Introduction

Hepatocellular carcinoma (HCC) is the sixth most common malignant tumor in the world and the third leading cause of carcinogenic death. Almost 910,000 people are diagnosed with HCC worldwide every year, and most of the patients have already developed into advanced stages at the time of diagnosis.1 In the USA, the 5-year survival rate of patients with HCC is less than 15%, and in patients without and with distant metastasis is 28% and 3%, respectively.2

Resistance to anoikis is a prerequisite for the metastasis of epithelial cancer cells. As a typical characteristic of the first appearing intrahepatic metastasis, HCC acquires anoikis resistance improving survival in the bloodstream, initiating secondary tumor formation, resulting in greater resistance to anticancer agents.3 Induction of anoikis in tumor cells has been beneficial for the treatment of metastatic diseases.4

Immunotherapy has benefited a number of patients with cancer, and has been approved for the treatment of advanced HCC. Several immunotherapy clinical trials related to programmed death protein-1 (PD-1)/programmed death ligand-1 (PD-L1) for HCC are already underway. However, ~75% of patients with HCC are unresponsive to immune checkpoint inhibitors (ICIs).5 6 Therefore, it is particularly important to accurately identify the ICI-responsive HCC population for immunotherapy to improve the therapeutic efficacy or to discover ways to transform the immunologically “cold” tumors into “hot” tumors and expand the HCC population suitable for immunotherapy.

Coagulation factor X (FX) is synthesized primarily in the liver, and activated to FXa by proteolysis, which then initiates the coagulation cascade. FXa and thrombin act as cellular activators via binding to protease-activated receptors (PARs), triggering different responses in several cell types, including fibroblasts, platelets, endothelial cells and cancer cells (online supplemental figure 1). FXa was reported to enhance the migration of breast, colon and lung cancer cells by activating PAR-1/2.7 The coagulation system is also a major player in the innate immune defense system response and correlates closely with immune system activity. FXa synthesized by monocytes and macrophages promotes fibrosarcoma and melanoma metastasis and immune escape by activating PAR-2.8 Conversely, metastasis of fibrosarcoma and melanoma are inhibited by targeted inhibition of FXa or knockout of FXa in macrophage by gene hybridization technology.8 Interestingly, the FXa inhibitor rivaroxaban exhibits a synergistic antitumor effect with the immune checkpoint PD-L1 inhibitor in fibrosarcoma and colon cancer.8 Other studies have demonstrated that FXa inhibitors have no significant effects on the growth of human breast and pancreatic cancers in immunodeficient mice,9 10 which may indicate that the function of FXa depends on immune cells response in vivo. Therefore, we speculate that the effects of FXa on tumors may depend not only on the immune system, but also on other mechanisms.

HCC is characterized by initial intrahepatic metastasis. Portal vein tumor thrombus is a common complication of advanced HCC, which causes multiple metastases inside the liver after shedding from the primary tumor.11 In this process, HCC cells need to overcome anoikis and simultaneously avoid immune surveillance to survive and metastasize. On the other hand, entry of metastatic cancer cells into the circulation, shedding of procoagulant extracellular vesicles, and recruitment/activation of inflammatory cells may further amplify cancer-associated thrombosis. The mechanism of tumor immune escape is predominantly associated with the modification of tumor cells and the alteration of the tumor immune microenvironment (TIME).12 13 The types and functions of immune cells and cytokines in tumors often exhibit changes, which lead to tumor cells escaping from immune surveillance.14 15 Recently, a pan-cancer analysis of human tumors revealed the link between coagulation/fibrinolysis and TIME.16 Activation of coagulation is conducive to the recruitment and activation of polymorphocytes and macrophages, thereby initiating the inflammatory response.17 These findings highlight the important role of the coagulation system in the tumor microenvironment and tumor immune escape. Additionally, coagulation is abnormally activated in the majority of patients with HCC, and FXa has proved to be closely related to metastasis of HCC.18 19 However, it remains unknown how FXa induces anoikis resistance in HCC and how FXa permits HCC to escape from immune surveillance. Moreover, whether FXa can be used as a biomarker to predict HCC immunotherapy efficacy, and whether inhibitors of FXa can enhance HCC immunotherapy remain unexplored. Here, we report that FXa activates PD-L1 expression to promote HCC immune escape both in vivo and in vitro, especially when cancer cells are in suspension. By binding to its receptor PAR-2, FXa promotes transcription of PD-L1 by activating transcription factor STAT2. Notably, combinations of PD-1/PD-L1 blockade and FXa inhibition may present a more efficient anti-HCC clinical option.

