Melatonin Maintains Macrophage M1 Phenotype to Reverse LPS-stimulated Tumor Immune Tolerance

Background: Lipopolysaccharide (LPS) is a potent trigger of macrophage-mediated inammation and its repeated stimulation results in immune tolerance. This study is to explore the cellular mechanisms of LPS-mediated tumor immune tolerance and to investigate whether melatonin can reverse this tolerance. Methods: The effect of melatonin and LPS on macrophages was assessed by cell proliferation, morphological changes, phagocytosis and autophagy in vitro. The tumor-preventing effect of melatonin and LPS were evaluated in the urethane-induced lung carcinoma model and in the H22 liver cancer allograft model. Immunouorescence, immunohistochemistry and ELISA were used to examine protein expression. The related targets and pathways of melatonin were predicted by comprehensive bioinformatics, and the clinical association of bacterial infections and survival was evaluated in cancer patients by meta-analysis. Results: In vitro (cid:0) Raw264.7 macrophages were polarized toward the M1 phenotype by single LPS administration but toward the M2 phenotype by repeated LPS administration. Interestingly, combination treatment with repeated LPS and 10 µM melatonin prevented macrophage polarization toward the M2-like phenotype and exerted lasting antitumor ecacy. In the urethane-induced lung carcinoma model, repeated LPS administration stimulated macrophage polarization toward the M2 phenotype and promoted lung carcinogenesis, which was abrogated by macrophage depletion, while melatonin alone or in combination with repeated LPS challenge restored M1-like macrophages and prevented carcinogenesis. In the H22 liver cancer allograft model, melatonin maintained the macrophage phenotype and promoted the tumor-suppressing effect of repeated LPS challenge. Furthermore, we found that macrophages repeatedly stimulated with LPS had a high level of surface lipid rafts that mediated PI3K/AKT and JAK2/STAT3 signaling and prevented both LPS sensitivity and immune response by self-expression of PD-L1 and surface expression of PD-1 receptor on NK cells, whereas melatonin decreased surface lipid rafts and PI3K/AKT and JAK2/STAT3 signaling. Finally, we conducted a comprehensive bioinformatics analysis of melatonin-relevant targets and pathways involved in M2 macrophage polarization and evaluated the clinical associations of bacterial infections and survival in cancer patients. Conclusions: This study suggests a function of melatonin in regulating macrophage polarization to maintain LPS-stimulated tumor immune surveillance.

cells. [8] The 10-year datasets from rst-line anti-CTLA4 therapy show an unprecedented long-term survival in 20% of terminal metastatic melanoma patients that had never been seen before with other approaches to cancer treatment, indicating an exciting success in cancer immunotherapy that can provide hope of becoming "super-survivors" to incurable cancer patients [9.10] . Nevertheless, response rates with the most promising immunotherapy, such as PD-1 inhibitors, which represent immune-checkpoint-blockade(ICB)mediated rejuvenation of exhausted T cells, only exhibit a modest 20% overall survival bene ts in many solid tumors, whereas CAR-T cell therapy was reported to result in cerebral edema and cytokine release storms [7.11] . In addition, a more recent clinical report showed that PD-1 inhibitors could speed up tumor growth and promote tumor hyperprogression in 9% of cancer patients, indicating a need to optimize cancer immunotherapeutic approaches. [12] Current immunotherapies mainly activate tumor-in ltrating T lymphocytes and natural killer cells without regard to environmental changes that can cause various states of T cell dysfunction, such as anergy, tolerance, exhaustion, and senescence. [12.13] In fact, antitumor T cell immunotherapies originally exhibited responses but subsequently became resistant during prolonged antigen exposure due to "immune exhaustion" induced by the immunosuppressive microenvironment that is shaped by tumor-associated in ammatory cells. [14] Obviously, effective tumor immune rejection relies not only on antigen exposure-induced adaptive immune responses but also on the innate immune surveillance-regulated microenvironment. [15] Tumor-in ltrated macrophages (TIMs) are a major in ammatory cell in ltrating the tumor microenvironment and are responsible for the immunosuppressive microenvironment and tumor progression. [16] Recent studies have shown that the inhibition of macrophage-mediated phagocytosis is an essential mechanism for tumor immune evasion. [17] In the clinic, TIMs are positively associated with high tumor grade and poor prognosis in various cancers. In mouse cancer models, TIM depletion or reeducation can reverse their tumor-promoting functions. [18.19] In addition, some recent reports have shown that innate macrophages also play important roles in the intratumor in ltration of CD8 cytotoxic T cells and the establishment of the long-lived memory lymphocytes. Therefore, targeting TIMs to reawaken innate immunity has emerged as a new cancer immunotherapy strategy. [20] . It is well known that activated macrophages are divided into antitumor M1 and protumor M2 phenotypes. [21] LPS can polarize macrophage toward the M1 phenotype, but repeated stimulation results in immune tolerance. [22] Melatonin is a neurohormone secreted by the pineal gland, This study explored the cellular mechanisms of LPS-mediated tumor immune tolerance and investigated whether melatonin can reverse this tolerance. [23] Our results are the rst to indicated a vital role of macrophage polarization in LPS tolerance and also suggests a new mechanism by which bacterial infections increase the risk of carcinogenesis.
