Infiltration of CD11b⁺ CD33⁺ MDSCs in PDAC tissues

To explore the mechanisms of inflammatory lymphocytes absence in PDAC tissues with PNI, we hypothesized that immune-suppressive cells in PDAC tissues such as MDSCs mediated the effects. So, we first detected the presence of MDSCs in PDAC tissues by immunofluorescence and found that CD11b⁺ CD33⁺ MDSCs were present in PDAC tissues (Fig. 2A), CP tissues and ANPTs of PDAC tissues surrounding the nerve (Fig. S2) but absent in normal pancreatic tissues far from PDAC tissues in PDAC (data not shown). The numbers of CD11b⁺ CD33⁺ MDSCs in PDAC tissues were more than those in CP or ANPTs (Fig. 2B). These results indicate that MDSCs in PDAC might be involved in PNI of PDAC by suppressing the infiltration of inflammatory lymphocytes.

Open in a separate window

Figure 2.

Analysis of MDSCs in tumor tissues and PBMC of patients with PDAC. (A) In situ immunofluorescence analysis of CD11b⁺ CD33⁺ myeloid-derived suppressor cells in PDAC tissues. Double immunofluorescence staining for CD11b (green) and CD33 (red) and merged images (yellow). Scale bar = 50 μm. (B) CD11b⁺ CD33⁺ MDSCs in various tissues including normal pancreatic tissues of PDAC (n = 30), CP tissues (n = 10), ANPT (n = 30) and PDAC tissue (n = 30) were counted and the data were shown as mean ± SD and statistically analyzed by ANOVA. (C, D) The numbers of MDSCs was examined in the PB of patients with PDAC and in healthy donors by multi-parameter flow cytometry. PBMCs were obtained from patients with PDAC (n = 36) and from healthy donors (n = 13), and then stained for MDSCs using fluorochrome-labeled antibodies against CD11b, CD14, HLA-DR, CD33 and CD15. Monocyte-like MDSCs were defined as the CD11b⁺ CD14⁺ HLA-DR^−/low population and neutrophil-like MDSCs were defined as the CD11b⁺ CD33⁺ CD14⁻ CD15⁺ population in phenotype. (C) Representative multi-parameter dot plots of monocyte-like MDSCs are shown as CD11b⁺ CD14⁺ HLA-DR^−/low population in PBMCs from healthy donors and from patients with PDAC. Numbers in plots indicate the percentages of gated populations and the percentages in bracket indicate the frequency of monocyte-like MDSCs in PBMCs. The numbers of monocyte-like MDSCs in PBMCs were calculated as “the frequency of cells in PBMCs × (the numbers of WBC cells in blood − the numbers of granulocytes in blood).” The numbers of monocyte-like MDSCs in PBMCs from healthy donors was compared with those from patients with PDAC and the data was statistically analyzed by unpaired t test. (D) Representative multi-parameter dot plots of neutrophil-like MDSCs are shown as CD11b⁺ CD33⁺ CD14⁻ CD15⁺ population in PBMCs from healthy donors and from patients with PDAC. Numbers in plots indicate the percentages of gated populations and the percentages in bracket indicate the frequency of neutrophil-like MDSCs in PBMCs. The numbers of neutrophil-like MDSCs in PBMCs were calculated as “the frequency of cells in PBMCs × (the numbers of WBC cells in blood − the numbers of granulocytes in blood).” The numbers of neutrophil-like MDSCs in PBMCs from healthy donors was compared with that from patients with PDAC and the data was statistically analyzed by unpaired t test. (E) Representative multi-parameter dot plots of neutrophil-like MDSCs (nMDSCs) are shown as CD45⁺ HLA-DR⁻ CD11b⁺ CD33⁺ CD14⁻ CD15⁺ population in purified tissue-infiltrating immune cells from CP tissues (n = 8) and PDAC tumor tissues (n = 10). Numbers in plots indicate the percentages of gated populations and the percentages in bracket indicate the frequency of neutrophil-like MDSCs in tissue-infiltrating immune cells. (F) nMDSCs in tissue-infiltrating immune cells from normal pancreatic tissues of PDAC (n = 10), CP tissues (n = 8) and PDAC tumor tissue (n = 10) were calculated as “(the frequency of cells in tissue-infiltrating immune cells × the numbers of isolated tissue-infiltrating immune cells)/the weight of tissue for cell isolation” and the data were shown as mean ± SEM and statistically analyzed by ANOVA.

