Systemic but not MDSC-specific IRF4 deficiency promotes an immunosuppressed tumor microenvironment in a murine pancreatic cancer model

To study the role of IRF4 in pancreatic cancer we made use of a transplantable, orthotopic tumor model using the tumor cell line T110299, which originate from a genetically engineered, spontaneous mouse model with PDAC-characteristic driver mutations (Kras^G12D Tp53^R172H). We recently reported that this model is particularly enriched for myeloid cells such as MDSC [31]. Three weeks after orthotopic tumor induction, tumor weight of Irf4^−/− mice was significantly increased compared to IRF4 sufficient Irf4^flox/flox controls (Fig. 1a, b). Moreover, the survival of tumor-bearing Irf4^−/− mice was significantly reduced (Fig. 1c). We noted no significant differences in histologic tumor morphology between both genotypes (Fig. S1). T110299 tumors from both control and Irf4^−/− mice showed a moderately differentiated tubular to tubulo-papillary adenocarcinoma pattern with similar amounts of desmoplastic stroma and moderate chronic inflammatory infiltrates, resembling human pancreatic ductal adenocarcinoma. To study the immunological consequences of IRF4 deficiency, immune cell frequency of T110299 tumor-bearing mice was analyzed three weeks after tumor induction by flow cytometry. In the spleen of Irf4^−/− mice, PMN-MDSC, M-MDSC and CD4⁺ T cell frequencies were significantly increased, whereas the frequency of CD8⁺ T cells was reduced (Fig. 1d). In the TME, PMN-MDSC frequency was significantly increased in Irf4^−/− mice, whereas the frequency of CD8⁺ T cells was dramatically reduced, indicating a profound immunosuppressive TME in mice deficient for IRF4 (Fig. 1e) that may account for the more rapid tumor progression in these mice.

To confirm these findings in the human disease, the effect of IRF4 expression on the survival of metastasized PDAC patients was analyzed using the Cancer Genome Atlas (TCGA) data set. Metastasized PDAC patients were categorized based on their IRF4 expression level in two groups (upper and lower quartile) and survival of these selected groups was analyzed. The median survival of patients with low IRF4 expression was 15.9 months versus 18.7 months of patients with high expression levels. There was no significant difference in the overall survival in regard to the IRF4 expression (Fig. 1f). However, χ²-test of survival after 1 year revealed a significantly reduced survival of patients with lower IRF4 expression (χ² = 5.33, p = 0.02) (Fig. 1g).

To better understand the role of IRF4 in the initiation of the immunosuppressive and myeloid cell-enriched TME, IRF4 expression of both monocytic and polymorphonuclear cells in tumor-free and T110299 tumor-bearing mice was analyzed by flow cytometry. IRF4 was expressed homogenously in monocytic cells in blood, spleen as well as tumor, and was independent of the tumor status. In contrast, IRF4 expression in polymorphonuclear cells in blood, spleen and tumor was absent (Fig. 2a, b).

As this is in contrast to published data [29], we further validated this finding by alternative approaches and used conditional myeloid-specific IRF4 knockout mouse models. Irf4^flox mice have been designed to express green fluorescent protein (GFP) instead of IRF4 upon successful Cre-mediated recombination under the same physiological control as IRF4 [18]. Therefore, the conditional IRF4 knockout mice can also act as reporter mice. LysM^CreIrf4^fl/fl mice have an IRF4 deletion in polymorphonuclear cells, monocytes, macrophages and partly dendritic cells [32], whereas Ly6G^CreIrf4^fl/fl mice have the IRF4 deletion in polymorphonuclear cells only [33].

In line with our previous observation, M-MDSC of the LysM^CreIrf4^fl/fl mice were positive for GFP in the blood, spleen and tumor. No GFP expression was detectable in PMN-MDSC or M-MDSC of the Ly6G^CreIrf4^fl/fl mice. In contrast, in LysM^CreIrf4^fl/fl mice, a small fraction of approximately 15% of GFP⁺ PMN-MDSC was detectable in blood, spleen and tumor (Fig. 2c). As the GFP expression in those PMN-MDSC may derive from its precursor cell expressing LysM, we characterized GFP expression of GMP in LysM^CreIrf4^fl/fl mice. Indeed, 12% of GMP expressed GFP indicating that IRF4 is expressed early in the myeloid cell development (Fig. 2d). To further characterize the cell-specific expression pattern of IRF4 in myeloid cells, bone marrow cells were stimulated overnight with the known IRF4 inducers IL-4 and IL-13. As expected, CD11b⁺ CD11c⁺ MHC-II⁺ bone marrow-derived dendritic cells (BMDC) responded to cytokine stimulation by upregulating IRF4 expression. IRF4 expression was only inducible in CD11b⁺Ly6C^highLy6G⁻ M-MDSC-like cells, but not in CD11b⁺Ly6C^intLy6G⁺ PMN-MDSC-like cells (Fig. 2e). Taken together, the data demonstrate that IRF4 is expressed in M-MDSC and myeloid precursor cells but is absent in mature Ly6G-expressing PMN-MDSC.

