Cascaded immunotherapy with implantable dual-drug depots sequentially releasing STING agonists and apoptosis inducers

Abstract

Non-nucleotide stimulators of interferon gene (STING) agonists hold promise as immunotherapeutic agents for postsurgical adjuvant treatment of tumors. However, their limited effect duration hampers therapeutic effectiveness, necessitating prolonged administration of multiple doses that heightens infection risk and impacts patient compliance. Here, we develop an implantable dual-drug depot in a sandwich-like configuration, with a non-nucleotide STING agonist (MSA-2) in the outer layers of 3D-printed scaffolds and an immunogenic apoptosis inducer (doxorubicin, DOX) in the inner layer of electrospun fibers. We discover that MSA-2 can elicit endoplasmic reticulum stress-mediated and general immunogenic apoptosis of cancer cells. The stimulations with tumor-associated antigens and damage-associated molecular patterns from cancer cells, along with proinflammatory factors secreted by matured dendritic cells and M1-polarized macrophages, can depolymerize intracellular microtubules guiding activated STING trafficking towards lysosomes for degradation. Collectively, the dual-drug depots can initiate a long-lasting cascaded immunotherapy and chemotherapy, suppressing postsurgical tumor recurrence and metastasis.

Introduction

Surgical resection is one of the most effective and common methods to treat solid malignant tumors^1,2,3. Despite significant advancements in surgical techniques, the issue of postoperative tumor recurrence and metastasis remains a pressing clinical challenge that requires sustained attention. This problem is primarily attributed to the presence of positive resection margins^3,4,5. To eradicate residual cancer cells, chemotherapy is often performed as one of the most important adjuvant therapies^6,7. Certain chemotherapeutic agents can also serve as immunogenic apoptosis inducers capable of initiating antitumor immune responses, thereby offering a promising strategy for enhancing tumor immunotherapy^8,9,10. Damage-associated molecular patterns (DAMPs), such as calreticulin (CRT), high mobility group protein 1 (HMGB1), and adenosine 5’-triphosphate (ATP), generated by the cancer cells undergoing immunogenic apoptosis, can promote antigen-presenting cell (APC) maturation and activation^10,11. The matured or activated APCs can, on one hand, express and secrete cytotoxic factors such as tumor necrosis factors (TNFs) and proinflammatory factors. On the other hand, they can present tumor-associated antigens (TAAs) to T cells through the major histocompatibility complex (MHC) on cell membranes, thereby inducing T cell activation and proliferation^10,12. The priming effector T cells can infiltrate into the residual tumor tissue post-resection, thereby exhibiting recognition and eradication of lingering cancer cells¹³. However, solely relying on the immunogenic apoptosis of cancer cells is inadequate to achieve an optimal therapeutic outcome^14,15. It is imperative to identify and develop strategies to enhance the antitumor immune response and effectively suppress postsurgical tumor recurrence and metastasis.

Recent research has demonstrated that activating the stimulator of interferon genes (STING) signaling pathway in cancer cells or APCs can augment tumor immunogenicity and efficiently elicit innate and adaptive antitumor immune responses^14,15,16,17. As prototypical STING agonists, cyclic dinucleotides (CDNs) can bind to STING and trigger a signaling cascade reaction to produce type I interferons (I-IFNs, particularly IFNβ), which can mature or activate APCs^18,19,20. However, natural CDNs are negatively charged and highly hydrophilic, resulting in poor cell membrane permeability, which disfavors their internalization by cancer calls or APCs. Furthermore, CDNs are susceptible to phosphodiesterase that can hydrolyze and inactivate them, thus constraining their effectiveness in immune activation^21,22. Given this, great endeavors have been dedicated to the exploration and synthesis of non-nucleotide STING agonists over the past few years^17,22,23. Recently, Addona’s group reported the discovery of MSA-2, a previously unidentified non-nucleotide small molecule STING agonist¹⁷. The carboxy group in MSA-2 can undergo acid-responsive protonation and hydrophobization, enabling specific interaction with cell membranes and efficient cellular uptake within the subacid tumor microenvironment (TME). Although the internalized MSA-2 can effectively trigger STING activation and lead to rapid induction of substantial IFNβ expression, its duration of action is limited, which can be evidenced by the observation that intratumor levels of IFNβ peak within 2 h to 4 h post-administration but subsequently decline rapidly, returning close to the initial level within one day¹⁷. The above-mentioned scenarios provide compelling evidence supporting the notion that implementing a combined treatment strategy incorporating MSA-2 and immunogenic apoptosis inducers has substantial potential as a strategic approach to stimulate an enhanced antitumor immune response. Concretely, abundant IFNβ proteins secreted by the cancer cells and APCs with the intracellular STING pathway activated by MSA-2, in conjunction with TAAs and DAMPs derived from the doxorubicin (DOX)-induced immunogenically apoptotic cancer cells, could more strongly mature and activate APCs, thereby causing efficient immune responses against tumors²⁴. This hypothesis was proposed prior to the current study, while we have discovered significantly more intriguing phenomena and unexpected collaborative mechanisms between MSA-2 and DOX, which can elicit surprising and enduring antitumor immune responses.

Previous studies have reported that following a single-dose administration of STING agonists, the intratumor IFNβ level can exhibit an initial increase within several hours but subsequently revert to the baseline within 24 hours^17,22. However, the long-term administration of multiple doses to maintain drug concentration may augment the susceptibility to infection and exacerbate injection-related discomfort, thereby compromising patient compliance. Moreover, the application of systemic administrations has been constrained by the adverse effects associated with normal tissue or organ toxicity and inadequate drug accessibility at the tumor site^25,26. The persistent maintenance of antitumor immune responses at the targeted site poses another significant challenge. Therefore, there is a demand for the development of localized therapeutic platforms that possess favorable drug-loading capacity and exhibit sustained drug-release properties. Promising advancements in three-dimensional (3D) printing techniques have facilitated the fabrication of implantable scaffolds, comprising biocompatible materials such as collagen, sodium alginate (SA), gelatin, and others. These techniques have demonstrated remarkable success in various biomedical applications^{3,27,28,29,30}. The fabrication of 3D-printed drug-loaded scaffolds enables precise construction and dosage control³¹. Additionally, 3D-printed scaffolds can release drugs in a controlled manner that effectively extends the duration of drug efficacy in tumor treatment^3,32. Nevertheless, a single drug-carried platform alone is insufficient to achieve the desired therapeutic effect, particularly when a sequential release of dual or multiple drugs is required. Electrospun fiber (patch), another drug-loaded polymer-based platform also renowned for its exceptional drug-loading capacity and precise control over drug release, has been employed in the in-situ administration^5,33. Besides, both 3D-printed scaffolds and electrospun fibers have been demonstrated to offer a certain degree of protection against in-vivo enzymatic degradation for loaded drugs³⁴. Therefore, an implantable composite composed of 3D-printed scaffolds and electrospun fibers can effectively encapsulate multiple drugs with distinct release profiles, enabling personalized customization based on drug functionalities and therapeutic regimens.

In this work, we present a sandwich-like dual-drug depot (MS-DF-MS) consisting of two layers of MSA-2-loaded 3D-printed scaffolds and one patch of DOX-included electrospun fibers, which synergistically enhances the long-lasting antitumor immune responses and suppresses postoperative tumor recurrence and metastasis based on efficient chemotherapeutic and immunotherapeutic capabilities (Fig. 1). The 3D-printed scaffold, composed of Ca²⁺-linked networks of gelatin and SA encapsuling MSA-2, is constructed to confer the platform with innate immune activation ability through STING agonism. The DOX-loaded poly (lactic-co-glycolic acid) (PLGA) electrospun fibers are incorporated into the 3D-printed scaffolds in a sandwich-like configuration. By leveraging the physical barrier effect provided by the exterior scaffolds and capitalizing on the disparity in solubility, DOX can be consistently released at a slower rate compared to MSA-2 from this dual-drug depot. Surprisingly, we discover that the preferentially released MSA-2 not only effectively activates the STING-IFNβ pathways in cancer cells and APCs but also simultaneously facilitates endoplasmic reticulum stress (ERS)-mediated and general immunogenic apoptosis of cancer cells, thereby enhancing the generation of DAMPs. Through achieving sustained release of DOX, it can concurrently exert direct cytotoxicity and induce immunogenic apoptosis in cancer cells. In addition, the DOX released in a hysteretic manner also exhibits a potentiation of the STING-IFNβ pathway activation, potentially ascribed to the liberation of endogenous dsDNA fragments, such as mitochondrial DNA (mtDNA)³⁵. The most important finding in our study is the sustained activation of the STING-IFNβ pathway in cancer cells and APCs, effectively induced by the enhanced microtubule depolymerization, thus addressing the limitations associated with short-lasting duration observed when using STING agonists alone, as well as the poor efficiency of immune activation found with immunogenic apoptosis inducers employed alone. This mechanism facilitates a persistent antitumor immune response following the sequential release of MSA-2 and DOX from our developed dual-drug depots. Collectively, our engineered and manufactured dual-drug depots can be surgically implanted at the tumor resection site to initiate a sequential release of STING agonists and immunogenic apoptosis inducers, thereby facilitating a cascaded self-reinforcement of long-term immunotherapy. This approach offers an efficient strategy for suppressing postsurgical tumor recurrence and metastasis.

Fig. 1: Schematic illustration of dual-drug depots for preventing postoperative tumor recurrence.

a Preparation of the DOX-loaded electrospun fibers. b Fabrication of the dual-drug depots composed of two pieces of MSA-2-loaded 3D-printed scaffolds sandwiching one patch of DOX-loaded electrospun fibers. c Tumor excision with positive margins, dual-drug depot implantation, controlled drug release, and mechanisms for the cascade-reinforced cycle of immunotherapy and chemotherapy inhibiting postoperative tumor recurrence.

Results

Characterization and properties of dual-drug depots

The dual-drug depots were constructed by initially fabricating DOX-loaded PLGA electrospun fibers using an electrospinning device (Fig. 1a). Subsequently, the bioink consisting of gelation, CaCl₂, SA, and MSA-2 was used for the fabrication of scaffolds enabling the controlled delivery of MSA-2 via a 3D printing extrusion system. Finally, one fabricated patch of electrospun fibers was interposed between two layers of 3D-printed scaffolds (Fig. 1b). Figure 2a visually demonstrates the successful fabrications of DOX-loaded electrospun fiber patch, shape-stable MSA-2-loaded 3D-printed scaffolds, and sandwich-like dual-drug depots consisting of a DOX-loaded fiber patch as the inner layer and two MSA-2-loaded scaffolds as the outer layers, as evidenced by the digital photographs captured from both top and side views. The characteristic hues of orange and light yellow effectively demonstrate the successful and uniform loading of DOX and MSA-2, respectively. Following fabrications, the microtopographies of DOX-loaded fibers, MSA-2-loaded scaffolds, and dual-drug depots were observed using scanning electron microscopy (SEM), and the representative SEM images are presented in Fig. 2b. The microstructure of 3D-printed scaffolds displays a porous architecture with pore sizes ranging from 10 to 60 μm, which is attributed to the abundant molecular networks of the bioink used in printing and provides high interfacial areas for MSA-2 release. The inner patch consists of interlaced fibers with diameters ranging from 200 to 800 nm and averaging around 500 nm (Fig. 2c). Fluorescence microscopy technique was also employed to validate the uniformity in size and successful DOX loading of the electrospun fibers (Fig. 2d).

Fig. 2: Properties of dual-drug depots.

Schematic diagrams, digital photographs (a), and representative SEM images (b) of the electrospun fiber patch, the 3D-printed scaffold, and the pseudo-color scaffold-fiber-scaffold sandwich-like composite (yellow, 3D-printed scaffolds; red, electrospun fibers) (n = 3 independent experiments in b). c Diameter distribution of electrospun fibers calculated based on SEM images. d Representative fluorescence images of DOX-loaded electrospun fibers (n = 3 independent experiments). Rheological behaviors (e), thixotropies (solid squares, G’, storage modulus; hollow squares, G”, loss modulus) (f), and Tanδ values (the ratio of G”/G’) (g) of the bioink with or without Ca²⁺ or MSA-2 as a function of oscillatory strain at a constant shear frequency of 1 Hz. Relative cumulative release profiles of MSA-2 (orange dots) (h) and DOX (red dots) (i) from dual-drug depots at pH 7.4 (solid dots) (j) or pH 6.5 (hollow dots) (k) (n = 3 independent experiments). Data are shown as mean ± SD. Source data are provided as a Source Data file.

The optimal formulas of bioinks were determined through an investigation into the rheological properties prior to the advent of 3D printing. The pure gelatin solution exhibits a viscosity-predominating gelation behavior at a relatively higher concentration of 6.25% (w/v), as illustrated in Supplementary Fig. 1a, wherein the storage modulus (G’) is lower than the corresponding loss modulus (G”). With the addition of SA, G’ surpasses G” and reaches its maximum value at the SA concentration of 5.25% (w/v), followed by an unexpected decline at an SA concentration of 5.50% (Supplementary Fig. 1b). However, the overall viscoelasticity remains low, making it unsuitable for the preparation of bioink for 3D printing. To further improve the rheological properties, Ca²⁺, a typical divalent cation capable of electrostatically crosslinking anionic polysaccharide or protein molecules, was introduced. Supplementary Fig. 1c demonstrates a continuous increase in both G’ and G” until the Ca²⁺ concentration reaches 0.15% (w/v), indicating suitable viscoelasticity for 3D printing. In addition, Supplementary Fig. 1d presents the gelation behaviors of bioinks containing gelatin and SA, regardless of the presence of Ca²⁺ or MSA-2, with elasticity dominating. This is evidenced by that G’ of each tested bioink is higher than the corresponding G”, and the corresponding Tanδ values (the ratio of G”/G’) consistently remain below 1, across a wide range of shear frequencies spanning two orders of magnitude from 0.1 to 10 Hz. Moreover, the G’ and G” values of all the tested bioinks rise with the increasing shear frequency. Especially, the Tanδ value of the bioink without Ca²⁺ decreases dramatically as the shear frequency increases (Supplementary Fig. 1e), demonstrating an enhancement in gelation. Thus, all three types of bioink appeared suitable for 3D printing. Given this, the rheological modulus variations were further detected as the oscillatory strain increased from 1% to 200%. The results in Fig. 2e demonstrate that, as the oscillatory strain continuously increases, the G” progressively surpasses G’ for each tested bioink, indicating the onset of viscosity-predominating gelation behaviors. Similar variation trends are presented in Supplementary Fig. 1c, wherein all the Tanδ values gradually converge towards and even exceed 1, indicating the emerging dominance of viscosity. In line with the variation trends in Supplementary Fig. 1b, the Tanδ value of the Ca²⁺-deficient bioink consistently exhibits a notably higher level and is closer to 1 compared to the other two types of bioink (Supplementary Fig. 1f). Besides, in the case of the bioink lacking Ca²⁺, the points of intersection between G’ and G” as well as the Tanδ value reaching 1 occur at the lowest shear strain. The introduction of Ca²⁺ was therefore indispensable for the gelation of the printable bioink, whereas MSA-2 had minimal impact on its hydromechanics, rendering it suitable for 3D printing. The shear-thinning property of the bioink under certain strains presents significant potential for injection from a printing syringe, but the post-injection gel restorability is crucial for the formation of hydrogel-based scaffolds. Then, the restorability-revealing thixotropy of MSA-2-contained bioink was measured. As depicted in Fig. 2f, g, a remarkable shear-thinning phenomenon occurs at a high strain of 200%, followed by a rapid recovery of elasticity-dominating viscoelasticity when the strain decreases to 1%. The complex viscosity serves as an indicator of the intra-gel interaction force strength and also demonstrates a similar changing trend of viscoelasticity versus strain (Supplementary Fig. 1g). These results reflect that the extruded bioink can construct shape-stable scaffolds with excellent performance. After the bioink formula determination by rheological measurement, the 3D-printed multilayer scaffolds consisting of interwoven continuous hydrogel sticks were successfully fabricated, and the sandwich-like scaffold-fiber-scaffold composites were also impeccably constructed. (Supplementary Fig. 1h and Fig. 2a). Additionally, the macro- and micro-structures of single-drug or dual-drug depots remain stable after lyophilization (Supplementary Fig. 1h and Fig. 2b).

