Metabolically hijacked Listeria monocytogenes synergizes STING activation and dual cell death for enhanced antitumor immunotherapy

Abstract

Background

Listeria monocytogenes (LM) holds promise as a microbial vector for cancer immunotherapy, yet its clinical translation requires precise engineering to achieve an optimal balance between attenuation and immunogenicity. This study aimed to develop an engineered LM strain that induces metabolic reprogramming while enhancing antitumor efficacy through multimodal immune activation.

Results

We encapsulated cepharanthine nanoparticles (CEP NPs) into a double-gene attenuated LM strain via electroporation, resulting in the induction of metabolic reprogramming and the emergence of a ‘zombie-like’ phenotypic state (LM^e@CEP). This modification disrupted folate metabolism to induce bacterial metabolic dormancy, while triggering a cyclic dinucleotide (CDN) surge that amplified STING-mediated immunogenicity. The CEP payload localized to tumors and exerted dual cytotoxic effects by simultaneously inducing ferroptosis and apoptosis in situ. These direct tumoricidal effects synergize with the immunostimulatory capacity of LM^e@CEP, effectively modulating the tumor immune microenvironment and exerting robust antitumor effects. Preclinical validation across multiple murine tumor models and ex vivo human tumor specimens confirmed the therapeutic versatility of the platform. Furthermore, results in STING^−/− mice corroborated that LM^e@CEP’s antitumor effectiveness partly depends on STING pathway activation, and its therapeutic potential may be further enhanced by PD-L1 blockade, thereby mitigating STING-driven immune suppression.

Conclusions

These findings establish metabolic hijacking of bacterial vectors as a paradigm-shifting strategy that integrates direct tumor killing with multimodal immune activation, overcoming key barriers in microbial immunotherapy.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12951-025-03945-3.

Introduction

Immunotherapy has indeed transformed the landscape of cancer treatment; however, it is still hampered by challenges such as suboptimal response rates, the emergence of resistance, associated toxicities, and prohibitive costs [1, 2]. These limitations underscore the urgent need for the development of novel, innovative strategies to address these issues.

In recent years, as research on the use of bacteria in the field of immunotherapy has continued to expand, many studies have shown that bacterial therapy can stimulate the immune system and amplify immune effects, thus leading to the elimination of distant tumor tissues and preventing cancer recurrence [3, 4]. Although bacterium-mediated immunotherapy has not been widely used in clinical practice, animal experiments have demonstrated that bacterial therapy can effectively induce tumor regression and even cure tumors [5, 6].

Several types of parthenogenetic or specialized anaerobic bacteria, such as Salmonella, Bifidobacterium, Clostridium, Listeria, and Escherichia coli, can intrinsically target tumors and induce pathogenicity [7]. One of these bacteria, Listeria monocytogenes (LM), is a gram-positive bacterium that is widely known for its ability to infect humans and produce a variety of symptoms, including gastroenteritis, meningitis and encephalitis. LM has several advantages in tumor immunotherapy [8]. First, LM can induce strong innate and adaptive immune responses. In tumor therapy, the immune response is very important because it can recognize and attack tumor cells. LM can enhance the immune response by activating host immune cells, such as macrophages and T lymphocytes, thereby improving the efficacy of tumor therapy [9, 10]. In addition, LM can promote the killing of tumor cells by inducing the production of cytotoxins such as tumor necrosis factor (TNF) and interferon (IFN). These cytokines can stimulate other immune cells, such as natural killer (NK) cells, which can further expand the scope of tumor destruction [11]. Importantly, LM infection triggers only a modest humoral immune response, with the possibility of reinfection, which provides favorable conditions for the repeated use of LM-based immunotherapy to achieve enhanced cytotoxic T lymphocyte responses [12].

LM is a widely used bacterium with strong immunotherapeutic potential; however, although its attenuation is widely recognized, numerous issues remain. Currently, the main strategies for reducing Listeria virulence are based on deletion of virulence or metabolic genes and subsequent backfilling, which has been demonstrated to result in biosafety issues while also achieving a significant reduction in virulence [13–15]. Therefore, balancing the need to reduce Listeria virulence with the need to increase its immunomodulatory ability in the tumor immune microenvironment has become a key obstacle to its clinical translation [16].

Cepharanthine, a bisbenzylisoquinoline alkaloid derived from Stephania cephalantha Hayata, has been clinically employed in Asia for over 70 years, with an established safety profile devoid of serious adverse effects. Pharmacological studies reveal its multi-target properties, including antitumor, anti-inflammatory, immunomodulatory, and antimicrobial activities [17]. The compound exhibits particularly promising antitumor efficacy by inducing apoptosis to inhibit tumor growth, increasing sensitivity to chemo and radiotherapy, suppressing metastasis, and boosting antitumor immunity [18]. Notably, cepharanthine exhibits significant antibacterial activity against intracellular pathogens such as Mycobacterium tuberculosis [19], although its exact antimicrobial mechanisms have yet to be fully elucidated. Given this multifaceted pharmacological profile, combining broad-spectrum antitumor activity, immunomodulatory capacity, and favorable clinical safety, cepharanthine represents a compelling candidate for development as a novel multifunctional antitumor agent.

Here, ultrasmall cepharanthine nanoparticles (CEP NPs) were loaded into LM via electroporation, with the objective of reducing the virulence of the bacteria by interfering with their growth and metabolic processes, ultimately results in the induction of metabolic reprogramming, producing a Listeria phenotype reminiscent of a “zombie-like” state. Specifically, internalized CEP NPs interfered with LM folate metabolism, resulting in elevated cyclic dinucleotides (CDNs) levels, which significantly enhanced STING-dependent antitumor immune responses. Notably, CEP NPs possess antitumor activity, and their release in situ synergized with the immunomodulatory effects of zombie-like LM, thereby enhancing the antitumor effects of the treatment(Scheme 1). The developed oncolytic zombified bacteria strategy has the potential to markedly enhance current engineering approaches for using bacteria as agents in cancer immunotherapy. This innovative method could boost the effectiveness of immunotherapies, minimize additional toxicity and the risk of symptomatic infections, and introduce new mechanisms of action that may surpass the constraints of traditional cancer treatments.

Scheme 1.

Schematic Illustration of Zombie-like Listeria monocytogenes (LM^e@CEP) for Synergistic Antitumor Therapy

Results

Preparation and characterization of LM^e@CEP

To use Listeria monocytogenes as a carrier for antitumor drug delivery, we first attempted to transform cepharanthine, a potential natural antitumor agent, into LM by the technique of electrotransformation(Fig. 1a). CEP nanoparticles (NPs) were obtained via a droplet-confined/cryodesiccation-driven crystallization method. Representative transmission electron microscopy (TEM) images revealed that the CEP NPs were spherical particles with a diameter of 20–40 nm that were suitable in size for loading into LM (Fig. 1b). The drug loading of LM^e@CEP was 0.46%. CEP in LM^e@CEP exhibited faster release in a medium with a pH of 6.5 compared to a release medium with a pH of 7.4, indicating that acidic conditions accelerate CEP release (Fig. S1). An investigation of the effects of CEP NP loading into LM (LM^e@CEP) on the morphology of LM revealed that while the bacterial cell wall boundaries of LM^e@CEP became blurred, the integrity of the cells was preserved (Fig. 1c). To track the intracellular localization of CEP NPs, we pre-labeled them with 5-nm dodecylamine-coated gold nanoparticles (Au NPs) as a definitive tracer and introduced them into LM using electroporation (Fig. 1d, e). TEM and EDS analysis revealed the presence of Au signals inside the bacterial cells (Fig. 1f). This observation confirmed the successful internalization of the CEP NPs in LM. To gain further insight into the impact of CEP NP loading on LM viability, we loaded CEP NPs into LM at various concentrations. CEP NPs at a concentration of 1.25 mg/ml significantly inhibited LM activity, and the observed effect was concentration-dependent (1.25–5.25 mg/ml) (Fig. S2a& b). We also examined the integrity of the cell membrane using SYTOX green fluorescent dye. Compared with those of LM^e, the SYTOX green fluorescence signals of LM^e@CEP were nearly two orders of magnitude greater (Fig. 1g), indicating a potential alteration in cell membrane permeability induced by CEP NPs. Thiazole orange staining analysis also indicated that loading CEP NPs into LM via electrotransfer resulted in a significant decrease in bacterial membrane integrity (Fig. 1h). In addition, the occurrence of ferroptosis-like cell death has been confirmed in microbes [20]. We thus evaluated the level of lipid peroxidation (LipROS) in the bacteria. Compared with that in the LM^e group, the level of LipROS in the LM^e@CEP group was significantly greater (p < 0.0001) (Fig. 1i). These changes may be the result of direct effects of CEP NPs on the structure of the bacterial cell wall and cell membrane; alternatively, they may be indirect effects resulting from the intracellular metabolic changes they trigger.

Fig. 1.

