A Notch-Dependent Inflammatory Feedback Circuit between Macrophages and Cancer Cells Regulates Pancreatic Cancer Metastasis

Pancreatic ductal adenocarcinoma (PDAC) is a highly devastating disease with a poor prognosis because of its invasive and metastatic characteristics (1). In advanced stages, patients show invasion of the (retro) peritoneum, liver, and other organs (2). To date, the major drivers of PDAC invasion and metastasis remain poorly understood. Therefore, better understanding of the molecular mechanisms underlying cancer invasion and metastasis would facilitate identification of new targets and development of therapies for PDAC.

The Notch signaling pathway is a highly conserved molecular cell signaling pathway that plays important roles in cell differentiation, survival, proliferation, stem cell renewal, and cell fate determination during development and morphogenesis (3). Notch signaling consists of receptors, ligands, and DNA-binding proteins. Four transmembrane Notch receptors (Notch1–4) and two families of Notch ligands [Jagged (JAG) 1 and 2, and Delta-like (DLL) 1–4)] have been found in mammals (4). Recent studies have established that Notch signaling is one of the most activated pathways in cancer cells (5). Notch signaling in cancer cells acts in either an oncogenic or tumor-suppressive manner that depends on the cellular context (6). Available data regarding the role of Notch signaling in pancreatic cancer are contradictory (7). Notch suppresses pancreatic intraepithelial neoplasia, the proposed precursor lesion of cancer formation in a mouse model of pancreatic cancer (8), but Notch may also play oncogenic roles in pancreatic tumorigenesis (9, 10). Positive expression of Notch ligands JAG1 and DLL4 closely correlates with severe clinicopathologic characteristics and a poor prognosis of patients with PDAC (11). Therefore, the exact role of Notch signaling in PDAC remains to be identified. In addition, because PDAC metastasis depends on bidirectional interactions between cancer cells and the microenvironment (12), whether Notch signaling participates in regulating the cross-talk between the different compartments of the tumor microenvironment remains to be explored.

We have previously reported that miR-124 is a tumor-suppressive miRNA involved in pancreatic cancer metastasis (13). Recently, we identified that the Notch receptor ligand JAG1 is also a directly regulated target of miR-124 in PDAC. Our preliminary results suggested that the function of JAG1 in PDAC metastasis is dependent on tumor-associated macrophages (TAM), which suggests Notch signaling–dependent cross-talk between pancreatic cancer cells and macrophages.

Therefore, in this study, we further investigated the role of Notch signaling in pancreatic cancer metastasis and uncovered a Notch signaling–dependent inflammatory feedback circuit between pancreatic cancer cells and macrophages, which promotes and maintains metastasis of PDAC.

Human pancreatic cancer samples were collected from patients with PDAC after surgical resection at Huashan Hospital, Fudan University (Shanghai, China), from January 2003 to December 2005, as reported previously (13). Written-informed consent was obtained from all patients in accordance with institutional guidelines before sample collection. The study was approved by the committee for ethical review of research at Fudan University Shanghai Cancer Center (approval no. 050432-4-1212B). Patient characteristics are provided in Supplementary Table 1.

Human pancreatic cancer cell lines AsPC-1, Capan-1, SW1990, BxPC3, MIA PaCa-2, PANC-1, and SU.86.86 were obtained from the American Type Culture Collection. Murine pancreatic cancer cell line Panc02 was purchased from ScienCell. Murine KPC1199 pancreatic cancer cell line was kindly provided by Professor Jing Xue from Shanghai Jiaotong University, Shanghai. All cell lines were maintained in complete growth medium as recommended by the manufacturer. The cultured cells were maintained in a humidified 5% CO₂ atmosphere at 37°C. Cell lines underwent routine testing for Mycoplasma every 3 months (last confirmed negative date, May 10, 2020). The genetic identity of the cell lines was confirmed by short tandem repeat profiling. The cell lines were used for experiments within ten passages after thawing.

See the Supplementary Materials and Methods section for detailed experimental procedures. The antibodies used in this study are listed in Supplementary Table S2, and the primer sequences used are listed in Supplementary Table S3.

