Close this search box.

Stimulation of tumoricidal immunity via bacteriotherapy inhibits glioblastoma relapse – Nature Communications

Design of Salmonella delivery vehicles targeted to the GBM microenvironment

Three essential components are required for intracranial administration of bacterial delivery systems: (1) accurate targeting of the tumor microenvironment, (2) effective cell invasion, and (3) excellent safety and controllability. Safety is a prerequisite for the successful application of autolyzing SDVs in the treatment of intracranial disorders. The Salmonella strain VNP20009 (with ΔmsbB, ΔpurI, and Δxyl deletions) is a well-known bacterial delivery system, but it is not safe enough for intracranial delivery due to the distinctive features of intracranial immunity30,31. We designed two genetic circuits that can activate invasion protein A (invA) expression under hypoxic conditions to enhance SDV invasion and induce Lysin E (LysE) expression, which causes bacterial lysis and subsequent release of lysogenic components in GBM cells (Fig. 1a). Exploiting the potential of P-selectin on the surface of megakaryocyte-derived exosomes to bind to CD44, which is highly expressed on the surface of GBM cells, we achieved precise active targeting to tumor cells (Supplementary Fig. 1). In pursuit of a synergistic therapy employing SDVs, we orchestrated the expression of exosome-localized brain acid soluble protein 1 (BASP1) and gasdermin D (GSDMD), where BASP1 anchors the GSDMD to the surface of the exosome membrane. Further, the construction of SLINs achieved by encapsulating L-arabinose in GSDMD-EXO through electroporation, SLINs can be degraded intracellularly releasing L-arabinose which could further induce SDVs to express LysE. This fusion protein was strategically positioned on the exosomal membrane, followed by coupling of the exosomes with the SDVs surface via an azide-alkyne cycloaddition reaction. This strategy culminated in the creation of an IASNDS, as depicted in Fig. 1b. We next drew inspiration from the model of phage-mediated bacterial lysis to strengthen this conceptual framework. L-arabinose is added to exosomes to form SLINs, which initiate SDV autolysis. The structural configuration not only guarantees the safety of intracranial SDVs application but also augments the capacity of SDVs to induce immunogenic cell death (ICD). This, in turn, increases the infiltration of immune cells within the GBM microenvironment and amplifies the innate immune response associated with the SDV system.

Fig. 1: Mechanism, preparation, and characterization of the hydrogel-based autolysing bacterial delivery system.
figure 1

a Schematic depicting the strategic transformation of Salmonella into the SDV and the induction of ICD in GBM cells via Salmonella VNP20009. b Illustration detailing the preparation of the IASNDS and synergistic localized treatment with the hydrogel for surgical cavity application. c Visualization of fluorescence (GFP+) in the SDV (green) under both normoxic and hypoxic conditions for 24 hours. Red dashed lines represent GL261 cell boundaries. (n = 3 independent experiments). d Statistical analysis showing the intensity of GFP fluorescence in normoxic and hypoxic cultures. (n = 12 images from three independent experiments). (exact P value: P =  2.42423E-09); ****P < 0.0001. e, f In vivo distribution assessment of 1 × 10^7 CFU SDVs (GFP) injected into intracranial tumor-bearing mice via the tail vein. Brain GBM tissue sectioning was performed, with the SDVs depicted by a white arrow. g Additional fluorescence staining of brain tissue sections to observe the SDV distribution (GFP, shown in red) in intracranial GBM tissues (white arrows). (n = 3 independent experiments). h Visualization of the green fluorescence distribution of the SDVs (GFP+) under pBAD-LysE (-) and pBAD-LysE (+) conditions induced by 500 μM L-arabinose for 24 hours in the cell culture medium (n = 3 independent experiments). The intracellular GFP distribution in the pBAD-LysE (+) condition indicates initiation of SDV autolysis in GL261 cells. Data are shown as the means ± SEMs. The statistical comparisons in d were performed with two-tailed, unpaired Student’s t tests, with asterisks indicating significant differences. c, f, g, and h show representative images of the corresponding independent biological samples. (n = 3 independent biological samples). The scale bar in c is 1 μm, the scale bar in f is 1.5 mm, the scale bars in g are 50 μm (middle) and 10 μm (right), and the scale bar in h is 1 μm. Source data are provided as a Source Data file.

