Blood levels of acute-phase protein α1-acid glycoprotein (AGP, orosmucoid) increase in patients with cancer. Although AGP is produced from hepatocytes following stimulation by immune cell–derived cytokines under conditions of inflammation and tumorigenesis, the functions of AGP in tumorigenesis and tumor progression remain unknown. In the present study, we revealed that AGP contributes directly to tumor development by induction of programmed death ligand 1 (PD-L1) expression and IL6 production in macrophages. Stimulation of AGP induced PD-L1 expression in both human monocyte–derived macrophages through STAT1 activation, whereas AGP had no direct effect on PD-L1 expression in tumor cells. AGP also induced IL6 production from macrophages, which stimulated proliferation in tumor cells by IL6R-mediated activation of STAT3. Furthermore, administration of AGP to AGP KO mice phenocopied effects of tumor-associated macrophages (TAM) on tumor progression. AGP decreased IFNγ secretion from T cells and enhanced STAT3 activation in subcutaneous tumor tissues. In addition, AGP regulated PD-L1 expression and IL6 production in macrophages by binding with CD14, a coreceptor for Toll-like receptor 4 (TLR4), and inducing TLR4 signaling. These results provide the first evidence that AGP is directly involved in tumorigenesis by interacting with TAMs and that AGP might be a target molecule for anticancer therapy.
AGP-mediated suppression of antitumor immunity contributes to tumor progression by inducing PD-L1 expression and IL6 production in TAMs.
Acute-phase proteins are produced from hepatocytes by immune-cell derived cytokines such as IL6 and TNFα. Increased blood concentrations of these proteins are observed in patients with inflammatory diseases and malignant tumors (1). The concentrations of acute-phase proteins such as C-reactive protein (CRP) and α1-acid glycoprotein (AGP, also known as orosomucoid) are also increased in patients with malignant tumors and high concentrations of acute-phase proteins are significantly associated with worse clinical courses in several malignant tumors (2–4). AGP, a glycoprotein with a molecular weight of approximately 44,000 Da, is mainly biosynthesized in the liver and functions as an endogenous carrier for basic drugs in vivo (5–7). The blood concentrations of AGP increase two to five fold (1–2.5 mg/mL versus. 0.5 mg/mL under normal conditions) in the presence of inflammation and various tumors including lung cancer, hepatocellular carcinoma, and melanoma (8–10). Recently, AGP was reported to function as an immune modulator inhibiting excessive immune responses by regulating macrophage activation and suppressing neutrophils under inflammatory conditions (11–13). Moreover, an association between lymphocyte activity and high AGP concentration was observed in immune-suppressed cancer patients (14). Thus, the finding of several studies have suggested the significance of AGP in tumor progression; however, the physiologic function of AGP in tumors remains poorly understood.
The tumor microenvironment includes not only tumor cells but also host-derived nontumor cells, including macrophages and fibroblasts, and close interactions between them provide a favorable environment for tumor development (15–17). Among them, macrophages are highly infiltrated in various tumor tissues and are termed tumor-associated macrophages (TAM). Various factors in the tumor microenvironment affect macrophage polarization into the protumor (M2-like) phenotype (18). Studies using human pathologic specimens reported that a high density of CD163-positive TAMs was associated with a worse clinical course in various solid tumors (19). TAMs also contribute to tumor proliferation and metastasis by secretion of various cytokines, and immunosuppression via the expression of immunosuppressive molecules such as programmed death ligand-1 (PD-L1), IL10, indoleamine 2,3-dioxygenase 1 (IDO1), and Siglec-15 (20–22). IHC studies using human tumor samples showed PD-L1 expression in tumor cells in less than 30% of cases, whereas PD-L1–positive TAMs were observed more frequently in patients with several tumors, including lung cancer (23–26). In vivo studies using mouse models showed that not only PD-L1 but also PD-L2 on TAMs were involved in immune suppression in the tumor microenvironment (27, 28).
AGP is a ligand of CD14 and CCR5, which are expressed on macrophages. We previously reported that AGP induces CD163 expression via CD14/toll-like receptor-4 (TLR4) signaling in macrophages (11). Thus, we hypothesized that AGP might induce a protumor microenvironment by affecting TAM activation. Therefore, in the present study, we examined the effects of AGP on tumor development using human cells (macrophages and tumor cells) and tumor-bearing mouse models using AGP-deficient mice.
Materials and Methods
Preparation of AGP
AGP was purified from the supernatant of human plasma fraction V provided by KM Biologics Co., Ltd. as described previously (29). The supernatant was diluted with acetate buffer and injected into an ion-exchange column (GE Healthcare) at a flow rate of 3 mL/minute on an AKTAprime Plus System (GE Healthcare), with AGP eluted in the first peak. The secondary structure of AGP was confirmed by circular dichroism with a Jasco J-720 spectropolarimeter (JASCO).
