2.1 TAM infiltration and IL-1β secretion are associated with immunosuppressive TME

To investigate whether tumor-infiltrated Mφ could produce an immunosuppressive microenvironment in ovarian cancer, a representative “cold” tumor, we determined the CD68 expression, a classic marker of Mφ, in clinical tumor specimens of high-grade serous ovarian cancer (HG-SOC) patients with immunohistochemistry (IHC) staining (n = 41). The images revealed that Mφ were massively distributed at both inter-tumor and intra-tumor (Figure 1A) of ovarian cancer, and over 80% of ovarian cancer tissues held abundant Mφ distribution (n = 41; Figure S1A). Due to the high frequency of Mφ and the close spatial location between tumor cells and Mφ, we prospected that the crosstalk between the two might contribute to the immunosuppressive TME. We then used a human inflammation array to explore mediators of the communication by determining the components released when ovarian cancer cell line SKOV3 was co-cultured with Phorbol-12-myristate-13-acetate (PMA)-primed THP1 Mφ, a human leukemia monocytic cell line. IL-1β was identified as one of the most abundant cytokines in the culture medium (Figure 1B). Similar findings were observed when other tumor cell lines, including the lung cancer cell line A549 and the glioma cell line U87MG, were co-cultured with PMA-primed THP1 Mφ (Figure S1B,C). To explore whether the increased IL-1β contributed to immune suppression, a TCGA dataset was analyzed, and we found that IL-1β was negatively correlated with T cell activation (Figure S1D). To further verify this notion, we examined CD68, IL-1β, CD8, and granzyme B, a cytotoxic executor, in clinical tumor tissues of HG-SOC patients with fluorescence multiplex immunohistochemical analysis. The results showed that both the frequency of Mφ and IL-1β levels were negatively associated with granzyme B expression in CD8⁺ T cells (Figure 1C,D). Additionally, to verify whether IL-1β could directly restrain T-cells activation and function, the primary human T cells were activated with anti-CD3/28 antibodies under the administration of different concentrations of IL-1β. The result showed that no significant changes were directly triggered by IL-1β on the expression of CXCR3, granzyme B, and IFN-γ of T cells (Figure S1F–I). As anti-CD3/CD28 antibodies activated T cells cannot specifically target tumor cells in vitro without tumor antigen presentation during their development and maturation, chimeric antigen receptor T (CAR-T) cells were used as a model for their specific antitumor abilities. To evaluate whether IL-1β could affect the tumor-killing capacity of T cells via modulating tumor cells, SKOV3 cells that express neoantigen New York esophageal squamous cell carcinoma-1 (NY-ESO-1) higher, compared to the other two ovarian cancer cell lines A2780 and OVCAR8 (Figure S1E), were pretreated with IL-1β or without and then co-cultured with NY-ESO-1 CAR-T cells at a ratio of 1:2 to mimic the tumor-specific killing of cytotoxic T cells. In comparison with non-treatment SKOV3, IL-1β-primed SKOV3 tumor cells were more tolerant when directly co-cultured with NY-ESO-1 CAR-T cells (Figure 1E,F). The expression of granzyme B (Figure 1G,H) and IFN-γ (Figure 1I,J) in CAR-T cells was suppressed in the presence of IL-1β. Taken together, these data illustrate that IL-1β released when Mφ communicates with tumor cells induces immunosuppression without directly inhibiting functional cytokines expression of T cells.

