Synthesis and characterization of RDPNs@diABZIs. MeβCD-based supramolecular polyrotaxanes (MSPs) with diselenide-bridged axle polymer by end-capping with two NIR fluorescence probes IR783 (IR783-PEG-MeβCD-PDS) was synthesized, as detailed in the experimental section (Supplementary Fig. 1)36,37. As a control, MSPs with dicarbonyl-bridged axle polymer by end-capping with two NIR fluorescence probes IR783 (IR783-PEG-MeβCD-PTD) were also synthesized. Eventually, polyrotaxane IR783-PEG-MeβCD-PDS was self-assembled into redox-responsive diselenide-bridged polyrotaxanes nanoformulations (RDPNs) in aqueous solution, with nonresponsive polyrotaxanes nanoparticles (NPNs) as control. STING agonists diABZIs were further encapsulated into the hydrophobic core of the nanoparticles (RDPNs@diABZIs) through simple ultrasonication38,39. H nuclear magnetic resonance (1H NMR) spectra were used to characterize the successful synthesis of polyrotaxane (Supplementary Figs. 2–5). The peaks related to the protons on MeβCD were observed in the 1H NMR spectrum of IR783-PEG-MeβCD-PDS, confirming the formation of a mechanically interlocked molecule (Supplementary Fig. 5). Gel permeation chromatography (GPC) curves were shown that the average molecular weight (Mn) of IR783-PEG-MeβCD-PDS was determined to be 3.94 kDa, 2.26 kDa higher than PEG-PDS (Mn = 1.68 kDa), providing direct evidence for the formation of a polyrotaxane with approximately 17 MeβCD units (Fig. 2a). Similar to the above results, the control polyrotaxane IR783-PEG-MeβCD-PTD was also successfully synthesized (Supplementary Fig. 6). Thermogravimetric analysis also confirmed the successful preparation of IR783-PEG-MeβCD-PDS through host-guest chemistry (Fig. 2b). The UV-visible spectrum of IR783 and IR783-PEG-MeβCD-PDS has showed similar curves, indicating that IR783 as a stopper was successfully capped (Fig. 2c and Supplementary Fig. 7). Compared with the maximum absorption peak of IR783-PEG-MeβCD-PDS in dimethyl formamide (DMF), RDPNs in aqueous solution have exhibited negligible blue shift of 20 nm, which may be caused by the solvent (Fig. 2d and Supplementary Fig. 6). The diABZIs-loaded RDPNs (RDPNs@diABZIs) were prepared with a diABZIs loading efficiency of 5.67%. Dynamic light scattering (DLS) results revealed that the hydrodynamic sizes of RDPNs, NPNs, and RDPNs@diABZIs were 94.09 ± 6.8 nm, 131.71 ± 9.4 nm, and 142.7 ± 8.5 nm, respectively (Fig. 2e). Additionally, the surface zeta potentials of these constructed NPs were found to be negatively charged, which were − 1.05 ± 0.17 mV, -0.68 ± 0.11 mV, and − 1.35 ± 0.21 mV, respectively (Fig. 2f). The degradation behavior of RDPNs was investigated under oxidative conditions (H2O2), mimicking the in vivo TME condition. As revealed by transmission electron microscopy (TEM) images, compared with the uniform spherical structure of RDPNs under PBS conditions, RPDNs aggregated and swelled after 12 h of exposure to H2O2 solution and fully collapsed into small fragments within 24 h. (Fig. 2g). Additionally, DLS results also revealed that as the concentration of hydrogen peroxide increased, the hydrodynamic sizes were gradually increased, indicating the RPDNs aggregated and swelled (Fig. 2h). On the contrary, the nonresponsive poly-MeβCD nanoparticles (NPNs) showed negligible changes in morphology and hydrodynamic sizes, and no degradation in the oxidative condition after 24 h of incubation (Supplementary Fig. 8). To investigate whether the degradation of RPDNs can induce the release of diABZIs, high-performance liquid chromatography (HPLC) was used to monitor the release behavior. RPDNs showed an increasing release of diABZIs as the concentration of hydrogen peroxide increased (Fig. 2i and Supplementary Fig. 9). They showed a burst release of approximately 78.5% of the cumulative total diABZIs in the oxidative condition after 6 h of incubation, fitting well with the degradation behavior.
