Targeting TNF-α–producing macrophages activates antitumor immunity in pancreatic cancer via IL-33 signaling

Diverse TAMs coexist in human PDA. Analysis of transcriptome data demonstrated that human PDA is characterized by excessive accumulation of heterogeneous macrophages (Figure 1A and Supplemental Figure 1A; supplemental material available online with this article; https://doi.org/10.1172/jci.insight.153242DS1), consistent with previous reports (14, 20). This macrophage expansion and infiltration starts at the early stages of disease (8, 20) (Supplemental Figure 1B), and increased levels of macrophages correlated with poor prognosis in PDA patients (Figure 1B). To further characterize human PDA macrophage populations, we used scRNA-Seq data from human PDA patients (n = 6) and 2 cancer-adjacent normal pancreas samples (n = 2) (data from ref. 27, where it is noted that more than 96% of cells in this grouping were identified as monocytes or macrophages). Overall, all the macrophages showed patient-specific heterogeneity (Supplemental Figure 1C). We observed distinct clusters and distribution of genes associated with processes such as proinflammation, matrix remodeling, metabolism, and immunosuppression using dimensionality reduction with t-distributed stochastic neighbor embedding. Overall, we found that, consistent with other reports (23, 28), PDA TAMs displayed a more alternatively activated (M2-like) polarization with the expression of genes associated with the M2 phenotype and immunosuppression such as SPP1, CD163, CXCR4, HIF1, TGF-β1, and multiple MHCII molecules (Figure 1C). Likewise, the vast majority of PDA TAMs robustly expressed the M2 marker CLEC7A (Dectin-1), which in concert with its ligand galectin 9 promotes strong immune suppression from macrophages that can be blocked to promote antitumor immunity in PDA (29). Further, PDA TAMs displayed shifts in metabolism (Supplemental Figure 1D). The majority of TAMs displayed a heterogeneous expression of transcripts associated with higher immunosuppressive metabolism and signaling such as KYNU (tryptophan metabolism), NLRP3 (inflammasome/IL-18), ADK (adenosine metabolism), and STAT3 (Supplemental Figure 1D). Notably, tryptophan production and kynurenine production are known to inhibit T cell proliferation and cytotoxic activity (30). Likewise, TAMs also expressed high levels of ADK, which is involved in adenosine metabolism, where adenosine metabolites promote antiinflammatory macrophage phenotype. In contrast, transcripts suggesting a classically activated M1-like polarization such as CCR7, IL-6, and IL-2RA were low, while NOS2 (iNOS), a robust M1 marker, was extremely low (Figure 1C), further supporting the conclusion that the majority of TAMs in PDA possess a more alternatively activated phenotype. Collectively, this culminated with strong statistical enrichment for immunosuppressive TAM function (Figure 1D) and also high enrichment for TNF-α production pathways (Figure 1E), which is consistent with high TNF-α levels in the PDA stroma (Supplemental Figure 1E), further suggesting that PDA TAMs could be a major source of TNF-α in PDA and are predominantly immunosuppressive.

Human PDA has heterogeneous macrophage populations.Figure 1

Human PDA has heterogeneous macrophage populations. (A) TCGA data set analysis demonstrating that transcripts identifying myeloid cells and TAMs in PDA (i.e., CD68, CD11b, and CD14) are overexpressed in PDA. *P < 0.0001. (B) Transcripts associated with CD68+ macrophages correlate with poor prognosis in PDA patients (TCGA data set analysis). TPM, transcripts per million. (C) Distribution of factors that regulate macrophage polarization and function within the heterogeneous macrophage populations in PDA showing the strongest signal for immunosuppressive factors. (D and E) Gene Ontology pathway analysis of human PDA CD68+ macrophages showing decreased antitumor immune activation (i.e., immunosuppressive behavior) (D) and increased pathways enriched in TNF-α production (E).