The rapid proliferation of GBM cells occurs simultaneously with the delayed development of the tumor vasculature, creating a hypoxic environment within the tumor32,33. The hypoxic state frequently precipitates necrotic tissue formation in the central region of the GBM tumor, thereby establishing a distinctive hallmark that sets it apart from other benign intracranial tumors. In this study, to investigate the capability of SDVs to invade GBM cells, we discovered that autolysing SDVs labeled with a green fluorescent protein (GFP+) proliferated and generated Salmonella-containing vesicles (SCV) in GBM cells (Supplementary Fig. 2). To further validate the response of the SDVs to hypoxia (1% oxygen), the efficacy of the SDVs (GFP+) was subsequently investigated in hypoxic and normoxic (21% oxygen) environments, and the results showed that under hypoxic conditions, the survival rate of the SDVs was increased and the fluorescence intensity was enhanced (Fig. 1c, d).

To gain insight into the spatial distribution of SDVs in a syngeneic orthotopic mouse GBM model, we intravenously injected SDVs and VNP20009 (Fig. 1e, f). Compared to VNP20009, the SDVs (red) showed significant aggregation at the tumor site, and dispersion within normal brain tissue was not evident (white arrows, Fig. 1g, bottom). The mechanism of SDVs accumulation and survival in GBM is due to the introduction of FNR-invA, and the local immunosuppressive microenvironment also inhibited the immunological clearance of SDVs. The observed increased porosity in tumor tissue within the SDVs group seems to be associated with the occurrence of pyroptosis in GBM cells, leading to clearance by the immune system. In contrast to the parental Salmonella strain devoid of Lysin E (LysE), the SDV engineered to express Lysin E (LysE+) from bacteriophage phiX174 induced lysis and promoted GFP release into the tumor cytoplasm upon exposure to L-arabinose (Fig. 1h). Notably, the half maximal effective concentration (EC50) of L-arabinose was determined to be 91.5767 μM (Supplementary Fig. 3). In summary, these findings imply that our designed SDV is capable of targeting the hypoxic region of the GBM and inducing lysis to release bacterial components.

Design of Salmonella lysis-inducing nanocapsules

Pyroptosis, an important form of ICD, plays a vital role in the fight against infections and endogenous danger signals34,35. Cytotoxic lymphocytes rely on gasdermin-mediated pyroptosis to kill tumor cells, suggesting that pyroptosis is also intimately involved in anticancer immune responses. Among the gasdermin family proteins, the membrane perforin gasdermin D (GSDMD) serves as a potent executor of pyroptosis, and its functions and mechanisms of action in the activation of immune cells have been demonstrated36,37. Here, GSDMD was attached to the exosome surface by the exosome membrane protein BASP1 to increase the content of GSDMD in tumor cells (Fig. 2a). A significant increase in GSDMD levels in megakaryocytes (MEG01) (an approximately 9.94-fold increase) was observed by PCR analysis (Fig. 2b). This increase in GSDMD protein expression was substantiated further through Western blotting (Fig. 2c). Subsequent harvesting of exosomes produced by MEG01 corroborated the presence of exosome-specific proteins on the surface of GSDMD-EXOs and demonstrated overexpression of GSDMD (Fig. 2d, e). Comparing the difference between SDV equipped with GSDMD (+) SLIN and GSDMD (−) SLIN, respectively, we found that GSDMD ( + ) SLIN in combination with SDV could significantly increase cellular pyroptosis (Supplementary Fig. 4).

Fig. 2: Design and characterization of Salmonella lysis-inducing nanocapsules.
figure 2

a Design of the gene construct illustrating GSDMD overexpression and attachment to the exosome membrane surface. b Verification of GSDMD overexpression to evaluate transfection efficiency in the MEG01 cell line through PCR (n = 4 independent experiments). (exact P value: P = 2.75356E-06); ****P < 0.0001. c Western blot analysis showing GSDMD-N protein expression in MEG01 (GSDMD+) cells (n = 3 independent experiments). d The collected exosomes were confirmed to express exosome-specific proteins by Western blotting (n = 3 independent experiments). e Western blot results confirming GSDMD-N protein expression in exosome-derived (GSDMD-EXO) samples (n = 3 independent experiments). f Transmission electron microscopy reveals the SLIN morphology and particle size distribution (g) using dynamic light scattering (DLS). (= 3 independent experiments). h Kinetic profile of L-arabinose release from SLINs in PBS, pH 7.4 (n = 3 independent experiments). i Particle size changes of SLINs loaded in a 10 nm dialysis bag and coincubated with 1 × 10^6 CFU SDVs for 7 days at 4 °C. (n = 3 independent experiments). Data are presented as the mean ± S.D. The statistical comparisons in b were performed with two-tailed, unpaired Student’s t tests, with asterisks indicating significant differences. cf show representative images of the corresponding independent biological samples. The scale bar in f is 50 nm. Source data are provided as a Source Data file.