The lyophilized outer scaffolds are designed to release and deliver MSA-2 through molecular diffusion, which is accelerated by the gradual dissolution or degradation processes such as rehydration, swelling, dissolution, and hydrolysis of soluble or degradable SA and gelatin (Supplementary Fig. 2). Additionally, the inner PLGA fibers facilitate the release and delivery of DOX via molecular diffusion, driven by the progressive hydrolysis of ester bonds (Supplementary Fig. 2). To validate the controlled and sequential release behaviors of the two drugs loaded in sandwich-like composites, the in-vitro release profiles of MSA-2 and DOX at pH 7.4 (simulating the microenvironment of normal tissues) and 6.5 (mimicking the subacid TME) were investigated. As shown in Fig. 2h, i, a delayed increase of relative cumulative release over time for MSA-2 in the outer scaffolds and DOX in the inner fiber patch can be observed. Due to the protonation effect, MSA-2 with a carboxyl group exhibits higher hydrophobicity at pH 6.5 than at pH 7.4, resulting in faster release of MSA-2 at pH 7.4¹⁷. Conversely, DOX with an amidogen group shows higher hydrophobicity at pH 7.4 than at pH 6.5, leading to a faster release of DOX at pH 6.5. However, regardless of pH level (7.4 or 6.5), the absolute amount of MSA-2 released is greater than that of DOX (Fig. 2j, k). The likely cause of this outcome involves two underlying mechanisms, one being the consistently higher water solubility of MSA-2 compared to DOX at pH levels ranging from 6.5 to 7.4, and the other being the physical barrier effect of external scaffolds that restrict the release of DOX within the interior fibers. Collectively, these phenomena indeed demonstrate the effective and controlled release of MSA-2 and DOX from our dexterously designed and constructed dual-drug depots in an ordinal manner.

In-vitro immunogenic apoptosis induction and immune activation

Before evaluating the antitumor effect of our constructed drug depots, we conducted an initial investigation into the cytocompatibility and cytotoxicity of the carrier materials without any drugs. Following the co-incubation with blank fiber patches, blank 3D-printed scaffolds, or blank sandwich-structured scaffold-fiber-scaffold composites (S-F-S) for 24 h or 48 h, B16F10 cells were scarcely impacted, as evidenced by the limited discrepancies observed in the live/dead-stained fluorescence images (Supplementary Fig. 3a) and cell viabilities (Supplementary Fig. 3b) between any blank material-treated groups and control group (without any treatments). After confirming the high cytocompatibility of our used carrier materials, we further validated the antitumor efficacy of drug-loaded sandwich-like composites, i.e., two layers of MSA-2-loaded scaffolds sandwiching one patch of blank fibers (MS-F-MS), two layers of blank scaffolds sandwiching one patch of DOX-loaded fibers (S-DF-S), and two layers of MSA-2-loaded scaffolds sandwiching one patch of DOX-loaded fibers (MS-DF-MS). B16F10 cells were co-incubated with S-F-S, MS-F-MS, S-DF-S, or MS-DF-MS for 24 h (Fig. 3a), and then the live/dead staining (Fig. 3b) and cell viability (Fig. 3c) detection experiments were carried out. The results suggest that all the drug-loaded composites exhibit a certain degree of cytotoxicity toward B16F10 cells. Specifically, MS-F-MS, S-DF-S, and MS-DF-MS demonstrated approximately 25%, 50%, and 75% efficacy in killing cancer cells (Fig. 3c), respectively. The presence of a typical chemotherapeutic drug DOX undoubtedly contributed to the high cytotoxicity exhibited by S-DF-S and MS-DF-MS. According to our previous report, a concentration of 6 μM free DOX in the culture medium was found to induce approximately 75% cancer cell death following a co-incubation for 12 h⁵. In contrast, the treatment with S-DF-S (9 μM DOX in the culture medium) resulted in only about 50% cancer cell death after 24-h treatment, thus confirming the controlled release behavior of DOX loaded within the inner fibers.

Fig. 3: In-vitro STING-IFNβ pathway activation and immunogenic apoptosis.

a Schematic diagram of the experimental design for assessing the effects of drug depots on in-vitro activation of STING-IFNβ pathway in cancer cells and induction of immunogenic apoptosis. Representative calcein AM (live)/PI (dead)-stained fluorescence images (b) and cell viabilities (c) of B16F10 cells treated with indicated composites for 24 h (n = 3 independent experiments). Representative flow cytometry pseudo-color plots (d) and representative immunofluorescence images with quantified 3D surface plots (e) reflecting the fluorescence intensities and distributions of pIRF3 in B16F10 cells treated with indicated composites for 6 h (n = 3 independent experiments). f Representative western blots of crucial proteins separately involved in general apoptosis (BAX), ERS-induced apoptosis (Casp12 and C-Casp12), and STING-IFNβ pathway (pIRF3) in B16F10 cells treated with indicated composites for 6 h (n = 4 independent experiments). Representative immunofluorescence images with quantified 3D surface plots showing the fluorescence intensities and distributions of CRT on cell membranes (g), the release of HMGB1 (h), and the release of ATP (i) of B16F10 cells treated with indicated composites for 12 h (n = 3 independent experiments). j Relative mRNA level of ifnb1 in B16F10 cells treated with indicated composites for 6 h (n = 3 independent experiments). Data are shown as mean ± SD. Statistical significances were evaluated by one-way ANOVA with Tukey’s multiple comparisons post hoc test. Source data are provided as a Source Data file.

Surprisingly, MS-F-MS exhibited significant cytotoxicity against B16F10 cells, and the cell-killing efficiency of MS-DF-MS was roughly 2.5-fold higher than that of S-DF-S (Fig. 3c). Although MSA-2 has been demonstrated as an effective STING agonist that can induce the expression and secretion of IFNβ, thereby activating antitumor immune response, its direct cytotoxic effect on cancer cells remains unreported¹⁷. Nonetheless, The STING-mediated phosphorylation of interferon regulatory factor 3 (IRF3) has been demonstrated to induce B-cell lymphoma-2-associated X (BAX) activation, leading to the subsequent cytochrome C release and ultimately triggering apoptosis³⁶. Moreover, recent reports have also proved a robust correlation between activated STING and ERS, the latter of which typically leads to apoptosis via caspase 12 (Casp12) activation^36,37,38. Taking this into consideration, we estimated the activation of STING-IFNβ pathway in B16F10 cells undergoing the co-incubations with or without MSA-2, using phospho-IRF3 (pIRF3) immunofluorescence staining-based flow cytometry (Fig. 3d and Supplementary Fig. 4) and microscopy (Fig. 3e) techniques. As depicted in Fig. 3d and Supplementary Fig. 4, the B16F10 cells co-incubated with MSA-2-included composites exhibit higher intracellular pIRF3 abundance than those without any treatments or treated with MSA-2-excluded composites. The immunofluorescence images (Fig. 3e) also demonstrate consistent findings, wherein the MS-F-MS and MS-DF-MS groups exhibit a broader distribution and higher intensity of pIRF3⁺ fluorescence. The presence of pIRF3 can be clearly observed surrounding and within the nuclei of B16F10 cells treated with MSA-2-loaded composites (Fig. 3e), indicating their translocation into the nuclei to trigger IFNβ expression. Additionally, consistent results for pIRF3 immunofluorescence were observed in CT26, A549, and MCF7 cells (Supplementary Fig. 5). Then, we employed western blot analysis to estimate the alterations in protein levels of pIRF3, BAX, and cleaved (activated) casp12 (C-Casp12) in the B16F10 cells subjected to various treatments (Supplementary Fig. 6, Fig. 3f, and Supplementary Fig. 7). As presented in Fig. 3f and Supplementary Fig. 7, the pIRF3 abundances in the B16F10 cells treated with composites containing MSA-2 are higher than those in the B16F10 cells co-incubated in the absence of MSA-2, aligning precisely with the canonical STING-IFNβ pathway. Consistent with the inference, the co-incubation with MSA-2-loaded composites led to the upregulations of BAX and C-Casp12 in B16F10 cells, with the highest levels of BAX and C-Casp12 observed in the B16F10 cells treated with MS-DF-MS. Furthermore, consistent patterns of variation in BAX and C-Casp12 were observed across CT26, A549, and MCF7 cells (Supplementary Fig. 8), all of which were subjected to identical treatments as the B16F10 cells.

The drug-loaded composites were further assessed for their abilities to enhance the release of DAMPs such as ATP, CRT, and HMGB1, which are recognized as the markers of immunogenic apoptosis. As anticipated, B16F10 cells treated with DOX-contained composites exhibit prominent expression of CRT on their cell membranes (Fig. 3g and Supplementary Fig. 9), as well as conspicuous releases of HMGB1 (Fig. 3h) and ATP (Fig. 3i) into the culture medium supernatants, which is consistent with previous reports demonstrating that DOX is a typical inducer of immunogenic apoptosis^8,9. Unexpectedly, the introduction of MSA-2 further promotes the expression of CRT on cell membranes (Fig. 3g and Supplementary Fig. 9), as well as the releases of HMGB1 (Fig. 3h) and ATP (Fig. 3i). These findings suggest that the STING-mediated apoptosis induced by MSA-2 may also possess immunogenic properties. Besides, similar results were found in CT26, A549, and MCF7 cells (Supplementary Fig. 10) undergoing the same treatment as B16F10 cells. Apart from confirming the successful induction of immunogenic apoptosis, we also investigate the mRNA level of IFNβ in B16F10 cells co-incubated with different composites to further validate the activation of the STING-IFNβ pathway. As presented in Fig. 3j, the treatment with MSA-2-loaded composites indeed increases the intracellular Ifnb1 levels. Additionally, the B16F10 cells undergoing co-incubation with MS-DF-MS exhibit the highest expression of Ifnb1, which is about 30-fold or 150-fold higher than those treated with MS-F-MS or without any composites (Fig. 3j). This implies that the inclusion of DOX or the elicitation of immunogenic apoptosis may trigger or reinforce the STING-IFNβ pathway activation, possibly attributed to the generation and release of endogenous dsDNA fragments, such as mtDNA³⁵.

To investigate the impact of the debris derived from immunogenically apoptotic cancer cells on the maturation of DCs and M1 polarization of macrophages, B16F10 cells were initially co-incubated with various composites, followed by the collection of B16F10 cell debris for the subsequent 24-h treatment of JAWSII or RAW264.7 cells (Fig. 4a). Echoing the findings of immunogenic apoptosis and STING-IFNβ pathway activation presented in Fig. 3d,j, the co-incubation with cell debris derived from the B16F10 cells treated with DOX-integrated composites significantly increased the proportions of CD45⁺CD80⁺ matured DCs (Fig. 4b, d; 9.8% in S-DF-S group and 23.0% in MS-DF-MS group) and CD45⁺CD86⁺ M1 macrophages (Fig. 4c, e; 36.2% in S-DF-S group and 54.8% in MS-DF-MS group) compared to the JAWSII (Fig. 4b, d; 1.5% in S-F-S group and 3.8% in MS-F-MS group) and RAW264.7 (Fig. 4c, e; 19.0% in S-F-S group and 20.6% in MS-F-MS group) cells co-incubated with the B16F10 cell debris gathered after the treatment without DOX. Undeniably, the effects of the cell debris of B16F10 cells treated with the dual-drug depots (MS-DF-MS) on DC maturation and M1 polarization are unparalleled (Fig. 4b–e), which is a rational outcome stemming from the highest levels of immunogenic apoptosis and STING-IFNβ pathway activation in the B16F10 cells treated with MS-DF-MS. Conspicuously, the cell debris of single MSA-2-included composite-treated B16F10 cells does not significantly improve the DC maturation or M1 polarization, indicating the inadequate activation of APCs only stimulated with the IFNβ secreted from STING-IFNβ-activated cancer cells. Due to the ubiquitous expression of STING across various cell types, it is plausible that the MSA-2 released from our implantable drug depots could elicit IFNβ secretion not only in cancer cells but also in APCs^20,39. Actually, numerous studies have chosen to initiate the antitumor immune response by directly stimulating the STING-IFNβ pathway in APCs rather than cancer cells, potentially attributing to these rationales^14,15,40. Taking this into consideration, we further investigated the effect of single MSA-2-contained composites (MS-F-MS) on the activation of APCs, both with and without a 6-h treatment employing B16F10 cell debris obtained after the co-incubation with MS-DF-MS, which might possess the highest capacity to induce immunogenic cancer cell debris among all drug depots (Fig. 4a). As presented in Fig. 4f, g, no significant differences can be observed in the mRNA levels of IFNβ in JAWSII or THP1 cells under basal conditions or following the treatment with S-F-S or S-F-S plus B16F10 cell debris. These findings suggest that while the stimulation with cell debris derived from immunogenically apoptotic cancer cells can mature DCs or polarized macrophages, it has limited effectiveness in activating the STING-IFNβ pathway in APCs. As anticipated, the treatment with MS-F-MS facilitates the expressions of Ifnb1 in JAWSII cells and IFNB1 in THP1 cells by approximately 5 folds, compared to the treatments excluding MSA-2 (Fig. 4f, g). Unanticipatedly, after the supplementation of immunogenic B16F10 cell debris hardly individually influencing the expression of IFNβ in APCs, the increase in IFNβ mRNA level induced by MS-F-MS is significantly heightened by 2 folds and 4 folds in JAWSII cells (Fig. 4f) and THP1 cells (Fig. 4g), respectively. One of the latest related reports has demonstrated that the activation of NF-κB can inhibit the intracellular degradation of activated STING, thereby enhancing signaling pathways related to STING³⁹. NF-κB is a typical proinflammatory cytokine playing a pivotal role in immune response and can be activated by the signals transduced by toll-like receptors (TLRs), some of which can be activated via binding TAAs and DAMPs^41,42,43. To elucidate the intriguing finding, we investigated the suppression effect of immunogenic cancer cells on the intracellular degradation of activated STING in APCs. The results show that the co-incubation with MS-F-MS and immunogenic B16F10 cell debris efficiently inhibits the degradation of STING activated by the released MSA-2 in JAWSII (Fig. 4h, i) and THP1 (Fig. 4j, k) cells, indicating the potential for achieving a latent long-lasting antitumor immune response through dual-drug depot implantation. The promising results obtained using classic monocyte lines have encouraged us to further investigate using murine bone marrow-derived dendritic cells (BMDCs) and macrophages (BMDMs). The corresponding results are presented in Supplementary Fig. 11, wherein BMDCs and BMDMs demonstrate comparable outcomes to JAWSII and RAW264.7 cells, respectively, thereby significantly reinforcing the aforementioned findings.

