Tumor-infiltrating myeloid cells induce tumor cell resistance to cytotoxic T cells in mice (original) (raw)
Effect of PNT on tumor cell recognition by CTLs. To evaluate the effect of PNT on tumor cells, we used either the PNT donor 3-morpholinosydnonimine hydrochloride (SIN-1) or PNT. We selected the doses of the compounds that caused modest increases in the level of NT in tumor cells without substantial toxicity to EL-4 target cells. PNT at 0.1 mM caused less than 10% EL-4 cell death, with a clearly detectable increase in the NT level (Supplemental Figure 1; supplemental material available online with this article; doi:10.1172/JCI45862DS1). As effector cells we used OT-1 transgenic CD8+ T cells recognizing the chicken OVA–derived peptide SIINFEKL in the context of H-2Kb. EL-4 target cells were treated with either SIN-1 or PNT for 10 minutes, washed, loaded with control or OVA-derived specific peptides, and then used in cytotoxicity assays. Pre-treatment of EL-4 cells with SIN-1 (Figure 1A) or PNT (Figure 1B) significantly (P < 0.01) reduced the ability of OT-1 CTLs to kill the OVA peptide–loaded targets as measured in a chromium release assay. To confirm these findings, we employed an additional cytotoxicity assay, which is based on the flow cytometric analysis of the proportion of specific target cells remaining in culture after incubation with CTLs. PNT significantly (P < 0.05) reduced the CTL killing of target cells loaded with specific (OVA) but not with a control irrelevant peptide (Figure 1C). The effect described above was observed only when the tumor cells were treated with SIN-1 or PNT prior to addition of specific peptide. If target cells were first loaded with the peptide and then treated with SIN-1 or PNT, the CTL-mediated killing was not affected (Figure 1, D–F). To assess the possible effect of PNT on CTL recognition of naturally processed antigens, we used EG-7 tumor cells (EL-4 cells transduced with OVA), which are recognized by OVA-specific OT-1 CTLs (Figure 1, G–I). SIN-1 and PNT treatment of EG-7 cells significantly (P < 0.01) reduced their susceptibility to be lysed by CTLs (Figure 1, G–I). To extend these findings to a different experimental system, we used B16-F10 melanoma cells and Pmel-1 transgenic CD8+ T cells that recognize an H-2Db–restricted epitope from the melanosomal antigen gp100 corresponding to amino acid positions 25–33 expressed in B16-F10 cells (31). Pre-treatment of B16-F10 target cells with PNT substantially reduced the ability of activated CTLs to kill tumor cells (Supplemental Figure 2).
PNT makes tumor cells resistant to CTL-mediated lysis. (A) NO donor (1 hour pre-treatment with 1 mM SIN-1) inhibited killing of EL-4 cells subsequently washed and loaded with specific peptides (SP) by CTLs in chromium release assay. CP-EL-4 cells loaded with control peptide. (B) Pre-treatment of EL-4 cells with 0.1 mM PNT for 10 minutes inhibited killing of target cells by CTLs. In A and B, 3 experiments in duplicate were performed with similar results. Mean ± SEM of 1 experiment is shown. (C) Killing of EL-4 cells that were labeled with 2 doses of CFSE (high and low) and pre-treated with 0.1 mM PNT by CTLs. After washing EL-4 cells were loaded with a SP (high dose) or CP (low dose). Target cells were mixed at 1:1 ratio and were incubated with OT-1 T cells for 5 hours. Data are representative results of 3 experiments. Mean ± SEM of 3 experiments *P < 0.05. (D–F) Experiments were performed as described in A–C, except that EL-4 target cells were first loaded with SP or CP and then treated with SIN-1 or PNT. (D and E) Three experiments in duplicate were performed, with similar results. Mean ± SEM of 1 experiment is shown. (F) Cumulative data (mean ± SEM) of 3 experiments are shown. (G–I) Experiments were performed essentially as described in A–C, except that EG-7 cells were used as targets instead of peptide-loaded EL-4 cells. (G and H) Three experiments in duplicate were performed, with similar results. Mean ± SEM of 1 experiment is shown. (I) Cumulative data (mean ± SEM) of 3 performed experiments are shown. *P < 0.01.