The pivotal step for the release of immune-activating components from SDVs depends on the initiation of autonomous lysis after cellular invasion. The release of L-arabinose in GBM cells induced the activation of the pBAD promoter within the SDV, which initiated the LysE expression. This circuit was achieved by encapsulating L-arabinose in GSDMD-EXO via electroporation, forming nanocapsules named SLINs with 4.82% drug loading capacity measured by HPLC. Further characterization of these SLINs revealed that their morphology did not change significantly after loading with L-arabinose, and the particle size was approximately 100 nm (Fig. 2f, g). SLINs were able to maintain relatively good stability over one week, and the content of L-arabinose remained in the desirable range (Fig. 2h). SLINs were placed in dialysis bags with a diameter of 10 nm and incubated with 1 × 10^7 bacterial CFU (Colony-Forming Units) to observe their stability, and their size did not change significantly within one week (Fig. 2i). Hence, SLINs are highly stable prior to their invasion of GBM cells, without the early release of the loaded L-arabinose.

The IASNDS invades GBM cells and initiates autolysis

Free GSDMD does not diffuse into GBM cells mainly due to its large molecular weight and negative surface charge, and GSDMD is unable to trigger cellular pyroptosis owing to concealment of its pore-forming domains38. To address this challenge, Salmonella typhimurium was used for intracellular delivery of GSDMD. Due to the properties of the intracellular bacterium in the SDV in the IASNDS, the SLINs on its surface are brought inside the GBM cell, and after inducing pyroptosis, they further activate innate and adaptive immunity to combat postoperative relapse (Fig. 3a).

Fig. 3: Mechanism of immunogenic cell death induced by the autolysing SDV system.
figure 3

a Illustration depicting the IASNDS system invading GBM cells, triggering cell death, activating antigen-presenting cells, and eliciting an immune response. b IASNDS cellular entry, intracellular autocleavage initiation, and tumor cell pyroptosis induction mechanism. c Transmission electron microscopy characterizing the structures of the SDVs and IASNDS, highlighting SLINs coupled to the IASNDS surface (white arrow). (n = 3 independent experiments). d GFP release from SDVs into the cytoplasm (green, black arrow) and the presence of lysed Salmonella membranes inside cells (red, white arrow) after coculture of 1 × 10^6 CFU of the IASNDS with QL01#GBM cells for 48 hours (n = 3 independent experiments). e Detection of the GFP fluorescence intensity inside tumor cells after coincubation of 1 × 10^6 CFU of the IASNDS with QL01#GBM cells for 48 hours (n = 9 images from three independent experiments). (exact P value: P = 2.491E−12); ****P < 0.0001. f, g Intratumor cell fluorescence intensity over 24 hours in QL01#GBM cells, illustrating IASND intracellular release behavior (n = 9 images from three independent experiments). (exact P value: P = 7.8029E-11); ****P < 0.0001. h Altered cell membrane permeability and cytoplasmic efflux post IASNDS treatment of QL01#GBM cells (green, white arrow), along with dispersed bacterial membranes after SDV cleavage (red, black arrow). (n = 3 independent experiments). Data are presented as the mean ± S.D. The statistical comparisons in e was performed with two-tailed, unpaired Student’s t tests, with asterisks indicating significant differences. c, d, f, and h are representative images of the corresponding independent biological samples. The scale bars in c are 150 nm (middle) and 50 nm (bottom), the scale bar in d is 0.5 μm, and the scale bars in f and h are 1 μm. Source data are provided as a Source Data file.