41000100002406. Liposome-encapsulated clodronate (LEC) was prepared as described previously. [24] Cell Culture and Assay Raw264.7 macrophages and the Lewis lung carcinoma (LLC) cells were from ATCC, purchased from the Chinese Academy of Sciences and grown in RPMI1640 medium supplemented with 10% (v/v) fetal bovine serum (FBS) in a humidi ed atmosphere containing 5% CO 2 and 95% air at 37℃. Raw264.7 macrophages were seeded in 24-well plates and stimulated with by 10 ng/ml LPS or 10 ng/ml IL-10 for 24 h to obtain M1-like (M1) and M2-like (M2) macrophages. To collect cell-conditioned media, M1-like or M2-like cells were cultured in serum-free medium for 24 h, centrifugated to remove the cells and further ltered to remove debris for supernatant collection as M1 and M2 cell-conditioned media(M1-CM and M2-CM,respectively). The supernatant levels of IFN-γ, TNF-α, NO, PD-L1, IL-10 and TGF-β1 were determined by ELISA kits, according to the manufacturer's protocols. [25] The results were calculated from linear curves obtained by using the Quantikine kit standards.
For proliferation analysis, LLC cells at 1 × 10 5 cells/mL were seeded in a 96-well plate and treated with M1 or M2 cell-conditioned media for 48 h, M1-like or M2-like cells were also treated with LPS or melatonin alone or in combination for 7 d (changing the medium every 2 days), and living cells were examined by MTT reduction assay, according to our previous method. [24] For the morphological assessment, the cells were analyzed by a Laser holographic cell imaging and analysis system (HoloMonitor M4, Phiab, Sweden). [25] For phagocytic ability assessment, neutral red phagocytosis was detected. For autophagy analysis, the cells were stained using PE-conjugated anti-LC3-B or anti-p62 antibodies. For apoptosis analysis, the binding of ANXV-FITC to phosphatidylserine was measured by an automated cell counter LLC cell immune clearance LLC cell immune clearance was assessed using a calcein-release assay, according to our previous method. [24] Brie y, NK cells (DX5+) were puri ed from the ICR mouse spleens using the MACS separation system (Miltenyi Biotec, Bergisch Gladbach, Germany), stimulated with IL-2 (10 ng/ml) for 24 h in the presence of M1-like or M2-like cell-conditioned media and harvested as attacking cells. Mitomycin Ctreated LLC cells were labeled with 10 µM calcein-AM as target cells and were placed into a 96-well plate with CD8 + T cells at 100:1, 50:1, and 25:1 (NK cells:LLC cells) ratios for 6 h at 37 °C. The supernatants were transferred from each well to another 96-well plate, and the uorescence was measured using a Synergy2 multimode microplate reader (BIO-TEK). Maximum release was obtained from detergentreleased LLC cells, and spontaneous release was obtained from LLC cells incubated in the absence of CD8 + T cells (n = 5).