The above-described results suggested that PNT might affect the binding of specific peptides to MHC class I on tumor cells. To directly test this hypothesis, EL-4 cells were treated with PNT, washed, and then loaded with H-2Kb–matching OVA-derived specific or control peptides followed by staining with an Ab recognizing the OVA-specific epitope in the context of H-2Kb. Pre-treatment of tumor cells with PNT substantially reduced the peptide binding and formation of the pMHC complex as measured by flow cytometry, whereas PNT had no effect on the pMHC complexes when tumor cells were treated with PNT after the peptide loading (Figure 2A). Treatment with PNT did not affect the expression of the H-2Kb molecule on the tumor cells when applied either before or after loading with the peptide (Figure 2B). We also tested the effect of SIN-1 on binding of two distinct survivin-derived HLA-A2 peptide epitopes (32) to human T2 target cells. Pre-treatment of the T2 cells with SIN-1 significantly reduced binding of the both tested peptides (Figure 2C). Next, we evaluated whether direct NT modification of peptide epitope would reduce its binding capacity to MHC. The results showed that nitration of the tyrosine in position 3 of a peptide epitope derived from telomerase reverse transcriptase (TERT) significantly reduced its binding to HLA-A2 in T2 cells (Figure 2D), demonstrating that nitration of either the MHC molecule or the peptide epitope can affect binding. Thus, PNT alters binding of the peptides to the MHC class I on the surface of tumor cells by impacting on the formation of pMHC complexes, diminishing effective recognition of the tumor cells by CTLs. However, in natural circumstances the pMHC complexes are assembled intracellularly in the ER and subsequently are transported to the cell surface for presentation to CTLs. Our data indicated that PNT blocked recognition by CTLs of the naturally processed peptide in EG-7 and B16-F10 cells. However, our results with exogenous peptide pulsing indicate that PNT did not affect the already-made pMHC complexes. These observations suggest that PNT exerts its main effect in EG-7 and B16-F10 cells in the recognition of naturally processed antigen during the assembly of the pMHC complex. To test this hypothesis, we used Lewis lung carcinoma (LLC) and B16-F10 melanoma cells transfected with a single-chain H-2Kb-SIINFEKL construct. In these cells the H-2Kb-SIINFEKL pMHC complex is synthesized as a fusion protein that does not require antigen processing and epitope assembly (33). The cells were treated with PNT at different concentrations, and binding of the anti-pMHC Ab was evaluated. At a concentration of 0.3 mM, PNT caused more than a 10-fold increase in the level of NT staining of tumor cells (Figure 2, E and F). However, PNT treatment did not affect the pMHC expression on the cell surface (Figure 2, E and F). Moreover, treatment of LLC single-chain H-2Kb-SIINFEKL cells with PNT did not affect their killing by OVA-specific CTLs (Figure 2G). These data support the hypothesis that PNT affects the formation of pMHC complexes, preventing CTL killing of the tumor cells. In contrast, PNT did not affect the ability of NK cells to kill their targets (Supplemental Figure 3).
PNT affects binding of the peptides to MHC class I. (A) Pre-treatment of target cells with PNT decreased binding of the specific peptide. EL-4 cells were treated with 0.1 mM PNT before or after loading with specific or control peptides at the indicated concentrations. Specific peptide loaded on MHC class I was detected by florescence-conjugated anti-SIINFEKL bound to H-2Kb. Typical result of 1 of 5 performed experiments is shown. (B) Effect of PNT treatment on the expression of MHC class I (H-2Kb) molecules on EL-4 tumor cells. The MFI of H-2Kb expression is shown. Data represent mean ± SEM from 3 performed experiments. (C) Effect of pre-treatment of T2 human cells with SIN-1 on the binding of HLA-A2–matching human survivin-derived peptides. “Background” indicates T2 cells incubated without peptides. Mean ± SEM of 3 experiments is shown. P < 0.05, untreated versus SIN-1–treated cells for each experimental point. (D) Binding of non-modified HLA-A2–matched human TERT-derived PVYAETKHFL and nitrated PVY(NO2)AETKHFL peptides to T2 cells. Mean ± SEM of 3 experiments is shown. P < 0.05, non-modified versus nitrated peptides for each experimental point. (E and F) PNT does not affect expression of pMHC on cells expressing single-chain H-2Kb-SIINFEKL protein. LLC (E) or B16-F10 (F) cells expressing single-chain H-2Kb-SIINFEKL were treated with PNT at indicated concentrations and then labeled with anti-NT or pMHC Abs. For each cell line, 2 experiments with the same results were performed. (G) PNT does not affect CTL killing of tumor cells expressing single-chain H-2Kb-SIINFEKL protein. Experiment was performed as described in F. LLC and LLC-H-2Kb-SIINFEKL cells were mixed at a 1:1 ratio and were untreated or pretreated with PNT, then used as targets for CTLs. Two experiments were performed, and cumulative results are shown.