The IASNDS arrives in the cell and forms Salmonella-containing vesicles (SCVs), which can rapidly multiply and prevent cellular clearance (Supplementary Fig. 5). SLINs can be degraded by GBM cells and release GSDMD and L-arabinose, slowly initiating the SDV lysis process, which can further activate caspase 1 to cleaved caspase 1. Then, cleaved caspase 1 converts GSDMD into pore-forming domain N-terminal GSDMD to bind to the GBM membrane and subsequently trigger cellular pyroptosis (Fig. 3b). As a type of ICD, pyroptosis releases tumor antigens, cytokines, and chemokines that significantly activate antitumor immune responses.

Click chemistry was used to achieve efficient coupling of SLINs to the SDV surface (Fig. 3c). The SDV surface was carefully functionalized with azido-PEG4-NHS ester, facilitating subsequent coincubation with DBCO-C6-NHS-modified SLINs. An average of approximately 97 SLINs are bound to the surface of each SDV by this coupling method. Tethering SLINs to the surface of the SDV can catalyze the cleavage process of the SDV by enabling the intake of large amounts of L-arabinose when the IASNDS enters GBM cells (Fig. 3d, e). Dynamic monitoring of the green fluorescence intensity in the cells revealed that when cocultured with 1 ×10^8 CFU of the IASNDS, the SDV slowly initiated the autocleavage process in the cells within 24 hours, with increasing fluorescence intensity of GFP (Fig. 3f, g, Supplementary Fig. 6). Bafilomycin A1 was employed to inhibit lysosomal function, the result showed that lysis of SDVs was significantly inhibited, thus suggesting that degradation of SLINs is achieved via the lysosomal pathway (Supplementary Fig. 7). After approximately 48 h of coculture, the intensity of green fluorescence in the cells reached a maximum, which indicated complete autolysis of the SDV; at the same time, there was a large amount of cytoplasmic leakage in the GBM cells, confirming the formation of pores on the cell surface after the polymerization of GSDMD-N (Fig. 3h, Supplementary Fig. 8).

GBM cell pyroptosis and antitumor immune activation

Pyroptosis of GBM cells is a vital step in the recruitment and activation of antitumor immune cells and is central to the prevention of GBM recurrence after surgery. We found that the IASNDS was indeed able to induce pyroptosis in GBM cells, and the GBM cells continued to expand and form balloon bubbles until the cell membrane ruptured (Fig. 4a, b, Supplementary Fig. 9). 3D tumor spheres and neurospheres were designed to reveal changes in the tumor invasion behavior of the IASNDS (Fig. 4c). Relative to that of both the control and SLIN-treated groups, the invasiveness of the GBM spheres was markedly attenuated following 24 hours of IASNDS treatment. This manifested as a 50.8% and 61.4% reduction in invasion distance, respectively, compared to that of the control group (Fig. 4d, e, Supplementary Fig. 10).

Fig. 4: Validation of IASNDS-induced GBM pyroptosis and ICD-related changes in vitro.
figure 4

a Visualization of cellular pyroptosis changes induced by the IASNDS under confocal microscopy, showing membrane pores and vacuole formation in QL01#GBM cells (n = 3 independent experiments). b Quantification of vacuoles in cells from different treatment groups to assess pyroptosis efficiency (n = 6 images from three independent experiments). (exact P values: SLIN vs. SDV P = 6.28201E−05, IASNDS vs. SDV P = 4.22634E−06); ****P < 0.0001. ce 3D neurosphere and QL01#GBM tumor sphere invasion experimental design, confocal images of the extent of GBM sphere invasion over time (n = 3 independent experiments). fh Flow cytometry analysis of stained cells (SYTOX, PI). (n = 6 independent experiments). (exact P values of g: Ctrl vs. SLIN P = 5.45783E−08, SDV vs. IASNDS P = 1.141E−12; exact P values of h: Ctrl vs. SLIN P = 1.24E−13, SDV vs. SLIN P = 7.5E−14,  SDV vs. IASNDS P = 1.9E−14); ****P < 0.0001. i, j HMGB1 and ATP release from QL01#GBM cells after 24 hours of PBS, SDV, SLIN, and IASNDS treatment, analyzed by flow cytometry (n = 3 independent experiments). (exact P value of i: SDV vs. IASNDS P = 5.8384E-06; exact P value of j: SDV vs. IASNDS P = 1.38256E-05); ****P < 0.0001. k PCR detection of cytokine and chemokine expression trends after 24 hours of PBS, SDV, SLIN, and IASNDS treatment of QL01#GBM cells. (n = 3 independent experiments). Data are presented as the mean ± S.D. The statistical comparisons in b, e, gj were performed using one-way ANOVA with Tukey’s post hoc test, with asterisks indicating significant differences (ns = no significance, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001). a, d show representative images of the corresponding independent biological samples. The scale bar in a is 5 μm, and the scale bar in d is 1 mm. Source data are provided as a Source Data file.

