NLRP3 Inflammasome Activation Expands the Immunosuppressive Myeloid Stroma and Antagonizes the Therapeutic Benefit of STING Activation in Glioblastoma

Abstract

Glioblastoma (GBM) is the most common and deadly primary brain malignancy and is clinically refractory to immunotherapy. Active NLRP3 inflammasome signaling and IL-1β secretion have been observed in GBM, and NLRP3-driven myeloid-derived suppressor cell (MDSC) recruitment can mediate cancer immune evasion. Agonists of the cytosolic double-stranded DNA-sensing stimulator of IFN gene (STING) pathway can mediate proinflammatory conversion of cancer MDSCs; however, secretion of the NLRP3 products IL-1β and IL-18 has also been observed in certain myeloid populations following STING activation. In this study, we aimed to determine both the potential mechanistic synergy between STING and NLRP3 agonists, and the effects of this innate immune combination on the GBM tumor immune landscape. We find that STING activation does not prime pro-IL-1β expression for activated NLRP3 inflammasome secretion. In subcutaneous GL261 GBM, we show that NLRP3 activation expands the immunosuppressive myeloid stroma primarily via granulocytic MDSC recruitment and antagonizes the benefit of STING activation. In brain GL261, we find that NLRP3 activation expands granulocytic MDSCs but does not antagonize the therapeutic benefit of STING activation. Finally, we report that mesenchymal subtype GBM tumors have elevated neutrophil, IL-1β, and NLRP3 gene expression, a setting where our data suggest that NLRP3 activation could counteract STING agonists.

Significance: NLRP3 inflammasome signaling, which suppresses antitumor immunity in some cancers, has been observed in GBM tissues. NLRP3 activation in GBM induces granulocyte-dependent tumor immunosuppression and antagonizes the therapeutic efficacy of STING activation.

Figure 1

STING-induced NF-κB activation does not prime the NLRP3 inflammasome. A, THP1-Dual cells were incubated overnight with vehicle, MLRR (10 μg/mL), or LPS (1 μg/mL). The supernatant was harvested and NF-κB induction measured using the InvivoGen QUANTI-Blue secreted embryonic alkaline phosphatase detection system. B and C, BMDMs were incubated overnight with treatments as described in A. Cells were divided and processed for both IL-1β qPCR (B) and flow cytometry (C) as described in methods and stained intracellularly with eFluor780 αIL-1β proform to assess pro–IL-1β production. D, BMDMs were incubated overnight with priming drugs LPS or MLRR as described in A. Cells were then collected and treated for 2 hours in NLRP3 activator nigericin (10 µmol/L), inhibitor MCC950 (1 µmol/L), or combination as indicated. The supernatant was harvested and IL-1β secretion measured via R&D Systems murine IL-1β ELISA kit. Statistical significance was calculated using the Student t test. ns, not significant; ****, P < 0.0001. MFI, mean fluorescence intensity.

Figure 2

In vivo characterization of STING and NLRP3 inflammasome activation in GL261. A, Mice received subcutaneous injection of 1.0 × 10⁶ GL261 cells in 30% Matrigel and then were injected intratumorally with vehicle, STING agonist cdGMP (25 μg), NLRP3 agonist nigericin (25 μg), or combination on days 10, 14, and 18. Tumors were then processed for tumor harvest and flow cytometry as described in methods (B, C, H–K) or measured for survival (D and E). Data shown represent (B) GL261 harvested tumor weights. C, Analyzed CD45⁺ cell frequency of total single cells. D, GL261 tumor volumes and (E) survival data. F–I, Overall composition and fold changes of cell densities vs. vehicle-treated tumors for analyzed CD45⁺ (F and G) myeloid and (H and I) lymphoid cell populations. J, Mice received intracranial injection of 5.0 × 10⁴ GL261 cells in PBS in the right striatum via a guide bolt system, then were treated intratumorally via the guide bolt on day 7 with vehicle, STING agonist cdGMP (5 μg), NLRP3 agonist nigericin (5 μg), or combination, and survival monitored (K). Error bars represent mean ± SEM. Statistical significance was calculated using the Student t test or log-rank test. ns, not significant; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. DC, dendritic cells; SubQ, sub-cutaneous.