To detect how the different macrophage phenotypes affect NK cells, NK cells were cultured in the lower chamber at a concentration of 5 × 10 6 cells/ml and were stimulated with IL-2, and M1-like or M2-like cells were added in the upper chamber at 2 × 10 6 cells/ml in the presence or absence of anti-IL-10, anti-TGF-β1, MβCD, WP1066 or LY294002. After coincubation for 24 h at 37 °C, the supernatant was centrifugated for the PD-L1 assay, adherent cells in the upper compartment were removed by a cotton swab, and the lter inserts were incubated in medium supplemented with 5 mg/ml DAPI for 30 min at 37 °C and analyzed for cell migration using an inverted uorescence microscope. NK cells in the lower chamber were collected for surface PD-1 receptor assay using PE-conjugated anti-PD-1 antibody.

Cytolytic assay
The cytolytic activity of LLC cells was assessed by CFSE-7AAD staining. Brie y, LLC cells were incubated with CFSE -labeled M0, M1-like or M2-like cells at 20:1 and 10:1 (M0, M1 or M2 cells: LLC cells) ratios for 6 h. Then, 7AAD was added to the cell suspensions and incubated on ice for 15 min. The percentages of 7AAD + cells among CFSE + cells were analyzed using an automated cell counter and analysis system.

Urethane-induced lung carcinogenesis model
Urethane (600 mg/kg body weight), alone or in combination with liposome-encapsulated clodronate (LEC, 4 mg/mouse) was injected intraperitoneally (i.p.) into ICR mice once a week for eight weeks, according to our previous protocol. [26] Following the rst urethane injection, mice received melatonin (20 mg/kg/day) via intragastric administration once a day or LPS (1 mg/kg/day) via intravenous tail injection once a week alone or in combination for twelve weeks. At thirteen weeks after the rst urethane injection, orbital venous blood was collected for serum assays of IFN-γ, IL-2, TNF-α,PD-L1, IL-10 and TGF-β1 using an ELISA kit. The mice were sacri ced, and cell-free alveolar uid was collected by inserting a cannula into the trachea with three sequential injections of 1 mL PBS, followed by centrifugation, for cytokine assay (IFN-γ, IL-2, TNF-α, ROS, IL-10 and TGF-β1), while the separated cells were resuspended in 0.9% sterile saline for total cell counts. Macrophages in the suspensions were enriched by magnetic cell sorting utilizing anti-F4/80-coated beads, and macrophage immunophenotypes were analyzed by FITCconjugated anti-mouse CD86 and CD163 staining using an automated cell counter and analysis system (Nexcelom Cellometer X2, Nexcelom, USA).
Spleen NK cells (DX5+) were separated using the autoMACS separation system for assays of surface PD-1 receptor and memory NK cell rate (NKG2c + NKG2a-) using an automated cell counter and analysis system.
The average numbers of lung carcinomas per mouse were calculated. A portion of each lung was preserved in 10% buffered formalin and routinely embedded in para n. Lung sections were stained by immunohistochemistry and immuno uorescence according to our previous method. [27] After overnight incubation with the primary antibodies (anti-PD1, anti-iNOS, and anti-CD31), the slides were incubated with the FITC-conjugated goat anti-mouse IgG for 30 minutes. The total immunohistochemical and immuno uorescence scores were calculated by the intensity score and proportion score by excluding the primary antibody and IgG matched serum, respectively, as positive and negative controls.
In addition, the lung vascular integrity was assayed by the Evans blue dye extrabarrier technique according to our previous method. [28] Tumor allograft model H22 cells were used for tumor allograft experiments. Two hundred microliters of saline containing 1 × 10 6 cells were injected subcutaneously into the lateral axilla of mice to establish tumor allografts. One day after tumor inoculation, in vitro LPS-induced M1 or IL-10-induced M2 cells (2 × 10 6 cells in 200 µL saline) were injected intravenously into mice once a week for three weeks; simultaneously, mice received melatonin (20 mg/kg) via intragastric administration once a day and LPS (1 mg/kg) or LEC (4 mg/mouse) via intravenous tail injection once a week alone or in combination for 3 weeks. Tumor size was monitored twice a week with calipers and calculated as the length x width 2 /2. On the twenty-second day after tumor inoculation, orbital venous blood was collected for serum assays of IFN-γ, IL-2, TNF-α,PD-L1, IL-10 and TGF-β1. The mice were euthanized, the tumors were extracted and weighed, peritoneal macrophages were enriched by magnetic cell sorting utilizing anti-F4/80-coated beads, and macrophage immunophenotypes were analyzed. Spleen NK cells were separated using the autoMACS separation system for assays of surface PD-1 receptor and memory NK cell rate. The complete assay procedure was similar to the methods in the urethane-induced lung carcinogenesis model.
