inoculation. Two-tailed unpaired Student’s t-test was applied (n.s.=not significant, *P=0.01–0.05, **P=0.001–0.01, ***P<0.001, results are shown as mean±s.e.m). One from two independent experiments is shown (n=6 mice/group).

The TME induces immunoproteasome expression in melanoma cells

The cytokines and chemokines secreted by immune and stroma cells within the melanoma microenvironment play an important role in recruiting immune cells and in communication with tumor cells. To assess the effect of immunoproteasome absence in the TME on the proteasome composition in tumor cells, we inoculated B16-GFP cells into WT, LMP7-deficient, and TKO mice, respectively. On day 15 after tumor inoculation, we measured the expression of immunoproteasomes within the GFP⁺ melanoma cells. The induction of the immunoproteasome subunit LMP7 was observed only in tumor cells derived from WT but not from LMP7 or TKO mice (Supplementary Fig. S7). This implies that immune cells with high expression of IFNγ in the TME of WT mice directly impact the proteasome activity of neighboring tumor cells. This observation was confirmed in in vitro experiments, as IFNγ increased the expression of all three immunoproteasome subunits in B16-F10 cells (Supplementary Fig. S7). Human melanoma cells overexpressing immunoproteasomes have been shown to have a clear tendency to increase their immunogenic antigen repertoire (24). Together, this may affect the magnitude and quality of CTL-mediated anti-cancer responses, as the immunoproteasome expressed in melanoma cells might be implicated in generation of neo-antigens.

Pro-tumoral function of immunoproteasomes during inflammation-driven carcinogenesis

Although the immunoproteasome was originally described as a multi-subunit catalytic complex required for optimizing the generation of peptides for MHC class I antigen processing, we and others have shown that the pharmacologic blockade of immunoproteasomes efficiently inhibits the development of colitis and colitis-associated cancer (CAC)(13,15,16). Over the past few years, findings from several laboratories have demonstrated that immunoproteasomes orchestrate the onset of immunopathology in diseases such as Crohn`s disease (CD) and rheumatoid arthritis (11,25). To better understand the role of immunoproteasomes in regulating the TME during chronic inflammation, we induced CAC in WT mice and mice lacking immunoproteasomes by treating them with the carcinogen azoxymethan (AOM) and dextran sodium sulfate (DSS). As expected, at day 80 after AOM/DSS treatment, we found enhanced development of cancer-associated neoplasia in WT mice, as indicated by the measurement of body weight, tumor numbers, colon length, and the H&E staining. In contrast, animals lacking immunoproteasomes did not show any signs of pathology. Although some minor alterations in the morphology of intestinal epithelial cells were observed, TKO mice exhibited less pronounced weight loss and shortening of the colon, as well as almost no detectable tumors (Fig. 3A–D).

Figure 3. Immunoproteasome-deficient mice are protected from colitis-associated carcinogenesis.

(A-D) WT and TKO mice (n=6 mice/group) were treated with three rounds of DSS after intraperitoneal AOM administration for CAC induction. On day 80, colonic tissue was isolated. (A) Relative weight loss was monitored for 80 days. (B) Bar graphs of numbers of colorectal tumors and (C) colon length. (D) Histology of distal colon tissue (Scale bars: 100μm). Data were pooled from two independent experiments with n=3 mice/experiment and were analyzed by two-tailed unpaired Student’s t-test (**P=0.001–0.01, ***P<0.001, mean±s.e.m).

To examine the role of the immunoproteasome in shaping the TME during chronic inflammation in more detail, we analyzed the immune cells in colonic lamina propria of WT and TKO mice at day 30 after induction of AOM/DSS-mediated carcinogenesis. Along with reduced tumor numbers, a lower percentage of IFNγ- and IL17A-producing CD4⁺ T cells and a lower frequency of infiltrating neutrophils were found in TKO animals compared to WT mice (Fig. 4A–C). Quantitative RT-PCR analysis revealed that many pro-tumorigenic factors involved in the development of CAC, such as the enzyme COX-2, the cytokines IL6 and IL1β, which directly enhance the proliferative capacity of epithelial cells, as well as pro-tumorigenic chemokines CXCL1, CXCL2, and CXCL3, governing the recruitment of neutrophils, were downregulated in TKO mice at day 30 (Fig. 4D). It is known that activated neutrophils produce high IL1β and reactive oxygen species (ROS), directly contributing to the pathogenesis of CAC by inducing DNA damage in epithelial cells (26,27). Neutrophil maturation in the bone marrow of TKO mice was normal (Supplementary Fig. S8), suggesting that impaired recruitment of neutrophilic granulocytes into the intestinal lamina propria, rather than formation of these cells in the absence of immunoproteasomes, is responsible for the observed phenotype. We observed partially defective NF-κB and ERK signaling pathways in LPS/IFNγ-treated APCs lacking immunoproteasomes (Supplementary Fig. S9). These signaling cascades are involved in the transcription control of pro-inflammatory mediators that are associated with CAC development, and their activation is known to be dependent on optimal function of the proteasome/ubiquitin system (28). The NF-κB pathway was intact in thymocytes derived from naïve immunoproteasome-deficient mice, indicating that only inflammation-driven, but not tonic activity of NF-κB is affected in the absence of immunoproteasomes (Supplementary Fig. S9). Together, these findings demonstrate pro-tumoral activity of immunoproteasomes in the context of chronic inflammation.