ANOVA and the Student t test were used to determine the statistical significance of differences between samples. Results are expressed as the mean ± SD. Note that χ² and Fisher exact tests were used to analyze the association between miR-124 expression and clinicopathologic parameters. Overall survival (OS) was defined as the interval between the dates of surgery and death. The Kaplan–Meier method was used to compare OS among patients in different groups. The log-rank test was used to estimate differences in survival. Univariate and multivariate analyses were based on the Cox proportional hazards regression model using SPSS 15.0 software (SPSS, Inc.). A value of P < 0.05 was considered as significant.

We have previously reported that miR-124 is downregulated and a tumor-suppressive miRNA involved in pancreatic cancer metastasis (13). Here, we identified that the Notch ligand JAG1 was a regulated target of miR-124 in human pancreatic cancer cells. JAG1 mRNA contains two putative target sequences for miR-124 in its 3′-untranslated region (UTR; Fig. 1A). When we employed a dual luciferase reporter system, we found that coexpression of miR-124 significantly suppressed firefly luciferase reporter activity of the wild-type JAG1 3′-UTR, but not that of a mutant 3′-UTR in human pancreatic cancer cells (Fig. 1A; Supplementary Fig. S1A and S1B). These results indicated that miR-124 suppressed JAG1 expression through the miRNA-binding sequences in the 3′-UTR of JAG1. Next, we found that miR-124 overexpression decreased endogenous expression of JAG1 mRNA and protein (Fig. 1B; Supplementary Fig. S1C), whereas downregulation of miR-124 with an anti–miR-124 inhibitor increased JAG1 expression (Fig. 1C; Supplementary Fig. S1D). To confirm that miR-124 inhibits pancreatic cancer cell invasion by targeting JAG1, we established stable JAG1-silenced PANC-1 transfectants by infecting the cells with a lentivirus encoding shJAG1 (Supplementary Fig. S1E). We found that silencing JAG1 in pancreatic cancer cells significantly decreased liver metastasis in vivo (Fig. 1D), which was similar to the phenotype induced by miR-124. In contrast, ectopic expression of JAG1 using a JAG1 expression vector encoding the entire coding sequence of JAG1, but lacking its 3′-UTR, significantly abrogated the tumor-suppressive effect induced by miR-124 (Supplementary Fig. S1F and S1G; Fig. 1D), which indicated that JAG1 is a functional target of miR-124. Because we have previously shown that high miR-124 expression levels are significantly associated with prolonged OS (13), we next confirmed that a high level of JAG1 expression correlated with shortened survival of patients with PDAC (Fig. 1E and Supplementary Table S4), which was validated in The Cancer Genome Atlas (TCGA) dataset (Fig. 1F). However, when we evaluated the function of JAG1 in mediating the invasion suppression of miR-124 in pancreatic cancer in vitro, we unexpectedly found that silencing JAG1 had little effect on pancreatic cancer cell invasion (Fig. 1G). Furthermore, miR-124 inhibited pancreatic cancer cell invasion in vitro. However, this effect was not abrogated by overexpression of JAG1 (Fig. 1H). These findings were confirmed in two other human pancreatic cancer cell lines (Supplementary Fig. S1H and S1I). Therefore, it is likely that the effect of JAG1 on metastasis in vivo is associated with communication between tumor and microenvironmental cells.

