Bioactive nanoparticles for boosting tumor immunotherapy via Ca2+ interference mediate TME reprogramming and PD-L1 depletion

The ecacy of antitumor immunotherapy is often limited by the highly tumor immunosuppressive microenvironment (TME). Here, a previously unknown strategy is proposed to synergize the “Ca 2+ interference” mediate TME reprogramming and Ca 2+ -activating PD-L1 depletion for immunotherapy. We discovered that calcium containing nanoparticles could induce signicant “Ca 2+ interference” effect and simultaneously resetting tumor-associated macrophages toward M1 phenotype and promoting in situ antigen release. Making them as powerful immunostimulants for TME reprogramming without notable toxicity. Meanwhile, we designed a circular aptamer-DNAzyme conjugate (cAD) with tumor-targeting ability and its catalytic shear activity can be specically activated by Ca 2+ , which reduces potential autoimmune disorders of PD-L1 antibody. By simple integrating the ultra-high pH sensitive calcium peroxide nanoparticles with cAD into one nanosystem, we demonstrate the nanosystem not only arrests primary tumor progression, but also prevents lung metastasis. In addition, it offers a long-term immunological memory function, which can protect against tumor rechallenge. Therefore, this work provided an alternative strategy for boosting tumor immunotherapy.


Introduction
Immunotherapy is rapidly rising as a promising strategy for the treatment of tumor 1, 2 . Nevertheless, its e ciency is limited by immunosuppressive tumor microenvironment (TME) and poor immunogenicity of tumor cells 3,4 . Although multiple factors in TME contribute to the failure of immunotherapy, one dominant reason is the recruitment of suppressive immune cells, such as tumor-associated macrophages (TAMs), regulatory T cells (Tregs) and cancer-associated broblast (CAFs) etc. al, which always lead to immune evasion by limiting T cell in ltration and functions 5 . Besides, tumor cells usually escape the immune surveillance via concealing neoplasma antigen (poor immunogenicity) 6 . Therefore, selectively reprogramming immunosuppressive TME and immunogenicity is recognized as an extremely promising strategy for immunotherapy. One relatively e cient approach is to develop nanomedicine to deliver immunomodulatory and therapeutic drugs to tumors 7,8 . Despite the advances in nano immunotherapy systems, compatibility of loaded drugs including the types and loading amount need to be precisely adjusted. Furthermore, distinct targets of loaded drugs require harsh differentiated drug release between immune and tumor cells, thus putting forward an extremely formidable task for the design and fabrication of nano carriers, which introduces more uncertainty for their further clinical application.
There is growing evidence that metal ions are critically implicated in regulating both innate and adaptive immune response, one possible way to overcome the above challenges is "metal ions interference". Particular Ca 2+ , as a second messenger, is a pivotal player in immune cell activation and phenotypic transformation 9 . For example, with the increase of cytosolic Ca 2+ , the central transcriptional regulators for lymphocyte and macrophage activation include NFAT, NF-κB, the kinase Jnk and calmodulindependent kinase are highly activated [10][11][12] . Furthermore, the effector function of NK, macrophages and CTL, as well as their proliferation and cytotoxicity to tumor cells largely depend on cytosolic Ca 2+ .
Interestingly, we and other groups also found that Ca 2+ overloading in tumor cells could lead to the enhanced release of multiple damage-associated molecular patterns(DAMPs), such as up regulation of "eat me" signal calreticulin (CRT), release high mobility group protein 1 (HMGB1), which signi cantly improved the immunogenicity of tumor cells 13,14 . Given the multiple immunostimulatory effect, we propose that calcium containing nanoparticles may serve as attractive materials for developing tumor immunomodulation platform. Besides, the activation of adaptive immune response often leads to the upregulation of PD-L1 expression in tumors, which needs further immune checkpoint blockade 15 . But the general application of PD-L1 antibody often leads to signi cant autoimmune disorders. Therefore, it's also highly desired to locally suppress the expression of PD-L1 on tumor cells.
Herein, we propose applying dual-targeted immunomodulatory strategy using "Ca 2+ interference" to stimulate antitumor immune response and activatable gene silencing in tumor cells to deplete PD-L1.
Notably, TAMs, as the major tumor-in ltrating immune cell population (over 50% of the cells in tumor), are believed to play critical roles in immunosuppressive TME. Moreover, as a professional phagocyte, TAMs are also believed as the main reservoir of nanoparticles in tumor 16 . As expected, we demonstrated that nano enable "Ca 2+ interference" conferred strong "re-educating" effect on TAMs, and surprisingly, stimulated both innate and adaptive immune responses in tumors.