Figure 4. Impaired recruitment of immune cells in the intestine of immunoproteasome-deficient mice during development of colitis-associated carcinogenesis.

(A-C) WT and immunoproteasome-deficient TKO mice were treated with two cycles of DSS after peritoneal AOM injection. On day 30 after AOM/DSS administration, colonic tissue was isolated and analyzed by flow cytometry and RT-qPCR. (A) Representative contour plots show the frequencies of IL17A⁺ and IFNγ⁺CD4⁺ T cells. (B) Representative contour plots and (C) bar graphs show the frequency of infiltrating neutrophils. Cells were gated on SSC^high granulocyte gate (B, C). One of two similar experiments with n=3 mice/group/experiment is shown. (D) Representative bar graphs (n=6 mice/group) show RT-qPCR analysis of colonic tissue analyzed on day 30 after AOM/DSS administration. One representative RT-qPCR analysis out of three experiments is shown. Data were analyzed by two-tailed unpaired Student’s t-test (*P=0.01–0.05, mean±s.e.m).

Patients with UC exhibit increased immunoproteasome-regulated mediators

Ulcerative colitis (UC) is a chronic inflammatory disease associated with high risk to develop colorectal cancer (CRC)(29). Because immunoproteasome-induced factors play a central role in both chronic colonic inflammation and CRC, we asked if we could find a signature of genes that were regulated by immunoproteasomes in human UC patients. We observed that UC patients with a severe course of disease displayed different patterns of chemokines in the inflamed colon compared to control tissue and Crohn’s disease (CD) patients. Microarray analysis revealed that in UC patients, but not in CD patients, the gene signature associated with recruitment of innate immune cells, including neutrophil-attracting chemokines CXCL1, CXCL2, and CXCL3, (which were downregulated in immunoproteasome-deficient mice) was pronounced (Fig. 5A). The genes characteristic for adenoma/adenocarcinoma were significantly enriched in UC patients compared to control colonic tissue (Fig. 5B). We also found a significant upregulation of immunoproteasome subunits in the chronically inflamed intestines of UC patients compared to control gut tissue (Fig. 5C).

Figure 5. Ulcerative colitis patients show an enrichment of adenocarcinoma and innate immune genes.

(A, B) Transcriptional analysis of colonic tissue derived from UC and CD patients compared to healthy controls (n=10 patients/group). (A) Heatmap showing the chemokine expression signature. Array intensity values were z-transformed and plotted. (B) Genes with an absolute fold-change of at least 2 in UC vs. control samples were probed by GSEA. Enrichment of adenocarcinoma genes enhanced in UC patients was compared to control group. Adenocarcinoma-associated genes for colon cancer were extracted from the publicly available gene set (see Methods). (C) Relative mRNA expression for immunoproteasome genes Psmb8, Psmb9, and Psmb10 in the large intestine of UC patients compared to control tissue assessed by RT-qPCR analysis (same cohort as in A, B). Data were analyzed by two-tailed unpaired Student’s t-test (**P=0.001–0.01, mean±s.e.m). (D, E) HT-29 colon cancer cells were stimulated with the indicated cytokines for 2 days in absence or presence of ONX 0914. (D) Representative bar graphs indicating gene expression of immunoproteasome gene Psmb8 and (E) chemokines Cxcl1, Cxcl2 and Cxcl3, obtained by RT-qPCR analysis. Three independent experiments were performed.