Fig. 4: In-vitro activation of immune response.

a Schematic diagram illustrating the experimental operation of evaluating the activation or polarization of APCs treated with B16F10 cell debris generated with the co-incubation of drug depots in the presence or absence of MS-F-MS. Representative flow cytometry zebra plots (b, c) and quantifications (d, e) of DC maturation assessed using JAWSII cells (b, d) and macrophage polarization estimated employing RAW264.7 cells (c, e) undergoing indicated treatments for 24 h (n = 3 independent experiments). Relative mRNA levels of (f) ifnb1 in JAWSII cells and (g) IFNB1 in THP1 cells subjected to indicated treatments for 6 h (n = 3 independent experiments). Representative flow cytometry half-overlapped histograms (h, j) and quantifications (i. k) of STING⁺ fluorescence in JAWSII (h, i) or THP1 (j, k) cells undergoing indicated treatments for 6 h (n = 3 independent experiments). Data are expressed as mean ± SD. Statistical significances were estimated by one-way ANOVA with Tukey’s multiple comparisons post hoc test. Source data are provided as a Source Data file.

In-vivo antitumor immune response following implantation of dual-drug depots

The in-vitro results prompted us to further investigate the potential of MS-DF-MS in eliciting a robust in-vivo antitumor immune response. Before conducting in-vivo experiments, we verified the biodegradability of our sandwich-structured implants, as depicted in Supplementary Fig. 2c–f. This not only ensures biosafety but also guarantees effective drug release and delivery. As illustrated in Fig. 5a, a tumor-bearing mouse model was established by the subcutaneous injection of B16F10 cells. Once the tumor volumes reached about 500 mm³, orthotopic implantation of MS-DF-MS composites was performed at the tumor site. Following the treatment for varying durations, the tumors were harvested and analyzed to evaluate the efficacy of antitumor immune response. As shown in Fig. 5b, the protein level of IFNβ exhibited a continuous and remarkable increase from 7.2 ng g^–1 to 26.1 ng g^–1 in tumors during the initial 12 h of treatment with MS-DF-MS, followed by a subsequent decrease to a relatively stable level at approximately 12.0 ng g^–1 in tumors that remains higher than the untreated group (0 h). Of note, the literature reporting MSA-2 as a STING agonist has demonstrated that the intratumor injection of free MSA-2 can effectively induce the expression of IFNβ in tumors, with a gradual increase within 2 h to 4 h post-administration and a return to the baseline level within 24 h¹⁷. However, our findings demonstrated that the designed and customized MS-DF-MS implants effectively delayed the time of peak IFNβ levels in tumors by 3 to 6 times, which occurred after 12 h post-implantation. Furthermore, the sustained release of MSA-2 facilitated prolonged intratumor activation of the STING-IFNβ pathway, maintaining elevated levels of intratumor IFNβ even up to 48 h post-implantation. Due to the activation of STING-IFNβ pathway and the stimulation with TAAs and DAMPs derived from the immunogenically apoptotic cancer cells, the immunosuppressive TME can undergo remodeling featured by the maturation or polarization of APCs^14,15,21,40. The innate immune-activated TME enables effective recruitment and activation of T cells, subsequently inducing their priming and proliferation to enter the effector phase characterized by the robust secretion of interferon γ (IFNγ)^14,15,40. As presented in Fig. 5c, the implantation of MS-DF-MS results in a continuous increase of IFNγ level from 5.7 ng g^–1 to 23.1 ng g^–1 in tumors within the 48-h treatment period, indicating a successful and strong adaptive immune response. Additionally, the high-abundance IFNγ could not only exhibit cytotoxicity towards cancer cells but also contribute to the reconstruction of the intratumor immune microenvironment^13,44. Then, we performed flow cytometry assays to further evaluate the activation of innate and adaptive immune in TME, with the corresponding gating strategies for intratumor DCs, macrophages, and T cells separately shown in Supplementary Fig. 12, Supplementary Fig. 13, and Supplementary Fig. 14. The results depicted in Fig. 5d,f obviously demonstrate a progressive increase in the proportion of matured DCs (CD80⁺CD86⁺) in intratumor DCs from 7.7% to 39.1% within 24 h post-implantation. Subsequently, with the prolonged treatment using MS-DF-MS, the proportion of matured DCs remains relatively stable at a range of 30%–40% in intratumor DCs. Besides, the proportion of MHC-class-II-positive (MHCII⁺) DCs in intratumor DCs exhibits a steady increase, rising from 4.2% to 33.3%, with an incremental rate of approximately 10% as the treatment duration progresses from 0 h to 12 h, 24 h, and finally reaching 48 h (Fig. 5e, g). These findings indicate the effective initiation of the antigen-presenting process in intratumor DCs. Similarly, the M1 polarization (CD86⁺) proportion in intratumor macrophages displays a significant time-dependent increase from 18.4% to 72.0% (Fig. 5h, j), while the variation in the proportion of M2 polarization (CD206⁺) presents an opposite trend evidenced by a sharp decline from 65.2% to 11.7% (Fig. 5i, k), both of which further reflect the successful reshaping of innate immune microenvironment and the activation of APCs in the tumors from mice treated with our dual-drug depots. Consistent with the increasing trend of the intratumor IFNγ abundance, the proportions of CD8⁺ T cells (Fig. 5l, n) and CD4⁺ T cells (Fig. 5m, o) in tumor-infiltrating T cells also exhibit a rise from 7.4% to 42.1% and from 6.0% to 36.7%, respectively. These findings once again suggest the effective activation of intratumor effector T cells that possess specific recognition and elimination capabilities against cancer cells.

Fig. 5: In-vivo antitumor immune response immediately following implantation.

(a Schematic diagram illustrating the B16F10 tumor-bearing mouse mode construction, tumor resection with positive margins, implantation of MS-DF-MS, and treatment for various durations. b, c Secretion levels of IFNβ (b) and IFNγ (c) in tumors from mice treated for indicated time periods (n = 3 mice). Representative flow cytometry zebra images (d, e) and quantified percentages (f, g) of matured DCs (d, f) and MHCII⁺ DCs (e, g) in intratumor DCs (n = 3 mice). Representative flow cytometry zebra images (h, i) and quantified percentages (j, k) of M1 (h, j) or M2 (i, k) polarization of intratumor macrophages (n = 3 mice). Representative flow cytometry zebra images (l, m) and quantified percentages (n, o) of CD8⁺ (l, n) or CD4⁺ (m, o) T cells in tumor-infiltrating T cells from mice undergoing indicated treatments (n = 3 mice). Data are presented as mean ± SD. Statistical significances were assessed by one-way ANOVA with Tukey’s multiple comparisons post hoc test. Source data are provided as a Source Data file.

To further elucidate the intricate mechanisms underlying intratumor activation and potentiation of the STING-IFNβ pathway, we conducted RNA sequencing and transcriptomic analysis on tumor samples obtained from mice in two distinct cohorts: an untreated control group and a treatment group receiving a 12-h MS-DF-MS therapy. The principal component analysis (PCA) in Supplementary Fig. 15a shows a distinct transcriptomic separation between the MS-DF-MS group and the control group along the first principal component (PC1). Also, the heatmap in Supplementary Fig. 15b reveals weaker intergroup correlations but stronger intragroup correlations among the samples. The upset plot including Venn diagrams in Fig. 6a reflects that a total of 20542 genes were co-expressed by these two groups, while 1396 and 1534 genes were exclusively expressed by the MS-DF-MS group and the control group, respectively. The heatmap, which incorporates interactive features and clustering analysis (Fig. 6b), visually demonstrates pronounced disparities in gene expression between these two groups. The differentially expressed genes (DEGs) included in this analysis were determined based on statistical significance (P < 0.05) and fold change criteria (FC > 2 or FC < 0.5). The differential expression analysis was subsequently performed on the DEGs between the MS-DF-MS group and the control group. The volcano plot in Fig. 6c illustrates a total of 608 DEGs, comprising 426 significantly upregulated (P < 0.05 and FC > 2) genes and 182 significantly downregulated (P < 0.05 and FC < 0.5) genes, in the DEGs (MS-DF-MS group compared to control group). Based on this discovery, we carefully selected 20 genes of interest (GOIs) involved in the latent mechanism of immunogenic apoptosis and immune response triggered by implantation with the dual-drug depots, and plotted a heatmap to visually represent the intricate expression variations of these GOIs (Fig. 6d). The expression of Il10 gene, which encodes a characteristic immunosuppressive cytokine predominantly secreted by TAMs or M2 macrophages, is remarkably downregulated in the tumors sampled from the mice treated with MS-DF-MS (Fig. 6c, d). Except for Il10, the remaining genes associated with the activations or enhancements of the STING-IFNβ pathway, innate and adaptive immune response, apoptosis, and necrosis exhibit elevated expression levels in the MS-DF-MS group (Fig. 6c, d). The subsequent analysis is presented in Fig. 6e, wherein a circos plot illustrates two principal enriched pathways involving the 20 genes of interest (GOIs), encompassing “immune and inflammatory response” as well as “cell proliferation, necrosis, and apoptosis”. Additionally, the functional relationships among the proteins encoded by these GOIs are elucidated through a protein-protein interaction (PPI) network (Fig. 6f), which demonstrates that the genes (e.g., Irf7, Ddx58, Tlr3 and Irf9) encoding the proteins with the indirect or direct induction effects on the expression of I-IFNs, as well as the gene Ifnb1 directly encoding IFNβ, play crucial roles and possess abundant interactions with other GOIs during the treatment with MS-DF-MS^45,46,47,48.

Fig. 6: Mechanisms of intratumor STING-IFNβ pathway activation and intensification.

a–f Transcriptomic analysis based on RNA sequencing of B16F10 tumors in the MS-DF-MS group and control group subjected to the treatment for 12 h (n = 3 mice). a Upset plot with Venn diagram illustrating the gene count and overlap across samples. b Interactive heatmap and clustering analysis of DEGs (P < 0.05 and FC > 2 or FC < 0.5). c Volcano plot illustrating significantly upregulated (P < 0.05 and FC > 2) and significantly downregulated (P < 0.05 and FC < 0.5) genes (MS-DF-MS group versus control group) and the GOI in DEGs. Heatmap displaying expression variations (d), circos plot with pathway enrichments (e), and PPI network (f) among the GOI in significantly regulated DEGs. Representative immunofluorescence images (g), fluorescence colocalization profiles with correlation coefficients (R_pearson) (h), and quantifications of R_pearson (i) for STING and lysosome (LAMP1) in B16F10 cells with indicated treatments for 6 h. M1, RAW264.7 cells undergoing co-incubation (24 h) with B16F10 cell debris generated by the treatment (24 h) with MS-DF-MS (n = 3 independent experiments). j–l Representative flow cytometry half-overlapped histograms (j) and quantifications (k) of STING⁺ fluorescence, and representative immunofluorescence images of microtubules (α-tubulin) (l) in B16F10 cells with indicated treatments (n = 3 independent experiments). m Schematic diagram exhibiting mechanisms of the STING signaling enhancement by activating NF-κB to depolymerize microtubules, suppress the STING trafficking into lysosomes, and prevent the STING degradation in cancer cells, dendritic cells, or macrophages. Data are expressed as mean ± SD. Statistical significances were assessed using unpaired two-tailed Student’s t test in a–f and one-way ANOVA with Tukey’s multiple comparisons post hoc test in i and k. Source data are provided as a Source Data file.