Materials and methods

Detailed methods are provided in Supporting Materials and methods (online supplemental file 2).

Patient tissue samples

For immunohistochemical staining and survival analysis, Cohort I and II were obtained from Lanzhou University Second Hospital. Cohort I includes 162 patients who had undergone surgical resection without chemotherapy or radiation, patient liver specimens were used for analysis of FXa level and its effect on prognosis of patients with HCC. Cohort II includes 21 HCC tissues, which were used to verify the correlation between FXa expression and tumor-infiltrating T cells via immunohistochemical staining. Cohort III includes 10 patients with HCC, who had surgical resection without chemotherapy or immunotherapy at Lanzhou University Second Hospital, and were used to measure the PD-L1 messenger RNA (mRNA) levels in HCC tissues and paired liver tissues. The diagnosis was performed according to the WHO tumor classification. Tumor stage was determined according to the eighth edition of the Union for International Cancer Control tumor staging. All research was conducted in accordance with both the Declarations of Helsinki and Istanbul. All specimens were collected with informed consent.

Mouse models and animal treatments

NOD/SCID and C57BL/6 mice (male, 4 weeks old) were purchased from Weitonglihua (Beijing, China) and Gansu Veterinary Institute (Lanzhou, China), respectively. All animals were housed in an specific pathogen free (SPF) condition in our laboratory.

All mouse models were established as previously described.20 Briefly, mice had operations at 6 weeks, weight ranges from 20 g to 22 g. Human HCC cell MHCC97-H and mouse HCC cell Hepa1-6 were injected into the spleen of individual mice at concentrations of 1×106/100 µL phosphate-buffered saline, respectively. After 28 days of treatment, the mice had vivo bioluminescence assays performed in order to evaluate the tumor size based on fluorescence intensity and size. Mice were sacrificed after assay and the tumors were weighed and minced for further analysis.

Rivaroxaban (Bayer Pharma AG) treatments were started 7 days after implantation with a dose of 2.6 mg/kg/day per mouse by chronic intragastric administration. And PD-1 antibody (Hengrui Medicine, Jiangsu) treatments were administrated at 9, 12, 15, 18 days after HCC cell implantation at a dose of 10 mg/kg by intraperitoneal injection as previously described.21 For vivo bioluminescence, mice were given intraperitoneal injections with 15 mg/mL D-Luciferin, sodium salt 100 µL and placed into the imaging chamber for X-ray and bioluminescence imaging.

Bioinformatic analyses

The Pearson correlation between immune cells and FXa expression were calculated using “GSVA” package based on RNA sequencing date in the TCGA database (https://portal.gdc.cancer.gov/).22 23 UCSC Genome Browse (http://genome.ucsc.edu) and JASPAR database (https://jaspar.genereg.net/) were used to search for PD-L1 promoter sequences and analyze transcription factors of PD-L1 and identify the binding sites.

Clinical study design and participants

In a single center, open­label, phase 3 trial, patients with unresectable HCC who had not previously received systemic treatment were randomly to receive either targeted therapy plus immunotherapy (control group) or targeted therapy plus immunotherapy and rivaroxaban (rivaroxaban group) until unacceptable toxic effects occurred or there was a loss of clinical benefit. The primary endpoints were overall survival in the intention-­to-treat population according to Response Evaluation Criteria in Solid Tumors, V.1.1 (RECIST V.1.1). This trial is registered with Chinese Clinical Trial Registry, number ChiCTR20000040540 (Clinical_Trial_Protocol, online supplemental file 3).

Statistical analysis

Data are presented as the mean±SD. Means between two groups were compared by Student’s t-test (n≥10) or Mann-Whitney test (n<10), and multiple comparisons were made by analysis of variance. Χ2 tests were used in comparing immunohistochemistry (IHC)-positive rates. Survival data are presented as Kaplan-Meier survival curves, and differences between groups were evaluated using the log-rank test. Pearson correlation analysis was performed to determine correlations among groups. P value<0.05 was considered statistically significant.