Further analysis of the human PDA macrophage transcriptome supports findings from mice demonstrating that macrophages are either tissue resident or bone marrow derived (20, 27); pancreatic resident macrophages originate from the fetally derived yolk sac and are maintained independently of circulating macrophages, while infiltrating macrophages arise from circulating CD14+ monocytes. Indeed, PDA TAMs are a mixed population displaying markers of both populations. However, the majority of TAMs across all patients displayed high gene expression of CD14, C1QB, and MHCII (e.g., HLA-DR; Figure 1C) typically associated with bone marrow–derived macrophages (20, 28). For instance, about 80% of TAMs that originate from infiltrated hematopoietic stem cell–derived monocytes are MHCIIhi (20), and monocytes that infiltrate tissue and differentiate into macrophages can be distinguished by high expression of C1QB (31) (Figure 1C). In contrast, genes associated with resident macrophages such as CX3CR1 or MerTK were also expressed but were noticeably less abundant, suggesting that despite their heterogeneity the majority of TAMs originated from bone marrow.

Immunosuppressive TAMs in genetically engineered murine models of PDA. To examine how well human PDA TAMs are represented in various murine models of PDA, we extended our investigation to the analysis of scRNA-Seq data from 3 different genetically engineered murine models of PDA: the KIC (KrasLSL-G12D/+ Ink4afl/fl Ptf1aCre/+), KPC (KrasLSL-G12D/+ Trp53LSL-R172H/+ Ptf1aCre/+), and KPfC (KrasLSL-G12D/+ Trp53fl/fl Pdx1-Cre) systems (Figure 2), as described previously (32). Overall, TAM populations observed in human PDA were also well represented in these murine models. Consistent with human data, murine PDA TAMs displayed high expression of M2-associated and immunosuppressive transcripts, such as Spp1, C1qb, Arg1, Tgfb1, and multiple MHCII molecules (e.g., H2-Aa, H1-Ab1, H2-Dma, H2-Dmb1, H2-Dmb2, and H2-Eb1; Figure 2 and Supplemental Figure 2). Furthermore, similar to human data, M1 markers (e.g., Nos2, Ccr5, Il-6) were found at a lower frequency (Supplemental Figure 2), consistent with previous data showing low levels of iNos+ TAMs in murine PDA (15). Further, Cd14-, C1qb-, and Cxcr4-expressing macrophages were present in lower numbers in the normal pancreas but expanded in early disease and were maintained throughout late-stage disease. Likewise, the presence of Spp1+ and Arg1+ macrophages was extremely low in normal pancreas, but their number increased in PDA. Together, these data demonstrate robust TAM heterogeneity in murine PDA similar to that seen in human PDA, with dominant immunosuppressive features and the majority showing high expression of markers indicating they are bone marrow–derived in origin.

Cross-species examination of macrophages from KPC, KIC, and KPfC geneticallFigure 2

Cross-species examination of macrophages from KPC, KIC, and KPfC genetically engineered murine models of PDA demonstrates similarities to human macrophages. Visualization of transcript distributions in murine macrophage populations using t-distributed stochastic neighbor embedding (t-SNE) shows that expression patterns in KPC, KIC, and KPfC macrophages, like those in human PDA macrophages, suggest robust immunosuppressive behavior.