GM-CSF-driven bone marrow cultures are an established model to study MDSC in vitro [14]. To evaluate the role of IRF4 in the development and function of MDSC, bone marrow cells of wild-type and Irf4^−/− mice were cultured for seven days in the presence of GM-CSF, and cell composition was analyzed. The frequency of MDSC-like (Gr1⁺MHC-II^low) cells was significantly increased in bone marrow cultures from Irf4^−/− mice, while both MHC-II^int and MHC-II^high BMDC were significantly reduced (Fig. 3a, b). Next, T cell suppressive capacity of bone marrow-derived MDSC was measured in T cell co-cultures in the presence of anti-CD3/anti-CD28 beads. T cell proliferation was measured and revealed that IRF4-deficient bone marrow culture cells suppressed the proliferation of both CD4⁺ and CD8⁺ T cells more potently, as compared to wild-type controls (Fig. 3c). However, as the difference in MDSC composition may influence the overall suppressive activity, FACS-sorted Gr1^highMHC-II^low MDSC-like cells from wild-type and Irf4^−/− mice were individually analyzed in a T cell suppression assay. Sorted Gr1⁺ MDSC-like cells of wild-type and Irf4^−/− mice exhibited similar T cell suppressive capacities, arguing against a direct role of IRF4 in controlling the suppressive capacity of MDSC-like cells (Fig. 3d).

The two conditional IRF4 knockout mouse models described above were used to study the intrinsic role of IRF4 in myeloid cells in vivo. KPC-derived T110299 tumors were orthotopically induced in Ly6G^CreIrf4^fl/fl mice, LysM^CreIrf4^fl/fl mice as well as Irf4^fl/fl control mice (Fig. 4a). While there was no significant difference in tumor growth between Ly6G^CreIrf4^fl/fl and control mice, tumor weight was significantly increased in LysM^CreIrf4^fl/fl mice, as compared to Irf4^fl/fl control mice (Fig. 4b). Flow cytometric analysis of MDSC frequencies three weeks after tumor induction showed no difference in PMN-MDSC frequency in blood, spleen and tumor of LysM^CreIrf4^fl/fl mice and Ly6G^CreIrf4^fl/fl mice compared to Irf4^fl/fl controls; however, M-MDSC frequency was moderately increased in the spleen of LysM^CreIrf4^fl/fl mice (Fig. 4c). Compared to Irf4^fl/fl controls, there was no significant difference in CD4⁺ or CD8⁺ T cell frequencies in LysM^CreIrf4^fl/fl mice or Ly6G^CreIrf4^fl/fl mice (Fig. 4d). Of note, both myeloid-specific IRF4-deficiencies had no influence on the survival of PDAC-bearing mice (Fig. 4e).

In conclusion, this study demonstrates that only early myeloid precursor cells and M-MDSC express IRF4, whereas mature PMN-MDSC lack IRF4 expression. In our hands, specific deletion of IRF4 in PMN-MDSC has no influence on PMN-MDSC expansion, their T cell suppressive capacity, tumor growth or survival. In contrast, myeloid cell-specific deletion of IRF4 in LysM^CreIrf4^fl/fl mice slightly accelerates tumor growth, which was accompanied by increased M-MDSC frequency in the spleen. No impact on PMN-MDSC numbers or survival is seen in these mice. These findings argue against a pro-tumoral effect of IRF4 in PMN-MDSC. Hence, as observed in mice with global IRF4 deficiency, the IRF4-mediated effects in other immune cells likely account for the immunosuppressive TME with its dense MDSC infiltration.

