Targeting Glycolysis in Macrophages Confers Protection Against Pancreatic Ductal Adenocarcinoma

Abstract

Inflammation in the tumor microenvironment has been shown to promote disease progression in pancreatic ductal adenocarcinoma (PDAC); however, the role of macrophage metabolism in promoting inflammation is unclear. Using an orthotopic mouse model of PDAC, we demonstrate that macrophages from tumor-bearing mice exhibit elevated glycolysis. Macrophage-specific deletion of Glucose Transporter 1 (GLUT1) significantly reduced tumor burden, which was accompanied by increased Natural Killer and CD8+ T cell activity and suppression of the NLRP3-IL1β inflammasome axis. Administration of mice with a GLUT1-specific inhibitor reduced tumor burden, comparable with gemcitabine, the current standard-of-care. In addition, we observe that intra-tumoral macrophages from human PDAC patients exhibit a pronounced glycolytic signature, which reliably predicts poor survival. Our data support a key role for macrophage metabolism in tumor immunity, which could be exploited to improve patient outcomes.

Keywords: glycolysis; immunometabolism; macrophage; pancreatic ductal adenocarcinoma.

Figure 1

Macrophages from tumor-bearing mice are highly glycolytic. (a) Pancreata from Day 28 tumor-bearing orthotopically-transplanted (OT) mice were assessed for their myeloid cell populations by flow cytometry. Scatter plot of macrophages (CD45+ Lin- MHCIIlow-int CD24+ CD11b+ Ly6G- Ly6Clow F4/80+) that infiltrate the pancreas/tumor, as a percentage of total CD45+ cells, from OT (black) and age-matched sham controls (grey) (sham, n = 10; OT, n = 8, three independent experiments pooled). (b) OT mice were fed either KI20227 250 ppm (red) or standard rodent chow (black) beginning Day 1 of tumor cell transplantation. Representative scatter plot of the total flux on Day 26 as calculated from the BLI image shown (KI20227, n = 9; control, n = 10, two independent experiments). BLI images for all figures were acquired at 15-s exposure, and color scale set with lower limit at 3 × 10⁶ and upper limit at 5 × 10⁷ photons/sec. (c) Kaplan–Meier survival analysis of KI20227-fed mice (n = 9) compared with standard chow-fed mice (n = 10). (d) Sorted peritoneal (PEC) macrophages from OT and sham control mice were rested at 37 °C for 1 h and assessed for their glycolytic capacity in a live metabolic flux assay. Shown is a representative Seahorse trace of OT-PEC macrophages (black) compared with sham (grey) (d), and a bar graph of glycolytic capacity as calculated from the Seahorse trace (n = 3 mice pooled per group, five independent experiments). (e) PEC macrophages from OT and sham controls were assessed at the transcript level for GLUT1, HK1, HK2, GPI, PFKB1, ALDOA, PGK, PKM2, LDHA, and HIF1α (n = 3 mice pooled per group, three independent experiments). Data are means ± SEM, * p < 0.05; ** p < 0.01; *** p < 0.001 by unpaired student’s t-test with 95% confidence interval.

Figure 2

Macrophage-specific deletion of GLUT1 confers resistance to tumor growth. (a) Sorted PEC macrophages from control (LysM-cre and/or GLUT1fl/fl alone) and GLUT1ΔmΦ mice were rested at 37 °C for 1 h and assessed for their glycolytic capacity in a live metabolic flux assay. Shown is a representative Seahorse trace of control (black) compared with GLUT1ΔmΦ (blue) (n = 3 mice pooled per group, three independent experiments). (b) Ratio of area under mass peak of each metabolite in the glycolysis pathway derived from metabolomics profiling of control (black) and GLUT1ΔmΦ (blue) PEC macrophages (control, n = 3; GLUT1ΔmΦ, n = 4, n = 3 mice pooled per sample). (c) Orthotopic tumors from control and GLUT1ΔmΦ mice were monitored in vivo over time using BLI. Shown is a representative scatter plot of the total flux on Day 28 as calculated from the BLI images (control, n = 17; GLUT1ΔmΦ, n = 25, pooled from four independent experiments). (d) Shown is a Kaplan–Meier survival analysis of GLUT1ΔmΦ mice (n = 21) compared with controls (n = 18), pooled from two independent experiments. (Data are means ± SEM, * p < 0.05; *** p < 0.001 by unpaired student’s t-test with 95% confidence interval.