It is worth noting that in the tumors treated with MS-DF-MS, there is an observed upregulation of ligands or receptors involved in the NF-κB activation (Fig. 6c–f), which can be initiated by the intracellular signals transduced through TLRs or TNF receptors (TNFRs)^39,49,50,51. Furthermore, the Kyoto encyclopedia of genes and genomes (KEGG) pathway analysis in Supplementary Fig. 16a unequivocally indicates a significant enrichment of the DEGs in the signaling pathways involving TLRs, TNFs, and NF-κB. The most recent report in this field has demonstrated that the NF-κB activation can cause the depolymerization of intracellular microtubes, which function as the essential tracks for guiding activated STING trafficking towards lysosomes and subsequent degradation³⁹. Our previous findings have confirmed that the stimulation with cell debris contained DAMPs and TAAs, produced from the B16F10 cells treated with MS-DF-MS, can activate the signal transduction through TLRs, thus effectively suppressing the degradation of STING in APCs and maintaining a durable innate immune response. Given that NF-κB activation can also be caused by the intracellular signals originating from the interaction between TNFRs and TNFs, the latter of which can be secreted by activated M1 macrophages, we employed M1-polarized RAW264.7 cells to co-incubate B16F10, JAWSII, and RAW264.7 cells in the presence or absence of MS-F-MS for 6 h. Before co-incubation, we generated the B16F10 cell debris through a 24-h treatment with MS-DF-MS, followed by exposing the immunogenic and proinflammatory factor-enriched cell debris to native RAW264.7 cells for an additional 24 h to induce M1 polarization. Subsequently, the intracellular levels of STING in the co-incubated B16F10, JAWSII, and RAW264.7 cells with or without the activation of STING-IFNβ pathway were assessed using immunofluorescence staining-based flow cytometry and microscopy techniques. In addition, the intracellular distributions or existence patterns of lysosomes and microtubules in the co-incubated B16F10 cells were also investigated using the immunofluorescence microscopy technique. As depicted in Fig. 6g, the treatment exclusively with MS-F-MS leads to a heightened abundance of lysosomes in B16F10 cells, consistent with previous studies demonstrating that activated STING can induce autophagy characterized by the extensive formation of lysosomes^35,39. By integrating the STING⁺ and LAMP1⁺ (lysosomes) fluorescence signal distributions in Fig. 6g, along with the colocalization analysis presented in Fig. 6h, i, wherein the colocalization correlation coefficient (R_pearson) of MS-F-MS group is closest to 1.0, it can be inferred that the STING proteins activated by the MSA-2 released from MS-F-MS predominantly localize within lysosomes, indicating their imminent degradation. However, the co-incubation with M1 macrophages effectively prevents the STING degradation by lysosomes, as evidenced by the reduction in lysosome fluorescence signal (Fig. 6g), the segregation of STING⁺ and LAMP1⁺ fluorescence signal distributions (Fig. 6g), and the decrease in R_pearson from 0.647 to 0.289 (Fig. 6h, i). A similar trend of intracellular STING degradation can also be observed in the flow cytometry analysis results (Fig. 6j, k), in which the introduction of M1 macrophages into the co-incubation system of B16F10 cells and MS-F-MS significantly promotes the STING⁺ fluorescence intensity by about 50%. In addition to cancer cells, we also used flow cytometry technique to analyze the alterations in fluorescence intensities of STING in DCs and macrophages undergoing the co-incubation with MS-F-MS, either alone or in combination with M1 macrophages. As shown in Supplementary Fig. 17a–d, the levels of STING in JAWSII and RAW264.7 cells are obviously reduced by the treatment with MSA-2-loaded composites. Nevertheless, the decreasing STING levels noticeably increase following the stimulation with cytokines secreted by M1-polarized RAW264.7 (Supplementary Fig. 17a–d). These robust findings confirm that our designed and constructed implantable drug-loaded composites can effectively address the issue of short-term immune activation caused by the injected free MSA-2, not only through the controlled and sustained drug release behaviors but also through the cascaded and self-reinforced cycle involving immunogenic apoptosis and subsequent immune response.

To gain a more profound and more comprehensive understanding of the integrative impact of MSA-2 and DOX released in a sequential manner from our prepared dual-drug depots on tumors, we performed the gene ontology (GO) (Fig. 7a, b) and KEGG (Supplementary Fig. 16a, b) analyses on the DEGs. As shown in Fig. 7a, the GO terms enriched in the biological processes are related to DNA damage, general apoptosis, ERS-induced apoptosis, and autophagy. Relevantly, the latter two processes align with our findings pertaining to MSA-2 and the self-intensified reconstruction of intratumor immune microenvironment in this study. Besides, the GO terms are also significantly enriched in the biological processes and molecular functions involving the activation of the STING-IFNβ pathway as well as innate and adaptive immune responses (Fig. 7b). The KEGG enrichment analyses in Supplementary Fig. 16a, b not only validate the intratumor activation of NF-κB and the correlations among signaling pathways of TLRs, TNFRs, and NF-κB, thereby consolidating our interferences regarding the inhibitory effect of NF-κB activated through the stimulations with immunogenically apoptotic cancer cells and matured or polarized APCs on the intracellular autogenous degradation of activated STING, but also provide additional support for the findings of GO enrichment analyses. To further elucidate the induction or facilitation effects of our fabricated dual-drug depots on the DNA damage, apoptosis, STING-IFNβ pathway, as well as innate and adaptive immune response in TME, we conducted gene set enrichment analysis (GSEA) on the DEGs. The results presented in Fig. 7c and Supplementary Fig. 16c unequivocally demonstrate that the implantation with MS-DF-MS precisely leads to immunogenic apoptosis, activates STING-IFNβ pathway, and elicits a potent antitumor immune response in tumors.

Fig. 7: Transcriptomic enrichment analyses revealing in-vivo immunogenic apoptosis and antitumor immune response.

a, b GO term enrichments of DEGs in B16F10 tumors from MS-DF-MS group and control group undergoing the treatment for 12 h (n = 3 mice). a GO terms enriched in the biological processes concerned with immunogenic apoptosis. b GO terms enriched in the biological processes and molecular functions associated with antitumor immune response. c GSEA based on Reactome enrichments indicating upregulated pathways (P < 0.05 and NES > 1) involving immunogenic apoptosis and antitumor immune response (n = 3 mice). Statistical significances were assessed using unpaired two-tailed Student’s t test. Source data are provided as a Source Data file.

In-vivo antitumor efficacy and immunogenic apoptosis

The previous findings have confirmed the robust abilities of our drug-loaded implants to induce immunogenic apoptosis in cancer cells and trigger a durable immune response against tumors. Considering these results, we further assessed the efficacy of our drug depots in suppressing postoperative tumor recurrence using a mouse model. As shown in Fig. 8a, the surgical resection is fulfilled on the B16 tumor-bearing mice, when the tumors reached an average volume of around 500 mm³. Specifically, to replicate the clinical scenario of a tumor-positive incisal margin, each tumor was resected with a residual volume of approximately 10 mm³. Subsequently, the orthotopic administration was performed by either injecting free drugs or implanting drug depots, followed by meticulous suturing. The tumor volumes were measured and recorded bidaily. The variation curves of recurrent tumor volumes are depicted in Supplementary Fig. 18a and Fig. 8b, demonstrating a significant suppression of tumor regrowth following the implantation with drug depots containing MSA-2, DOX, or their combination. Reaffirming the in-vitro findings, S-F-S exhibits no discernible antitumor efficacy. Undoubtedly, the postsurgical implantation with MS-DF-MS displays the strongest inhibition of tumor recurrence among all treatments, including the in-situ injection of free MSA-2 and DOX (free M + D), which only shows limited suppressive effect on tumor regrowth possibly due to the fast drug diffusion and metabolism (Supplementary Fig. 18a and Fig. 8b). After a 10-day treatment period, the tumors and major organs were sampled for subsequent investigations, and the mice that exhibited complete tumor eradication by MS-DF-MS were used to establish a rechallenge model through subcutaneously inoculating B16F10 tumors into the contralateral sides of orthotopic tumors. The tumor volumes of mice subjected to the rechallenge experiment were also recorded bidaily for a duration of 30 days, during which the mice with tumor volumes exceeding 1500 mm³ would be euthanatized and included in the calculation of survival rates. The tumors inoculated in the mice treated with MS-DF-MS exhibit significant inhibition, as shown in Fig. 8c. After a 20-day post-rechallenge period, these tumors only experience an average volume increase of around 36 mm³. In contrast, the tumor volumes of the control group, consisting of age-matched normal mice, are approximately 40-fold larger than those of the MS-DF-MS group, approaching 1500 mm³ on day 20 after rechallenging (Fig. 8c). The survival curves depicted in Fig. 8d exhibit similar patterns, illustrating a complete absence of survival (0%) for the control group, while the MS-DF-MS group achieves full survival (100%) at the end of the recording period. These findings clearly demonstrate that our developed dual-drug depots possess the ability to initiate a durable antitumor immune response, thereby effectively preventing tumor recurrence following surgical resection.

Fig. 8: In-vivo antitumor efficacy and immunogenic apoptosis.

a Schematic diagram displaying the animal experimental procedures including B16F10 tumor-bearing mouse model construction, tumor resection with positive margins, in-situ implantation, sample harvest and analysis, and rechallenge by inoculating B16F10 tumors into the contralateral sides of orthotopic tumors. b Variations of orthotopic tumor volumes recorded every two days for 10 days following the surgery (n = 5 mice). Variations of contralateral tumor volumes (c) and survival rates (d) recorded bidaily for 20 days post-rechallenge (n = 3 mice). e Representative bright-field H&E-stained images, and representative Ki67, CRT, and HMGB1 immunofluorescence images of tumor sections from mice subjected to indicated treatments (n = 3 mice). Quantified percentages of necrosis area (f) and Ki67⁺ cells (g), and fold changes of CRT⁺ (h) or HMGB1⁺ (i) area per cell in tumor determined using tumor section images of indicated groups (n = 3 mice). Data are shown as mean ± SD. Statistical significances were evaluated through two-way (b and c) or one-way ANOVA (f–i) with Tukey’s multiple comparisons post hoc test, or log-rank (Mantel-Cox) test (d). Source data are provided as a Source Data file.

To further assess the effects of different treatments on tumor recurrence, the harvested tumors were subjected to histological examination using H&E staining, TUNEL labeling, as well as for Ki67 immunofluorescence staining. The necrosis rate in the tumor sections of the free M + D group is significantly higher (23.3%) compared to that of the control (10.7%) and S-F-S (9.7%) groups, as depicted in Fig. 8e, f, diverging from the tumor volume variation trends observed in Fig. 8b. Moreover, an additional elevation in the necrosis rate is noted among the groups implanted with drug depots, wherein the MS-DF-MS group exhibits particularly conspicuous tumor necrosis (66.9%). The Ki67-indicated (Fig. 8e, g) and TUNEL-labeled (Supplementary Fig. 18b) images, separately reflecting the cell proliferation and apoptosis in tumors, also demonstrate consistent trends found in Fig. 8e, g and Supplementary Fig. 18b, wherein the least proliferation and the most apoptosis are simultaneously displayed by the group treated with MS-DF-MS. To elucidate the underlying cause of the superior antitumor effect of MS-SF-MS compared to free M + D, we conducted separate measurements of drug concentrations in both tumor tissue and blood at specific time intervals. As depicted in Supplementary Fig. 19, the concentrations of MSA-2 and DOX in both tumor and blood reach their peak values one day after in-situ injection, followed by a sharp decrease of over 50%. By day 10 prior to sampling, the concentrations of MSA-2 and DOX in both tumor and blood within the free M + D group return to baseline levels, indicating limited drug retention and rapid drug metabolism. In contrast, the concentrations of MSA-2 and DOX in both tumor and blood within the MS-DF-MS group exhibit a progressive upward trend over time post-implantation, suggesting a sustained antitumor effect leading to a more remarkable outcome.

The potency of such treatments in inducing tumor immunogenic apoptosis was further assessed through immunofluorescence staining using antibodies specific for CRT (Fig. 8e, h) and HMGB1 (Fig. 8e, i). It can be found that the intratumor expression of CRT emitting the signal of “eat me” is predominantly upregulated by the treatments with implants containing DOX, echoing the observations presented in Fig. 2g, certainly with the highest level observed in the group treated with implants simultaneously incorporating DOX and MSA-2 (Fig. 8e, h). The fluorescence signals of HMGB1 in the tumor sections also exhibit consistent alterations, aligning with the cell experiment results shown in Fig. 2h. The treatments involving either MSA-2 or DOX significantly reduce the abundance of HMGB1 because of its leakage from immunogenically apoptotic cancer cells (Fig. 8e, i). Notably, the treatment with MS-DF-MS almost eliminates intratumor HMGB1. These findings suggest that our developed dual-drug depots possess remarkable potential in inducing tumor immunogenic apoptosis and effectively suppressing tumor recurrence.

In-vivo durable immune response against tumors

To further evaluate the reinforcement of long-lasting immune response against tumors, we employed immunofluorescence staining-based microscopy and flow cytometry techniques to analyze the immune-related indices of the sampled tumors and spleens. In Fig. 9a, b, it can be found that the MS-DF-MS group exhibits the highest level of MHC class I (MHCI) expression in tumors, primarily localized on cancer cell membranes for presenting antigens to T cells and thereby eliciting precise immune attack. Previous studies have demonstrated that the activated STING can enhance autophagy, further leading to the degradation of MHCI^35,52. Consequently, the intratumor expression abundance of MHCI in the MS-F-MS group exhibits a reduction of approximately 15% compared to that observed in the S-F-S group (Fig. 9b). However, the introduction of DOX not only independently upregulate MHCI but also synergistically promote its expression with MSA-2, as supported by the observation that the level of MHCI in tumors from the mice treated with MS-DF-MS remains 1.6-fold higher than the sum of intratumor MHCI levels observed in the MS-F-MS and S-DF-S groups (Fig. 9b).

Fig. 9: In-vivo long-term immune response against tumors.

Representative immunofluorescence images (a) and quantified fold changes (b) of MHCI in tumor sections from mice undergoing indicated treatments (n = 3 mice). Quantified percentages of matured (c) or MHCII⁺ (d) DCs in intratumor DCs, polarized M1 (e) or M2 (f) in intratumor macrophages, and CD8⁺ (g) or CD4⁺ (h) T cells in tumor-infiltrating T cells determined through flow cytometry analyses of tumors from mice in indicated groups (n = 3 mice). i–m Representative flow cytometry zebra images (i, k) and quantified percentages (j, l, and m) of CD8⁺ T cells (i, j), and T_EM (CD44⁺CD62^–) (k, l) or T_CM (CD44⁺CD62⁺) (k, m) cells in intrasplenic T cells (n = 3 mice). Data are presented as the mean ± SD. Statistical significances were assessed using one-way ANOVA with Tukey’s multiple comparisons post-hoc test. Source data are provided as a Source Data file.

The relevant gating strategies for flow cytometry analyses performed on the intratumor DCs and macrophages are illustrated in Supplementary Fig. 20a and Supplementary Fig. 21a, respectively. The proportion of matured DCs in intratumor DCs, as depicted in Supplementary Fig. 20b and Fig. 9c, increases from about 24.0% (control and S-F-S groups) to around 40.0% following the 10-day treatment with MSA-2 or DOX via various administration patterns. The proportion of MHCII⁺ DCs in intratumor DCs also exhibits a comparable upward trend, rising from approximately 20.0% to more than 30.0% after the administrations of free M + D, MS-F-MS, S-DF-S, or MS-DF-MS (Supplementary Fig. 20b and Fig. 9d). Furthermore, the MS-DF-MS group concurrently exhibits the highest proportions of matured DCs (67.8%, Fig. 9c) and MHCII⁺ DCs (63.7%, Fig. 9d) among intratumor DC populations. Slightly differently, the MS-F-MS-treated mice display a higher proportion of MHCII⁺ DCs in tumors than the S-DF-S group (Supplementary Fig. 20b and Fig. 9d), potentially attributed to the preferential induction of MHCII expression on DCs through the activation of STING-IFNβ pathway by the initially released MSA-2 rather than the stimulation with TAAs or DAMPs derived from cancer cells. It has been found that the immunogenic apoptosis of cancer cells is predominantly reliant on the subsequent release of DOX from our implants, thereby augmenting the generation of TAAs necessary for the presentation by MHCII on APCs. Thus, there exists a close cooperation between the sequentially released MSA-2 and DOX. Both the increase in M1 polarization (Supplementary Fig. 21b and Fig. 9e) and decrease in M2 polarization (Supplementary Fig. 21b and Fig. 9f) of intratumor macrophages exhibit similar trends to the maturation in intratumor DCs. Specifically, the polarization proportions of M1 and M2 in intratumor macrophages significantly separately vary from around 14.0% to over 34.0% (Fig. 9e) and from more than 68.0% to less than 50.0% (Fig. 9f), following the treatment with different drugs, certainly with the highest proportion (68.8%) of M1 polarization (Fig. 9e) and the lowest proportion (23.8%) of M2 polarization (Fig. 9f) simultaneously observed in the MS-DF-MS group. The results suggest that the implantation of our dual-drug depots can effectively establish an intensified innate immune response against tumors.