Cell lines and cell culture
Peripheral blood mononuclear cells (PBMC) were isolated by gradient separation (LymphoprepTM, Cosmo Bio) from buffy coats of healthy donors. Written informed consent was obtained from all donors of PBMCs and the experimental procedures were approved by the Kumamoto University Review Board (#1169). Monocytes were cultured in polystyrene dishes (Becton Dickinson) in DMEM (Wako) supplemented with 2% FBS, 1% penicillin/streptomycin (WAKO), 50 ng/mL macrophage colony-stimulating factor (M-CSF; WAKO), and 20 ng/mL GMCSF; WAKO) for 5 days to differentiate into human monocyte-derived macrophages.
Tumor-cell lines including ES2, a human ovarian cancer; SBC3, a human small-cell lung cancer; HepG2, a human liver cancer; SK-MEL-28, human melanoma; NCI-H358, a human non-small cell lung cancer; 786-O, a human renal cancer; MC38, a mouse colon cancer; and LM8, a mouse osteosarcoma cancer, were purchased from RIKEN Cell Bank or JCRB Cell Bank. A mouse Lewis lung cancer (LLC), was kindly gifted by Keizo Takenaga (Chiba Cancer Center). The cells were maintained in RPMI 1640 containing 10% FBS and 1% penicillin/streptomycin and were regularly tested using a Mycoplasma test kit (TAKARA). These cells were cultured for less than 3 months before reinitiating the cultures and routinely inspecting microscopically for a stable phenotype.
Primary murine peritoneal macrophages were obtained from peritoneal exudates of mice. Peritoneal macrophages were cultured in low-glucose DMEM supplemented with 2% FBS and 1% penicillin/streptomycin.
RNA sequencing transcriptome analysis
For RNA sequencing (RNA-seq) of macrophages and macrophages with AGP samples, total RNAs were prepared by TRIzol (Thermo Fisher) and quality of total RNA was confirmed by BioAnalyzer 2100 (RIN >9). Library DNAs were prepared according to the Illumina TruSeq protocol using TruSeqStandard mRNA LT Sample Prep Kit (Illumina) and sequenced by Illumina NextSeq 500 (Illumina) using NextSeq 500/550 High Outputv2 Kit (Illumina) to obtain single-end 75 nt reads. RNA-seq data analysis resulting reads were aligned to the human genome (UCSC hg19) using STAR ver.2.6.0a after trimmed to remove adapter sequence and low-quality ends using Trim Galore! v0.5.0 (cutadapt v1.16). Gene expression level measured as TPM was determined with RSEM v1.3.1. GO analysis and pathway enrichment analysis were performed using multiple databases PANTHER (http://www.pantherdb.org/), DAVID (https://david.ncifcrf.gov/), and Metascape (http://metascape.org/gp/index.html#/main/step1).
Human monocyte-derived macrophages were treated with human FcR-blocking reagent (Miltenyi Biotec, 120–000–442) for 10 minutes and then reacted with phycoerythrin (PE)-labeled antihuman PD-L1 (BioLegend catalog no. # 329707, RRID:AB_940358) antibody, allophycocyanin (APC)-labeled PD-L2 (BioLegend, 329608), and each isotype-matched control antibody (BioLegend) diluted in Cell Staining Buffer (BioLegend, 420201) for 20 minutes. The stained cell samples were analyzed on a FACSverse (Becton Dickinson) flow cytometer with FACSuite (Becton Dickinson) software. Mouse peritoneal macrophages were treated with mouse FcR-blocking reagent (Miltenyi Biotec, 130–092–575) for 10 minutes and then reacted with APC-labeled antimouse PD-L1 (BioLegend catalog no. # 124312, RRID:AB_10612741) antibody or isotype-matched control antibody (BioLegend, 400612) diluted in Cell Staining Buffer for 20 minutes. The stained cell samples were analyzed on a FACSverse flow cytometer with FACSuite software.
Western blot analysis
Cells were lysed in NP-40 lysis buffer [50 mmol/L Tris (pH 8.0), 150 mmol/L NaCl, 1 mmol/L EDTA, 1% NP-40] with both protease and phosphatase inhibitors. The cell lysates were run on a 10% SDS-polyacrylamide gel. The proteins in the SDS-polyacrylamide gel were transferred to polyvinylidene difluoride (PVDF) membranes (Millipore) and subjected to western blotting with the primary antibodies and secondary-detection antibodies. The probed proteins were detected using EzWsetLumi plus solution (ATTO), and the signals were detected with a LAS-4,000 image analyzer (Fujifilm). The primary and secondary antibodies used in the Western blot analyses were described in Supplementary Table S1.