We next observed that miR-124–overexpressing PANC-1–xenografted tumors exhibited decreased JAG1 expression that was accompanied by decreased infiltration of TAMs stained for CD68 and CD206 (Fig. 2A; Supplementary Fig. S2A). This observation was validated in human PDAC tissue when CD68 and CD206 were used to identify TAMs (Fig. 2B). Confocal fluorescence microscopy revealed that JAG1 was mainly expressed in cancer cells, but not in CD68⁺ cells (Supplementary Fig. S2B). Patients with high JAG1 expression were associated with an increase in infiltrated CD68⁺ macrophages (Fig. 2B). In addition, a high number of CD68⁺ macrophages correlated with decreased survival time of patients with PDAC in the Fudan cohort (Fig. 2C), which was validated in the TCGA cohort in which high CD68 expression was correlated with decreased survival time (Fig. 2D). These results suggested cross-talk between cancer cell–derived JAG1 and macrophage activation. To test our hypothesis, we cocultured human pancreatic cancer cells with M2-like macrophages generated by inducing human monocyte–derived macrophages with IL4 (50 ng/mL for 7 days; Supplementary Fig. S3A and S3B) using a Transwell chamber in vitro. JAG1 silencing had little effect on pancreatic cancer cell invasion when cultured alone, whereas JAG1 silencing resulted in a significant decrease in invading cells when the cancer cells were cocultured with macrophages (Fig. 2E; Supplementary Fig. S3C). In addition, JAG1 overexpression abrogated the invasion-suppressive effect induced by miR-124 in the coculture system (Fig. 2F; Supplementary Fig. S3D). We further confirmed that JAG1 overexpression significantly increased cancer cell invasion only in coculture with macrophages, whereas this effect was abolished by human JAG1–neutralizing antibody treatment in the coculture system (Fig. 2G; Supplementary Fig. S3E). Therefore, the results suggest that JAG1 affects cancer cell invasion dependent on the involvement of TAMs in pancreatic cancer. TAMs are derived from inflammatory monocytes recruited to the tumor microenvironment by adhesion to the blood vessel endothelium and extravasation into the tumor where they differentiate into tissue macrophages typically through interactions with cytokines and colony-stimulating factors (14). We next investigated the abilities of distinct pancreatic cancer cells to activate macrophages and their correlation with miR-124/JAG1 expression levels. We identified different abilities among the pancreatic cancer cell lines to activate macrophages in terms of M2 polarization as shown by CD206 expression and tumor-promoting cytokines (CCL17, CCL18, CCL22, and IL10; Supplementary Fig. S4A and S4B) and extravasation (Supplementary Fig. S4C). In addition, pancreatic cancer cells that exhibited higher abilities to activate macrophages had lower miR-124 expression and higher JAG1 expression (Supplementary Fig. S4D–S4F), which suggested the involvement of miR-124/JAG1 signaling in macrophage activation. To assess this possibility, we cocultured human monocyte–derived macrophages and pancreatic cancer cell line Su.86.86 with different expression levels of miR-124 or JAG1. We found that pancreatic cancer cells increased macrophage M2 polarization, whereas JAG1-silenced cells exhibited decreased abilities to activate macrophage M2 polarization (Fig. 2H; Supplementary Fig. S5A). In addition, miR-124–overexpressing cells exhibited decreased abilities to activate macrophages, whereas this effect was abrogated by JAG1 overexpression (Fig. 2H; Supplementary Fig. S5A). We also obtained conditioned medium (CM) from pancreatic cancer cells with different expression levels of miR-124 or JAG1 and further confirmed that miR-124–regulated Notch signaling determined macrophage activation in term of extravasation in pancreatic cancer (Fig. 2I; Supplementary Fig. S5B). We next used a Jagged1-neutralizing antibody and found that cancer cells activated M2 macrophage polarization, whereas treatment with the Jagged1-neutralizing antibody (Fig. 2J; Supplementary Fig. S5C) or treatment with gamma-secretase inhibitor (GSI) PF-03084014 (Supplementary Fig. S5D) abrogated macrophage activation induced by pancreatic cancer cells in the coculture system. We further observed that cancer cell–derived CM activated macrophages, whereas CM from cancer cells pretreated with the Jagged1-neutralizing antibody (Fig. 2K and L; Supplementary S5E and S5F) or GSI (Supplementary Fig. S5G and S5H) exhibited diminished abilities to activate macrophages. Therefore, our results identified miR-124–regulated autocrine JAG1/Notch signaling that activated macrophages in PDAC.