Preparation and characterization of zombified Listeria monocytogenes. (a) Schematic representation of Listeria monocytogenes (LM) loaded with CEP NPs by electroporation. (b) TEM images of CEP NPs. Scale bars: 200 nm (left) and 10 nm (right). (c) Scanning electron microscopy (top) and TEM (bottom) images of LM^e and LM^e@CEP. Scale bars: 500 nm (top) and 200 nm ((bottom)). (d) Schematic presentation of Au-labeled CEP NPs. (e) TEM and energy-dispersive X-ray spectroscopy (EDS) mapping images of Au-labeled CEP NPs. (f) TEM-EDS elemental mapping images of LM^e@Au-labeled CEP. (g-i) Representative flow cytometry histograms (right) and quantitative comparison (left) of SYTOX Green (g), Thiazole Orange (h) and LipROS (i) staining in LM^e and LM^e@CEP (n = 3). (j) Adhesion and invasion assays to compare the infectivity of LM^e and LM^e@CEP on B16F10 monolayers in vitro (n = 3). (k) Bacterial density (CFU/g) of excised organs 24 h after the intratumoral administration of LM^e or LM^e@CEP (n = 3). (l) Representative immunofluorescence images showing the localization of both LM^e and LM^e@CEP within tumor tissues from B16F10 tumor-bearing mice. Nuclei are stained blue, bacteria red, and macrophages green. Scale bars: 50 μm. The data are presented as the means ± SEMs; P values were determined via one-way ANOVA with Tukey’s multiple comparisons test

LM is a gram-positive intracellular pathogen that is able to invade and multiply within host cells and is distributed in various host tissues. Next, we investigated the attachment and invasion of LM^e@CEP in tumor cells. After electroporation, LM^e could still effectively colonize tumor cells, while both the adhesion and invasion abilities of the bacteria were obviously attenuated in LM^e@CEP (Fig. 1j and Fig. S3). To further verify the bacterial distribution and the bacterial burden in the major organs (the heart, liver, spleen, lungs, kidneys) and tumor tissue, LM^e and LM^e@CEP were locally injected into tumors, and the numbers of LM cells in different tissues were quantified at 24 h. As shown in Fig. 1k, LM was primarily found in the spleen and at tumor sites, while no bacteria were observed in the other organs. Compared with that in the LM^e group, LM^e@CEP colonization in the spleen and tumors was significantly lower (Fig. S4). These results suggest that there is a lower risk of LM infection after LM^e@CEP treatment than after LM or LM^e treatment. Finally, the intratumoral distribution of LM 24 h after injection was investigated by observing tumor tissue via immunofluorescence staining. As shown in Fig. 1l, some bacteria were engulfed by tumor-associated macrophages (TAMs). Nevertheless, a considerable proportion of persisted outside of phagocytic cells, and several bacteria may have entered tumor cells. Additionally, compared with those in the LM^e group, larger necrotic areas were observed in tumors in the LM^e@CEP group. These results implied that electroporation of CEP NPs into live-attenuated LM reduced the risk of systemic bacterial infection and resulted in remarkable antitumor potency.

Perturbation of bacterial metabolism via electroporation of CEP NPs

To further elucidate the effects of CEP NPs on bacterial metabolism and explore the possible underlying mechanism, UPLC-ESI-QTOFMS analysis coupled with multivariate data analysis was employed to analyze changes in bacterial metabolites. Principal component analysis (PCA) was further applied to identify differentially abundant metabolites between the treatment and control groups (Fig. S5a). Forty-four differentially abundant metabolites between the control and CEP-treated groups were identified (Table S1), suggesting that some crucial metabolic pathways involved in bacterial metabolism are altered after LM loading with CEP NPs. VIP (variable important in projection) analysis was performed and identified 14 metabolites significantly contributing to classification (Fig. S5b). Among these 14 significant metabolites, Pro-Trp, 2,4,6-tri-tert-butylaniline, and N-myristoylsphinganine were the top three markers contributing to the metabolic changes induced by the electroporation of CEP NPs in LM. Figure 2a depicts the changes in the abundance of each potential biomarker. Noticeable separation was observed between the LM^e@CEP and LM^e groups, with samples within the same group exhibiting clustering tendencies. On the basis of FDR < 0.05 and log2 FC > 1.5, four metabolites exhibited significantly decreased abundance in LM^e@CEP, whereas eleven metabolites exhibited significantly increased abundance. Next, Pearson correlation coefficients were used to identify associations between differentially abundant metabolites; R > 0.5 signified a positive correlation, whereas R<−0.5 indicated a significant negative correlation (Fig. S5c). Pathway analysis of biomarkers was conducted using MetaboAnalyst to reveal the biological impacts revealed by these metabolomic data. The results suggested that loading of LM with CEP NPs may influence nutrient uptake by LM, metabolic activity and ultimately viability, primarily mediated by changes in histidine metabolism; β-alanine, valine, leucine and isoleucine degradation; pyruvate metabolism; alanine, aspartate and glutamate metabolism; fatty acid degradation; and fatty acid elongation (Fig. 2b). These results suggest that electroporation of CEP NPs might drive LM to enter a “zombie” state.

Fig. 2.

Effects of electroporation of CEP NPs on bacterial metabolism. a Heatmap analysis displaying the abundance of significant metabolites across all clustered samples. b Metabolic pathways affected by the electroporation of CEP NPs in LM^e. c Molecular docking between DHFR (top), Thy (lower) and CEP NPs. d Quantification of intracellular c-di-AMP production in LM treated with CEP NPs by LC-MS/MS (n = 3). e-g Intracellular tetrahydrofolate (e, n = 5), dihydrofolate (f, n = 5) and 5-methyl-THF concentrations (g, n = 5). h Intracellular c-di-AMP concentration in LM^e@CEP after thymidine supplementation quantified by LC-MS/MS (n = 3). i Schematic diagram of thymidine metabolism in Listeria monocytogenes. ThyA, thymidylate synthase; DHFR, dihydrofolate reductase; THF, tetrahydrofolate. The data are presented as the means ± SEMs; P values were determined via one-way ANOVA with Tukey’s multiple comparisons test

Moreover, changes in bacterial metabolic pathways could lead to the production of different metabolic products. Histidine metabolism and β alanine have been implicated in bacterial virulence and the production of the bacterial second messenger cyclic di-AMP (c-di-AMP), which controls a wide range of cellular processes (Fig. S5d) [21]. In addition, disturbances in bacterial folate metabolism may be involved in the production of c-di-AMP [22]. To explore whether CEP NPs were able to directly influence critical enzymes in the folate metabolism pathway, molecular docking was carried out to predict the potential binding mode of CEP NPs with dihydrofolate reductase (DHFR) and thymidylate synthase (ThyA) using Discovery Studio. The docking results (Fig. 2c) revealed specific binding sites for CEP on both DHFR and ThyA, with the lowest binding energies recorded at −7.78 kcal/mol and − 8.27 kcal/mol, respectively. Thus, CEP NPs may affect the balance of folate metabolism and ultimately contribute to bacterial intracellular c-di-AMP accumulation. To confirm this hypothesis, we detected c-di-AMP levels in LM treated with CEP NPs. Significantly greater production of c-di-AMP was observed following CEP NP treatment (P < 0.0001) (Fig. 2d). Consistently, CEP NP-loaded LM also exhibited increased cyclic di-GMP (c-di-GMP) levels (Fig. S6). To further investigate the main folate substrates and folate products associated with folate metabolism, the concentrations of tetrahydrofolate, dihydrofolate and 5-methyl-THF were quantified. The levels of tetrahydrofolate and 5-methyl-THF in LM loaded with CEP NPs were significantly lower than those in LM^e (Fig. 2e-g). The enzyme dihydrofolate reductase (DHFR) plays a crucial role in the folate regeneration required for de novo thymidine synthesis. Therefore, we quantified the c-di-AMP levels of LM^e@CEP in the presence and absence of exogenous thymidine. After thymidine supplementation, the c-di-AMP level significantly decreased (Fig. 2h). These observations implied that CEP NP treatment inhibited thymidylate synthase (ThyA) or DHFR activity, thus disrupting thymidine biosynthesis and ultimately promoting bacterial intracellular c-di-AMP accumulation (Fig. 2i). These findings unveil the mechanism of CEP NPs, which disrupt bacterial folate metabolism via DHFR/ThyA inhibition to amplify bacterial cyclic dinucleotide levels, strategically engaging the STING pathway.

Potent antitumor effects of LM^e@CEP

To evaluate the therapeutic efficacy of intratumorally administered LM^e@CEP, a subcutaneous B16F10 melanoma cancer model and a 4T1-Luc murine breast cancer model were established as described in the Experimental Methods section (Fig. 3a). We evaluated the tumor suppression effects of LM^e@CEP compared to its individual components, LM^e and CEP NPs, in the treatment of B16F10 melanoma-bearing mice. Compared with CEP NPs (50 µg) and LM^e [5 × 10⁷ colony-forming units (CFUs)], LM^e@CEP (50 µg of CEP NPs with 5 × 10⁷ bacterial cells) presented the strongest tumor inhibition effect, and treatment with LM^e@CEP resulted in significant reductions in tumor growth and weight (Fig. 3b-d). In addition, compared with the saline and CEP NP treatments, the LM^e and LM^e@CEP treatments significantly improved the survival time of B16F10 tumor-bearing mice (Fig. 3e). However, during the treatment cycle, the weight of the mice in the LM^e group significantly decreased, whereas the weights of the mice in the saline-, CEP NP- and LM^e@CEP-treated groups did not significantly change, suggesting that the use of CEP NPs in zombified LM could significantly improve therapeutic biosafety (Fig. 3f). Meanwhile, histological examination using H&E staining of principal organs, such as the heart, liver, spleen, lungs, and kidneys, revealed no significant histopathological abnormalities(Fig. S7). Furthermore, critical markers of hepatic and renal function, including alanine aminotransferase (ALT), aspartate aminotransferase (AST), total bilirubin (TBIL), albumin (ALB), urea, creatinine (CREA), and uric acid (UA), remained within normal physiological ranges post-treatment (Fig. S8). Collectively, these findings underscore the favorable biosafety profile of LM^e@CEP, reinforcing its potential as a promising anticancer agent. Moreover, to assess immune toxicity and systemic inflammation after repeated intratumoral administration, serum samples were analyzed for inflammation markers like TNF-α, IL-6 and complete blood count (CBC) after treatments. The LM^e@CEP group showed low systemic inflammation, with even lower TNF-α and IL-6 levels compared to the LM^e-alone group, indicating a safer profile with less off-target inflammation(Fig. S9). No significant CBC abnormalities were noted (Table. S2). These findings suggest that CEP nanoparticle encapsulation in LM reduces systemic immune activation, enhancing biocompatibility and clinical potential. Furthermore, our findings indicate that the biodistribution profile of LM^e@CEP, even when delivered through intraperitoneal injection, significantly diverges from that of the LM^e group. While some bacterial colonization was observed in the liver and spleen at the 24 h, the bacterial load was considerably reduced in comparison to LM^e. By 72 h, no viable LM were detectable in any organs of the LM^e@CEP-treated mice (Fig. S10). This rapid clearance implies that CEP loading significantly impairs the pathogen’s capacity to establish a sustained systemic infection. To further investigate antitumor activity, sectioned tumors were stained with H&E, TUNEL and Ki67 to assess cell proliferation and apoptosis. Cell necrosis, apoptosis and anti-proliferative effects were most pronounced in the LM^e@CEP group. Proliferation analysis revealed that LM^e@CEP significantly increased the apoptosis rate compared with that of the CEP NP- and LM^e-treated groups (Fig. 3g-i). To investigate the broad spectrum of tumor inhibition effects in LM^e@CEP, we also tested its antitumor effect in a 4T1-Luc BALB/c breast cancer mouse tumor model. We observed that tumor growth was significantly delayed in the mice treated with CEP NPs, LM^e or LM^e@CEP, whereas tumors grew at a significantly greater rate in the saline-treated mice (Fig. 3j). Consistent with the results obtained in the B16F10 mouse melanoma model, LM^e@CEP significantly inhibited 4T1-Luc tumor growth compared with that in the CEP NP and LM^e groups (Fig. 3k, l). The weights of the tumors in the LM^e@CEP group were lower than those in the other groups (Fig. 3m). The survival of LM^e@CEP-treated 4T1-Luc tumor-bearing mice was significantly greater than that of the mice in the other groups (Fig. 3n). Taken together, these results substantiate the potential of LM^e@CEP as a promising modality for cancer therapy, highlighting its superior antitumor efficacy relative to that of free LM^e or CEP NPs.