Myeloid cells are the principal producers of PNT in lung cancer and induce tumor cell resistance to CTLs. Next, we investigated what cells could release PNT into the tumor microenvironment. We evaluated tumor tissues from 15 patients with lung adenocarcinoma, 2 patients with large cell lung carcinoma, 8 patients with breast ductal carcinoma, and 13 patients with pancreatic ductal carcinoma. All subjects underwent surgical resection of the primary tumors between 2008 and 2010. Immunohistochemistry of NT staining was used as a marker of PNT production. Pancytokeratin AE1/AE3 and CD33 Abs were used to confirm the epithelial and myeloid origins of the cells, respectively. Typical staining examples are shown in Supplemental Figure 4. NT measurements in tumor and myeloid cells were scored on a 4-level scale (0 to 3): 0: no positive cells; 1: few slightly positive cells; 2: less than 50% positive cells; 3: almost all cells positive. In lung cancer patients, the rate of NT staining in myeloid cells (2.07 ± 0.21) was significantly higher than in tumor cells (0.58 ± 0.23, P = 0.004) or normal epithelial cells (0.76 ± 0.20, P = 0.008). Similar results were obtained in pancreatic cancer patients: myeloid cells (1.69 ± 0.21), tumor cells (0.15 ± 0.15, P = 0.001), epithelial cells (0.69 ± 0.20, P = 0.02). In breast cancer patients, the NT staining rate in myeloid cells was also higher than in tumor cells (2.25 ± 0.25 vs. 1.5 ± 0.38, P = 0.06). No difference was found in NT staining rates between myeloid and normal epithelial cells in breast cancer patients (2.1 ± 0.22, P = 0.36). Thus, myeloid cells were the predominant source of NT in lung, pancreatic, and breast cancer patients, while the normal ductal epithelial cells adjacent to breast cancer also showed moderate to strong NT positivity.
In LLC and EL-4 tumor tissues, practically all of the PNT production was associated with Gr-1+ cells, which are primarily MDSCs, and to a lesser extent with F4/80+ macrophages (Figure 3A). These results raised the question as to whether myeloid cells were able to cause tumor cell resistance to CTLs. Gr-1+ MDSCs and F4/80+ macrophages were isolated from spleens or tumor tissues of tumor-bearing mice and incubated with EL-4 tumor cells overnight. After that time, the myeloid cells were removed, and binding of the fluorescently labeled SIINFEKL peptide to the tumor cells was evaluated. Gr-1+ immature myeloid cells (IMCs) from the spleen of naive mice did not affect peptide binding to the tumor cells, whereas MDSCs from spleen or tumor tissues had a significant inhibitory effect on the peptide binding to tumor cells (Figure 3B). The effect of macrophages on the peptide binding was much smaller and statistically not significant (Figure 3B). No effect of the myeloid cells on the expression of MHC class I on the surface of the tumor cells was detected (data not shown). To assess the ability of MDSCs to cause tumor cell resistance to CTLs, we preincubated EL-4 cells overnight with either MDSCs or control Gr-1+ IMCs. Myeloid cells were then removed by bead purification, and the tumor cells were loaded with specific or control peptides and used as targets in a lysis CTL assay. Preincubation of tumor cells with MDSCs from spleens or tumors of EL-4 tumor–bearing mice did not affect the killing of EL-4 target cells loaded with control peptide but significantly reduced the killing of specific peptide–loaded target cells as compared with tumor cells preincubated with IMCs (Figure 3C).