Cytokine array analysis was performed using a human cytokine array kit, panel A (R&D Systems), according to the manufacturer's protocol.
High-quality total RNA was isolated from cultured cells or tumor tissues fractured by a MagNA Lyser System (Roche Diagnostics GmbH) with TRIzol reagent (Thermo Fisher Scientific) and NucleoSpin RNA clean-up system (MACHEREY-NAGEL GmbH & Co). Genomic DNA was digested with amplification grade DNase I (Invitrogen, #18068–015). cDNA was synthesized using a PrimeScript RT Master Mix (Takara Bio Inc., RR036). qRT-PCR was performed with TB Green Premix Ex Taq II (Takara Bio Inc., #RR820) on a ViiA7 system (Thermo Fisher Scientific). Each melting curve was analyzed to confirm that the PCR signal was derived from a single PCR product. The amplification conditions were an initial denaturation at 95°C for 20 seconds, followed by 40 cycles of denaturation at 95°C for 1 second and annealing/extension at 60°C for 20 seconds. A minimum of 3 separate samples was used and the expression levels were calculated from at least 2 technical replicates. mRNA expression levels were estimated using the 2ΔΔCt method and the mRNA levels were normalized to those of β-actin mRNA. The primer sequences are described in Supplementary Table S1.
All cytokines (IL6, TNFα, IL10, IL1β, CCL2, and IFNγ) were detected by specific sandwich ELISA sets (BD Biosciences) according to the manufacturer's protocol.
Cell viability assays
Tumor cells were plated in 96-well plates, cultured for 24 hours, and incubated with AGP-treated macrophage-conditioned medium (AGP-MCM). After incubation for 48 hours, 10 μL of WST-8 solution was added to each well and the cells were incubated for 2 hours. The number of surviving cells was determined by measuring the absorbance at 450 nmol/L. Cell viability was calculated as the percentage of the control value.
T-cell isolation and coculture with macrophages
A pan–T-cell isolation was performed according to the manufacturer's recommendations (Miltenyi Biotec, human: 130–096–535; mouse: 130–095–130) and a coculture experiment was performed according to the Minimal Information About T cell Assays (MIATA) guidelines (30). AGP-treated macrophages were cocultured with autologous T cells in the presence of anti-CD3 (eBioscience, 16–0037–85, RRID:AB_468855) and CD28 (eBioscience, 302923, RRID:AB_2291210) antibody at a 1:1 ratio for 48 hours. IFNγ production from T cells was measured by ELISA.
AGP-deficient (AGP−/−) mice in the C57BL/6N background were obtained from the Center for Animal Resources and Development (CARD) in Kumamoto University, while wild-type (WT) mice in the C57BL/6N background and C3H/HeN or C3H/HeJ mice were purchased from CLEA Japan. The mice were housed in a temperature-controlled room with a 12-hour light/dark cycle. All animal experiments were approved by the Ethics Committee or Animal Experiments of Kumamoto University (A2019–015) and conducted in accordance with the Institutional Animal Care and Use Committee.
Murine tumor-bearing model
WT or AGP−/− mice (6–8 weeks of age) were subcutaneously injected with MC38 (5 × 105 cells/mouse) or LLC (3 × 105 cells/mouse) cells suspended in 100 μL of RPMI 1640 medium. Three days later, AGP (2 mg/mL) was intraperitoneally administered to AGP−/− mice every day. The mice were sacrificed on day 14 or 21, followed by the determination of subcutaneous tumor development and a part of the tumor tissue was used for immunohistologic analysis and qRT-PCR.
C3H/HeN or C3H/HeJ mice were subcutaneously injected with LM8 cells (5 × 105 cells/mouse) suspended in 100 μL of RPMI 1640 medium. Three days later, AGP (2 mg/mL) was intraperitoneally administered to C3H mice every day. The mice were sacrificed on day 14, followed by the determination of subcutaneous tumor development.
50 μL clodronate liposome (Hygieia Bioscience) was treated to tumor-bearing C3H/HeN mice by intravenous injection every 4 days.
