STING agonist-conjugated metal-organic framework induces artificial leukocytoid structures and immune hotspots for systemic antitumor responses

ABSTRACT

Radiotherapy is widely used for cancer treatment, but its clinical utility is limited by radioresistance and its inability to target metastases. Nanoscale metal-organic frameworks (MOFs) have shown promise as high-Z nanoradiosensitizers to enhance radiotherapy and induce immunostimulatory regulation of the tumor microenvironment. We hypothesized that MOFs could deliver small-molecule therapeutics to synergize with radiotherapy for enhanced antitumor efficacy. Herein, we develop a robust nanoradiosensitizer, GA-MOF, by conjugating a STING agonist, 2′,3′-cyclic guanosine monophosphate–adenosine monophosphate (GA), on MOFs for synergistic radiosensitization and STING activation. GA-MOF demonstrated strong anticancer efficacy by forming immune-cell-rich nodules (artificial leukocytoid structures) and transforming them into immunostimulatory hotspots with radiotherapy. Further combination with an immune checkpoint blockade suppressed distant tumors through systemic immune activation. Our work not only demonstrates the potent radiosensitization of GA-MOF, but also provides detailed mechanisms regarding MOF distribution, immune regulatory pathways and long-term immune effects.

INTRODUCTION

Radiotherapy (RT) is one of the most widely used cancer treatments in both curative and palliative settings [1], with approximately half of all cancer patients treated with RT during their disease courses. RT utilizes ionizing radiation to generate reactive radicals and damage DNA, thereby killing fast-proliferating cancer cells [2]. While RT successfully eliminates cancerous cells, it is also highly toxic to normal tissues. Thus, RT is limited by cumulative radiation dose and localized radiation to avoid excessive damage to normal tissues. As a result, RT is mostly applied in a local-regional setting and cannot target distant metastases [3]. Furthermore, recurrent and metastatic tumors can develop radioresistant phenotypes, making RT much less effective for these tumors [4,5].

To improve its efficacy, RT has traditionally been combined with chemotherapy. However, the combination of RT and chemotherapy is also highly toxic to normal tissues, causing debilitating side effects to patients. In the past three decades, significant efforts have been devoted to the search for non-chemotherapy radiosensitizers to augment the efficacy of RT without causing serious adverse effects [6,7]. Of particular interest are high-Z nanoradiosensitizers, which are non-toxic and can enhance RT by absorbing more radiation energy [8,9]. They are administered intratumorally (i.t.) to achieve a high local concentration for effective radiosensitization and cancer cell killing [10,11].

Despite preclinical efficacy in enhancing RT-mediated cell killing, non-chemotherapy radiosensitizers have not been approved for clinical use by the US Food and Drug Administration [12]. This lack of clinical success has been attributed to the immunosuppressive properties of RT in the complex tumor microenvironment (TME) [13]. RT often induces an immunosuppressive TME due to increased recruitment of myeloid-derived suppressor cells (MDSCs), upregulation of the nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κβ) and exhaustion of lymphoid cells [14,15]. This immunosuppressive TME is less responsive to RT and prevents the patient's immune system from attacking tumor cells.

Given these limitations of existing radiosensitizers, we have identified nanoscale metal-organic frameworks (MOFs) as efficient and non-toxic nanoradiosensitizers via a unique radiotherapy-radiodynamic therapy (RT-RDT) process [11,16]. The high-Z-metal-based secondary building units (SBUs) in MOFs efficiently absorb X-rays to enhance RT effects (primarily by enhancing hydroxyl radical generation), whereas the photosensitizing ligands in the MOFs are excited for the RDT process (via singlet oxygen generation) [9,17,18]. MOF-mediated RT-RDT causes immunogenic cell death (ICD) of cancer cells to synergize with toll-like receptor agonists and immune checkpoint inhibitors [19–21]. We have also used MOFs for neutrophil reprogramming [21]. As MOFs are known to be a unique class of porous material [22,23], we surmised that MOFs could deliver other small molecule therapeutics to synergize with RT-RDT for enhanced antitumor efficacy. There is also a need to understand the detailed mechanisms of MOF-mediated TME regulation, including MOF distribution in tumors, immune regulation pathways and long-term immune effects after RT-RDT treatment.