MDSCs caused tumor cell resistance to CTLs. (A) NT (brown) and Gr-1+ or F4/80+ (red) staining in LLC and EL-4 tumors. Scale bars: 100 μm. (B) MDSCs reduced MHC class I binding ability to specific peptide (4 μg/ml) of EL-4 cells after overnight culture at a 1:1 ratio with indicated myeloid cells. Different concentrations of peptide were tested and showed similar results. PNT was used as a positive control. Percentage of change from MFI in untreated EL-4 cells set as 100% is shown. Spl, spleen; Tu, tumor; IMC, IMCs from the spleen of naive mice. Data are mean ± SEM from 4 experiments. *P < 0.05 versus control. (C) MDSCs inhibit CTL killing of target tumor cells after overnight incubation. Myeloid cells were removed, and then EL-4 cells were used as targets in CTL assay as described in Figure 1C. Result of 1 typical experiment and cumulative data (mean ± SEM) of 3 performed experiments are shown. *P < 0.05 versus IMC. (D) PNT caused nitration of tyrosine in MHC class I molecule (H-2Kb) in tumor cells. Immunoprecipitation of whole cell lysates from EL-4 cells treated with PNT was performed as described in Methods. Lysates precipitated with IgG showed no bands (not shown). (E and F) Colocalization of MHC class I and NT in tumor cells. (E) EL-4 cells treated with PNT and stained as indicated. (F) EL-4 cells were stained with blue trackers and cultured with MDSCs as described in B, myeloid cells were removed, and EL-4 cells were stained. Scale bars: 10 μm. Two experiments with the same results were performed.
To confirm that the PNT and MDSCs can indeed cause tyrosine nitrosylation of the MHC class I molecules in tumor cells, we cultured EL-4 cells with PNT and prepared whole cell lysates. H-2Kb molecules were immunoprecipitated, and membrane blots were probed with an anti-NT Ab. NT+ staining was detected only in cells treated with PNT (Figure 3D). These results were further confirmed by confocal microscopy, where colocalization of MHC class I and NT was found on the tumor cells treated with PNT (Figure 3E). The expression of NT on tumor cells was also detected after overnight incubation with MDSCs but not with the control IMCs. Colocalization of NT with MHC class I was readily detectable (Figure 3F).
The fact that PNT produced by MDSCs caused modification of MHC molecules on tumor cells raised the question of why MDSCs themselves were able to present peptides to T cells and cause T cell tolerance (28, 34). It is known that neutrophils and macrophages have an elaborate system of antioxidants that protect them from excesses of ROS and RNS (35). To test this hypothesis directly, we treated EL-4 cells and MDSCs side-by-side with the same amount of PNT (0.1 mM). PNT did not affect the expression of MHC class I (H-2Kb) on EL-4 cells or MDSCs. However, it dramatically reduced binding of SIINFEKL peptide to EL-4 cells but not to MDSCs (Figure 4A).
Effect of MDSCs on the protection of tumor cells from CTLs depends on ROS production. (A) MDSCs are resistant to PNT. MDSCs or EL-4 cells were treated for 10 minutes with 0.1 mM PNT. The levels of H-2Kb expression and NT after treatments were examined by flow cytometry (left panels). Cells were labeled with SIINFEKL peptide (SIIN) conjugated with FITC; peptide binding was performed by flow cytometry, and MFI was compared. Four experiments with similar results were performed. (B) Binding of FITC-SIINFEKL peptide was measured in EL-4 cells incubated with untreated MDSCs isolated from EL-4 tumor–bearing mice, in the presence of 250 nM CDDO-Me, or MDSCs isolated from EL-4 tumor–bearing gp91phox–/– mice. As a control EL-4 cells were treated with CDDO-Me (250 nM). The background was set as 100% of peptide binding to untreated EL-4 cells. The graph shows the percentages of changes in MFI compared with background. Different concentrations of peptide were tested and showed similar results. One concentration, 4 μg/ml, is shown. Mean ± SEM of 4 experiments is shown. *P < 0.05 versus background. (C) EL-4 cells isolated from culture with MDSCs were loaded with control or specific peptides as target cells in CTL assays described in Figure 1C. MDSCs were from EL-4 tumor–bearing mice treated for 5 days with 150 mg/kg CDDO-Me or control diets. Representative 1 experiment and mean ± SEM of 3 performed experiments are shown. *P < 0.05 versus IMC-treated EL-4 cells. (D and E) Binding of peptides to tumor cells after treatment with CDDO-Me. EL-4 tumors were established in congenic (CD45.1+) mice. Mice were treated with control or CDDO-Me diets for 5 days. EL-4 (CD45.2+) cells were isolated by magnetic beads, and H-2Kb expression on tumor cells was examined by flow cytometry (D). Peptide binding was analyzed by incubation with FITC-SIINFEKL and evaluated by flow cytometry (E). Data are representative of 3 experiments with similar results.