After confirming the durable maturation or polarization of intratumor APCs, we subsequently assessed the activation and priming of effector T cells in mice undergoing 10-day treatments. The related gating strategies for flow cytometry analyses conducted on the tumor-infiltrating T cells and intrasplenic T cells are separately depicted in Supplementary Fig. 22a and Supplementary Fig. 23. As presented in Supplementary Fig. 22b and Fig. 9g, the proportion of CD8⁺ T cells in tumor-infiltrating T cells rises from about 7.0% to more than 14.0% upon the administrations involving MSA-2 or DOX in different states. A similar upward trend is evident in the proportion of CD4⁺ T cells among tumor-infiltrating T cells, exhibiting a notable increase from approximately 6.0% to over 16.0%, as illustrated in Supplementary Fig. 22b and Fig. 9h. Consistent with the observed variation trend of MHCII⁺ DC proportion within intratumor DCs, the MS-F-MS group also exhibits an elevated proportion of CD8⁺ T cells among tumor-infiltrating T cells compared to the S-DF-S group (Supplementary Fig. 22b and Fig. 9g), further substantiating the inference that the activation of STING-IFNβ pathway, rather than the stimulation with TAAs or DAMPs from immunogenically apoptotic cancer cells, can more readily induce higher expression of MHCII, which is responsible for the antigen presentation to T cells and the subsequent priming of effector T cells. Undoubtedly, the MS-DF-MS group exhibits a simultaneous presence of the highest proportions of CD8⁺ (40.5%, Fig. 9g) and CD4⁺ (42.5%, Fig. 9h) T cells among tumor-infiltrating T cells. Besides the increase found in CD8⁺ T cells within the tumors from mice treated with MSA-2, DOX, or both, a consistent trend of alteration is also evident in the intrasplenic CD8⁺ T cells, the proportion of which exhibits an augmentation from less than 7.0% to more than 16.0% (Fig. 9i). After 10 days post-implantation with MS-SF-MS, the proportion of CD8⁺ T cells among intrasplenic T cells displays the highest value (42.2%) compared to other treatment groups (Fig. 9j). Additionally, the proportions of CD44⁺CD62L^– effector memory T (T_EM) cells and CD44⁺CD62L⁺ central memory T (T_CM) cells within intrasplenic T cells also demonstrate similar changing patterns (Fig. 9k–m), increasing from approximately 10.0% and around 17.0% to over 12.0% and more than 20.0%, respectively, as presented in Fig. 9k. Also consistently, the highest proportions of T_EM and T_CM cells among intrasplenic T cells are concurrently observed in the MS-DF-MS group. These collective data presented herein suggest that the implantation with our developed drug depots, capable of sequentially releasing MSA-2 and DOX, can elicit robust and enduring antitumor adaptive immune responses, thereby effectively suppressing postoperative tumor recurrence and metastasis.

To significantly demonstrate the pivotal role of drug release sequence in therapeutic efficacy, we developed another implantable dual-drug depot, DF-MS-DF. This implant differs entirely from MS-DF-MS and consists of two patches of electrospun fibers loaded with DOX, strategically sandwiching a 3D-printed scaffold loaded with MSA-2 (Supplementary Fig. 24a). The comparative analyses of in-vivo antitumor efficacies and immune response between DF-MS-DF and MS-DF-MS is presented in Supplementary Fig. 24b–o and Supplementary Figs. 25–27 (gating strategies in flow cytometry analysis). It is evident that MS-DF-MS exhibits greater abilities to inhibit tumor recurrence and elicit antitumor immune activation compared to DF-MS-DF. This superiority reflects the enhanced antitumor effect resulting from preferential release of MSA-2 from the 3D-printed scaffolds.

It is imperative to assess the in-vivo biosafety of the in-situ treatments subsequent to surgical resection of tumors. Alongside tumor volume measurements, the body weights of mice undergoing surgery were also recorded. As shown in Supplementary Fig. 28a, the body weights of all mice exhibit consistent and slightly increasing trends during the 10-day treatment period, with no significant differences observed among the control, S-F-S, free M + D, MS-F-MS, S-DF-S, MS-F-MS, and MS-DF-MS groups. The blood biochemistry indices, including alanine transaminase (ALT, Supplementary Fig. 28b), uric acid (UA, Supplementary Fig. 28c), urea (Supplementary Fig. 28d), creatinine (CREA, Supplementary Fig. 28e), and creatine kinase MB (CKMB, Supplementary Fig. 28f), demonstrate a lack of significant variations among these experimental groups. Specifically, ALT serves as an indicator of liver function, while UA, urea, and CREA are biomarkers reflecting kidney function. Additionally, CKMB is used to evaluate cardiac function. The histological assessments, presented in Supplementary Fig. 29, reveal few abnormalities or necrotic changes in the H&E-stained tissue sections obtained from the major organs, including heart, liver, spleen, lung, and kidney of mice undergoing various treatments. These findings demonstrate that our implemented treatments, particularly the in-situ implantation with MS-DF-MS, exhibit minimal toxicological or adverse effects in vivo.

In-vivo antitumor efficacy and immune response against metastatic tumors

The promising results regarding the inhibition of orthotopic tumor recurrence through cancer cell immunogenic apoptosis and the induction of antitumor activation have strongly motivated us to further investigate the therapeutic efficacy on metastatic tumors, leveraging the long-lasting antitumor immune response. We established a mouse model with co-bearing orthotopic and contralateral B16F10 tumors, as depicted in Fig. 10a, and administered identical treatments to the orthotopic B16F10 tumor-bearing mice. Supplementary Fig. 30a and Fig. 10b illustrate the variation curves of distant tumor volumes, demonstrating a significant inhibition of distant tumor growth after the implantation of drug depots containing MSA-2, DOX, or their combination into the resection site of primary tumors. As anticipated, the orthotopic implantation with MS-DF-MS exhibits the most potent inhibition of distant tumor growth, surpassing all other treatments, including intratumor injection of free M + D, which shows limited inhibitory influence on distant tumor growth due to rapid drug diffusion and metabolism (Supplementary Fig. 30a and Fig. 10b). To further evaluate the antitumor efficacy of different treatments on distant tumor growth, the sampled contralateral tumors underwent histological examination using H&E staining, Ki67 immunofluorescence staining, and TUNEL staining. The necrosis rate in the distant tumor sections of the free M + D group (22.8%) is comparable to that of the control (17.1%) or S-F-S (17.4%) group, while a significant increase in the necrosis rate is observed among the groups implanted with drug depots. The MS-DF-MS group shows the highest tumor necrosis rate (37.5%). The Ki67-immunofluorescence stained (Fig. 10c, e) and TUNEL-labeled (Supplementary Fig. 30b) images consistently demonstrate that the group treated with MS-DF-MS exhibits the least proliferation and the most apoptosis. Besides, the biosafety of our treatments on mice was assessed and confirmed in this mouse model (Supplementary Fig. 31).

Fig. 10: In-vivo antitumor efficacy and immune response against metastatic tumors.

a Schematic diagram illustrating the experimental procedures involving the establishment of orthotopic and contralateral B16F10 tumors co-bearing mouse model, surgical resection of orthotopic tumors with positive margins followed by in-situ implantation, and subsequent sample collection and analysis. b Variations of contralateral tumor volumes recorded at two-day intervals for 10 days post-surgery (n = 5 mice). c Representative bright-field H&E-stained images and Ki67 immunofluorescence images of contralateral tumor sections from mice subjected to indicated treatments (n = 3 mice). Quantified percentages of necrosis area (d) and Ki67⁺ cells (e) in contralateral tumor determined using tumor section images of indicated groups (n = 3 mice). Quantified percentages of matured (f) or MHCII⁺ (g) DCs in intratumor DCs, polarized M1 (h) or M2 (i) in intratumor macrophages, and CD8⁺ (j) or CD4⁺ (k) T cells in tumor-infiltrating T cells determined through flow cytometry analyses of contralateral tumors from mice in indicated groups (n = 3 mice). l Secretion levels of IFNγ in contralateral tumors from mice subjected to the indicated treatments (n = 3 mice). m Schematic diagram exhibiting the experimental procedures, including the construction of orthotopic B16F10 tumors bearing and pulmonary metastasis mouse model, surgical resection of orthotopic tumors with positive margins followed by in-situ implantation, and subsequent sample collection and analysis. Representative digital photographs and bright-field H&E-stained images of lung sections (n) and the corresponding assessment of pulmonary metastasis rate (o) in mice from indicated groups (n = 3 mice). Data are presented as the mean ± SD. Statistical significances were assessed using one-way ANOVA with Tukey’s multiple comparisons post-hoc test. Source data are provided as a Source Data file.

Subsequently, to assess the long-lasting immune response against tumors, we utilized immunofluorescence staining-based microscopy and flow cytometry techniques to analyze immune-related markers in distant tumors. Supplementary Fig. 32a, b clearly show that the MS-DF-MS group displays the highest level of MHCI expression in distant tumors, facilitating antigen presentation to T cells and promoting precise immune attack. By integrating the findings from Fig. 6 and Supplementary Fig. 17, we observe that the activation or proinflammatory polarization of APCs leads to a substantial production of proinflammatory factors, including TNFs, which in turn hinder autophagy in cancer cells and consequently weaken MHCI degradation. The MHCI abundance in distant tumors from MS-F-MS group shows a notable increase, about 11-fold higher than that observed in the control, S-F-S, or free M + D group (Supplementary Fig. 32b). The gating strategies for flow cytometry analyses performed on the DCs and macrophages in distant tumors are presented in Supplementary Fig. 33a and Supplementary Fig. 34a, respectively. The proportion of matured DCs in the DCs in distant tumors increases from around 17.0% to approximately 28% following a 10-day treatment with MSA-2 or DOX via various administration patterns (Supplementary Fig. 33b and Fig. 10f). The proportion of MHCII⁺ DCs in distant tumors also shows an upward trend, increasing from about 15.0% to over 20.0% after the administrations of free M + D, MS-F-MS, S-DF-S, or MS-DF-MS (Supplementary Fig. 33b and Fig. 10g). Moreover, the MS-DF-MS group concurrently displays the highest proportions of matured DCs (37.3%, Fig. 10f) and MHCII⁺ DCs (42.2%, Fig. 10g) among the DC populations in distant tumors. The increase in M1 polarization (Supplementary Fig. 34b and Fig. 10h) as well as the decrease in M2 polarization (Supplementary Fig. 34b and Fig. 10i) of the macrophages in distant tumors demonstrate similar trends to the maturation observed in the DCs within distant tumors. Particularly, the polarization proportions of M1 and M2 in macrophages in the distant tumors significantly change from around 19.0% to over 30.0% (Fig. 10h) and from more than 55.0% to less than 30.0% (Fig. 10i), respectively, following the treatment with different drugs. The MS-DF-MS group simultaneously exhibits the highest proportion of M1 polarization (51.3%) and the lowest proportion of M2 polarization (25.8%). The aforementioned findings demonstrate that the implantation of our dual-drug depots can effectively induce a robust innate immune response, thereby inhibiting metastatic tumors.

Following the confirmation of DC maturation and macrophage M1 polarization, we further evaluated the activation and priming of effector T cells in distant tumors. The relevant gating strategies for flow cytometry analyses performed on the distant tumor-infiltrating T cells are presented in Supplementary Fig. 35a. The proportion of CD8⁺ T cells among distant tumor-infiltrating T cells increases significantly from approximately 12.0% to more than 19.0% following the administrations involving MSA-2 or DOX in different states, as demonstrated in Supplementary Fig. 35b and Fig. 10j. Similarly, there is a noticeable rise in the proportion of CD4⁺ T cells among distant tumor-infiltrating T cells, increasing from around 12.0% to over 25.0%, as depicted in Supplementary Fig. 35b and Fig. 10k. As anticipated, the MS-DF-MS group exhibits the highest proportions of both CD8⁺ (40.7%, Fig. 10j) and CD4⁺ (50.7%, Fig. 10k) T cells among distant tumor-infiltrating T cells. Furthermore, we constructed a pulmonary metastasis mouse model by tail intravenous injection (i.v.) of B16F10 cells 7 days post-inoculation of B16F10 cells (Fig. 10m). As presented in Fig. 10n, the distribution and total area of tumor nodes in the mouse lung decreased with MSA-2 or DOX treatment, with few tumor nodes observed in the lung tissue section from the MS-DF-MS group. Remarkably, the in-situ implantation of MS-DF-MS significantly inhibits the pulmonary metastasis rate by about 70% (Fig. 10o). The data presented herein provide compelling evidence that our developed drug depots, which release MSA-2 and DOX in a sequential manner, can effectively target and attack metastatic tumors by inducing a robust and long-lasting adaptive immune response.

Discussion

Despite notable advancements in surgical techniques, surgical resection remains a persistent clinical challenge for managing solid malignant tumors due to the high incidence of postoperative tumor recurrence and metastasis, primarily attributed to positive resection margins. To effectively eliminate residual cancer cells, chemotherapy is frequently employed as a pivotal adjuvant therapy. Apart from exerting direct cytotoxic effects on cancer cells, specific chemotherapeutic agents like DOX can also function as immunogenic apoptosis inducers, thereby triggering antitumor immune responses. However, the immunogenic apoptosis in tumors alone falls short of achieving a satisfactory therapeutic efficacy. It is imperative to foster and advance approaches that enhance antitumor immune responses and prevent tumor recurrence and metastasis following surgical intervention. Previous studies have unveiled that the activation of the STING-IFNβ pathway in cancer cells or APCs can facilitate the immune-mediated tumor-killing effect. Nevertheless, the limited cell membrane permeability and the susceptibility to hydrolysis mediated by phosphodiesterase greatly restrict the efficacy of common CDN-based STING agonists in immune activation. Consequently, considerable efforts have been devoted to the investigation and synthesis of non-nucleotide STING agonists in recent years. The recently reported non-nucleotide small molecule STING agonist, MSA-2, holds promise for augmenting the antitumor immune response. The unique acid-responsive protonation and hydrophobization of MSA-2 confer it with enhanced cellular uptake and bioaccessibility in the subacid TME, while exhibiting low toxicity and minimal side effects. However, like most STING agonists, administration of free MSA-2 solely elicits an initial activation of antitumor immunity within a few hours but fails to induce a sustained and durable antitumor immune response, potentially attributed to the rapid autophagy of activated STING. At the outset of this study, our primary objective was to simply combine MSA-2 and DOX in a therapeutic regimen, aiming to address the inherent limitations associated with their individual use, leveraging their well-evidenced activities and functions. Remarkably, our investigations have unveiled the unforeseen impacts of MSA-2 and DOX when administered independently for tumor treatment and shed light on the synergistic mechanisms underlying their collaborative actions.

