Fig 2: Characterization and verification of VNP-GD-induced tumor cell pyroptosis.a Direct observation of the pyroptosis of 4T1 cells under confocal microscope after different treatments with PBS, VNP, GD (GSDMD protein cage), VNP-GD (GSDMD protein cage-conjugated VNP), and VNP-GD+EI-NP (GSDMD protein cage-conjugated VNP + EI-NP) for 24 h. 4T1 cells were further stained with Annexin V for imaging. Scale bar: 15 µm. The experiments were repeated three times independently. Flow cytometry analysis of cell uptake of SYTOX green (b) and PI (c) in 4T1 tumor cells after incubation with PBS, VNP, GD (GSDMD protein cage), VNP-GD (GSDMD protein cage-conjugated VNP), and VNP-GD+EI-NP (GSDMD protein cage-conjugated VNP + EI-NP) for 24 h. d Data analysis of the PI-positive cells after different treatments. Data are presented as mean ± s.d. (n = 3 biologically independent samples) and analyzed with one-way ANOVA followed by Dunnett’s multiple comparisons test. VNP-GD + EI-NP vs. VNP-GD: **P = 0.0055; VNP-GD + EI-NP vs. GD: ****P < 0.0001. LDH (e) and HMGB1 (f) release after the treatments of PBS, VNP, GD (GSDMD protein cage), VNP-GD (GSDMD protein cage-conjugated VNP), and VNP-GD+EI-NP (GSDMD protein cage-conjugated VNP + EI-NP) for 24 h. Data are presented as mean ± s.d. (n = 3 biologically independent samples) and analyzed with one-way ANOVA followed by Dunnett’s multiple comparisons test. (e VNP-GD+EI-NP vs. VNP-GD: **P = 0.0068; VNP-GD+EI-NP vs. GD: ***P = 0.0001; f VNP-GD+EI-NP vs. VNP-GD: *P = 0.0119; VNP-GD + EI-NP vs. GD: ***P = 0.0007). Source data are provided as a Source Data file.

Fig 3: Preventing ESCRT-dependent cell membrane repair enhanced pyroptosis and augmented antitumor immune response.a Flow cytometry assay of dendritic cell maturation in the lymph nodes after treatments with PBS, aPD-1, VNP@Gel, GD/EI-NP@Gel, VNP-GD@Gel, VNP-GD/EI-NP@Gel and VNP-GD/EI-NP@Gel+aPD-1 (n = 4 mice). b Flow cytometry assay of CD8+ T cell infiltration in the tumor tissue after treatments as mentioned above, n = 4 mice. c Dendritic cell maturation in the lymph nodes after different treatments. Data are shown as mean ± s.d. and analyzed with one way ANOVA followed by Tukey’s multiple comparisons test, n = 4 mice (VNP-GD@Gel vs. VNP-GD/EI-NP@Gel: ***P = 0.0008; VNP-GD@Gel vs. VNP-GD/EI-NP@Gel+aPD-1: ***P = 0.0002; VNP-GD/EI-NP@Gel vs. VNP-GD/EI-NP@Gel+aPD-1: nsP = 0.4116). d The number of CD8+ T cells per mg tumor tissue after different treatments. Data are shown as mean ± s.d. and analyzed with one way ANOVA followed by Tukey’s multiple comparisons test, n = 4 mice (VNP-GD@Gel vs. VNP-GD/EI-NP@Gel: *P = 0.0215; VNP-GD@Gel vs. VNP-GD/EI-NP@Gel+aPD-1: ****P < 0.0001; VNP-GD/EI-NP@Gel vs. VNP-GD/EI-NP@Gel+aPD-1: **P = 0.0042). e The number of Granzyme B+CD8+ T cells per mg tumor tissue after different treatments, n = 4 mice (VNP-GD@Gel vs. VNP-GD/EI-NP@Gel: **P = 0.0065 VNP-GD@Gel vs. VNP-GD/EI-NP@Gel+aPD-1: ****P < 0.0001; VNP-GD/EI-NP@Gel vs. VNP-GD/EI-NP@Gel+aPD-1: **P = 0.0028). HMGB1 (f), IFNγ (g), and TNFα (h) expressions in the tumor tissues after different treatments detected by ELISA. Data are shown as mean ± s.d. and analyzed with one way ANOVA followed by Tukey’s multiple comparisons test, n = 4 mice. (f, VNP-GD@Gel vs. VNP-GD/EI-NP@Gel: **P = 0.0080; VNP-GD@Gel vs. VNP-GD/EI-NP@Gel+aPD-1: **P = 0.0017; VNP-GD/EI-NP@Gel vs. VNP-GD/EI-NP@Gel+aPD-1: nsP = 0.5319. g, VNP-GD@Gel vs. VNP-GD/EI-NP@Gel: *P = 0.0256; VNP-GD@Gel vs. VNP-GD/EI-NP@Gel+aPD-1: ***P = 0.0002; VNP-GD/EI-NP@Gel vs. VNP-GD/EI-NP@Gel+aPD-1: *P = 0.0127. h, VNP-GD@Gel vs. VNP-GD/EI-NP@Gel: **P = 0.0047; VNP-GD@Gel vs. VNP-GD/EI-NP@Gel+aPD-1: ****P < 0.0001; VNP-GD/EI-NP@Gel vs. VNP-GD/EI-NP@Gel+aPD-1: **P = 0.0053.) i Luminex-based quantification of cytokines and chemokines, including IL-1α, IFN-γ, TNF-α, MCP-1, IL-12, IL-1β, IL-6, IL-27, IL-17A, GM-CSF (n = 5 mice). Source data are provided as a Source Data file.