Figure 3

The pro-inflammatory NLRP3-IL1β axis is suppressed in GLUT1ΔmΦ macrophages. (a–c) Intracellular flow cytometry (ICFC) of pro-inflammatory cytokines was performed on control (black) and GLUT1ΔmΦ (blue) pancreata cell suspensions restimulated with 100 ng/mL LPS for 4 h. Shown are scatter plots of the % pro-IL1β+ (a), % IL6+ (b), % TNFα+ (c) of total macrophages (control, n = 16; GLUT1ΔmΦ, n = 16, pooled from three independent experiments). Representative flow plots are shown in Supplementary Figure S3b. (d) Immunofluorescence staining of ASC specks was performed in FACS-sorted control (black) and GLUT1ΔmΦ (blue) PEC macrophages stimulated with 1 μg/mL of LPS for 3 h, and 10 μM of nigericin for a further 30 min (+stim, bottom row) or without (-stim, top row). Shown are representative confocal images with nuclei (blue) and ASC (green) overlaid. Each ASC speck is identified as a dot-like perinuclear stain as shown by the representative white arrow in the bottom left image. Magnification bar = 10 μM. Scatter plots shown are the quantification of % ASC speck+ cells for every field of view in unstimulated (open circles) and stimulated cells (filled circles) (n = 3 mice pooled per group, two independent experiments). (e) FACS-sorted control (black) and GLUT1ΔmΦ (blue) PEC macrophages were stained with Caspase-1-FAM-FLICA to assess caspase-1 activity by flow cytometry. Shown is a representative bar graph of the MFI of caspase-1 (n = 3 mice pooled per group, two independent experiments). (f) Representative immunoblot of sorted control and GLUT1ΔmΦ pancreatic tumor macrophage cell lysates probed for pro-IL1β and pro-caspase 1 (n = 6–8 mice pooled per group, two independent experiments). Data are means ± SEM, * p < 0.05; ** p < 0.01; *** p < 0.001; n.s. not significant, by unpaired student’s t-test with 95% confidence interval.

Figure 4

NK cells and CTLs mediate anti-tumor immunity in GLUT1ΔmΦ mice. (a–g, i–l) Flow cytometry was performed in. control and GLUT1ΔmΦ mice to assess NK and CTL populations. Flow cytometry gating strategy shown in Supplementary Figure S4b. Shown are scatter plots of the % CD4+ T cells (a), % CD8+ T cells (b), and % NK cells (c) of the total CD45+ population in the pancreata of control (black) and GLUT1ΔmΦ (blue) mice (n =14 mice per group, pooled from four independent experiments). Shown are scatter plots of % IFNγ+ (d), % Perforin+ (e), % Granzyme B+ (f) (n = 16 mice per group, pooled from three independent experiments, representative flow plots are shown in Supplementary Figure S4c) and % CD69+ (g) (n = 6 mice per group) of total NK cells in the pancreata of control (black) and GLUT1ΔmΦ (blue) mice. (h) NK cells were depleted in vivo by injecting 250 μg of either isotype (open bars) or anti-NK1.1 (filled bars) antibody i.p. per mouse, once a week. Shown are bar graphs of the tumor weight measured at Day 28 of control and GLUT1ΔmΦ mice (n = 5 mice per group, representative of two independent experiments). (i–l) Shown are scatter plots of % IFNγ+ (i), % Perforin+ (j), % Granzyme B+ (k) (n = 16 mice per group, pooled from three independent experiments) and % CD69+ (l) (n = 6 mice per group_of total CD8+ T cells in the pancreata of control (black) and GLUT1ΔmΦ (blue) mice. (m) CD8+ T cells are depleted in vivo by injecting 500μg of either isotype (open bars) or anti-CD8α (filled bars) antibody i.p. per mouse, once a week. Shown are bar graphs of the tumor weight measured at Day 27 of control and GLUT1ΔmΦ mice (at least n = 5 mice per group). Data are means ± SEM, * p < 0.05; ** p < 0.01; *** p < 0.001 by unpaired student’s t-test with 95% confidence interval. (n) Scatter plot of % IFNγ+ CD8+ OTI cells after co-culture with either unpulsed or OVA-pulsed macrophages isolated from the pancreata of control (black) and GLUT1ΔmΦ (blue) mice (n ≥ 8 mice per group). Data are means ± SEM, *** p < 0.001 by one-way ANOVA with Bonferroni post-hoc test.

