Macrophage PI3Kγ Drives Pancreatic Ductal Adenocarcinoma Progression

Inflammation plays a major role in cancer immune suppression and progression, perhaps especially in pancreatic cancer (1–4). Pancreatic ductal adenocarcinoma (PDAC), a devastating disease with one of the poorest 5-year survival rates among all cancers, is the third leading cause of cancer death in the United States (5–6). Although 10% to 15% of patients are candidates for gross total surgical resection, local recurrence and metastasis are frequent, and the overall 5-year survival rate among patients with pancreatic cancer is only around 7% (5, 6). Because standard therapies have only a modest impact on survival (7, 8), novel therapeutic and diagnostic strategies are urgently needed.

Pancreatic carcinomas, like other solid tumors, are characterized by an abundant inflammatory cell infiltrate, including T cells, B cells, mast cells, macrophages, and neutrophils, yet these cells produce an ineffective antitumor immune response (9–11). Myeloid lineage cells, including macrophages, neutrophils, mast cells, and myeloid-derived suppressor cells, play important roles in the initiation of pancreatic carcinoma and in establishing an immune-suppressive microenvironment that dampens effective antitumor T-cell responses in pancreatic carcinomas (9–11). Strategies to reverse immune suppression in PDAC by inhibiting myeloid cell trafficking or signal transduction (12–16), by reactivating adaptive immune responses, by vaccination with attenuated intracellular bacteria or DNA plasmids coding for PDAC-associated antigens (16–18), or by the application of checkpoint inhibitors (19–21) have demonstrated that appropriate targeting of the immune system may lead to new and effective treatment strategies for pancreatic cancer. However, continued investigation into the mechanisms by which innate immune cells contribute to PDAC progression will help to refine approaches to improve therapy for this disease.

For this reason, we sought to identify mechanisms regulating the contribution of myeloid cells to immune suppression and progression of PDAC. We report here that the macrophage lipid kinase PI3Kγ promotes immune-suppressive polarization in macrophages, leading to immune suppression, tumor progression, metastasis, and fibrosis in PDAC. The Class I PI3K lipid kinases phosphorylate PtdIns (4,5)P2 on the 3′ hydroxyl position of the inositol ring to produce PtdIns (3,4,5)P3, which regulates metabolic and transcriptional pathways during inflammation and cancer (22, 23). The Class IA isoforms PI3Kα and PI3Kβ are widely expressed in endothelial, epithelial, and tumor cells, whereas the Class IA isoform PI3Kδ is primarily expressed in T and B lymphocytes and the structurally unique Class IB isoform PI3Kγ is expressed mainly in myeloid cells, where it is the major PI3K isoform (22, 23). PI3Kγ promotes the adhesion and migration of granulocytes and monocytes, and mice lacking the PI3Kγ catalytic subunit exhibit defects in neutrophil and monocyte accumulation in inflamed and tumor tissues (24, 25). Using orthotopic and genetically engineered mouse models of PDAC, we found that inhibition of PI3Kγ unexpectedly slowed PDAC tumor growth and enhanced survival and responsiveness to standard-of-care chemotherapy by altering macrophage transcriptional profiles and thereby activating T cells, indicating that targeting PI3Kγ therapeutically could provide long-term control of this devastating malignancy.

Macrophages play particularly significant roles in PDAC tumor progression, as they can promote tumor initiation, immune suppression, and metastasis (9–15). To investigate the role of the macrophage in spontaneous and experimentally induced pancreatic adenocarcinomas, we immunostained pancreata from normal and tumor-bearing mice to detect the presence of F4/80⁺ macrophages. We examined pancreata from normal mice, as well as LSL-Kras^G12D; Pdx-1Cre mice (KC mice), which develop extensive PanIns but not invasive carcinoma, LSL-Kras^G12D/+; LSL-Trp53^R172H/+; Pdx-1Cre mice (KPC mice), which develop invasive carcinomas and widespread metastases, and orthotopically implanted LSL-Kras^G12D/+; LSL-Trp53^R172H/+Pdx-1Cre (LMP) tumors (26–32). Although normal pancreata exhibited very few F4/80⁺ macrophages, pancreata from KC and KPC mice exhibited extensive infiltration by macrophages, as did orthotopic LMP tumors (Fig. 1A). Quantification of CD11b⁺ myeloid cells and F4/80⁺ macrophages in spontaneous and orthotopic LMP tumors demonstrated that macrophage infiltration increases, whereas T-cell and B-cell infiltration decreases, during tumor development (Fig. 1B; Supplementary Fig. S1A–S1D). We also observed increases in CD68⁺ macrophage content in invasive PDACs compared with healthy pancreata from patients (Fig. 1C and D), as has been previously reported (26), indicating that increased macrophage infiltration occurs in both murine and human PDACs.