Fig. 3.

LM^e@CEP for in vivo treatment across various tumor models. a Schematic representation of the therapeutic strategy using LM^e@CEP during B16F10 or 4T1-Luc tumor challenge. b Individual growth curves of B16F10 tumors in mice subjected to the indicated treatments (n = 5). c, Photographs of B16F10 tumors from the four groups (n = 5). d, e Tumor weights (d, n = 5) and survival curves (e, n = 5) of C57BL/6 mice receiving the indicated treatments. f Body weights of C57BL/6 mice during the indicated treatment periods (n = 5). g-i Representative images of H&E g, TUNEL immunofluorescence h and Ki67 immunohistochemistry (i) staining of B16F10 tumors subjected to the indicated treatments. Scale bar: 50–100 μm. j Bioluminescence images of 4T1-Luc model mice on days 7, 14 and 21 (n = 5). k Individual growth curves of 4T1-Luc tumors subjected to the indicated treatments (n = 5). l Photographs of 4T1-Luc tumors from the four groups (n = 5). m, n Tumor weights (m, n = 5) and survival curves (n, n = 5) of BALB/c mice subjected to the indicated treatments. The data are presented as the means ± SEMs; P values were determined via one-way ANOVA with Tukey’s multiple comparisons test

LM^e@CEP orchestrates a robust antitumor immune response

To systematically characterize the antitumour mechanisms of the developed zombified bacteria, mice bearing B16F10 tumors were intratumorally injected with CEP NPs, LM^e or LM^e@CEP, and the spleens, tumors and tumor-draining lymph nodes (TDLNs) were harvested for flow cytometry-based measurement of immune cells. A marked enlargement of TDLNs on the injected side was observed (Fig. S11), and the spleens in the treated groups were significantly larger than those in the saline group (Fig. S12), suggesting that a systemic immune response occurred in these lymphatic organs. Mature dendritic cells (DCs) are pivotal in orchestrating the antitumor immune response, primarily due to their ability to cross-present tumor antigens [23]. Compared with saline treatment, intratumoral treatment with CEP NPs, LM^e or LM^e@CEP significantly increased the frequency of mature DCs (CD11c⁺CD86⁺CD80⁺) in tumors. Compared with mice subjected to the other treatments, LM^e@CEP-treated mice presented greater proportions of mature DCs in their tumors (Fig. 4a). This increase was accompanied by a significant increase in the level of surface MHC class I (Fig. S13). In addition, to track the spatiotemporal distribution of LM^e@CEP following intratumoral injection. In vivo fluorescence imaging was conducted at 2 h and 8 h post-injection. The results revealed strong fluorescence signals at the primary tumor site at 2 h, indicating substantial local accumulation of the bacteria. Importantly, significant fluorescence was also detected in the TDLNs as early as 2 h, demonstrating active trafficking. At the 8 h time point, while the fluorescence intensity at the tumor site had decreased, a clear signal persisted within the TDLNs(Figure S14). These findings implied that the administration of LM^e@CEP may enhance the cross-presentation of tumor-associated antigens in TDLNs and facilitate the recognition of CD3⁺CD8⁺ T cells by tumor cells. We then analyzed the adaptive immune responses of CD8⁺ and CD4⁺ T cells. Notably, local administration of LM^e@CEP increased the recruitment of CD8⁺ T cells to the tumor and draining lymph nodes and increased the ratio of cytolytic CD8⁺ to helper CD4⁺ T cells (Fig. 4b, Fig. S15). Moreover, compared with CEP NP or LM^e treatment, LM^e@CEP treatment significantly increased the levels of the cytokines IFN-γ and TNF-α secreted by CD8⁺ and CD4⁺ T cells in the TDLNs (Fig. 4b-d; Fig. S16). These data suggest that intratumor LM^e@CEP treatment induces DC maturation and promotes T-cell activation, facilitating antitumor T-cell immunity.

Fig. 4.

LM^e@CEP activates both adaptive and innate immune responses. a Representative flow cytometry analysis (left) and the percentage (right) of mature DCs (CD80⁺ CD86⁺ gated on CD11c⁺) in B16F10 tumor-draining lymph nodes (TDLNs) (n = 5). b Representative flow cytometry analysis of CD4 and CD8 staining (left) and the ratio of CD8/CD4 (right) T cells in the TDLNs (n = 5). c Representative flow cytometry analysis of intracellular IFN-γ staining of CD8⁺ T cells (left) and the percentage of stained cells (right) in the TDLNs (n = 5). d Representative flow cytometry analysis of intracellular TNF-α staining of CD8⁺ T cells (left) and the percentage of stained cells (right) in the TDLNs (n = 5). e Representative flow cytometry analysis of Tregs (CD4⁺CD25⁺Foxp3⁺) (left) and quantification of Tregs (right) in the TDLNs (n = 5). f Representative flow cytometry analysis (left) and the percentage (right) of macrophages (CD45⁺CD11b⁺CD11c⁺ F4/80⁺) in tumors from B16F10 tumor-bearing mice (n = 5). g Representative immunofluorescence images demonstrating F4/80 and CD206 expression in tumors from B16F10 tumor-bearing mice (green indicates F4/80-positive staining, red indicates CD206-positive staining). Scale bars: 50 μm. The data are presented as the means ± SEMs; P values were determined via one-way ANOVA with Tukey’s multiple comparisons test

Since CD3⁺CD4⁺CD25⁺FOPX3⁺T regulatory cells (Tregs), M2-type tumor-associated macrophages (TAMs), and myeloid-derived suppressor cells (MDSCs) are known to be major immunosuppressive cell types in the tumor microenvironment that strongly suppress antitumor T-cell immunity [24–26], we also analyzed the proportions of Tregs, TAMs and MDSCs after the administration of different treatments. Compared with untreated mice and mice treated with CEP NPs or LM^e alone, mice treated with LM^e@CEP presented the lowest proportion of regulatory T cells in the TDLNs and spleen (Fig. 4e, Fig. S17). Notably, we observed a decrease in the frequency of CD11b⁺F4/80⁺ macrophages within the tumor microenvironment and TDLNs following intratumoral administration of LM^e@CEP (Fig. 4f, Fig. S18), whereas the proportion of M2-like TAMs (CD206⁺CD11b⁺F4/80⁺) obviously decreased (Fig. 4g, Fig. S19). Further analysis of MDSCs revealed that the proportion of MDSC cells (CD11b⁺Ly6G⁻Ly6C⁺) was profoundly lower in tumors and TDLNs from mice injected with LM^e@CEP than in those from mice treated with CEP NPs or LM^e alone, whereas the proportion of polymorphonuclear myeloid-derived suppressor cells (PMN-MDSCs, CD11b⁺Ly6G⁺Ly6C⁻) did not appreciably differ across treatments (Fig. S20 &21). Taken together, the above results suggest that zombified bacteria are able to reprogram the tumor microenvironment by activating both innate and adaptive immunity while simultaneously modulating immunosuppressive cells. These effects synergistically contribute to the remarkable antitumor performance of the zombified bacteria.

The antitumor activity of LM^e@CEP is partially mediated by STING activation

As mentioned above, electroporation of CEP NPs obviously promotes bacterial accumulation of intracellular c-di-AMP and c-di-GMP, which are known to be natural stimulator of interferon genes (STING) agonists. In immunogenic tumors, the STING pathway enhances the activation and maturation of DCs, which are pivotal for T-cell priming and the subsequent adaptive immune response [27]. Therefore, to test whether LM^e@CEP promotes the activation and functional regulation of antigen-presenting cells by activating the STING pathway, we coincubated LM^e@CEP with BMDCs in vitro. Compared with CEP NPs or LM^e alone, LM^e@CEP resulted in greater efficiency in inducing phenotypic maturation of BMDCs (CD11c⁺ CD80⁺ CD86⁺, CD11c⁺ CD80⁺ MHC-II⁺, and CD11c⁺ CD86⁺ MHC-II⁺) (Fig. 5a, b), as indicated by the significant upregulation of the CD80, CD86 and MHC class II molecules (Fig. 5c, d, e). Moreover, activation of the STING pathway promotes the expression of interferon genes, thereby inducing interferon synthesis and release. Consistent with this idea, the secretion of IFN-α, IFN-β and the related inflammatory cytokines TNF-α and IL-6 was significantly increased in the supernatant of BMDCs after LM^e@CEP treatment (Fig. 5f), mechanistically indicating STING pathway engagement. To further substantiate the capacity of LM^e@CEP to activate STING signaling, the expression levels of proteins downstream in the STING pathway were analyzed. Notably, Western blotting analysis revealed that the levels of p-Sting, p-TBK1 and p-IRF3 were significantly increased in the LM^e@CEP group (Fig. 5g). Therefore, we speculate that, to some extent, the antitumor effect of LM^e@CEP is facilitated through the enhanced secretion of the bacterial c-di-AMP and c-di-GMP to activate STING, thereby inducing a type I IFN immune response (Fig. 5h). Similarly, the secretion of IFN-α, IFN-β and the related inflammatory cytokines TNF-α and IL-6 was also significantly greater in mouse tumors from the LM^e@CEP group than in those from the other treatment groups (Fig. 5i). These results demonstrated that LM^e@CEP can effectively stimulate the maturation and activation of BMDCs, promote the expression of interferon genes by activating the STING signaling pathway, induce the synthesis and release of interferon, and enhance the antitumor immune response in vivo.