To achieve local PD-L1 suppression, we designed a circular aptamer-DNAzyme conjugate (cAD) loaded in the designed calcium nanoparticle, which can be speci cally released in tumor site for e ciently PD-L1 gene silencing. Comparing with traditional siRNA-based gene silencing strategy, cAD holds high stability in vivo and can be spontaneously uptaken by tumor cells, due to the circular structure and functional aptamer sequence 17 . Besides, DNAzyme are catalytic nucleic acids that can mimic the function of endonucleases for cleavage of speci c mRNA with multiple turnovers, which may be more effective for PD-L1 gene silencing on a per molecule basis 18,19 . However, DNAzymes always need su cient ion cofactor supply for activation, which just can be solved by our "Ca 2+ interference" strategy 20,21 . Since Ca 2+ release was also speci c in tumor site in response to the relatively low pH, this integrated system showed high tumor speci city for local PD-L1 suppression.
In this study, we designed nanoparticles that contain a calcium core, cAD and a hydrophilic shell (SDS-PEG2000) for e cient immunotherapy. This immunotherapeutic nanoplatform exhibits multifunctional characteristics, including (i) improve stability in blood circulation, thus leading to the enhanced tumor accumulation of NPs 22 ; (ii) super sensitive pH response, leading to the site-speci c release of cAD in TME, cAD target to tumor cells for Ca 2+ -activating PD-L1 depleting; (iii) calcium cores ingested by TAMs achieve e cient "Ca 2+ interference", leading to the dephosphorylation of p38, activation of NF-κB to promote nuclear translocation, and release of NLRP3 in ammasomes, obtaining a synergetic activation of both innate and acquired immune responses 23,24 ; (iv) "Ca 2+ interference" in tumor cells that can activate calmodulin and serve as an inducer to the release of multiple tumor antigens (CRT, HMGB1), further promoting tumor cell immunogenicity. We demonstrated that the simple nanoplatform exhibited potent e cacies as a treatment against both primary and metastatic tumors without obvious side effects.

Results
CaNP resets tumor-associated M2 macrophages to M1 phenotype Calcium nanocarriers (calcium peroxide (CaO 2 ), calcium phosphate (Cap) and honeycomb calcium carbonate (HOCN)) were synthesized according to the methods previously reported [25][26][27] . Monodispersed nanoparticles with sizes of ~150 nm, ~50 nm and ~200 nm were obtained, respectively (Fig.1A). CaO 2 nanoparticle (CaNP) was selected for realizing "Ca 2+ interference" due to its excellent acid sensitivity, most e cient Ca 2+ release performance and distinctive positive electrical property, which facilitates subsequent cAD loading (Fig.1B, C and Supplementary Fig.1A, B). As a second messenger, Ca 2+ is believed to play critical roles in immune cell activation and phenotypic transformation. Here, we asked whether CaNP would induce the polarization of TAMs. Mice bone-marrow-derived macrophage cells (BMDMs) were harvested and treated following the operation procedure in Fig.1D. With prolonged incubation time, a gradual increase of Ca 2+ levels in BMDM-M2 was observed (Supplementary Fig. 2A, B), suggesting the e cient endocytosis of BMDM-M2, and CaNP showed a neglectable cytotoxicity on BMDM-M2 ( Supplementary Fig. 2C). More important, we found that CaNP treatment group visually indicated a signi cant increased in M1-phenotype (CD80 + ) and a decreased in M2-phenotype (CD206 + ) compared with Cap NP and H 2 O 2 treatment group (Fig.1E), indicating that CaNP had a synergetic enhancement on "re-educating" TAMs. The decrease in immunosuppressive factor interleukin-10 (IL- 10) and the increase in immune activating factor interleukin 12p70 (IL-12p70) secretion of CaNP treatment group further con rmed the successful phenotypic regulation (Fig.1F).
With the change of macrophage phenotype, its phagocytosis to B16 cells also showed an obvious enhancement in vitro ( Fig.1G). At the same time, two main chemokines including CXCL9 and interferon-γ (IFN-γ) were observed a signi cant increase in CaNP treatment BMDM-M2 (Fig.1H). Speci cally, CXCL9 is considered to be an important chemokine for the recruitment of CD8 + T cells into tumors. Therefore, we used transwell to investigate the recruitment effect of CaNP treated BMDM-M2 on CD8 + T cells in vitro ( Fig.1I, left). As expected, CaNP treated group induced obviously CD8 + T cells in ltration in the below room, almost as the positive control (LPS/ IFN-γ treated group) (Fig.1I, right). What's more interesting, besides the recruitment of CD8 + T cells, the proliferation of CD8 + T cells was also promoted by CaNP treatment. In brief, 5,6-carboxy ourescein diacetate succiniidyl ester (CFSE) labeled CD8 + T cells were cocultured with pretreated BMDM-M2, after CaNP treatment, the number of CD8 + T cells increased from 10.7% to ~30.1% and induced more obvious cell proliferation than other groups (Fig.1J). These data demonstrate that CaNP could e ciently regulate the phenotype and function of TAMs via "Ca 2+ interference", which is essential for stimulating antitumor immune response.