Herein, we report a highly effective nanoradiosensitizer, GA-MOF, by conjugating a MOF with an agonist of stimulator of interferon genes (STING), 2′,3′-cyclic guanosine monophosphate–adenosine monophosphate (GA), for synergistic RT and immunotherapy. Coordination between GA and Hf₁₂-SBU prolonged GA retention in tumors and enhanced STING activation. GA-MOF showed low preclinical toxicity, and when combined with low-dose RT, demonstrated superb anticancer efficacy in colon cancer, pancreatic cancer, and head and neck cancer models. We determined the uptake of MOF particles by tumor cells and different immune populations, and revealed the formation of immune-cell-rich nodules, termed artificial leukocytoid structures (ALS), in the tumors after i.t. injection of GA-MOF. Low-dose RT turned these ALS into immunostimulatory hotspots in the tumors and facilitated antitumor immunity. Further combination of GA-MOF plus RT with an immune checkpoint blockade (ICB) not only improved local cancer eradication, but also suppressed distant tumors via systemic immune activation.

RESULTS

MOF enhances RT and enables sustained release of GA

DBP-Hf MOF was synthesized via a solvothermal reaction between HfCl₄ and 5,15-di(p-benzoato)porphyrin (H₂DBP) in N,N-dimethylformamide (DMF) with acetic acid (AA) as a modulator (Fig. S1) [22]. The DBP-Hf MOF consisted of Hf₁₂ SBUs and photosensitizing DBP ligands for RT-RDT effects [11]. The AA-capped DBP-Hf MOF was modified with trimethylsilyl trifluoroacetate (TMS-TFA) to afford a TFA-capped DBP-Hf MOF (abbreviated as MOF in the following text) (Fig. S1) [19]. This AA/TFA exchange created a dynamic surface for GA coordination [23] to afford GA-MOF by replacing the weakly bound TFA on Hf₁₂-SBUs with GA (Figs S2 and S3). Liquid chromatography-mass spectrometry (LC-MS) analysis gave a GA loading of 0.74 wt% in GA-MOF, corresponding to 1 GA molecule per 14.9 Hf₁₂-SBUs. Transmission electron microscopy (TEM) revealed the same nanoplate morphology for MOF and GA-MOF (Fig. 1a and b). Powder X-ray diffraction (PXRD) patterns showed that GA-MOF maintained the same crystalline hexagonal close-packed (hcp) topology as MOF (Fig. 1c). Dynamic light scattering (DLS) revealed hydrodynamic sizes of 100.0 ± 2.2 and 133.3 ± 4.5 nm for MOF and GA-MOF, respectively (Fig. 1d). The binding strength between GA and MOF was quantified by isothermal titration calorimetry (ITC), affording an association constant (K_GA) of (1.50 ± 0.79) × 10⁷ M⁻¹ (Fig. 1j).

RT enhancement by MOF was evaluated in four different cell lines, including murine colon cancer MC38 and CT26, pancreatic cancer Panc02, and head and neck cancer SCC7, by a growth rate inhibition (GR) assay [24]. The growth inhibition factors (GIFs) were calculated based on real-time growth rates of cancer cells with increasing doses of X-ray (Fig. S4). The cells treated by MOF without RT [denoted MOF(−)] showed minimal growth delay (Fig. 1e and h and Table S1). With X-ray irradiation, MOF treatment [denoted MOF(+)] significantly reduced cell proliferating rates. The GIF values at the 10% GR level were 1.46, 1.17, 1.23 and 1.31 for MC38, CT26, Panc02 and SCC7, respectively. These results suggest that MOF is a biocompatible and efficient nanoradiosensitizer.

We recently showed that high inorganic phosphate concentrations inside cells could trigger the release of coordinated molecules from the SBUs [21,23]. We used LC-MS to determine the release profiles of GA-MOF in 0.1 × phosphate-buffered saline (PBS) (1.18 mM phosphate), 1 × PBS (11.8 mM phosphate) and 10 × PBS (118 mM phosphate) at 37°C to mimic different physiological conditions [25,26]. Incubation of GA-MOF in 0.1 × PBS released <50% of GA in 6 days, while incubation in 1 × PBS and 10 × PBS released 80% and 92% of GA, respectively (Fig. 1i). Since the cytoplasm has ∼10-fold higher free phosphate concentration (∼10 mM) than the interstitial fluid or plasma (∼1 mM) [25,26], GA-MOF slowly releases GA in the interstitial fluid but rapidly releases GA upon endocytosis into cells [27,28]. The release of GA was found to be pH-independent (Fig. S1). These findings suggest that GA can be released from Hf₁₂-SBUs by the intracellular/extracellular phosphate gradient (Fig. 1k) [21,23].

