Human PC cell lines BxPC-3, Panc-1, Miapaca-2 and healthy pancreatic cell line HPDE6-C7 were obtained from Cancer Institute of Chinese Academy of Sciences (Beijing, China). All cell lines were maintained in an appropriate cell culture medium (see online supplementary additional file 2: table S1), supplemented with 10% fetal bovine serum (HyClone, Logan, USA) and 100 µg/mL each of penicillin and streptomycin (Gino, Hangzhou, China) in 5% CO₂ at 37℃. All cell lines were used within 6 months of thawing. The animal experiment abided by national guidelines for the care and use of animals and approved by the ethics committee of Huzhou Central Hospital (HzhcLS2018-0807). C57BL/6 (B6) mice were purchased from Experimental Animal Center of Weitonglihua Co., Ltd. (Beijing, China), and were maintained in laminar airflow cabinets under specific pathogen-free conditions, and provided free access to irradiated pellet food and sterilized water. An ectopic inoculation method was used to establish a tumor xenograft model (six per group). Anti-PD1 blocking antibody (BioXcell Life Sciences RMP1-14, West Lebanon, USA) was used at a dose of 200 mg per mouse and performed intraperitoneally on the indicated days. Recombinant human IFN-γ (Wanbang Biopharma, Shanghai, China) was diluted in 2.5 mL of PBS (4×10⁵ U/bottle) and given intraperitoneally with 25 µL (1×10⁴ U) per mouse on continuous 5 days per week. Tumor tissues were harvested on indicated day after tumor implantation (day 21 after tumor implantation). Tumor volumes and survival time of each mouse was documented in the duration of the experiments. Tumor volumes were calculated using the formula: volume=length×width²/2. The endpoint for tumor growth and survival analysis was day 42 after tumor implantation.

Rat anti-mouse CD4 (clone GK1.5), rabbit anti-mouse CD68 (clone 2449C) and rabbit anti-mouse CXCR1 (clone 1122A) antibodies were purchased from R&D Systems Inc. Rat anti-mouse CD8 antibody (clone 53-6.7) was purchased from BD Pharmingen. Mouse anti-human CD68 (clone Y1/82A) and mouse anti-human PD-L1 (clone MIH1) antibodies were purchased from eBioscience Inc. Mouse anti-human CXCL8 (clone E8N1), rat anti-mouse CXCR2 (clone SA045E1) and mouse anti-human CXCR2 (clone 5E8/CXCR2) antibodies were purchased from BioLegend. All antibodies were used according to the manufacturer’s instructions. Intracellular cytokine staining was performed after the whole cells incubated with soluble anti–IL-2 (clone JES6-5H4; BioLegend) or anti–TNF-α (clone MP6-XT22; eBioscience) antibodies and monensin at 37°C and 5% CO₂ for 6 hours. Flow cytometry was performed with FACS Fortessa (BD Biosciences), analyzed by FlowJo V.9.0 software. In all flow cytometry assays, background versus positive staining was ascertained using isotype controls and fluorescence minus one (FMO) controls. Boxed populations quantified were identified as positive in relation to isotype and/or FMO controls.

Tissue sections from isolated mice tumor xenografts and human PCs were deparaffinized in xylene and rehydrated with ethanol, followed by antigen repair with EDTA buffer (PH 8.0) boiling for 10 min, then samples were incubated with 2% BSA for 30 min and added primary antibody at 4°C overnight. After three times PBS washes, HRP-conjugated secondary antibody (corresponding to species of primary antibody) was added for 30 min. Following three times PBS washes, fluorophore was added for 10 min. The fluorophores used for dual-immunofluorescence assays were FITC for CD68 and CY3 for CXCR1/CXCR2 (figure 1C), while the fluorophores used for three-marker immunofluorescence assays were CY3 for CD68, FITC for CXCR2 and CY5 for CXCL8 (figure 2). After DAPI staining for 10 min, immunofluorescence of tissue sections was evaluated on inverted fluorescent microscope (BX-51, TR32000 Olympus, Japan) and analyzed by Image J 2X software (Rawak Software, Germany). Immunoblotting for detection of CXCL8 expression in cells and tumor tissues was performed as previously described.24 The primary antibodies used for immunoblotting were mouse anti-human CXCL8 (1:1000; BioLegend) and mouse anti-human GAPDH (1:1000, clone 1E6D9; ProteinTech). PE-conjugated goat anti-mouse IgG (1:2000; Santa Cruz) was used as secondary antibody. Reactive bands were detected using enhanced chemiluminescence (ECL; Amersham Bioscience, USA) immunoblotting detection reagent and analyzed by Image Lab V.5.2 software (Bioscience, USA).