Pancreatic cancer cell–derived JAG1 functions in an autostimulatory fashion, and therefore we next attempted to identify paracrine cytokines secreted by pancreatic cancer cells, which activate macrophages. We compared the cytokine profiles of CM from Su.86.86 cells treated with or without JAG1 siRNA using a human cytokine antibody array. The array followed by qPCR validation confirmed downregulation (decrease in expression by ≥1.5-fold) of nine cytokines in siJAG1-treated Su.86.86 cells compared with their parental cells (Fig. 3A). We next hypothesized that deregulated cytokines may contribute to macrophage activation. To test this hypothesis, siRNAs for each of the nine cytokines were transfected into Su.86.86 cells, and we subsequently examined their effects on macrophage activation. We found that the IL8 siRNA resulted in a decreased effect of Su.86.86 cells on macrophage M2 polarization as detected by flow cytometric analysis of CD206 expression (Supplementary Fig. S6A). Moreover, CCL2 siRNA resulted in a decreased ability of macrophage extravasation (Supplementary Fig. S6B), whereas IL1α and uPA siRNAs resulted in a significantly decreased ability of macrophage adhesion (Supplementary Fig. S6C), which suggested that these cytokines contributed the most to macrophage activation. We next attempted to confirm the roles of the four cytokines in macrophage activation. We first confirmed that the four cytokines, IL8, CCL2, IL1α, and uPA, were all decreased in CM from JAG1-knockdown Su.86.86 cells (Fig. 3B). In addition, their levels in Su.86.86 CM were all determined by miR-124–regulated Notch signaling (Fig. 3B). We further evaluated whether miR-124 directly regulates these cytokines. Using the miRNA target prediction program, we found that only the 3′-UTR of CCL2 mRNAs contain miR-124–binding site (Supplementary Fig. S6D). Pancreatic cancer cells were then cotransfected with miR-124 mimics or inhibitors with the luciferase reporter construct containing the wild-type CCL2 3′-UTR. We found that coexpression of miR-124 could not suppress the firefly luciferase reporter activity of the 3′-UTR of CCL2, which suggested that miR-124 did not affect its expression directly (Supplementary Fig. S6D). Because Notch-dependent processes require release of the Notch intracellular domain (N^IC), which then directly translocates into the nucleus to activate Notch-target gene transcription at CBF1/RBPJκ DNA-binding sites (15), we next identified four phylogenetically conserved N1^IC-binding sites located in CCL2 promoter areas (Supplementary Fig. S6E). Further chromatin immunoprecipitation (ChIP) analysis showed that N1^IC bound to the CCL2 promoter in pancreatic cancer cells (Supplementary Fig. S6F and S6G). Next, we confirmed the function of these four Notch-dependent cytokines in terms of macrophage activation. We found that recombinant human IL8 (rhIL8) significantly induced macrophage M2 polarization as shown by increases in CD206 expression and tumor-promoting cytokines (Fig. 3C; Supplementary Fig. S7A). Recombinant human CCL2 (rhCCL2) significantly increased macrophage extravasation (Fig. 3D), whereas recombinant human IL1α (rhIL1α) and recombinant uPA significantly promoted macrophage adhesion (Fig. 3E and F). Furthermore, addition of their corresponding neutralizing antibodies to the CM of pancreatic cancer cells significantly abolished the effects induced by the CM from cancer cells (Fig. 3C–F; Supplementary Fig. S7B–S7G). Next, we confirmed that IL8 knockdown in cancer cells abolished the Notch signaling–mediated promotion of macrophage M2 polarization (Fig. 3G; Supplementary Fig. S7H), whereas CCL2, IL1α, and uPA knockdown abolished the Notch signaling–mediated promotion of macrophage recruitment as detected by extravasation and adhesion assays, respectively (Fig. 3H–J; Supplementary Fig. S7I–S7K). Because TAMs generally show an alternative form of activation, referred to as M2 macrophages that produce high amounts of protumoral cytokines, we performed qPCR assays to screen a panel of cytokines related to M2 macrophages in macrophages treated with CM from Su.86.86 cells with different Notch signaling activities (16). As expected, we observed a significant correlation between the M2-related cytokine profile and Notch signaling. Among all tested cytokines, IL6 was the most responsive cytokine in Su.86.86 CM–activated macrophages (Fig. 3K). We further investigated this finding by an ELISA and confirmed that Notch signaling promoted IL6 production from macrophages, which was abolished by IL8 knockdown (Fig. 3L; Supplementary Fig. S7L). Taken together, our results suggested that Notch signaling–dependent cytokine secretion contributed to macrophage activation.

