Impact of Surgery-Induced Myeloid-derived Suppressor Cells and the NOX2/ROS Axis on Postoperative Survival in Human Pancreatic Cancer

Abstract

Preclinical studies imply that surgery triggers inflammation that may entail tumor outgrowth and metastasis. The potential impact of surgery-induced inflammation in human pancreatic cancer is insufficiently explored. This study included 17 patients with periampullary cancer [pancreatic ductal adenocarcinoma (PDAC) n = 14, ampullary carcinoma n = 2, cholangiocarcinoma n = 1] undergoing major pancreatic cancer surgery with curative intent. We analyzed the potential impact of preoperative and postoperative immune phenotypes and function on postoperative survival with >30 months follow-up. The surgery entailed prompt expansion of monocytic myeloid-derived suppressor cells (M-MDSC) that generated NOX2-derived reactive oxygen species (ROS). Strong induction of immunosuppressive M-MDSC after surgery predicted poor postoperative survival and coincided with reduced functionality of circulating natural killer (NK) cells. The negative impact of surgery-induced M-MDSC on survival remained significant in separate analysis of patients with PDAC. M-MDSC-like cells isolated from patients after surgery significantly suppressed NK cell function ex vivo, which was reversed by inhibition of NOX2-derived ROS. High NOX2 subunit expression within resected tumors from patients with PDAC correlated with poor survival whereas high expression of markers of cytotoxic cells associated with longer survival. The surgery-induced myeloid inflammation was recapitulated in vivo in a murine model of NK cell-dependent metastasis. Surgical stress thus induced systemic accumulation of M-MDSC-like cells and promoted metastasis of NK cell-sensitive tumor cells. Genetic or pharmacologic suppression of NOX2 reduced surgery-induced inflammation and distant metastasis in this model. We propose that NOX2-derived ROS generated by surgery-induced M-MDSC may be targeted for improved outcome after pancreatic cancer surgery.

Significance: Pancreatic cancer surgery triggered pronounced accumulation of NOX2+ myeloid-derived suppressor cells that inhibited NK cell function and negatively prognosticated postoperative patient survival. We propose the targeting of M-MDSC as a conceivable strategy to reduce postoperative immunosuppression in pancreatic cancer.

FIGURE 1

Pancreatic cancer surgery triggers inflammation and expansion of M-MDSC. A, Flow chart of included IPEP study participants. B, Study design and available samples. Absolute numbers of neutrophils (C), monocytes (D), and lymphocytes (E) in patients undergoing surgery for periampullary cancer with whole blood samples (differential counts) analyzed before (pre-op) and 1, 3–5 and 28 days after pancreatic cancer surgery (day 1, days 3–5, and day 28, respectively; N = 17 for pre-op, N = 16 for day 1, N = 15 for days 3–5, N = 11 for day 28). F, Plasma levels of IL6 in healthy donors (HD) and in pre- and post-op samples from patients (N = 16 for pre-op and for day 1, N = 15 for days 3–5, N = 9 for day 28, N = 6 for HD). G–I, CD14⁺HLADR^low M-MDSC in peripheral blood of healthy donors and patients before and after surgery analyzed by flow cytometry. G, Representative FACS plot of the frequency of CD14⁺HLADR^low cells among CD14⁺ cells. Absolute numbers (H) and frequencies (I) of M-MDSC (N = 17 for pre-op, N = 16 for day 1 and days 3–5, N = 10 for day 28, N = 12 for HD). J, M-MDSC levels of ROS as measured by DCFDA MFI (N = 16 for pre-op, for day 1 and for days 3–5, N = 10 for day 28, N = 12 for HD). Preoperative and postoperative samples were compared using mixed-effects analysis followed by Šídák multiple comparison test (P-values indicated by *). HD and cancer samples were compared using Kruskal–Wallis test followed by Dunn multiple comparison test (P-values indicated by ^#). *^/#P < 0.05, ^##P < 0.01, ***^/###P < 0.001, ****^/####P < 0.0001.

FIGURE 2

Pancreatic cancer surgery entails reduced levels of NK cells with abridged expression of the activating NK cell receptor NKp30. Absolute numbers (A, D) and frequencies (B, E) of CD16⁺ (A, B) and CD56^bright (D, E) NK cells in preoperative (pre-op) and postoperative (day 1, days 3–5, and day 28) samples from patients with periampullary cancer and healthy donors (HD). Frequencies of NKp30^high-expressing cells among CD16⁺ (C) and CD56^bright (F) NK cells (N = 16 for pre-op, N = 15 for day 1, N = 14 days 3–5, and N = 10 for day 28 for A and D and N = 16 for pre-op, day 1 and days 3–5, N = 10 for day 28, N = 10 for HD for B–C and E–F). Preoperative and postoperative samples were compared using mixed-effects analysis followed by Šídák multiple comparison test (P-values indicated by *). HD and cancer samples were compared using Kruskal–Wallis test followed by Dunn multiple comparison test (P-values indicated by ^#). ^#, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