Total RNA was extracted from cultured cells, sorted CXCR1⁺/CXCR2⁺ macrophages and tumor xenografts from mice using the RNeasy Mini Kit (Qiagen, Hilden, Germany). Complementary DNA was synthesized using M-MLV reverse transcriptase (Takara Bio, Dalian, China). Real-time fluorescence quantitative PCR (RTFQ-PCR) was performed with specific primers for detecting murine arginase-1, iNOS, PD-L1, GAPDH and human CXCL8, GAPDH (see online supplementary additional file 2: table S2). The primers were obtained from Wuhan Servicebio Technology Co., Ltd., and the assays were performed according to the manufacturer’s protocol. The PCR reactions were performed in a Bio-Rad Hard Shell 96-well microplate on a Bio-Rad CFX96 Real-Time System C1000 Touch Thermal Cycler using iTaq Universal SYBR Green Supermix (Bio-Rad) and amplified for 40 cycles. CXCL8 levels of serum samples and cell culture supernatants were detected by using the human CXCL8 sandwich ELISA kits (Dakewe Biotech, Shenzhen, China) according to the manufacturer’s instructions. Normal ranges for human CXCL8 were defined by the range of values spanning (means±2 SDs) obtained from the CXCL8 level data of serum from 12 healthy donors.

Peripheral blood CXCR1⁺/CXCR2⁺ macrophages (1×10⁴) which electronically sorted from tumor-bearing mice were placed in the top chamber of a Transwell (Millipore, USA). BxPC-3 or Panc-1 cells were plated in the bottom chamber. Blocking antibodies for CXCL8, CXCR1 and CXCR2 (50 µg/well) were added at the beginning of the experiment. The total number of macrophages migrating from the top to the bottom chamber after 12 hours was calculated. The migrated cells counting method was performed as previously described.24

Serum samples (n=44) from patients with PC were obtained at the time of hospitalization in Huzhou Central Hospital from May 2016 to January 2019, following informed consent. In these patients, 36 patients underwent radical resection and the surgical resection specimens (n=36) were collected during the operation. All fresh specimens containing pancreatic tumor or adjacent normal pancreatic tissue were selected for further study. There are 25 men and 19 women in these patients (male:female=1.32:1); the median age was 52 years (range, 37–84 years). Serum samples from healthy donors (n=12) were obtained after informed consent and following approved protocols (ChiCTR1800017665).

SPSS V.19.0 was used for clinical data statistical analysis. The χ² test and Fisher’s exact test were used for quantitative data (table 1). Graphs for experimental data were created using GraphPad Prism software. Data in graphs were expressed as mean±SD. P values were calculated in each respective figure using paired or unpaired t-test for intergroup comparison and log-rank test for survival analysis. P value <0.05 was considered statistically significant.