CCL2 is overexpressed by distinct cell populations at primary and metastasis sites. Our data analysis in both human and murine PDAs suggests that the majority of immunosuppressive TAMs originate from bone marrow–derived monocytes. Loss or inhibition of CCR2 is known to significantly reduce levels of circulating monocytes and, as a result, numbers of bone marrow–derived TAMs (20, 33), and reducing TAMs may be beneficial as a therapeutic strategy against PDA (e.g., 14, 17, 18, 34). Therefore, to examine the expression of the CCR2 ligand CCL2 in PDA, we first performed correlation analysis with human patient data and observed a significant correlation between CCL2 and CD14 (Figure 3A), as well as other macrophage markers such as CXCR4, CD206, CD163, and MHCII transcripts (not shown), further linking expression of CCL2 in tumors with myeloid cell infiltration. Furthermore, examination of CCL2 levels using IHC demonstrated overexpression in the primary tumor as well as in a metastatic lesion of human PDA (Figure 3B), where CCL2 was overexpressed starting in early disease and continued to be expressed highly throughout disease progression (Supplemental Figure 3A). Similarly, elevated expression was seen in pancreatic intraepithelial neoplasia (PanIN) and primary and metastatic PDA in KPC mouse samples (Figure 3C). In murine and human PDA, IHC revealed CCL2 expression in carcinoma cells and the stromal compartment (Figure 3, B and C), suggesting that multiple cell populations in PDA can recruit circulating monocytes. To confirm this, we performed immunofluorescent staining for CCL2 and α-SMA, a marker for a dominant subtype of carcinoma-associated fibroblasts, or myofibroblastic CAFs (myCAFs), in human PDA, and observed CCL2 localization with carcinoma cells, α-SMA+ CAFs, and α-SMAlo or α-SMA– cells in the stroma (Figure 3D). Therefore, we sought to further evaluate the relative CCL2 contribution of carcinoma cells and CAFs. Interestingly, while primary carcinoma cells secreted CCL2, matched metastatic lines (i.e., lines from metastatic lesions in the same animals) showed greater mRNA (Supplemental Figure 3B) and CCL2 secretion (Figure 3E). Yet primary CAFs from KPC tumors secreted substantially higher CCL2 protein when compared with carcinoma cells (Figure 3F).

Both iCAFs and myCAFs secrete high levels of CCL2 in PDA tumor microenvironFigure 3

Both iCAFs and myCAFs secrete high levels of CCL2 in PDA tumor microenvironments. (A) Human TCGA data set analysis shows a strong correlation between expression levels of CCL2 and CD14. (B) CCL2 is overexpressed in human PDA. IHC analysis demonstrates high expression of CCL2 in both human primary PDA tumors and metastatic lesion (lung). (C) CCL2 is also overexpressed in the genetically engineered KPC model of PDA at all stages. IHC shows high expression in PanINs, primary tumor, and metastatic lesions (liver). (D) CCL2 is colocalized with carcinoma cells and both α-SMAlo and α-SMAhi cells in the stroma of both human and murine PDA. Scale bars: 50 μm (BD). (E) Metastatic cancer cells secrete higher levels of CCL2 than carcinoma cells derived from primary tumors (CCL2 levels were measured in culture supernatant by ELISA for paired primary and metastatic cell lines derived from KPC mice). (F) CAFs secrete higher CCL2 than carcinoma cells. CCL2 levels in culture supernatant of CAFs (grown on 2D culture plates with serum, i.e., myCAFs) and carcinoma cells were measured by ELISA (n = 5–7 KPC cell lines; P value was derived by Mann-Whitney test). (G and H) CAFs are the major CCL2 contributors in PDA. Violin plots of Ccl2 transcripts for individual cell populations in KPC PDA show that iCAFs and myCAFs both express higher levels of Ccl2 compared with other tumor cell populations. The expression of Ccl2 in both myCAFs and iCAFs was validated experimentally by quantitative PCR (n = 5–7 KPC cell lines; P value was derived by Mann-Whitney U test). (I) Strong correlations between CCL2 and both iCAF marker genes (CXCL1, LIF) and myCAF marker genes (ACTA, CTGF) in human TCGA data sets demonstrating that CCL2 is highly expressed by both iCAFs and myCAFs.