Fig. 5.

LM^e@CEP activates the STING pathway and facilitate the efficacy of anti-PD-L1 immunotherapy. a, b Representative flow cytometry zebra plots (left) and percentage (right) of CD45⁺CD11c⁺CD80⁺CD86⁺ BMDCs a and CD45⁺CD11c⁺CD86⁺MHC-II⁺ BMDCs (b) treated with CEP NPs, LM^e and LM^e@CEP for 24 (n = 3). c-e Quantitative comparisons of MHC-II (c), CD80 d and CD86 e expression in CD45⁺CD11c⁺ BMDCs treated as described in a and b for 24 h (n = 5). in a, b for 24 h (n = 5). f ELISA of IFN-α, IFN-β, TNF-α and IL-6 levels in the supernatant of BMDCs treated as described in a and b for 24 h (n = 3). g Immunoblot images showing STING pathway activation in BMDCs treated as described in a and b for 10 min. h Schematic model depicting the roles of LM^e@CEP in the STING pathway in BMDCs. i ELISA of IFN-α, IFN-β, TNF-α and IL-6 levels in tumors from B16F10 tumor-bearing mice (n = 5). j Photographs of B16F10 tumors from the three groups (n = 5). k, l Tumor growth curves (k, n = 5) and the final weights (l, n = 5) of wild-type (WT) and STING knockout (STING^−/−) C57BL/6 mice subcutaneously implanted with B16F10 cells subjected to the indicated treatments. m Photographs of B16F10 tumors from the three groups (n = 5). n, o Tumor growth curves (n, n = 5) and the final weights (o, n = 5) of C57BL/6 mice subjected to the indicated treatments. The data are presented as the means ± SEMs; P values were determined via one-way ANOVA with Tukey’s multiple comparisons test

To further verify whether the antitumor effect of LM^e@CEP is closely related to its activation of the STING pathway, we assessed the therapeutic efficacy of LM^e@CEP in STING knockout (STING^−/−) and wild-type (WT) mice. The results indicated that treatment with LM^e@CEP led to significant reductions in tumor growth and weight, whereas its antitumor activity was notably diminished in STING^−/− mice (Fig. 5j-l). While STING activation is crucial for antitumor immunity, emerging evidence reveals a paradoxical effect: STING-driven upregulation of PD-L1, which may induce immune suppression and limit therapeutic efficacy [28]. To address this limitation, we evaluated the combinatorial effect of LM^e@CEP with anti-PD-L1 antibody (αPD-L1 Ab) in WT tumor-bearing mice. Remarkably, the concurrent administration of LM^e@CEP and αPD-L1 Ab resulted in synergistic tumor suppression, with a significant reduction in tumor volume and weight observed on day 21 compared to compared to the LM^e@CEP monotherapy group (Fig. 5m-o). These findings collectively suggest that the antitumor efficacy of LM^e@CEP is partially dependent on STING pathway activation, and its therapeutic potential may be further enhanced by PD-L1 blockade, thereby mitigating STING-driven immune suppression.

LM^e@CEP triggers dual cell death to enhance tumor immunogenicity

Although the effects of LM^e@CEP on antitumor immunity have been well documented, it is unclear whether LM^e@CEP is able to directly induce tumor cell death. To address this question, we initially developed a three-dimensional tumor spheroid model using human nasopharyngeal carcinoma class 2Z (CNE-2Z) cells in a low-adhesion culture to simulate the tumor microenvironment. Subsequently, we assessed the antiproliferative efficacy of various treatments by measuring the growth volume of the spheroids. As shown in Fig. 6a, LM^e@CEP notably decreased spheroid growth volume more effectively than CEP NPs or LM^e (Fig. 6a). To further explore the potential of LM^e@CEP to directly induce tumor cell death, we examined its impact on ferroptosis in tumor spheroids using C11-BODIPY ^581/591 to measure lipid peroxidation, employing C11-BODIPY^581/591 as a lipid peroxidation probe. Results showed that both CEP and LM^e@CEP markedly increased lipid peroxidation levels in tumor spheroids, as indicated by the intensity of green fluorescence. Notably, CEP appears to be the principal component responsible for inducing ferroptosis in tumor cells (Fig. 6b). Consistently, we quantitatively evaluated lipid peroxidation levels associated with ferroptosis by employing flow cytometry to analyze two-dimensional cultured tumor cell models. The findings were consistent with those obtained from three-dimensional tumor spheroid experiments. Both CEP NPs and LM^e@CEP markedly elevated lipid peroxidation and facilitated ferroptosis across various tumor cell types (Fig. 6c and Fig. S22&S23). Furthermore, pathways associated with key components of the ferroptosis process, including glutathione peroxidase 4 (GPX4), solute carrier family 7, (cationic amino acid transporter, y + system) member 11 (SLC7A11), and long-chain acyl-CoA synthetase 4 (ACSL4), were evaluated via Western blot analysis. As expected, LM^e@CEP treatment increased the expression of the positive regulator ACSL4, whereas the levels of the negative regulators GPX4 and SLC7A11 were significantly reduced (Fig. 6d). Moreover, consistent with the typical effects of ferroptosis, TEM imaging revealed mitochondrial shrinkage and swollen vacuoles in CEP NP- and LM^e@CEP-treated cells (Fig. 6e). In addition, cepharanthine has previously been reported to promote tumor cell apoptosis and to serve as a valuable adjunct treatment in the context of cancer therapy [29]. Herein, we also investigated whether LM^e@CEP treatment, in addition to inducing tumor cell ferroptosis, was capable of promoting the apoptosis of tumor cells. As expected, treatment of human epithelioma-2 (Hep-2), CNE-2Z and B16F10 tumor cells with both CEP NPs and LM^e@CEP obviously induced cell apoptosis (Fig. S24).

Fig. 6.

LM^e@CEP triggers ferroptosis and apoptosis in tumor cells and enhances their immunogenicity. a Microscopy images showing the morphology of CNE 2Z tumor spheres. Spherical colonies of various sizes of CNE 2Z tumor spheres treated with the indicated treatments for 24 h. b Representative fluorescence images showing the induction of ferroptosis in CNE 2Z tumor spheres by the indicated treatments after 24 h. CNE 2Z tumor spheres were stained with C11-BODIPY ^581/591. Scale bars: 5 μm. c Representative flow cytometry histograms (left panel) and the percentage (right panel) of Lipid peroxidation levels in CNE 2Z cells treated with indicative treatments for 24 h and then stained with C11-BODIPY^581/591 (n = 3). d Immunoblot images showing GPX4, SLC7A11 and ACSL4 levels in CNE 2Z cells treated with the indicated treatments for 24 h. e TEM images of CNE 2Z cells treated with the indicated treatments for 24 h. The red arrows indicate mitochondria. Scale bars: left, 2 μm; right, 0.5 μm. f Schematic representation of the co-culture system for tumor cells and BMDCs. g ATP and HMGB1 h levels in B16F10 cells treated with CEP NPs, LM^e or LM^e@CEP for 24 h (n = 3). i Phagocytosis (left) and quantitative assay (right) showing effective uptake of B16F10 cells by BMDCs at a co-culture ratio 1:1 after 2 h. Representative flow cytometry zebra plots of the uptake of SYTOX green-labeled dead cell material by BMDCs (CD11c⁺SYTOX⁺ double-positive cells). The data are presented as the means ± SEMs; P values were determined via one-way ANOVA with Tukey’s multiple comparisons test

Recent studies have demonstrated that the induction of ferroptosis and apoptosis can augment the immunogenicity of tumor cells, consequently facilitating the activation of dendritic cells, which play a crucial role in initiating adaptive immune responses [30, 31]. Therefore, to delineate the relationship between of LM^e@CEP-induced tumor cell death and the modulation of the tumor immune microenvironment, we conducted in vitro co-cultures with primary BMDCs and B16F10 tumor cells under different treatments (Fig. 6f). First, the exposure of damage-associated molecular pattern (DAMP) molecules, such as ATP and HMGB1, on tumor cells, which is an important step in triggering immunogenic cell death in the context of tumor immunotherapy, was detected by using an ELISA kit. As shown in Fig. 6g and h, LM^e@CEP treatment resulted in the highest levels of ATP and HMGB1 release, whereas treatment with CEP NPs alone resulted in the same trend but slightly lower levels. Next, the capacity of BMDCs to phagocytose tumor cells in vitro was assessed. Similarly, LM^e@CEP-treated BMDCs exhibited the strongest phagocytic activity, and CEP NPs also facilitated the engulfment of tumor cells, but this effect was not observed with LM^e treatment alone (Fig. 6i). Taken together, these results indicate that, beyond inducing apoptosis, LM^e@CEP can initiate ferroptosis in tumor cells, subsequently activating dendritic cells and enhancing tumor immunogenicity, which collectively contribute to the improved overall efficacy of cancer immunotherapy.