GA-MOF delivers GA and elicits robust STING activation

The binding of GA to MOF did not affect the cellular uptake of MOF. MOF and GA-MOF showed similar Hf uptake in MC38, CT26 and SCC7 cells (Fig. S5) by inductively coupled plasma-mass spectrometry (ICP-MS). However, GA-MOF significantly enhanced GA uptake in vitro. We used cyanine5-conjugated GA (GA/Cy5) to track GA uptake by murine macrophage Raw264.7 cells by confocal laser scanning microscopy (CLSM). GA/Cy5 showed rapid internalization of fluorescence signal in 2 hours, but the signal gradually decreased and almost completely disappeared in 24 hours (Fig. S5). In contrast, the intracellular fluorescence of GA/Cy5-MOF gradually increased and was 34.5-fold stronger than that of the GA/Cy5 group at 24 hours (Fig. 2a). The inhibition experiment showed that only chlorpromazine (but not rottlerin and nystatin) reduced the GA-MOF uptake of Raw264.7 cells, which suggests that Raw264.7 cells take up GA-MOF mainly through clathrin-mediated endocytosis (Fig. S5).

STING activation was assessed using human monocyte THP-1 cells with a luciferase reporter gene linked to interferon (IFN)-responsive elements [29]. After 24-hour incubation, GA-MOF showed a half-maximal effective concentration (EC₅₀) for the IFN regulatory factor (IRF) response of 2.34 ± 1.44 μM, which was 3-fold lower than that of GA (EC₅₀ = 6.98 ± 1.15 μM) (Fig. 2b). This result indicates that GA-MOF elicits stronger STING activation than GA.

STING activation was supported by phosphorylation of STING (p-STING) by CLSM. Bone-marrow-derived dendritic cells (BMDCs) and bone-marrow-derived macrophages (BMDMs) from C57BL/6 mice were incubated with GA or GA-MOF (0.27 μM GA/50 μM Hf) for different lengths of time. GA treatment showed a peak signal of p-STING in BMDCs and BMDMs in 2 hours, while GA-MOF treatment showed sustained STING phosphorylation for up to 8 hours (Fig. 2c and Fig. S6). GA-MOF also showed stronger activation of IRF-3 (Fig. S6) and improved phagocytosis capacity of innate immune cells (Fig. S7).

STING activation leads to the downstream secretion of type-I IFN and proinflammatory cytokines [30]. GA-MOF treatment increased IFN-β secretion in BMDCs, BMDMs and Raw264.7 cells over other groups, indicating stronger activation of the STING-IFN axis in immune cells (Fig. 2d, Fig. S8). Low-dose RT at 2 Gy did not affect IFN secretion, and immune cells were viable upon MOF(+) treatment (Fig. S9). Both GA-MOF(+)-and GA-MOF(−)-treated primary immune cells showed an increased secretion of interleukin-6 (IL-6) and tumor necrosis factor (TNF-α) (Fig. 2e, f and Fig. S8). GA-MOF(+) stimulated the secretion of IFN-β and TNF-α by Raw264.7 cells, but did not upregulate the secretion of immunosuppressive IL-10 and transforming growth factor beta (TGF-β) (Fig. S8). These results demonstrate that GA-MOF induces a more potent and sustained STING activation than GA, leading to enhanced secretion of immunostimulatory cytokines either with or without RT.

GA-MOF retains GA in tumors and induces artificial leukocytoid structures

To assess the potential of GA-MOF as an immune agonist and a radiosensitizer, we evaluated the toxicity, pharmacokinetics (PK) and tumor retention of GA-MOF in preclinical models. We injected PBS, GA or GA-MOF at a GA or MOF dose of 1 or 100 mg/kg into the subcutaneous space of Sprague Dawley (SD) rats. The SD rats in all groups showed steady weights and had no health problems throughout the experiment (Fig. 3a and Fig. S10). Compared to GA, GA-MOF showed longer plasma retention of GA, as quantified by a competitive enzyme-linked immunosorbent assay (ELISA) kit (Fig. 3b). The area under the curve (AUC) of GA-MOF in plasma was 1.48-fold higher than that of GA. We quantified the tumor retention of GA after i.t. injection of GA or GA-MOF into subcutaneous MC38 tumors in C57BL/6 mice. GA-MOF slowed the release of GA and retained 3-fold more GA in the tumors over a 7-day period (Fig. 3c). These results demonstrate increased GA retention in tumors by GA-MOF.

During preclinical evaluations of SD rats with subcutaneously injected GA-MOF, we observed the formation of dark subcutaneous nodules covered with fascias (Fig. 3d and Fig. S10). We observed similar nodules in MC38 tumors in C57BL/6 mice with i.t. injected GA-MOF. To characterize these nodules, we first performed histological staining of the subcutaneous nodules in rats. Due to the dark red color of DBP, we could directly observe MOF by optical microscopy. Hematoxylin and eosin (HE) staining showed the presence of MOF (brown) in the centers and cells and connective tissues in the surrounding areas of the nodules (Fig. 3d). Ionized calcium-binding adapter molecule 1 (IBA-1) and cluster of differentiation 3 (CD3) immunohistochemistry (IHC) staining revealed that the majority of the cells in the nodules were innate immune cells, with some T cells at the nodule margins (Fig. 3e and f and Fig. S11). We hypothesized that GA-MOF acted as a foreign object to attract immune cells into the injection site and activated the local immune system through STING activation [31–33]. Interestingly, we also found significantly less MOF and more cells in the nodules on day 14 (Fig. S11) than on day 7 (Fig. 3d), which suggests active and dynamic transport/clearance of MOF particles by immune cells.