The infiltration of bone marrow–derived granulocytes and monocytes into tumors depends in part on the gamma isoform of the Class I PI3K lipid kinases (24, 25). Therefore, we asked whether PI3Kγ might promote inflammation associated with tumor progression in PDAC. Immunostaining of clinical biopsies of human invasive PDACs to detect expression of PI3Kγ (green) and CD68⁺ macrophages (red) showed that CD68⁺ macrophages, but not other cells in pancreatic tumors, uniformly express PI3Kγ (yellow; Fig. 1E). Western blotting analyses showed that murine bone marrow–derived and tumor-associated macrophages (TAM) express PI3Kγ, whereas CD19⁺ B cells and CD90⁺ T cells express PI3Kδ, and murine LSL-Kras^G12D/+; LSL-Trp53^+/−; Pdx-1Cre (p53 2.1.1) and LSL-Kras^G12D/+; LSL-Trp53^R172H/+; Pdx-1Cre (LMP, K8484, DT6606, or DT4313) pancreatic carcinoma cells express PI3Kα but not PI3Kγ (Fig. 1F; Supplementary Fig. S2A). These results indicate that PI3Kγ is a biomarker of PDAC-associated macrophages that might have a functional role in PDAC inflammation.

To investigate the functional role of macrophage PI3Kγ in pancreatic cancer, Panc02, K8484, or p53 2.1.1 pancreatic tumor cells were orthotopically implanted in wild-type (WT) or PI3Kγ-deficient (p110γ^−/−) mice, according to the depicted schema (Fig. 2A). Whereas WT macrophages expressed PI3Kγ, neither p110γ^−/− macrophages nor PDAC cells expressed this kinase (Fig. 2B; Supplementary Fig. S2A). Although normal p110γ^−/− pancreata were similar in size as those from WT mice, pancreatic tumors from p110γ^−/− mice were significantly smaller than pancreatic tumors in WT mice, as p110γ^−/− tumors were reproducibly half the size of WT tumors (Fig. 2C–E). Notably, the incidence of PDAC metastases, including kidney, diaphragm, and liver metastases, was also significantly reduced in p110γ^−/− animals (Fig. 2F and G), indicating that macrophage PI3Kγ contributes to tumor spread.

We previously showed that PI3Kγ regulates the trafficking of bone marrow–derived CD11bGr1^lo monocytes and CD11bGr1^hi granulocytes into lung tumors by stimulating integrin α4β1-mediated adhesion to endothelium (24, 25, 33). To determine if PI3Kγ regulates myeloid cell trafficking in pancreatic tumors, we performed flow cytometry to quantify immune cell populations in tissues from PDAC-bearing p110γ^−/− versus WT animals and in PI3Kγ inhibitor–treated versus vehicle control–treated animals. Whereas PI3Kγ inhibition only slightly suppressed recruitment of CD11b⁺Gr1⁺ myeloid cells to tumors, it had little effect on the recruitment of CD11b⁺Gr1-F4/80^hi macrophages into PDACs (Fig. 2H and I; Supplementary Fig. S2B). Importantly, PDACs from p110γ^−/− mice exhibited significantly more CD4⁺ and CD8⁺ T-cell content than PDACs from WT mice (Fig. 2J and K). In contrast, no differences were observed in the number of T cells in the spleen, lung, or liver of p110γ^−/− and WT mice (Supplementary Fig. S2C and S2D). As p110γ is primarily expressed in myeloid cells, and p110δ is the major PI3K isoform in T cells (Fig. 1F), these results suggest that PI3Kγ inhibition in myeloid cells promotes enhanced CD8⁺ T-cell mobilization into PDACs, thereby enhancing antitumor immunity.