Tumor tissue samples were fixed in 10% neutral buffered formalin and embedded in paraffin wax. After the sections had been deparaffinized in xylene and rehydrated in a graded ethanol series, they were immersed in EDTA solution (pH8.0) and heated in a pressure cooker for antigen retrieval in the staining of Iba-1, PD-L1, and CD8. For the p-STAT3 staining, the sections were immersed in citrate buffer (pH 6.0) and irradiated in a microwave oven for 5 minutes and then allowed to cool down at room temperature for 20 minutes. Next, the sections were immersed in EDTA solution (pH 8.0) with and irradiated in a microwave oven for 5 minutes. The primary antibodies used were anti-Iba1 (WAKO, 019–19741), anti-PD-L1 (R&D Systems, AF1019), anti-CD8 (Cell Signaling Technology, 98941), and anti-p-STAT3 (Cell Signaling Technology, 9145). Following the reaction of the primary antibodies, the samples were incubated with horseradish peroxidase (HRP)–labeled secondary antibodies (Nichirei). The immunoreactions were visualized using the DAB substrate system (Nichirei). Double immunostaining was performed to identify the colocalization of Iba1 and PD-L1 or p-STAT3. First, the sections were stained with anti–PD-L1 or p-STAT3 and visualized with DAB as described above. After the sections were rinsed with PBS, they were heated in a microwave for antigen retrieval. Anti-Iba1 antibody was used as the second antibody and visualized with HistoGreen (Linaris). The number of p-STAT3–positive tumor cells were calculated by subtracting the number of Iba-1 and p-STAT3 double-positive macrophages from all p-STAT3–positive cells.
Data were analyzed using GraphPad Prism 7 software (RRID:SCR_002798). All data from animal and cell culture studies represent at least 2 or 3 independent experiments. Data are expressed as means ± SD. Differences between groups were examined using the Mann–Whitney U test and the nonrepeated measures ANOVA to determine statistical significance. P values <0.05 were considered to indicate statistically significant differences.
AGP induced PD-L1/L2 expressions in macrophages but not in tumor cells
As IFNγ induces PD-L1 and PD-L2 expression in both macrophages and tumor cells (31, 32), we first tested the effect of AGP on the PD-L1/L2 expression in human monocyte-derived macrophages (HMDM). IFNγ significantly induced both PD-L1 and PD-L2 expression, as reported previously (Fig. 1A and B; ref. 31). AGP also significantly induced PD-L1 expression but did not affect PD-L2 expression (Fig. 1A and B). Furthermore, the PD-L1 expression induced by AGP stimulation was enhanced in a dose-dependent manner and the expression was induced at 6 hours after AGP stimulation (Supplementary Fig. S1A and S1B). In contrast, when we applied AGP to several human tumor-cell lines derived from cancers previously reported have elevated blood concentrations of AGP (8–10), no marked change in the PD-L1 expression was observed in any line (Fig. 1C). These results indicated that AGP specifically enhanced PD-L1 expression in macrophages.
AGP-induced PD-L1 overexpression on macrophages was mediated by STAT1 activation
IFNγ-induced PD-L1 expression is regulated by the STAT1 signaling cascade in both macrophages and tumor cells (33, 34). Thus, we measured the effect of AGP on STAT activation in HMDMs, observing that AGP significantly induced STAT1 activation, similar to the IFNγ stimulation observed in HMDMs (Fig. 2A). The activation induced by AGP stimulation was enhanced in a dose-dependent manner starting at 6 hours after AGP stimulation (Supplementary Fig. S1C and S1D). In addition, the AGP-induced PD-L1 expression was inhibited by both fludarabine and nifuroxazide, famous STAT1 inhibitors (Fig. 2B; Supplementary Fig. S1E), suggesting that AGP induces PD-L1 expression in macrophages through STAT1 activation, similar to IFNγ stimulation. Interestingly, AGP also induced STAT3 activation in HMDMs, whereas IFNγ did not affect STAT3 activation (Fig. 2A). The activation induced by AGP stimulation was enhanced in a dose-dependent manner starting at 3 hours after AGP stimulation (Supplementary Fig. S1C and S1D). Furthermore, WP1066, a STAT3 inhibitor, did not affect AGP-induced PD-L1 expression (Fig. 2C), indicating that the STAT3 activation induced by AGP was not related to the mechanism of PD-L1 expression.
In contrast, AGP did not induce STAT1/3 activation in several tumor-cell lines, whereas IFNγ significantly induced STAT1 activation in all tumor cells (Fig. 2D). These results indicated that AGP-induced PD-L1 expression was regulated by STAT1 signaling in macrophages and that AGP was not correlated with PD-L1 expression in tumor cells.
AGP induced the production of protumor cytokines from macrophages
Since STAT3 activation is induced by stimulation by cytokines such as IL6 and IL10, we tested the hypothesis that AGP-induced STAT3 activation was mediated by indirect mechanisms related to IL6 or IL10 by examining the effect of AGP on cytokine expression in HMDMs using a cytokine array. As shown in Fig. 3A, AGP stimulation increased IL6, growth-related oncogene α (GROα), and CXCL-8 production. Similar results were observed by both ELISA and qRT-PCR (Fig. 3B and C). RNA-seq transcriptome analysis showed upregulation of cytokines and chemokines related to interferon and lipopolysaccharide (LPS)-associated pathways as well as IL6, GROα, and CXCL8 (Supplementary Fig. S2A–S2D), suggesting that these three cytokines may be candidate molecules that activate STAT3 in macrophages. Therefore, we next measured the effects of recombinant IL6, GROα, and CXCL8 on STAT3 activation in HMDMs. As shown in Fig. 3D, IL6 significantly enhanced STAT3 activation, as reported previously, whereas neither GROα nor CXCL8 showed an effect on STAT3 activation. In contrast, IL6 did not affect either STAT1 activation or PD-L1 expression in HMDMs (Fig. 3D). Furthermore, IL6R-neutralizing antibodies inhibited STAT3 activation induced by AGP stimulation in IL6R expressing HMDMs (Fig. 3E; Supplementary Fig. S3A) but had no inhibitory effect on PD-L1 expression (Fig. 3E), indicating that IL6 produced by macrophages following AGP stimulation bound to IL6R on the macrophages, followed by the induction of STAT3 activation in macrophages (Fig. 3F).