PNT is a product of the interaction between NO and superoxide. To establish a causal relationship between the PNT production by MDSCs and their effect on tumor cells, we used two experimental approaches. First, MDSCs were generated from tumor-bearing mice lacking gp91phox, a component of the NADPH complex responsible for the generation of ROS. MDSCs from these mice are not able to produce ROS in response to various stimuli (36). Second, we used the triterpenoid 2-cyano-3,12-dioxooleana-1,9(11)-dien-28-oic acid methyl ester (CDDO-Me; bardoxolone methyl). CDDO-Me is a compound that previously was demonstrated to inhibit ROS and PNT production by MDSCs (37). Both gp91phox–/– MDSCs and MDSCs treated with CDDO-Me were unable to inhibit the peptide binding to MHC molecules by EL-4 tumor cells (Figure 4B). Treatment of tumor cells with CDDO-Me alone did not affect the peptide binding (Figure 4B). The CDDO-Me–treated MDSCs demonstrated significantly decreased protection of tumor cells from killing by CTLs (Figure 4C). To assess the possible effect of CDDO-Me in vivo, we established EL-4 tumors (CD45.2+ cells) in congenic CD45.1+ mice. The tumor-bearing mice (s.c. tumor diameter, 1.5 cm) were treated for 5 days with CDDO-Me, CD45.2+ tumor cells were isolated, and specific peptide binding was measured. Short treatment with CDDO-Me did not substantially affect tumor size (data not shown) or expression of MHC class I on tumor cells (Figure 4D). However, tumor cells from mice treated with CDDO-Me had a substantially higher binding of the peptide than those from control untreated mice (Figure 4E). We asked whether treatment of mice with CDDO-Me made tumor cells more susceptible to CTL killing. Two tumor models with defined antigens were used: EG-7 thymoma, recognized by OT-1 CTLs, and B16-F10 melanoma recognized by gp100-specific Pmel-1 CTLs. Mice with established s.c. tumors were treated with CDDO-Me for 5 days, followed by tumor resection. Tumor cells were then used as targets in CTL assays. CDDO-Me treatment did not affect expression of MHC class I on tumor cells (Figure 5A). However, it enhanced CTL-mediated killing of both EG-7 (Figure 5, B and D) and B16-F10 (Figure 5, C and D) tumors. Thus, MDSCs generated in tumor-bearing mice produce large amounts of PNT, which affects the binding of the peptides to MHC class I on tumor cells, resulting in a decreased sensitivity of the tumor cells to CTL lysis.
Effect of CDDO-Me treatment on CTL recognition of tumor cells. B16-F10 and EG-7 tumors were established s.c. in congenic (CD45.1+) C57BL/6 mice. When tumors reached 1 cm in diameter, mice were treated with CDDO-Me diet for 5 days. B16-F10 tumor cells were isolated after collagen digestion and negative selection using anti-CD45 Abs and magnetic beads. EG-7 cells were isolated using anti-CD45.2 Ab and magnetic beads. Cells were then labeled with CFSE and used for CTL assay. (A) MHC class I (H-2Kb) expression in tumor cells isolated from nontreated (solid line) and CDDO-Me–treated (dotted line) mice. (B) CTL assay with tumor cells isolated from EG-7 tumor–bearing mice. Targets: EG-7 (high CFSE dose) and EL-4 (control, low CFSE dose); effector cells: OT-1 CTLs. (C) CTL assay with tumor cells isolated from B16-F10 tumor–bearing mice. Targets: B16-F10 (high CFSE dose) and LLC (control, low CFSE dose); effectors cells: pmel-1 CTLs. (D) Cumulative results of the experiments. Mean ± SD is shown. Each group included 2–3 mice. *P < 0.05.