C57BL/6 wild-type mice were purchased from Janvier, France. Irf4^flox mice (B6.129S1-Irf4^tm1Rdf/J) were a kind gift from Prof. Bopp (Institute of Immunology, Universitätsmedizin Mainz), Ly6G^Cre mice (C57BL/6-Ly6g^{(tm2621(Cre−tdTomato)}Arte)) were a kind gift from Prof. Gunzer (Institute for Experimental Immunology and Imaging, University of Duisburg-Essen), LysM^Cre mice (B6.129P2-Lyz2^tm1(cre)Ifo/J) were a kind gift from PD Dr. Lech (Institute of Clinical Biochemistry, Klinikum der Universität München). FLP1 recombinase expressing FLPe mice (B6;SJL-Tg^{(ACTFLPe)9205Dym)}/J) were purchased from Jackson Laboratory (Sulzfeld, Germany). All mice were kept on C57BL/6 background with a 12-h light/dark cycle, water ad lib. and regular chow diet (sniff, Soest, Germany) at the Klinikum der Universität München, Munich, Germany. Experiments were performed according to national ethical guidelines approved by the local government (Regierung von Oberbayern, Munich, Germany; file number 55.2-1-54-2532-175-12). LysM^Cre were cross-bred with Irf4^flox mice to obtain LysM^CreIrf4^flox, and Ly6G^Cre were cross-bred with Irf4^flox mice to obtain Ly6G^CreIrf4^flox. Both mouse strains were kept on homozygous Irf4^fl/fl background. Exons 1 and 2 of Irf4 in B6.129S1-Irf4^tm1Rdf/J mice are flanked by two FRT sites. To generate global IRF4-deficient mice, B6;SJL-Tg^{(ACTFLPe)9205Dym}/J mice were cross-bred with B6.129S1-Irf4^tm1Rdf/J mice and, as described before, Irf4^−/− mice were obtained [18]. IRF4 sufficient mice originating from those breedings were used as littermate controls. Genotypes of all mice were routinely analyzed by PCR.

DNA from ear or tail biopsies was extracted and analyzed as described before [35]. Briefly, biopsies were incubated in 75 µl alkaline lysis buffer (25 mM NaOH, 0.2 mM EDTA in H₂O) for 30 min at 95 °C and reaction was stopped by adding 75 µl neutralization buffer (40 mM Tris-HCl in H₂O). Supernatant containing genomic DNA was subsequently used for genotyping with locus specific primer pairs listed in supplementary table 1 by using genotype-specific cycling programs, as summarized in supplementary table 2.

Primary cells were cultured in RPMI-1640 medium (Sigma-Aldrich, Taufkirchen, Germany), supplemented with 10% fetal calf serum (FCS), 2 mM l-glutamine, 100 U/l penicillin, 0.1 mg/ml streptomycin, 100 mM non-essential amino acids (all gibco®, Thermo Fisher Scientific, Karlsruhe, Germany), 1 mM sodium pyruvate and 50 mM 2-mercaptoethanol (both Sigma Aldrich). T110299 tumor cells had been isolated from tumors of genetically-engineered Ptf1a-Cre Kras^G12D p53^fl/R172H (KPC) mice and kindly provided by Prof. Siveke (West German Cancer Center (WTZ), University Hospital Essen). T110299 cells were cultured in DMEM high glucose media (Sigma-Aldrich), supplemented with 10% FCS, 2 mM l -glutamine, 100 U/l penicillin and 0.1 mg/ml streptomycin (all gibco®). All cells were kept in a humidified incubator at 37 °C and 5% CO₂ and were regularly tested for mycoplasma contamination.

Bone marrow cells were isolated by flushing femur and tibia. 2 × 10⁶ cells per 10 ml were seeded in a 10 cm cell culture round plate in primary cell medium supplemented with 20 ng/ml GM-CSF (Peprotech, London, United Kingdom). After three days, 10 ml primary cell medium supplemented with 20 ng/ml GM-CSF was added. Five days after cell isolation, 66% of the medium containing 20 ng/ml GM-CSF was exchanged. If indicated, cells were stimulated overnight with 10 ng/ml IL-4 and 10 ng/ml IL-13 (both Peprotech, London, United Kingdom).

Orthotopic tumors were induced by surgical implantation, as described before [36]. Briefly, 6–12 weeks old mice were anesthetized, and by surgical incision of the skin and peritoneum, the pancreas was carefully mobilized. After the injection of 2 × 10⁵ T110299 cells in 25 µl PBS, the pancreas was relocated, and the incision was closed by surgical suture. Mice were monitored daily and distressed mice were sacrificed. Tumor weight of sacrificed animals was measured and normalized to average tumor weight of WT animals in the respective experiment.

Spleens and tumors were isolated from the mice and blood was drawn. Spleens were processed through a 70 µm cell strainer. Erythrocytes from spleen and blood were removed using the red blood cell lysis buffer (BD Biosciences, Heidelberg, Germany). Tumor tissue was minced into pieces and mechanically dissociated using the mouse Tumor Dissociation Kit with the gentleMACS™ Dissociator application (both Miltenyi Biotech, Bergisch Gladbach, Germany), according to the manufacturer´s instructions. The cell suspension was separated from tissue debris by sequentially using 100 µm and 70 µm cell strainers. For functional assays, untouched T cells were isolated using the Pan T cell isolation Kit II, and for MDSC isolation the Myeloid-Derived Suppressor Cell Kit was used (both Miltenyi Biotec). In brief, in a two-step separation process, PMN-MDSC (CD11b⁺ Ly6C^int Ly6G⁺) were isolated with anti-Ly6G beads followed by M-MDSC (CD11b⁺ Ly6C^high Ly6G⁻) isolation using anti-Gr1 beads. The purity of isolated cells was > 95% for T cells and between 75 and 90% for MDSC.