AGP-stimulated macrophages induced tumor cell growth by activating the IL6/STAT3 pathway
IL6 also functions as a STAT3 activator in tumor cells (35). STAT3 activation is associated with cell growth, stem cell properties, and chemoresistance in tumor cells (36). Cell–cell interaction with TAMs induces STAT3 activation in neighboring tumor cells (37). Therefore, we speculated that AGP enhances STAT3 activation induced by cell–cell interactions between TAMs and tumor cells and measured the indirect effect of AGP on STAT3 activation in tumor cells. All of the human tumor-cell lines used in the present study showed IL6R expression or STAT3 activation during incubation of IL6 with sIL6R (Supplementary Fig. S3A and S3B). As shown in Fig. 4A and Supplementary Fig. S4, the STAT3 activation was enhanced at 3 hours after the stimulation of AGP-MCM in tumor cells compared with both control and unstimulated MCM, whereas AGP had no direct effect on STAT3 activation. The AGP-MCM–induced STAT3 activation was suppressed by IL6R neutralizing antibody (anti-IL6R ab; Fig. 4B), whereas neither GROα nor CXCL8 affected STAT3 activation in tumor cells (Supplementary Fig. S5A), indicating that the IL6 produced by macrophages following AGP stimulation enhanced STAT3 activation in tumor cells. In addition, AGP-MCM significantly induced tumor proliferation compared with that for MCM (Fig. 4C), whereas AGP had no direct effect on tumor proliferation (Supplementary Fig. S5B). Anti-IL6R ab also inhibited tumor proliferation induced by AGP-MCM (Fig. 4D), indicating that the IL6 produced by macrophages following AGP stimulation also bound to IL6R on the tumor cells, followed by the induction of STAT3 activation and cell proliferation in tumor cells (Fig. 4E).
AGP enhanced tumor development in mouse LLC and MC38 tumor model
We next examined the significance of AGP in tumor development in AGP-deficient mice. First, we confirmed the effect of AGP on PD-L1 expression in both mouse peritoneal macrophages and mouse tumor-cell lines. As shown in Supplementary Fig. S6A and S6B, AGP enhanced PD-L1 expression in mouse macrophages, similar to human macrophages, whereas AGP did not affect PD-L1 expression in mouse tumor-cell lines, similar to human tumor-cell lines. Mouse tumor-cell lines (LLC and MC38) were injected subcutaneously into WT and AGP−/− mice and tumor development was compared. The results showed significantly reduced subcutaneous tumor weight in AGP−/− mice compared with that in WT mice (Fig. 5A). Furthermore, the administration of AGP to AGP−/− mice significantly increased tumor weight in both tumor-bearing models (Fig. 5B), thus indicating that AGP was involved in tumor development.
AGP treatment induced an immunosuppressive microenvironment
As shown in Fig. 5C, AGP administration increased the numbers of PD-L1–positive cells in subcutaneous tumor tissues of AGP−/− mice in both tumor-bearing models, whereas the number of Iba-1 and CD8-positive cells was not changed by AGP administration. Furthermore, administration of AGP in the MC38-bearing model also significantly increased the numbers of PD-L1 and Iba-1 double-positive cells in subcutaneous tumor tissues of AGP−/− mice (Fig. 5D). In contrast, the numbers of PD-L1 and Iba-1 double-positive cells tended to increase after AGP administration in the LLC-bearing model (Fig. 5D), whereas the number of pSTAT3-positive cells was significantly increased by AGP administration in the LLC-bearing model (Fig. 5E). Furthermore, the both numbers of Iba-1 and pSTAT3 double-positive macrophages and pSTAT3-possitive tumor cells were significantly increased by the administration of AGP (Fig. 5F), suggesting that the activation of STAT3 in the tumor cells and macrophages was induced by AGP. Since PD-L1 abrogates lymphocyte activation, we next evaluated both IFNγ and Granzyme B (GzmB) expressions in MC38 tumor tissues as a lymphocyte activation marker. AGP treatment decreased the IFNγ mRNA expression and the number of IFNγ-positive CD8+ cells and GzmB-positive CD8+ cells in subcutaneous tumor tissues (Fig. 5G–I). As the MC38 cell is an immunogenic tumor (hot tumor) model and LLC is nonimmunogenic tumor (cold tumor) model, the number of CD8-positive cells infiltrating subcutaneous tumor tissues in the MC38-bearing model was high compared with that in the LLC-bearing model, suggesting that AGP was involved in tumor development by inducing PD-L1 expression on tumor-infiltrated macrophages in the MC38-bearing model and that AGP was involved in tumor development by other mechanisms in the LLC-bearing model.