A similar approach, involving the integration of DOX-containing cancer cell-targeted nanoparticles and a classic STING agonist 2’,3’-cGAMP directly co-loaded into an injectable hydrogel, has been reported to yield promising results in the prevention of postoperative recurrence and metastasis²⁴. However, due to its nucleotide nature, 2’,3’-cGAMP is susceptible to biodegradation in the complex extracellular physiological environment or within the endo-lysosomes during internalization. Therefore, we selected the recently reported non-nucleotide small molecule STING agonist MSA-2 to further improve the antitumor effect through immunogen-generated chemotherapy and STING-IFNβ pathway-activated immunotherapy. MSA-2 demonstrates remarkable stability, even when administered orally. A distinguishing feature of MSA-2 compared to traditional STING agonists is its carboxy group, which can undergo acid-responsive protonation and hydrophobization processes, facilitating specific interactions with cell membranes and efficient cellular uptake within the TME¹⁷. Moreover, our drug depot is not a single delivery system; rather, we have designed and constructed a sandwich-like drug depot simultaneously containing MSA-2 and DOX. MSA-2 is included in the two 3D-printed outer layers consisting of gelatin and SA crosslinked with Ca²⁺, while DOX is loaded in one patch of PLGA electrospun fibers as the inner layer. By virtue of the physical barrier effect provided by the exterior scaffolds and the disparity in solubility, DOX can be consistently released at a slower rate compared to MSA-2 from this dual-drug depot (MS-DF-MS). The sandwich-like structure of our implants (featuring a PLGA inner layer surrounded by 3D-printed hydrogel-based outer layers with millimeter-sized pores) permits contact with the release solution or interstitial fluid. Although the physical barriers can partially impede the degradation of PLGA fibers and the release of drugs from the inner layers, they do not completely prevent it. Notably, the release of DOX from the inner fiber patch, while delayed, begins promptly after the release of MSA-2 from the outer layers, effectively initiating a self-reinforced synergistic antitumor effect. In contrast, alternative dual-drug-loaded implants with sequential drug release properties, such as core-shell-structured 3D-printed scaffolds, typically do not initiate subsequent drug release until significant degradation of the materials carrying the initially released drug has occurred. This delay necessitates an extended period for the exclusive release of the preferentially released drug. Moreover, by the time the local cumulative concentration of the subsequently released drug reaches a level sufficient to achieve antitumor effects, the initially released drug may have already been eliminated or metabolized. These factors significantly challenge the synergistic mechanism involving MSA-2 and DOX that we propose and substantiate. The preferentially released MSA-2 can activate STING-IFNβ pathway in the remaining cancer cells and APCs, and its action time can be to some extent prolonged due to its slow-release behavior. In addition, as one of the pivotal discoveries in our study, the released MSA-2 can also lead to both ERS-mediated and general immunogenic apoptosis of cancer cells, the latter of which was originally intended to be triggered solely by the subsequently released DOX. Then, the sustainedly and slowly released DOX can not only directly kill cancer cells through its inherent cytotoxicity but also trigger the immunogenic apoptosis of cancer cells, the latter of which can produce DAMPs and TAAs. The TAAs are then presented by the matured or activated APCs to T cells, thereby initiating a serious of antitumor immune responses. Another important finding in our study is that DOX can also cause the activation of STING-IFNβ pathway in cancer cells, possibly through the release of endogenous fragments of dsDNA, such as mtDNA. The most notable discovery in our study is the synergistic effects and related mechanisms of STING agonists and immunogenic apoptosis inducers, which are released in an ordered manner, on the extended activation of STING-IFNβ pathway in cancer cells and APCs. This combined approach can effectively induce a sustained antitumor immune response. The TAAs and DAMPs derived from immunogenically apoptotic cancer cells, as well as the proinflammatory factors (especially TNFs) secreted by matured DCs and M1-polarized macrophages, can robustly activate the NF-κB signaling pathways in both cancer cells and APCs. This activation leads to the depolymerization of microtubules, which act as crucial tracks for guiding activated STING trafficking towards lysosomes for subsequent degradation.

In summary, our fabricated dual-drug depots enable a sequential release of STING agonists and immunogenic apoptosis inducers at the residual tumor site, serving as an effective adjuvant therapy strategy following surgical excision, and finally initiating a cascade-reinforced cycle of immunotherapy and chemotherapy to effectively inhibit postsurgical tumor recurrence and metastasis. Our developed MS-DF-MS implant can be surgically implanted at the tumor resection site, which is particularly beneficial in cases where complete removal of the tumor is not possible, such as in rectal or breast cancer. In these situations, patients typically receive adjuvant chemotherapy after surgery to eliminate any remaining tumor cells. The dual-drug depot proposed in this study offers an additional treatment option to achieve this objective. However, it may not be suitable for patients with multiple positive lymph nodes or multiple metastases. Moreover, not all cancer types exhibit comparable sensitivity to STING agonists or DOX. Resistance to DOX has been documented in breast cancer, hepatocellular carcinoma, and various other malignancies. The underlying mechanisms involve p53 mutations, ABC transporter amplification, and long non-coding RNAs^53,54. Similarly, resistance to STING agonists has also been observed⁵⁵. Therefore, from a clinical perspective, the proposed drug depots represent an alternative or supplementary treatment choice specifically tailored for a certain subgroup of patients; nevertheless, this subgroup is quite prevalent and could greatly benefit from this therapeutic approach. Future research will primarily focus on further enhancing the efficacy of implantable dual-drug depots to broaden their applicability as an adjunctive therapy for a wider spectrum of post-surgical tumors.

Methods

The animal experiments were conducted in compliance with the Guidelines for Care and Use of Laboratory Animals established by Sun Yat-sen University and received approval (SYSU-IACUC-2023-000513) from the Animal Ethics Committee of Sun Yat-sen University.

Materials

PLGA (#PLG50-10, intrinsic viscosity = 0.7–1.0 dL g^–1) was purchased from Polymtek (Shenzhen, China). Sodium alginate (SA, #S100128), gelatin (#G274269), β-mercaptoethanol (BME, #M301574), penicillin G sodium (#P433335), and streptomycin sulfate (#S432673) were obtained from Aladdin (Shanghai, China). Doxorubicin hydrochloride (DOX, #BD32885) and carboxymethylcellulose (CMC, #BD01294049) were bought from Bidepharm (Shanghai, China). Hexafluoro-isopropylalcohol (HFIP, #B603683), propidium iodide (PI, #B608469), 4’,6-diamidino-2-phenylindole dihydrochloride (DAPI, #B608184), beta-mercaptoethanol (β-ME, #B601534), and phorbol 12-myristate 13-acetate (PMA, #B612870) were purchased from Boer (Shanghai, China). MSA-2 was bought from Topule (Wuhan, China). Mycoplasma detection kits (#D101) were purchased from Vazyme (Nanjing, China). Gibco^TM Roswell Park Memorial Institute-1640 (RPMI-1640) medium (#11875093), Gibco^TM Dulbecco’s modified eagle medium with 4.5 g L^-1 glucose (DMEM, #11965092), Invitrogen^TM Trizol reagent for total RNA isolation (#15596026), eBioscience^TM anti-mouse CD80 (B7-1) antibody labeled with PE (#12-0801-81), and eBioscience^TM anti-mouse CD206 (MMR) antibody labeled with PE (#12-2061-80) were bought from Thermo (USA). Alpha-minimum essential medium (α-MEM, #CL0002) was obtained from SparkJade (Qingdao, China). Murine granulocyte macrophage colony stimulating factor (GM-CSF, #Z03300), murine macrophage colony stimulating factor (M-CSF, #Z03275), and murine interleukin 4 (IL4, #Z02996) were bought from GenScript (Nanjing, China). Certified fetal bovine serum (FBS, #04-001-1) was purchased from Biological Industries (Israel). Rat-derived anti-mouse CD45 monoclonal antibody labeled with PE/Cyanine7 (#103114), rat-derived anti-mouse/human CD11b monoclonal antibody labeled with PerCP/Cyanine5.5 (#101227), rat-derived anti-mouse F4/80 monoclonal antibody labeled with APC (#123115), rat-derived anti-mouse CD86 monoclonal antibody labeled with FITC (#105006), rat-derived anti-mouse CD8a monoclonal antibody labeled with PE (#162303), rat-derived anti-mouse CD4 monoclonal antibody labeled with APC (#100515), rat-derived anti-mouse/human CD44 monoclonal antibody labeled with FITC (#103005), rat-derived and anti-mouse CD62L monoclonal antibody labeled with APC (#104411) were obtained from BioLegend (USA). Rat-derived anti-mouse MHCII monoclonal antibody labeled with FITC (#35-5321) was purchased from Tonbo Biosciences (San Diego, USA). 2 mM calcein-acetoxymethyl ester solution (calcein-AM, #C2012), immunostaining permeabilization solution with triton X-100 (#P0096), rabbit-derived anti-CRT monoclonal primary antibody (#AF1666), donkey-derived anti-rabbit IgG labeled with Alexa Fluor 555 (#A0453), antifade mounting medium with DAPI (#P0131), ATP assay kits (#S0026), radio immunoprecipitation assay (RIPA) lysis buffer containing protease inhibitors (#P0013B), protease and phosphatase inhibitor cocktail for general use (#P1045), 5× (#P0015L) and 1× (#P0015A) sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) protein sample loading buffers, protein ladder with pre-stained color (#P0076), skim milk powder (#P0216), bovine serum albumin (BSA, #ST025), primary antibody dilution buffer (#P0023A), goat-derived anti-rabbit IgG labeled with HRP (#A0208), secondary antibody dilution buffer (#P0023D), rabbit-derived anti-BAX polyclonal antibody (#AF0057), rabbit-derived anti-STING polyclonal antibody (#AF8073), rabbit-derived monoclonal antibody against α-tubulin (#AG0126), rabbit-derived anti-MHCI monoclonal antibody (#AG2146), enhanced chemiluminescence (ECL) kits (#P0018AS), and red blood cell lysis buffer (#C3702) were purchased from Beyotime (Shanghai, China). IFKine^TM Green donkey-derived anti-rabbit IgG (#A24221) was obtained from Abbkine (Wuhan, China), Bicinchoninic acid (BCA) protein assay kits (#BCA02) was bought from Dingguo Biology (Guangzhou, China). Cell counting kits-8 (CCK-8, # K1018) were obtained from ApexBio (USA). Murine HMGB1 ELISA kits (#XY9M0382) were purchased from X-Y Biotechnology (Shanghai, China). Normal donkey serum (#SL050) was bought from Solarbio (Beijing, China). PreScript III reverse transcriptase kits (#R421B) and 2× robust SYBR Green kits (#S018B) were obtained from EnzyValley (Guangzhou, China). Rabbit-derived anti-phospho-IRF3 (pIRF3) polyclonal antibody (#AF2436) was purchased from Affinity Biosciences (Changzhou, China). Rabbit-derived anti-caspase12 (Casp12) monoclonal antibody (#A22864) was obtained from ABclonal (Wuhan, China). Rabbit-derived anti-GAPDH polyclonal antibody (#10494-1-AP), mouse-derived anti-vinculin monoclonal antibody (#66305-1-Ig), and rat-derived anti-mouse LAMP1 monoclonal antibody labeled with APC (#APC-65050) were obtained from Proteintech (Wuhan, China). Powered tris-buffered saline with tween 20 (TBST, #JXF003) was purchased from GenXion (Guangzhou, China). Murine IFNβ (#TAEm-386) and IFNγ (#TAEm-913) ELISA kits were bought from ANRC (Tianjin, China). Hematoxylin&eosin (H&E) staining kits (#G1076), TUNEL cell apoptosis detection kits (#G1502), rabbit-derived anti-Ki67 polyclonal antibody (#GB111141), and goat-derived anti-rabbit IgG labeled with Cyanine 3 (#GB21303) or Alexa Fluor 488 (#GB25303) were purchased from Servicebio (Wuhan, China). Rabbit-derived anti-HMGB1 monoclonal antibody was obtained from Huabio (Hangzhou, China).

Other reagents with analytical purity were used as received. All the water used in our study was ultrapure with an electrical resistivity not exceeding 18.2 MΩ cm^-1, which was meticulously prepared using a laboratory water purification system (Milli-Q Integral 3, Merck Millipore, Germany).

Construction and characterization of dual-drug depots

Manufacturing and characterization of electrospun fibers

An electrospinning device was used to fabricated electrospun fibers as per the methods slightly modified based on our previous reports^3,5,25,27. To be specific, PLGA and HFIP were mixed at a 1:10 weight/volume (w/v) ratio and stirred overnight to acquire a homogeneous solution. Later, DOX (9 μM) was added into the PLGA solution, and the mixture was stirred for about 1 h until a homogeneous viscous solution was obtained. The solution was then electrospun at a constant voltage of 12 kV, a preset flow rate of 2.16 mL h^–1, and an optimized distance of 12 cm between the coaxial needle and the fiber receiving roller. After that HFIP was volatilized adequately, the electrospun fibers were stored under dry conditions for future use. After cutting the desiccated DOX-contained electrospun fiber (DF) patch into the required size, we used a cellphone (iPhone 12pro max, Apple, USA) and a fluorescence microscope (Ni-U, Nikon, Japan) with a Texas red-labeled optical filter to acquire the digital photograph and microscopic fluorescence images, respectively. Meanwhile, the microtopography of DF was investigated with a scanning electron microscope (SEM, Quanta 400 F, Oxford, UK). Then, the diameter of fibers was determined using the software Fiji (Version 1.53c, National Institutes of Health, USA) based on the representative SEM images.