As recent studies have highlighted CAF heterogeneity in PDA (27, 32, 35, 36), we further dissected CCL2 secretion from 2 prominent CAF populations, myCAFs and inflammatory CAF (iCAFs), that we generated from KPC tumors as described previously (35). Remarkably, both CAF phenotypes showed significantly higher expression of CCL2 compared with primary or metastatic carcinoma cells, which is consistent with scRNA-Seq data and moderate correlations between iCAF or myCAF transcript markers and CCL2 in human PDA (Figure 3I). However, iCAFs expressed significantly higher levels of CCL2 than myCAFs (Figure 3H and Supplemental Figure 3D), suggesting a hierarchy for recruiting monocyte-derived macrophages to immunosuppressive niches in order of iCAFs > myCAFs > metastatic cells > primary carcinoma cells. However, it is notable that both iCAFs and myCAFs expressed profoundly more CCL2 cytokine than other cell populations from PDA tumors (Figure 3, F–H, and Supplemental Figure 2D), suggesting they both have a robust capacity to recruit monocytes via CCL2. In fact, this behavior was observed for a number of factors known to promote PDA progression and/or therapeutic resistance (e.g., collagens, proteoglycans, hyaluronan synthesis, lysyl oxidase, IGF-1, IL-6, CSF1, etc.; Supplemental Figure 4), while others, such as CXCL12, appeared more phenotype specific. Thus, compared with the broader cell populations in PDA (vs. comparing levels solely between iCAFs and myCAFs), it is clear that many key factors are elevated in both iCAFs and myCAFs relative to other cell types but sometimes to different degrees (i.e., both are significant sources of key ECM proteins, cytokines, and growth factors), adding further complexity to dissecting distinct roles of CAF phenotype, particularly spatially and temporally within complex tumor microenvironments.

Ccr2 deletion delays PDA progression and reduces metastasis to improve overall survival. To better understand the influence of CCL2/CCR2 signaling and the impact of bone marrow–derived TAMs in PDA, we generated KPC mice lacking Ccr2 genes. Ccr2-deleted mice have been reported to have reduced capacity for monocyte recruitment from bone marrow (37). To generate KPC-CCR2–knockout mice, KC mice (KrasLSL-G12D/+ Pdx1-Cre) were crossed with mice with global knockout of Ccr2 and then bred with Trp53LSL-R172H/LSL-R172H mice (Figure 4A). The progeny were born in the expected Mendelian ratio, with no obvious functional defects. The deletion of Ccr2 in KPC-CCR2–/– was confirmed by PCR (Supplemental Figure 5A). As expected, the loss of CCR2 led to a profound reduction of circulating CD11b+ myeloid cells (>80% reduction; Supplemental Figure 5B), which resulted in a concomitant decrease in the PDA stroma (Supplemental Figure 5C). Interestingly, examination of full survival data demonstrated that KPC-CCR2–/– mice survived significantly longer compared with KPC littermates (median survival 168 vs. 120 days, respectively; Figure 4B). Histological examination of pancreatic tumors from early-stage (10–11 weeks old) KPC and KPC-CCR2–/– mice showed that KPC-CCR2–/– mice had more disease-free normal pancreatic tissue and lower-grade PanINs than the KPC group (Figure 4C), consistent with the right-shifted survival curve (Figure 4B), suggesting a slower progression of disease. Consistent with this finding, depletion of circulating myeloid cells also reduced the proliferation of carcinoma cells in early- and late-stage disease (Supplemental Figure 5D). Examination of PDA in both cohorts showed regions of well-differentiated, moderately differentiated, poorly differentiated, and necrotic regions; however, KPC-CCR2–/– mice again showed less high-grade tumor as well as significantly less metastatic burden, but no differences in local invasion, α-SMA+ CAF frequency, or fibrillar collagen deposition, when compared with KPC mice (Figure 4C and Supplemental Figure 5, D–G), suggesting that reduced levels of CCL2-recruited TAMs in PDA slow disease progression and reduce metastatic burden, resulting in longer survival of KPC mice.

Genetic deletion of Ccr2 reduces disease severity.Figure 4

Genetic deletion of Ccr2 reduces disease severity. (A) Schematics of the KPC mouse model of pancreatic cancer: KrasLSL-G12D/+ p53LSL-R172H/+ Pdx1-Cre. Deletion of Ccr2 was attained by crossing of KPC with Ccr2–/– (global) mice, referred to as KPC-CCR2–/–. (B) Kaplan-Meier analysis comparing survival of KPC (n = 37) and KPC-CCR2–/– (n = 21) mice demonstrates that KPC-CCR2–/– mice survive longer than KPC mice. P = 0.018 by log-rank test. (C) KPC-CCR2–/– animals show delayed PDA onset. Comparative H&E staining and histopathology of pancreata from KPC and KPC-CCR2–/– mice in early disease (10–11 weeks) (n = 3 in each group) and at the endpoint (KPC, n = 22; KPC-CCR2–/–, n = 19) show less advanced disease throughout the pancreas of KPC-CCR2–/– mice. (D and E) KPC-CCR2–/– mice show decreased metastasis. Scale bars: Early: 100 μm (50 μm for the boxed areas) and Late: 200 μm (50 μm for the boxed areas). (D) Representative H&E-stained images of liver and lung metastatic lesions from KPC and KPC-CCR2–/– cohorts. Scale bars: 200 μm (50 μm for the boxed areas). (E) Percentage of metastasis in various organs of KPC (n = 22) and KPC-CCR2–/– (n = 19) animals. P value by Fisher’s exact test.