In addition, the tumor vascular integrity was assayed with the Evans blue dye extrabarrier technique according to our previous method. [25] For the immune rechallenge study, the tails of 1/2 of the 40 mice were injected subcutaneously with 5 × 10 5 H22 cells suspended in 50 µl saline. One day after tumor inoculation, the mice received melatonin (20 mg/kg) via intragastric administration once a day and LPS (1 mg/kg) or LEC (4 mg/mouse) via tail intravenous injection once a week alone or in combination for 2 weeks. Fifteen days after tumor implantation, the tumor-bearing tail was cut off to remove the primary tumor, and the mice were rechallenged with subcutaneous injections of 1 × 10 6 H22 cells in 200 µl saline in the anks, while 10 normal mice were challenged with identical H22 cells. Tumor size was monitored twice a week with calipers. At thirty-six days, the same detections as above were carried out.
The regulatory mechanism of melatonin on neutrophils The gene expression pro les GSE5099 were obtained from the Gene Expression Omnibus (GEO) database, Up-and downregulated genes related to tumor-associated macrophages were identi ed using GEO2R, and the human structures of these differential proteins were collected from the protein data bank (PDB) for docking analysis. The chemical structure of melatonin was obtained from PubChem, and the docking exercise was conducted using the online software systemsDock with the autoremoval of unspeci ed protein structures. Docking scores over 5 were regarded as the potential targets for melatonin. Gene ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses were performed for the potential targets using the Database for Annotation, Visualization and Integrated Discovery (DAVID) and the online software Omicshare. The protein-protein interactions (PPIs) among these potential targets were constructed using the STRING database, and the hub genes were identi ed using Cytoscape.

Network Meta-analysis
We systematically searched PubMed and Web of Science to identify eligible studies published from Jan 1, 1990 to Apr 1, 2019. OR and hazard ratios (hrs) with 95% con dence intervals (cis) were used to evaluate the risk of bacterial infections in carcinogenesis and cancer survival. Fixed and random-effect meta-analyses were conducted based on the heterogeneity of the included studies. To minimize case selection bias, sensitivity analyses were performed.

Statistical analyses
The data were statistically analyzed using GraphPad Prism, Version 5.0 (San Diego, CA, USA) and are presented as the mean ± SD. The differences between two groups were evaluated using a t-test. A P value of less than 0.05 was considered statistically signi cant. Meta-analyses were performed using the xedeffects inverse-variance method using RevMan 5.3. Heterogeneity was calculated using the I 2 statistics and chi-square Q test, and a P-value of heterogeneity < 0.10 or I 2 >50% indicated signi cant heterogeneity. Usually, a xed-effect model was used, and the random-effects model was adopted when the heterogeneity was signi cant.

M2 macrophages had an opposing effect on LLC cells compared to that of M1 macrophages
It was reported that cancer-associated macrophages have M2-like characteristics and exert tumorpromoting actions, unlike M1-like macrophages, which show antitumor functions. [29] To explore these properties, we induced the polarization of Raw264.7 macrophages into M1-like cells and M2-like cells. As expectedly, compared to the characteristics of M1-like cells, M2-like cells had high levels of surface CD163 expression (Fig. 1A, 1B) with morphological changes indicated by the cell distribution (Fig. 1D, 1E ). In neutral red phagocytosis, there was no difference between M1 and M2 cells (Fig. 1C); however, M2like cells had a reduction in ROS indicated by intracellular uorescence of DCFH-DA (Fig. 1F) and an increase in autophagy indicated by LC3-B and p62 immuno uorescence (Fig. 1H). When cocultured with M1 cell-conditioned media (M1-CM), LLC cells showed reduced proliferation (Fig. 2B) and increased apoptotic rates (Fig. 2C), which were not affected by M2 cell-conditioned media (M2-CM). Consistent with these results, M1-like cells led to LLC cell lysis (Fig. 2D), which was not in uenced by M2-like cells, and M1 cell-conditioned media promoted LLC cell immune clearance (Fig. 2D) which was prevented by M2-CM.