ResultsFXa expression correlates with prognosis of patients with HCC

The role of FXa in HCC progression has not been explored. Here, we showed that FXa expression was upregulated in HCC tissues compared with normal liver tissues (figure 1A,B), and its expression positively correlated with portal venous emboli (p<0.0001), pathological Tumor-Node-Mteastasis (pTNM) stage (p<0.0001) and recurrence (p=0.039) (table 1). Kaplan-Meier survival analyses demonstrated that patients with higher FXa expression have shorter overall survival (OS) (figure 1C). Cox regression analysis indicated that FXa expression, portal venous emboli, pTNM stage and recurrence are risk factors for poor prognosis in HCC (online supplemental table 1). These observations indicated that FXa may serve as a prognostic molecular marker and promoter of tumor metastasis.

Figure 1

FXa promotes metastasis of HCC in vitro and in vivo. (A) FXa is upregulated in HCC tissues compared with normal tissues. (B) Representative staining of FXa proteins in HCC tissues by IHC. FXa protein was primarily located in the cytoplasm. Scale bars, 100 µm (upper panels) and 20 µm (lower panels). (C) Kaplan-Meier analyses of overall survival based on FXa expression levels; high expression of FXa predicts a poor prognosis of patients with HCC in 162 cases (Cohort I) (p<0.0001). (D–E) FXa depletion inhibits the invasion and migration of MHCC97-H and HCCLM3 cells, and overexpression of FXa in huh-7 cells promotes invasion and migration. Data are mean±SEM from at least three independent experiments. (F) FXa depletion inhibits metastasis of MHCC97-H cells in vivo. Bioluminescence images of tumors were visualized using an in vivo imaging system (left panels). The gross specimens and ratio of liver weight/ body weight were analyzed, which in the sh-FXa group was lower than that in the control (medium panels); H&E staining showed the number of liver tumors, and metastasis lesions in the sh-FXa group were lesser than that in the control (the red arrows represent liver tumors) (right panels). Data represent mean±SEM (n=3). *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. FXa, factor Xa; HCC, hepatocellular carcinoma; IHC, immunohistochemistry.

Table 1

The correlation between FXa expression and clinicopathologic characteristics of hepatocellular carcinoma (Cohort I)

FXa promotes HCC metastasis through anoikis resistance

To further investigate the influence of FXa on HCC progression, MHCC97-H and HCCLM3 cells were used for FXa knockdown, and huh-7 for FXa overexpression (online supplemental figure 2A,B). In vitro invasion and migration of HCC cells were significantly inhibited by downregulation of FXa, and enhanced after FXa overexpression (figure 1D,E). Furthermore, the metastasis ability of HCC cells was also inhibited in immune-deficient mice by FXa downregulation (figure 1F). The results suggest that FXa promotes metastasis of HCC.

To investigate the mechanism by which FXa promoted HCC metastasis, HCC cells were cultured as a single-cell suspension in poly-HEMA-coated dishes to mimic the status of tumor cells in the blood and lymphatic circulation. After suspension for 24 hours and 48 hours, FXa was significantly upregulated in HCC cells compared with adherent culture (figure 2A). Next, we determined apoptosis of HCC cells under adherent or suspension conditions. We found that the apoptosis rate was increased after cell suspension, but such increases in HCC cells were significantly lower than that in L02 (figure 2B), suggesting that HCC cells exhibit increased anoikis resistance compared with normal cells. To determine if FXa may contribute to anoikis resistance, we depleted FXa in HCC cell. Suspension-induced apoptosis was significantly increased in HCC cells by FXa knockdown (figure 2C). Meanwhile, Fas and caspase-3/7, the extrinsic apoptosis pathway markers were also dramatically increased in suspension status by FXa knockdown (figure 2D). However, the intrinsic apoptosis pathway markers Bcl-2, Bax and caspase-9 showed no significant changes (online supplemental figure 3). In conclusion, FXa may promote metastasis by conferring anoikis resistance on HCC cells.

Figure 2

FXa inhibits the anoikis of HCC cells. (A) Compared with liver cells L02, FXa expression was upregulated in MHCC97-H, HCCLM3 and huh-7 cells grown in suspension for 24 hours and 48 hours. (B) Flow-cytometric analysis of apoptotic cell content. The apoptosis rate of L02 and HCC cells was increased after suspending cells for 24 hours and 48 hours, while the apoptosis of HCC cells was lower than normal liver cells. Both early (annexin V-FITC (AV) + /propidium iodide (PI) −) and late (AV + /PI +) apoptotic cells were counted. (C–D) FXa depletion increases apoptosis of HCC cells and the extrinsic apoptotic markers Fas, cleaved caspase-3/7 were also upregulated on suspension condition; while apoptosis rate and apoptotic markers were both inhibited when FXa overexpressed in huh-7 cells. *p<0.05, **p<0.01, ***p<0.001. ns, not significant. FXa, factor Xa; HCC, hepatocellular carcinoma.