Inhibition of lung metastasis by LM^e@CEP

The immunotherapeutic potential of bacterial vectors stems from their dual capacity to deliver exogenous antigens and elicit endogenous danger signaling. Pathogen-associated molecular patterns (PAMPs) from invading microbes engage pattern recognition receptors (PRRs) on antigen-presenting cells, while infection-induced cellular stress generates DAMPs that function as endogenous adjuvants [32]. This immunogenic milieu synergistically activates dendritic cell maturation, enabling robust cross-presentation of tumor antigens to CD8⁺ T lymphocytes. As described above, “mixed” tumor cell death after LM^e@CEP treatment could induce antitumor immune responses through the activation of immune cells, followed by the formation of immunological memory, whereas memory CD8⁺ T-cell subsets mediate long-lived antitumor immunity. Analysis of memory CD8⁺ T-cell subsets in the TDLNs after various treatments revealed a higher ratio of CD8⁺ central memory (TCM, CD8⁺CD44⁺CD62L⁺) to effector memory (TEM, CD8⁺CD44⁺CD62L⁻) T cells in the LM^e@CEP group than in the other groups (Fig. S25). These findings suggest that LM^e@CEP-induced tumor cell death evoked robust antitumor immunity and generated lasting immune memory. In addition, LM^e@CEP treatment significantly up-regulated histocompatibility complexes I (MHC-I) levels in both B16F10 and 4T1-Luc cells (Fig. S26). Thus, to test whether the immunogenicity of tumor cells treated with LM^e@CEP enhances tumor immunotherapy, we established mouse models of lung metastases from breast cancer or melanoma. In a mouse model of lung metastatic melanoma tumors, mice were intravenously injected with B16F10 melanoma cells, followed by subcutaneous injection of different types of treated tumor cells, as shown in Fig. 7a. Compared with the other treatments, weekly subcutaneous administration of LM^e@CEP-treated tumor cells significantly inhibited the progression of B16F10 lung metastases. Compared with the saline, CEP NPs and LM^e groups, the LM^e@CEP group presented the greatest inhibition of lung metastasis (Fig. 7b, c). Moreover, there was no significant weight loss in the mice during the treatment of mice with lung metastases (Fig. 7d). Similarly, a mouse model of metastatic breast cancer was established by intravenous injection of 4T1-Luc tumor cells into BALB/c mice on day 0. On days 1, 7 and 14, these mice received subcutaneous injections of tumor cells treated with saline, CEP NPs, LM^e or LM^e@CEP. 4T1-Luc mouse breast cancer cells were transformed with a gene expressing luciferase, allowing us to monitor lung metastases in vivo via bioluminescence imaging (Fig. 7e). Compared with saline injection, weekly subcutaneous injection of CEP NPs-, LM^e- or LM^e@CEP-treated tumor cells significantly reduced the tumor burden (Fig. 7f, g). No obvious change in body weight was observed (Fig. 7h). Also, In addition, pathology analysis of the lungs revealed a similar trend (Fig. 7i, j). Overall, these results indicated that immunization of mice with tumor cells pretreated with LM^e@CEP significantly enhanced the antitumor immune response to inhibit tumor metastasis.

Fig. 7.

LM^e@CEP inhibits the metastasis of B16F10 and 4T1-Luc tumors. (a) Schematic presentation of the therapeutic strategy involving the use of LM^e@CEP to protect against metastatic B16F10 or 4T1-Luc tumor challenge. (b, c) Photographic images of lungs (b) and quantification of lung metastasis nodules (c) in B16F10 tumor-bearing mice on day 21 post-tumor challenge after the indicated treatments (n = 5). (d) Changes in the body weights of B16F10 tumor-bearing mice during the indicated treatment periods (n = 5). (e) In vivo bioluminescence imaging of lung metastases established by 4T1-Luc cells on days 14 and 21 (n = 5). (f, g) Photographic images of lungs (f) and quantification of lung metastasis nodules (g) in 4T1-Luc tumor-bearing mice on day 21 post-tumor challenge after the indicated treatments (n = 5). (h) Changes in the body weights of 4T1-Luc (h) tumor-bearing mice during the indicated treatment periods (n = 5). (i, j) Images of H&E-stained lungs slices from B16F10 (i) and 4T1-Luc (j) tumor-bearing mice subjected to the indicated treatments. The data are presented as the means ± SEMs; P values were determined via one-way ANOVA with Tukey’s multiple comparisons testwjshan@tmmu.edu.cn

Validating the antitumor effects of LM^e@CEP in human samples

Immune cells, particularly T cells, are among the mainstays of cancer immunotherapy, but investigations of their interactions with tumor cells at the individual patient level are limited. To validate the immunotherapeutic efficacy of LM^e@CEP in individual patients, we investigated the production and function of LM^e@CEP-treated tumor-reactive T cells in a coculture system with isolated human head and neck tumor tissues and autologous lymphocytes from peripheral blood. Prior to the start of coculture, tumor tissue excised from patients was cut into small pieces, and peripheral blood mononuclear cells (PBMCs) were isolated from peripheral blood and coincubated with various treatments for 24 h (Fig. 8a). We first examined the secretion of interferon γ (IFN-γ) from T cells in the culture system, and as shown in Fig. 8b, the secretion of IFN-γ from CD4 and CD8 T cells was significantly greater in the cells from the LM^e@CEP treatment group than in those from the other treatment groups. This result further confirmed that LM^e@CEP could effectively activate STING and promote interferon expression. In addition, the levels of the cytotoxicity-associated molecule Granzyme B in CD8⁺ effector T cells were also detected. Compared with those in the other groups, CD8⁺ T cells in the LM^e@CEP group produced the highest levels of granzyme B, suggesting that LM^e@CEP effectively enhanced the ability of cytotoxic CD8⁺ T cells to kill tumor cells (Fig. 8c). Immunofluorescence staining also revealed that LM^e@CEP treatment significantly promoted CD8⁺ T-cell infiltration into tumors (Fig. 8d).

Fig. 8.

LM^e@CEP activates immune-mediated killing of tumor cells in patients. a Schematic representation of the co-culture system with nasopharyngeal tumor tissue excised from patients and PBMCs were isolated from peripheral blood. b Representative flow cytometry zebra plots (left) and percentages (right) of IFN-γ⁺CD4⁺ T cells and IFN-γ⁺CD8⁺ T cells in tumors from patients (n = 3). c Representative flow cytometry zebra plots (left) and percentages (right) of Gran B⁺CD8⁺ T cells in tumors from patients (n = 3). d Representative immunofluorescence images demonstrating CD3 and CD8 expression in patient tumors (green indicates CD3 staining, red indicates CD8 staining). Scale bars: 50 μm. e, f Representative flow cytometry zebra plots e and percentages f of macrophages (CD45⁺CD11b⁺CD68⁺CD163⁺) in tumors from patients (n = 3). g Representative immunofluorescence images demonstrating CD68 and CD206 expression in patient tumors (green indicates CD68 staining, red indicates CD206 staining). Scale bars: 50 μm. h Representative images of H&E staining of tissue sections from patient tumors subjected to the indicated treatments for 24 h. The data are presented as the means ± SEMs; P values were determined via one-way ANOVA with Tukey’s multiple comparisons test

Moreover, the status of TAMs in the tumor microenvironment was analyzed. The results revealed that after treatment with LM^e@CEP, there was a significant reduction in the number of CD68⁺CD163⁺ M2 TAMs in the LM^e@CEP group compared with the CEP NP and LM^e groups (Fig. 8e, f). Immunofluorescence results also revealed that the number of M2-type macrophages in the tumors of the LM^e@CEP group was significantly reduced (Fig. 8g). These data mechanistically demonstrate that LM^e@CEP orchestrates a phenotypic inversion of the tumor immune microenvironment, transitioning from a protumorigenic state to an immunologically “hot” phenotype. In addition, H&E staining of tissue sections from human tumors revealed a large area of tumor tissue necrosis in the LM^e@CEP group, confirming that LM^e@CEP has a strong ability to promote the killing of human tumor cells ex vivo (Fig. 8h). Overall, these findings demonstrate that LM^e@CEP exhibits significant antitumor efficacy in human tumor models, underscoring its translational potential.

Discussion

Although bacterium-mediated tumor immunotherapy holds great promise, there are many obstacles to its development, including the potential cytotoxicity of bacteria, the inability to completely lyse cancer cells, and the possibility of mutation [7]. In this study, we adopted an ad hoc strategy involving electroporation of CEP NPs into live-attenuated LM strains, which greatly improved the safety and efficacy of bacterial cancer immunotherapy. CEP NPs obviously inhibited bacterial viability and interfered with the metabolism of LM, leading to its transformation into a harmless “zombie” state. Moreover, loading with CEP NPs interfered with the thymidine synthesis pathway of LM by inhibiting folate metabolism, thereby promoting elevated levels of the bacterial second messenger cyclic di-AMP (c-di-AMP) and consequently enhancing the STING-dependent antitumor immune response. In addition, zombified LM could be used to deliver CEP NPs into the tumor microenvironment, which induce the apoptosis and ferroptosis of tumor cells. Thus, in this work, we generated in an oncolytic zombified bacterium, LM^e@CEP, which represents an innovative tool to enhance cancer immunotherapy.

Bacterium-based cancer immunotherapy has shown great promise for cancer treatment because of its ability to surpass the immune threshold and induce strong immune responses against cancer cells. Although live attenuated strains of the bacterium LM, which is a facultative pathogen, have been successfully used for the treatment of numerous malignancies, the potential risk of infection remains nonnegligible. This is especially true for patients who are already in an immunosuppressed state, for whom the use of live attenuated pathogens could be problematic and high risk [33]. These limitations prompted us to further improve the safety of LM-based bacteriotherapy. On the basis of a double-deficient strain, LM ∆actA/∆inlB, which lacks the actin assembly-inducing protein (ActA) and internalin B (InlB), we first identified a natural product called CEP that could further reduce bacterial virulence by destroying the bacterial cell membrane and disturbing normal bacterial metabolism, turning LM into zombie-like bacteria. Reservoirs of CEP NPs accumulated in the zombified bacterial cells. After reaching the tumor microenvironment, the zombified bacteria induced an aggressive immune response, while the stored CEP NPs simultaneously leached out to induce tumor cell apoptosis and ferroptosis. The inability of certain strains of heat-killed LM to effectively induce CD8⁺ T-cell responses has been well documented [34, 35], suggesting that the entry of LM into the host cytoplasm is a prerequisite for the effective activation of CD8⁺ T cells. The advancement of LM-based tumor immunotherapies therefore necessitates the utilization of attenuated LM rather than killed LM. In contrast to live viral or bacterial oncolytic therapies that may induce listeriosis under certain circumstances, our oncolytic zombified bacteria may prove to be a safer option for in vivo applications, as we have effectively reduced pathogenicity while enhancing immunogenicity.