Increased numbers of neutrophil-like MDSCs in patients with PDAC

MDSCs are one of several important immuno-suppressive cells during cancer development, yet the details of MDSC subsets are unknown. Human MDSCs are heterogeneous in phenotype and usually include monocyte-like MDSCs (mMDSCs) and nMDSCs.^1,21 We investigated the numbers of different MDSC subsets in PBMCs from patients with PDAC by multi-parameter flow cytometry analysis and compared it to healthy donors. Representative flow cytometry dot plots demonstrated the presence of CD11b⁺ CD14⁺ HLA-DR⁻ mMDSCs in PBMCs of both patients with PDAC and healthy donors (Fig. 2C). Collective data confirmed that the numbers of mMDSCs in PBMCs of patients with PDAC (n = 36, mean = 6.10 × 10⁶/L, SEM = 0.51 × 10⁶/L, p = 0.0844) was comparable to that in PBMCs of healthy donors (n = 13, mean = 7.25 × 10⁶/L, SEM = 0.46 × 10⁶/L) (Fig. 2C). Furthermore, representative flow cytometry dot plots showed the existence of a multitude of CD11b⁺ CD33⁺ CD14⁻ CD15⁺ nMDSCs in PBMCs of the patient with PDAC (Fig. 2D). The numbers of nMDSCs in PBMCs of patients with PDAC (n = 36, mean = 61.29 × 10⁶/L, SEM = 11.37 × 10⁶/L, p = 0.0068) were increased significantly compared with those in healthy donors (n = 13, mean = 12.20 × 10⁶/L, SEM = 3.00 × 10⁶/L) (Fig. 2D). More importantly, we found the existence of CD45⁺ HLA-DR⁻ CD11b⁺ CD33⁺ CD14⁻ CD15⁺ nMDSCs by multi-parameter flow cytometry analysis in infiltrating immune cells of PDAC tissues (Fig. 2E) and there were no infiltrating immune cells found in normal pancreatic tissues far from PDAC tissues (data not shown). Collective data indicated that their numbers in PDAC tissues (n = 10, mean = 18.62 × 10⁴/g, SEM = 3.27 × 10⁴/g, p = 0.0131) were higher than that in CP tissues (n = 8, mean = 9.41 × 10⁴/g, SEM = 0.87 × 10⁴/g) (Fig. 2F). These data collectively demonstrate that the numbers of nMDSCs but not mMDSCs are significantly higher in patients with PDAC, suggesting a distinct profile of MDSC subsets during PDAC development.