Cancer-cell stiffness is enhanced by depletion of cholesterol with RDPNs. Contributing to abnormal generation and cross-linking of extracellular matrix proteins, tumors are typically harder than their corresponding normal tissues40–42. Interestingly, membranes of cancer cells are often softer than non-malignant cells43–45. Cancer cells soften cell membranes via increasing cholesterol, which reduces the force exerted by T-cell synapses to help cancer cells evade T-cell killing. MeβCD, a cholesterol-depletion molecule, could increase cancer cell membrane stiffness (Fig. 3a). We quantified the cholesterol levels in murine tumor tissues, normal tissues, and related cells using Amplex Red cholesterol detection kit. The total cholesterol levels in tumor tissues isolated from 4T1-bearing mice increased by 2.67 times compared to adjacent muscle tissues (Fig. 3b and Supplementary Fig. 10). Compared to HC11 cells (murine breast epithelial cells), 4T1 cells (murine breast cancer cells) showed a significant increase in cholesterol levels (Fig. 3c). Additionally, elevated cholesterol levels in different types of tumor tissues were also analyzed (Supplementary Fig. 11). These results provide evidence that a common feature of cholesterol levels increase in different cancer types compared with normal cells.
To investigate whether MeβCD released from RDPNs in tumors or H2O2 can delete cholesterol, we quantified the cholesterol levels after treatment with indicated treatments. Firstly, we validated whether using MeβCD reduces cancer cell membrane cholesterol levels in vivo and in vitro. The membrane cholesterol level of cancer cells isolated from different cancer-type tumor-bearing mice was reduced after intratumoral injecting MeβCD, while the intracellular cholesterol level showed negligible changes (Supplementary Fig. 12). Therefore, plasma membrane cholesterol levels could be controlled via MeβCD in various mouse cancer cell lines.
To verify whether RDPNs also can reduce membrane cholesterol levels, tumor-bearing mice were used as models that were treated with MeβCD, RDPNs, or NPNs intratumorally (i.t). We found that the membrane cholesterol level of 4T1 cancer cells treated with RDPNs dropped markedly to only 47.8% relative to PBS group (G1). In contrast, the membrane cholesterol level of 4T1 cancer cells treated with NPNs showed negligible changes (Fig. 3d). Moreover, the results in vitro were consistent with in vivo (Fig. 3e). Further, the membrane cholesterol level of 4T1 cancer cells were visualized or quantitatively analyzed via staining the Filipin Ⅲ, a fluorescent dye that specifically binds to cholesterol (Figs. 3f, g and Supplementary Fig. 13). The fluorescence images and the flow cytometry analysis all revealed the outstanding cholesterol-depleting capability of RDPNs comparable to the same concentration of MeβCD. These results provided evidence that the diselenide-structured RDPNs showed a burst release of MeβCD compared to carbide-structured NPNs in cancer cells or tumors and that RDPNs can reduce membrane cholesterol levels as MeβCD.
To investigate whether cancer-cell stiffness could be enhanced by depleting membrane cholesterol, we directly measured single-cell cortical stiffness using Atomic Force Microscopy (AFM)46 (Fig. 3h). We found that the cortical stiffness of 4T1 cancer cells treated with MeβCD and RDPNs decreased by 3.02 times and 2.89 times, respectively, relative to PBS group (G1). Moreover, similar results were observed in other cancer cells (Supplementary Fig. 14). These results demonstrated that cholesterol depletion in the cell membrane contributes to cancer cell stiffness.
Cancer-cell stiffness by RDPNs enhances T-cell-mediated cytotoxicity through T-cell forces. Cancer-cell softness impairs T-cell forces further preventing membrane pore formation by perforin47. Cancer-cell stiffening via cholesterol depletion allowed for accelerating the speed of pore formation by synergistic mechanical force and perforin (Fig. 3i). To investigate whether cytotoxicity of different stiffness of cancer cells is related to perforin concentrations in the absence of activated T cells, differently pre-treated 4T1 cells were co-cultured with varying concentrations of perforin (Fig. 3j). Interestingly, in the absence of activated T cells, the cell viability of cancer cells of different stiffness was independent of the concentration of perforin. Next, we co-cultured activated Pmel CD8+ T cells with pre-treated 4T1 cells. Similar to pre-treated with MeβCD, RDPNs-pretreated 4T1 cells accelerated the speed of pore formation compared to native 4T1 cells (G1) (Fig. 3k). These findings indicated that cancer cell stiffening enhances T cell force-mediated cytotoxicity, mediated by increased membrane pore formation by perforin.
In vitro cytotoxicity of RDPNs. Compared with MeβCD cytotoxicity against L929 cells, cytotoxicity of L929 cells treated with RDPNs or NPNs was negligible (Fig. 4a and Supplementary Figs. 15, 16). We next monitored the cytotoxicity of 4T1 cells treated with RDPNs or NPNs. The result revealed greatly enhanced cytotoxicity in the group treated with RDPNs than the group treated with NPNs (Fig. 4a). Further, the analysis of live/dead staining and Annexin V-FITC/propidium iodide (PI) flow cytometry were consistent with the above results (Supplementary Figs. 17, 18).