SYTOX Green was used for staining to evaluate the integrity of the cell membrane. When cells undergo pyroptosis, the permeability of the cell membrane increases and SYTOX binds to nucleic acids and exhibits a significant increase in fluorescence intensity. As shown in Fig. 4f, g, the SYTOX fluorescence intensity was significantly increased after administration with the IASNDS treatment, with 15.14, 2.25, and 2.71 times higher than those of the Ctrl, SDV, and SLIN groups, respectively. In addition, propidium iodide (PI) staining further confirmed that the cells underwent programmed death and exhibited cell membrane damage (Fig. 4h, Supplementary Fig. 11).

The therapeutic mechanism of action of the IASNDS involves the activation of GSDMD, which catalyzes the formation of pores on the cell surface and results in cellular swelling, ultimately facilitating the release of cellular contents. To validate this, our assessment included the quantification of lactate dehydrogenase (LDH) levels within the culture medium. IASNDS treatment caused an approximately 3.54-fold increase in LDH levels relative to those of the SDV group (Supplementary Fig. 12). Notably, elevated secretion of high mobility group box 1 (HMGB1) and adenosine triphosphate (ATP) after IASNDS treatment signified the occurrence of ICD within tumor cells (Fig. 4i, j). Further studies showed that IASNDS treatment caused a significant increase in the levels of cytokines and chemokines associated with immune activation of THP-1 cells after co-incubation with QL01#GBM cells (Fig. 4k). In conclusion, the pronounced amplification of the signaling cascade induced by this bacteriotherapy underscores the efficacy of the IASNDS system in recruiting and activating a variety of immune cells, including phagocytes, natural killer cells, and T cells.

Smart hydrogels enable the release of ATP-responsive CpG oligonucleotides

Since the brain tissue is filled with flowing cerebrospinal fluid which is unfavorable to the local application of the drug, hydrogel as a carrier for local drug delivery can ensure the sustained release in the operative cavity, avoiding the toxicity to the normal brain tissue. Intracellular SDVs curtail the extracellular release of nucleic acids after bacterial lysis, potentially posing challenges to achieving enduring immunotherapeutic effects with the SDVs. This limitation implies that the activation of adaptive immunity may not be durably sustained once host cells are cleared by phagocytes. In light of these considerations, we devised a strategy involving an immune-activating hydrogel (termed “Gel”) to couple CpG ODN with a nucleic acid aptamer. This approach capitalizes on the release of ATP resulting from IASNDS-induced cellular pyroptosis, thereby fostering sustained activation of immune cells (Fig. 5a). Notably, the gel, which is cross-linked using UV irradiation, confines the gel to the surgical cavity, effectively circumventing potential interference with cerebrospinal fluid reflux and mitigating potential side effects.

Fig. 5: ATP-responsive hydrogel design and characterization.
figure 5

a Schematic of the ATP-responsive hydrogel used to encapsulate the IASNDS. b PAGE analysis of HAMA-Apt synthetic conjugates formed from reaction mixtures of ATP nucleic acid aptamer (lane 1) and Apt with HAMA at molar ratios of 100:1 (lane 2), 50:1 (lane 3), 25:1 (lane 4), 12:1 (lane 5), 6:1 (lane 6), and 1:1 (lane 7). (n = 3 independent experiments). c PAGE analysis of the supernatant after binding of HAMA-Apt to CpG ODN with molar ratios of HAMA-Apt to CpG ODN of 1:0 (lane 8), 1:1 (lane 7), 1:2 (lane 6), 1:4 (lane 5), 1:8 (lane 4), 1:16 (lane 3), 1:32 (lane 2) and 1:64 (lane 1). (n = 3 independent experiments). d Representative scanning electron microscope image of the hydrogel (n = 3 independent experiments). e Representative photographs of hydrogels before and after gel formation. f Variation in the hydrogel modulus with time, measured at an angular frequency of 1 rad s-1. (n = 3 independent experiments). g Analysis of hydrogel variation with frequency, measured at a fixed strain of 0.5%. (n = 3 independent experiments). h Schematic of ATP-induced dehybridization of double-stranded BHQ3-modified Apt and Cy5.5-modified CpG structures, with fluorescence spectra showing the fluorescence recovery of Cy5.5 under ATP induction. (n = 3 independent experiments). i Schematic showing the release of Cy5-labeled CpG from the hydrogel. j Cumulative amount of CpG ODN released from the hydrogel under different concentrations of ATP. (n = 3 independent experiments). Data are presented as the mean ± S.D. The images in bd are representative images of the corresponding independent biological samples. The scale bars in d are 20 μm (left) and 2 μm (right). Source data are provided as a Source Data file.