We found that IL6 production was increased the most in response to Notch signaling, which suggested that IL6 contributes to the protumorigenic function of macrophages induced by Notch signaling. To confirm this possibility, we first found that treatment of pancreatic cancer cells with recombinant human IL6 (rhIL6) enhanced cancer cell migration and invasion in a dose-dependent manner (Supplementary Fig. S8A and S8B). Previous studies have reported that TAMs enhance pancreatic cancer cell invasion and metastasis by inducing epithelial–mesenchymal transition (EMT; ref. 17). Thus, we investigated the effect of IL6 on EMT. Treatment with rhIL6 downregulated E-cadherin expression, whereas vimentin expression was upregulated in pancreatic cancer cells (Supplementary Fig. S8C). These results were confirmed by Western blotting that indicated that the expression of the epithelial marker E-cadherin was decreased, whereas the expression of mesenchymal markers (vimentin, snail, slug, and twist) was increased and accompanied by an increased amount of phosphorylated STAT3 (p-STAT3; Supplementary Fig. S8D). To determine whether IL6 contributes to the protumorigenic function of activated macrophages, we employed an anti-IL6 antibody for neutralization assays. We found that CM from activated macrophages increased the migration and invasion of pancreatic cancer cells, whereas addition of the anti-human IL6 antibody to the culture system of pancreatic cancer cells abolished these effects (Fig. 4A; Supplementary Fig. S8E), which was affected by Notch signaling when manipulating miR-124 or JAG1 expression (Supplementary Fig. S8F). Immunofluorescence staining and Western blot assays confirmed that CM from activated macrophages increased EMT of pancreatic cancer cells, whereas IL6 neutralization abolished these effects (Fig. 4B; Supplementary Fig. S8G). We further evaluated the association of IL6 expression with survival of patients with PDAC. We found that patients with high IL6 expression in the tumor stroma exhibited a shortened survival time compared with those with low IL6 expression (32.1 vs. 12.3 months, respectively; log-rank test, P = 0.001, Fig. 4C). This association in patients with PDAC was validated in the TCGA cohort (Fig. 4D). Therefore, these results suggest that activated macrophages promote EMT and invasion of pancreatic cancer cells via IL6.

We further found that rhIL6 treatment induced EMT and invasion of pancreatic cancer cells accompanied by decreased miR-124 expression. Further analyses confirmed that the expression of primary and mature miR-124 was decreased after exposure to rhIL6 in a dose- and time-dependent manner (Fig. 4E and F). Moreover, the repression of miR-124 expression was reversed by STAT3 knockdown, which suggested that STAT3 mediates the repression of miR-124 by IL6 (Fig. 4G). Next, to determine whether STAT3 regulates miR-124 at the transcriptional level, we investigated STAT3-binding sites in the miR-124 promoter. MiR-124 is present in three genomic loci [miR-124-1 (8p23.1), miR-124-2 (8q12.3), and miR-124-3 (20q13.33)]. Inspection of the miR-124 genomic region revealed four phylogenetically conserved STAT3-binding motifs located in chromosomes 8 and 20 (Fig. 4H). ChIP analysis showed that STAT3 bound strongly to the miR-124 promoter in pancreatic cancer cells (Fig. 4I and J). In addition, luciferase reporters, including each STAT3BS of miR-124, showed significantly decreased activities in STAT3-overexpressing pancreatic cancer cells. However, these effects were reversed as the STAT3BSs were mutated (Fig. 4K). To finally confirm that STAT3 protein bound to the miR-124 promoter, an electrophoretic mobility shift assay was performed using an oligonucleotide probe from the miR-124 locus. As a result, STAT3 specifically bound to the fluorescein amidite–labeled probe. Similar results were observed in a supershift assay using an anti-STAT3 antibody (Fig. 4L). Collectively, our results suggested that activated macrophage-derived IL6 repressed miR-124 expression at the transcriptional level by STAT3.