Phenotype and identification of CD13^hi or CD13^low neutrophil-like MDSC subsets

Considering the existence of abundant nMDSCs in PBMCs and tumor tissues of patients with PDAC, we further analyzed the phenotype of these MDSCs by multi-parameter flow cytometry and first focused on the expression of CD13, a neutrophil-related marker. Representative flow cytometry dot plots clearly demonstrated two distinct subpopulations of CD11b⁺ CD33⁺ CD14⁻ CD15⁺ nMDSCs in patients with PDAC, CD13^hi and CD13^low cells (Fig. 3A), which have not been reported before. We proposed that CD13^hi and CD13^low nMDSCs might be two different novel MDSC subsets during cancer development. Monocytes defined as CD11b^hi CD14^hi CD15⁻ cells served as control cells in the experiment (Fig. 3B). By flow cytometry analysis, both CD13^hi nMDSCs and CD13^low nMDSCs exhibited neutrophil-like characteristics in FSC/SSC, different from CD11b^hi CD14^hi CD15⁻ monocytes (Fig. 3C). CD13^hi nMDSCs expressed higher levels of CD11b, CD33 and CD16 and lower levels of CD66b than CD13^low nMDSCs, although they both expressed low levels of HLA-DR, CD14 and CD124 (Fig. 3C). Collective data demonstrated that the numbers of CD13^hi nMDSCs in PBMCs of patients with PDAC were significantly more than those in healthy donors (PDAC: n = 36, mean = 53.42 × 10⁶/L, SEM = 10.29 × 10⁶/L; Healthy donors: n = 13, mean = 10.02 × 10⁶/L, SEM = 3.05 × 10⁶/L, p = 0.0079) (Fig. 3D) and there were the increasing trend numbers of CD13^low nMDSCs in PBMCs of PDAC patients compare with healthy donor (PDAC: n = 36, mean = 13.28 × 10⁶/L, SEM = 2.94 × 10⁶/L; Healthy donors: n = 13, mean = 6.19 × 10⁶/L, SEM = 1.92 × 10⁶/L, p = 0.0834) (Fig. 3E). Overall the numbers of CD13^hi nMDSCs in PBMCs of patients with PDAC were higher than those of CD13^low nMDSCs (Figs. 3D and E). Consistently, both CD45⁺ HLA-DR⁻ CD11b⁺ CD33⁺ CD14⁻ CD15⁺ CD13^hi and CD13^low nMDSC subsets were detected in PDAC tumor tissues (Fig. 4A). The numbers of CD13^hi nMDSC subsets in PDAC tumor tissues (n = 10, mean = 16.77 × 10⁴/g, SEM = 3.08 × 10⁴/g, p = 0.037) were more than those in CP tissues (n = 8, mean = 8.08 × 10⁴/g, SEM = 0.92 × 10⁴/g) (Fig. 4B). Thus, we identified two different nMDSC subsets according to the expression levels of CD13 in PDAC. To demonstrate whether these two nMDSCs subsets can differentiate toward the other, isolated PBMC cells from PDAC patients were stimulated with inflammatory cytokines or not and flow cytometry analysis demonstrated that the percentages of these two populations were similar irrespective of whether TNF-α or IL-1α was used (Fig. 3), indicating that CD13^hi nMDSC subsets did not lose CD13 or CD13^low nMDSC subsets did not acquire CD13 even in response to inflammatory stimuli such as TNF-α or IL-1α.

Open in a separate window

Figure 3.

Phenotype and identification of neutrophil-like MDSC subsets in PBMC of patients with PDAC. (A) Phenotype of neutrophil-like MDSCs (nMDSCs) in PBMC of patients with PDAC was analyzed by multi-parameter flow cytometry. PBMCs were obtained from patients with PDAC, and then stained using fluorochrome-labeled antibodies against CD11b, CD14, HLA-DR, CD33, CD15, CD13 (one of neutrophil markers) and CD16. Representative multi-parameter dot plots of CD13 expression of nMDSCs from patients with PDAC are shown. nMDSCs can be divided into two subsets including CD13^hi and CD13^low nMDSCs according to CD13 expression. Numbers in plots indicate the percentages of gated populations. (B) Monocytes in PBMC of patients with PDAC were defined as CD14^hi CD11b^hi CD15⁻ population. Representative dot plots of monocytes in PBMC of patients with PDAC are shown. Numbers in plots indicate the percentages of gated populations. (C) Three populations including monocytes, CD13^hi nMDSCs and CD13^low nMDSCs in PBMC of patients with PDAC are compared in phenotype. Representative histograms of several markers expression including FSC/SSC, CD11b, CD33, CD16, HLA-DR, CD13, CD66b, CD124, ROS and CD14 are shown. Numbers in the right of histograms indicate gMFI of several markers expression. (D and E) The numbers of nMDSC subsets in PBMCs were calculated as “the frequency of cells in PBMCs × (the numbers of WBC cells in blood − the numbers of granulocytes in blood).” The numbers of CD13^hi nMDSCs (D) or CD13^low nMDSCs (E) in PBMCs from healthy donors (n = 13) were compared with those from patients with PDAC (n = 36) and the data was statistically analyzed by unpaired t test.