In the HE-stained slides of nodules, ∼16.7% of cells in the nodules carried MOF. Most of these cells were macrophages or monocytes according to their morphologies (Fig. 3e). In the HE-stained slides of draining lymph nodes (LNs), we observed migrating macrophages/monocytes carrying MOF contents (Fig. 3f, Fig. S11). We can depict the fate of GA-MOF after injection into subcutaneous space or tumors based on these observations. After injection, GA-MOF induces immune cell infiltration and triggers foreign body clearance mechanisms. Innate immune cells, including neutrophils, macrophages and monocytes, form the nodules (ALS) surrounding the injected GA-MOF. The phagocytes carry GA-MOF back into the draining LNs for clearance and antigen presentation [34]. T cells are also known to be involved in this clearance and regenerative process [35]. GA may be released once GA-MOF is uptaken by these innate immune cells to activate their STING-related pathways, either in the nodules or in the draining LNs. However, it remains unclear if STING activation contributes to nodule formation.

RT transforms artificial leukocytoid structures into immune hotspots

Different from the subcutaneous space in rats, MOF after i.t. injection can access both tumor cells and immune cells. We performed flow cytometry staining to determine MOF uptake by different cell populations in MC38 tumor-bearing C57BL/6 mice. The fluorescence of DBP (BV711 channel) from MOF was used to differentiate MOF⁻ and MOF⁺ cells in the tumor. One day after i.t. injection of MOF, 65.4% MOF⁺ cells were non-immune cells (CD45⁻) and 34.6% MOF⁺ cells were leukocytes (CD45⁺, Fig. 4a). The leukocytes exhibited 1.4-fold higher MOF signal than non-immune cells (Fig. S12). Approximately 97.9% MOF⁺ leukocytes were myeloid cells (MOF⁺CD45⁺CD11b⁺, Fig. 4b), with 41.9% monocytes (Ly6C⁺), 33.9% macrophages (F4/80⁺) and 23.5% neutrophils (Ly6G⁺) (Fig. 4c). These results are consistent with the histological observations of subcutaneous nodules in rats (Fig. 3d and h).

To understand the impact of RT on the nodules, we irradiated the injected tumors with 4 Gy X-ray for 3 daily fractions (Fig. S13), excised the nodules from the tumors 14 days from the first RT, and performed flow cytometric staining. Both MOF and GA-MOF triggered nodule formation in the injected tumors (Fig. S14). These nodules were immunologically distinct from the tumors (Fig. 4d). Compared to GA-MOF(−), MOF(−) and MOF(+), GA-MOF(+) resulted in smaller tumor sizes and a more inflammatory TME with significantly more infiltration of myeloid cells (CD45⁺CD11b⁺), helper T cells (CD45⁺CD3⁺CD4⁺), effector T cells (CD45⁺CD3⁺CD8⁺CD44⁺CD62L⁻) and central memory T cells (CD45⁺CD3⁺CD8⁺CD44⁺CD62L⁺) and downregulation of regulatory T cells (T_reg, CD45⁺CD3⁺CD4⁺FOXP3⁺CD25⁺) (Fig. 4d, Fig. S14). These results suggest that GA-MOF(+) produces immunologically ‘hot’ nodules (immune hotspots) to suppress tumor growth.

IHC staining of GA-MOF(+)-treated tumors revealed similar nodule structures to the subcutaneous nodules in rats (Figs 3d–f and 4e). The MOF center was surrounded by IBA-1⁺ innate immune cells and CD3⁺ T cells. Colonies of tumor cells were surrounded by a large number of immune cells (Fig. S15), indicating an active and T-cell-dependent antitumor response. In contrast, MOF(+)-treated tumors did not exhibit immune infiltration near the nodules, with isolated MOF islands trapped in the tumor interstices (Figs S14–S16). We also inoculated MC38 tumors in STING-knockout (STING^−/−) B6 mice, and i.t. or subcutaneously injected GA-MOF to observe nodule formation. Compared to wild-type (WT) mice, STING^−/− mice showed little infiltration of lymphoid cells in both intratumoral and subcutaneous nodules (Fig. S17). This result indicates that STING activation is required for the formation of immune hotspots in the tumors.