First, we analyzed the effect of PD1 blockade therapy in PC xenografts using BxPC-3 cells. BxPC-3 tumors grow rapidly after tumor inoculation into the axilla of mice, contributing to 100% mortality by day 27 (figure 3A and online supplementary additional file 1: figure S1a). BxPC-3 tumors constitutively overexpressed PD-L1 (figure 3C and online supplementary additional file 1: figure S2, left panel) and enhanced PD1 expression on circulating CD4⁺ or CD8⁺ T cells by day 14 (figure 3B and online supplementary additional file 1: figure S3a), leading us to hypothesize that the PC-induced PD1/PD-L1 signal pathway on T cells contributes to immune escape in BxPC-3 tumors. Subsequently, we administered anti-PD1 blocking monoclonal antibodies (mAbs) to mice on the day of tumor inoculation and the mice showed 100% survival without evidence of tumor development (figure 3A). This indicated a role for PD1/PD-L1 signaling in immune escape by PC. To detect whether anti-PD1 treatment could lead to the regression of established BxPC-3 tumors, anti-PD1 mAbs were administered 7 days after tumor inoculation, at which time the PC tumors were well established. We found the delayed treatment modestly decreased the tumor growth and lead to a minor statistically significant improvement in survival (figure 3D). Although delayed anti-PD1 treatment showed minor therapeutic efficacy on tumor regression, BxPC-3 tumor-bearing mice after delayed anti-PD1 treatment had significantly more interleukin-2–positive (IL-2⁺) and tumor necrosis factor-α–positive (TNF-α⁺) CD8⁺ T cells infiltrating the tumors than untreated control mice (figure 3E and online supplementary additional file 1: figure S4a), suggesting that delayed anti-PD1 treatment could induce immune activation of tumor-bearing mice but it was insufficient to mediate tumor regression.

We also detected the antitumor effects of PD-1 blockade in mice inoculated with Panc-1, another PC cell line which overexpresses PD-L1 in tumor xenografts (figure 3C and online supplementary additional file 1: figure S2, right panel) and demonstrated upregulated PD1 expression on peripheral T cells (figure 3B and online supplementary additional file 1: figure S3b). The therapeutic efficacy of anti-PD1 treatment was similar to that observed in mice with BxPC-3 tumors when PD-1 blockade therapy was initiated on the day of tumor inoculation (figure 3F and online supplementary additional file 1: figure S5a). The delayed treatment initiated on day 7 had no significant impact on tumor growth and survival rate (figure 3G and online supplementary additional file 1: figure S5b); however, immune activation was observed (figure 3H and online supplementary additional file 1: figure S4b).

To explore the underlying mechanisms by which PC may limit the efficacy of immune-based therapy, we analyzed TAMs in the tumors and peripheral blood of tumor-bearing mice. We found that CXCR1/2⁺CD68⁺ macrophages were increased in the peripheral blood of mice as early as day 7 after tumor inoculation (figure 1A). The proportion of CXCR1/2⁺CD68⁺ macrophages in peripheral blood increased following the growth of BxPC-3 and Panc-1 tumors (figure 1). By day 21, CXCR1/2⁺CD68⁺ macrophages represented a sizeable fraction of the monocytes in BxPC-3 and Panc-1 tumors compared with healthy tissue (figure 1C). The expansion of CXCR1/2⁺CD68⁺ macrophages in Panc-1 tumors was less significant than in BxPC-3 tumors, and the expansion of CXCR2⁺CD68⁺ macrophages greater than CXCR1⁺CD68⁺ macrophages in PC.

To investigate whether the CXCR1/2⁺CD68⁺ macrophages expanded in PC tumor-bearing mice were suppressive cells, we sorted CXCR1/2⁺CD68⁺ cells from tumor-bearing mice and evaluated their ability to produce arginase-1 and inducible nitric oxide synthase (iNOS), mediators of macrophages immune suppression. We found that higher levels of arginase-1 were observed in tumor-infiltrating CXCR1/2⁺CD68⁺ macrophages which from BxPC-3 and Panc-1 tumor-bearing mice, whereas the levels of iNOS were decreased (figure 1D), suggested M2 polarization. Both circulating and tumor-infiltrating CXCR1/2⁺CD68⁺ macrophages from BxPC-3 and Panc-1 tumor-bearing mice expressed higher levels of PD-L1 compared with macrophages from healthy mice (figure 1E and online supplementary additional file 1: figure S6). We also detected the fraction of tumor-infiltrating T-cell subsets, such as Th1 (CD4⁺IL-2⁺, CD4⁺TNF-α⁺), Th2 (CD4⁺IL-4⁺, CD4⁺IL-10⁺), Treg (CD4⁺FoxP3⁺), CD8⁺IL-2⁺ and CD8⁺TNF-α⁺ cells. We observed a significant reduction of Th1, CD8⁺IL-2⁺ or CD8⁺TNF-α⁺ cells, whereas we observed an increase of Th2 or Treg cells in the tumor bed (see online supplementary additional file 1: figure S7).