Open in a separate window

Figure 4.

Identification of neutrophil-like MDSC subsets in tumor tissues of patients with PDAC. (A) Representative multi-parameter dot plots of CD45⁺ HLA-DR⁻ CD11b⁺ CD33⁺ CD14⁻ CD15⁺ CD13^hi nMDSCs or CD13^low nMDSCs are shown in purified tissue-infiltrating immune cells from CP tissues (n = 8) and PDAC tumor tissues (n = 10). Numbers in plots indicate the percentages of gated populations and the percentages in bracket indicate the frequency of nMDSC subsets in tissue-infiltrating cells. (B) nMDSC subsets in tissue-infiltrating immune cells from normal pancreatic tissues of PDAC (n = 10), CP tissues (n = 8) and PDAC tumor tissue (n = 10) were calculated as “(the frequency of cells in tissue-infiltrating immune cells × the numbers of isolated tissue-infiltrating immune cells)/the weight of tissue for cell isolation” and the data were shown as mean ± SEM and statistically analyzed by ANOVA.

Properties of CD13^hi and CD13^low nMDSCs

To investigate the properties and functions of the two novel subsets of nMDSCs, we sorted CD13^hi and CD13^low nMDSCs for further analysis, with CD11b^hi CD14^hi CD15⁻ monocytes served as control cells or antigen-presenting cells. By morphology, both CD13^hi nMDSCs and CD13^low nMDSCs showed the characteristics of neutrophils and were polymorphonuclear (Fig. 5A). It is well known that MDSCs usually express high levels of Arg 1, iNOS or ROS. Western blot analysis showed that both subsets of MDSCs in patients with PDAC expressed Arg 1 but not iNOS, and expression of Arg 1 in CD13^hi nMDSCs was higher than CD13^low nMDSCs, while monocytes did not express Arg 1 or iNOS (Fig. 5B, Fig. 5C and data not shown). Flow cytometry analysis showed that both subsets of MDSCs in patients with PDAC expressed similar levels of ROS (Fig. 3C). In terms of cytokines secretion, CD13^hi nMDSCs produced more IL-6 and IL-1β than CD13^low nMDSCs after LPS stimulation and both of them produced very low levels of TNF-α, IL-10, IL-12, IL-2, IL-4 and IFNγ irrespective of LPS stimulation or not (Fig. 5D and data not shown). As control cells, monocytes produced high levels of TNF-α, IL-6, IL-10 and IL-1β after LPS stimulation (Fig. 5D). Overall the major difference between CD13^hi nMDSCs and CD13^low nMDSCs was the level of Arg 1 expression.

Open in a separate window

Figure 5.

Properties of two neutrophil-like MDSC subsets in PBMC of patients with PDAC. PBMCs were obtained from patients with PDAC, and stained using fluorochrome-labeled antibodies against CD11b, CD14, CD33, CD15 and CD13. Then, three populations including CD11b^hi CD14^hi CD15⁻ monocytes (Mono), CD11b⁺ CD33⁺ CD14⁻ CD15⁺ CD13^hi nMDSCs (Hi) and CD11b⁺ CD33⁺ CD14⁻ CD15⁺ CD13^low nMDSCs (Low) in PBMC of patients with PDAC are sorted by Moflo. (A) Representative cytospin images of sorted monocytes and nMDSCs with Wright-Giemsa staining are shown (Scale bars = 50 μm). (B and C) Expression of Arg 1 and β-actin are detected for by Western blot in sorted cells. Representative images are shown (B) and expression levels of Arg 1 protein are normalized by β-actin (C). Data are shown as a representative of three independent experiments. (D) Sorted cells are stimulated with or without 500 ng/mL LPS for 24 h, and then supernatants are collected and assayed for cytokine production by CBA assays. Data are shown as mean ± SD of triplicate cultures of one representative experiment. Similar results are obtained in three independent experiments. **p < 0.01 vs. respective “Ctrl” group; #p < 0.01 between “Hi” group and “Low” group after LPS stimulation.