To determine whether PI3Kγ inhibition improves the survival of mice with spontaneous pancreatic tumors, we crossed p110^−/− mice with KC and KPC mice (27, 28). PI3Kγ deletion significantly delayed tumor progression and extended the survival of both KC and KPC mice (Fig. 3A and B). The median survival time of control KC mice was 36 weeks, whereas median survival of p110γ^−/− KC mice was 43.9 weeks (Fig. 3A). Similarly, the median survival time of control KPC animals was 26 weeks, whereas the median survival time of p110γ^−/− KPC animals was 35.5 weeks (Fig. 3B). Remarkably, 20% of p110γ^−/− KPC mice lived for more than 1 year, whereas none of the WT mice survived beyond 40 weeks. Analysis of pancreata from KPC animals by immunohistochemistry indicated that most KPC animals exhibited extensive claudin⁺, cytokeratin19⁺ invasive carcinoma (30) with little normal amylase⁺ pancreas remaining (Fig. 3C–E). WT animals almost uniformly exhibited metastases in organs that included liver, heart, lung, and kidney (Fig. 3F). In contrast, KPC; p110γ^−/− animals exhibited extensive areas of normal amylase⁺ pancreas, little claudin⁺ and cytokeratin19⁺ invasive carcinoma (Fig. 3C–E), and little evidence of metastasis (Fig. 3F). Together, these results illustrate that PI3Kγ drives tumor development and spread in the pancreas, and PI3Kγ inhibition prevents these events and significantly extends survival.

Because PI3Kγ promotes PDAC growth and spread, we speculated that pharmacologic inhibitors of PI3Kγ could favorably affect the outcome of pancreas cancer. To test this, we evaluated the effect of the investigational PI3Kγ/δ inhibitor TG100-115, which has an IC₅₀ of 83 nmol/L for p110γ, 238 nmol/L for p110δ, and >1,000 nmol/L for p110α and p110β (34), in mouse models of PDAC. mCherry-labeled LMP cells were orthotopically implanted into the pancreata of WT mice, and mice were then treated with the PI3Kγ inhibitor TG100-115 or chemically matched control compound, according to the schema depicted in Fig. 4A. Intravital imaging revealed that control-treated mice exhibited large pancreatic tumors and multiple metastases, whereas TG100-115–treated mice had significantly smaller pancreatic tumors and few, if any, metastases (Fig. 4A; Supplementary Fig. S3A and S3B). Quantification of tumor size by area of fluorescence and by tumor weight demonstrated that TG100-115 substantially suppressed PDAC growth to approximately half of that of control-treated animals (Fig. 4B and C). TG100-115 also quantitatively suppressed metastasis to the diaphragm, liver, colon, and other organs (Fig. 4D). Either early or late treatment with TG100-115 suppressed growth of p53 2.1.1 orthotopic tumors (Fig. 4E; Supplementary Fig. S4A–S4C), indicating PI3Kγ inhibition can suppress the growth of late-stage tumors as well as early-stage tumors. Histologic analysis of tumors revealed that TG100-115–treated tumors exhibited enhanced tumor necrosis compared with control-treated tumors (Fig. 4F); however, TG100-115 had no direct affect on viability of any PDAC cell line (Supplementary Fig. S4D and S4E). As PI3Kγ inhibition suppresses tumor growth without directly targeting tumor cells, these results suggested that PI3Kγ inhibitors might effectively combine with standard-of-care chemotherapy, which targets tumor cells directly.

To evaluate the combinatorial effect of PI3Kγ inhibition and chemotherapy on PDAC growth and progression, we treated mice bearing orthotopic PDAC tumors with TG100-115 and gemcitabine, according to the depicted schema (Fig. 4G). TG100-115 and gemcitabine similarly suppressed tumor growth as single agents, whereas the combination of the two treatments suppressed tumor growth even further (Fig. 4G) and significantly inhibited metastasis (Fig. 4H). Importantly, combined TG100-115 and gemcitabine treatment prolonged survival of mice with implanted PDACs from a median survival of 28 days for control animals to 38 days for the combination of gemcitabine and TG100-115 (Fig 4I). In addition, long-term TG100-115 treatment significantly increased survival of KPC mice with spontaneous pancreatic tumors and was accompanied by substantial reductions in metastasis (Fig. 4J–L). These results indicate that PI3Kγ inhibitors provide long-term survival benefits in these models of pancreas cancer.