GA-MOF elicits antitumor effects and activates the TME

We evaluated the antitumor efficacy of GA-MOF in subcutaneous MC38, CT26, Panc02 and SCC7 models. Since different tumor models had different sensitivity to RT, we used 2 Gy by 3 fractions for CT26, 3 Gy by 3 fractions for Panc02 and SCC7, and 4 Gy by 3 fractions for MC38. GA(+) moderately slowed tumor growth with tumor growth inhibition (TGI) values of 53%–64% in the four tumor models. MOF(+) significantly slowed tumor growth with TGI values of 71%–90%. The simple combination of GA and MOF(+) gave a TGI of 79% in the MC38 model (Fig. S18). In stark contrast, GA-MOF(+) treatment synergized RT-RDT effects and STING activation to greatly enhance antitumor efficacy with TGI values of 82%–98% (Fig. 5a–d, Table S2). One out of eight CT26 tumor-bearing mice and five out of seven MC38 tumor-bearing mice were cured after GA-MOF(+) treatment. GA-MOF(−) barely inhibited tumor growth (Fig. S18). Superb antitumor efficacy and steady body weights (Fig. S18) of GA-MOF(+)-treated mice demonstrated GA-MOF as a biocompatible nanoradiosensitizer for synergistic radiosensitization and STING activation. To reveal early changes in the TME, we quantified cytokines in tumor lysates and performed immune profiling of MC38 tumors by flow cytometry 3 days post-treatment (day 13). GA-MOF(+) induced significantly more secretion of IFN-β than GA(+) or PBS(+), but downregulated TGF-β in the tumors (Fig. S19). MOF(+)- and GA-MOF(+)-treated tumors showed >9-fold higher infiltration of dendritic cells (DCs) (CD45⁺CD11b⁺CD11c⁺) than other groups (Fig. 5e, Fig. S20), likely resulting from the enhanced DC differentiation from monocytes and upregulation of CD11c in granulocytes by RT-RDT [11,21,36]. Furthermore, GA-MOF(+) induced 3.4-fold higher DC infiltration and 1.5-fold higher total leukocyte infiltration than MOF(+). The upregulation of major histocompatibility complex class II (MHCII) and CD45, together with the downregulation of naïve T cell marker in the GA-MOF(+) group (Fig. 5e) indicated enhanced antigen presentation and T cell maturation.

To identify the major effector immune populations, we performed depletion experiments with neutralizing antibodies. Depletion of B cells or NK cells did not alter the efficacy of the treatment compared to isotype controls (Fig. S18). However, when the T cells of tumor-bearing mice were depleted by the anti-CD3 antibody, the efficacy of GA-MOF was significantly compromised (Fig. S18). Thus, the GA-MOF-mediated antitumor effect is dependent on T cell responses.

To understand the synergistic effects of STING activation and RT-RDT effects on local tumor regression, we performed TME screening by extracting mRNA from SCC7 tumors for NanoString analysis. Gene set analysis (GSA) revealed that GA-MOF(+) showed higher scores for innate and adaptive immune responses than MOF(+) and GA-MOF(−) (Fig. 5f). In particular, GA-MOF(+) upregulated IFN, proinflammatory cytokines, phagocytosis-related transporter functions, antigen processing and presentation, DC and macrophage functions, and T/B cell functions.