AGP enhanced macrophage-related immune suppression and tumor growth
We next conducted in vitro studies to test whether or not AGP was involved in immune suppression via macrophage activation. We found that AGP treatment reduced the IFNγ production from activated T cells under coculture conditions with both HMDMs (Fig. 6A) and mouse peritoneal macrophages (Fig. 6B), suggesting that AGP suppressed T-cell activation by inducing the PD-L1 expression on infiltrated macrophages in tumor tissues. Next, we determined PD-L1 expression in both mouse macrophages and MC38 cells stimulated by serum from AGP−/− mice treated with or without AGP. As shown in Fig. 6C and Supplementary Fig. S6C, the serum of AGP-treated mice showed an enhanced PD-L1 expression in macrophages compared with the serum of control mice, whereas no marked difference in the PD-L1 expression was found between the 2 serum samples of MC38 cells and LLC cells (Fig. 6D; Supplementary Fig. S6D), thereby demonstrating the immunosuppressive effect of AGP on T cells by enhancing the PD-L1 expression on infiltrated macrophages in tumor tissues.
AGP also induced IL6 production in mouse macrophages, similar to human macrophages (Fig. 6E). We examined the effect of AGP-treated MCM on STAT3 activation and tumor proliferation in LLC cells. As shown in Fig. 6F and G, AGP-MCM enhanced both the STAT3 activation and cell proliferation in tumor cells compared with both control and MCM, whereas AGP had no direct effect on the STAT3 activation and cell proliferation (Fig. 6F; Supplementary Fig. S6E). In addition, the IL6 production and STAT3 activation were confirmed to be enhanced in MCM prepared by stimulation with serum from AGP-treated tumor (LLC and MC38)-bearing mice (AGP treatment) compared with MCM prepared by stimulation with serum from AGP-untreated tumor (LLC and MC38)-bearing mice (Control; Fig. 6H–J; Supplementary Fig. S6F–S6H). These findings indicate the involvement of AGP in tumor progression in the tumor (LLC and MC38)-bearing mouse models via the induction of IL6 production by macrophages, followed by enhanced STAT3 activation in both tumor cells.
The AGP–CD14 complex induced PD-L1 and IL6 expression via TLR4 signal
TLR4 and CCR5 are known as AGP receptors (6, 13). Among those receptors, TLR4 recognizes the AGP-CD14 complex whereas CCR5 directly recognizes AGP molecules (13, 38). We next examined the effect of neutralizing antibodies and inhibitors of these receptors on AGP-induced expression of PD-L1 and IL6 to identify the receptor for AGP in macrophages. As shown in Fig. 7A and B, the CD14-neutralizing antibody significantly suppressed AGP-induced PD-L1 expression, whereas maraviroc (a CCR5 antagonist) did not affect AGP-induced PD-L1 expression. In addition, the CD14-neutralizing antibody suppressed both STAT1 and STAT3 activation induced by AGP treatment (Fig. 7A) and inhibited AGP-induced IL6 production (Fig. 7C). CD14/TLR4 signaling is also correlated with MAPK activation in macrophages (39). As shown in Fig. 7D, AGP induced MAPK activation, such as p38, Erk1/2, and JNK, an activation that was inhibited by coincubation with CD14-neutralizing antibody (Fig. 7E). These data indicate that TLR4 signaling induced by AGP-CD14 complex binding to TLR4 was correlated with PD-L1 expression and IL6 production in macrophages.
Since the CD14-neutralizing antibody suppressed AGP-induced IL6 production in HMDMs (Fig. 7C), we hypothesized that the culture supernatant of AGP-treated macrophages during incubation with CD14-neutralizing antibody would reduce both STAT3 activation and tumor proliferation in tumor cells by inhibiting IL6 production by macrophages (Fig. 7F). As shown in Fig. 7G, AGP-MCM enhanced STAT3 activation in tumor cells. Under these conditions, the CD14-neutralizing antibody suppressed STAT3 activation in tumor cells. Furthermore, the CD14-neutralizing antibody also suppressed tumor proliferation induced by treatment with AGP-MCM (Fig. 7H), indicating that the IL6 produced from macrophages by AGP-CD14/TLR4 signaling was correlated with both STAT3 activation and tumor proliferation in tumor cells.