FXa inhibits CD8+ T cell-mediated antitumor immunity in patients with HCC

In addition to promoting anoikis resistance of HCC cells, it has not been explored whether FXa was also involved in immune response in HCC. By analyzing the TCGA databases, we found that FXa expression was inversely correlated with numbers of tumor-killing immune cells, while positively correlated with Th17 cells in HCC (figure 3A). Importantly, enrichment of T cells was negatively correlated with FXa levels (R=−0.206, p<0.001) in HCC (figure 3B,C). Next, we investigated the tumorous FXa expression and counts of tumor-infiltrating lymphocytes (TILs) in HCC tissues from 21 patients (online supplemental table 2). The results showed that FXa level was negatively correlated with the number of CD8+ TILs (p=0.0487) in HCC, but exhibited no significant correlation with CD4+ TILs (p=0.4981) (figure 3D). Further analyses demonstrated that patients with high levels of FXa and low CD8+ TILs have a shortest OS time, whereas the prognosis in patients with low levels of FXa and high CD8+ TILs was best (p=0.0146) (figure 3E). Moreover, we analyzed the efficacy of 12 cases who received immunotherapy based on the RECIST V.1.1 standard. Results indicated that patients who achieved partial response showed low expression of FXa, however, FXa levels had no significant correlations with the prognosis of patients with HCC receiving immunotherapy (p=0.0831) (figure 3F,G). The results suggest that the FXa level of HCC may be involved in the exhaustion of CD8+ TILs.

Figure 3

FXa negatively correlated with T cells in HCC. (A) TCGA databases analysis showed that FXa was negatively correlated with numbers of tumor-killing immune cells, while positively correlated with Th17 cells in HCC; (B–C) TCGA data analysis revealed that enrichment of T cells was inversely correlated with FXa level in HCC (R=−0.206, p<0.001); (D) immunohistochemical analysis of FXa expression and tumor-infiltrating T cells content in 21 HCC samples (Cohort II). FXa level was inversely correlated with CD8+ T cells in patients with HCC (p=0.0487), while showing no significant correlation with CD4+T cells (p=0.4981) (scale bar: 20 µm) (E) Kaplan-Meier analyses of overall survival based on the combination of FXa expression and the number of identified infiltrating cells, the prognosis of high CD8+T cells infiltration (n=11) was better than low infiltration (n=10) (p=0.0184); and high CD8+T cells infiltration combined with low FXa level (n=7) indicates the best prognosis of patients with HCC, while low CD8+T cells infiltration combined with high FXa level indicates the worst prognosis (n=6) (p=0.0146); (F–G) analyses of correlation between FXa level and immunotherapy efficacy in 12 patients with HCC. Patients with better response to ICI had a low expression of FXa (n=3); while there was no significant correlation between FXa level and prognosis of patients receiving ICI (p=0.0831). FXa, factor Xa; HCC, hepatocellular carcinoma; ICI, immune checkpoint inhibitor; PD, progressive disease; PR, partial response; SD, stable disease; TPM, transcripts per million.

To evaluate whether FXa-mediated survival benefit in HCC was dependent on T cells, healthy human peripheral blood mononuclear cells (PBMCs) were co-cultured with HCC cells. After co-culturing with PBMCs for 24 hours, the apoptosis rate was significantly increased in HCC cells depleted of FXa compared with control cells under both adherent and suspension culture conditions, while it was decreased after FXa overexpression (figure 4A). These results indicate that FXa depletion enhances T cell-mediated tumor cell killing.