Although myeloid cells can establish immunosuppressive microenvironments by inhibiting CD8⁺ T-cell recruitment and/or survival in tumors (9–11), p110γ^−/− PDACs exhibited increased CD8⁺ cell influx (Fig. 2J and K). Therefore, we investigated whether PI3Kγ inhibition alters the immune response expression profile of PDACs. Genes associated with immune suppression, chronic inflammation, or tumor angiogenesis, including Arg1, Tgfb, Il1b, Il6, and Vegfa, were strongly expressed in myeloid cells isolated from WT PDAC tumors (Supplementary Fig. S5A). Genetic (p110γ^−/−) and pharmacologic (TG100-115) inhibition of PI3Kγ both significantly inhibited expression of these genes in orthotopic PDAC tumors and in TAMs purified from PDAC tumors (Fig. 5A and B; Supplementary Fig. S5B and S5C). In contrast, the expression of immunostimulatory factors, including Il12 and Ifng, was significantly enhanced in tumors and TAMs from p110γ^−/− and PI3Kγ inhibitor–treated animals (Fig. 5A and B; Supplementary Fig S5B and S5C). However, no significant difference in gene expression was observed in macrophages isolated from the uninvolved spleen or liver of WT and p110γ^−/− PDAC tumor–bearing mice (Supplementary Fig. S5D and S5E). To determine whether PI3Kγ directly controls these transcriptional changes in macrophages, we evaluated the effect of PI3Kγ inhibition in IL4-polarized in vitro cultured macrophages, which model tumor-derived macrophages, as they express immune-suppressive cytokines and factors, including Arg1, Il10, and Tgfb, and inhibit expression of immune stimulatory genes including Il12b and Ifng (35–39). Genetic or pharmacologic inhibition of PI3Kγ not only inhibited immunosuppressive gene expression but also stimulated expression of Il12b, Ifng, and Nos2 (Fig. 5C), indicating that PI3Kγ directly regulates a macrophage transcriptional switch from immune suppression toward immune stimulation.

To determine if selective blockade of PI3Kγ in macrophages was sufficient to disrupt tumor growth, we performed adoptive transfer studies by implanting tumor-derived macrophages in tumors grown in host mice. Tumor-derived macrophages were isolated from WT and p110γ^−/− p53 2.1.1 tumors, mixed with freshly cultured p53 2.1.1 tumor cells and implanted in WT or p110γ^−/− host mice. Tumor growth was significantly inhibited when p110γ^−/− macrophages were implanted in either WT or p110γ^−/− host mice (Fig. 5D). In contrast, tumor growth was enhanced when WT macrophages were implanted in p110γ^−/− host mice (Fig. 5D). Importantly, tumors implanted with p110γ^−/−, but not WT, macrophages exhibited a 3-fold increase in CD8⁺ T-cell content (Supplementary Fig. S5F).

Our studies show that PI3Kγ inhibition alters macrophage transcriptional programming, leading to increased CD8⁺ T-cell recruitment and reduced PDAC growth (Fig. 2; Supplementary Fig. S5F). We speculated therefore that PI3Kγ inhibition indirectly activates T-cell–mediated antitumor immune responses in vivo. Accordingly, we found that T cells isolated from p110γ^−/− PDAC tumors exhibited enhanced expression of the T_H1 cytokine Ifng and downregulated expression of the immune-suppressive cytokines Tgfb and Il10 (Fig. 5E). To exclude the possibility that PI3Kγ within T cells directly controls T-cell activation, we examined the effect of p110γ inhibition on T-cell proliferation ex vivo. T cells were isolated from naïve (Fig. 5F) or PDAC-bearing (Fig. 5G) WT and p110γ^−/− mice, and proliferation was stimulated by incubation with anti-CD3, anti-CD3 + CD28, or IL2 + anti-CD3 + CD28 antibodies. No differences were observed in the proliferative capacity of WT and p110γ^−/− T cells, whether from naïve or tumor-bearing mice. To determine if the antitumor effect of p110γ inhibition required the presence of cytotoxic CD8⁺ T cells, we crossed WT and p110γ^−/− mice with CD8^−/− animals and performed PDAC tumor studies. As only 3% of live cells in WT PDACs were CD8⁺ T cells (Fig. 2J), it was not surprising that CD8 deletion had little effect on WT PDAC tumor size (Fig. 5H). Importantly, CD8 deletion ablated the tumor-suppressive effect of p110γ inhibition (Fig. 5H). Taken together, these results indicate that macrophage PI3Kγ suppresses T-cell–mediated immune surveillance by promoting expression of immunosuppressive cytokines by TAMs. Furthermore, our studies indicate that PI3Kγ inhibition indirectly stimulates T-cell–mediated antitumor immune responses, leading to growth suppression in PDAC.