AGP-induced tumor development was abrogated in mice with TLR4 mutations
C3H/HeJ mice have a loss-of-function mutation in TLR4 (40); thus, we measured the difference in the effect of LPS and AGP on TLR4 signaling between C3H/HeJ and C3H/HeN mouse-derived macrophages. As shown in Fig. 8A, the LPS-induced TLR4 signaling (p38, Erk, and JNK activation) was downregulated in C3H/HeJ mouse-derived macrophages compared with that in C3H/HeN mouse-derived macrophages. Similar results were observed for AGP stimulation (Fig. 8A), indicating that AGP was directly correlated with TLR4 signaling activation. Both PD-L1 expression and IL6 production induced by AGP stimulation were also downregulated in C3H/HeJ mouse-derived macrophages compared with those in C3H/HeN mouse-derived macrophages (Fig. 8B and C), indicating that the activated TLR4 signaling induced by AGP stimulation was correlated with PD-L1 expression and IL6 production in macrophages. We next compared the effect of AGP on tumor progression in an LM8 mouse sarcoma-bearing mice model between C3H/HeJ and C3H/HeN mice. As shown in Fig. 8D, neither STAT3 activation and PD-L1 expression were changed by direct AGP stimulation in LM8 cells. However, AGP-MCM from C3H/HeN mouse peritoneal macrophages enhanced STAT3 activation in LM8 cells compared with AGP-MCM from C3H/HeJ mouse peritoneal macrophages (Fig. 8E). In addition, AGP-MCM from C3H/HeN mouse peritoneal macrophages induced tumor proliferation in LM8 cells, whereas AGP-MCM from C3H/HeJ mouse peritoneal macrophages had no effect on tumor proliferation (Fig. 8F), suggesting that AGP-CD14/TLR4 signaling was correlated with tumor proliferation in mouse tumor cells. Therefore, LM8 cells were injected subcutaneously into C3H/HeN and C3H/HeJ mice and the tumor development was compared. Treatment of C3H/HeN mice with AGP significantly increased the tumor weight, whereas treatment of C3H/HeJ mice with AGP had no effect on the tumor weight (Fig. 8G), indicating that AGP-CD14/TLR4 signaling was correlated with tumor development.
AGP is well known as an acute-phase protein, and its blood concentrations are significantly increased in patients with cancer. Therefore, AGP has been postulated to play a physiologic role in tumor progression. To date, there is limited evidence regarding the indirect effect of AGP on tumor progression, in which AGP influences the therapeutic efficacy of anticancer agents because AGP is a major binding protein for these drugs; thus, based on its concentration, AGP modulates drug pharmacokinetics and pharmacodynamics (41). Paclitaxel, an anticancer drug, binds strongly to AGP and shows reduced antitumor activity depending on the blood concentration of AGP (42). In the present study, we demonstrated the first evidence that AGP is directly correlated with tumor immunosuppression and tumor proliferation by acting on macrophages in the tumor microenvironment.
Recent reports have shown that not only genetic and epigenetic changes in tumor cells (intrinsic factors) but also the tumor microenvironment constructed by various cells surrounding cancer cells (extrinsic factors) are involved in tumor onset and progression. This is evidenced by the fact that the therapeutic strategies for targeting infiltrated cells other than tumor cells in tumor tissues are effective against many cancers in the clinical setting (43–45). Among those therapies, immune checkpoint molecule inhibitors such as PD-1/PD-L1 antibody show high effectiveness for several cancers (46, 47); thus, immunotherapy is attracting attention as a ‘fourth' treatment following surgery, radiation therapy, and chemotherapy.
Since many TAMs infiltrate tumor tissues and closely interact with tumor cells and fibroblasts in the tumor microenvironment, TAMs have recently been recognized as important key players in tumor progression. The results of the present study revealed that AGP, an acute-phase protein with immunosuppressive functions, was correlated with immune tolerance through the induction of PD-L1 expression in TAMs, followed by the suppression of IFNγ secretion from T cells, whereas AGP had no direct effect on tumor cells. Inhibition of STAT1 signaling, a major signaling pathway for PD-L1 expression in both tumor cells and macrophages, suppressed AGP-induced PD-L1 expression in macrophages (Fig. 2B), suggesting that AGP was directly involved in tumor immunosuppression by acting on TAMs in the tumor microenvironment.