Figure 4

FXa inhibits CD8+ T cell-mediated antitumor immunity. (A) Flow-cytometric analysis of the effects of FXa on apoptosis of HCC cells after co-culturing with PBMCs in suspension for 24 hours. Compared with negative control, PBMCs exhibited a killing effect on HCC cells, and FXa depletion dramatically increased apoptosis of HCC cells; while FXa overexpression decreased apoptosis of HCC cells. Both early (AV + /PI −) and late (AV + /PI +) apoptotic cells were counted. (B–C) CD3+ CD4+ CD25+ FOXP3+ Treg cells and CD3+ CD8+ T cells percentages following co-culture of HCC cells and PBMCs for 24 hours in suspension state. (B) CD3+ CD4+ CD25+ FOXP3+ Treg cells percentage (third quadrant blue labels) in the FXa depletion or overexpression group had no significant alterations; (C) CD8+T cells percentage (third quadrant blue labels) in the FXa depletion group was higher than that in the control, while FXa overexpression decreased the CD8+T percentage. (D–E) ELISA analysis of TNF-a, IFN-γ, perforin and granzyme B levels in supernatants from 24 hours co-cultures in suspension. FXa depletion in MHCC97-H and HCCLM3 cells significantly increased the TNF-a, IFN-γ, perforin and granzyme B levels in supernatants, and FXa overexpression in huh-7 cells decreased these levels. Data represent mean±SEM. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. ns, not significant. AV, annexin V-FITC; FXa, factor Xa; IFN, interferon; HCC, hepatocellular carcinoma; PBMCs, peripheral blood mononuclear cells; PI, propidium iodide; TNF, tumor necrosis factor; Treg, regulatory T cells.

To further verify the above notion, we evaluated the effects of FXa on the number and function of TILs by co-culture experiments. The counts of CD3+ CD4+ CD25+ FOXP3+ regulatory T cells (Treg) exhibited no significant changes in FXa knockdown groups compared with controls on suspended status, whereas CD3+ CD8+ T cells were increased dramatically (figure 4B,C). However, the counts of both TILs did not change significantly under adhesion condition after FXa depletion (online supplemental figure 4A,B), which indicates that FXa may specifically inhibit CD8+ T cells on suspended status. In addition, CD8+ TILs mediated antitumor immunity depends on certain cytokines, among which tumor necrosis factor-a (TNF-a), interferon-γ (IFN-γ), interleukin-2 (IL-2), granzyme B and perforin are the most important antitumor cytokines. Similarly, all cytokines except IL-2 were significantly increased in FXa knockdown groups compared with the controls under suspended (figure 4D,E) and adhesion conditions (online supplemental figure 4C,D). Collectively, these results indicate that FXa may attenuate CD8+ T cell-mediated antitumor immunity to promote HCC immune escape.

FXa promotes HCC progression by activating PAR-2

It is well known that PAR-2 is the main receptor of FXa and regulates tumor progression in fibrosarcoma and colon cancer.8 19 24 We first explored whether PAR-2 was the potential downstream effector of FXa-induced anoikis resistance and antitumor immunity in HCC. PAR-2 levels were downregulated by knockdown of FXa in HCC cells, and significantly upregulated in cells in suspension (online supplemental figure 5A), indicating that PAR-2 level is positively correlated with FXa expression. We also examined PAR-1 and PAR-3 levels, however, FXa knockdown or overexpression showed no significant effect on their expression (online supplemental figure 5A). To evaluate the effects of PAR-2 on HCC progression, we knocked down PAR-2 expression by lentivirus (online supplemental figure 5B). Migration and invasion of HCC cells was inhibited by downregulating PAR-2 (figure 5A,B). Meanwhile, the apoptosis rate was increased after PAR-2 depletion in suspended cell cultures when compared with the control group (figure 5C), rather than in the adhesion state (online supplemental figure 5C). Furthermore, rescue experiments revealed that PAR-2 knockdown attenuated migration and invasion ability induced by FXa overexpression (figure 5D,E). The apoptosis rate of HCC cells was reversed by FXa overexpression in suspended cells (figure 5F), rather than in adherent cell cultures (online supplemental figure 5D). Collectively, these results indicated that FXa may inhibit anoikis and attenuate CD8+ T cell-mediated antitumor immunity by activating PAR-2.

Figure 5

FXa promotes HCC progression by activating PAR-2. (A–B) PAR-2 depletion in MHCC97-H and HCCLM3 cells inhibits invasion and migration, and (C) promotes apoptosis of HCC cells following suspension for 24 hours. Both early (AV + /PI −) and late (AV + /PI +) apoptotic cells were counted. (D–E) The enhanced invasion and migration ability of HCC cells by FXa overexpression was reversed by PAR-2 depletion. (F) FXa overexpression decreased apoptosis rates in suspended huh-7 cell cultures and was reversed by PAR-2 depletion. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. AV, annexin V-FITC; FXa, factor Xa; HCC, hepatocellular carcinoma; PAR, protease-activated receptor; PI, propidium iodide.