Preparation and characterization of 3D-printing bioink

Briefly, 6.25% gelatin (w/v) was firstly added into ultrapure water, and the mixture was stirred at 60 °C for 30 min until the gelatin was completely dissolved. After cooling to room temperature, the gelatin solution was sufficiently mixed with 0.15% (w/v) CaCl₂ and 35 μM MSA-2. Finally, 5.25% (w/v) SA was introduced, and the mixture was thoroughly agitated to obtain the 3D printing bioink. The rheological property of the bioink was measured with a Rheometer (MCR 302e, Anton Paar, Austria) equipped with a parallel plate with the diameter of 24.994 mm paralleling to a temperature-controlled sample-loaded platform at 25 cC, and the test gap was fixed at 1.0 mm. Under frequency sweep mode, the oscillatory strain was maintained at a fixed value of 1%, while recording the G’ and G” moduli of each sample as the angular frequency increasing from 0.1 to 10 Hz. Subsequently, the shear frequency was set to 1 Hz during the strain sweep process, and the G’ and G” moduli of each sample were recorded as the strain varied within the range of 1–200%. To further validate the feasibility of the bioink preparation method, an oscillatory thixotropy test at a constant frequency of 1 Hz and three-stage strains (1–200–1%) with each stage lasting for 60 s was performed on the samples with the optimal formulas acquired from the sweep modes of frequency and strain.

Fabrication and characterization of 3D-printed scaffold and sandwich-like composite

The 3D-printed scaffold was fabricated according to our reported methods with a few modifications^3,27,29. Simplify 3D software (Version 4.1, USA) was used to design the wanted scaffold shape and the printing parameters. Under 0.5–1.5 kg cm⁻² compressed air pressure, the scaffold was fabricated by extruding bioink at a filling rate of 33% and printing speed of 30 mm s⁻¹. Concurrently, the prepared and cut electrospun fiber patch was sandwiched as the middle layer of the sandwich-like composite. Lastly, the obtained composite was frozen at −80 °C for 2 h and subsequently freeze-dried overnight. After lyophilization, the MSA-2-loaded scaffold (MS) and composite composed of two MS sandwiching one DF (MS-DF-MS) were photographed with a cellphone (iPhone 12 pro max). The corresponding microstructures were observed and captured with a SEM (Quanta 400 F).

In-vitro drug release

The in-vitro drug release properties of MSA-2 or DOX from the sandwich-structured composite was evaluated in PBS at pH 6.5 or 7.4 at 37 °C. The single drug-loaded depots (S-DF-S or MS-F-MS) in dialysis bags with cut-off molecular weight of 3 kD were immersed into 10 mL PBS. Afterwards, 200 μL released solution was collected from each individual release system as schedule, and 200 μL fresh PBS with the corresponding pH value was refilled into the system to keep the constant volume. The absorption values of the collected PBS solutions containing DOX or MSA-2 were measured at 480 nm and 326 nm, respectively, using a microplate reader (Synergy H1MF, BioTek, USA). The release kinetic of DOX or MSA-2 was analyzed based on the corresponding quantification of the cumulative drug release.

Cell lines

A murine melanoma cell line B16F10 (#TCM36), a murine colorectal cancer cell line CT26 (#TCM37), a human non-small cell lung cancer line A549 (#SCSP-503), a human breast cancer cell line MCF7 (#SCSP-531), a murine macrophage cell line RAW264.7 (#SCSP-5036), and a human macrophage cell line THP1 (#SCSP-567) were obtained from National Collection of Authenticated Cell Cultures (Shanghai, China). A murine immature dendritic cell line JAWSII (#CRL-11904) was obtained from the American Type Culture Collection. All cell lines were authenticated via short tandem repeat (STR) profiling. Routine authentication was further ensured through the examination of cell morphology using an inverted microscope and the analysis of growth curves during the culturing process. Additionally, upon receipt of all cell lines, mycoplasma testing was conducted using a mycoplasma detection kit, and regular mycoplasma detection tests were performed to prevent potential contamination.

Cell culture

B16F10 cells were cultured in RPMI-1640 medium supplemented with 10% (v/v) FBS and 1% (w/v) penicillin G sodium/streptomycin sulfate (P/S). RAW264.7 cells were cultured in DMEM supplemented with 10% (v/v) FBS and 1% (w/v) P/S. THP1 cells were cultured in RPMI-1640 medium supplemented with 10% (v/v) FBS, 0.05 mM β-ME, and 1% (w/v) P/S. JAWSII cells were cultured in α-MEM supplemented with 20% (v/v) FBS, 5 ng mL^–1 GM-CSF, and 1% (w/v) P/S. All cells were cultured in an incubator (Forma^TM Steri-Cycle^TM i160, Thermo, USA) at 37 °C with a humidified atmosphere containing 5% CO₂.

In-vitro evaluation of cytocompatibility and cytotoxicity

Cytocompatibility of blank implants

To evaluate the cytocompatibility of the fabricated macrocarriers without drugs, B16F10 (5 × 10⁴ cells well^–1) were seeded in a 24-well plate and pre-cultured for 12 h to facilitate cellular adhesion. Afterwards, the cells were treated with blank fiber, blank scaffold, or blank sandwich-structured (scaffold-fiber-scaffold) composite for 24 h or 48 h. After co-incubation, both the transwell inserts with polyethylene terephthalate (PET) permeable membrane supports (pore size of 8 μm, base area of 0.33 cm², #TCS013024, Jet Bio-Filtration, Guangzhou, China) and blank carrier materials were carefully removed, and the cells were reacted with CCK-8 work solutions for another 2.5 h in an incubator (37 °C). The absorbance at a wavelength of 450 nm was measured using a microplate reader for each well, and the cell viability was assessed following the guidelines provided by the manufacturer. The cytocompatibility of the used materials was further evaluated through a live/dead cell staining experiment. Briefly, B16F10 cells were seeded into a 24-well plate at a density of 5 × 10⁴ cells per well, followed by the introduction of different carrier materials. After the co-incubation for 24 h or 48 h, the live and dead cells were stained with calcein-AM and PI at 37 °C for 30 min, respectively. The stained cells were observed and imaged with an inverted fluorescence microscope (Ti2-U, Nikon, Japan).

Cytotoxicity of drug depots

To investigate the cytotoxicity of drug-loaded composites, B16F10 cells were seeded in a 24-well plate (5 × 10⁴ cells well^–1) with transwell inserts. After cell adhesion, S-F-S, MS-F-MS, S-DF-S, or MS-DF-MS was introduced. After the co-incubated with different composites for 24 h, a CCK-8 assay and live/dead cell staining experiment were conducted as mentioned above.

In-vitro evaluation of immunogenic apoptosis

Apoptosis induced by DOX or MSA-2

DOX is a typical chemotherapeutic agent exhibiting the capacity to cause immunogenic apoptosis through specific mechanisms^8,9. In recent reports, the STING-mediated phosphorylation of IRF3 was demonstrated to activate BAX, further releasing cytochrome C, and finally triggering apoptosis³⁶. Besides, recent reports also showed a link among STING, ERS, and apoptosis^36,37,38. As is well known, ERS generally results in apoptosis through the activation of Casp12. To confirm the apoptosis induction effect of DOX or MSA-2 on cancer cells, we conducted western blot assays to assess the levels of BAX and Casp12 in B16F10 cells treated with various materials containing DOX or MSA-2. Initially, B16F10 cells were seeded in 6-well plates (2 × 10⁵ cells well^–1) containing transwell inserts with PET permeable membrane supports (pore size of 8 μm, base area of 4.67 cm², #TCS020006, Jet Bio-Filtration) and treated with S-F-S, MS-F-MS, S-DF-S, or MS-DF-MS for 6 h. Then, the cells were washed with pre-warmed PBS and lysed with 100 μL pre-cooled (4 °C) RIPA lysis buffer supplemented with 1 mM PMSF on ice for 10 min. The protein samples were gathered using cell scrapers and shifted into pre-cooled 1.5 mL Eppendorf (EP) tubes, followed by centrifugation (12000 g) at 4 °C for 5 min. The protein contents of harvested supernatants were measured using a BCA protein assay kit and uniformed by dilution with the lysis buffer. After mixing with 5× loading buffer, the sample solutions were heated at 95 °C and oscillated at 1200 rpm using a shaking incubator (MSC-100, Thermo, USA) for 20 min. The sample solutions with the identical concentration of denatured proteins were loaded onto a 12-well SDS-PAGE gradient (4–20%) gel, followed by electrophoresis with a vertical electrophoresis system (#165-8004, Bio-Rad, USA) and a basic power supply (#164-5050, Bio-Rad) at a constant voltage of 80 V. When the bromophenol blue band approached the bottom of the gel, the electrophoresis was stopped. The proteins were transferred from the gel to a methanol-activated polyvinylidene fluoride (PVDF) membrane with a pore size of 0.45 μm (IPVH00010, Merck Millipore) using a power supply at a constant current of 400 mA and a blotting module (#170-3935, Bio-Rad) for 1 h at 4 °C. Subsequently, the PVDF membrane was immersed in a TBST solution containing 5% (w/v) BSA and blocked by shaking (50 rpm) at room temperature for 2 h. The membrane was then washed with TBST solutions thrice and co-incubated with the buffer containing the primary antibody against BAX (1:1000), Casp12 (1:1000), or GAPDH (1:10000) overnight while shaking (50 rpm) at 4 °C, followed by another triple washing step with TBST. Afterwards, the washed membranes were immersed in the buffer containing an HRP-conjugated secondary antibody (1:2000) and gently agitated (50 rpm) at room temperature for 2 h. Following the last washing procedure with TBST, the protein blots were visualized employing an ECL kit and imaged with a multi-fluorescence and chemiluminescence imaging system (G:BOX Chemi XX6, Syngene, UK). Quantification of the resulting western blots was performed using Fiji.

CRT distribution on cell membrane

The cell surface distribution of CRT was detected using an immunofluorescence assay. Initially, B16F10 cells were seeded at a density of 5 × 10⁴ cells per well in 24-well plates and cultured for adhesion. Subsequently, the cells were co-incubated with S-F-S, MS-F-MS, S-DF-S, or MS-DF-MS for 12 h in an incubator. After the removal of transwell inserts and materials, the cells were gently washed twice with PBS (pre-warmed at 37 °C), followed by the fixation with a 4% (w/v in PBS) paraformaldehyde (PFA) solution for 20 min. The fixed cells were further washed with PBS and co-incubated with 5% (v/v) donkey serum in PBS for 1 h at room temperature. Following another round of washing with PBS, the cells were then co-incubated with a CRT primary antibody (1:500) for another 1 h at room temperature. The unbound primary antibody was removed by moderate washing with PBS, and the cells were subsequently co-incubated with a donkey-derived secondary antibody conjugated with Alexa Fluor 555 (1:500) for 1 h at room temperature. In the end, the washed cells were stained using an antifade mounting medium with DAPI. The immunofluorescence images were acquired employing an inverted fluorescence microscope (Ti2-U, Nikon, Japan).

HMGB1 and ATP release

The culture media of B16F10 cells undergoing different treatments mentioned above were harvested and centrifugated (300 g) at 4 °C for 5 min to completely remove the suspended cells. Afterwards, the contents of HMGB1 and ATP in the collected media were measured using an ATP assay kit and an HMGB1 ELISA assay kit in accordance with their manufacturer’s instructions, respectively.

In-vitro DC maturation and macrophage polarization

The B16F10 cells seeded in 12-well plates (1 × 10⁵ cells well^–1) were co-incubated with S-F-S, MS-F-MS, S-DF-S or MS-DF-MS for 12 h, followed by the collection of cancer cell debris. JAWSII and RAW 264.7 cells separately already growing in different 12-well plates (1 × 10⁵ cells well^–1) were co-incubated with the collected cell debris for 24 h. After that, both JAWSII and RAW264.7 cells were harvested with cell scrapers and washed with PBS. The washed RAW264.7 cells were subsequently permeabilized using triton X-100 for 15 min at room temperature. The sufficiently resuspended JAWSII cells were co-incubated with CD45 antibodies labeled with PE/Cyanine7 (1:100), CD86 antibodies conjugated with FITC (1:50), and CD80 antibodies labeled with PE (1:300) for 1 h at room temperature. The resuspended RAW264.7 cells were also co-incubated with CD45 antibodies (1:100) and CD86 antibodies (1:50) for 1 h. Lastly, both the DC maturation of JAWSII cells and the M1 polarization of RAW264.7 cells were assessed with a flow cytometer (CytoFLEX S, Beckman, USA). The acquired data was analyzed using the FlowJo software (Version 10.8.1, USA). Furthermore, the aforementioned experiments were also conducted on BMDCs and BMDMs, both of which were derived from murine bone marrow and pre-cultured for a week in specialized conditioned media. BMDCs were cultured in RPMI-1640 medium supplemented with 10% (v/v) FBS, 1% (w/v) P/S, 50 μM BME, 20 ng mL^–1 GM-CSF, and 10 ng mL^–1 IL4. BMDMs were cultured in DMEM supplemented with 10% (v/v) FBS, 1% (w/v) P/S, 50 μM BME, and 20 ng mL^–1 M-CSF.

STING-IFNβ pathway activation

Intracellular protein level of pIRF3

pIRF3 is the most direct regulatory factor of IFNβ in STING-IFNβ pathway^40,56,57. According to the method used to evaluate the levels of BAX and Casp12 in B16F10 cells with different treatments, the level of pIRF3 in B16F10 cells was determined using a western blot assay, which was conducted with a primary antibody against pIRF3 (1:2000). Then, to further assess the abundance and distribution of pIRF3 in B16F10 cells with different treatments, immunofluorescence staining was performed followed by flow cytometry and microphotography assays. After the co-incubations following the same method employed for western blot experiments, the treated or untreated B16F10 cells were gathered and washed thrice with pre-cooled PBS, followed by the fixation with 4% (w/v in PBS) PFA solutions and permeabilization with triton X-100 at room temperature for 15 min. After a subsequent round of washing, B16F10 cells were fixed and permeabilized before being blocked with 5% (v/v) donkey serum solutions at room temperature for 1 h. The blocked and washed cells were then co-incubated with a primary antibody against pIRF3 (1:300) at room temperature for an additional 1 h. After removal of the primary antibody, the cells were washed, followed by the co-incubation with IFKine^TM Green donkey-derived IgG (1:500) at room temperature for 2 h. In the end, the pIRF3⁺ fluorescence of rinsed B16F10 cells was determined with a flow cytometer (CytoFLEX S, Beckman), and the acquired data was subsequently analyzed using the FlowJo software (Version 10.8.1).

In addition to flow cytometry analysis, the fluorescence microscopy technique was also used to acquire microscopic images of B16F10 cells growing on glass coverslips in 24-well plates. Following different co-incubations, the washed B16F10 cells were fixed, permeabilized, and subsequently blocked with donkey serum diluents. Then, the cells were washed and co-incubated with a pIRF3 primary antibody (1:500) at 4 °C overnight. After thorough removal of the primary antibody, the B16F10 cells were washed and sequentially co-incubated with the IFKine^TM Green donkey-derived anti-rabbit IgG (1:500) at room temperature for 2 h, followed by secondary antibody removal and washing. Each rinsed coverslip was carefully transferred onto a glass slide, using a solitary droplet of antifade mounting medium that contained DAPI. Lastly, the immunofluorescence pictures were captured employing an inverted fluorescence microscope (Ti2-U, Nikon).

