Small-molecule GSDMD agonism in tumors stimulates antitumor immunity without toxicity

Abstract

Gasdermin-mediated inflammatory cell death (pyroptosis) can activate protective immunity in immunologically cold tumors. Here, we performed a high-throughput screen for compounds that could activate gasdermin D (GSDMD), which is expressed widely in tumors. We identified 6,7-dichloro-2-methylsulfonyl-3-N-tert-butylaminoquinoxaline (DMB) as a direct and selective GSDMD agonist that activates GSDMD pore formation and pyroptosis without cleaving GSDMD. In mouse tumor models, pulsed and low-level pyroptosis induced by DMB suppresses tumor growth without harming GSDMD-expressing immune cells. Protection is immune-mediated and abrogated in mice lacking lymphocytes. Vaccination with DMB-treated cancer cells protects mice from secondary tumor challenge, indicating that immunogenic cell death is induced. DMB treatment synergizes with anti-PD-1. DMB treatment does not alter circulating proinflammatory cytokine or leukocyte numbers or cause weight loss. Thus, our studies reveal a strategy that relies on a low level of tumor cell pyroptosis to induce antitumor immunity and raise the possibility of exploiting pyroptosis without causing overt toxicity.

Keywords: GSDMD; GSDMD agonist; antitumor immunity; cancer; checkpoint blockade; gasdermin; immunogenic cell death; immunotherapy; pyroptosis; tumor.

Figure 1.. High-throughput screening for small-molecule GSDMD agonists

(A) GSDMD-induced liposome leakage assay using time-resolved terbium (Tb³⁺)/dipicolinic acid (DPA) fluorescence. (B) Percentage activation of liposome leakage by screened compounds, assayed at 25 μg/mL (~50 μM for most compounds). Cutoff was 50% activation relative to detergent. (C) The 24 identified compounds. EC₅₀s of liposome leakage, GSDMD binding dissociation constants (K_D) by microscale thermophoresis (MST), and cell death activity are shown. N.D., not detected. The data are shaded by the in vitro EC₅₀s of the compounds (red for < 1.0 μM, yellow for 1.0 – 5.0 μM, green for > 5.0 μM, and dark green for N.D.) The selected hit C-185 is labeled in red. (D) Chemical structure of compound C-185, also known as DMB. (E) Dose-response curve with EC₅₀ of DMB in the liposome leakage assay. (F) The binding curve for Alexa 647-labeled GSDMD with DMB by MST. See also Figure S1.

Figure 2.. DMB induces pyroptotic cell death in a GSDMD-dependent manner

(A) Time course of PI positivity in WT and GSDMD KO THP-1 cells after treatment with DMSO, DMB (5 and 20 μM), or LPS + nigericin. (B) EC₅₀ determination for WT THP-1 cells using PI positivity at 4 h after treatment with different concentrations of DMB (red) or DMSO (blue). (C) Quantification of PI uptake 1, 2, and 4 h after treatment with DMSO or DMB (5 and 20 μM) in GSDMD KO THP-1 cells, GSDMD KO THP-1 cells reconstituted with WT GSDMD, or GSDMD KO THP-1 cells reconstituted with D275A uncleavable GSDMD mutant. (D and E) Time course of LDH release (D) and extracellular ATP measured by a luciferase-based assay (E) after treatment of WT and GSDMD KO THP-1 cells with DMSO, DMB (5 and 20 μM), or LPS + nigericin. Note discontinuous x axis in (D). (F and G) Quantification of PI uptake (F) and LDH release (G) after treatment with DMSO, DMB (5 and 20 μM), or LPS + nigericin, with or without pretreatment with DSF in human PBMCs. (H and I) Quantification of PI uptake (H) and LDH release (I) after treatment with DMSO, DMB (5 and 20 μM), or LPS + nigericin in WT and GSDMD KO primary mouse BMDMs. (J and K) Quantification of PI uptake (J) and LDH release (K) after treatment of WT and caspase-1/11 KO iBMDMs with DMSO or DMB (5 and 20 μM) in mouse iBMDMs. Error bars represent SEM of 3 independent experiments. Statistics were measured by Student’s t tests (C, F, H, and J) or two-way ANOVA (A, D, G, I, and K). NS, not significant; ****p < 0.0001. See also Figure S1.

Figure 3.. Cleavage-independent and selective activation of GSDMD pore formation by DMB