We discovered that CaNP could induce signi cant "Ca 2+ interference" effect and have a synergetic enhancement on resetting TAMs toward M1 phenotype, which was also con rmed by the highest expression of iNOS in CaNP group compared with CaP and H 2 O 2 treatment ( Fig. 2A and Supplementary   Fig.3A left). And then, we investigated the underling molecular mechanism. According to the previous studies, mitogen-activated protein kinase (MAPK) and nuclear factor-κB (NF-κB) signal molecules are highly activated in M1 macrophages, and they also play the key role in regulation of proin ammatory factors, such as IL-12p70. Therefore, a mainly MAPK members (p38) was rstly analyzed. What excited us is that only calcium containing nanoparticle (CaNP and CaP) induced the phosphorylation of p38 in BMDM-M2 macrophages ( Fig. 2A and Supplementary Fig.3A right). To further conform this, we used cyclosporin A (CsA), a Ca 2+ signaling inhibitor to pre-treatment with BMDM-M2, the effect of CaNP on p38 phosphorylation was signi cantly blocked ( Fig. 2B and Supplementary Fig. 3B), suggesting that "Ca 2+ interference" promotes the development of M1 phenotype through p38 molecule. Secondly, we tested the in uence of "Ca 2+ interference" on NF-κB by examining its nuclear translocation. Compared with the control group, we found CaNP treatment resulted in the colocalization of NF-κB with nucleus in BMDM-M2 macrophages, while the nuclear translocation could be signi cantly blocked by CsA (Fig. 2C). The above results illustrated that both the activation of MAPK (phosphorylation of p38) and nuclear translocation of NF-κB are involved for "Ca 2+ interference"-mediated TAMs resetting. Besides the phenotypic changes of TAMs, our previous research showed that "Ca 2+ interference" could increase the reactive oxygen species (ROS) level of mitochondrion (25), and ROS-mediated mitochondrial dysfunction triggers the NLRP3-in ammasome pathway for activating the innate immune response of macrophages 28 . According to this, IL-1β secretion of BMDM-M2 macrophages with different treatments was investigated. As expected, a signi cant increase of IL-1β secretion was observed in BMDM-M2 treated with CaNP, while showing an obvious blockade after catalase (CAT) treatment (Fig. 2D), indicating that "Ca 2+ interference"-mediated IL-1β secretion is ROS depended. Active caspase-1 converts pro-IL-1β to mature IL-1β that stimulates in ammation 29 . Therefore, the increase of caspase-1 expression in CaNPtreated BMDM-M2 further con rmed the result of IL-1β secretion (Fig. 2E), which can stimulate the strong innate immune response of macrophages.
For further investigation of the molecular mechanisms involved in the CaNP-induced macrophage polarization. We obtained the differentially expressed genes in CaNP+CAT treated BMDM-M2 cells by transcriptome analysis. Several genes related to in ammation suppression phenotype were signi cantly downregulated (i.e. Arg1, CD163, Il4ra, Il13ra2), while upregulating the expression of in ammation-related genes (i.e. Tnfsf15, Tlr12, Pik3cb) ( The design, composite and preparation diagram of cAD is shown in Fig. 3A. In brief, phosphorylated cellspeci c aptamer (Sgc8) 30 and Ca 2+ -speci c DNAzyme named EtNa 31 were hybridized with complementary sequences, and then two gaps were sealed with T4 DNA ligase (See Supplementary table  1 for speci c sequence information). The results of agarose gel electrophoresis illustrated the successful synthesis of CAD (Fig. 3B). We then investigated whether cyclization enhances the stability of EtNa in biological medium. As shown in Fig.3C, D, Sgc8 and EtNa were signi cantly degraded after incubation in exonucase I or DMEM-10% fetal bovine serum, while the integrity of each conjugate was still evident for CAD under the same conditions. (Fig. 3D). Next, we examined whether cyclization affected the performance of EtNa and Sgc8. Fig. 3E showed that cAD had an enzyme cleavage effect, and the cleavage efficiency is Ca 2+ concentration-dependent. Furthermore, we performed fluorescent detection to investigate the cleavage efficiency of cAD (Fig. 3F), and both the enzyme cleavage effect of EtNa and cAD showed a strong Ca 2+ activated property. Interestingly, cAD displayed a signi cant increase in the cleavage efficiency than that of EtNa ( Fig. 3G), indicating that the enhanced stability of cAD improves the a nity with the substrate. The cyclization of Sgc8 and EtNa to obtain cAD not only brings the enhanced stability and improved cleavage efficiency, but also makes cAD have a high tumor speci city for tumor cell speci c PD-L1 suppression. Therefore, the tumor cell speci c reorganization and targeting ability of cAD was con rmed via ow cytometry and confocal microscope (Fig. 3H).