We established a Notch-dependent feedback circuit between pancreatic cancer cells and macrophages, which contributed to pancreatic cancer cell invasion. We next examined whether disrupting the circuit suppressed metastasis. To this end, we first established a xenografted mouse model by injecting human Su.86.86 cells into the subcapsular region of the NOD/SCID mouse spleen. The mice were then treated with control IgG or neutralizing antibodies against human Jagged1 via tail vein injection. We found that mice treated with the neutralizing antibodies developed significantly fewer liver metastases than IgG-treated mice. They also displayed decreased numbers of metastatic colonies and a decrease in the ratio of the metastatic area to the total area (Fig. 5A and B). IHC performed on Su.86.86-xenografted tumors with human-specific antibodies showed that expression of IL8, IL1α, CCL2, and uPA was decreased after treatment with the neutralizing antibody against human Jagged1 (Fig. 5C). In addition, neutralizing antibody treatment led to decreased infiltration of TAMs stained by a mouse-specific CD68 antibody (Fig. 5D). To further examine the cross-talk between cancer cells and macrophages in vivo, we established a Panc02-grafted mouse model by injecting murine Panc02 cells into the subcapsular region of the C57BL/6 mouse spleen. The mice were then treated with control IgG or neutralizing antibodies against murine Jagged1, IL1α, CCL2, uPAR, or IL6 via tail vein injection after 7 days. We identified that mice treated with the neutralizing antibodies developed significantly fewer liver metastases than IgG-treated mice (Fig. 5E and F). The Kaplan–Meier analysis demonstrated that the corresponding neutralizing antibody treatments increased the survival of mice compared with IgG treatment (Fig. 5G). Consistent with the suppression of liver metastasis by disrupting the circuit, IHC revealed an increase in miR-124 expression, decreases in JAG1, IL1α, CCL2, uPA, IL6, and p-STAT3 expression, and a decrease in tumor macrophage infiltration in tumors after treatment with the neutralizing antibodies targeting different cytokines of the circuit (Fig. 5H; Supplementary Fig. S9). To validate the results from the Panc02-grafted mouse model, we also established a murine KPC1199–grafted mouse model and then treated the mice with neutralizing antibodies against murine Jagged1, IL1α, CCL2, uPAR, or IL6. As shown in Supplementary Fig. S10A and S10B, mice treated with the neutralizing antibodies developed significantly fewer liver metastases than IgG-treated mice. Because we proposed cross-talk between cancer cells and tumor microenvironment cells (although specifically macrophages), the presence of other immune cell types was also evaluated. Murine Panc02 cells with JAG1 knock down were orthotopically transplanted into the pancreas of immunocompetent C57BL/6 mice. The tumor-infiltrating immune cell content was then analyzed by flow cytometry. As shown in Supplementary Fig. S11A, notably increased infiltration of CD8⁺ and CD4⁺ T cells and B cells, and decreased infiltration of TAMs and myeloid-derived suppressor cells (MDSC) were observed among murine pancreatic cancer cells with JAG1 knockdown. We also performed Sirius Red staining to detect the degree of fibrosis and stained α-SMA+ fibroblasts by IHC. The results suggested that Notch signaling had little effect on fibrosis and cancer-associated fibroblast accumulation in PDAC (Supplementary Fig. S11B and S11C). Taken together, our study suggests that the circuit is a therapeutic target to inhibit PDAC metastasis and that any step of the circuit can be targeted.

We further evaluated the feedback circuit in human PDAC. We determined the expression of other components of the circuit, including IL8, IL1α, CCL2, uPA, IL6, p-STAT3, and E-cadherin, by IHC in the serial sections used in Fig. 2B (Fig. 6A, top). We observed significant correlations where JAG1 expression was positively associated with IL8, IL1α, CCL2, uPA, IL6, and p-STAT3 expression and CD68⁺ TAM infiltration and was negatively with E-cadherin and miR-124 expression (Fig. 6B, bottom; Supplementary Fig. S12). In addition, increased IL6 and p-STAT3 expression was correlated with decreased miR-124 expression in PDAC tissue (Fig. 6B; Supplementary Fig. S12). Finally, we confirmed that increased IL8, IL1α, CCL2, uPA, and p-STAT3 expression was correlated with shortened survival, whereas increased E-cadherin expression correlated with prolonged survival time in patients with PDAC (Fig. 6B). Therefore, these results confirmed that the Notch-dependent feedback circuit is perturbed in human PDAC, which contributes to poor survival of patients with PDAC (Fig. 6C).