Early studies confirmed the ability of LM-based immunotherapies to elicit a strong and sustained antitumor immune response [33]. The resulting zombified LM loaded with CEP NPs was shown to produce high levels of cyclic di-AMP and activate the cGAS-STING signaling pathway, leading to the secretion of type I interferons. However, microbial stress responses and viability are mediated by c-di-AMP in many Firmicutes [22, 36]. Little is known about how environmental cues influence nucleotide dynamics in bacteria. We therefore speculated that exposure to CEP NPs could interfere with bacterial metabolism and lead to the accumulation of c-di-AMP in LM. The results of our metabolomic analyses revealed that exposure of LM to CEP NPs altered the activity of several crucial metabolic pathways, which might be expected to reduce viability. Mechanistically, CEP NPs transported into the cell may bind to the catalytic sites of thymidylate synthase (ThyA) and DHFR; thus, CEP NPs may act as inhibitors, disrupting thymidine biosynthesis and ultimately enhancing c-di-AMP production. Notably, Type I IFNs induced by LM can increase acute pathogenic effects and even hinder the development of adaptive immunity. Therefore, the modulation of cyclic di-AMP production by LM-based tumor immunotherapy may be promising but will require a delicate balance between the ability to impair and enhance antitumor immunity [37, 38]. Encouragingly, the zombified LM^e@CEP resulted in faster and stronger immune activation, modulated the immune microenvironment and retarded tumor growth in multiple mouse tumor models.

As a crucial conserved cytosolic DNA-sensing mechanism involved in innate immunity, the STING signaling pathway regulates crosstalk between tumor cells and immune cells, thus serving as an increasingly important target for antitumor immunotherapy [39]. Nevertheless, although several small-molecule STING agonists have been used as anticancer therapies in preclinical and clinical trials [40–42], owing to the dense extracellular matrix and high interstitial fluid pressure of tumor tissues, some STING agonists have difficulty reaching and penetrating into the local tumor microenvironment, which greatly limits their immunotherapeutic efficacy in vivo [43, 44]. Additionally, STING agonists alone usually cannot achieve satisfactory tumor inhibition effects. The combination of STING agonists with other established cancer chemotherapies and immunotherapies is becoming a mainstream approach in cancer immunotherapy [45, 46]. To date, various approaches for delivering STING agonists to enhance their cytosolic delivery and elicit STING-dependent antitumor immune responses have been explored [47–49]. In the majority of these approaches, cargoes are delivered passively, limiting tumor cell penetration efficiency and insufficiently activating antigen-presenting cells. In our study, we utilized zombified LM as a self-propelling vector to efficiently deliver bacterial cyclic dinucleotides and CEP cargo to the tumor immune microenvironment. As expected, LM^e@CEP potently suppressed the growth of both B16F10 and 4T1-Luc tumors implanted subcutaneously. Simultaneously, findings in STING^−/− mice corroborated that the antitumor efficacy of LM^e@CEP is partially reliant on the activation of the STING pathway. The administration of combination therapy with PD-L1 antibody blockade mitigates the STING-mediated upregulation of PD-1/PD-L1, thereby augmenting the antitumor effectiveness of LM^e@CEP. Moreover, it significantly repressed lung metastasis while avoiding systemic infection and toxicity.

Mechanistic analysis in vivo and in vitro revealed that our oncolytic zombified bacteriotherapy successfully stimulated proinflammatory activation of TAMs and increased cytotoxic T-cell infiltration in immune-relevant tissues. Concurrently, the production of IFN-α and IFN-β within the tumor microenvironment obviously increased, which was important for shaping the immune response to tumors and crucial for activating the host STING pathway. One limitation of our study is that the activation of other relevant immune signaling pathways, such as Toll-like receptor (TLR) (TLR2 and TLR4) and NOD-like receptor (NLR) (NOD1 and NLRP3) pathways, was not analyzed in depth [7, 50]. Previous studies have confirmed that bacteria also have the ability to activate these immune pathways. In addition to STING activation, a significant increase in the production of additional proinflammatory factors (TNF-α and IL-6) was observed following LM^e@CEP treatment. Presumably, the antitumor effect of our oncolytic zombified bacteria was achieved through the activation of multiple pathways and complementary mechanisms.

However, despite compelling preclinical results demonstrating the therapeutic potential of previous bacterium-mediated cancer treatments, clinical trials have repeatedly failed to demonstrate efficacy in patients [51–54]. Translating results from mouse models to human patients is often a great challenge. The sensitivity and response of mice to a certain stimulus may also differ from those of humans, with promising preclinical results in animals but a lack of activity in humans. In our study, we sought to determine whether the antitumor and immunomodulatory effect of the developed oncolytic zombified bacteria were maintained in human tumor samples. Notably, LM^e@CEP treatment also induced the death of tumor cells while efficiently activating antitumor immunity in human head and neck cancer samples. Our work highlights a zombified bacterium-mediated antitumor immunotherapeutic strategy for future therapeutic approaches and clinical applications.

In conclusion, we have developed a bacterium-based immunotherapy approach, namely, oncolytic zombified bacteria, by electroporating CEP NPs derived from a bioactive natural product with anticancer activity into attenuated LM. The obtained zombified bacteria (LM^e@CEP) could act as immune-stimulating agents, and after being administered to tumors, they could sequentially activate innate and adaptive immunity. LM^e@CEP could serve as an effective therapeutic agent for tumor treatment, as evidenced in different types of mouse tumor models and human tumor samples. Biological activation of STING signaling via zombified bacteria plus tumoricidal cargo exhibited synergistic antitumor effects, which optimized the bacteriotherapy approach and demonstrated the potential of this strategy for cancer immunotherapy.

Materials and methods

Materials

CEP was purchased from Abcam (Ab141915). The double-gene attenuated LM strain (LM-actA/inlB) was provided by Professor Guo Fu, Xiamen University (Xiamen, China). D-Luciferin Sodium was purchased from TargetMol (T19743). C11-BODIPY ^581/591 dye was purchased from Thermo Fisher (D3861). GoInVivo™ Purified anti-mouse CD274 (B7-H1, PD-L1) Antibody was purchased from Biolegend(124329). Mouse IL-4 recombinant protein and recombinant murine GM-CSF were purchased from PeproTech (214 − 14 and 315-03). A commercial ELISA kit for murine IFN-γ, IFN-β, TNF-α, and IL-6 was obtained from Meimian in Jiangsu, China.The rabbit anti-STING antibody [EPR25090-107] (ab288157) was purchased from Abcam. The phospho-IRF3 (Ser386) rabbit pAb (530857), phospho-NAK/TBK1 (Ser172) rabbit mAb (R30260), GPX4 rabbit pAb (381958), ACSL4 rabbit mAb (R24265), SLC7A11 rabbit mAb (R382036) and GAPDH rabbit pAb (380626) were purchased from Zenbio. Phospho- STING (Ser366) rabbit pAb (AF7416), TBK1 rabbit pAb (DF7026) and IRF3 rabbit pAb (DF6895) were purchased from Affinity. All antibody dilutions were performed according to the manufacturer’s instructions.

Synthesis and characterization of CEP NPs and LM^e@CEP

CEP NPs were synthesized using a brilliant droplet-confined/cryodesiccation-driven crystallization approach [55]. Briefly, a trichloromethane solution containing CEP powder was mixed with Pluronic F127 (1.0 wt%) and then converted into an O/W emulsion by sonification (500 v, 2 kHz, 20% power). The emulsion was immediately immersed in liquid nitrogen and freeze-dried at −80 °C for 24 h. CEP NPs were harvested and characterized. The morphology of the CEP NPs was visualized via transmission electron microscopy (TEM, Talos F200S, Thermo Fisher Scientific, NL).

The LM used in the treatments was prepared from overnight cultures. The number of cells was determined by measuring the OD at 600 nm (1 OD = 1 × 10⁹ CFUs). Bacterial suspensions adjusted to the desired concentration were washed three times in sterile PBS. To improve the controlled delivery of the formulation, PBS or PBS-soluble CEP NPs (2 mg/mL) and LM (2 OD) were mixed together at a ratio of 1:1(V: V) and then electroporated with an electroporation system for 6 cycles (Voltage 1800 v, Capacitance 10 µF; resistance, 60 Ω; cuvette, 2 mm; interval, 5–10 s). The samples were subsequently incubated at 4 °C for 12 h. Electroporated LM (LM^e) and electroporated LM loaded with CEP NPs (LM^e@CEP) were harvested and characterized. After electroporation, the bacterial suspension underwent purification to eliminate debris and lysates. This involved centrifuging at 8,000 × g for 10 min at 4 °C to separate intact bacteria, removing the supernatant, and washing the pellet twice with sterile PBS. A brief ultrasonication (5-second pulses, 50% amplitude, on ice) was then used to break up aggregates, followed by another centrifugation to obtain a clean pellet of viable bacteria. The final pellet was resuspended in fresh PBS for further experiments and animal injections. The microstructures of LM^e and LM^e@CEP were visualized using a scanning electron microscope (SEM, SU8010, Hitachi, Japan) equipped with a cryo system (PP3010T, Quorum Technologies Ltd., UK). The morphologies of LM^e and LM^e@CEP were visualized by transmission electron microscopy (TEM, Talos F200S, Thermo Fisher Scientific, NL).

The release behavior of LM^e@CEP

LM^e@CEP were placed in separate dialysis bags (MWCO = 3000 Da) using PBS with pH 6.5 and 7.4 as the release medium. The bags were then placed in a shaking incubator at 37 ℃ for incubation. Samples were collected at 0.5, 1, 2, 4, 8, 10, 12 and 24 h, and CEP content was determined via HPLC-MS.