DNAzymes need su cient ion cofactor supply for activation, which just can be solved by our "Ca 2+ interference" strategy. In turn, Ca 2+ interference also can active cAD for PD-L1 depleting. A signi cant Ca 2+ interference effect was observed in B16 tumor cells after treated with CaNP, and the level of Ca 2+ in tumor cells increased with the prolongation time of CaNP incubation (Fig. 3I). Interestingly, we also observed that the lysosomal escape of cAD was signi cantly enhanced with the addition of CaNP ( Supplementary Fig.4), which is advantageous to the PD-L1 mRNA cleavage. Accordingly, the expression level of PD-L1 in B16 tumor cells treated with CaNP+cAD was signi cantly reduced compared with those treated with CaNP or cAD alone (Fig. 3J). This phenomenon was also con rmed by qRT-PCR (Fig. 3K, primer sequences are shown in Supplementary table 2). Notably, with Ca 2+ interference effect induced by CaNP, the PD-L1 level and mRNA level of cAD signi cantly reduced, further indicating the Ca 2+ activating property of cAD. Furthermore, our previous studies have shown that the increase of intracellular Ca 2+ promotes the release of multiple tumor antigens, such as the exposure of CRT and the release of HMGB1 in CT26 cells 27 , and this effect of Ca 2+ interference was also con rmed in B16 tumor cell for further improving the tumor cell immunogenicity (Fig. 3L, M). These results indicated that cAD combined with CaNP can not only improve the tumor cell immunogenicity, which is conducive to the occurrence of antitumor immunity, but also reduce the expression of PD-L1 in tumor cells and restore the function of T cells.

CaNP@cAD-PEG effectively inhibites tumor growth
After evaluating the performance of CaNP on phenotypic transformation of TAMs and cAD on tumor cell speci c PD-L1 depleting, respectively, we aimed to combine the two into one nanoparticle for e cient delivery in vivo. Firstly, cAD was loaded onto the surface of CaNP due to its positive electrical property, and the optimal loading ratio of CaNP to cAD is 10:1 (V: V, CaNP: 5 mg/mL, cAD: 10 μM, Supplementary  Fig.5A). After cAD loading, a hydrophilic shell (DSPE-PEG2000) was then modi ed to obtain CaNP@cAD-PEG (Fig. 4A). According to the result of TEM (Fig.4B), compared with the hollow structure of CaNP, a sharp contrast rise was observed in CaNP@cAD-PEG, and the particle size was ~180 nm in diameter, a little larger than that of CaNP (~150 nm, Fig. 1A and Supplementary Fig.5B). The successful preparation of CaNP@cAD-PEG was also con rmed by the dynamic change of surface charge during preparation process and the result of element co-localization by Mapping (Fig. 4C, D). The site-speci c release of cAD from CaNP@cAD-PEG in TME is critical for simultaneously regulating the phenotype and function of TAMs and reducing the expression of PD-L1 in tumor cells. As part of the design, CaNP with ultra-high pH sensitivity was used as the calcium core, therefore, CaNP@cAD-PEG could achieve e cient super cial disintegration at pH 6.5 within 3 h. Compared with pH 7.4, the morphology of CaNP@cAD-PEG signi cantly changed, and seem likely returned to CaNP (Fig. 4B), indicating the release of cAD. The release rate was calculated to 75%, which is as high as pH 5.0 (Fig. 4E). In addition, the release of Ca 2+ at pH 6.5 was also much lower than that at pH 5.0, further con rming the super cial disintegration of CaNP@cAD-PEG. At pH 5.0, more than 85% of total Ca 2+ was released from CaNP@cAD-PEG (Fig.4F), which provides plenty of Ca 2+ for regulating the phenotype of TAMs and initiating the shearing effect of cAD (Fig.4G). Interestingly, a signi cant charge reversal effect of CaNP@cAD-PEG was found after incubating at pH 6.5 for 3 h, from -14.1 mV to 16.3 mV ( Supplementary Fig.5C), which is also bene t to the rapid phagocytosis by the main cells (TAMs and tumor cells) in tumor tissue, and play the role of calcium interference.