M2 macrophages had the more lipid rafts and suppressed NK cells It is well known that high TIMs and low NK cells may be associated with an poor prognostic survival in cancer patents [30] . To explore how macrophages affect NK cells, we detected differences between M1-like and M2-like macrophages and their in uence on NK cells. M2 cells had the more lipid rafts, as indicated by membrane cholesterol (Fig. 2I) ang produced more PD-L1 (Fig. 2G ) and JAK2/STAT3 activation (Fig. 2G) compared to those of M1 cells. M1-CM promoted immune clearance of LLC cells by NK cells (Fig. 2E), which was prevented by M2-CM. Consistent with these results, M2 cells promoted NK cell expression of PD-1 surface receptor (Fig. 2F ), and MβCD and anti-TGF-β1 antibody but not anti-IL-10 antibody prevented M2 cells from inducing PD-L1 production and NK cells from expressing PD-1, whereas JAK2/STAT3 signal blockade by WP1066 but not PI3K/AKT signal blockade decreased M2 cell lipid rafts and PD-L1 production (Fig. 2G,2H).

Melatonin reverses the macrophage tolerance of LPS
Macrophages may be tolerant to LPS, an effect that lasts for a period of time. [31] To explore how macrophages become tolerance to LPS, we observed the effects of melatonin and LPS on macrophage polarization. As expected, repeated LPS treatment led to macrophage tolerance of LPS (Fig. 1A) accompanied by a shift from the M1 to M2 phenotype, the increased lipid rafts and the JAK2/STAT3 activation (Fig. 2G, 2H), while no more than three stimulations induced macrophage polarization toward the M1 phenotype (Fig. 1A). Unexpectedly, when administered at the dose (10 µM), which had a small effect on cell viability in M1 cells ( Fig. 2A), melatonin improved the morphological changes indicated by cell distribution (Fig. 1D, 1E) and induced apoptosis in M2 cells ( Fig. 2A) with the increased ROS (Fig. 1F) and the decreased autophagy (Fig. 1H), accompanied by suppression of lipid rafts and JAK2/STAT3 signaling. Importantly, the combination treatment with LPS and melatonin resulted in the polarization of macrophages toward M1 cells (Fig. 1A). Furthermore, we found that both of lipid raft depletion by MβCD and JAK2/STAT3 signal blockade by WP1066 attenuated the effect of melatonin on M2 cells (Fig. 2F,  2H).
Melatonin maintains macrophage M1 phenotype to reverse LPS-mediated carcinogenesis in a urethane-induced lung cancer model To con rm the roles of macrophages in carcinogenesis, we investigated macrophage phenotypes in a urethane-induced mouse lung cancer model, which is usually used for studying basic lung tumor biology and nding new tumor intervention strategies. In this study, the mice received melatonin or LPS for twelve weeks. At thirteen weeks, lung cancer nodes were visible to the naked eye (Fig. 3A). The number of lung cancer nodes was 26.2 ± 4.1, regardless of the heterogeneity of tumor histology in the control group (Fig. 3B). As expected, macrophage in ltration in alveolar cavities was positively correlated with lung carcinogenesis in the control group (Fig. 3C, 3D, 3E, 3F), and both of melatonin treatment and macrophage depletion induced by LEC prevented lung carcinogenesis (Fig. 3A, 3B). Unexpectedly, LPS alone did not prevent lung carcinogenesis but promoted these incidents (Fig. 3A, 3B). Immunophenotypes showed that the in ltrated macrophages were similar to IL-10-treated Raw264.7 macrophages and expressed more surface CD163 in the control group (Fig. 3E, 3F ), indicating an M2 phenotype. ELISA showed that the levels of IFN-γ, IL-2, and TNF-α decreased, while the levels of IL-10 and TGF-β1 increased in serum and in alveolar cavities in control mice compared to those of normal mice (Fig. 3G, 3H), indicating immune tolerance. These changes were promoted by repeated LPS administration and were attenuated by melatonin. Importantly, the combination of LPS and melatonin maintained the M1 phenotype (Fig. 3D, 3F) and reverses the lung carcinogenesis-promoting effect of LPS (Fig. 3A, 3B), accompanied by a reduction in serum PD-L1 levels (Fig. 3G ), spleen NK cell surface PD-1 expression (Fig. 4C), lung tissue immunohistochemical staining of PD-1, iNOS, immuno uorescence staining of CD31 (Fig. 4A, 4B), and permeability to evans blue dye (Fig. 4E), as well as an increase in spleen memory NK cells (Fig. 4D) and NK cell surface PD-1 expression rates (Fig. 4C), indicating a better immune restoration and lung vascular integrity.