Between MOF(+) and GA-MOF(+)-treated tumors, 51 genes passed the threshold of the nCounter PanCancer Immune Profiling Panel with ≥1.5-fold expression changes and adjacent p values (p_adj) < 0.05 (Figs S21 and S22). Compared to MOF(+), GA-MOF(+) increased gene expression downstream of IFN, including radical S-adenosyl methionine domain-containing protein 2 (Rsad2, 6.09-fold) and IFN-stimulated gene 20 (Isg20, 2.47-fold), and enhanced infiltration and migration of DCs (Fig. 5f), as demonstrated by upregulation of tyrosine-protein kinase Lyn (1.92-fold), solute carrier family 11-member 1 (Slc11a1, 2.91-fold), C-X-C chemokine receptor type 4 (Cxcr4, 3.88-fold) and C-C chemokine ligand 5 (Ccl5, 1.84-fold) [37–40]. GA-MOF(+) treatment upregulated Cd14 (3.42-fold), indicating enhanced differentiation of monocytes into DCs [41]. The expressions of cytokine/chemokine-related genes also showed significant changes. GA-MOF(+)-treated tumors upregulated TNF superfamilies, including Tnfrsf13b, Tnfrsf1b, Tnfsf12 and Tnfaip3, by >1.8-fold over the MOF(+) group. Similar upregulation was observed in C-X-C chemokine ligand 1 (Cxcl1, 4.13-fold), Ccl4 (13.06-fold), Il1b (6.72-fold), Pro-Platelet basic protein (Ppbp, 6.97-fold), chemokine receptor-like 2 (Ccrl2, 8.63-fold), Cxcl2 (12.84-fold) and Ccl3 (10.61-fold). Inflammasome pathways were also activated with upregulation of Nlrp3 (4.01-fold), Il1b, IL1 receptor type 2 (Il1r2, 6.42-fold), IL1 receptor-associated kinase 3 (Irak3, 3.87-fold) and nitric oxide synthase 2 (Nos2, 4.18-fold) in GA-MOF(+)-treated tumors. These changes in cytokines/chemokines reflected inflammatory responses of macrophages and active recruitment of leukocytes [42]. STING activation could reprogram tumor vasculature [43,44]. Angiogenesis-related genes were activated in GA-MOF(+)-treated tumors, including Serpinb2 (9.78-fold), thrombomodulin (Thbd, 2.81-fold) and kinase insert domain receptor (Kdr, 2.90-fold). The complement system is essential for foreign body clearance and the removal of damaged cells in the innate immune system [45]. Upregulation of Serping1 (2.74-fold) and complement component 3a receptor 1 (C3ar1, 2.24-fold) demonstrated dynamic regulation of the complement system, which might be related to cancer cell apoptosis, nodule formation and MOF trafficking.

GA-MOF plus αPD-L1 elicits robust systemic immune responses

GA-MOF(+) treatment induces intratumoral immune hotspots and synergizes STING activation with RT-RDT for regression of local tumors and amelioration of the TME. Activated immune systems and improved antigen presentation provided excellent contexts for combination with ICB [46]. We established two bilateral tumor models to evaluate the antitumor efficacy and abscopal effects of GA-MOF(+) in combination with a monoclonal anti-PD-L1 antibody (αPD-L1). We treated the primary tumors on the right flanks similarly to the single lateral model and left the distant tumors untreated. The tumor-bearing mice received intraperitoneal (i.p.) injections of αPD-L1 (Fig. S13).

In the bilateral CT26 model, PBS(+) or αPD-L1(+) moderately controlled local tumors with TGI values of <61% and had no effects on distant tumors (Fig. 6a and b, Table S3). GA-MOF(+) exhibited excellent control of primary tumor growth with a TGI of 95% but had no effect on distant tumors. The addition of αPD-L1 to GA-MOF(+) not only enhanced local tumor regression to give a TGI of 98% but also controlled distant tumors with a TGI of 66%. The combination of GA(+) and αPD-L1 also moderately controlled both local and distant tumors with TGI values of 75% and 41%, respectively. This result supports the role of ICB in activating cytotoxic T cells for systemic antitumor effects.

In the bilateral MC38 model, PBS(+) slightly inhibited the growth of primary and distant tumors with TGI values of 66% and 30%, respectively (Fig. 6c and d, Table S3). The αPD-L1(+) treatment enhanced local tumor control with a TGI of 82% and moderately controlled distant tumors with a TGI of 47%. The addition of GA to αPD-L1(+) increased TGI values to 88% and 64% for local and distant tumors, respectively. GA-MOF(+) showed excellent regression of primary tumor growth with a TGI of 99% but had a modest impact on distant tumors with a TGI of 67%. Combination of GA-MOF(+) with αPD-L1 completely eradicated primary (100% TGI) and regressed distant tumors with a TGI of 93% and a 50% cure rate. The combination regimen did not show side effects in both models (Fig. S23).

Flow cytometric profiling of immune cells at the late stage (day 25) showed that GA-MOF(+) enhanced i.t. infiltration of adaptive immune cells, particularly memory T cells, CD8⁺ cytotoxic T cells and NKT cells (Fig. 6e, Fig. S24). GA-MOF(+) plus αPD-L1 further enhanced systemic immune responses with increased effector T cells and memory T cells in the distant tumors (Fig. 6f). The reduced T_reg populations in the primary and distant tumors of GA-MOF(+) and GA-MOF(+) plus αPD-L1 groups (Fig. 6f) implied a more active T-cell-dependent antitumor effect. Interestingly, the percentages of myeloid cells, including the GR1⁺ population and macrophages, increased in the distant tumors of GA-MOF(+) and GA-MOF(+) plus αPD-L1 groups (Fig. 6f). The enhanced infiltration of innate immune cells is consistent with amelioration of immunosuppressive TMEs and active immune clearance of tumor cells.