The TEM-EDS analysis

5 nm gold nanoparticles coated with dodecylamine were purchased from Suzhou Yansheng Bio-tech Co.,Ltd. The TEM- energy-dispersive X-ray spectroscopy (EDS) analysis of Au-labeled CEP NPs and LM^e@Au-labeled CEP were determined by transmission electron microscope (TEM, Talos F200S, Thermo Fisher Scientific CDLtd, Czech).

Cell culture and treatments

The B16F10 murine melanoma cell line was obtained from the Cell Bank of the Chinese Academy of Sciences. The 4T1-Luc murine breast adenocarcinoma cell line, CNE 2Z human nasopharyngeal carcinoma cell line and Hep2 human laryngocarcinoma cell line were obtained from the BeNa Culture Collection. All cell lines were cultured in complete media (DMEM supplemented with 10% (v/v) FBS and 1% (v/v) penicillin/streptomycin) at 37 °C, 5% CO₂ in a humidified atmosphere. Cells exposed to CEP (5 µg/mL), LM^e (1 × 10⁶ CFUs), and LM^e@CEP (5 µg/mL CEP at 1 × 10⁶ CFUs LM^e) treatment for 24 h were set as the control, CEP, LM^e and LM^e@CEP groups, respectively. 2 h after above factors were treated, gentamicin was separately added for further incubation.

Animal treatments and tumor models

All animal experiments were performed according to protocols approved by the Institutional Animal Care and the Animal Ethics Committee of Army Medical University, Chongqing, China (Approval No. AMUWEC2020995). Female C57BL/6 mice and female BALB/c mice (aged 6–8 weeks; weighing 18–22 g) were purchased from SiPeiFu (SPF) Biotechnology (Beijing, China). STING-deficient (STING^−/−) C57BL/6 J mice were gifted by Professor Zhexue Qin from the Third Military Medical University. To establish subcutaneous tumor models, 1 × 10⁶ B16F10 cells in sterile PBS were subcutaneously inoculated on the right side of each C57BL/6 female mouse. When tumor volumes with an average volume of 100 mm³, mice were randomly divided into four groups (n = 5 per group) as follows: saline, the CEP NPs (CEP), the LM^e (LM^e), and the CEP NPs + LM^e (LM^e@CEP), respectively. Thereafter, the mice were intratumorally injected with 100 µL of saline, CEP NPs (50 µg), LM^e (5 × 10⁷ CFUs), and LM^e@CEP (50 µg CEP at 5 × 10⁷ CFUs LM^e) two or three times weekly for two weeks. Therapeutic blockade of PD-L1 was performed by twice weekly intraperitoneal injections of 200 µg PD-L1 antibodies or control IgG2b mAb. The body weight and survival time were recorded. Simultaneously, the tumor size was monitored by digital caliper and the tumor volume (mm³) was calculated by using the formula: V = Length × Width²/2. After 21 days of treatment, mice were sacrificed, and the tumor tissues were collected for assessing antitumor effect. Other major organ samples (lymph nodes, heart, liver, spleen, and kidney) were also obtained for future analyzation. Additionally, blood samples were collected and analyzed the levels of serum aspartate transaminase (AST), alanine aminotransferase (ALT), total bilirubin (TBIL), albumin (ALB), urea (UREA), creatinine (CREA) and uric acid (UA).

To set up orthotopic 4T1 tumor modes, 1 × 10⁶ 4T1-Luc cells in sterile PBS were subcutaneously injected into the second breast fat pad on left side of each female BALB/c mouse. When tumor volumes with an average volume of 50 mm³, mice were randomized into four groups (n = 5 per group), and received corresponding treatments as above. Simultaneously, survival time and tumor size were recorded. Additionally, tumor growth was monitored through the bioluminescence images on days 7, 14 and 21. The mice were intraperitoneally injected with 150 µL D-luciferin sodium (15 mg/mL, diluted in PBS). To monitor the distribution of the Cy5.5-labeled LM^e@CEP in tumor and tumor-draining lymph nodes (TDLNs), Cy5.5-labeled LM^e@CEP was intratumorally injected to C57BL/6 J mice and sacrificed at 2 h and 8 h. An imaging instrument (IVIS Spectrum, Caliper Life.

Sciences, USA) was used to monitor the distribution and analyzed with Living Image 4.2 software (PerkinElmer). After 21 days of treatment, mice were sacrificed, and the tumor tissues in different treatments were collected for photography and tumor weight.

To set up tumor metastasis models, 1 × 10⁵ B16F10 or 4T1-Luc cells were implanted into C57BL/6 or BALB/c female mice via the tail vein on day 0. The mice were subsequently randomized into four groups as described above and subcutaneously received injections of “death” tumor cells (1 × 10⁴ B16F10 or 4T1-Luc was induced in vitro by treatment with LM^e@CEP for 24 h), followed by treatment with 100 µL of saline, CEP NPs (50 µg), LM^e (5 × 10⁷ CFUs), and LM^e@CEP (50 µg CEP and 5 × 10⁷ CFUs LM^e) at 1, 7, and 14 days, intraperitoneally. The body weights of the tumor metastasis model mice were recorded. Additionally, the tumor growth was monitored through the bioluminescence images on days 14 and 21. After 21 days of treatment, the mice were sacrificed, and the lungs were collected for photography, tumor metastasis nodules and H&E staining.

Histology and immunohistochemical staining

Fresh tissues (tumors and major organs) with different treatments were harvested and fixed in 4% paraformaldehyde. Subsequently, the samples embedded in paraffin and then cut into 4 μm slices for hematoxylin and eosin (H&E) and immunohistochemical staining.

For immunohistochemical staining, B16F10 tumor slices were incubated with anti-Ki67 (GB111141, Servicebio) overnight at 4°C. Subsequently, slices were stained with horseradish peroxidase (HRP)-labeled secondary antibody (GB23303, Servicebio), followed by treatment with 3,3’-diaminobenzidine (DAB) (G1212, Servicebio). The slices were imaged under a light microscope (Olympus BX51, Japan).

Immunofluorescence staining

To detect the intratumoral distribution of LM^e or LM^e@CEP, B16F10 tumor-bearing mice were intratumorally injected with Cy5.5-labled LM. 24 h later, tumor tissues were collected and fixed with 4% paraformaldehyde. Then the tissues were embedded and tissue slices (4 μm) were prepared with the RM2016 (leica). After blocking with 5% BSA (Sigma-Aldrich, USA), tissues were incubated with F4/80 primary antibody (1:200 diluted) overnight at 4 °C. The next day, the samples were washed with PBS and then incubated with indicated secondary antibodies (AlexaFluor^®488-conjugated IgG). Finally, the cell nucleus was stained with DAPI (1 mg/mL).

To detect the polarization of macrophages within the tumor microenvironment in vivo, fresh tumor tissues from B16F10 tumor-bearing mice were harvested, fixed and embedded. After blocking with 5% BSA, tumor sections were incubated with F4/80 and CD206 primary antibody (1:200 diluted) and indicated secondary antibodies (AlexaFluor^®555- or AlexaFluor^®488-conjugated IgG). Finally, the slices were stained with 1 mg/ml DAPI to label the tumor cell nucleus. The fluorescence images were captured with a LSM800 microscope (Zeiss).

Western blot

To evaluate the expression of target proteins, cells were harvested in RIPA lysis buffer supplemented with 1% (v/v) PMSF protease inhibitor or 1% (v/v) phosphatase inhibitor. Then, the protein concentrations were determined via a BCA protein assay kit (Beyotime Biotechnology, Jiangsu, China) and denatured with loading buffer at 100 °C for 5 min. The denatured samples (30 µg/lane) were separated via 8–12% SDS-PAGE and then transferred to polyvinylidene difuoride membranes (PVDF, Millipore Corp, USA). The membranes were blocked and probed at 4 °C overnight with specific primary antibodies for STING, p-STING, IRF3, p-IRF3, TBK1, p-TBK1, GPX4, ACSL4, SLC7A11 and GAPDH. Then, HRP-conjugated secondary antibodies (Cell Signaling Technology, MA, USA) were added and incubated for 2 h at room temperature. Finally, the membrane-bound immune complexes were spotted using ChemiDoc™ Touch Imaging System (Bio-Rad, California, USA) with Pierce ECL Plus Substrate (Thermo Fisher Scientific).

Flow cytometry analysis

Analysis of immune cells for flow cytometry analysis, on day 21, tumors and lymph nodes of B16F10 tumor-bearing mice received various treatments were harvested. Then, the tumor samples were mechanically disrupted and digested with a Tumor Dissociation Kit (mouse tumor) (130-096-730, Miltenyi Biotechnology, USA), and the lymph nodes were prepared to single cell suspensions by mechanical grinding. The cell suspensions were centrifugated at 1,500 rpm for 5 min, and red blood cells were removed using ammonium-chloride-potassium (ACK) lysis buffer for 5 min, followed by washing twice with Flow Cytometry Staining Buffer (FACS). Next, the single cell suspensions from tumors and lymph nodes of various groups were collected and stained with specific fluorescent conjugated antibodies for flow cytometry analysis(CytoFLEX, Beckman, USA). Fluorescence minus one (FMO) controls for each sample were utilized as gating references for each fluorophore. Compensation for all fluorophores was conducted using compensation beads (BD, catalog # 552845). A detailed gating strategy for all immune cell subsets(e.g., lymphocyte, macrophages, dendritic cells, PBMCs etc.) analyzed is provided in Fig S27. All flow cytometry antibodies used in this study, including their clones, manufacturers, catalog numbers, and dilutions, are detailed in Table S3 for reproducibility and transparency. All the data were analyzed with FlowJo V10.

Measurement of lipid peroxidation

The fluorescent dye C11-BODIPY ^581/591 (Thermo Fisher Scientific, USA) is used to assess lipid peroxidation in tumor cells and bacteria. The samples were incubated with 1 µM C11-BODIPY ^581/591 at 37 °C for 30 min. For flow cytometry analysis, the cells were digested and resuspended in PBS. After that, the samples were subsequently detected via flow cytometry (CytoFLEX LX, Beckman Colter, USA) with the FL1 channel. For fluorescence imaging, tumor cells were washed with PBS after incubating. The cells were subsequently captured the images via fluorescent microscope (CYTATION 5, BioTeK, USA).