To test if malignant epithelial cells were able to upregulate the expression of immunoproteasomes, we treated HT-29 cells, a human colorectal adenocarcinoma cell line, with diverse stimuli. The assembly of immunoproteasomes is known to be accomplished by IFNγ treatment, but it remains unclear whether colon cancer cells are capable of expressing immunoproteasomes. Only IFNγ, but not other stimuli tested, was able to induce the expression of the immunoproteasome subunit LMP7 in HT-29 cells (Fig. 5D). To assess the relative expression of pro-tumorigenic chemokines in the presence of immunoproteasomes, we treated HT-29 cells with IFNγ, TNFα and IL1β. TNFα and IL1β are known to synergistically induce the expression of CXCL1–3 (30). We found that upregulation of all three chemokines was completely abrogated when HT-29 cells were treated with the immunoproteasome inhibitor ONX 0914 (Fig. 5E). Thus, immunoproteasomes are not only important regulators of TME, but they are also active in cancer cells to promote recruitment of innate immune cells and to support carcinogenesis. Our data provide insight into CAC pathogenesis and suggest that the treatment of patients with a severe course of UC with specific immunoproteasome inhibitors may reduce the high risk of developing colon cancer.

Discussion

The function of the immunoproteasome in cancer development is not as well understood as its role in regulating innate immunity and cytokine production. One report has demonstrated that deficiency of LMP7 in breast cancer cells suppresses tumor invasion and metastasis (31). Another study revealed that the high expression of immunoproteasomes in human melanoma is associated with a favorable response to treatment with immune checkpoint inhibitors (24). Although, the primary activity of immunoproteasomes has been attributed to optimal processing of viral and intercellular bacterial proteins for MHC class I antigen presentation, novel findings suggest that this enzymatic molecule is also an important regulator of the neo-antigen repertoire that can be more efficiently presented when the expression of immunoproteasome is higher (10).

In this study, we describe opposing roles for immunoproteasomes in association with shaping the TME. The immunoproteasome, which contains three proteolytic subunits LMP2, MECL-1, and LMP7, appears to play a crucial role in modulating immune responses during infection and inflammation (32). We show here that during chronic colonic inflammation, the immunoproteasome supports the progression of colonic tumors by elevating production of pro-tumorigenic factors in immune and cancer cells, as well as by recruiting innate immune cells into the inflamed gut. The capacity of immunoproteasomes to alter the gene expression of pro-inflammatory and pro-tumorigenic mediators likely results from the post-transcriptional regulation of key transcription factors such as NF-κB, STAT3, and IRF4 in immune cells (16,33), but also by the upregulation of immunoproteasome activity in the gut epithelium. The experiments with colon cancer cells support the novel concept that immunoproteasomes are not only active in immune cells but also in neoplastic epithelial cells during CAC development. IL6, IL1β, and NF-κB-controlled pro-tumorigenic chemokines CXCL1–3 were downregulated in TKO compared to WT animals, providing evidence that immunoproteasomes act as a functional link between chronic inflammation and cancer development in the large intestine. In contrast to inflammation-driven carcinogenesis, we observed that, in the absence of immunoproteasomes, the lack of efficient CTL-mediated anti-tumor immunity and reduced expression of APC-derived cytokines such as IL12 led to the enhanced growth of melanoma tumors. We also observed an upregulation of immunoproteasomes in melanoma cells, a phenomenon which was dependent on the presence of IFNγ in the TME. The immunogenicity and presence of neo-antigens are shown to be increased in human melanoma cells with high immunoproteasome expression (24). Collectively, the role of immunoproteasomes during development of carcinogenesis depends on the tumor-specific environment. By promoting recruitment and activation of immune cells, immunoproteasomes upregulate cytokine and chemokine networks that perpetuate inflammatory reactions leading to development of CAC. In contrast to the intestinal inflammatory milieu, the immunoproteasome has an anti-tumorigenic potential in melanoma tumors by supporting APC function and T cell-mediated anti-tumor immunity (Fig. 6). Our findings demonstrate that the role of immunoproteasomes in TME interactions should be considered in a more differentiated way. Future studies will be needed to elucidate the function of the immunoproteasome in different cancer types in order to develop adequate therapeutic strategies.

Figure 6. Schematic overview showing opposing activities of immunoproteasomes in two different cancer types.

Immunoproteasomes promote development of colon cancer associated with chronic inflammation by recruiting immune cells and by regulating the expression of pro-tumorigenic cytokines and chemokines in inflamed tissues. In contrast, in melanoma, immunoproteasomes are crucial anti-tumorigenic molecules supporting the function of APCs and T lymphocytes, leading to optimal CTL-mediated anti-tumor immunity.

Synopsis:

Data demonstrate that the immunoproteasome is essential for shaping the tumor microenvironment. Opposing pro- and anti-tumor functions of the immunoproteasome are identified, highlighting the importance of understanding its role in different cancer types.