Examination of the metastatic distribution and burden in KPC mice demonstrated that while 91% of KPC mice displayed metastatic dissemination, only 40% of KPC-CCR2–/– presented with metastasis (Figure 4, D and E; Table 1; and Supplemental Table 1). In the liver, 57% of mice in the KPC cohort displayed metastatic lesions, in contrast to 25% in KPC-CCR2–/– mice (Figure 4E). KPC-CCR2–/– mice also exhibited a lower percentage of diaphragm metastasis (38% vs. 20%) and a nonsignificant trend of decreasing lung metastasis (41% vs. 30%). Last, we note that recent studies have established a role for elevated fibronectin (FN) in promoting the pre-metastatic niche, in part through the recruitment of bone marrow–derived macrophages (38, 39). Therefore, to determine whether our observed reduction in metastasis was due, at least in part, to a decrease in pre-metastatic niche formation, we measured the levels of FN in the 8- to 11-week-old mice. No differences in FN levels were observed in the liver or lungs (Supplemental Figure 5H), suggesting that the observed decrease was not due to a difference in key ECM in the pre-metastatic niche formation but rather could have been due to decreased macrophages. Taken together, these data suggest that the depletion of CCR2-mediated macrophage infiltration profoundly decreases metastatic disease.

Table 1

Comparison of metastatic burden in KPC and KPC-CCR2–/– mice

CCR2 inhibition mitigates the immunosuppressive environment and renders PDA susceptible to ICB. Given collective findings by us and others showing a predominant immunosuppressive TAM phenotype in PDA and our data from KPC-CCR2–/– mice, we sought to deplete bone marrow–derived monocytes to test the hypothesis that depletion of infiltrated TAMs is an avenue for the development of new combination immunotherapies for PDA. To test our hypothesis, we enrolled the KPC mice with 4 to 8 mm tumors in the longest axis by high-resolution small-animal ultrasound into treatment cohorts using a rolling enrollment model. The enrolled mice were treated with (a) standard-of-care chemotherapy with gemcitabine (Gem), (b) gemcitabine in combination with ICB in the form of anti–PD-1 and anti–CTLA-4 (Gem+ICB), or (c) a combination of gemcitabine, ICB, and CCR2 inhibition (Gem+ICB+CCR2i) (Figure 5A). To our knowledge, CCR2 inhibition in combination with ICB immune therapies has not been previously tested in PDA, especially using autochthonous disease that arises in the KPC model, and we focused on preclinical drug combinations that could be clinically viable (i.e., that include a standard-of-care chemotherapy). In order to inhibit CCR2, we tested a new orthosteric inhibitor, CCX598 (a potent third-generation CCR2 antagonist from ChemoCentryx). CCR2 inhibitor was given orally every day, a dosing scheme that achieves the desired concentration in the circulation to obtain receptor coverage (Supplemental Figure 6A). Importantly, and consistent with our hypothesis, the Gem+ICB+CCR2i combination therapy significantly increased the survival of KPC mice compared with Gem or Gem+ICB treatments (Figure 5B), while the addition of ICB did not improve outcomes from gemcitabine alone, consistent with other reports showing that without a stroma-targeting approach chemotherapy plus ICB is not impactful in PDA (12, 40). Concomitantly with this behavior, in primary tumors, total myeloid cells, F4/80+ macrophages, CD206+ immunosuppressive TAMs, MDSCs, which are robustly immunosuppressive in PDA (18), and neutrophils were all significantly decreased, while numbers of CD8+ T cells were concurrently significantly increased (Figure 5, C–F, and Figure 6, A and B). We note that the observed decrease in neutrophils in autochthonous disease is in contrast to findings observed with grafted tumor systems using a distinct CCR2 inhibitor (41). Interestingly, Gem+ICB+CCR2i combination treatment also increased the frequency of iNOS+ cells, a marker of classically activated macrophages, in the tumor microenvironment (TME) (Supplemental Figure 6B). Along with these shifts in the immune landscape, we also identified decreased tumor cell proliferation as evidenced by reduced Ki67+ staining and enhanced apoptosis in the Gem+ICB+CCR2i compared with the Gem and Gem+ICB groups (Figure 6, C and D). Combination therapy also resulted in increased vascular patency (Supplemental Figure 6C) without decreasing general fibrosis (as discerned from levels of α-SMA+ CAFs and fibrillar collagen in the PDA stroma; Supplemental Figure 6, D and E), demonstrating that reducing TAMs derived from bone marrow not only renders PDA susceptible to ICB, but can also surmount aspects of the vascular collapse phenotype that play a key role in driving drug-free sanctuaries in PDA (42).