Figure 4

NK cells and CTLs mediate anti-tumor immunity in GLUT1ΔmΦ mice. (a–g, i–l) Flow cytometry was performed in. control and GLUT1ΔmΦ mice to assess NK and CTL populations. Flow cytometry gating strategy shown in Supplementary Figure S4b. Shown are scatter plots of the % CD4+ T cells (a), % CD8+ T cells (b), and % NK cells (c) of the total CD45+ population in the pancreata of control (black) and GLUT1ΔmΦ (blue) mice (n =14 mice per group, pooled from four independent experiments). Shown are scatter plots of % IFNγ+ (d), % Perforin+ (e), % Granzyme B+ (f) (n = 16 mice per group, pooled from three independent experiments, representative flow plots are shown in Supplementary Figure S4c) and % CD69+ (g) (n = 6 mice per group) of total NK cells in the pancreata of control (black) and GLUT1ΔmΦ (blue) mice. (h) NK cells were depleted in vivo by injecting 250 μg of either isotype (open bars) or anti-NK1.1 (filled bars) antibody i.p. per mouse, once a week. Shown are bar graphs of the tumor weight measured at Day 28 of control and GLUT1ΔmΦ mice (n = 5 mice per group, representative of two independent experiments). (i–l) Shown are scatter plots of % IFNγ+ (i), % Perforin+ (j), % Granzyme B+ (k) (n = 16 mice per group, pooled from three independent experiments) and % CD69+ (l) (n = 6 mice per group_of total CD8+ T cells in the pancreata of control (black) and GLUT1ΔmΦ (blue) mice. (m) CD8+ T cells are depleted in vivo by injecting 500μg of either isotype (open bars) or anti-CD8α (filled bars) antibody i.p. per mouse, once a week. Shown are bar graphs of the tumor weight measured at Day 27 of control and GLUT1ΔmΦ mice (at least n = 5 mice per group). Data are means ± SEM, * p < 0.05; ** p < 0.01; *** p < 0.001 by unpaired student’s t-test with 95% confidence interval. (n) Scatter plot of % IFNγ+ CD8+ OTI cells after co-culture with either unpulsed or OVA-pulsed macrophages isolated from the pancreata of control (black) and GLUT1ΔmΦ (blue) mice (n ≥ 8 mice per group). Data are means ± SEM, *** p < 0.001 by one-way ANOVA with Bonferroni post-hoc test.

Figure 5

The GLUT1 inhibitor WZB117 attenuates tumor burden in vivo. WZB117 (250 ppm) was fed to mice beginning Day 1 of orthotopic transplantation of tumor cells. Disease progression was monitored in vivo, and immune parameters were assessed in the pancreatic tumors. (a) Scatter plot of the total flux on Day 26 of mice fed with standard (“Std”) rodent chow (black) (n = 9) or WZB117-incorporated chow (magenta) (n = 10) as calculated from BLI images. (b) The same mice from (a) were euthanized and shown is a scatter plot of their tumors weighed at Day 34. Tumor images are shown on the right hand side. Scale bar = 1 cm. (c–g) Shown are scatter plots of the % NK of total CD45+ cells (c), % IFNγ+, % Perforin+, and % Granzyme B+ of NK cells (d), % CD8+ T of total CD45+ cells (e), % IFNγ+, % Perforin+, and % Granzyme B+ of CD8+ T cells (f), % pro-IL1β+, % IL6+, % IL12p40+, and % TNFα+ (g) of total macrophages. Data representative of two independent experiments, with n = 9–10 mice per group. (h) WZB117 was compared head to-head with standard-of-care gemcitabine. Gemcitabine was injected 50 mg/kg i.p., twice a week, beginning the first week. Shown is a scatter plot of tumors from mice fed standard rodent chow (“untreated”) (black), WZB117-incorporated chow (magenta), or treated with gemcitabine (brown) weighed at Day 28. Data representative of two independent experiments, n = 10 mice per group. Data are means ± SEM, * p < 0.05; ** p < 0.01; *** p < 0.001 by unpaired student’s t-test with 95% confidence interval.