We next investigated the mechanisms by which TAMs traffic to the PC tumor bed to mediate their immunosuppressive effect. We detected whether PCs produce and secrete ligands for CXCR1/2 and observed that BxPC-3, Panc-1 and Miapaca-2 PC cell lines overexpressed CXCL8 mRNA and protein, and the culture supernatant taken from these cell lines contained high concentrations of CXCL8 protein compared with a healthy pancreatic cell line HPDE6-C7 (figure 4A). In tumor-bearing mice, BxPC-3 and Panc-1 tumors overexpressed CXCL8 protein and high concentrations of CXCL8 protein was also detected in the serum (figure 4B).

To confirm whether CXCL8 induced a chemokine gradient used for CXCR1/2⁺CD68⁺ macrophages trafficking to the tumor bed, we used an in vitro migration assay where BxPC-3 and Panc-1 cells were plated in the bottom chamber, and sorted CXCR1/2⁺CD68⁺ cells isolated from the blood of BxPC-3 and Panc-1 tumor-bearing mice were added to the top chamber of a 96-well Transwell plate. Subsequently, blocking antibodies toward CXCL8, CXCR1/2 and CXCL8+CXCR1/2 were added. After 12 hours, the migration of CXCR1/2⁺CD68⁺ cells was enhanced in the presence of BxPC-3 and Panc-1 cells (p<0.0001 and p<0.0001, respectively) (figure 4C,D). Using CXCL8 or CXCR2 blocking antibody significantly inhibited the migration of CXCR2⁺CD68⁺ cells in the Panc-1 (p=0.0031 and p=0.0003, respectively) and BxPC-3 groups (p=0.0011 and p<0.0001, respectively) when used alone, and their combined effects were additive (p<0.0001 in Panc-1 group and p<0.0001 in BxPC-3 group, respectively) (figure 4D). Furthermore, the blocking antibody for CXCR1 significantly prohibited the migration of CXCR1⁺CD68⁺ cells in the Panc-1 and BxPC-3 groups (p=0.0108 and p=0.0035, respectively), and CXCL8 blockade shown no inhibitory effects, similar to untreated controls (p=0.6437 in Panc-1 group and p=0.8180 in BxPC-3 group, respectively). The blockade of CXCR1 and CXCL8 simultaneously was equivalent to anti-CXCR1 used alone (p=0.3846 in Panc-1 group and p=0.3819 in BxPC-3 group, respectively) (figure 4C), demonstrating that CXCL8, as a ligand of CXCR1/2, mediates a chemokine gradient effect for CXCR2⁺CD68⁺, but not CXCR1⁺CD68⁺ macrophages trafficking to the tumor bed. The same results mirrored in the in vivo experiments (figure 4E,F and online supplementary additional file 1: figures S8 and S9).

Our study revealed a critical role for CXCR2⁺CD68⁺ macrophages in murine PC progression and CXCL8 is a key chemokine in recruiting CXCR2⁺CD68⁺ macrophages to tumors. We next explored whether human PCs use the same axis. We have demonstrated that human PC cell lines secrete high concentrations of CXCL8 protein into the culture supernatants previously; thus, we measured the levels of CXCL8 in serum from patients with PC (n=44) and healthy donors (n=12). We found that patients with PC had higher levels of CXCL8 compared with healthy donors (patients with PC: 146.0±16.11; healthy donor: 44.42±3.747, p=0.0019) (figure 5A, left panel). We divided patients with PC into high/low (19/25) CXCL8 level groups according to the median value of CXCL8 levels of the total patients with PC, with a follow-up of 24.3 months, the survival rate of the two groups had no significant differences (HR 1.248; 95% CI 0.7017 to 2.414; p=0.4391) (figure 5B, left panel).