Motivated by these results, we moved on to explore the feasibility in B16 tumor-bearing mice. Firstly, the biodistribution of CaNP@cAD-PEG was investigated. After 12 hours of tail vein injection, main tissues (heart, liver, spleen, lung, kidney, tumor) were harvested for inductively coupled plasma mass spectrometry (ICP-MS) of Ca 2+ . The result stated the tumor accumulation characteristics of CaNP@cAD-PEG ( Supplementary Fig. 6A), likely due to the enhanced permeability and retention (EPR) effect of tumor. Then, in vivo antitumor e ciency was evaluated according to the schematic diagram (Fig. 4H). The tumor volume and body weight of the mice with different treatments were recorded during the period (Fig.  4K, L and Supplementary Fig. 6B). Compared with the rapid increase of tumor in saline group, there was signi cant antitumor activity in the CaNP group (tumor inhibition rate: 35%), indicating the stimulating antitumor immune response induced by Ca 2+ interference. Importantly, the group treated with CaNP@cAD-PEG exhibited the remarkable highest tumor growth inhibition effect (tumor inhibition rate: 82%) than other groups (Fig. 4K, L). Tumor images of different groups further demonstrated that CaNP@cAD-PEG group presented the best antitumor e cacy ( Fig. 4I and Supplementary Fig. 6C). Without abnormalities in blood biochemical indexes (The relevant results and description are shown in Supplementary Fig. 6D), CaNP@cAD-PEG signi cantly extended the survival rate of mice to 85%. In contrast, none of the saline group survived on day 30 (Fig. 4J). Furthermore, pathological analysis revealed that CaNP@cAD-PEG treatment resulted in a large amount of apoptosis/necrosis as well as a signi cant reduction of PD-L1 in tumor (Fig. 4M, N and Supplementary Fig. 6E) without damaging to normal organs (Supplementary Fig. 6E). Importantly, the cardiotoxicity induced by PD-L1 antibody was signi cantly decreased via an alternative Ca 2+ -activating PD-L1 depleting strategy (Supplementary Fig.   6F).
Furthermore, CD8 + T cells was signi cantly increased in CaNP@cAD-PEG treatment as compared to other controls ( Fig. 5C and Supplementary Fig.7D). In addition, immuno uorescence staining visually indicated the marked increase in M1 macrophages ( Supplementary Fig.8A up), natural killer cell (NK, Supplementary Fig. 8A down) and CD8 + T cells in the tumors after CaNP@cAD-PEG treatment (Fig. 5D).
The remarkable increase of NK bene ts from the innate immune response activated by Ca 2+ interference.
The signi cant immune response after CaNP@cAD-PEG treatment was also con rmed by the enlarge of lymph nodes (Supplementary Fig. 8B) and DCs, CD8 + and CD4 + T cell in ltrating in lymph nodes ( Supplementary Fig. 8C). Altogether, those results con rmed the effective activation of innate and adaptive immune responses after CaNP@cAD-PEG treatment.
Inspired by the memory-like property of CD8 + T lymphocytes, the immune memory response against tumor relapse was investigated via a distal tumor model (Fig. 5E). According to the tumor growth curve, tumor tissue images and TUNEL straining, CaNP@cAD-PEG treatment could signi cantly inhibit the growth of distal tumors (Fig. 5F, G and Supplementary Fig.9A, B) without abnormalities in body weight ( Supplementary Fig. 9C). Furthermore, the effector memory T (T EM ) cells in spleen were much higher in CaNP@cAD-PEG group than that in other control groups ( Fig. 5H and Supplementary Fig.9D), indicating that CaNP@cAD-PEG treatment induced effectively immune memory effect. To further evaluated the systemic effect of the immune response, a mouse lung metastasis model was performance (Fig. 5I). According to the results in Fig. 5J, CaNP@cAD-PEG treatment displayed much cleaner and smoother lungs in comparison with other groups, signi cantly suppressing 90% of lung metastasis. These results implied the potential of CaNP@cAD-PEG for long-term antitumor relapse capability and metastasis prevention.