The initial phase of this endeavor involved the integration of an ATP-specific aptamer (Apt) into the molecular structure of hyaluronic acid methacryloyl (HAMA) via an amide bond. The linkage of Apt and HAMA was evaluated via polyacrylamide gel electrophoresis (PAGE), which demonstrated a significant band shift when the molar ratio of Apt to HAMA reached 12:1 (Fig. 5b). Subsequently, the binding efficacy of HAMA-Apt and CpG was assessed, revealing heightened efficacy at a 1:8 ratio of the two components (Fig. 5c). Further evaluation of the properties of the IASNDS-loaded gel revealed its capacity to maintain remarkable porosity and robust gel formation (Fig. 5d, e). After 10 seconds of UV irradiation, both the G′ (storage modulus) and G″ (loss modulus) values exhibited notable time-dependent increases. Notably, the G′ value surpassed the G″ value by a factor of approximately ten, signifying successful gel formation (Fig. 5f, g).

The double-stranded structure of the gel can dissociate in the presence of ATP due to Black Hole Quencher 3 (BHQ3, a quencher of Cy5.5) and the proximity of CpG-Cy5.5 to the quenching reaction, suggesting that the CpG ODN can be released in response to elevated ATP levels (Fig. 5h). Further studies confirmed that CpG ODN could indeed be liberated in the presence of ATP, and the responsiveness of gel release was positively correlated with the ATP concentration (Supplementary Fig. 13). Notably, the addition of various ATP concentrations to the solution post gel formation yielded distinct fluorescence intensity variations, and higher concentrations of ATP accelerated the release of CpG ODN (Fig. 5i, j). Co-incubation of ATP-responsive hydrogel in artificial cerebrospinal fluid simulating the in vivo environment revealed that the degradation was related to the ATP concentration, with the higher the concentration the faster rate of degradation (Supplementary Fig. 14). Furthermore, investigations involving coincubation of gels and macrophages revealed that solutions containing CpG ODN induced the upregulation of TLR4, TLR5, and TLR9 (Supplementary Fig. 15). In conclusion, our synthesized gel effectively promotes the progressive release of CpG ODN in response to the corresponding ATP release triggered by pyroptosis, and the gel can activate the innate immune response.

Cavitary injection of IASNDS@gel to prevent tumor recurrence

The efficacy of the antitumor intervention in vivo was evaluated utilizing the GL261 syngeneic orthotopic mouse model. Ten days after inoculation of a Luci+GL261 cell suspension, tumor-bearing mice were subjected to random grouping and received various treatments after surgical resection (Fig. 6a). To establish the GBM excision mouse model, a portion of the intracranial GBM in mice was removed with forceps under a microscope on day 10, and then the tumor was suctioned by using a suction device to form a cavity that could accommodate 30 μL volume, hydrogel with different formulations were subsequently injected into the cavity (Supplementary Figs. 16 and 17). During surgical resection, distinct formulations, including Ctrl (PBS), Gel, SLIN@gel, SDV@gel, and IASNDS@gel, were administered to the respective groups (Fig. 6b).