Inflammation and tumor cell resistance to CTLs. To assess the biological significance of these findings, we modeled enhanced inflammatory conditions in vivo using the secreted proinflammatory cytokine IL-1β. We established several LLC cell lines: cells expressing OVA (LLC-OVA), cells expressing the secreted form of IL-1β (LLC-IL-1β), and cells expressing both OVA and IL-1β (LLC-IL-1β-OVA). Production of IL-1β was verified by ELISA of tumor cell supernatants. The IL-1β–transfected cell lines produced 200–300 pg/ml/24 hours, whereas in untransfected cells the IL-1β production was below the detectable level (Supplemental Figure 5A). Overexpression of OVA or IL-1β did not affect proliferation of tumor cells (Supplemental Figure 5B) or their ability to form colonies in semisolid medium (Supplemental Figure 5C). The expression of IL-1β by LLC-OVA cells did not affect their killing in vitro by OVA-specific CTLs (Supplemental Figure 6). However, the IL-1β–producing LLC cells grew faster than control LLC cells after s.c. injection into syngeneic C57BL/6 mice (Supplemental Figure 7). Therefore, we adjusted the concentration of tumor cells to provide for a comparable rate of tumor growth between the parental and IL-1β–producing LLC cells. Even when it was adjusted to achieve an equal tumor size, the IL-1β–producing tumors induced a much higher number of MDSCs in spleens and in tumor tissues than their parental cells (Figure 6A). Although expansion of macrophages was observed in the spleen, no differences in the proportion of tumor-associated macrophages was seen between the LLC-OVA and LLC-IL-1β-OVA tumor–bearing mice (Figure 6B). MDSCs generated in mice bearing LLC-OVA and LLC-IL-1β-OVA tumors did not differ in the production of ROS (Figure 6C). The level of NO production was lower in MDSCs from the LLC-IL-1β-OVA mice than from the LLC-OVA mice (Figure 6D). However, large numbers of NT+ MDSCs and macrophages were detected in the tumor tissues of the LLC-IL-1β-OVA–bearing mice (Figure 6, E and F, as compared with Figure 3A). Thus, overexpression of IL-1β in tumor cells resulted in the accumulation of a large number of PNT-producing MDSCs in tumor tissues.
Experimental model of tumor-associated inflammation. (A and B) LLC-OVA or LLC-IL-1β-OVA tumors were established in C57BL/6 mice. The percentages of MDSCs (A) and macrophages (B) were determined in spleens and tumors by flow cytometry. Data are mean ± SEM for 3 experiments. *P < 0.05. (C) Measurement of ROS in splenic MDSCs using the oxidation-sensitive dye DCFDA. Cells were incubated with DCFDA (2 μM) with or without PMA (300 nM) for 30 minutes in serum-free media. Cells were then washed and detected by flow cytometry. Cumulative results of 3 experiments are shown. (D) NO production by MDSCs was measured by detection of nitrite concentrations. Cumulative results of 3 experiments are shown. (E and F) NT staining in LLC and LLC-IL-1β tumors. Double staining of NT+ (brown) and either Gr-1+ or F4/80+ (red) cells in tumor tissues. Scale bars: 100 μm. (E) The percentages of NT+ cells in LLC and LLC-IL1β tumor tissues analyzed by Aperio software. Ten fields (800 × 600 μm2 each) were selected from each tumor, and mean ± SEM is shown. Four experiments with the same results were performed. *P < 0.01. (G and H) Antitumor effect of T cell therapy. Mice were injected s.c. with different numbers of LLC-OVA (G) or LLC-IL1β-OVA (H) cells, which provided for similar tumor sizes 2 weeks after inoculation. On days 18 and 23, 8 × 106 activated OT-I T cells were injected i.v. Tumors were measured. Each group included 9–12 mice. Data are mean ± SEM. In G the differences were significant on day 23. (P < 0.05).
We used these models to evaluate the effect of adoptive transfer of CTLs on the growth of tumors with induced inflammation. In preliminary experiments we determined the number of IL-1β–producing tumor cells that resulted in a rate of tumor growth similar to that of tumor cells with a low IL-1β level. The tumor cells were injected on day 0, and activated OVA-specific T cells were transferred to the mice on days 18 and 23 in this advanced model. As expected, transfer of the activated OVA-specific OT-1 T cells into the LLC-OVA tumor–bearing mice resulted in a significant (P < 0.05) delay in tumor progression compared with the LLC-bearing mice (Figure 6G). However, this effect was completely lost in the mice bearing tumors that produce IL-1β (Figure 6H).