Analysis of IBA-1 and CD3 markers revealed spatial distributions of innate and adaptive immune cells in the primary and distant tumors. GA-MOF(+) plus αPD-L1 further increased T cell infiltration without impacting the innate immune populations in distant MC38 tumors (Fig. 7a, Fig. S25). This result indicates that GA-MOF(+) plus αPD-L1 reverses immunosuppression to enhance immune cell infiltration and antitumor responses.

We also performed enzyme-linked immunosorbent spot (ELISpot) assays to assess antigen-specific immune responses of T cells (Fig. S26). Splenocytes were isolated from bilateral CT26-bearing BALB/c mice and bilateral MC38-bearing C57BL/6 mice and stimulated with tumor-specific peptide antigens AH1 (SPSYVYHQF) and KSP (KSPWFTTL), respectively. In both models, GA-MOF(+) plus αPD-L1 showed significantly more spot-forming cells (SFCs) than GA-MOF(+) or GA(+) plus αPD-L1, indicating enhanced adaptive immunity to produce more IFN-γ-generating splenocytes and stronger immune memory effects (Fig. 7b and c). The enhanced recognition of tumor antigens correlated well with improved antigen presentation (Fig. 5f) and increased infiltration of memory T cells in the tumors after GA-MOF(+) plus αPD-L1 treatment (Fig. 6e and f). These results show that the synergistic combination induces an inflammatory and T cell infiltrated TME and activates systemic antitumor immune responses for abscopal effects.

DISCUSSION

Although RT has been in clinical practice for over a century, non-toxic radiosensitizers have yet to be approved by the US Food and Drug Administration for clinical use [47]. Chemotherapeutics such as cisplatin, paclitaxel and gemcitabine are used in combination with RT to enhance antitumor efficacy, but these chemoradiotherapy regimens cause debilitating adverse effects [48,49]. Oxygen mimetics and hypoxia-targeting reagents have been reported to enhance RT in preclinical models, but they failed to demonstrate clinical benefits in human patients [6].

By absorbing more X-rays, high-Z particles increase ionizing events, generation of reactive radicals and DNA damage by RT [50]. This mode of radiosensitization requires very high i.t. concentrations of high-Z particles to be effective, which cannot be achieved with existing systemic drug delivery [51–54]. As a result, the nanoradiosensitizers in clinical testing are administered by i.t. injection to achieve high local concentrations for cancer killing [55]. However, little is known about the trafficking and clearance of these particles and their induced immune effects [56–60].

As a bifunctional particle, GA-MOF comprises MOF for radiosensitization and the endogenous STING agonist GA for STING activation. The STING pathway elicits antitumor functions such as type-I IFN-mediated T-cell-dependent tumor eradication, modulation of the vasculature and augmentation of adaptive immunity by developing tertiary lymphoid structures [43,61,62]. In this work, we determined the uptake of MOF-based nanoradiosensitizers by different cell populations and their spatial-temporal distributions in tumors. One day post i.t. injection, two-thirds of the MOF is uptaken by cancer cells, and the rest of the MOF is uptaken by innate immune cells, including neutrophils, monocytes and macrophages (Fig. 4a–c). MOF-mediated RT-RDT is non-toxic to terminally differentiated immune cells (Fig. S9) but significantly inhibits cancer cell proliferation (Fig. 1e–h). The MOF particle distribution in cancer cells ensures efficient radiosensitization for cancer cell killing, while MOF uptake by immune cells provides an excellent avenue to deliver innate immune agonists. Indeed, GA-MOF prolongs GA retention in tumors to enhance STING activation of innate immune populations and facilitate antitumor immunity. GA-MOF treatment caused little health burden to murine subjects, exemplified by healthy morphology in major organ histology (Fig. S27), normal liver and kidney functions (Fig. S28), and little deposition of MOF particles in peripheral tissues (Fig. S29).

GA-MOF(+) induces intratumoral immune hotspots and efficiently inhibits tumor growth in vivo. We first observed nodule formation after subcutaneous injection and i.t. injection of GA-MOF to rats and mice, respectively (Fig. 3d). These nodules contained MOF in the center and innate and adaptive immune cells in the periphery, effectively forming a separated immune space from the rest of the tumor (Fig. 4d and e). These ALS are dominated by innate immune cells (Fig. S14). Although both MOF and GA-MOF can trigger nodule formation in the tumors, only GA-MOF(+) results in an immunologically ‘hot’ TME with immunostimulatory signatures (Figs 4d and 5e and f). The immune hotspots in the tumors disrupt the immunosuppressive TME to elicit antitumor immunity. GA-MOF(+) treatment suppresses tumor growth in four tumor models, including immunologically ‘cold’ Panc02 pancreatic cancer and SCC7 head and neck cancer models. Thus, GA-MOF(+) creates immune hotspots to turn ‘cold’ TME ‘hot’ for enhanced anticancer effects.