In the present study, AGP also induced IL6 production by macrophages (Fig. 3A–C) and produced IL6-enhanced STAT3 activation in tumor cells, followed by the induction of tumor proliferation, whereas AGP had no direct effect on tumor proliferation by STAT3 activation (Fig. 3A). Furthermore, STAT3 activation in tumor cells and tumor proliferation were suppressed in the presence anti-IL6R ab (Figs. 3E, 4B, and D), suggesting that IL6 produced by macrophages following AGP stimulation was involved in tumor proliferation. In addition, AGP induced STAT3 activation in macrophages and this phenomenon was suppressed in the presence anti-IL6R ab. However, anti-IL6R ab did not affect AGP-induced PD-L1 expression in this time. Therefore, the combination therapy of anti–PD-1/PD-L1 antibody and anti-IL6R antibody may lead to synergistic therapeutic effects in cancer.
CD163, an M2 phenotype marker in macrophages, is reportedly induced by AGP stimulation and its CD163 expression was suppressed by cotreatment with anti-IL6R ab (11), suggesting that AGP indirectly contributed to tumor development through IL6-induced M2 polarization in the tumor microenvironment. IL6 is also known to promote metastasis through the transformation of normal fibroblasts into cancer-associated fibroblasts (CAF) or induction of the epithelial–mesenchymal (EMT) transition and worsen patient prognosis (48–50), indicating that IL6 produced by TAMs induced by AGP might create a microenvironment suitable for tumor cells via the induction of CAF or mesenchymal cells, leading to poor patient prognosis.
As mentioned above, although AGP induced IL6 production and STAT1/PD-L1 pathway in macrophages, the AGP receptor related to these functions remains unknown. However, the present study also revealed the receptor for AGP related to these functions. CD14, a marker of monocyte and macrophage and coreceptor for TLR4, and CCR5, a chemokine receptor, are known as a receptor for AGP (6, 13). In the present study, both STAT1 activation and PD-L1 expression in macrophages induced by AGP were remarkably suppressed by treatment with CD14-neutralizing antibody (Fig. 7A). Treatment with CD14-neutralizing antibody also remarkably suppressed both AGP-induced STAT3 activation and IL6 production in macrophages (Fig. 7A and C), strongly suggesting that CD14/TLR4 is a receptor for AGP related to those functions in macrophages. Similar results were observed in C3H/HeJ mouse-derived macrophages with a loss-of-function mutation in TLR4 (Fig. 8B and C). This is the first evidence that the CD14/TLR4 signaling pathway is directly correlated with PD-L1 expression and IL6 production in macrophages. Although the organ-protective effect of AGP has been reported in various inflammation models (11, 51, 52), suggesting that AGP improves inflammation by direct effects on parenchymal cells, our results suggest that AGP may contribute to the improvement of inflammation by interacting with the CD14/TLR4 pathway in macrophages. In fact, AGP administration had no notable effect on tumor progression in clodronate-treated C3H/HeN mice (Supplementary Fig. S7), indicating that AGP is involved in tumor development by interacting with macrophages.
Furthermore, AGP was also correlated with tumor progression in a tumor-bearing mouse model (Fig. 5A and B). In the MC38-bearing (immunogenic tumor) model, AGP increased the number of PD-L1–positive TAMs, whereas the number of CD8-positive T cells did not change in subcutaneous tumor tissues (Fig. 5C and D). AGP also decreased IFNγ expression in subcutaneous tumor tissues (Fig. 5F), indicating the protumor function of AGP with tumor immunosuppression.
In conclusion, the results of our study revealed the first evidence that AGP is directly involved in tumorigenesis by interacting with TAMs, which are known to be major constituents of the tumor microenvironment.
No disclosures were reported.
K. Matsusaka: Data curation, formal analysis, investigation, writing–original draft. Y. Fujiwara: Conceptualization, data curation, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. C. Pan: Data curation, formal analysis, validation, investigation, visualization. S. Esumi: Data curation, formal analysis, investigation, visualization, methodology. Y. Saito: Data curation, formal analysis, investigation, methodology. J. Bi: Resources. Y. Nakamura: Resources. A. Mukunoki: Resources. T. Takeo: Resources. N. Nakagata: Resources. D. Yoshii: Data curation, formal analysis, investigation, methodology. R. Fukuda: Data curation. T. Nagasaki: Data curation. R. Tanaka: Data curation. H. Komori: Resources. H. Maeda: Resources, funding acquisition. H. Watanabe: Conceptualization, resources, funding acquisition, project administration. K. Tamada: Resources. Y. Komohara: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, investigation, methodology, writing–original draft, project administration, writing–review and editing. T. Maruyama: Conceptualization, supervision, funding acquisition, writing–original draft, project administration, writing–review and editing.
The authors thank Ms. Takana Motoyoshi (K. I. Stainer, Inc.) and Mr. Takenobu Nakagawa (Kumamoto University) for their technical assistance. This work was supported by JSPS KAKENHI (grant numbers: 16H05162, 16K09247, 15K15185, 16K15320, and 20H03459).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.