Discussion
The e cacy of antitumor immunotherapy is often limited by the highly tumor immunosuppressive microenvironment (TME). TAMs are believed as the dominant driver that contribute to the immunosuppressive TME via producing immunosuppressive cytokines, suppressing cytotoxic T cell activity while promoting regulatory T cells, and inhibiting B cell signaling 32,33 . Multiple drugs have been exploited to achieve TAM polarization, such as regorafenib, zoledronic acid and miR-155. Despite early preclinical studies and ongoing clinical trials of the conventional small-molecule drugs or nucleic acid drugs, developing new therapeutics with e cient M2→M1 phenotypic conversion, adequate ability to ingested by TAMs in vivo, as well as resulting optimal therapeutic e cacy are still needed 34,35 . Metal ions are essential fundamental of all forms of life which nearly half of all enzymes required for their function and are critically involved in virtually all biological processes. Particularly, increasing evidence has mounted to suggest that metal ions are critically implicated in regulating both the innate and adaptive immune response 9,24,36 . In this study, nanoscale calcium successfully led to the burst increase of cytosolic Ca 2+ in TAMs/tumor cells via bypassing the ion channels. "Ca 2+ interference" in TAMs showed the dephosphorylation of p38, activation of NF-κB, promoting nuclear localization, thus leading to an e cient M2→M1 phenotypic conversion (23). Furthermore, "Ca 2+ interference" also promoted the release of NLRP3 in ammasomes, synergetic leading to activation of both innate and acquired immune responses 37,38 . Besides TAMs, "Ca 2+ interference" in tumor cells showed the activation of calmodulin and served as an inducer to the exposure of CRT and the release of HMGB1 for further promoting tumor cell immunogenicity 39 . Owing to the super pH-sensitivity of calcium nanoparticles, the hydrophilic shell (SDS-PEG2000) on the surface showed improved stability in blood circulation and led to the enhanced tumor accumulation of NPs, while site-speci c separating in TME, exposed the electropositive calcium core, which is bene t for the enhanced endocytosis by TAM/tumor cells.
Besides the immunosuppressive tumor microenvironment, another major challenge faced by tumor immunotherapy is the fact that the programmed-death/programmed-death ligand 1 (PD-1/PD-L1) system 40,41 . Furthermore, activation of adaptive immune response often leads to the upregulation of PD-L1 in tumors. Antibodies against PD-L1 have shown a great deal of clinical activity in various cancer types 42,43 . However, PD-L1 inhibitors are widely distributed in normal cells and often cause adverse autoreactive immune responses and associated side effects 44,45 . Hence developing a precise PD-L1 inhibition strategy for tumor cells is still in great demand 46 . In this study, to achieve targeted gene silencing in tumor cells, a circular aptamer-DNAzyme conjugate (cAD) were designed 47 . Firstly, DNAzyme are catalytic nucleic acids that can mimic the function of endonucleases for cleavage of speci c mRNA with multiple turnovers, which may be more effective for PD-L1 gene silencing on a per molecule basis. Secondly, cAD had a speci c tumor cell targeting ability via splicing a tumor targeting aptamer (sgc8), which guaranteed the speci city of PD-L1 depletion in tumor cells; Thirdly, the closed loop "dumbbell" structure of cAD brought extremely stability to the DNAzyme, even under the action of nuclease, cAD also displayed an equivalent catalytic activity, thus leading to the higher e cient gene silencing e ciency. More importantly, DNAzymes always need su cient ion cofactor supply for activation, which just can be solved by our "Ca 2+ interference" strategy. Since Ca 2+ release was also speci c in tumor site in response to the relatively low pH, this integrated CaNP@cAD-PEG showed high tumor speci city for local PD-L1 suppression. The catalytic shear activity of cAD was speci cally activated by Ca 2+ interference, which in turn reduces potential autoimmune disorders of PD-L1 antibody.
In summary, the rational combination of "Ca 2+ interference" and targeted depleting checkpoint PD-L1 using Ca 2+ -activating DNAzyme yielded improved e cacy in primary and metastatic tumors without obvious side effects. This previously unknown combination provided an alternative strategy for boosting tumor immunotherapy. In addition, this study serves as a proof of concept for Ca 2+ induced immunotherapy, which can be easily tailored to other transition metal ions, such as Mn 2+ , Zn 2+ and Ni 2+12, 48 . Last, this closed loop "dumbbell"-engineering strategy not only improves in vivo stability of DNAzyme, but also can be targeted to various cells types as well as be activated by various metal ions 49 .

Methods
Cell line and cell culture. Synthesis of CaO 2 nanoparticle (CaNP): Firstly, 3% hydrogen peroxide solution and 2 mol/L calcium chloride solution were prepared as the original solution. 1 mL of deionized water and 1 ml of calcium chloride were added to 60 ml of absolute methanol in turn. After stirring vigorously for 10 min, 1 ml of hydrogen peroxide was added. After stirring for 5 min, ammonia was added to the mixing system drop by drop until the color changed to Cambridge blue. Centrifugation at 8000 g for 20 min, repeated washing with methanol twice, and nally dispersed in deionized water.
Synthesis of circular aptamer-DNAzyme conjugate (cAD): 8 μL of EtNa (10 μM) and Sgc8 (10 μM) were added to 18 μL of DNase/Rnase-Free water, and then 10 × T4 DNA Ligase Buffer was added. After incubation at 95 °C for 5 min, the T4 DNA Ligase was rapidly cooled to 16 °C, and then added 2 μL of T4 DNA Ligase. After incubation at 16 °C for 12 hours, the above mixture was heated at 75 °C for another 5 min to obtian cAD.