In this study, we showed that miR-124–regulated Notch signaling activated macrophages to a tumor-supporting M2-like phenotype. In turn, the activated macrophages had enhanced Notch signaling by IL6–STAT3-induced miR-124 silencing and the EMT phenotype. Therefore, our study identified a Notch signaling–dependent inflammatory feedback circuit between pancreatic cancer cells and macrophages, which contributes to PDAC metastasis.

Initial reports generally indicated that Notch signaling favors proinflammatory macrophages. JAG1 is one of the ligands responsible for autoamplification of Notch signaling in proinflammatory macrophages (18). When the Notch signaling pathway is blocked pharmacologically or genetically, some important proinflammatory functions are compromised (19). Hyperactivation of Notch signaling specifically in macrophages of the tumor mass has also been shown to suppress tumor growth in an animal model of cancer (20). Macrophages deficient for canonical Notch signaling show TAM phenotypes, and the compromised Notch pathway activation leads to M2-like TAMs (21). Recent evidence has also indicated the involvement of cancer cell–derived JAG1 in driving TAMs to a protumoral phenotype by activating Notch signaling in TAMs (13, 22). All of these previous studies suggest that Notch signaling is involved in various types of macrophage activation in a context-dependent manner. In our study, we identified a positive correlation between JAG1 and TAM infiltration in PDAC. However, blockade of Notch signaling in macrophages had little influence on macrophage activation, whereas Notch signaling inhibition in cancer cells prevented the M2 polarization of macrophages. These data suggest that JAG1 triggers autoamplification of Notch signaling in cancer cells, which finally contributes to activation of macrophages to a tumor-supporting M2-like phenotype in PDAC.

The tumor microenvironment is enriched with factors that recruit circulating monocytes and favor the generation of TAMs resembling M2 macrophages. In this study, we identified Notch signaling–dependent cytokine secretion in cancer cells, which contributed to macrophage activation. Among the cytokines, we identified Notch-dependent CCL2 derived from PDAC cells, which contributed to monocyte extravasation. This finding is consistent with the previous consensus that the CCL2/CCR2 axis plays a significant role in tumor monocyte/macrophage recruitment (23), and increased bone marrow monocyte mobilization correlates with increased PDAC infiltration by CCR2⁺ macrophages and is associated with poor survival of patients with PDAC (24). Recent studies have shown that CCR2 inhibition reduces tumor myeloid cells and unmasks a checkpoint inhibitor effect (25, 26). Our study suggested that Notch signaling contributed to macrophage activation and targeting Notch signaling increased infiltration of CD8⁺ T cells and decreased infiltration of TAMs and MDSCs in murine pancreatic cancer, which provide the basis to combine Notch and PD-1 blockade in a future study of PDAC treatment. Another Notch-dependent cytokine that contributed to monocyte recruitment was IL1α that is produced by tumor cells or adjacent stromal cells and is associated with the development of many different human cancers (27). High IL1α levels are also associated with distant metastases and a poor survival rate of patients with PDAC (28). Our study identified a new mechanism by which IL1α promotes cancer metastasis by recruiting monocytes into cancer tissue. A recent study also identified the uPA–uPAR system as a central player in mediating proteolysis during cancer invasion and metastasis (29). Elevated uPA expression is related to adverse outcomes of patient with different types of cancers including PDAC (30). Recent studies have also suggested a role of the uPA–uPAR system in mediating different types of immune responses (31). uPAR-deficient mice fail to recruit neutrophils and macrophages to the site of bacterial infection (32, 33), which suggests a role of uPA in the recruitment of monocytes/macrophages. Here, we showed that recombinant human uPA increased monocyte adhesion, whereas an anti-uPAR antibody abrogated the increased adhesion induced by CM from cancer cells. Therefore, our study revealed a new mechanism of the metastasis-promoting ability of uPA, whereby PDAC cell–derived uPA recruits monocytes into cancer tissue. In addition to Notch-dependent CCL2, IL1α, and uPA contributing to monocyte recruitment and macrophage infiltration in cancer, we also identified a function of Notch-dependent IL8 in determination of the M2 macrophage polarization. To date, polarized activation of macrophages has been studied extensively, and five main M2 stimuli have been identified, including colony-stimulating factor (CSF)-1, chemokines, and cytokines such as IL4, IL13, TGFβ, and IL10. Recently, studies have suggested that other cytokines, including GM-CSF and IL8 (34, 35), also contribute to macrophage M2 polarization. Therefore, our study combined with others suggest a significant role of IL8 in M2 polarization, which extends our understanding of autoamplified Notch signaling in cancer cells in the shaping of a pro-oncogenic inflammatory microenvironment.