FIGURE 3

High levels of M-MDSC and low levels of NKp30^high NK cells after pancreatic cancer surgery correlate and predict poor survival. A, Forest plots for postoperative survival were generated by log-rank tests where patients were dichotomized on the basis of Youden index for high or low absolute number of M-MDSCs, CD56^bright NK cells, CD16⁺ NK cells, monocytes, lymphocytes, CD8⁺ T cells, regulatory T cells, CD4⁺ T cells, DC, neutrophils, or plasma levels of IL6, high expression level of NKp30⁺ cells among CD16⁺ (NKp30^high CD16⁺) and CD56^bright (NKp30^high CD56^bright) NK cells one day after surgery or by age or sex. Kaplan–Meier survival comparisons by high (red) or low (blue) absolute numbers of M-MDSC (B) and NKp30^high cells among CD16⁺ NK cells (C) one day after surgery by log-rank test. D, Heat map of correlations (Pearson r, above the diagonal) between immune cell populations and IL6 among patients undergoing pancreatic cancer surgery. P-values corresponding to each correlation are shown below the diagonal. Correlation between absolute numbers of M-MDSC and frequencies of NKp30^high cells among CD16⁺ NK cells at all analyzed timepoints (E), or within the first week after surgery (green dots: pre-op, black dots: 1 day, red dots 3–5 days and orange dots 28 days after surgery; F). Statistics by linear regression. Survival impact of high (red) or low (black) intratumoral NCF2 expression in pancreas cancer samples from the IPEP study (n = 14; G) or the Kaplan–Meier plotter database (n = 852; H), using the best cut-off feature of the portal and log-rank statistics. I, Expression analysis of Single Cell portal PDAC samples showed that the majority of NCF2-expressing cells were confined to a specific immune cell cluster.

FIGURE 4

M-MDSC suppress NK cells by generating ROS and show enhanced NK cell suppression after surgery. A–C, NK cells were co-cultured with or without PANC-1 cells and freshly isolated healthy donor (HD) monocytes in the presence or absence of 100 µmol/L HDC, 200 U/mL catalase (Cat), 10 µmol/L GSK2795039 (GSK) or 1% DMSO (control). NK cell activation and degranulation were measured by flow cytometry and are shown by frequencies of NK cells expressing CD69⁺CD16⁺ (A) or CD107a⁺CD16⁺ (B), respectively. C, NK cell cytotoxicity against PANC-1 cells was determined by frequency of LIVE/DEAD stained PANC-1 cells. NK cell cytotoxicity against PANC-1 cells was measured in four independent experiments, that each included two to three combinations of NK cells and monocyte from different donors. D, ROS formation was measured by CL from CD33⁺ cells isolated 3–5 days after surgery in the absence of stimulation (PDAC Ctrl, orange), in the presence of rituximab (PDAC RTX, red), in the presence of rituximab and HDC (PDAC RTX + HDC, green) and from 2 HDs stimulated with rituximab (mean ROS formation shown, HD RTX, black). NK cells were cocultured with K562 cells and CD33⁺ cells isolated from HDs or patients before surgery (pre-op) or 3–5 days after surgery (days 3–5) in the presence (HDC) or absence of HDC. NK cell activation and degranulation were measured by flow cytometry as frequencies of CD16⁺ NK cells expressing CD69 (E) or CD107a (F) in NK:K562:Myeloid cell cocultures. NK cell cytotoxicity was measured in 10 independent experiments that were normalized by setting the mean NK cell cytotoxicity against K562 in the absence of myeloid cells to 1 in each experiment. Statistics by mixed-effects analysis followed by Šídák multiple comparison test (P-values indicated by *). HD and cancer samples were compared using Kruskal–Wallis test followed by Dunn multiple comparison test (P-values indicated by ^#).*, P < 0.05; **^/##, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

FIGURE 5

NOX2 inhibition prevents surgery-induced metastasis in a murine model. A, Schematic outline of the experimental model of surgery-induced metastasis. Blood samples were collected from naïve (Ctrl) and sponge-bearing (Inf) WT and NOX2-KO mice 1 week after a surgical procedure (sponge implantation). A group of WT mice received intraperitoneal injections with histamine dihydrochloride (WT HDC) every other day starting one day before surgery and continuing until 3 days after tumor cell challenge. Frequency of CD11b⁺Ly6C⁺M-MDSC (B) and intracellular ROS levels in M-MDSC (C) were measured by flow cytometry. At 8 days postsurgery, mice were intravenously inoculated with B16F10 cells. D, Metastatic lesions in lungs were enumerated 19–21 days after tumor cell challenge. Differences in inflammatory monocytes, DCFDA levels, and tumor numbers in vivo were calculated using one-way ANOVA following Holm-Šídák multiple comparisons test. The frequency of M-MDSC and number of metastatic foci was evaluated in four independent experiments for WT mice with or without surgery-induced inflammation, and in two experiments for KO mice and HDC treated mice. DCFDA levels of M-MDSC were measured in three independent experiments for WT mice, two experiments for KO mice, and one experiment for HDC-treated mice. In each experiment, 4–5 mice per group were included. Fold change was calculated by dividing raw data with the mean of raw data in the WT mice control group in each experiment. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.