2.2 IL-1β secreted from Mφ enhances PD-L1 expression in tumor cells

To further investigate the mechanisms underlying IL-1β-mediated immunosuppression, we focused on how IL-1β reshaped the gene expression profile of tumor cells. With RNA-sequencing, IL-1β enhanced a range of immune inhibitory molecules expression on tumor cells, including CD274, PDCD1LG2, and IDO1. Among these molecules, the level of CD274 gene, which encodes PD-L1, was consistently elevated in all tumor cells, including the ovarian cancer cell line SKOV3 (Figure 2A), the lung cancer cell line A549 (Figure S1J), and the glioma cell line U87MG (Figure S1K). To explore whether the elevated PD-L1 could be verified in vitro, primary human peripheral blood mononuclear cell-derived Mφ (PBMC Mφ) were co-cultured with human ovarian cancer cell lines, SKOV3 and OVCAR8, in the presence or absence of IL-1 receptor antagonist (IL-1RA). The results showed that PD-L1 expression in tumor cells was upregulated, based on elevated mRNA and protein levels (Figure 2B–E). However, IL-1RA blocked this upregulation (Figure 2B–E). Similar results were observed in PMA-primed THP1 Mφ co-culturing with SKOV3/OVCAR8/A549/U251/U87MG (Figure S2A–O). To further validate the IL-1β-PD-L1 axis, mouse tumor tissues were employed to stain IL-1β and PD-L1. The results showed a positive correlation between PD-L1 expression and IL-1β secretion in mouse ovarian cancer ID8 tumor tissues (Figure 2F,G) and mouse lewis l LLC tumor tissues (Figure S3A,B). Consistent with this, IHC staining with anti-IL-1β and anti-PD-L1 in 41 surgically resected human ovarian cancer samples demonstrated that IL-1β was abundant in tumor tissues and positively correlated with the level of PD-L1 (Figure 2H,I). In addition, IL-1β treatment triggered dose-dependent increases in PD-L1 expression in SKOV3, A549, U251, and U87MG cells directly (Figure S4A–D). Importantly, a neutralizing antibody targeting PD-L1 dramatically normalized the cytotoxicity of NY-ESO-1 CAR-T cells suppressed in the presence of IL-1β (Figure 2J,K). Thus, these results confirmed that IL-1β induced an immunosuppressive feature via improving PD-L1 expression.

To determine whether IL-1β was secreted by Mφ or tumor cells, we knocked down IL-1β in tumor cells in the co-culture system (Figure S5A–D). With real-time polymerase chain reaction (PCR) and flow cytometry detection, CD274 mRNA and its protein levels before and after IL-1β knockdown were similar in all four cell lines (Figure S5E–P), which revealed that the IL-1β released in the co-culture medium was dominantly derived from Mφ. To validate the notion in tumor tissues, the images of the above fluorescence multiplex immunohistochemical analysis were reused to analyze IL-1β distribution. The results showed that more than 80% of IL-1β was co-localized with CD68-positive Mφ (Figure S6A,B). Consistently, the majority of IL-1β was expressed in F4/80-positive Mφ in ID8 and LLC mouse tumor tissues with immunofluorescence (IF) staining (Figure S6C–E).

Taken together, these results demonstrate that IL-1β is secreted by Mφ and is pivotal in regulating PD-L1 expression by tumor cells in the TME.

2.3 Nuclear factor-κB (NF-κB) is involved in the upstream signaling of PD-L1 expression following IL-1β stimulation

Next, we sought to explore the mechanism underlying the IL-1β-induced increase in PD-L1 expression. NF-κB and mitogen-activated protein kinase/extracellular regulated protein kinase (MAPK/ERK) involved in the regulation of PD-L1 expression have been identified as two major signaling pathways downstream of IL-1β.^24-26 Thus, we asked whether NF-κB or MAPK/ERK signaling pathway contributed to the IL-1β-PD-L1 axis. Three ovarian cancer cell lines SKOV3, OVCAR8, and A2780 were administrated with or without IL-1β, and we found that IL-1β increased the phosphorylation of P65 (NF-κB activation) and ERK, beginning at 30 min after treatment, compared to non-treatment groups (Figure 3A–C). Furthermore, pharmacologic inhibition of NF-κB activation (BAY11-7085) dampened the IL-1β-induced upregulation of PD-L1 transcription in all three tumor cell lines, while ERK inhibition (SCH772984) did not work (Figure 3D–F). Consistent with the transcription, the cell membrane (Figure 3G–I) and total protein (Figure 3J–L) of PD-L1 were reduced following treatment with the NF-κB inhibitor but not with the ERK inhibitor. Finally, A549, U251, and U87MG were used to verify the essential role of NF-κB in the IL-1β-induced PD-L1 elevation (Figure S7A–L). Collectively, we propose that NF-κB functions as a key mediator in the IL-1β-induced PD-L1 expression.