Figure7d summarizes our hypothesis for immune hotspot formation and its role in antitumor immunity. As a foreign object, MOF triggers the complement pathway and attracts immune cells. Adsorption of albumin and other interstitial proteins on the MOF surface triggers immune cell recognition and infiltration [56]. Significant upregulation of complement-related genes, chemokines and adhesion molecules causes active and continuous recruitment of leukocytes into MOF-treated tumors (Fig. 5f, Fig. S22), leading to the formation of ALS. Upon low-dose X-rays, MOF(+) induces ICD of cancer cells and generates autologous tumor-specific antigens in situ (Fig. S5) [46,63]. The released GA from GA-MOF induces robust and sustained STING activation (Fig. 2b–f) and facilitates DC infiltration (Fig. 5e), antigen processing and antigen presentation (Fig. 5f).

By inducing ALS formation, GA-MOF not only serves as a nanoradiosensitizer to kill cancer cells and expose tumor-specific antigens in situ, but also acts as an exogenous nano-adjuvant to induce immune cell infiltration and enhance antigen presentation [64]. STING activation can also induce vascular disruption and reprogram the tumor protective mechanisms that otherwise inhibit immune cell infiltration. Most importantly, sustained STING activation results in STING-IFN-T-cell-dependent responses (Figs 6e and f, and 7b and c). GA-MOF(+) treatment improved T cell function via antigen presentation, as evidenced by enhanced infiltration of memory T cells (Fig. 6e and f) and improved tumor antigen-specific antitumor responses (Fig. 7b and c). The combination of GA-MOF(+) with αPD-L1 extends this local treatment regimen to systemic antitumor effects. We used bilateral tumor models to mimic patients with distant metastases that are not amenable to localized RT treatment. Blockade of the PD-1/PD-L1 axis not only reinvigorates T cells and eradicates local tumors (Fig. 6a and c), but also reduces the infiltration barrier of T cells to distant tumors and elicits robust abscopal effects (Fig. 6b and d).

Our current study has limitations. The subcellular localization of GA and GA-MOF may not be accurate because we are using GA/Cy5 to visualize GA and the particle, which has different hydrophobicity from GA. The interactions between different immune cells, especially early infiltrated phagocytes, DCs and resident T cells in ALS, were not elucidated. STING polarization of innate immune populations and the contribution of vascular reprogramming also merit further study. Detailed genomic analysis will also help to elucidate these immune pathways and interactions. A detailed toxicology study on large animals such as monkeys or dogs is needed to confirm the biocompatibility of MOF-based nanoradiosensitizers.

In summary, we designed a bifunctional nanoplatform for simultaneous STING activation and radiosensitization. GA-MOF prolongs GA retention in tumors and elicits strong and sustained STING activation. GA-MOF forms ALS that are rich in immune cells, which are converted to immune hotspots upon X-ray irradiation. GA-MOF(+) greatly ameliorates the immunosuppressive TME and demonstrates great local tumor control in four tumor models. Further combination with αPD-L1 elicits a robust abscopal effect, extending this local tumor control to systemic antitumor responses. With the ability to induce ALS, MOFs can not only enhance RT but also provide a unique mechanism for immune cell infiltration and in situ vaccination. MOFs thus represent a unique nanoplatform for developing innovative cancer treatments.

MATERIALS AND METHODS

For detailed materials and methods, please see the Supplementary Notes. The study protocol involving animals was reviewed and approved by the Institutional Animal Care and Use Committee at the University of Chicago (ACUP-72408).

ACKNOWLEDGEMENTS

We thank Geoffrey T. Nash, Jianming Mao and Xuanyu Feng for their experimental help and discussions. We acknowledge Dr. Christine Labno for help with confocal imaging and whole slide scanning, Dr. Shihong Li for help with the histology study, Dr. Eugene Xu and Chun-yi Chiang for help with live imaging and Hongyuan Jiao for help with NanoString.

FUNDING

We acknowledge funding from the National Cancer Institute (1R01CA253655) and the University of Chicago Medicine Comprehensive Cancer Center (NIH CCSG: P30 CA014599).

Author contributions

T.L. and W.L. conceived the project. T.L. and X.J. designed and performed the majority of the experiments and data analysis. Y.F., E.Y., J.L., and L.T. assisted with experiments and figure production. T.L., X.J., and W.L. wrote the manuscript.

Conflict of interest statement. W.L. is the founder of Coordination Pharmaceuticals Inc., which licensed the nMOF-mediated RT-RDT technology from the University of Chicago.

REFERENCES

Author notes