We next detected CXCR2 expression on peripheral and tumor-infiltrating CD68⁺ macrophages and found that patients with PC have a higher percentage of CXCR2⁺CD68⁺ cells in the peripheral blood compared with healthy donors (p<0.0001) (figure 5A, middle panel). Furthermore, CXCR2⁺CD68⁺ macrophages represent a higher fraction in tumor tissues compared with adjacent normal pancreatic tissue (p<0.0001) (figure 5A, right panel). We divided patients with PC into high/low CXCR2⁺CD68⁺ expression groups in peripheral blood (18/26) and tumor (13/23) according to the median value respectively, and all groups were compared with clinicopathological parameters (age, sex, tumor size, tumor site, histological grade, T classification, N classification, M classification, clinical stage) by χ² test. We found that CXCR2 expression on peripheral CD68⁺ macrophages was closely correlated with T classification (χ²=6.631, p=0.010), N classification (χ²=4.400, p=0.036) and clinical stage (χ²=5.107, p=0.024). There was no significant correlation between CXCR2 expression and age, sex, tumor size, tumor site, histological grade and M classification (table 1). The same results were observed for CXCR2 expression on tumor-infiltrating CD68⁺ macrophages (see online supplementary additional file 2: table S3). As CXCL8 levels seemed to have no relationship with the overall survival of patients with PC, we next explored whether increased CXCR2 expression on peripheral and tumor-infiltrating CD68⁺ macrophages correlated with diminished overall survival in patients with PC. With a follow-up of 24.3 months, the median survival of patients with high (n=18) versus low (n=26) CXCR2 expression on peripheral CD68⁺ macrophages was 17.0 months versus 24.0 months (HR 3.358; 95% CI 3.138 to 14.99; p<0.0001) (figure 5B, middle panel). Furthermore, high (n=13) versus low (n=23) CXCR2 expression on tumor-infiltrating CD68⁺ macrophages showed a median survival of 18.0 months versus 24.0 months (HR 2.416; 95% CI 1.588 to 8.434; p=0.0053) (figure 5B, right panel). These results demonstrated that CXCL8 is an essential chemokine for the recruitment of CXCR2⁺CD68⁺ macrophages in murine PC tumors as well as in human patients with PC. Furthermore, the expansion of CXCR2⁺CD68⁺ macrophages contributes to human PC progression and poor prognosis.

CXCL8 and CXCR2⁺CD68⁺ macrophages are capable of promoting PC progression and limiting antitumor T-cell responses and our previous studies showed that IFN-γ suppresses CXCL8 in BxPC-3 and Panc-1 cell lines24 (see online supplementary additional file 1: figure S10); thus, we sought to investigate whether anti-PD1 therapy combined with using IFN-γ was more effective at reducing tumor progression. We focused on the BxPC-3 model because PD1 blockade therapy is highly efficacious and has a significant expansion of CXCR2⁺CD68⁺ macrophages. We tested the expression and secretion of CXCL8 in IFN-γ treated tumor-bearing mice and found that the expression of CXCL8 mRNA and protein decreased on days 7, 14 and 21 after IFN-γ treatment. These results mirrored the concentrations of CXCL8 protein in the serum of tumor-bearing mice (figure 2A). Mice were inoculated with BxPC-3 cells on day 0, and treatment begun on day 7, when the tumors were well established. Anti-PD1 or IFN-γ treatment alone was ineffective at preventing tumor growth (figure 2B, left and middle panel). In contrast, mice treated with anti-PD1+IFN-γ experienced significant reductions in tumor growth (figure 2B, right panel) (see online supplementary additional file 1: figure S11). Kaplan-Meier survival analysis in BxPC-3 tumor-bearing mice showed a modest but significant difference in median survival between no treatment and the anti-PD1 groups (median survival: 24.0 days vs 30.0 days; HR 3.704; 95% CI 1.539 to 24.05; p=0.0145), while mice receiving IFN-γ treatment did not have an increase in the median survival (median survival: 24.0 days vs 26.0 days; HR 1.885; 95% CI 0.6659 to 7.179; p=0.2303). Conversely, mice treated with anti-PD1+IFN-γ experienced a significant increase in median survival (median survival: 24.0 days vs 39.0 days; HR 8.103; 95% CI 8.199 to 249.2; p<0.0001) (figure 2C).