Figure 5

The GLUT1 inhibitor WZB117 attenuates tumor burden in vivo. WZB117 (250 ppm) was fed to mice beginning Day 1 of orthotopic transplantation of tumor cells. Disease progression was monitored in vivo, and immune parameters were assessed in the pancreatic tumors. (a) Scatter plot of the total flux on Day 26 of mice fed with standard (“Std”) rodent chow (black) (n = 9) or WZB117-incorporated chow (magenta) (n = 10) as calculated from BLI images. (b) The same mice from (a) were euthanized and shown is a scatter plot of their tumors weighed at Day 34. Tumor images are shown on the right hand side. Scale bar = 1 cm. (c–g) Shown are scatter plots of the % NK of total CD45+ cells (c), % IFNγ+, % Perforin+, and % Granzyme B+ of NK cells (d), % CD8+ T of total CD45+ cells (e), % IFNγ+, % Perforin+, and % Granzyme B+ of CD8+ T cells (f), % pro-IL1β+, % IL6+, % IL12p40+, and % TNFα+ (g) of total macrophages. Data representative of two independent experiments, with n = 9–10 mice per group. (h) WZB117 was compared head to-head with standard-of-care gemcitabine. Gemcitabine was injected 50 mg/kg i.p., twice a week, beginning the first week. Shown is a scatter plot of tumors from mice fed standard rodent chow (“untreated”) (black), WZB117-incorporated chow (magenta), or treated with gemcitabine (brown) weighed at Day 28. Data representative of two independent experiments, n = 10 mice per group. Data are means ± SEM, * p < 0.05; ** p < 0.01; *** p < 0.001 by unpaired student’s t-test with 95% confidence interval.

Figure 6

The macrophage glycolytic signature is a predictor of poor survival in human PDAC. (a–c) Gene expression and clinical survival data was extracted from the TCGA cohort, a publicly available dataset. Expression of GLUT1 (a), HIF1α (b), and HK2 (c) were thresholded at various percentiles and the binary states (patients above the cut-off percentile was defined as “high”, and those below defined as “low”) for each marker were used in a log-rank survival analysis. Shown are Kaplan-Meier survival curves of GLUT1 (a), HIF1α (b), and HK2 (c) for which the percentile cut-offs yielded significant difference between the high and low patient populations. (d) Schematic of the sequential, multiplex, immunofluorescence staining protocol of myeloid markers (CD68 and CD163) and metabolic markers (GLUT1, HIF1α, and HK2) used to stain FFPE sections from multi-center patient cohorts. (e) Shown are representative images from matched normal and patients at different stages of PDAC (Stage III not shown, as we had only one stage III patient among our cohorts). CD68 (magenta), GLUT1 (yellow), CD163 (aqua), HK2 (green), HIF1α (red), DAPI (blue). Scale bar = 100 µM. (f–m) The expression of GLUT1, HK2, and HIF1α on CD68+ macrophages was objectively quantified. CD68+ cells co-expressing each glycolytic marker (called “subsets”) were defined as a percentage of total CD68+ cells in individual patients. Shown are pie charts of the mean of the %GLUT1+ (f) in total cells, % CD68+GLUT1+ subset (i) and % CD68+ GLUT1+HK2+HIF1α+ (l) in adjacent “normal” versus PDAC (collectively) as a representation of total CD68+ cells. Shown are box plots of the mean ± s.d. of the %GLUT1 (g), % CD68+GLUT1+ subset (j) and % CD68+ GLUT1+HK2+HIF1α+ (m) across the different stages of disease compared with adjacent “normal” controls. Similar to Figure 6a, the percentage composition of the GLUT1+ (h), CD68+GLUT1+ subset (k) and CD68+ GLUT1+HK2+HIF1α+ subset (n) were thresholded at various percentages and the binary states (patients above the cut-off percentage was defined as “high”, and those below defined as “low”) were used in a log-rank survival analysis. Shown are Kaplan-Meier survival curves for the two subsets in which the percentage cut-offs yielded significant difference between the high and low patient populations.