Additionally, the same treatments and subsequent detections were performed on CT26, A549, and MCF7 cells to assess the variations in intracellular protein levels of pIRF3 within these cells.

Intracellular mRNA level of IFNβ

B16F10 cells were initially cultured in 6-well plates (2 × 10⁵ cells well^–1), followed by the co-incubation with S-F-S, MS-F-MS, S-DF-S, or MS-DF-MS in transwell inserts. After the treatment for 6 h, the total RNA in each well was separately isolated using Trizol on ice and purified through sequential steps involving chloroform, isopropanol, and 75% (v/v) ethanol at 4 °C. Following the determination of contents and purities of RNA samples, cDNA was synthesized with a reverse transcription kit and a polymerase chain reaction (PCR) thermocycle instrument (ProFlex^TM Base, Thermo). Subsequently, a real-time quantitative PCR (RT-qPCR) based on the cDNA and designed primers (table S1) was performed using an SYBR Green kit and a RT-qPCR instrument (QuantStudio™ 5, Thermo). For each sample, the 2^Ct (Ct = amplification cycle threshold) value of Ifnb1 was normalized to that of Gapdh as an internal reference, and the corresponding mRNA level of Ifnb1 was expressed as a fold change relative to control group.

To evaluate the mRNA level of IFNβ in APCs, B16F10 cells were pre-treated with MS-DF-MS for 24 h to generate DAMPs and TAAs. Then, the cancer cell debris was collected and shifted into transwell inserts in 6-well plates containing JAWSII or THP1 cells (2 × 10⁵ cells well^-1) with or without S-F-S or MS-F-MS. After co-incubation for 6 h, the mRNA level of ifnb1 (JAWSII cells) and IFNB1 (THP1 cells) were determined with a RT-qPCR assay mentioned above.

Inhibition of STING degradation by facilitation of microtubule disaggregation

To better understand the cascade-reinforced cycle of immunogenic apoptosis and antitumor immune response, we investigate the interplay among STING activation, activated NF-κB-induced microtubule disaggregation, and STING degradation in cancer cells, dendritic cells, or macrophages³⁹.

Firstly, the protein levels of STING in APCs with or without the activation of STING-IFNβ pathway or the stimulation using cancer cell debris were detected. DAMPs- and TAAs-contained cancer cell debris were prepared by the co-incubation of MS-DF-MS and B16F10 cells in 6-well plates (2 × 10⁵ cells well^–1) for 24 h. Subsequently, the B16F10 cell debris was used to treat JAWSII or THP1 cells in 6-well plates (2 × 10⁵ cells well^–1) containing transwell inserts with or without MS-F-MS for 6 h. The JAWSII or THP1 cells, whether treated or untreated, were collected and washed with pre-cooled PBS. Subsequently, they were fixed with PFA solutions and permeabilized with triton X-100 at room temperature for 15 min. After an additional round of washing, the fixed and permeabilized cells were blocked with donkey serum diluents at room temperature for 1 h. The blocked and washed cells were subsequently co-incubated with a primary antibody against STING (1:200) at room temperature for another 1 h. After the primary antibody was completely removed, the cells were washed and co-incubated with the IFKine^TM Green donkey-derived IgG (1:500) at room temperature for 1 h. The STING⁺ fluorescence of washed cells was ultimately measured using a flow cytometer (CytoFLEX S, Beckman), and the obtained data were subsequently analyzed with the FlowJo software (Version 10.8.1). Besides, the protein levels of STING were also assessed in BMDCs and BMDMs undergoing identical treatments.

Secondly, the protein levels of STING were evaluated in cancer cells, dendritic cells, or macrophages undergoing or not undergoing the activation of STING-IFNβ pathway or the stimulation with the culture supernatants of M1 macrophages. Meanwhile, the distributions or existence patterns of lysosomes and microtubules within the same cells were also examined. To achieve the M1 polarization, RAW264.7 cells were subjected to a 24-h co-incubation with B16F10 cell debris, which had been generated via a 24-h treatment with MS-DF-MS. Afterwards, the M1 culture supernatants were gathered and used to treat the pre-incubated B16F10 cells, JAWSII cells, or RAW264.7 cells with or without MS-F-MS for 6 h. All the cells, whether treated or untreated, were digested, gathered, and washed with PBS before undergoing fixation and permeabilization. Subsequently, the intracellular STING levels were evaluated using flow cytometry analysis, following the aforementioned methods. The fluorescence analyses based on microscopic images were also performed on B16F10 cells cultured on glass coverslips at the bottoms of 24-well plates (5 × 10⁴ cells well^–1) containing transwell inserts. After con-incubating with M1 macrophages for 6 h, the B16F10 cells were washed, fixed, permeabilized, and subsequently blocked using a diluent containing donkey serum. To assess the intracellular abundances and colocalizations of STING and lysosomes, the washed B16F10 cells were sequentially co-incubated with a rabbit-derived antibody against STING (1:200) and a rat-derived anti-mouse APC-conjugated LAMP1 antibody (1:500) separately at room temperature for 1 h. After thorough antibody removal, the B16F10 cells were washed and subsequently co-incubated with a 1:500 dilution of IFKine^TM Green donkey-derived anti-rabbit IgG. To estimate the distributions and spatial arrangements of microtubules, the B16F10 cells post-washing were co-incubated with an antibody against α-tubulin (1:50) at room temperature for 1 h, followed by antibody removal and washing. The washed B16F10 cells were co-incubated with a donkey-derived IgG conjugated with Alexa Fluor 555 (1:500) at room temperature for 2 h. Following fluorescence labeling, each washed coverslip was meticulously shifted onto a glass slide carrying a single droplet of antifade mounting medium that contained DAPI. All immunofluorescence images were finally acquired using an inverted fluorescence microscope (Ti2-U, Nikon).

Animal experiments

An experimental melanoma animal model was established using male C57BL/6 mice aged 6–8 weeks, which were obtained from the Laboratory Animal Center of Sun Yat-sen University (Guangzhou, China). The mice were housed in cages with a maximum occupancy of 5 mice per cage, and maintained at a temperature of approximately 25 °C in a humidity-controlled environment. The lighting was regulated on a 12-h light/dark cycle, while standard food and water were provided ad libitum. Specifically, the lower right flank of each mouse was subcutaneously inoculated with 100 μL B16F10 cell suspension (2 × 10⁷ cells in 1 mL PBS). After inoculation, the length (L) and width (W) of each subcutaneous tumor were recorded to calculate the corresponding tumor volume (V) using the formula: V = L × W²/2⁵⁸. Two weeks later, almost every tumor volume increased to approximately 500 mm³_.

To sufficiently confirm the in-vivo activation of STING-IFNβ pathway and assess the corresponding immune response, each mouse was anesthetized with an intraperitoneal injection (10 mL kg^–1) of a sterile PBS solution containing 0.6% (w/v) pentobarbital sodium, implanted with MS-DF-MS at the tumor site, and treated for 0 h, 6 h, 12 h, 24 h, or 48 h. Afterwards, the treated mice were euthanized, and their tumors were sampled for further analyses.

To investigate the antitumor effect of this in-situ implantation as a postsurgical adjuvant therapy, an experimental postsurgical recurrence animal model was established based on the tumor-bearing mice. To be specific, after the mice were anesthetized, each tumor was excised with a remaining volume of about 10 mm³. The free drugs or drug-loaded depots were administrated through injection or implantation to the location of the remaining tumor, subsequently completing suturing and disinfection. According to the forms of postoperative therapy protocols, the mice were categorized into five distinct experimental groups separately treated with S-F-S, free M + D (MSA-2, 22.5 mg kg^–1; DOX, 11.25 mg kg^–1), MS-F-MS (MSA-2, 22.5 mg kg^–1), S-DF-S (DOX, 11.25 mg kg^–1), MS-DF-MS (MSA-2, 22.5 mg kg^–1; DOX, 11.25 mg kg^–1), or DF-MS-DF (DOX, 11.25 mg kg^–1; MSA-2, 22.5 mg kg^–1). The mice undergoing surgical resection exclusively were defined as the control group. The free M + D was prepared by dissolving MSA-2 and DOX into a 1% (w/v) CMC solution with the assistance of ultrasonication. Following the in-situ implantation or injection, the tumor volumes and body weights of mice were recorded bidaily. In line with the animal ethics guidelines of Sun Yat-sen University, mice with tumor volumes exceeding 1500 mm³ would be euthanized, and their inclusion in the calculation of survival rates is warranted. Additionally, their tumor volumes prior to sacrifice, which might be larger than 1500 mm³, would be taken into account in the final recording and presented accordingly. On the 10th day post-surgery, the peripheral blood and main organs (liver, heart, spleen, lung, and kidney) were sampled from the anesthetized mice, and the recurrent tumors were additionally harvested for subsequent experimental analyses.

To evaluate the antitumor efficacy and immune response against metastatic tumors, a subcutaneous inoculation of 100 μL B16F10 cell suspension (1 × 10⁷ cells in 1 mL PBS) was performed on the lower left flank of each mouse, which had been inoculated with B16F10 cells in the right flank 7 days prior. This established an orthotopic (primary) and contralateral (distant) B16F10 tumor co-bearing model. Subsequently, the mice underwent the same treatments as mentioned above, except for DF-MS-DF implantation. On the 10th day post-surgery, distant tumor and major organ samples were harvested for further analysis. Additionally, a pulmonary metastasis mouse model was established through intravenous injection of 50 μL B16F10 cell suspension (8 × 10⁶ cells in 1 mL PBS) into the tail vein, following inoculation of B16F10 cells for 7 days. On the 10th day after surgery, lung samples were obtained for histological evaluation.

Transcriptomic analysis

The transcriptomes of the control and MS-DF-MS groups were subjected to sequencing analysis, which was performed using the tumors sampled after treatment for 12 h and quick-frozen in liquid nitrogen. The RNA extraction, purification, fragmentation, reverse transcription, PCR amplification, library construction, and sequencing procedures were fulfilled at LC-Bio Technology (Hangzhou, China). The acquired data underwent rigorous analysis and visualization following the methodologies outlined in our previous reports^3,59,60.

Histologic evaluation

The tumors harvested on day 10 after treatment were fixed in 4% (w/v in PBS) PFA solutions at room temperature for 2 days, subsequently dehydrated using gradient ethanol solutions, and finally embedded in paraffin. The embedded tissues were sliced into sections with a constant thickness of 6 μm using a paraffin microtome (RM 2235, Leica, Germany). After melting at 60 °C, deparaffination with xylene, and rehydration using counter-gradient ethanol solutions, the tissue sections were subjected to the staining procedures with H&E and TUNEL assays. The tumor sections were rehydrated, antigen-repaired, and then pre-treated with a blocking solution containing 5% (v/v in PBS) normal goat serum prior to immunofluorescence staining. Following washing with PBS, the tumor sections were separately co-incubated with the rabbit-derived primary antibodies against CRT (1:500), HMGB1 (1:100), and Ki67 (1:1000) at 4 °C overnight. Afterwards, the tumor sections were washed and co-incubated with goat-derived secondary antibodies conjugated with different fluorescent molecules under the dark. After another round of washing, the tumor sections were mounted with an antifade medium containing DAPI. Lastly, the bright-field and fluorescence images of the stained tissue sections were observed and captured using an inverted fluorescence microscope (Ti2-U, Nikon) and quantified with Fiji.

ELISA and flow cytometry analysis for tissues

Each harvested tumor or spleen was cut into pieces on ice, and a certain weight of tissue pieces was homogenized in a predetermined volume of pre-cooled PBS to obtain the corresponding single-cell suspension, followed by a filtration using cell strainers with an average pore size of 70 μm. After centrifugation at 1000 g and 4 °C for 15 min, the supernatants were used to investigate the contents of IFNβ and IFNγ with ELISA assay kits, while the cell precipitates were resuspended in red blood cell lysis buffer and lysed for 15 min at room temperature. Before the co-incubation with various rat-derived monoclonal antibodies labeled with distinct fluorescent dyes, the washed and suspended cells were fixed in 1% (w/v in PBS) PFA solutions at 4 °C overnight and then blocked with 5% (v/v in PBS) normal rat serum solutions at room temperature for 2 h, followed by washing with PBS. To estimate the maturation DCs in tumors, the suspended cells were co-stained with antibodies against CD45 (1:100), CD11c, CD80 (1:300), CD86 (1:50), and MHCII (1:200). To assess the polarization of macrophages in tumors, the suspended cells were co-stained with antibodies against CD45 (1:100), CD11b (1:100), F4/80 (1:100), CD86 (1:50), and CD206 (1:200). Notably, in order to achieve optimal binding of the anti-CD206 antibody, the suspended cells were pre-permeabilized using triton X-100. To analyze the phenotypes of T cells in tissues, the suspended cells were co-stained with antibodies against CD45 (1:100), CD3, CD4 (1:100), and CD8 (1:50) for the cell suspensions from tumors and co-stained with antibodies against CD45 (1:100), CD3, CD4 (1:100), CD8 (1:50), CD44 (1:200), and CD62L (1:100) for cell suspensions from spleens, respectively. Following the last washing step with PBS, flow cytometry measurements were performed on the cells from tissues, and the subsequent analyses were conducted with the FlowJo software (Version 10.8.1).

Biosafety evaluation

To assess the safety of our executed treatments, particularly those involving the implanted materials, we employed an automatic biochemical analyzer (3100, Hitachi, Japan) to perform the serological test on the serum obtained by centrifugating (1000 g) the blood at 4 °C for 20 min. Then, we also conducted the histological analyses of the harvested organs through H&E staining and microscopic examination.

Statistics and reproducibility

All statistical analyses were carried out employing GraphPad Prism (version 9.5), and data visualization was conducted using Origin (version 2021). The data are presented as the mean value with its corresponding standard deviation (mean ± SD), the latter of which is represented by error bars. Unless otherwise specified, three or more experimental units or subjects were used to calculate each mean and SD. The significance of differences between two groups, with one independent nominal-level variable, was evaluated using an unpaired two-tailed Student’s t test. The significance of differences among three or more groups, with one or more independent nominal-level variables, was assessed separately using one-way and two-way ANOVA with Tukey’s multiple comparisons post hoc test. The significance of differences among the inter-group data from survival results was analyzed using log-rank (Mantel-Cox) test. Asterisks denoting different levels of significance are used as follows: *P < 0.05; **P < 0.01; ***P < 0.001; and ****P < 0.0001. The symbol “n.s.” is used to indicate no statistical significance. No statistical method was used to predetermine sample size. No data were excluded from the analyses. The experiments were randomized. The investigators were not blinded to allocation during experiments and outcome assessment.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.