FXa/PAR-2 activation of PD-L1 transcription by STAT2 enhances HCC immune escape

T-cell exhaustion is one of the mechanisms leading to the immune escape of tumor cells, and the immune checkpoint ligand PD-L1 is involved in T-cell exhaustion. Inhibition of PD-1/PD-L1 signaling can restore the activity and function of CD8+ TILs, and prevent the immune escape of tumor cells. Here, we measured the PD-L1 mRNA levels in 10 patient with HCC specimens (online supplemental table 2). We found that PD-L1 mRNA levels in cancer tissues were dramatically increased when compared with levels in normal tissues (online supplemental figure 6A). Furthermore, PD-L1 was significantly inhibited by FXa and PAR-2 downregulation in HCC cell lines, and upregulated by FXa overexpression in huh-7 cells in suspension conditions (online supplemental figure 6B,C and G,H). Collectively, the results indicate that PD-L1 expression is dysregulated in HCC, which may be attributed to FXa and PAR-2.

Mechanisms of FXa/PAR-2 regulation of PD-L1 transcription have not been explored. PD-L1 abundance is regulated by genomic alteration, epigenetic modification, transcriptional regulation, post-transcriptional regulation, and post-translational modification.25 Considering that PD-L1 mRNA is abnormally expressed in HCC and FXa/PAR-2 depletion inhibits PD-L1 mRNA levels, we explored the regulatory mechanism of PD-L1 at the transcriptional level. The JASPAR database was used to explore the potential transcription regulators of PD-L1, and a correlation heat map of transcription factors and PD-L1 was produced based on the TCGA sequencing database. Results indicate that STAT1/2/3 have the highest correlations with PD-L1, which might be the potential regulatory transcription factors of PD-L1 (online supplemental figure 6D,E). Based on the TCGA database, we further analyzed the relationships between PD-L1 expression and transcription factors. STAT1/2/3 levels were positively correlated with PD-L1 expression (online supplemental figure 6F). Previous research has also revealed that inhibition of PAR-2 induces apoptosis of cervical cancer by interfering with STAT3 signaling,26 and STAT3 is an important transcription factor of CD274, which promotes tumor immune escape by increasing the transcription of PD-L1. Therefore, we verified whether FXa/PAR-2 also promoted PD-L1 transcription by activating STAT1/2/3 in HCC. FXa and PAR-2 synergistically activated the expression of STAT2, but not STAT1 or STAT3 (online supplemental figure 6G,H).

When phosphorylated, STATs are activated to become transcriptional activators and enter the nucleus to bind to regulatory elements of target genes to promote their transcription. Our results indicate that STAT2 was phosphorylated (p-STAT2) in suspension cell cultures, but not in adherent cells, and FXa and PAR-2 knockdown dramatically decreased phosphorylation levels (online supplemental figure 7A,B). Furthermore, immunofluorescence also revealed that after suspending stimulation, p-STAT2 was upregulated and mainly expressed at the nucleus (online supplemental figure 7C). Accordingly, we confirmed that suspension stimulation-induced phosphorylation and nucleus translocation of STAT2 in HCC cells, and FXa/PAR-2 were positively correlated with p-STAT2 levels.

To determine whether p-STAT2 can directly bind to the PD-L1 promoter, we evaluated several binding sites using the JASPAR database (online supplemental figure 7D and table 5). Then, we performed chromatin immunoprecipitation-quantitative polymerase chain reaction (ChIP-qPCR) analysis of the top five binding sites on the JASPAR database in MHCC97-H and HCCLM3 cells (online supplemental figure 7E). Results demonstrated that the site (−1803/–1789) and site (−425/–411) of the PD-L1 promoter region were the main binding sites of p-STAT2 (online supplemental figure 7F). Subsequently, we conducted dual luciferase reporter assay in MHCC97-H and HCCLM3 cells to further validate the transcriptional activation effect of STAT2 on CD274. According to the ChIP-qPCR results, we mutated the third site and the fifth site, respectively, in the PD-L1 promoter region that STAT2 mainly bind. The luciferase activity of the wild-type CD274 promoter was significantly higher than that of the control group and that of the mutant five-site PD-L1 promoter (online supplemental figure 7G). These data suggested that p-STAT2 directly binds to the CD274 promoter to transcriptionally upregulate PD-L1 expression. Collectively, FXa may enhance the transcription of PD-L1 by binding to its receptor PAR-2 and activating STAT2 in suspended HCC cells, which ultimately promotes immune escape of HCC.