Immuno-suppressive activity of CD13^hi and CD13^low nMDSCs

The most important function of MDSCs is to suppress antigen-presenting cells induced T cell responses. We established monocytes-initiated alloreactive T cell proliferation by assaying the dilution of CFSE expression in CD4⁺ or CD8⁺ T cells and investigated the effects of CD13^hi and CD13^low nMDSCs on alloreactive CD4⁺ or CD8⁺ T cell proliferation. We found that both CD13^hi nMDSCs and CD13^low nMDSCs had a very weak ability to initiate alloreactive T cell proliferation, but could suppress monocytes-initiated alloreactive CD4⁺ or CD8⁺ T cell proliferation (Figs. 6A–C). The inhibition of monocytes-initiated alloreactive T cell proliferation by CD13^hi nMDSCs was much stronger than that by non-specific nMDSCs or CD13^low nMDSCs (Figs. 6A–C). Arg 1 specific inhibitor was able to abolish the suppression of monocytes-initiated alloreactive T cell proliferation by CD13^hi nMDSCs (Figs. 6A–C), suggesting the strong immuno-suppressive functions is mediated by Arg-1-related mechanism.

Open in a separate window

Figure 6.

Immunosuppressive function of neutrophil-like MDSC subsets in PBMC of patients with PDAC. Monocytes, non-specific nMDSCs (MDSCs), CD13^hi nMDSCs and CD13^low nMDSCs in PBMCs from patients with PDAC are sorted. CD3⁺ T cells in PBMCs from healthy donor are purified and stained with CFSE. Then, CD3⁺ T cells are co-cultured with monocytes and/or nMDSCs at the ratio of 10:1 (T cells: monocytes or nMDSCs). In some groups, Arg-1-specific inhibitor nor-NOHA is used. After 7 d, cells are collected and stained with fluorochrome-labeled antibodies against CD4⁺ and CD8⁺ for flow cytometry. Representative histograms of CFSE dilution in CD4⁺ T cells (A) or CD8⁺ T cells (B) in various groups are shown. Numbers in histograms indicate the percentages of CFSE^low cells in CD4⁺ or CD8⁺ T cells. (C) The results in triplicate cultures in each group were expressed as proliferating CFSE^low T cells (% Mono/T group) and the value in Mono/T group was set as 100%. The data are shown as mean ± SD in one representative experiment and analyzed by ANOVA. Similar results are obtained in three independent experiments. **p < 0.01.

Reagents

The antibodies for flow cytometry, including fluorochrome-labeled antibodies against human CD15 (HI98), CD16 (3G8), CD45 (HI30), CD13 (WM15), CD33 (WM53), CD11b (ICRF44), CD14 (HCD14), HLA-DR (L243), CD66b (G10F5), CD124 (G077F6), CD4⁺ (OKT4) and CD8⁺ (HIT8a) were from Biolegend (San Diego, CA). Antibodies against human Arginase 1 (sc-166920) and NOS 2 (sc-7271) were from Santa Cruz. CFSE were from Invitrogen (Eugene, OR). DCFDA cellular ROS detection assay kit was from Abcam (Cambridge, UK). CBA reagents of TNF-α, IL-6, IL-1β, IL-10, IL-12, IL-2, IL-4 and IFNγ were from BD Bioscience. Microbead-conjugated mAbs to human CD3 were from MiltenyiBiotec (BergischGladbach, Germany). Arginase-1-specific inhibitor nor-NOHA was from EMD Millipore (Billerica, MA).