Figure 3

In vivo analysis of the GL261 tumor immune microenvironment following STING and NLRP3 inflammasome activation. Mice received subcutaneous injection of 1.0 × 10⁶ GL261 cells in 30% Matrigel and then were injected intratumorally with vehicle (n = 6), STING agonist cdGMP (25 μg; n = 10), NLRP3 agonist nigericin (25 μg; n = 9), or combination (n = 10) on days 10, 14, and 18. Tumors were then processed for tumor harvest and flow cytometry as described in methods. Data shown represent Gr-MDSC (A), Mono-MDSC (B), and TAM (C) frequency as a percent of total analyzed CD45⁺ cells. D, Arginase, (E) CD206, (F) Ki67, and (G) iNOS expression on Gr-MDSC, Mono-MDSC, and TAM populations as indicated. Ratios of number of (H) CD8 T cells/FOXP3⁺ CD4 Tregs and (I) Gr-MDSC/CD8 T cells. Among NK cells, granzyme⁺ NK cells were identified and reported as (J) density per mg of tumor tissue and (K) frequency as a percent of total analyzed CD45⁺ cells. L, Granzyme and (M) Ki67 expression on CD4, CD8, and NK cell populations as indicated. Error bars represent mean ± SEM. Statistical significance was calculated using the Student t test. ns, not significant; *, P < 0.05; **, P < 0.01; ****, P < 0.0001. iNOS, inducible nitric oxide synthase.

Figure 4

In vivo analysis of intracranial GL261 tumor immune microenvironment composition following STING and NLRP3 inflammasome activation. A, Mice received intracranial injection of 5.0 × 10⁴ GL261 cells, then were treated with 5 μg of cdGMP and/or nigericin, and tumor-bearing hemispheres harvested 48 hours following treatment for spectral flow cytometry analysis. B, Live CD45⁺ cell densities reported as cells per tumor-bearing hemisphere. C, Live CD45⁺ cell frequency as a fraction of total single cells. D–G, Overall composition and fold changes of cell densities vs. vehicle-treated tumors for analyzed CD45⁺ (D and E) myeloid and (F and G) lymphoid cell populations. Error bars represent mean ± SEM. Statistical significance was calculated using the Student t test. ns, not significant; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. cDC1, type 1 conventional dendritic cells; Neg., negatively.

Figure 5

In vivo analysis of intracranial GL261 tumor immune microenvironment phenotypes following STING and NLRP3 inflammasome activation. Mice received intracranial injection of 5.0 × 10⁴ GL261 cells, then were treated with 5 μg of cdGMP and/or nigericin, and tumor-bearing hemispheres harvested 48 hours following treatment for spectral flow cytometry analysis. A, Gr-MDSC, (B) Mono-MDSC, and (C) microglia frequency as a percent of total analyzed CD45⁺ cells. D, CD206, (E) arginase, and (F) PD-L1 expression on Gr-MDSCs, Mono-MDSCs, and microglia as indicated. Ratios of the number of (G) CD8 T cells/FOXP3⁺ CD4 Tregs and (H) Gr-MDSC/CD8 T cells. Expression of (I) PD-1 and (J) LAG-3 on CD8 T cells. K, Granzyme expression on CD8 Teff, NK cells, and NKT cells as indicated. Among CD8 Teff, NK cells, and NKT cells, granzyme⁺ cells were identified and reported as (L) density as cells per tumor-bearing hemisphere and (M) frequency as a percent of total analyzed CD45⁺ cells. Error bars represent mean ± SEM. Statistical significance was calculated using the Student t test. ns, not significant; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001. MFI, mean fluorescence intensity.

Figure 6

TCGA analysis of neutrophil-related genes in patients with primary GBM. Bulk RNA-seq transcriptome profiling data were normalized, and neutrophil gene scores based on the expression of S100A8, S100A9, and ITGAX were assigned to each patient sample. A, Survival was analyzed comparing patients with low (bottom quartile) and high (top 75%) neutrophil gene scores. B, Survival was analyzed comparing patients with high (top quartile) and low (bottom 75%) expression of the gene IL1B. Neutrophil gene scores were compared across IDH mutation status (C) and transcriptome molecular subtypes (D). Expression of IL1B (E) and NLRP3 (F) genes were compared across transcriptome molecular subtypes. Error bars represent mean ± SEM. Statistical significance was calculated using the Student t test or log-rank test. CL, classical; ME, mesenchymal; PN, proneural; WT, wild-type.