FXa inhibition combined with PD-1 blockade therapy enhances HCC therapeutic efficacy in animal models

Inhibition of the PD-1/PD-L1 signaling is a viable strategy for normalizing the dysregulated tumor microenvironment (TME).27 Until now, anti-PD-1/PD-L1 treatments have exhibited durable antitumor activities in various cancers.6 28 29 However, the beneficiary population of anti-PD-1/PD-L1 treatments alone remains a small percentage, and most patients have no response.30 31 We hypothesized that FXa inhibition might synergize with PD-1/PD-L1 blockade to enhance the therapeutic efficacy of HCC.

To evaluate whether FXa depletion might have synergistic effects with PD-1/PD-L1 blockade on HCC progression, we used Heap1-6 mouse HCC cells, which highly express FXa (figure 6A). We established an intrahepatic metastasis animal model in immune-competent mice as previously described20 (figure 6B). FXa specific inhibitor rivaroxaban and PD-1 antibody were used for tumor immunotherapy studies in vivo. Results demonstrate that rivaroxaban and PD-1 antibody monotherapy, respectively, inhibited metastasis of HCC cells in vivo when compared with the control group. Moreover, the combination of rivaroxaban with anti-PD-1 therapy more potently suppressed tumor metastasis than either monotherapy alone (figure 6C–E). Consistent with the results from the co-culture system, FXa inhibition in transplanted tumors increased the absolute counts of CD8+ TILs but not FOXP3+ Treg cells, especially in rivaroxaban combined with anti-PD-1 therapy group (figure 6F,G). Meanwhile, both mRNA and protein level of PD-L1 were also decreased (figure 6H,I). These results reveal combining coagulation FXa inhibitors and PD-1/PD-L1 immune checkpoint blockade can definitely enhance the therapeutic efficacy for HCC and eventually expand the pool of beneficiaries of liver cancer immunotherapy.

Figure 6

FXa inhibition combined with PD-1 blockade enhances HCC therapeutic efficacy. (A) FXa protein is highly expressed in Hepa1-6 mouse HCC cells compared with H22 cells. (B) Animal model of intrahepatic metastasis and treatment strategy of Hepa1-6 HCC cells in C57BL/6 mice. (C) The bioluminescence images by in vivo imaging system showed that FXa inhibitor combined with PD-1 antibody significantly inhibited metastasis of HCC cells in vivo (n=5). (D) Gross specimens of tumors in liver indicated that the ratio of liver weight/body weight in tumor-burdened C57BL/6 mice was dramatically decreased in rivaroxaban and PD-1-antibody combined treatment group compared with that in other treatment alone. (E) H&E staining confirmed the liver metastasis tumors. Compared with the control group, tumor numbers were reduced in rivaroxaban and anti-PD-1 combined treatment group. Scale bars, 200 µm. (F–G) Immunohistochemical analysis of FOXP3+ Treg cells and CD8+ TILs count in tumors illustrated that the numbers of CD8+ TILs in the combined group were significantly higher than those of control and either treatment alone. However, FOXP3+ Treg cells had no significant alteration. Scale bars, 100 µm (upper panels) and 20 µm (lower panels). (H–I) FXa protein level was significantly inhibited in rivaroxaban and combined groups; PD-L1 protein and mRNA levels were significantly decreased in treatment groups, especially in the combined group. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001. FXa, factor Xa; HCC, hepatocellular carcinoma; mRNA, messenger RNA; PD-1, programmed cell death protein-1; PD-L1, programmed cell death ligand-1; TIL, tumor-infiltrating lymphocyte; Treg, regulatory T cells.