Blocking infiltration of bone marrow–derived TAMs increases responsivenessFigure 5

Blocking infiltration of bone marrow–derived TAMs increases responsiveness to immune therapy in KPC mice. (A) Schematic of the therapy regime. (B) Kaplan-Meier curve showing that Gem+ICB+CCR2i–treated animals have significantly longer survival compared with Gem and Gem+ICB cohorts. *P < 0.05. (CF) IHC/IF analysis demonstrates that Gem+ICB+CCR2i combination therapy results in significant decreases in total CD11b+ myeloid cells (C), F4/80+ macrophages (D), CD206+ immunosuppressive macrophages (E), and MDSCs (F). P values by Kruskal-Wallis and Dunn’s multiple-comparison tests; n =4–6 animals in each group. Scale bars: 50 μm.

Blocking infiltration of bone marrow–derived TAMs increases cytotoxic T celFigure 6

Blocking infiltration of bone marrow–derived TAMs increases cytotoxic T cell levels and carcinoma cell death in KPC mice. (A) IHC/IF analysis demonstrates that Gem+ICB+CCR2i combination therapy results in significant decreases in neutrophils within PDA. (B) IF staining shows significant increases in CD8+ cytotoxic T cells in the Gem+ICB+CCR2i treatment group. (C) Gem+ICB+CCR2i–treated animals have lower numbers of Ki67+ cells. (D) Gem+ICB+CCR2i therapy increases cell death as shown by IHC staining for cleaved caspase-3 (CC3+). Cell number or signal per field of view is shown. P values by Kruskal-Wallis and Dunn’s multiple-comparison tests; n = 4–6 animals in each group. The scale bar for the main images is 50 μm. The magnifications are 1.5×.

Combination of CCR2 inhibition and checkpoint blockade decreases metastasis in KPC mice. Consistent with decreased myeloid cells in primary tumors, CCR2 inhibition also blocked the accumulation of myeloid cells in metastatic organ sites (Figure 7, A and B). Therefore, to specifically determine the impact of combination Gem+ICB+CCR2i therapy on metastatic lesions, we performed a detailed necropsy and histopathological analysis of each of the 3 preclinical treatment cohorts. Analysis demonstrated that Gem+ICB+CCR2i combination therapy but not Gem or the Gem+ICB combination decreased total metastasis (Figure 7, C–F, and Supplemental Table 2). Forty-four percent of Gem+ICB+CCR2i had liver metastases compared with 66% in the Gem and 70% in the Gem+ICB cohorts (Figure 7D). Gem+ICB+CCR2i–treated mice also exhibited a profoundly lower percentage of lung metastasis (from 55 % to 11%; Figure 7E). This profound decrease in lung metastasis suggests a stronger role for marrow-derived macrophages in lung metastases. Notably, we also observed a concerning trend of increased diaphragm metastasis in the Gem+ICB, but a decreasing trend in the Gem+ICB+CCR2i cohort (Figure 7F). However, in the Gem+ICB+CCR2i cohort we again observed that decreasing myeloid cells led to a significant increase in CD8+ T cell population (Figure 7G) and tumor cell death (Figure 7H). We note the larger increases in cytotoxic T cell levels in metastatic sites compared with increases in primary tumors, suggesting that myeloid-targeting therapy has robust benefit for combating metastatic disease. Thus, overall these data show that, similarly to the primary disease, combination therapy can render metastatic disease susceptible to ICB, increase antitumor immunity, and increase cell death in established metastatic disease, resulting in an overall decrease in metastatic burden.