Activation of cGAS-STING Pathway by RDPNs@diABZIs in vitro. STING agonists diABZIs can activate the cyclic GMP-AMP synthase-stimulator of interferon genes (cGAS-STING) pathway (Fig. 4b)48,49. To verify whether RDPNs@diABZIs enhance secretion of type-I Interferon (IFN-I) in APCs, we detected the levels of IFN-β secretion. cGAMP, as a second messenger to activate STING, rarely elicited the secretion of IFN-β with a half-maximum effective concentration (EC50) of 84.61 ± 3.11 µM in THP-1 (Fig. 4c). This can contribute to its acts as a negatively charged molecule. Interestingly, similar to EC50 of 5.15 ± 0.59µM of STING agonists diABZIs, RDPNs@diABZIs also induced a concentration-dependent secretion of IFN-β in THP-1 cells with EC50 of 5.70 ± 0.71µM (Fig. 4c). Compared with cGAMP, the delivery of diABZIs or RDPNs@diABZIs increased the IFN-β secretion levels by order of magnitude. Similarly, an enhancement in stimulating IFN-β secretion was observed in Bone Marrow Dendritic Cells (BMDCs), the EC50 of diABZIs, RDPNs@diABZIs, and cGAMP was 5.71 ± 0.12 µM, 6.42 ± 0.21 µM, 92.36 ± 6.90 µM (Fig. 4d). Compared to cells treated with free cGAMP, BMDCs stimulated by RDPNs@diABZIs exhibited significantly increasing expression of co-stimulatory molecules CD80 and CD86 (Figs. 4e-h). Western blot analysis revealed a noticeable phosphorylation of STING (p-STING) and IRF-3 (p-IRF-3) in response to diABZIs and RDPNs@diABZIs (Figs. 4i, j). The expression levels of pro-inflammatory cytokines tumor necrosis factor-alpha (TNF-α), interleukin-6 (IL-6), and IL-1β also significantly increased (Figs. 4k-m). These results suggested that RDPNs@diABZIs can activate the cGAS-STING pathway and have the potential to stimulate T-cell activation.
Biodistribution and pharmacokinetic analysis. We evaluated the pharmacokinetics profiles by measuring the concentration of IR783. The elimination half-life (t1/2) of RDPNs and free IR783, calculated were 1.40 h and 0.31 h, respectively (Fig. 5a). Similarly, t1/2 of RDPNs@diABZIs and diABZIs were 1.5 h and 0.25 h, respectively (Fig. 5b). These provide evidence that nanomaterials can extend the blood circulation time of IR783 compared to free IR783. Additionally, the area under the curve (AUC) for RDPNs was about 10 times than IR783 (Fig. 5c). The concentration of IR783 in tumors sacrificed from injected with RDPNs was significantly enhanced compared with that injected with free IR783 (Fig. 5d). Therefore, we have reason to believe that RDPNs exhibit long-term retention characteristics in the tumor tissue.
Further, to investigate the biodistribution of RDPNs in the tumor-bearing mice model, we intravenously injected RDPNs or NPNs and monitored the fluorescence of IR783 using small animal in vivo imaging. In vivo fluorescence images and quantitative analysis all indicated enhanced tumor fluorescence signal with a peak at 12 h post-injection of RDPNs or NPNs (Figs. 5e, g). The fluorescence intensity gradually decreases after 24 h, but fluorescence is still visible at the tumor site at 48 h. The above results indicated a strong drug retention effect of RDPNs or NPNs. Furthermore, ex vivo images sacrificed at 48 h showed significant nanomedicines accumulation in the tumor site and little accumulation in the major organ (heart, liver, spleen, lung, Kidney), which indicated that nanomedicines were selectively accumulated to tumors (Figs. 5f, h).