Melatonin maintains the macrophage M1 phenotype to promote LPS-induced tumor suppression and to reverse repeated LPS-induced immunosuppression in the H22 liver cancer allograft model  (Fig. 5A, 5B ). The tumor-suppressing e cacy of LPS was promoted by melatonin, macrophage depletion and M1 cell injection, which decreased the intratumor permeability of evans blue dye (Fig. 5C) but was attenuated by M2 cell injection (Fig. 5A, 5B), which increased the intratumor permeability of evans blue dye. Immunophenotyping showed that the in ltrated macrophages were similar to IL-10-treated Raw264.7 macrophages and expressed more surface CD163 in the control group, indicating an M2 phenotype (Fig. 5D). The results of serum cytokines (Fig. 5G) and spleen NK cell analyses (Fig. 5E, 5F) were similar to those in the urethane-induced lung cancer model, indicating an M2 macrophage-polarizing effect of repeated LPS. The immunosuppressive e cacy of repeated LPSinduced macrophages was further con rmed in an H22 cell rechallenge immune study where tumor immune rejection was suppressed by repeated LPS stimulation, while it was promoted by melatonin. Importantly, the combination of LPS and melatonin synergistically stmulated immune rejection of H22 cells (Fig. 6A), accompanied by a reduction in peritoneal M2 macrophages (Fig. 6B), serum levels of PD-L1, IL-10 and TGF-β1 (Fig. 6E), and spleen NK cell surface PD-1 expression (Fig. 6C) and an increase in peritoneal M1 macrophages (Fig. 6B), serum levels of IFN-γ, IL-2, TNF-α (Fig. 6E), and spleen memory NK cell rate (Fig. 6D), indicating a sustained effect of melatnin on the M1 macrophage phenotype. The immune-rejecting e cacy of melatonin on tumor rechallenge could be aborgated by macrophage depletion, indicating an important role of macrophages in secondary immunity and immune memory.
Melatonin regulates macrophages by targeting the multiprotein network We queried 590 upregulated genes (LogFC ≥ 1.5, P < 0.05) and 994 downregulated genes (LogFC ≤-1.5, P < 0.05) related to M1-associated macrophages and obtained 181 targets. A total of 136 potential targets with a docking score > 6.0 (pKd/pKi) were selected for GO and KEGG analyses (Fig. 8, Fig. 7A). A hypergeometric distribution count > 4 and P < 0.05 were set as threshold criteria to identify the functional gene ontology and pathways. GO enrichment analysis indicated that the potential targets of melatonin were primarily associated with the "signal transduction", "innate immune response", "cell proliferation", "protein phosphorylation" and "apoptotic process" terms ( Fig. 8). KEGG enrichment analysis revealed that the potential targets of melatonin were signi cantly enriched in the "Pathways in cancer", "TNF signaling pathway" and "JAK/STAT signaling pathway" terms (Fig. 7C). The PPI network identi ed 4 key genes (JAK2, STAT3, PIK3CA, and AKT1) that were hub genes for melatonin (Fig. 7B). These results were con rmed by Western blot analysis (Fig. 2F, 2G, 2H).
Bacterial infection increases the risk of carcinogenesis and is adversely associated with survival times in cancer patents.