Immunohistochemical staining

Formaldehyde-fixed tissues were made into 4-μm sections and were mounted onto poly-L-lysine-coated slides. The slides were incubated with mouse anti-human CD3 antibody or anti-human CD20 antibody (1:30, DAKO, Glostrup, Denmark) in a humid incubator at 4°C overnight. The secondary antibody system (PV9000, Golden Bridge International, Beijing, China) was applied according to the manufacturer's instructions. A positive staining result was recorded when the membrane and/or cytoplasm of cells was stained yellow or brown. Three high-power fields (HPF) per slide were randomly chosen to measure the proportion of immunopositive cells in perineural cuffs.

Immunofluorescence double staining

Paraffin-embedded PDAC tissues were made into 4-μm sections. After de-paraffin, sections were subjected to antigen retrieval for 20 min, and then blocked by goat serum for 20 min. For co-localization of CD11b/CD33, rabbit anti-human CD11b or mouse anti-human CD33 (1:50, Jackson, West Grove, PA, USA) was added and maintained at 4°C overnight. After washing with PBS, FITC or rhodamine-conjugated anti-rabbit or mouse secondary antibodies (1:50, Abcam, Hong Kong) was added and incubated at room temperature for 30 min. Fluorescence was observed after drying by a microscope (Leica DM 2000, Leica Microsystems, Germany) and the photographs were acquired by software LAS V3.7 (Leica Microsystems, Germany). Five random HPFs per slide were chosen for cell counting and the average of CD11b⁺ CD33⁺ cells per slide was calculated.

Isolation of human peripheral blood mononuclear cells (PBMCs)

PBMCs were isolated from the peripheral blood of patients with PDAC and healthy donors by density gradient centrifugation at 800 × g for 30 min using Ficoll-Paque Plus (17–1440–03, GE Healthcare, Piscataway, NJ, USA). The buoyant layer was recovered and washed in RPMI-1640 medium plus 10% fetal bovine serum and antibiotics.

Isolation of tissue-infiltrating immune cells in PDAC and CP

Tumor tissues and normal pancreatic tissues from PDAC patients and CP tissues were rinsed with PBS twice and then minced into 1~3 mm³ pieces. The minced tissues were re-suspended in enzymatic cocktail containing 1 mg/mL collagenase (Sigma-Aldrich), 0.1 mg/mL hyaluronidase (Sigma-Aldrich) and 0.1 mg/mL DNase (Sigma-Aldrich) and incubated in 37°C for 1.5 h. The cell suspension was passed through a Falcon 70 μm cell strainer (BD Biosciences) to remove any large aggregates and debris and was then subjected to centrifugation at 300 g for 5 min. The cell pellets were re-suspended in 36% Percoll solution, added onto 63% Percoll solution and centrifuged at 1,000 g for 15 min. The cells in the middle layer were collected and washed twice with PBS. The purified cells were tissue-infiltrating immune cells for cell counting or flow cytometry analysis.

Flow cytometry and cell sorting

The phenotypic characteristics of isolated PBMCs were examined by flow cytometry. PBMCs were pretreated with 50 μg/mL rat immunoglobulin G (IgG) whole molecules (Pierce, Rockford, IL, USA) and then incubated with the following antibodies in PBS, i.e., APC-conjugated anti-CD13, FITC-conjugated anti-CD15, PE-conjugated anti-CD16, PE-Cy7-conjugated anti-CD11b, Percp-Cy5.5-conjugated anti-CD33, APC-Cy7- conjugated anti-CD14 and Pacific blue-conjugated anti-HLA-DR. After incubation at 4°C for 30 min, cells were washed once with PBS and cell phenotypes were analyzed by flow cytometry. Cytometric data were collected on an LSR II cytometer (Becton Dickinson, Franklin Lakes, NJ, USA) and analyzed using FlowJo Version 5.7.2 software (TreeStar).