Cells from bone morrow cultures were isolated by FACSorting. Cells were stained as described for FACS analysis. FVD^negGr1^highMHC-II^low and FVD^negGr1^lowMHC-II^high were sorted on a BD Aria III system (BD Bioscience, Heidelberg, Germany).

Prior to fluorochrome staining of single cell suspensions, FcR III/II blocking was performed using the anti-CD16/CD32 TrueStain fcX™ antibody (BioLegend, London, UK). Dead cells were stained for exclusion with fixable viability dye (FVD) (Thermo Fisher Scientific, Karlsruhe, Germany). For cell-specific surface staining, cells were labeled with CD4 (clone GK1.5), CD8 (clone 53–6.7), CD11b (clone M1/70), CD11c (clone N418), CD45 (clone 30-F11), Gr1 (clone RB6-8C5), Ly6C (clone HK1.4), Ly6G (clone 1A8), MHC-II (clone AF6-120.1; all BioLegend, London, UK). IRF4 (clone IRF4.3E4; BioLegend, London UK) or isotype control (clone RTK2071; both BioLegend, London, UK) were stained intracellularly using the one-step intracellular staining protocol of the eBioscience™ FoxP3/Transcription Factor Staining Buffer Kit (Thermo Fisher Scientific, Karlsruhe, Germany). Samples were acquired on a BD LSRFortessa system (BD Bioscience, Heidelberg, Germany), and data were analyzed with FlowJo X software (FLOWJO LLC, Ashland, OR, USA).

Isolated T cells were stained with 2.5 µM Carboxyfluorescein succinimidyl ester (CFSE; Thermo Fisher Scientific) in PBS for 4 min at room temperature and reaction was stopped with FCS. For the assessment of MDSC suppressive capacity, MDSC were co-cultured with anti-CD3/anti-CD28 stimulated CFSE-labeled T cells. For this, 5 × 10⁴ T cells (per well) were seeded into 96-well plates and cocultured with 0.31 × 10⁴, 0.63 × 10⁴, 1.25 × 10⁴, 2.5 × 10⁴ or 5 × 10⁴ MDSC. Each well was supplemented with 1 µl beads (Dynabeads™ Mouse T-Activator CD3/CD28, gibco®, Thermo Fisher Scientific, Karlsruhe, Germany). After 72 h, CFSE dilution of CD4⁺ and CD8⁺ T cells was analyzed by flow cytometry (BD Canto II system, BD Bioscience, Heidelberg, Germany). Unstimulated CFSE-labeled T cells only were used to set the threshold of proliferated T cells (CFSE^low).

Survival data, IRF4 expression level as well as information on metastasis status of PDAC patients from the TCGA cohort were retrieved via the Xena browser [37]. Two groups of patients with metastasized PDAC were analyzed: Patients with low IRF4 expression level (≤ the lower quartile) and patients with high IRF4 expression level (≥ the upper quartile). The survival of these two groups was analyzed in a Kaplan–Meier analysis and compared with a Wilcoxon test. The survival status after one year was displayed in a contingency table. As all expected cell frequencies were above five, χ² test was used to compare one-year survival of the two groups. All analysis on the TCGA data set was performed using IBM® SPSS® Statistics 25. Graphs were generated in Graphpad Prism 8.3.

Tumors were embedded in Tissue-Tek® O.C.T.™ (Sakura Finetek GmbH, Staufen, Germany), rapidly frozen in liquid nitrogen and stored at − 80 °C. Prior to analysis, samples were thawed, once washed in PBS, fixed in ROTI^®Histofix 4% paraformaldehyde (Roth, Karlsruhe, Germany) and subjected to automated routine histological tissue processing. After paraffin embedding, four µm thick whole tissue sections were stained using haematoxylin-eosine (HE) staining (haematoxylin: Waldeck, Münster, Germany; eosine: Sigma) in an automated tissue stainer (Tissue Tek Prisma, Sakura Finetek).

Data represent individual mice and are displayed as mean with standard error of the mean (SEM). To test for statistically significant differences between two groups, student’s t test (for expression data only) or Mann–Whitney U test was used. To compare more than two groups, we applied Kruskal–Wallis test followed by Dunn’s multiple comparison post hoc tests between selected samples. To analyze differences in the survival, Mantel–Cox Logrank test was conducted. For the statistical analysis of suppression assays, only the difference between the conditions with the highest concentration of MDSC was analyzed using a Mann–Whitney U test. Statistical analysis of murine data was performed in Graphpad Prism 8.3.