Recent progress in anticancer therapeutics has led to the development and clinical approval of T-cell checkpoint inhibitors for therapy of some solid tumors, including melanoma and lung carcinoma (18–21). To explore the possibility that PI3Kγ inhibitors might synergize with T-cell checkpoint inhibitors for the treatment of PDACs, we first evaluated expression of the programmed cell death protein-1 (PD-1) and its ligand, programmed cell death ligand-1 (PD-L1), in WT and p110γ^−/− PDAC tumors. Although PD-1 was expressed on CD8⁺ T cells from PDAC tumors, we observed no significant differences in expression levels in T cells from WT and p110γ^−/− tumors (Supplementary Fig. S6A). We did observe significant expression of the checkpoint ligand PD-L1 on macrophages but not tumor cells from PDACs, but no difference in expression between cells from WT and p110γ^−/− tumors (Supplementary Fig. S6B). We also observed no change in the presence of FOXP3⁺CD25⁺CD4⁺ T regulatory cells in WT and p110γ^−/− tumors (Supplementary Fig. S6C). To explore the relative effects of PI3Kγ inhibition and T-cell checkpoint inhibition as therapeutic strategies, we treated WT and p110γ^−/− animals bearing PDAC tumors with anti–PD-1 or isotype control antibodies. Both anti–PD-1 and p110γ inhibition substantially suppressed PDAC tumor growth (Supplementary Fig. S6D and S6E). However, the combination of anti–PD-1 with p110γ inhibition provided no added benefit over single-agent therapy (Supplementary Fig. S6D and S6E). We also treated WT animals bearing PDAC tumors with TG100-115, anti–PD-1, isotype control antibodies, or combination therapy; similarly, no benefit was observed with the combined treatment. These data support the conclusion that PI3Kγ inhibition promotes an antitumor T-cell response that, like anti–PD-1, relieves T-cell exhaustion. Future studies optimizing checkpoint inhibitor immunotherapy and tumor cell–targeted therapy in combination with PI3Kγ inhibition may provide further insights into the nature of immune suppression in PDAC cancer.

PI3Kγ inhibition mediated a striking suppression of metastasis in orthotopic and spontaneous PDAC models (Figs. 2–4). To determine whether macrophage PI3Kγ promotes PDAC invasion, we analyzed the effect of conditioned medium (CM) from IL4-stimulated WT or p110γ^−/− macrophages on migration of LMP and p53 2.1.1 PDAC cells. CM from WT macrophages stimulated robust chemotaxis, whereas CM from p110γ^−/− macrophages was less effective in stimulating cell migration (Fig. 6A; Supplementary Fig. S7A). Macrophage CM had no effect on cell proliferation (Supplementary Fig. S7B). To identify factors that could account for these differences in migration, we performed RNA sequencing on p110γ^−/− and WT IL4-stimulated macrophages and found that the mRNA expression of many chemotactic factors, including Ccl2, Scf, Hbegf, Pdgfa, and Pdgfb, was strongly suppressed in p110γ^−/− macrophages (Fig. 6B). To determine if these factors could stimulate PDAC invasion, we tested the effect of the PDGFR inhibitor imatinib (40, 41), the PDGF inhibitor Fovista (42), and the EGFR inhibitors erlotinib (43), lapatinib (44), and anti-CCL2 on CM-stimulated chemotaxis. Only imatinib and Fovista suppressed PDAC migration toward macrophage CM, indicating that PDGF and PDGFR are required for macrophage-stimulated PDAC chemotaxis (Fig. 6C and D; and Supplementary Fig. S7C and S7D). To determine which cytokines were sufficient to induce migration, we supplemented p110γ^−/− CM with PDGF-AA, PDGF-BB, SCF, or CCL2 and found that only PDGF-BB was sufficient to restore chemotactic migration in murine and human PDAC cells (Fig. 6E; Supplementary Fig. S7E and S7F). Notably, we found that PDGF-BB protein expression was reduced in p110γ^−/− CM and in PDAC tumors from p110γ^−/− mice (Fig. 6F and G). As p110γ^−/− CD11b⁺Gr1⁻ tumor-derived macrophages expressed reduced Pdgfb mRNA in vivo (Fig. 6H), our studies show that macrophage-derived PDGF-BB may promote PDAC metastasis in vivo and that PI3Kγ inhibitors may block metastasis by suppressing macrophage expression of PDGF-BB.