2.4 Tumor cell-secreted lactate is critical for the expression of IL-1β by Mφ

Although IL-1β secreted by Mφ upregulates PD-L1 expression in tumor cells, the molecule that induces IL-1β secretion from Mφ is still unknown. Thus, we sought to determine whether tumor cells can trigger IL-1β secretion from Mφ. SKOV3 and OVCAR8 tumor culture medium (TCM) and fresh medium were mixed in different ratios and used to culture THP1/PBMC Mφ. The results showed that the transcription of IL-1β was increased by TCM in a dose-dependent manner (Figure 4A–D), hinting that there were functional molecules, including proteins, nucleic acids, or other metabolites, that acted as inducers. More interestingly, we showed that boiled TCM (inactivated protein and nucleic acid) still facilitated IL-1β transcription and secretion (Figure 4E–L), suggesting that the molecules responsible for inducing IL-1β expression were secreted metabolites rather than proteins or nucleic acids. Surprisingly, pretreatment with a lactate receptor antagonist (LARA) suppressed IL-1β transcription and secretion (Figure 4E–L). Analysis of the TCM from A549, U251, and U87MG cells corroborated these findings (Figure S8A–I). Other metabolite receptor antagonists, including AhR, mGluR5, and A2AR inhibitors had little effect on IL-1β expression (Figure S9A,B). These results revealed that lactate was a potential candidate associated with boosting IL-1β expression. Of note, we found that IL-1β transcription and secretion were increased in a dose-dependent manner by direct lactate treatment (Figure 4M–P). Further investigation revealed that pro-caspase-1 was cleaved under lactate stimulation, while treatment with an NOD-like receptor thermal protein domain accociated protein 3 (NLRP3) inflammasome inhibitor MCC950 weakened this cleavage and decreased the secretion of IL-1β from THP1 Mφ (Figure 4Q–S). Taken together, we elucidate that tumor cell-derived lactate enhances IL-1β transcription and secretion by Mφ.

2.5 IL-1β-induced C-C motif chemokine ligand 2 (CCL2) recruits Mφ

We figured out a loop composed of tumor cells, lactate, Mφ, IL-1β, and PD-L1 above. What is more, the loop could be deteriorated by the replenishment of Mφ in the TME. Therefore, we further assessed the role of secreted IL-1β in promoting Mφ infiltration in TME. As CCL2 functions as the most critical chemokine in the recruitment of Mφ to tumor sites,^27-29 we analyzed transcriptome data from IL-1β-treated A549/ U87MG cells and found that CCL2 was one of the most highly upregulated chemokine genes (Figure S10A,B). This was confirmed by the observation of a significant positive correlation between IL-1β and CCL2 levels in ID8 tumor tissues (Figure 5A,B) and LLC tumor tissues (Figure S12A,B). To further clarify whether Mφ-derived IL-1β induced CCL2 expression, SKOV3/OVCAR8 and THP1 Mφ were co-cultured in the presence or absence of IL-1RA. The results showed that co-culturing with THP1 Mφ strongly increased CCL2 expression in tumor cells, and this elevation could be restricted under the treatment of IL-1RA (Figure 5C,D). Furthermore, a lactate-primed THP1 Mφ culture medium was used to stimulate SKOV3/OVCAR8 in the presence or absence of IL-1RA. We found that CCL2 secretion declined following the administration of IL-1RA, compared to the treatment with THP1 Mφ supernatant alone (Figure 5E,F). Furthermore, inhibition of NF-κB by BAY11-7085 blocked the IL-1β-induced CCL2 secretion from tumor cells in different types of cancer cells (Figures 5G,H andS11A–C). Last, transwell migration assays showed that IL-1β-primed SKOV3-TCM significantly increased the recruitment of THP1/PBMC Mφ, compared to control TCM and fresh medium (Figure 5I–K), and similar results were obtained using IL-1β-primed U87MG-TCM (Figure S12C–E). Taken together, these results show that IL-1β secreted by Mφ stimulates CCL2 secretion in tumor cells, which in turn boosts Mφ infiltration, thus further increasing PD-L1 expression in tumor cells and orchestrating tumor immunosuppressive microenvironment.