To purify monocytes or nMDSC subsets from PBMCs of patients with PDAC, isolated PBMCs were costained with APC-conjugated anti-CD13, FITC-conjugated anti-CD15, PE-Cy7-conjugated anti-CD11b, Percp-Cy5.5-conjugated anti-CD33 and PE-conjugated anti-CD14, and then CD11b^hi CD14^hi CD15⁻ monocytes, CD11b⁺ CD33⁺ CD14⁻ CD15⁺ MDSCs, CD11b⁺ CD33⁺ CD14⁻ CD15⁺ CD13^hi nMDSCs, or CD11b⁺ CD33⁺ CD14⁻ CD15⁺ CD13^low nMDSCs were sorted by a Moflo high-speed cell sortor (Dako). The purity was confirmed by flow cytometry to be >97%.

Wright-Giemsa staining

Sorted cells were fixed in 2% formaldehyde for 10 min and then washed twice with PBS. Re-suspended cells were smeared on slides by cytospin centrifugation. After being dried, the samples were stained with Wright-Giemsa dye solution (Nanjing Jiancheng Bioengineering Institute, Nanjing, China) for appropriate times to observe the cellular nuclear characteristics. The size and shape of cells were observed by a microscope (Leica DM 2000, Leica Microsystems, Germany) and the images were acquired by software LAS V3.7 (Leica Microsystems, Germany).

Western blotting

Cells were washed twice with cold PBS and lysed with M-PER Protein Extraction Reagent (Pierce, Rockford, IL) supplemented with protease inhibitor cocktail, and protein concentrations of the extracts were measured by BCA assay (Pierce). Proteins were separated on 10% SDS-PAGE gels, and transferred onto nitrocellulose membranes. The membranes were blocked in 5% non-fat milk and reacted with primary antibodies for Arginase 1 (sc-166920, Santa Cruz, CA, USA), NOS 2 (sc-7271, Santa Cruz, CA, USA) and β-actin (sc-1616, Santa Cruz, CA, USA). Then, HRP-conjugated secondary antibodies were added to the membranes. Protein expressions were detected by enhanced chemiluminescence (ECL) Western blotting detection reagent (Pierce, Rockford, IL).

Cytokine assay

Sorted monocytes or nMDSCs (2 × 10⁵ cells per mL) were cultured in RPMI-1640 medium with 10% (vol/vol) FCS in the presence or absence of 500 ng/mL LPS. After 24 h, the contents of TNF-α, IL-6, IL-1β, IL-10, IL-12, IL-2, IL-4 and IFNγ in cell supernatants were measured by cytometric bead array immunoassay (CBA; BD Biosciences) according to the manufacturer's protocol.

Assay of alloreactive T cell response

To analyze the ability of nMDSC subsets to initiate alloreactive T cell responses or suppress monocytes-induced alloreactive T cell responses, CD3 T cells from healthy donor PBMCs were purified using anti-human CD3 magnetic microbead and labeled with CFSE (Invitrogen). Then, CD3 T cells (1 × 10⁵/well) were co-cultured with sorted monocytes (1 × 10⁴/well) and/or non-specific nMDSCs, CD13^hi or CD13^low nMDSCs (1 × 10⁴/well) at the ratio of 10: 1 (T cells: monocytes or nMDSCs) in 96-well round-bottom plates in triplicate cells. In some groups, Arg-1-specific inhibitor nor-NOHA was used. After 7 d, cells were stained with anti-CD4-Percy-Cy5.5 and anti-CD8-PE and assayed by flow cytometry for CFSE dilution in CD4⁺ T cells and CD8⁺ T cells. The degree of T cell proliferation was determined by the percentages of CFSE^low cells in gated CD4⁺ or CD8⁺ T cells. The results were expressed as proliferating CFSE^low T cells (% Mono/T group) and the value in Mono/T group was set as 100%. In each experiment, blood sample from one PDAC patient was collected for purification of MDSC subsets or monocytes. And these experiments were repeated twice.