Discussion
Cancer may arise from alterations in different physiological processes and is refractory to cure due to unknown etiology and genetic heterogeneity. Based on self-healing ability, immunotherapy has been great expectation against distinct types of cancer. [32][33][34][35][36][37] Current immunotherapies targeting different cellular checkpoint controllers emerged as having either innate or acquired resistance due to the immunosuppressive tumor environment, guiding the direction of developing cancer immunotherapy. [38] The object of cancer immunotherapy is to stimulate a long-lasting immunosurveillance, maintaining the antitumor immunity. [39] Macrophages, as the rst-line immune responders in the innate immune system, are responsible for nonresolving in ammation in tumors. Understanding the roles of macrophages in the tumor immune response may help develop new immunotherapeutic strategies and enhance the response rate of immunotherapy. [40] Macrophages can exhibit both pro-and antitumorigenic properties, depending on their phenotype. Recent studies have demonstrated that targeting TIMs can reverse the immunosuppressive tumor microenvironment and stimulate robust tumor-speci c immune responses, which is consistent with the fact that immunosuppressive TIMs are abundant in the tumor microenvironment and are positively correlated with poor prognosis. Therefore, maintaining the antitumorigenic phenotype of macrophages rather than completely depleting TIMs represents a new cancer immunotherapy strategy. [41.42] LPS is a potent trigger of macrophage-mediated in ammation and has been recognized as a potent antitumor agent in animal tumor models. [42] However, its use in human cancer therapy was not very successful due to LPS-induced tolerance, a state of altered responsiveness in macrophages, which results in poor tumor response and is a major cause of secondary hospital infections. Previously, Boris et al reported that β-glucan could reversed the epigenetic state of LPSinduced immunological tolerance to reduce overall sepsis mortality. [22.43] In this study, we rst showed that the combination of LPS and melatonin could prevented macrophage polarization toward M2-like cells and therefore exerted a lasting antitumor e cacy, suggesting a novel effective strategy for reversing LPS tolerance.
Macrophages can be activated in response to different agents to become M1 and M2 macrophages and thus exert different functions. It is well known that LPS is able to polarize macrophages toward the M1 phenotype, which exerts the proin ammation and antitumor effects. [44] Consistent with these functions, in an animal model, LPS demonstrated a therapeutic effect on the transplanted tumor with inhibition of tumor size and growth. In small clinical trials, LPS also led to cancer remission and disease stabilization in cancer patients. [45.46] However, LPS is responsible for the biological properties of bacterial endotoxins, which are potent in ammagens and result in fever, septic shock, toxic pneumonitis, and respiratory symptoms; however, a cohort study found that long-term exposure to endotoxin was associated with a reduced risk of lung cancer. [46.47] In goats, LPS treatment triggered an excessive in ammatory response and elevated the body temperature to 40 °C. [48] In clinical trials, even pretreated with ibuprofen resulted in the unavoidable LPS-mediated clinical toxicities. [49] In addition, subsequent LPS-induced macrophage resistance after activation was also an obstacle that prevented durable tumor response to macrophages.
Despite intense investigations of various epigenetic and genetic changes in tolerant macrophages for many years, a unifying mechanism that is responsible for LPS tolerance remains elusive. [50][51][52] Jo¨rg et al used genome-wide transcriptional pro ling technology to show the hyporesponsiveness of most LPS target genes in tolerant macrophages. [53] In the present study, Raw264.7 macrophages were polarized toward M1-like cells by a single LPS challenge but toward M2-like cells by repeated LPS challenge, suggesting a critical process of LPS tolerance in macrophages. The meta-analysis showed that bacterial infection increases the risk of carcinogenesis and is adversely associated with cancer survival, suggesting a wide role of LPS tolerance in cancer progression. A mechanistic understanding of the role of LPS in tumor progression will provide unique therapeutic alternatives.