2.6 Blockade of IL-1β inhibits tumor growth and reverses the tumor immunosuppression

The above data clarified mechanisms underlying the IL-1β-centered immunosuppressive loop. Next, we asked whether the IL-1β blockade could reinvigorate the suppressed immune responses and synergistically inhibit the tumor growth with ICB in syngeneic mice models. To this end, immunocompetent C57BL/6 mice were subcutaneously injected with mouse ovarian cancer cell line ID8 under general anesthesia and administrated with anti-IL-1β neutralizing antibodies or isotype immunoglobulin G (IgG) twice a week. The mice were then injected with or without anti-PD-1 antibodies from Day 7. As expected, tumor growth was sloped down following anti-IL-1β antibodies treatment, similar to the administration with anti-PD-1 antibodies (Figure 6A–C). Interestingly, a combination of IL-1β and PD-1 blockade showed a better antitumor effect than monotherapy (Figure 6A–C). Subsequently, to assess how IL-1β blockade reshaped the TIME, we determined the levels of PD-L1, F4/80, and granzyme B in ID8 tumor tissues with IHC staining. The results showed that IL-1β blockade significantly decreased the expression of PD-L1 in tumor tissues (Figure 6D,E), and the frequency of Mφ was much less in anti-IL-1β antibody-treated tumor tissues, compared to non-treatment (Figure 6F,G). Moreover, both the blockade of IL-1β and PD-1 promoted the expression of granzyme B in mice tissues, and the combinational treatment achieved higher levels of granzyme B than monotherapy (Figure 6H,I). In addition, a synergistic antitumor effect was achieved when combining the IL-1β and PD-1 blockade in the LLC tumor-bearing mouse model (Figure S13A–C), and anti-IL-1β antibodies also reversed the suppressive TIME of LLC tumors (Figure S13D–I). Collectively, IL-1β is a promising therapeutic target when used in combination with ICB molecules due to its role in the malignant loop.

2.7 IL-1β is positively associated with the expression of PD-L1 and CCL2, while negatively associated with disease progression in human public databases

We next verified the modulatory effect of IL-1β on PD-L1 expression and the formation of the malignancy loop with bioinformatic analysis. Three Gene Expression Omnibus datasets were analyzed and revealed that IL-1β levels were positively correlated with PD-L1 and CCL2 expression in ovarian cancer, NSCLC, and glioma (Figure S14A,B,D,E,G,H). Higher levels of IL-1β were correlated with poorer prognosis in patients with ovarian cancer after 5 years of follow-up (Figure S14C). IL-1β levels in tumor tissues were negatively correlated with overall survival in patients with other two types of cancers, which was observed with Gene Expression Omnibus datasets (Figure S14F,I). Additionally, TCGA data about liver hepatocellular carcinoma, pancreatic ductal adenocarcinoma, and stomach adenocarcinoma were analyzed to confirm these results (Figure S15A–I). Moreover, a signature of negative regulation of IL-1β production was enriched in the ICB-responded lung cancer patients (Figure S16A) in Gene Expression Omnibus datasets (GSE126044). The relationship between IL-1β and PD-L1/CCL2 was further verified in patient-derived datasets, uncovering that enhancing therapeutic efficacy and sensitivity by combining anti-IL-1β with ICB was necessary for improving the current poor responsiveness of anti-PD-1/L1 treatment in solid tumors.

4.1 Cells

The ID8 cell line was a kind gift from the University of Kansas Medical Center. LLC, THP1, and SKOV3 cell lines were purchased from the National Collection of Authenticated Cell Cultures (Shanghai, China). The A2780, A549, U251, and U87MG cell lines were purchased from the American Type Culture Collection (Rockville). The OVCAR8 cell line was purchased from American Type Culture Collection. All cell lines were cultured following the manufacturer's guidelines.

4.2 Cell culturing

A549 (human lung cancer cell line), U251 (human glioma cell line), U87MG (human glioma cell line), LLC (mouse lung cancer cell line), and ID8 (mouse ovarian cancer cell line) cells were cultured in Dulbecco's modified Eagle's medium (DMEM; Gibco, 11965092). SKOV3 (human ovarian cancer cell line) cells were cultured in McCoy's 5A (modified) medium (Gibco, 12330031). THP1 (human monocyte cell line), A2780 (human ovarian cancer cell line), and OVCAR8 (human ovarian cancer cell line) cells were cultured in Roswell Park Memorial Institute (RPMI) 1640 medium (Gibco, 11875176). All media were supplemented with 10% fetal bovine serum (FBS) (Gibco, 10099141), and 100 μg/mL penicillin-streptomycin (Gibco, 15070063). All cells were maintained at 37°C in a humidified chamber with 5% CO₂.

4.3 Mice

Wild-type C57BL/6 female mice (6 weeks old, healthy) were obtained from Beijing HFK Biotechnology Co. Ltd. The mice were housed under pathogen-free conditions in the animal facilities of the Tongji Hospital of Huazhong University of Science and Technology (HUST). The study was approved by the ethics committee of animal experiments of Tongji Medical College. Eight-week-old female mice were used in all experiments.