The regulating mechanism underlying activated Notch signaling remains unknown, although activated Notch signaling has been identified in many cancers including PDAC. In this study, we found that the Notch ligand JAG1 was overexpressed in cancer cells and a directly regulated target of miR-124, which suggested a significant regulatory role of miR-124 in activated JAG1/Notch signaling. We have previously shown that miR-124 may be epigenetically silenced through tumor-specific methylation in PDAC (13). Here, we further demonstrated that activated macrophage-derived IL6 activated the oncogenic transcription factor STAT3 that directly repressed miR-124 genes via a conserved STAT3-binding site in their promoters. Therefore, our study suggests a miRNA-mediated Notch signaling regulatory mechanism that participates in pancreatic cancer metastasis.

Our study suggests that the regulatory loop is a therapeutic target to inhibit pancreatic cancer metastasis and can be affected at any step. In this study, mice injected with neutralizing antibodies against anti-JAG1, IL8, IL1α, CCL2, uPAR, or IL6 developed significantly fewer liver metastases and exhibited prolonged survival time. Therefore, our study provides many potential avenues for disrupting this circuit. GSIs are the best-studied small molecules that target the Notch pathway. There are several GSIs in different stages of clinical trials, including MK-0752 and RO4929097 (36). In preclinical studies, therapeutic anti-JAG1 antibodies have decreased cancer stem cells and bone metastasis in breast cancer (37, 38). Targeting TAMs with the CCR2 inhibitor PF-0413630 in combination with FOLFIRINOX was safe and tolerable in patients with advanced PDAC in a phase 1b trial (39). In addition, IL8 receptor CXCR1 and CXCR2 inhibitor Reparixin (40), anti-IL1α antibody MABp1 (41), uPA inhibitor upamostat (WX-671; ref. 42), and Siltuximab, an anti-IL6 monoclonal antibody, have been evaluated in patients with advanced solid tumors (43). Therefore, our findings provide potential therapeutic targets, and certain agents that have been used for inhibition of these targets to treat other diseases can be rapidly repurposed for PDAC treatment as information becomes available on their pharmacology, formulation, and toxicity.

Taken together, we have demonstrated an important role of a Notch-dependent inflammatory feedback circuit between pancreatic cancer cells and macrophages in PDAC metastasis. Importantly, we have provided potential therapeutic targets and robust preclinical evidence for PDAC treatment by interrupting the feedback with targeted inhibitors.

No disclosures were reported.

Y. Geng: Data curation, formal analysis, investigation, methodology, writing-original draft. J. Fan: Data curation, validation, investigation, visualization, methodology. L. Chen: Data curation, software, investigation, visualization, methodology. C. Zhang: Formal analysis, investigation. C. Qu: Data curation, software, investigation. L. Qian: Investigation, visualization, methodology. K. Chen: Data curation, software, formal analysis, investigation. Z. Meng: Resources, data curation, methodology. Z. Chen: Resources, data curation, supervision, visualization. P. Wang: Conceptualization, resources, data curation, software, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, project administration, writing-review and editing.

We thank Professor Jing Xue from Shanghai Jiaotong University, Shanghai, for kindly providing KPC1199 cell line. This study was supported by the National Natural Science Foundation of China (81370068, 81572376, 81622049, and 81871989 to P. Wang), Shanghai Science and Technology Committee Program (19XD1420900 to P. Wang), Shanghai Education Commission Program (17SG04 to P. Wang), and Shanghai Municipal Health Commission program [ZY(2018-2020)-CCCX-2005 to P. Wang].

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.