Pancreatic tumors are associated with extensive desmoplasia; this severe desmoplastic response impedes the effectiveness of chemotherapy and disrupts the normal functions of the pancreas (45). Pancreata from p110γ^−/− KC and KPC animals exhibited substantially less desmoplasia than their WT counterparts, as revealed by Masson’s Trichrome staining of pancreatic tumor sections (Fig. 7A). Similarly, pancreata from KPC animals treated with the PI3Kγ inhibitor TG100-115 exhibited less desmoplasia by Trichrome staining (Fig. 7B). Similar differences in desmoplastic response were seen in picrosirius red–stained pancreata from KPC orthotopic tumors grown in WT and p110γ^−/− animals (Fig. 7C and D). In addition, less collagen protein and gene expression was observed in orthotopic LMP tumors that were treated with the PI3Kγ inhibitor TG100-115 (Fig. 7E–G). To determine whether macrophage PI3Kγ regulates fibroblast-mediated collagen expression, we incubated primary murine fibroblasts with CM from basal and IL4-stimulated WT and p110γ^−/− macrophages. Less collagen mRNA was expressed in fibroblasts incubated with p110γ^−/− CM (Fig. 7H). As p110γ^−/− macrophages express less TGFβ1 and PDGF-BB than WT macrophages, we tested the effect of inhibitors of these factors in collagen expression assays. Only anti–PDGF-BB and imatinib suppressed collagen expression in fibroblasts, indicating that macrophage PI3Kγ likely regulates desmoplasia associated with PDAC by controlling expression of secreted PDGF-BB (Fig. 7I).

In summary, the studies presented here demonstrate the critical roles that PI3Kγ plays in regulating PDAC tumor growth and progression. Our studies demonstrate PI3Kγ promotes transcription of genes associated with the M2 immunosuppressive macrophage phenotype in PDACs, including immune-suppressive factors such as Arg1, Tgfb, and Il10 and chemokines and wound-healing factors that include PDGF-BB. By inhibiting PI3Kγ, the expression of these genes is constrained, thereby activating CD8⁺ T-cell–dependent tumor suppression and increasing survival.

Inflammation plays a critical role in pancreatic carcinoma progression and relapse from therapy (1–4). B cells, T cells, and myeloid cells can all be found in the pancreatic tumor microenvironment, yet these cells do not mount an appropriate antitumor adaptive immune response. It is now accepted that TAMs and myeloid-derived suppressor cells release immune-suppressive factors that inhibit T-cell–mediated antitumor responses (3, 38), and therapeutic approaches that are aimed at preventing myeloid-derived suppressive responses show some therapeutic efficacy in animal models of cancer (12–17). T-cell checkpoint inhibitors that target T-cell functions in cancer also hold some promise as novel therapeutics for pancreatic cancer (18–21). However, strategies that inhibit myeloid cell–mediated immune suppression may well boost the effect of checkpoint inhibitors, vaccines, and other therapeutic strategies in the highly immunosuppressive microenvironment of pancreatic tumors.

In the studies presented here, we have identified PI3Kγ as a critical regulator of the pathways that control immune suppression, metastasis, and desmoplasia in pancreatic cancer. We showed that PI3Kγ is expressed in human and murine pancreatic tumor macrophages and that selective deletion of this PI3K isoform suppresses orthotopic and spontaneous PDAC growth and metastasis. In addition, PI3Kγ inhibition significantly enhanced survival of mice bearing spontaneous PDACs by suppressing tumor growth and metastasis. Our studies demonstrate that PI3Kγ plays a key role in activating the immune-suppressive transcriptional signature of tumor-derived macrophages, in that PI3Kγ inhibition suppressed expression of Arg1, Tgfb, and Il10 and stimulated expression of Il12 and Ifng in vivo. These changes in myeloid gene expression signatures were associated with increased CD8⁺ T-cell recruitment to PDAC tumors, increased T-cell expression of IFNγ, decreased expression of TGFβ and IL10, and CD8⁺ T-cell–dependent tumor suppression. Our results are in agreement with recent studies that showed that PI3Kδ but not PI3Kγ is required for T-cell activation, but contrast with studies that showed that p110γ is required for T-cell recruitment by inflammatory chemokines (46–49). It is possible that T-cell recruitment to tumors in vivo depends on chemokines that activate only p110δ rather than p110γ. Finally, we show that PI3Kγ also regulates macrophage expression of PDGF-BB, which stimulates tumor cell chemotaxis and fibroblast production of collagen in vitro and in vivo. Recent studies revealed that p53 mutations induce constitutive PDGFR expression and signaling that drives tumor invasion and metastasis (50). Our studies suggest that macrophage-derived PDGF-BB may cooperate with this mutant pathway to promote metastasis but also show that targeting PI3Kγ can control this pathway by which PDACs spread. Our studies thus demonstrate that inhibitors of PI3Kγ offer promise as new therapeutic approaches to control tumor growth and progression, metastasis, and desmoplasia in this devastating malignancy.