Investigate the best feeding ratio of CaNP and cAD.
CaNP (5 mg/mL) and cAD (10 μM) were mixed by volume ratio of 1:1, 1:5,1:10,1:15, after incubating at room temperature for 4 hours, and 3% agarose gel was prepared to investigate the best feeding ratio of CaNP and cAD. CANP and cAD were used as the control.

Synthesis of CaNP@cAD-PEG.
CaNP@cAD-PEG was prepared by mixing cAD and CaNP in a volume ratio of 1:10. After incubation for 6 h at room temperature at 300 rpm/min, DSPE-PEG2000 was added in a mass ratio of 2:1 with CaNP. The mixture was shaken at room temperature for another 3 hours and then centrifuged at 8000 g for 20 min to obtain CaNP@cAD-PEG.
Characterization of Sgc8 aptamer targeting e ciency.
Sgc8 and random base sequences were labeled with FAM uorescence, and 1µM of Sgc8 and random sequences were incubated with B16 cells at 37 ℃ for 2 h, respectively. Meanwhile, Sgc8 was coincubated with HL-7702 cells as the control group, and the uorescence intensity of each group was detected by a ow cytometer (BD Accuri®C6, American) and data analysis was performed with FlowJo (Version 10).
Cleavage ability of cAD in vitro. cAD (100 nM) was incubated with an equal amount of PD-L1 mRNA (100 nM) at 37 ℃ for 12 h in Tris-HCl buffer (pH = 8.8) containing different concentrations of CaCl 2 (1 mM, 10 mM, 20 mM, 30 mM, 40 mM and 50mM), respectively. After incubation, the cleavage e ciency was veri ed by polyacrylamide gel electrophoresis.
Study on the stability of cAD in vitro. cAD (100 nM) was incubated with were added to 7.75 μL of DNase/Rnase-Free water, and then 10 × Exo I Buffer and 0.25 μL Exo I was added. After incubation at 37 °C for 1 h, 4 μL 6×DNA loading buffer was then added. The samples were then tested with 3% agarose gel. Meanwhile, Sgc8 and EtNa were also treated as control.
The morphology of CaNP and CaNP@cAD-PEG were detected via TEM (Tecnai G2 F20). The particle size potential is obtained by dynamic light scattering. 100 μg/mL of CaNP and CaNP@cAD-PEG nanoparticles in water solution were applied onto a 300-mesh carbon-copper grid. The excess solution was removed by lter paper. Images were recorded by a transmission electron microscope (JEM 1200EX, JEOL, Japan) operated at a voltage of 120 KV. 1 mL 50 μg/mL CaNP and CaNP@cAD-PEG in deionized water were used to measure the corresponding size distribution and zeta potential by a Zeta sizer (Nano ZS-90, Malvern, UK).

Characterization of CaNP@cAD-PEG in acidic environment
Morphology of CaNP@cAD-PEG in different pH values: CaNP@cAD-PEG (100 μg/mL) was fully dispersed into PBS buffer (pH 7.4, pH 6.5, pH 5.0) for 3 h, respectively. Then the above solution was dripped onto the copper network, and after removing the excess water, the morphology of nano particles was observed by a transmission electron microscope (JEM 1200EX, JEOL, Japan). 50 μg/mL of CaNP@cAD-PEG were incubated in PBS buffer (pH 7.4, pH 6.5 and pH 5.0) at 37 ℃ for 12 h, respectively. Subsequently, the above mixtures were centrifuged at 12000 g for 20 min, the supernatant was collected to incubate with PD-L1 mRNA (500 nM) at 37 ℃ for 12 h in Tris-HCl buffer (pH = 6.8), the cleavage e ciency was veri ed by polyacrylamide gel electrophoresis.
In vitro biodistribution of cAD in B16. qRT-PCR: cells were seeded at 6-well plates at 1×10 5 per well and cultured for 12 h, and then were incubated with different nanoparticles (CaNP, CaNP@cAD-PEG, CaNP@RcAD, cAD) for 12 h, 35 μg/mL nanoparticles or 300 nM cAD was applied, respectively. After washed with PBS, cells were incubated with fresh complete medium for another 12 h. And then, precipitate of cells was completely collected by trypsin digestion and centrifugation (2000 rpm, 5 min). Then mRNA was isolated with a Trizol Regent kit (Invitrogen). One microgramme of RNA sample was utilized to obtain cDNA. One microliter of cDNA sample and speci c primers were used to amplify PD-L1 cDNA according to the manufacturer's parameters (Custom gene qRT-CANPR Quantitation Kit) on a Real-Time CANPR machine.