Fig. 6: Suppression of postoperative recurrence and prolonged survival via intracavity injection of IASNDS@gel.
figure 6

a Illustration of the in vivo experimental design. b Upon reaching day 10 after initial inoculation, tumor-bearing GL261 mice underwent surgical tumor resection, followed by hydrogel injection. Subsequently, brain tissues from the excised mice were subjected to euthanasia. H&E (c) and immunofluorescence (d) staining was carried out to visualize residual GBM cells (labeled red with a Vimentin primary antibody). (n = 3 independent experiments). e In vivo bioluminescence images were captured (n = 3), and f the resulting signal intensity was quantified (n = 10). (exact P values: SLIN@gel vs. SDV@gel P = 1.05777E-09, SDV@gel vs. IASNDS@gel P = 6.4922E−11); ****P < 0.0001. g Mouse survival was assessed through Kaplan‒Meier survival curves for each treatment group (n = 10). Data analysis was performed employing the log-rank (Mantel‒Cox) test. (exact P values: Ctrl vs. Gel P = 2.55345E-08, SLIN@gel vs. SDV@gel P = 1.60341E−08, SDV@gel vs. IASNDS@gel P =  3.26493E−05); ****P < 0.0001. h The percentage of positive cells was tallied following immunohistochemical staining for Ki-67 in tumor tissues originating from various mouse groups (n = 5 images from three independent experiments). (exact P values: Ctrl vs. Gel P = 8.876E−05, SLIN@gel vs. SDV@gel P =  7.74109E−09); ****P < 0.0001. i Visualization of alterations in mouse brain tissues was achieved by applying WGA staining (red), enabling observation of tumor cell dynamics across the different treatment groups in mice (n = 3 independent experiments). Data are presented as the mean ± S.D. The statistical comparisons in f and h were performed using two-way ANOVA and one-way ANOVA by Tukey’s post hoc test, with asterisks indicating significant differences (***P < 0.001). The images in c, d, and i are representative images of the corresponding independent biological samples. The scale bar in c is 1.5 mm, and the scale bar in d and i is 50 μm. Source data are provided as a Source Data file.

Evaluation of the brain by hematoxylin and eosin (H&E) staining and immunofluorescence after surgery revealed an infiltration of GBM cells into the postsurgical parenchyma and a distribution of these cells around the margins of the cavity, with limited intracranial dissemination (Fig. 6c, d). Dynamic monitoring of tumor volume via bioluminescence imaging (BLI) of Luci+GL261 cells with various treatment intervals underscored the significant inhibition of tumor volume growth upon IASNDS@gel treatment (Fig. 6e, f). Without Gel, SDV or IASNDS alone did not demonstrate significant GBM suppression, primarily because drugs injected locally tend to be rapidly cleared with cerebrospinal fluid circulation. These findings imply that there is a substantial preventative effect of intracavitary injection of IASNDS@gel on GBM recurrence.

Comparative analysis revealed that the gel-alone treatment notably extended the survival of mice (by 2.14-fold) relative to the Ctrl group (Fig. 6g). However, the difference in survival time between the Gel and SLIN@gel groups was not statistically significant. In contrast, the survival time of mice receiving SDV@gel treatment was a remarkable 4.46 times longer than that of the Ctrl group. After up to 120 days of follow-up, 80% of the mice remained alive in the IASNDS@gel group. Immunohistochemical staining of tumor tissues after 30 days of treatment showed Ki-67 expression upon gel-based intervention, with the lowest tissue expression observed in the SDV@gel and IASNDS@gel groups (Fig. 6h, Supplementary Fig. 18). Further fluorescence staining of tumor tissues revealed a narrower distribution and increased vacuole formation in the IASNDS@gel group. These changes and the observed softening of the tumor texture seem to correlate with pyroptosis-mediated tumor cell death (Fig. 6i). These alterations may facilitate the infiltration of immune cells, consequently reshaping the immune microenvironment of GBM and potentially bolstering the immunotherapeutic effect.

Mechanisms of anti-GBM effects in vivo

Next, we explore the anti-GBM mechanisms of different therapeutic agents. GBM-bearing mice were euthanized, and tumor tissues were isolated on the 20th day (Fig. 7a). Notably, the excised tumor tissues from mice in the gel treatment group exhibited a marked increase in the number of infiltrated immune cells compared to that in the Ctrl group, suggesting a role of the gel in immune cell recruitment (Fig. 7b, Supplementary Fig. 19). In the IASNDS@gel group, the abundance of CD4+ T cells was significantly increased, with values 6.62, 4.05, 2.95 and 1.87 times higher than those in the Ctrl (PBS), Gel, SLIN@gel, and SDV@gel groups, respectively (Fig. 7c). Encouragingly, CD8+ T cells also showed the same trend, with a notable increase in the number of CD8+Granzyme B+ T cells (Fig. 7d, e, Supplementary Fig. 20). Remarkably, the IASNDS@gel group exhibited a substantial proportion (71.6%) of Granzyme B+ T cells among CD8+ T cells. These findings suggest that IASNDS@gel plays a pivotal role in initiating antitumor immune responses.