Blocking infiltration of bone marrow–derived TAMs decreases metastatic burdFigure 7

Blocking infiltration of bone marrow–derived TAMs decreases metastatic burden and increases antitumor immune responses in metastasis lesions. (A and B) IHC staining for CD11b shows that Gem+ICB+CCR2i therapy decreases myeloid cell recruitment into metastatic liver sites. P value by Kruskal-Wallis and Dunn’s multiple-comparison tests; n = 4–6 animals in each group. (CF) Gem+ICB+CCR2i therapy significantly decreases metastatic burden. Representative images of lung metastatic lesions in Gem, Gem+ICB, and Gem+ICB+CCR2i animals (C) and associated quantification of metastatic burden in the liver (D), lung (E), and diaphragm (F) in KPC mice treated with Gem (n = 9), Gem+ICB (n = 11), or Gem+ICB+CCR2i (n = 13). (G) IF staining shows that Gem+ICB+CCR2i combination increases CD8+ T cell infiltration at metastatic sites. P value by Mann-Whitney test; n = 4–5 animals. (H) CC3 IHC analysis shows that Gem+ICB+CCR2i therapy increases cell death in metastatic PDA. P value by Mann-Whitney test; n = 4–5 animals in each group. Scale bars: 50 μm.

TNF-α/IL-33 signaling resulting in increased CD8+ T cell infiltration. To identify the signal that led to increased antitumor immunity, we measured cytokine levels in treated tumors. Among all cytokines, IL-33 was the most highly enriched in the Gem+ICB+CCR2i–treated tumors (Figure 8A). We confirmed this result in all 3 treatment groups through IHC analysis that shows a heterogeneous expression in both the carcinoma and stromal compartments in both the GEM and GEM+ICB groups with higher levels in the stroma (Figure 8B). However, in the triple therapy group we again observed expression in both compartments, but a profound increase in IL-33 expression in carcinoma cells (Figure 8B). Increased levels of IL-33 were also observed in KPC-CCR2–/– animals (Figure 8C). These data suggest that blockade of bone marrow–derived macrophages in concert with chemotherapy and immune therapy leads to robust increases in IL-33 expression in pancreatic tumors.

Blocking TNF-α–producing TAMs increases alarmin IL-33 levels in the TME.Figure 8

Blocking TNF-α–producing TAMs increases alarmin IL-33 levels in the TME. (A) Protein cytokine array of tumor lysates and quantification of array spots showing increased IL-33 expression upon Gem+ICB+CCR2i treatment (n = 2 tumors in each group). (B) IHC of IL-33 in Gem-treated, Gem+ICB–treated, and Gem+ICB+CCR2i–treated tumors, showing robustly elevated levels following Gem+ICB+CCR2i treatment. (C) IHC and quantification show higher IL-33 levels in KPC-CCR2–/– animals compared with KPC. P value by Mann-Whitney test; n = 4–5 animals. Scale bars: 50 μm.