4.4 Generation of primary human Mφ

Primary peripheral blood monocytes were isolated from the peripheral blood of healthy donors recruited from the Cancer Biology Research Center of Tongji Hospital using Ficoll density gradient centrifugation. Briefly, peripheral venous blood was added to the upper layer of Ficoll solution in a ratio of 1:1 and centrifuged at 800 g × 20 min, after that, the PBMCs layer in the mixture was washed with phosphate buffer saline (PBS) three times at 500 g × 10 min. RBCs were depleted with ACK Lysis Buffer at 4°C for 5 min. PBMCs were collected and washed with PBS three times before being plated into a culture dish. Monocytes were cultured for 6 days in DMEM (Gibco, 11965092) supplemented with 10 ng/mL macrophage CSF (PeproTech, 300−25) and 10% FBS (Gibco, 10099141) to generate Mφ.

4.5 Generation of human CAR-T cells and tumor cell killing assay

Primary CD3 positive T cells were isolated from healthy donor PBMC via Dynabeads Untouched Human T Cells Kit (Thermo Fisher Scientific, 11344D) according to the instruction, and then activated with anti-CD3 (5 μg/mL) and anti-CD28 (5 μg/mL) antibodies for 2 days. The pre-activated T cells were transfected with a second-generation CAR-T virus targeting NY-ESO-1, which was generated by GENECHEM. One day later, the CAR-T cells were persistently activated with anti-CD3/CD28 antibodies for about 2 days and then expended with 20 ng/mL IL-2 for 5 days. SKOV3 tumor cells were pretreated with or without 10 ng/mL IL-1β for about 48 h and then co-cultured with those expanded CAR-T cells at a ratio of 1:2 for 5 h, followed by determining the apoptosis of tumor cells with annexin V and propidium iodide staining. The cytotoxicity of CAR-T cells was evaluated by granzyme B and IFN-γ expression via flow cytometry. For the assessment of the cytotoxicity of CAR-T cells affected by anti-PD-L1 antibody, the co-cultured media were administrated with 10 μg/mL anti-PD-L1 antibody.

4.6 Cell stimulation

4.7 Flow cytometry

Tumor cells were washed with PBS and stained with phycoerythrin (PE)-conjuncted PD-L1 antibody (29E.2A3, Biolegend) and then determined with flow cytometry (Beckman Coulter). Data were analyzed using Flow Jo software (Tree Star).

4.8 Reverse transcription-PCR (real-time PCR)

All the primers were purchased from Qingke Biologics, Inc.

4.9 ELISA quantitation of IL-1β and CCL2

To investigate which tumor-derived stimulator triggers IL-1β release from Mφ, the fresh TCM or boiled TCM or lactate was used to treat PBMC Mφ or THP1 Mφ for 72 h and then determined the IL-1β level via a human IL-1β ELISA Kit (DAKEWE, 1110122) according to the manufacturer's instructions.

To explore the CCL2 secretion, tumor cells were administrated with lactate-primed THP1 Mφ culture medium or recombinant IL-1β cytokine (10 ng/mL) for about 72 h, and then the CCL2 concentration in tumor supernatant was determined by a Human MCP-1/CCL2 ELISA Kit (Biolegend, 438807) according to the manufacturer's instructions.

4.10 In vitro gene silencing

Silencing of IL-1β in the tumor cell lines was achieved through in vitro small interfering RNA (siRNA) transfection according to the manufacturer's instructions, and the IL-1β siRNA sequence was 3′-CGATGCACCTGTACGATCA-5′. The IL-1β siRNA and negative control siRNA were purchased from Ribo Life Science. Tumor cells were seeded in six-well plates (2 × 10⁵ cells/well) and transfected with 10 nM of control siRNA or IL-1β siRNA. Silencing efficiency was assessed using real-time PCR.

4.11 Immunohistochemistry

Ovarian cancer tissue samples were obtained with informed consent from 41 treatment naïve HG-SOC patients at the Tongji Hospital of HUST, and the baseline information of these patients is applied in Table S1. The specimens were extracted during surgery, and tissues were formalin-fixed and paraffin-embedded. Standard IHC was performed using antibodies against granzyme B (Abcam, ab4059), CD68 (Abcam, ab125212), F4/80 (Servicebio, GB113373), PD-L1 (Abcam, ab238697), and IL-1β (Cell Signaling Technology, 12242S) using an Avidin-Biotin Complex Vectastain Kit (Zsgb-Bio, SP-9000) according to the manufacturer's protocols. Tissue sections were scanned after immunostaining, and protein expressions were measured using Image J software. The diagnosis of high-grade serous ovarian cancer (HG-SOC) was confirmed by two independent pathologists.