In related studies, we demonstrated that human and murine PDACs exhibit increased PI3Kγ-dependent BTK activation in CD11b⁺ FcγRI/III⁺ myeloid cells (51). BTK or PI3Kγ inhibition as monotherapy in early-stage PDAC, or in combination with gemcitabine in late-stage PDAC, slowed progression of orthotopic tumors in a manner dependent on T cells (51). However, we observed that BTK was not activated in p110γ^−/− macrophages and that the combination of BTK and PI3Kγ inhibitors had no additive effects in regulating macrophage gene expression or PDAC progression. These studies indicated that the two kinases regulate overlapping signal transduction pathways in macrophages. An increase in effector and memory CD8⁺ T-cell phenotypes was also observed in these studies, which is consistent with other reports about CD8⁺ T-cell responses to various immunotherapies (17).

We demonstrated that genetic and pharmacologic PI3Kγ inhibition was as effective as treatment with the checkpoint inhibitor anti–PD-1 in mouse models of PDAC. Although we were unable to observe additive effects of targeting PI3Kγ and PD-1 on tumor growth at this time, we did observe improved survival by combining PI3Kγ inhibitors and chemotherapeutic agents. Our studies show that targeted inhibition of PI3Kγ can combine with other therapeutic approaches targeting distinct components of the tumor microenvironment to effect long-term durable anti-PDAC tumor immune responses.

All studies with human tissues were approved by the Institutional Review Board for human subjects research of the University of California, San Diego (UCSD). Informed consent was obtained from all patients prior to surgery. The use of samples occurred under “exempt category 4” for research on deidentified biological specimens. All animal experiments were performed with approval from the Institutional Animal Care and Use Committees of the University of California, San Diego, and the University of Torino, Italy.

TG100-115 was from Targegen, Inc. Fovista was from Ophthotech.

The p53 2.1.1 pancreatic adenocarcinoma cell line was derived from primary PDAC tumors (Fvb/N) of male transgenic KC mice harboring null mutations in Trp53. The LMP pancreatic adenocarcinoma cell line was derived from a liver metastasis of a primary PDAC tumor from transgenic KPC mice in a C57BL6;129Sv background. The K8484 pancreatic adenocarcinoma cell line was derived from a primary PDAC tumor from transgenic KPC mice in a C57BL6 background. The Panc02 cell line from C57BL6 mice has been previously described (33). All cell lines were tested for Mycoplasma contamination and grown in DMEM/10% FBS/1.0% Penicillin-Streptomycin (Pen-Strep) on plastic plates (LMP, Panc02, K8484) or plastic plates coated with 50 μg/mL rat tail collagen I (p53 2.1.1; BD Biosciences). Cells used in these studies were authenticated by morphologic profiling and RNA sequencing and RT-PCR in 2012, whole-exome analysis in 2015 (p53 2.1.1), and PI3Kγ inhibitor sensitivity analyses in 2013–2015. Panc02 cells and LMP were acquired in 2009; K8484 cells and p53 2.1.1 cells were acquired in 2014.

Generation and characterization of p110γ^−/−, KC, and KPC mice have been described previously (23, 24, 28, 29). KPC animals in the C57BL6 background were crossed with Pik3cg^−/− animals in the C57BL6 background to generate Pik3cg^−/−; KC and Pik3cg^−/−; KPC mice in the Animal Facility at the Molecular Biotechnology Center, University of Turin.

Ten thousand p53 2.1.1 PDAC cells were orthotopically implanted into the pancreata of p110γ^+/+ (WT) or p110γ^−/− 8-to-10-week-old FVB/n mice. Five hundred thousand LMP PDAC cells were orthotopically implanted into the pancreata of 8-to-10-week-old C57BL6;129 mice, and five hundred thousand K8484 or Panc02 PDAC cells were orthotopically implanted into the pancreata of 8-to-10-week-old C57BL6 mice (n = 10). In some studies, WT and p110γ^−/− animals with tumors were treated by i.p injection with and without gemcitabine (150 mg/kg) on day 7 and day 14 (n = 10). In other studies, mice were treated by i.p injection b.i.d. with 2.5 mg/kg of PI3Kγ inhibitor (TG100-115) or with a chemically similar inert control twice daily from days 7 to 21 (n = 10). In some studies, WT and p110γ^−/− mice were treated with 100 μg anti–PD-1 or isotype control clone LTF-2 (BioXCell) administered by i.p. injection on days 7, 10, and 13 of tumor growth. Mice were sacrificed on day 21. For all tumor experiments, tumor volumes and weights were recorded at sacrifice. In other studies, the growth, metastasis, and survival of spontaneous PDAC tumors in p110γ^+/+ and p110γ^−/−LSL-Kras^G12D/+; Pdx-1Cre, and LSL-Kras^G12D/+; LSL-Trp53^R172H/+Pdx-1Cre animals were evaluated over 100+ weeks.