Western blot analysis: cells were treated according to the above same method. The cells were collected and washed twice with PBS. Cell precipitations were lysed for 1 h at 4 o C. The protein content was measured by Bradford assay. Western blot analysis was conducted using standard method. Each BALB/c mouse was subcutaneously injected with B16 cells (1×10 7 ) on the left armpit. The tumor bearing mice were randomly grouped (n=3) when the tumor volume reached 200 mm 3 . The mice were then intravenously injected with 100 µL of CaNP (10 mg/mL) and saline as control . After 24 h of adimination, mice were then sacri ced for the major organs (heart, liver, spleen, lung and kidney) and tumors. After concentrated hydrochloric acid digestion for 12 h, the deionized water is diluted and ltered to collect the solution for Ca 2+ detection by ICP-MS (Aglient 7800).
Biosafety evaluation of CaNP@cAD-PEG in vivo.
After treatment for 5 times, 5 mice from saline and CaNP@cAD-PEG groups were sacri ced by taking blood from the abdominal aorta and the obtained blood was used for blood chemistry analysis and blood routine analysis.
In vivo anti-tumor e ciency.
The tumor volume was calculated by the equation: Tumor volume = (width 2 ×length)/2. Relative tumor volume was the corresponding changes relative to the initial value measured before treatment. All mice were photographed and euthanized at two weeks post injection, and the tumors, lymph nodes and spleens were taken out and photographed. The collected organs and tumors were xed immediately in 10% paraformaldehyde solution, followed by standard dehydration and para n embedding. The embedded tissues were then sectioned into 4 μm slices and then subjected to standard H&E and Tunnel staining for histological analysis.
The tumors were harvested 2 days post the nal treatments and cut into small pieces to homogenate. After washing twice with PBS, Cell precipitations were lysed for 1 h at 4 o C. The protein content was measured by Bradford assay. Western blot analysis was conducted using standard method.

Detection of the immune memory effect of CD8 + T cells in spleen.
The mice were killed by dislocating executed 10 days post the distal tumor implantation, soaked in 75% ethanol for 10 min, and the limbs of the mice were xed on the foam plate with the abdomen facing up with a needle. Pinch the abdominal skin with forceps, scissors cut a small mouth, expose the abdominal wall, tear the skinwith two forceps in the small mouth, try to expose the abdomen; The abdominal wall was clamped with forceps, the abdominal cavity was cut open, the spleen was removed, and the spleen was removed with an elbow forceps clamp. The removed spleen was placed in PBS and ground with the frosted surface of a sterilized and pre-wetted glass slide; The ground single cell suspension was ltered through a 70 μm cell screen to remove large tissue mass. The centrifuge was centrifuged at 300×g for 5 min in a cryogenic refrigerated centrifuge. Add erythrocyte lysate (3 times volume), gently beat evenly, and lyse for 5-10 min at room temperature; Wash with PBS 7.2 for 2 times, count, resuspend with PBS 7.2 (containing 1 mM EDTA and 2% FBS), adjust the cell density and stained with anti-CD3-FITC, anti-CD8α-PerCP, anti-CD44-APC and anti-CD62L-PE antibodies. Flow cytometry acquisition was performed on a Cyan ow cytometer (BD LSR Fortessa) and data analysis was performed with FlowJo (Version 10).
Analyze of the treatment-induced cytokine secretion in vivo.
The blood samples were agglutinated for 30 min at room temperature. Centrifuged at 1000 g for 10 min.
Serum samples were collected for immediate testing. ELISAs (PT512 (Mouse TNF-α ELISA Kit) and P1580 (Mouse IFN-γ ELISA Kit), both from Beyotime) were performed according to standard protocols.
Immuno uorescence and immunohistochemical analysis of tumor tissues.
Frozen tumor sections were stained with immuno uorescence and immunohistochemical staining. The tumor was rst placed in 4% paraformaldehyde (PFA) for 24 h at 4 ℃, and then transferred to 15% and 30% sucrose solution (w/w) for dehydration. The tumor was implanted in the optimal cutting medium (O.C.T.) and the frozen section was placed in a cryostat microtome. The dye was rinsed with PBS, permeated, and then sealed at room temperature with 5% bovine serum albumin (BSA) for 1 h, followed by staining with different major antibodies: CD4, CD8, CD206, CD80, DAPI overnight at 4 ℃, as per manufacturer's instructions. After adding uorescently labeled secondary antibodies (goat anti-rat IgG and goat anti-rabbit IgG), the slides were analyzed by confocal microscopy.

Statistical analysis
Statistical analysis Statistics were performed using GraphPad Prism 8. Differences between two experimental groups were determined by two-tailed Student's t test and multiple groups by one-way