It is now established that effective adoptive transfer therapy requires a lymphodepletion that can be achieved with chemotherapy or by total body irradiation (TBI) (38, 39). Therefore, we repeated experiments with the adoptive transfer of OVA-specific CTLs into tumor-bearing recipients treated with TBI followed by bone marrow transplantation. During the first 12 days after adoptive transfer, CTLs almost completely stopped the growth of LLC-OVA tumors, whereas in the control group tumors continued to grow. This resulted in almost 4-fold differences in tumor size between these two groups of mice (Figure 7A). However, this antitumor effect of CTLs was completely absent in the LLC-IL-1β-OVA tumor–bearing mice (Figure 7B). We investigated whether the observed effect was the result of profound immune suppression caused by the systemic expansion of MDSCs in mice bearing IL-1β–expressing tumors. We selected a time point (7 days after TBI and T cell transfer) when the antitumor effect in the LLC-OVA–bearing mice was apparent and in the LLC-IL-1β-OVA mice was absent (Figure 7, A and B). T cells were isolated from LN or tumor tissues and either restimulated with specific peptide–loaded naive antigen-presenting cells or exposed to anti-CD3/CD28 Abs. T cells from LLC-OVA and LLC-IL-1β-OVA tumor–bearing mice showed similar responses to both the specific antigen and nonspecific stimulation as measured in an IFN-γ ELISPOT assay (Figure 7C) or by T cell proliferation (Figure 7D).
Inflammation reduced the effect of adoptive T cell therapy. (A and B) LLC-OVA (A) or LLC-IL-1β-OVA (B) tumors were established as described in Figure 6, G and H. All mice received TBI and bone marrow transplant on day 0. OT-I T cells were transferred to the treatment groups on day 1. Data are mean ± SEM. Each group included 9–12 mice. In A the differences were significant (P < 0.01). (C and D) T cell responses. Tumor-bearing mice received TBI with bone marrow transplant and T cell transfers as described in A and B. On day 7 T cells were isolated from LNs and tumors and mixed at a 1:1 ratio with irradiated syngenic control splenocytes and stimulated with either control or specific peptides, or anti-CD3/CD28 Abs. (C) IFN-γ production was measured in ELISPOT assays. The number of spots per 5 × 104 T cells was calculated. Each experiment was performed in triplicate and included 3 mice. Cumulative mean ± SEM is shown. (D) The proliferation of T cells isolated from spleens and tumors was determined by labeling of T cells with CFSE, followed by stimulation with specific or control peptides in the presence of irradiated naive splenocytes. The experiments were performed twice, with similar results. (E and F) The percentages of MDSCs (E) and macrophages (F) in spleen and tumor sites in mice 1 week after TBI and bone marrow transfer. (G) The number of NT+ cells in LLC-IL-1β-OVA tumors 7 days after TBI. Gr-1+ cells are red; NT+ cells are brown. Scale bars: 100 μm. Right panel: Cumulative results of the number of NT+ cells per 10 high-power fields (800 × 600 μm2). Each group included 3 mice.
We then measured levels of peripheral MDSCs and macrophages after TBI in these mice. Seven days after TBI, the numbers of MDSCs and macrophages in spleens of LLC-OVA– and LLC-IL-1β-OVA–bearing mice were dramatically and equally reduced to barely detectable levels (Figure 7, E and F). However, the number of MDSCs in LLC-IL-1β-OVA tumors was still significantly (P < 0.05) higher than in the LLC-OVA tumors (Figure 7E). Most importantly, the total number of PNT-producing myeloid cells in the tumor site was only marginally reduced 7 days after TBI (Figure 7G).
When T cell transfer to tumor-bearing mice was delayed by a week (to allow for partial reconstitution of the myeloid compartment after TBI), LLC-IL-1β-OVA tumors caused significantly stronger suppression than LLC-OVA tumors (Supplemental Figure 8).