(A) Chemical structure of biotinylated DMB (left) and biotinylated DMB pull-down of GSDMD in THP-1 whole cell lysates (WCL), which was markedly reduced by excess DMB. The immunoblot used anti-GSDMD antibody NBP2–33422 (Novus Biologicals). * indicates a non-specific band. Lysate loaded was 5% total, and each lane is 15% total. (B) Negative-staining EM images of liposomes incubated with DMB or GSDMD alone, cleaved GSDMD, or GSDMD plus DMB upon liposome reconstitution and detergent solubilization. Scale bars, 100 nm. Yellow arrowheads point to DMB-induced GSDMD pores. Cleaved GSDMD and DMB + GSDMD bottom left insets show enlarged images of the boxed areas. (C) Liposome leakage assay showing that DMB activates recombinant GSDMD similarly as cleavage. The NT-CT linker contained an engineered 3C protease cleavage site. (D) BRET assay using NT-fused YFP and CT-fused luciferase showing that DMB treatment reduced the intramolecular BRET ratio relative to DMSO treatment, suggesting that DMB increased the distance between the GSDMD-NT and -CT. (E) Nano-LC/MS/MS spectrum of the Cys191-containing human GSDMD peptide (aa 184–203; 2,057.00 Da) modified on Cys191 (red) by carbamidomethyl (an increase of 57.0214 Da). A triplet-charged precursor ion m/z 705.6812 (mass: 2,114.0435 Da) was observed. (F) The corresponding GSDMD peptide after GSDMD incubation with DMB, which was modified on Cys191 (red) by the quinoxaline moiety of DMB (an increase of 267.0330 Da). A triplet-charged precursor ion m/z 775.6767 (mass: 2,324.0591 Da) was observed. (G) DMB dose-response curves and EC₅₀ of liposome leakage after incubation with human GSDMA, GSDMB, GSDMC, GSDMD, or GSDME (0.3 μM). (H) PI uptake in HEK293T cells transfected with WT and Cys191 mutants of GSDMD-NT. C191F and C191R GSDMD-NT induced comparable cell death as WT GSDMD-NT. (I) LDH release from HEK293T cells transfected with full-length GSDMD and treated with DMB or DMSO. WT, the D275A mutant that cannot be cleaved by inflammatory caspases, and the C268G/D275A double mutant that also cannot be cleaved by ELANE were activated by DMB. By contrast, the C191R and C191F mutants were resistant to DMB activation. (J) Leakage of liposomes induced by adding DMB to full-length C191A, C191R, C191F, or WT GSDMD. WT GSDMD was also pretreated with DSF before DMB treatment. Only WT GSDMD that was not pretreated with DSF could be activated by DMB. Data represent mean ± SEM of 3 independent experiments performed in triplicate. Statistics were measured by Student’s t tests (H). NS, not significant; **p < 0.01, ***p < 0.001, ****p < 0.0001. See also Figures S2, S3, and S4.

Figure 4.. GSDMD-dependent pyroptosis and DAMP release in mouse tumor cells

(A) Phase contrast, SYTOX Green staining, and merged images of WT (left) and GSDMD KO clone 10 (right) EMT6 cells treated for 2 h with DMSO, DMB (5 and 20 μM), or mitomycin C (MMC), an apoptosis inducer. Red squares indicate zoomed-in regions. Red arrows point to pyroptotic bubbles. Scale bars, 20 μm. (B) Quantification of % of cells that took up SYTOX Green 1, 2, and 4 h after treatment of WT and GSDMD KO clone 10 EMT6 cells with DMSO, DMB (5 and 20 μM), or MMC. (C and D) Time course of LDH release (C) and ATP release (D) after treatment of WT and GSDMD KO EMT6 clone 10 cells with DMSO, DMB (5 and 20 μM), or MMC. (E) Phase contrast, SYTOX Green staining, and merged images of CT26 cells treated for 2 h with DMSO, DMB (5 and 20 μM), or mitomycin C (MMC). Red squares indicate zoomed-in regions. Red arrows point to pyroptotic bubbles. Scale bars, 20 μm. (F) ATP release over time from CT26 cells after treatment with DMSO, MMC, or DMB. Note discontinuous x axis. (G) % SYTOX Green uptake positivity in DMB-treated EMT6 cells that were pretreated or not with the inflammatory caspase inhibitor AC-FLTD-CMK. Error bars represent SEM of 3 independent experiments. Statistics were measured by Student’s t tests (B and G) or two-way ANOVA (C). NS, not significant; *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. See also Figure S5.

Figure 5.. DMB induces antitumor activity that depends on tumor cell GSDMD expression

(A–D) Mice bearing orthotopic WT (left), Gsdmd^−/− (middle), or Gsdmd^−/−Gsdme^−/− (right) EMT6 tumors were treated with vehicle or DMB (10 mg/kg) every week starting when tumors became palpable and analyzed for tumor volume (A), percentage of CD8⁺ TILs expressing GzmB or PFN (B), percentage of CD8⁺ TILs expressing IFN-γ or TNF-α after PMA and ionomycin stimulation ex vivo (C), and percentage of NK TILs with GzmB and PFN expression (D). n = 5 mice/group. (E and F) Mice bearing subcutaneous CT26 tumors were treated with vehicle or DMB (10 mg/kg) and analyzed for tumor volume (E) and percentage of CD8⁺ or NK TILs expressing GzmB or PFN and percentage of CD8⁺ TILs expressing IFN-γ or TNF-α after PMA and ionomycin stimulation ex vivo (F). n = 6 mice/group. (G–J) Subcutaneously implanted KP tumor cells in WT or Gsdmd^−/− mice treated with vehicle or DMB and analyzed for tumor volume (G and H), and percentages of CD8⁺ (I) and NK TILs (J) expressing GzmB or PFN or expressing IFN-γ or TNF-α after PMA and ionomycin stimulation ex vivo. WT mice, vehicle or DMB treatment, n = 7 mice/group; Gsdmd^−/− mice, vehicle treatment, n = 5 mice/group, or DMB treatment, n = 6 mice/group. All data are represented as mean ± SEM. For tumor volume analysis, the area under the tumor growth curves was compared. Two-tailed Student’s t tests were used to determine differences between two groups. *p < 0.05, **p < 0.01, ***p < 0.001. See also Figure S6.