Macrophages also showed very high enrichment of TNF-α production pathways (Figure 1E). Indeed, dual staining of KPC tumor showed that most of the TNF-α colocalized with F4/80+ macrophages (Supplemental Figure 7A), and analysis of RNA-Seq data showed strong expression from TAMs (Supplemental Figure 7B), validating the conclusion that macrophages are a primary source of TNF-α in PDA, consistent with previous findings (43, 44). Further, immunofluorescence (IF) analysis of both tumors from KPC-CCR2–/– mice and tumors treated with combination therapy showed a profound decrease in TNF-α levels (Figure 9, A and B), again supporting the conclusion that tumor-infiltrating bone marrow–derived TAMs are a primary source of TNF-α in PDA. Previously, TNF-α has been shown to modulate the expression of IL-33 in normal and diseased fibroblasts (45). Thus, next we hypothesized that this increase in IL-33 could be due to a decrease in TNF-α from macrophages. Since the major increase in IL-33 in the Gem+ICB+CCR2i group was observed from carcinoma cells, we sought to test whether TNF-α can directly regulate IL-33 expression in carcinoma cells. We treated primary KPC cell lines with recombinant TNF-α, which led to significantly decreased IL-33 at both the gene and protein levels (Figure 9, C and D). Interestingly, IL-33 is a member of the IL-1 family that has been shown to play a key role in innate and adaptive immunity (46, 47). IL-33 is normally released by damaged or necrotic cells and can act as an alarmin capable of activating either Th1 or Th2 response (47, 48). Furthermore, IL-33 was recently shown to activate tumor-infiltrated group 2 innate lymphoid cells (ILC2s) to regulate CD8+ T cell responses (49). Indeed, IL-33 expression correlated strongly with CD8A and granzyme B as well as genes like BATF3, IRF8, THBD, CLEC9, and XCR1 that are required for tumor antigen cross-presentation functionality of CD103+ dendritic cells (DCs) (Figure 9E and Supplemental Figure 7C). Therefore, to test whether decreased TNF-α and increased IL-33 do indeed lead to increased numbers of CD103+ DCs in the tumor microenvironment, we stained the tumor sections for both CD11c and CD103 and observed a significant increase in both DC markers (Figure 9F and Supplemental Figure 7D). Next, we subcutaneously implanted KPC cells in syngeneic mice with or without recombinant IL-33 (rIL-33) mixed into growth factor–reduced Matrigel. Consistent with evidence suggesting that increased IL-33 levels promote antitumor T cell responses, PDA tumor growth was profoundly inhibited in the rIL-33 group compared with control conditions (Figure 9G) with concomitant and robust increases in CD8+ T cell infiltration (Figure 9H). We note that, consistent with findings by Moral et al. (49), we did not observe IL-33R expression of CD8+ T cells (Supplemental Figure 7, E–G), suggesting that IL-33 signaling impacts T cells through an intermediary, such as ILC2s and CD103+ DCs. These data therefore show that the alarmin IL-33 promotes CD8+ T cell infiltration and antitumor response in PDA. Thus, we demonstrate that depletion of immunosuppressive TAMs and MDSCs results in decreased TNF-α, causing increased IL-33 levels in carcinoma cells that result in increased cytotoxic T cell response to combat metastatic PDA.

Increases in IL-33 induce increases in CD8+ cytotoxic T cell levels.Figure 9

Increases in IL-33 induce increases in CD8+ cytotoxic T cell levels. (A and B) IF staining shows significantly decreased TNF-α levels in KPC-CCR2–/– and Gem+ICB+CCR2i treatment groups compared with KPC and Gem groups. P value by Mann-Whitney test; n = 4–5 animals. (C and D) Treatment of KPC cells with recombinant TNF-α causes a decrease in IL-33 at the gene and protein levels. P value by Kruskal-Wallis and Dunn’s multiple-comparison tests; n = 3. (E) Strong correlations between IL-33 and markers of CD103+ DC levels and functionality (TCGA data set analysis). (F) IF staining shows a significant increase in CD103+ DCs in Gem+ICB+CCR2i treatment group. P value by Mann-Whitney test; n = 4–5 animals. (G) Analysis of tumor growth curves for control and rIL-33 mice shows that IL-33 decreases the size of subcutaneous PDA tumors and overall tumor volume at the endpoint (n = 3 per group). (H) IF staining and quantification demonstrate that IL-33 increases CD8+ T cell numbers in PDA tumors. P value by Mann-Whitney test; n = 3 animals per group. Scale bars: 10 μm.

留言 (0)

沒有登入
gif