Fig 4: Working mechanism and preparation and characterization of the hydrogel-based bacteria protein cage delivery system.a The GSDMD proteins were crosslinked into protein cages and then conjugated on the surface of attenuated Salmonella typhimurium (designated VNP-GD). The ESCRT inhibitor was loaded in the dextran nanoparticles (designated EI-NP). Two formulations, an injectable hydrogel and a cell patch, were developed to co-load VNP-GD and EI-NP to treat primary tumors through local administration and inoperable cancer through implantation. b The underlying mechanism of tumor pyroptosis triggered by VNP-GD and further enhanced by EI-NP. Firstly, after the invasion of the VNP-GD into the tumor, the GSDMD protein would be released upon GSH stimulation, and the abundant flagella on the surface of bacteria could activate the caspase 1 into cleaved caspase 1, which will further cleave the GSDMD protein to the N-terminal GSDMD that will multimerize and perforate the cell membrane, initiating cell pyroptosis. Secondly, the released ESCRT inhibitor from EI-NP could effectively block the calcium influx to inhibit the ESCRT III-mediated membrane repair to enhance the tumor pyroptosis. c Particle size and transmission electronic microscope (TEM) image (inserted) of the protein cages (scale bar = 100 nm). d Representative TEM image of the protein cage-conjugated VNP bacteria (VNP-GD, scale bar = 200 nm). The experiments were repeated three times independently. e Confocal images of the conjugation of the protein cage (labeled with Rhodamine B) on the surface of the VNP bacteria (labeled with Hoechst), scale bar = 10 μm. The experiments were repeated three times independently. f Cumulative release of the protein from the bacteria protein cage with or without the trigger of GSH (10 mM). Data are presented as mean ± s.d. (n = 3 biologically independent samples). g Particle size and transmission electronic microscope (TEM) image (inserted) of the EI-NP (scale bar = 500 nm). h ESCRT inhibitor release profile from the dextran nanoparticle (EI-NP) at predetermined time points. Data are presented as mean ± s.d. (n = 3 biologically independent samples). Source data are provided as a Source Data file.

Fig 5: Working mechanism of GSDMD-induced pyroptosis and ESCRT III-mediated cell membrane repair.a Schematic illustration of the pyroptosis-related signaling pathway. b Western blot assay of the pyroptosis signaling pathway in 4T1 cells after treatments of PBS, VNP, GD (GSDMD protein cage), VNP-GD (GSDMD protein cage-conjugated VNP), and VNP-GD+EI-NP (GSDMD protein cage-conjugated VNP + EI-NP). The experiments were repeated three times independently. c Schematic illustration of the calcium influx induced ESCRT III-mediated membrane repair during cancer cell pyroptosis. d Flow cytometry assay of the Ca2+ influx detection with Fluo-8 AM in 4T1 cells after treatment with PBS, VNP, GD, VNP-GD, and VNP-GD+EI-NP for 24 h (n = 3 biologically independent samples). e Direct observation of the ESCRT III-mediated 4T1 cancer cell membrane repair during pyroptosis after treatment with VNP-GD and VNP-GD+EI-NP for 24 h (n = 4 biologically independent samples). The cell membrane was stained with Annexin V. The 4T1 cells were genetically engineered to express mCherry-labeled charged multivesicular body protein 3 (CHMP 3) protein, which is the main component of ESCRT III machinery. Scale bar: 5 µm. The experiments were repeated three times independently. Source data are provided as a Source Data file.