Figure 6.. DMB induces immunogenic cell death, and its antitumor effect depends on the immune response

(A) Experimental scheme to analyze DMB treatment of EMT6 tumors in WT or NSG mice. (B) DMB did not affect EMT6 tumor growth in NSG mice. (C) EMT6 tumor cells (CD45⁻ CD3⁻) showed increased PI uptake in DMB-treated WT mice compared with vehicle-treated WT mice (left), but DMB did not significantly change PI uptake in CD45⁺ CD11b⁺ F4/80⁺ TAMs (right). (D) DMB treatment did not increase PI uptake of EMT6 tumor cells (CD45⁻ CD3⁻) in NSG mice. (E) Schematic of vaccination experiment using MMC- or DMB-treated EMT6 tumor cells as immunogens. BALB/c mice were vaccinated in the left flank with MMC- or DMB-treated EMT6 cells and challenged 8 days later by injecting untreated EMT6 cells in the right mammary fat pad. MMC treatment, n = 10 mice/group; DMB treatment, n = 12 mice/group. (F) Average tumor volumes (left) and individual tumor growth kinetics for MMC (middle) and DMB (right) groups. (G–I) Percentages of CD8⁺ (G), NK (H), and CD4⁺ (I) TILs expressing GzmB or PFN and producing IFN-γ or TNF-α after PMA and ionomycin activation ex vivo. All data are represented as mean ± SEM. For tumor volume analysis, the areas under the tumor growth curves were compared. Two-tailed Student’s t tests were used to determine differences between two groups. *p < 0.05, ****p < 0.0001. See also Figure S6.

Figure 7.. DMB inhibits growth of B16 tumors expressing human GSDMD

(A) Design of the human GSDMD-GFP fusion protein ectopically expressed in B16 cells that do not express endogenous GSDMD. Black lines mark the regions of the protein recognized by the anti-GSDMD (ZRB1274, Sigma-Aldrich) and anti-GFP (2956, Cell Signaling Technology) antibodies used in (C). GFP and GSDMD-GFP were expressed in B16 cells to generate B16-GFP and B16-GSDMD-GFP cells, respectively. (B) Time-lapse images of B16-GFP and B16-GSDMD-GFP cells treated with 10 μM DMB showing morphological changes by brightfield and PI uptake as an indicator of pyroptosis. (C) Immunoblot of B16-GFP or B16-GSDMD-GFP incubated with 10 μM DMB for indicated times. Anti-GAPDH (60004–1-Ig, Proteintech) is a loading control. GSDMD-GFP was not cleaved by DMB. (D) Experimental scheme for investigating the effect of DMB treatment on B16 tumors expressing human GSDMD-GFP or GFP. B16 clones were implanted subcutaneously (sc) on day 0, and mice were treated i.p. every 3 days for 6 injections (red arrows) with 10 mg/kg DMB or vehicle beginning 5 days later when all mice had palpable tumors. (E) Growth of B16-GSDMD-GFP tumors (clone 9, left; clone 10, right) after treatment with DMB or vehicle (n = 8 mice/group). Tumor growth curves show mean ± SEM. for each timepoint and statistical analysis was performed by two-tailed Student’s t test comparing the area under the tumor growth curves. (F) Kaplan-Meier survival curve of mice bearing B16-GSDMD-GFP tumors treated with DMB or vehicle (n = 16 mice/treatment group, combining clones 9 and 10). Survival was analyzed by log-rank test. 5 of the 16 mice survived in the DMB-treated group. (G) Immunofluorescence microscopy analysis of infiltrating CD3⁺ CD8⁺ T cells and MHCII⁺ CD11c⁺ dendritic cells in B16-GSDMD-GFP tumors after treatment with DMB or vehicle (n = 3 mice). Representative images are shown at left. Quantification is shown at right for mean ± SEM based on n = 12 images from 3 tumors for each group. Anti-CD3, anti-CD8, anti-MHCII, and CD11c antibodies were 100,235, 100,728, 107,619, and 117,309 from BioLegend. (H) Response of B16-GSDMD-GFP tumors according to tumor size at the time of the first DMB treatment, plotted against the tumor size 3 days after the last DMB treatment. Two sets of data were included, mice treated 5 days after implantation shown in (D) and (E) and mice treated 10 days after implantation when the tumors were larger. Mice in the first and second groups that survived for the duration of the experiment are indicated by red and pink dots, respectively, and those that died or had to be sacrificed are indicated by black and gray dots, respectively. Simple linear regression was used to model the relationship between tumor volumes at initiation and after all DMB treatments. A Pearson’s r value was used to assess correlation, and the statistical significance of the respective linear regression slope (Wald test) is shown. See also Figures S6 and S7.