4.12 Fluorescence multiplex immunohistochemical analysis

Fluorescence multiplex immunohistochemical analysis was performed according to the manufacturer's instruction. Briefly, paraffin-embedded HG-SOC patient tissues were cut into 4 μm consecutive sections, deparaffinized and antigen retrieved by a microwave oven for 15 min in EDTA buffer (pH 8.0). All the sections were incubated in four rounds of staining; in the order of CD8 (Abcam, ab237709), IL-1β (Cell Signaling Technology, 12242S), granzyme B (Abcam, ab125212), and CD68 (Abcam, ab4059), each using a separate fluorescent tyramide signal amplification system: Quadruple-Fluorescence immunohistochemical mouse/rabbit kit (Immunoway, RS0037). The information of HG-SOC patients (n = 5) involved in the analysis is summarized in Table S1.

4.13 Immunofluorescence

Ovarian and lung cancer mouse tissue samples were obtained from the C57BL/6 ovarian and lung cancer model. The specimens were formalin-fixed and paraffin-embedded. Standard IF was performed with antibodies against IL-1β (Cell Signaling Technology, 12242S), PD-L1 (R&D SYSTEMS, AF1019), and CCL2 (ABclonal, A7277) according to the manufacturer's instructions. After immunostaining, the expression of IL-1β, PD-L1, and CCL2 was measured using Image J software.

4.14 Western blotting

Western blotting was performed according to standard methods. Primary antibodies included the following: GAPDH (Abcam, ab9485), p-P65 (Cell Signaling Technology, 3033S), and PD-L1 (Abcam, ab238697). SuperSignal West Pico Chemiluminescent Substrate kit (Pierce Biotechnology, Inc.) was used to determine the blots on a ChemiDoc XRS+ machine (Bio-Rad Laboratories). BAY11-7085 (Selleck, S7352) and SCH772984 (Selleck, S7101) were used for the pretreatment of cancer cells.

4.15 Tumor models

A volume of 100 μL PBS containing 1 × 10⁷ mouse ovarian cancer cells ID8-VEGF^Hi was mixed with 100 μL Matrigel, and then 5 × 10⁶ cells (100 μL) within the mixture were injected subcutaneously into the right flank of the immune-competent female C57BL/6 mouse. Similarly, 2 × 10⁵ mouse lung cancer cells LLC were transplanted subcutaneously to the right flank of the C57BL/6 mouse. Ten micrograms of neutralizing antibodies against mouse IL-1β (BioXCell, BE0246) were administered intraperitoneally to each mouse twice a week from Day 0 (i.e., the day the tumor cells were administered), and 0.2 mg of neutralizing antibodies against PD-1 (BioXCell, BP0273) were administered intraperitoneally to each mouse twice a week starting 7 days after the injection of tumor cells. Appropriate IgG controls (BioXCell, BP0085) were used. Tumor growth was measured every 3−4 days using an identical caliper by the same surveyor, and tumor volume was calculated as L × W × W × 0.5, where W was the width and L was the length.

4.16 Gene Expression Omnibus (GEO) datasets and The Cancer Genome Atlas (TCGA) cohort

Data on the correlation between IL-1β and PD-L1, IL-1β and CCL2 expression in ovarian cancer (GSE12172), lung cancer (GSE19804), and glioma (GSE53733) were obtained from the Gene Expression Omnibus (1http://www.ncbi.nlm.nih.gov/geo). The correlation analyses between IL-1β and PD-L1, IL-1β, and CCL2 were performed via the R2 platform (2https://hgserver1.amc.nl/cgi-bin/r2/main.cgi?open_page=login) and the GEPIA 2 platform (3gepia2.cancer-pku.cn/#index). Survival analysis of the TCGA cohort was performed via the Kaplan–Meier plotter platform (4http://kmplot.com/analysis/index.php?p=service&cancer=ovar).

4.17 Statistical analysis

GraphPad Prism 5 software was used to graph our data. Statistical significance was determined by unpaired two-tailed Student's t-test, one-way analysis of variance (ANOVA), and linear regression analysis. A p-value of less than 0.05 was regarded as significant.