Fig. 7: Immunomodulatory efficacy of the Salmonella-loaded hydrogel system for mobilizing immunity against tumors via intracavity injection.
figure 7

a Schematic of the experimental design for assessing immunomodulatory efficacy. b Flow cytometry results showing CD3+ cells in brain tissues after GBM removal in each treatment group. c Quantitative analysis of CD3+CD4+ cytotoxic T cells (n = 6 independent experiments). (exact P value: SDV@gel vs. IASNDS@gel P =  1.57104E-08); ****P < 0.0001. d Quantitative analysis of CD3+CD8+ cytotoxic T cells (n = 6 independent experiments). (exact P value: SDV@gel vs. IASNDS@gel P = 1.32267E−10); ****P < 0.0001. e Flow cytometry results indicating CD3+CD8+Granzyme B+ T cells in GBM tissues from different treatment groups. f, g Flow cytometry analysis and statistical analysis of CD80+CD86+ cells in GBM tissues after different treatments (n = 6 independent experiments). h Heatmap displaying the expression profiles of pyroptosis-related proteins, cytokines, and chemokines in brain tumor tissues. (n = 3 independent experiments). i, j Statistical analysis of IFN-γ expression and TNF-α expression in GBM tissues under different treatment conditions (n = 6 independent experiments). (exact P values of i: SLIN@gel vs. SDV@gel P =  1.22239E−07, SDV@gel vs. IASNDS@gel P =  8.7998E−07; exact P values of j: SLIN@gel vs. SDV@gel P =  2.02309E−07; SDV@gel vs. IASNDS@gel P = 3.4876E−07); ****P < 0.0001. Data are presented as the mean ± S.D. The statistical comparisons in c, d, g, i and j were performed using one-way ANOVA with Tukey’s post hoc test, with asterisks indicating significant differences (ns = no significance, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001). Source data are provided as a Source Data file.

As mediators of innate and adaptive immunity, antigen-presenting cells (APCs) play a pivotal role in tumor immunotherapy39,40. As revealed in Fig. 7f and g, the IASNDS@gel group exhibited the highest number of CD80+CD86+ cells, surpassing the Ctrl (PBS), Gel, SLIN@gel, and SDV@gel groups by 5.68, 2.67, 2.22, and 1.43 folds, respectively. Further analysis revealed a significant increase in M1 macrophages (CD80+F4/80+) and a significant decrease in M2 macrophages (CD206+F4/80+) in the IASNDS@gel group (Supplementary Fig. 21).

Further reinforcing these findings, the expression levels of cytokines associated with innate and adaptive immunity were enhanced within the GBM tissues of mice receiving IASNDS@gel treatment, and the levels of interferon γ (IFN-γ) and tumor necrosis factor α (TNF-α) were particularly elevated (Fig. 7h–j). In particular, compared to the Ctrl group, CD8+IFNγ+ T cells were significantly higher in the IASNDS@gel treatment group, suggesting a more pronounced anti-tumor immune response within the GBM (Supplementary Fig. 22).

Furthermore, we investigated the effect of ATP-responsive gel on the effect of activated immunization, and the non-ATP-responsive IASNDS@gel immunization was significantly less vigorous than the ATP-responsive IASNDS@gel (Supplementary Fig. 23). On further analysis, the activation and increase of immune memory cells may play a very important role (Supplementary Fig. 24). We further re-challenged mice originally bearing GBM in the right side, and found that when the mice were treated with IASNDS@ gel for 20 days and then transplanted with GBM cells to the left side, the IASNDS@ gel treatment significantly inhibited the growth of GBM, and the survival time of the mice was also prolonged (Supplementary Fig. 25).

These findings demonstrate that the introduction of the autolysing bacterium-hydrogel system into the tumor cavity of GBM mice after surgery increased the frequency of responding immune cells. This orchestrated the establishment of a comprehensive tumor clearance system involving cytokines, immune cells, and the immune microenvironment. The synergistic response generated by this bacteriotherapy effectively inhibits the relapse of GBM.