RDPNs@diABZIs shift tumor immune microenvironment. The activation of the cGAS-STING pathway can prime innate and adaptive immune responses, which is beneficial for anti-tumor systemic immune responses50–52. An orthotropic 4T1 mouse breast tumor model was established. Surface expression of CD86 of dendritic cells (DCs, CD45+CD11c+CD86+) of the tumor-draining lymph node (TDLN) in RDPNs@diABZIs-treated mice was increased (Fig. 6a and Supplementary Fig. 19). Additionally, the total cell number and the fraction of activated DCs also were markedly increased, which exhibited approximately 2.23-fold activated DCs than those treated with PBS. (Fig. 6b and Supplementary Figs. 19, 20). The total cell number and the fraction of M1-like macrophages (CD11b+F4/80+CD86+) in the TME was significantly increased in mice-treated RDPNs@diABZIs than other groups (Figs. 6c, d and Supplementary Figs. 21, 22). On the contrary, RDPNs@diABZIs decreased the total cell number and the fraction of M2-like macrophages (CD11b+F4/80+CD206+) in TME compared with that received PBS only (Figs. 6e, f and Supplementary Figs. 23, 24). These results indicated repolarization or recruitment of macrophages with reduced immunosuppressive capacity. Notably, compared with those treated with others, the tumors in mice that received RDPNs@diABZIs increased the total number and the fraction of natural killer cells infiltrated (Fig. 6g and Supplementary Figs. 25–27). We further verified whether downregulating immune cells with immunosuppressive capacity and upregulating that with immunoactivated capacity in TME can potentiate anti-tumor immune responses. RDPNs@diABZIs led to intratumoral cytotoxic CD8+ (CD3+CD8+) and CD4+ (CD3+CD4+) T cells increased compared with other groups (Figs. 6h, j and Supplementary Figs. 28–30). Immunofluorescence staining of the CD8 marker in tumor slices retrieved after mice treated with RDPNs@diABZIs showed enhanced infiltration of CD8+ T cells (Supplementary Fig. 31). The mice that received RDPNs@diABZIs elicited 2.43-fold more CD8+ T cells and 1.85-fold more CD4+ T cells than that treated PBS only. Immune cell populations influence cytokines secretion levels, which also have critical roles in regulating the composition of immune cells. The concentrations of intra-tumoral TNF-α, IFN-γ, and granzyme B were markedly increased at 2 d after RDPNs@diABZIs treatment (Fig. 6k). Together, these results verified that RDPNs@diABZIs enhanced immunoactivated cells and decreased immunosuppressive cells in TME, which potentiates anti-tumor immune responses.
RDPNs@diABZIs trigger robust anti-tumor therapeutic efficacy. Based on the above results, we next evaluated the anti-tumor effect of RDPNs@diABZIs in vivo. An orthotropic 4T1-luc mouse breast tumor model was established, and mice were randomly divided into five groups: PBS (G1), NPNs (G2), RDPNs (G3), diABZIs (G4), and RDPNs@diABZIs (G5). Mice were treated when the tumor volume reached approximately 100 mm3 and were intravenously injected with indicated formulations on days 0, 2, and 4. Mice treated with NPNs did not suppress tumor growth (Figs. 7a-d). On the contrary, the tumor volume of mice treated with RDPNs was inhibited. These demonstrated that MeβCD released from RDPNs reducing membrane cholesterol promotes cancer cell stiffness and induces its apoptosis, but it can’t be released from NPNs. Undoubtedly, tumor growth rates markedly were inhibited in mice that received diABZIs only, which contributed to enhanced anti-tumor immune responses induced by diABZIs-activated cGAS-STING pathway. Notably, mice treated with RDPNs@diAZBIs showed robust tumor regression and long-term tumor-free survival of at least 2 months in about 62.5%. Consistent with the above results, mice treated with RDPNs@diAZBIs obtained more efficient therapeutic effects than other group according to the hematoxylin-eosin (H&E) and TUNEL staining results (Supplementary Figs. 32, 33). Next, blood chemistry analysis, including alanine aminotransferase (ALT), lactate dehydrogenase (LDH), aspartate transaminase (AST), urine acid (UA), creatine kinase (CK), and creatinine (CRE) were performed for long-term toxicity evaluation after treatments. The results showed all treatments did not induce significant side effects in the mice (Supplementary Fig. 34). Additionally, there were no significant pathological changes in major organs (heart, liver, spleen, lung, Kidney) H&E slices images (Supplementary Fig. 35).
We next explored the mechanism of tumor regression and durable anti-tumor effect in mice-treated RDPNs@diAZBIs. To test the immunological memory, the memory T cells (CD8+CD44highCD62Llow) of the spleen retrieved from mice-treated indicated treatments were analyzed using flow cytometry. Compared to the PBS group (G1), mice treated with RDPNs@diABZIs exhibited a 2.82-fold increase in the fraction of memory T cells, a better result than other groups (Figs. 7e, f and Supplementary Fig. 36). This evidence suggested that mice established strong immune memory after RDPNs@diABZIs treatment. To further determine whether the anti-tumor immune responses induced by RDPNs@diABZIs were durable, the mice that survived following the first inoculation were re-challenged with 4T1-luc tumor cells by intravenous administration. Noticeable tumor nodules bioluminescence in lung in vivo and ex vivo was observed in mice of the control group (Figs. 7g-j and Supplementary Fig. 37). In contrast, mice treated with RDPNs@diABZIs showed no tumor nodules bioluminescence in lung, indicating the establishment of durable anti-tumor immune responses. These results provided evidence that RDPNs@diABZIs elicited robust anti-tumor efficacy and induced durable anti-tumor immunological memory.