Freshly isolated mouse bone marrow cells from 9 WT and 9 p110γ^−/− mice were pooled into 3 replicate sets of WT or p110γ^−/− cells that were differentiated into macrophages for 6 days in RPMI + 20% FBS + 1% Pen-Strep + 50 ng/mL macrophage colony-stimulating factor (M-CSF). Each replicate set of macrophages was then incubated for 48 hours with M-CSF or M-CSF + IL4. Macrophages were removed from dishes, and RNA was harvested using a Qiagen Allprep Kit. One microgram of total RNA per sample was used for the construction of sequencing libraries. RNA sequencing was performed by the University of California, San Diego, Institute for Genomic Medicine Genomics Center as follows: RNA libraries were prepared for sequencing using standard Illumina protocols. mRNA profiles of M-CSF– and IL4-stimulated macrophage derived from WT and PI3K gamma null (p110γ^−/−) mice were generated by single read deep sequencing, in triplicate, using Illumina HiSeq2000. Sequence analysis was performed as previously described, and results are available to view online at the NCBI Gene Express Omnibus website (file number GSE58318).

For studies evaluating the effect of drugs on tumor growth, a sample size of at least 10 mice/group provided 80% power to detect mean difference of 2.25 SD between two groups (based on a two-sample t test with 2-sided 5% significance level). Prior to statistical analyses, data were examined for quality and possible outliers. Data were normalized to the standard where applicable. Significance testing was performed by one-way Anova with Tukey post-hoc testing for multiple pairwise testing or by parametric or nonparametric Student t test as appropriate. The Fisher exact test was used to query significant differences in rates of metastasis between groups.

Additional detailed methods are available online in the Supplementary Materials.

E. Hirsch has ownership interest (including patents) in Kither Biotech SrL Italy. J.A. Varner reports receiving a commercial research grant from Infinity Pharmaceuticals. No potential conflicts of interest were disclosed by the other authors.

Conception and design: M.M. Kaneda, A.V. Nguyen, N. Ralainirina, F. Novelli, E. Hirsch, J.A. Varner

Development of methodology: M.M. Kaneda, A.V. Nguyen, P. Foubert, P. Sun, A.M. Lowy, J.A. Varner

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M.M. Kaneda, P. Cappello, A.V. Nguyen, C.R. Hardamon, M.C. Schmid, E. Mose, M. Bouvet, A.M. Lowy, M.A. Valasek, F. Novelli, E. Hirsch, J.A. Varner

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M.M. Kaneda, P. Cappello, A.V. Nguyen, M.A. Valasek, R. Sasik, F. Novelli, J.A. Varner

Writing, review, and/or revision of the manuscript: A.M. Lowy, F. Novelli, J.A. Varner

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): P. Sun, E. Mose, J.A. Varner

Study supervision: E. Mose, M. Bouvet, A.M. Lowy, J.A. Varner

The authors thank Ophthotech for the gift of Fovista and Sanofi-Aventis/Targegen for the gift of TG100-115. They also thank Xiaodan Song, Joan Manglicmot, and Roberta Curto for technical assistance.

The authors acknowledge support from T32HL098062 (to M.M. Kaneda) and R01CA167426-03S1 (to A.V. Nguyen); from the Ministero della Salute Ricerca Sanitaria Finalizzata RF-2013-02354892, Associazione Italiana Ricerca sul Cancro (5 × mille no. 12182), and IG no. 15257; from the University of Turin-Progetti Ateneo 2014-Compagnia di San Paolo (PANTHER to P. Cappello; PC-METAIMMUNOTHER to F. Novelli); and from the NCI/NIH (R01CA167426, R01CA126820, and R01CA083133), Lustgarten Foundation, and AACR/Landon Foundation (to J.A. Varner).