Melatonin is a pleiotropic molecule and has numerous physiological and pharmacological actions. [54.55] Melatonin generally plays a check-and-balance role in immunity and in ammation. Melatonin is an immunostimulator that drives an activated immunocyte state in favor of effectively clearing pathogens under normal or immunosuppressive conditions, while it acts as an immunosuppressor that urges immunocytes to enter an inactivated state that suppresses in ammatory reactions under excessively in ammatory conditions. [56.57] Several studies found that melatonin had an exciting potential to override the immunosuppressive tumor environment. [58] Various reports have shown that melatonin promotes 80% survival in lethal LPS-treated mice and signi cantly reduces the LPS-treated mortality in mice and rats by correcting the LPS-induced in ammatory imbalance with decreased levels of NO and lipid peroxidation. [59.60] In addition, a recent study demonstrated that melatonin cold suppresses indoleamine 2,3-dioxygenase-1 (IDO1) (a key immunomodulatory enzyme associated with cancer immune escape) to overcome tumor-mediated immunosuppression. [61.62] Based on these results, we believe that the combination of LPS and melatonin can maintain macrophage sensitivity to LPS and simultaneously limit excessive pathogenic stimuli to avoid a "macrophage exhaustion" phenotype. Consistent with our hypothesis, in this study, the combination of LPS and melatonin resulted in optimal cancer prevention in a urethane-induced lung carcinogenic model and H22 liver cancer allograft model without signi cant side effects. Furthermore, we found that macrophages repeatedly challenged with LPS had a high level of surface lipid rafts and JAK2/STAT3 activation, which prevented both M1-like polarization and immune responses, whereas melatonin decreased surface lipid rafts and JAK2/STAT3 signaling. Bioinformatics analysis found that the potential targets of melatonin regulation of macrophages were primarily associated with the "in ammatory response", "signal transduction", "cell proliferation" "innate immune response" and "negative regulation of apoptotic process", indicating a function of melatonin targeting multi-protein networks, such as "Jak-STAT signaling pathway", "Toll-like receptor signaling pathway" and "Chemokine signaling pathway", whereby melatonin maintains M1 macrophage phenotypes to reverse LPS-stimulated immune tolerance. We used a pharmacological blocker to con rm the important role of Jak-STAT signaling in the melatonin-maintained macrophage phenotype, and JAK2/STAT3 signal blockade prevented M2-like macrophages from producing PD-L1 and NK cells from expressing PD-1, whereas lipid raft depletion decreased M2-like macrophage JAK2/STAT3 signaling, indicating an association of M2 macrophage functions and the feedback loop of lipid rafts -JAK2/STAT3-PD-L1.
Certainly, this is only one of the important feedback loops.
Historically, bacterial therapy as oncolytic agents has been recognized for malignant brain tumours, which showed an extended survival times for patients who developed infections at the site of resection of malignant gliomas, and cancer vaccines were assumed to be based on immunotoxins of bacterial origin. [63] It was previously reported that occupational exposure to endotoxin in organic material reduced the risk of lung cancer among workers employed ≤ 35 years but increased the risk of lung cancer among those employed > 50 years, implying a cancer-promoting action of endotoxin tolerance. [64.65] In fact, LPS (also referred to as endotoxin) as a cell wall component of gram-negative bacteria has been used for tumor destruction for many years, such as a report by Chicoine et al that intratumor injection of LPS could cause the regression of subcutaneously implanted mouse glioblastoma multiforme in animal tumor models. [66] Goto S et al also reported that three of ve evaluable human cancers showed a signi cant response to intradermal LPS administration, suggesting that LPS is a potent antitumor agent. [67] However, only one trial of LPS use in human cancer therapy declared a poor tumor response and unbearable side effects. In addition, LPS inhalation can also produce both a systemic and a bronchial in ammatory response. [68.69] Therefore, it is necessary for LPS use as an antitumor agent to balance its in ammatory response and tolerance. Our study provides a safe and effective strategy for the clinical application of LPS, and this therapeutic strategy is worth investigating.

Conclusions
In summary, melatonin regulation of immune balance plays an important role in the LPS-stimulated macrophage response, and our ndings suggest several potential clinical implications. First, M2 macrophage polarization correlates with LPS-stimulated "macrophage exhaustion", supporting therapeutic targeting of TIMs in an immunosuppressive environment. Second, alteration of the macrophage phenotype is superior to depletion of TIMs for reversing LPS tolerance and tumor immunosuppression, indicating the potential of reeducating these cells to reverse their protumor functions for antitumor properties. Third, macrophages, as the rst-line responders of innate immune surveillance, are needed not only for the initiation of tumor innate immune responses but also for the long-lasting preventive functions of NK cells against tumors, indicating the rst priority for the development of e cient immunotherapies. Certainly, whether reeducating TIMs by the combination of melatonin and LPS is the most effective approach for restoring antitumor immune responses needs to be fully evaluated, and the dose-effect and exposure-timing relations also need also to be further investigated.  The data present Mean ± SD, the experiments were repeated 3 times, and statistical signi cance was determined by a t-test. *P < 0.05, **P < 0.01 . Mel:Melatonin.   Serum cytokine levels (n=6). The data present Mean ± SD, the experiments were repeated 3 times, and statistical signi cance was determined by a t-test. *P < 0.05, **P < 0.01 Mel:Melatonin.