Apoptosis analysis

Following the manufacturer’s instructions, the Annexin V-AF488/7-AAD Apoptosis detection kit (Vazyme, Nanjing, China) was used to detect the cell apoptosis in vitro. Briefly, tumor cells were incubated with 5 µL of annexin AF488 and 5 µL of 7-AAD for 10 min at room temperature. After washing twice with PBS, the cells were evaluated via flow cytometry (CytoFLEX LX, Beckman Colter, USA), and the data were analyzed with FlowJo V10. In accordance with standard protocols, apoptosis in tumor tissues in vivo was assessed via TUNEL staining with an in-situ Cell Death Detection Kit (Promega, USA). The resulting fluorescence was detected using a fluorescence microscope (LSM800, Zeiss, Germany).

Generation of BMDCs

BMDCs were prepared as previously described [56]. Briefly, the intact femurs and tibias were isolated from female C57BL/6 mice, dipped briefly into 75% ethanol, and deposited in RPMI 1640 medium over ice. Both ends of each femurs and tibias were cut and flushed with RPMI 1640 supplemented with 10% (v/v) FBS and 1% (v/v) penicillin/streptomycin. Following suspension in ammonium-chloride-potassium (ACK) lysis buffer, bone marrow cells were collected after passing through a 40-µm cell strainer. Then cells were counted and resuspended at a density of 1 × 10⁶ cells per ml, cultured in complete RPMI 1640 medium containing 10% (v/v) heat-inactivated FBS, 1% (v/v) penicillin/streptomycin, 20 ng/mL recombinant murine granulocyte-macrophage colony stimulating factor (mGM-CSF), 10 ng/mL IL-4 and 50 µM β-mercaptoethanol. The culture medium was changed on days 3 and 6. Finally, loose and nonadherent cells were harvested for further experiments after 7 days.

In vitro phagocytosis and BMDC maturation assays

To evaluate the phagocytosis ability of BMDCs, target cells (B16F10) were treated with CEP (5 µg/mL), LM^e (1 × 10⁶ CFUs), or LM^e@CEP (5 µg/mL CEP and 1 × 10⁶ CFUs LM^e) for 24 h. At preset time points, B16F10 cells were collected, washed and incubated with SYTOX Blue (KeyGEN Biotech, Jiangsu, China) at 37 °C in the dark condition for 30 min and then co-cultured with BMDCs at a ratio of 1:1 for 2 h. Afterwards, the co-cultured cells were harvested, stained with APC-conjugated anti-CD11c (BioLegend, CA, USA), and analyzed by flow cytometry (CytoFLEX, Beckman, USA). For cytokine release and co-stimulatory marker characterization, BMDC cells were plated in 24-well plates at a density of 1 × 10⁵ cells per ml. Various treatments with CEP (1 µg/mL), LM^e (2 × 10⁵ CFUs), or LM^e@CEP (1 µg/mL CEP at 2 × 10⁵ CFUs LM^e) were added to the cells and incubated at 37 °C for 24 h. Afterwards, the supernatants medium was collected for detection cytokine release. Simultaneously, the cells were washed and scraped from 24-well plates, followed by staining with APC-Cy7-conjugated anti-CD45, APC-conjugated anti-CD11c, FITC-conjugated anti-MHC-II, PerCP-Cy5.5-conjugated anti-CD80, or PE-conjugated anti-CD86 antibodies at room temperature for 30 min. Then, the cells were washed twice with FACS buffer and imaged using a Beckman flow cytometer. All the data were analyzed with FlowJo V10.

Enzyme-linked immunosorbent assay (ELISA)

To detect the levels of cytokines (IFN-β, IL-6, TNF-α and IFN-α) in BMDCs in vitro, the supernatants medium of BMDCs treated with CEP NPs, LM^e and LM^e@CEP for 24 h were collected. And, the cytokines levels were detected with ELISA kits (Meimian, Jiangsu, China) according to the manufacturer’s instructions. To detect the levels of IFN-β, IL-6, TNF-α, along with IFN-α in tumor in vivo, murine B16F10 tumor samples were separated, homogenized and centrifuged, and the supernatants were detected through ELISA. The absorption at 450 nm was measured by a multifunctional reader (SpectraMax i3, Molecular Devices, USA), and the expression of cytokines was calculated by using the standard curve generated.

UPLC-ESI-QTOF-MS analysis and data processing

To identify the effects of CEP NPs on bacterial metabolism, untargeted metabolimic analysis using LC-MS/MS were performed at Shanghai Applied Protein Technology Co., Ltd. The nanomaterials were added to 100 µL of ultrapure water and 400 µL of acetonitrile containing 5 nM myristoyl-trimethyl D9 carnitine. The mixtures were homogenized for 30 s and centrifuged at 16,000 × g for 15 min at 4 °C. Fifty microliters of the supernatant were transferred to a new 1.5 ml tube, dried under vacuum, redissolved in HPLC-grade water containing 1.6% acetonitrile, and assayed using a Waters UPLC-Q/TOF MS system with an electrospray source. An Acquity BEH C18 column (50 × 2.1 mm, 1.7 μm, Waters Crop.) was applied for chromatographic separation. The flow rate was 0.5 mL/min, with an aqueous acetonitrile gradient containing 0.1% formic acid over a 10 min run. Mass spectrometry was performed in positive (ESI+) ionization mode. The source and desolvation temperatures were set at 120 °C and 350 °C, respectively. The capillary voltage and cone voltage were set to 3000 and 20 V, respectively. Nitrogen was used as both the cone gas (50 L/h) and desovation gas (600 L/h). Argon was used as the collision gas. A mass range of m/z 50–850 was acquired. The results were calculated according to individual standard curves established as follows: area analyte/area internal. MetaboAnalyst 4.0 was used to identify the pathways most relevant to the differentially abundant metabolites. The differentially expressed genes were subjected to Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway enrichment analysis using the omicsRE network platform. VIP > 1 and p < 0.05 were considered to indicate statistical significance. Online metabolic databases, including Metlin (http://metlin.scripps.edu), HMDB (http://www.hmdb.ca/) and PubChem (http://pubchem.Ncbi.nlm.nih.gov), and the exact masses of the metabolites were used to identify the differentially abundant metabolites.

Human samples

Nasopharyngeal tumor samples and PBMCs isolated from peripheral blood were harvested from the surgical specimens of three patients who underwent primary surgery at Southwest Hospital. The samples were obtained with patient consent in accordance with the principles of good clinical practice (Chongqing, China, KY202243) and the Declaration of Helsinki after study approval was obtained from the internal review board of Southwest Hospital.

The tumor samples were cut into approximately 1 mm³ fragments, and PBMCs were isolated from the peripheral blood. Then, the tumor fragments and PBMCs were cocultured at 37 °C in 24-well plates in 500 µL of culture medium composed of RPMI 1640 containing L-glutamine (VivaCell, C3010–0500) supplemented with 10% FBS (Azaood, AD00004–500), 1% MEM nonessential amino acid solution (Solarbio, N1250–100), and 100 U/mL penicillin/streptomycin (Beyotime, C0222). Stimulation was performed with CEP (5 µg/mL), LM^e (1 × 10⁶ CFUs), or LM^e@CEP (5 µg/mL CEP and 1 × 10⁶ CFUs LM^e). The above procedures were carried out under sterile conditions. After 24 h, the PBMCs were collected for flow cytometry analysis, and the tumor fragments were collected for H&E staining, immunofluorescence analysis and gene sequencing analysis.

For flow cytometry analysis, PBMCs were stained at 4 °C for 30 min with antibodies specific for the following human proteins (from BioLegend): CD45 (368511), CD11c (301607), CD11b (301329), HLA-DR (327007), CD68 (333813), CD86 (374205), CD163 (326513), CD3 (300305), CD4 (344607), CD8 (300925), IFN-γ (502508), and granzyme B (396409). After two washes with FACS buffer, the cells were immediately subjected to flow cytometry (CytoFLEX, Beckman, USA). The data were analyzed with FlowJo V10.

Statistical analysis

All data are expressed as the means ± standard errors of the means (SEMs). The statistical significance of differences between groups was analyzed by one-way ANOVA using GraphPad Prism 9 software (La Jolla, California, USA). A probability value of 0.05 was considered to indicate statistical significance. All in vivo and in vitro experiments were performed at least in triplicate.

Author contributions

Tao Liu performed experiments, analyzed data, wrote the manuscript. Qunfang Yang, Jianhua Huang and Lanfang Zhang performed a part of experiments, wrote the manuscript. Zhihao Zhao conceived and performed metabolomic analysis. Maoling Huang and Xiaohong Chen assisted with data analysis. Ya Liu and Haigang Zhang provided valuable discussions. Lin Zhang edited the manuscript, conceived and designed this study. Cheng Zhong conceived and designed this study. Wenjun Shan wrote and edited the manuscript, provided funding, conceived and designed this study.All authors reviewed the manuscript.

Funding

This work was supported by the National Natural Science Foundation of China (grant numbers 32471458, 82473933), the Natural Science Foundation of Chongqing, China (cstc2020jcyj-msxmX0411), Chongqing Talents Program (grant number cstc2022ycjh-bgzxm0111) and the Science and Technology Research Program of Chongqing Municipal Education Commission (Grant No. KJZD-K202312802, KJZD-M202512802 ).

Data availability

The authors declare that the data supporting the findings of this study are available within the article and its Supporting Information or from the corresponding author upon reasonable request.

Declarations

Ethics approval and consent to participate

All animal experiments were approved by the Institutional Animal Care and Use Committee of the Army Medical University (Chongqing, China, AMUWEC2020995). The human samples were obtained with patient consent in accordance with the principles of good clinical practice (Chongqing, China, KY202243) and the Declaration of Helsinki after study approval was obtained from the internal review board of Southwest Hospital.

Patient consent statement

The study was approved by the internal review board of Southwest Hospital, Chongqing, China. All patients provided written informed consent.

Competing interests

The authors declare no competing interests.