FXa inhibition combined with ICIs enhances HCC therapeutic efficacy in patients with advanced HCC

Between April 30, 2021, and October 9, 2023, 36 patients were enrolled. 18 patients to the rivaroxaban group (2 patients are not evaluable, 5 data missing) and 18 to the control group (3 patients are not evaluable, 3 data missing) for efficacy and safety analyses. Base­line characteristics were generally well-balanced between treatment groups (online supplemental table 6). CT results demonstrated a statistically significant difference in terms of the clinical response between the two groups (online supplemental figure 8A,B). Furthermore, rivaroxaban improved OS (online supplemental figure 8C); median survival was not reached for rivaroxaban versus 10 months (95% CI, 7.26 to 12.74) for control. Given that results for OS were statistically significant, objective response rates were sequentially evaluated (online supplemental table 7). The objective response rates were 27.3% (95% CI, 9.7% to 56.6%) with rivaroxaban and 16.7% (95% CI, 4.7% to 44.8%) with the control group, disease control rates were 81.8% (95% CI, 52.3% to 94.9%) and 58.3% (95% CI, 31.9% to 80.7%), respectively, according to independent assessment with RECIST V.1.1 (p<0.001). No patients in both treatment groups had a complete response.

Adverse events were reported in all rivaroxaban recipients (11 (100%) of 11) and (12 (100%) of 12) control recipients (online supplemental table 8). The most common clinically relevant grade 3 or 4 treatment-emergent events were increased γ-GGT (one patient (9.1%) in the rivaroxaban group vs one patient (8.3%) in the control group), hypertension (one patient (9.1%) vs 1 (8.3%)), and thrombocytopenia (one patient (9.1%) vs two patients (16.7%)). Clinically relevant non-major bleeding occurred in one patient in the rivaroxaban and zero control groups, respectively (cumulative incidence, 9.1% vs 0%).

Discussion

Our study suggests that through binding to its receptor PAR-2, FXa expressed in HCC cells promotes transcription of PD-L1 by activating transcription factor STAT2, which ultimately leads to anoikis resistance and immune escape of HCC. FXa-activated PAR-2 endows HCC cells with ability to survive in the circulating blood by inhibiting the extrinsic apoptosis pathway. Additionally, FXa upregulates PD-L1 expression, which decreases the number of TILs and attenuates the function of CD8+ TILs, enabling HCC cells to escape from the immunity. Metastatic HCC cells are detached from primary tumor lesions to enter the circulation, and eventually metastasize to the liver or distant organs. In this process, such HCC cells must overcome anoikis and simultaneously avoid immune surveillance. As we demonstrated, the FXa signaling pathway and its downstream effectors play a central role in anoikis resistance and immune escape in HCC. More importantly, inhibition of FXa suppressed HCC growth and metastasis, and FXa-inhibition combined with anti-PD-1 treatment had a synergized antitumor effect on HCC.

FXa is a key serine protease of the coagulation cascade acting as the convergence point of the intrinsic and extrinsic clotting pathways, leading to the thrombin formation. Importantly, FXa is mainly synthesized and secreted by the liver, and FXa has been demonstrated to regulate the growth and metastasis of solid tumor, such as melanoma and endometrial cancer.7 However, the role of FXa in HCC has remained elusive. Here, we have demonstrated that high FXa expression predicted a poor prognosis of HCC patients. Meanwhile, our results revealed that FXa also promoted the growth and metastasis of HCC cells in vitro and in vivo. Thus, FXa has great potential to become a promising biomarker of HCC prognosis and also a potential target for therapeutic suppression.

Anoikis is a programmed cell death caused when cells detach from the extracellular matrix and other cells. Anoikis is mediated by different pathways, including the mitochondrial (intrinsic) and cell surface death receptor (extrinsic) pathways. The intrinsic pathway is activated by cellular stress responses or apoptotic signals, and regulated by the Bcl-2 family through controlling mitochondrial permeability. The extrinsic pathway is mainly activated by extracellular signals through binding to death receptors such as Fas and TNFR1. Both pathways activate endonucleases eventually, leading to DNA fragmentation and cell death.32 Resistance to anoikis allows tumor cells to survive after detachment and to form metastatic lesions. Here, we revealed that FXa activates its receptor PAR-2, which confers resistance to anoikis in HCC cells by inhibiting the extrinsic apoptosis pathway. Previous studies suggested that FVIIa-TF interaction at physiological concentrations inhibited activation of caspase-3 and induced anoikis resistance of Baby Hamster Syrian Kidney cells, and FXa generation positively regulated this process.33 Here we first demonstrated that FXa can also inhibit anoikis of HCC cells through PAR-2. Few studies have explored the relationship between FXa and PAR-2 in HCC.