We hypothesized that if MDSC-derived PNT was indeed responsible for the tumor cell resistance to CTLs, then the blockade of PNT with CDDO-Me would improve the effect of adoptive T cell therapy even in conditions of induced enhanced inflammation. Five-day treatment of mice bearing either LLC-OVA or LLC-IL-1β-OVA tumors with a CDDO-Me–containing diet caused a significant (P < 0.05) decrease in the number of NT+ cells infiltrating the tumors (Figure 8A). We evaluated the antitumor effect of CDDO-Me on T cell therapy of tumor-bearing mice. LLC-OVA–bearing mice were treated with OVA-specific CTLs. In addition, the mice received 3 cycles of the CDDO-Me treatment (5 days each, with 3-day intervals, starting from day –2). CDDO-Me alone had only a moderate effect on tumor growth. In contrast, the addition of CDDO-Me to the T cell transfer resulted in a significant antitumor effect (P = 0.01) (Figure 8B). Experiments were stopped at that time, since mice in all control groups had to be sacrificed because tumor size in the control groups exceeded 2.5 cm in diameter (the maximum allowed by University of South Florida Health Sciences Center Animal Care and Use Committee regulations). Next, we evaluated the effect of CDDO-Me and T cells in LLC-IL-1β-OVA tumor–bearing mice. LLC-IL-1β-OVA–bearing mice were treated with TBI and OVA-specific CTLs. CDDO-Me alone had only a moderate effect on the growth of tumors. However, in contrast to the results described in Figure 7, the addition of CDDO-Me to the T cell transfer halted tumor progression for 4 weeks (Figure 8C) (P = 0.018). Since tumor cells expressing H-2Kb-SIINFEKL fusion protein (LLC-H-2Kb-SIINFEKL) were resistant to the negative effect of PNT (Figure 2, E–G), we asked whether this tumor would be more sensitive to T cell therapy than LLC-OVA. In contrast to the LLC-OVA tumor model, even in the absence of TBI or CDDO-Me treatment, OT-I T cells were able to completely reject LLC-H-2Kb-SIINFEKL tumors (Figure 8D). Although these results are striking, it is very difficult to formally exclude the possibility that the effect of the treatment was caused, at least partially, by the higher expression and stability of the peptide-MHC fusion protein in these cells than pMHC complexes assembled in LLC-OVA. OT-I T cells showed more potent killing of LLC-H-2Kb-SIINFEKL than LLC-OVA tumor cells (Figure 2G and Supplemental Figure 6).
Inhibition of PNT production improves the antitumor effect of adoptive T cell transfer. (A) The number of NT+ cells per 10 high-power fields (800 × 600 mm2) in LLC-OVA and LLC-IL-1β-OVA tumors 5 days after treatment with 150 mg/kg CDDO-Me or control diets. Cumulative results of 3 mice per group.*P < 0.05. (B**) Combined effect of CDDO-Me and T cell transfer on growth of LLC-OVA tumor. Tumor-bearing mice were treated with control and CDDO-Me diets for 5 days with 3 days interval started on day –2. OT-1 T cells were injected on days 1 and 8. Each group included 4–5 mice. The differences between CDDO-Me + OT-1 and other groups were significant (P = 0.01). (C) Combined effect of CDDO-Me and T cell transfer on growth of LLC-IL-1β-OVA tumor. All mice received TBI and bone marrow transfer on day 0. Each group included 4 mice. The differences between CDDO-Me + OT-1 and other groups were significant (P < 0.05). (**D**) Effect of T cell transfer on tumor growth of LLC-H-2Kb-SIINFEKL tumor. Each group included 4–5 mice. (**E** and **F**) Combined effect of CDDO-Me and T cell transfer on growth of B16-F10 (**E**) or H-2Kb– B16-F10 (**F**) tumors. Pmel-1 T cells were injected on days 1 and 8 into mice bearing B16-F10 melanoma (**E**) and OT-1 T cells to mice bearing H-2Kb– B16F-10 melanoma. Each group included 4–5 mice. The differences between CDDO-Me + T cells and other groups were significant (_P_ = 0.04). In **E** but not in **F** (_P_ > 0.1). (B–F**) Mean ± SEM is shown. Tumor-bearing mice were treated as described in B. (G) Schematic of proposed effect of myeloid cells on tumor cell resistance to CTLs. Prf, perforin; GrzB, granzyme B. Red circles: processed antigen; brown: nitrated amino acids.
To formally assess the role of CDDO-Me in increased tumor cell sensitivity to T cell therapy, we used H-2Kb– B16-F10 melanoma cells (40) and OT-1 CTLs that recognize irrelevant OVA-derived peptide. Adoptive transfer of Pmel-1 T cells slowed the growth of wild-type B16-F10 melanoma, and CDDO-Me significantly (P = 0.04) enhanced this effect (Figure 8E). As expected, no effect of OT-1 T cell transfer was observed in mice bearing H-2Kb– melanoma. CDDO-Me had effect on tumor growth. However, the combination of OT-1 T cell transfer with CDDO-Me treatment did not enhance the effect of the therapy (Figure 8F).