We next explored whether the efficacy of combinatorial therapy was related to the downregulation of CXCL8 and reduction of CXCR2⁺CD68⁺ macrophages in tumor-bearing mice. We found that the percentage of circulating CXCR2⁺CD68⁺ macrophages in mice were no different between control, anti-PD1 or IFN-γ used alone and the combined treatment groups (figure 2D). The expression of CXCL8 was modestly increased in BxPC-3 tumors after anti-PD1 treatment but was significantly decreased following anti-PD1+IFN-γ treatment; however, tumor-infiltrating CXCR2⁺CD68⁺ macrophages represented a minor increased fraction in tumor tissue with PD1 blockade therapy but were largely diminished after anti-PD1 +IFN-γ treatment (figure 2E).

The CXCR2⁺CD68⁺ macrophages expanded in tumor-bearing mice were suppressive cells; thus, we further investigated whether these cells would change the immunosuppressive nature after anti-PD1+IFN-γ treatment. To this end, we sorted CXCR2⁺CD68⁺ macrophages from tumor-bearing mice and evaluated their capacity to produce arginase-1 and iNOS. We observed that increased, but not significant, levels of arginase-1 were detected in tumor-infiltrating CXCR2⁺CD68⁺ macrophages after anti-PD1 treatment, which significantly decreased following anti-PD1+IFN-γ treatment. The levels of iNOS in tumor-infiltrating CXCR2⁺CD68⁺ macrophages were no different between no treatment and anti-PD1 treated mice; however, iNOS was significantly increased in these cells from anti-PD1+IFN-γ treated tumor-bearing mice (figure 2F), suggesting these cells showed M1 polarization. We also observed higher levels of PD-L1 expression in circulating and tumor-infiltrating CXCR2⁺CD68⁺ macrophages from tumor-bearing mice after anti-PD1 treatment, whereas the expression of PD-L1 was significantly decreased in these cells following anti-PD1+IFN-γ treatment (figure 2G,H). We also observed that the fraction of tumor-infiltrating Th1 (CD4⁺IL-2⁺, CD4⁺TNF-α⁺), CD8⁺IL-2⁺ and CD8⁺TNF-α⁺ cells minor increased after anti-PD1 or IFN-γ treatment alone while significantly increased following anti-PD1+IFN-γ treatment; whereas Th2 (CD4⁺IL-10⁺) and Treg (CD4⁺FoxP3⁺) cells minor decreased after anti-PD1 or IFN-γ treatment alone while significantly reduced following anti-PD1+IFN-γ treatment in tumor bed (see online supplementary additional file 1: figures S12 and S13). Together, these results suggest that IFN-γ improves the efficacy of anti-PD1 therapy. IFN-γ could suppress the expression of tumor-derived CXCL8 to inhibit tumor-induced trafficking of CXCR2⁺CD68⁺ macrophages to the tumor bed. In addition, the peripheral and tumor-infiltrating CXCR2⁺CD68⁺ macrophages alter the immunosuppressive properties, while the T-cell subsets turn to a state of immune activation after anti-PD1 combined with IFN-γ treatment.