Direct inhibition of the TLR4/MyD88 pathway by geniposide suppresses HIF‐1α‐independent VEGF expression and angiogenesis in hepatocellular carcinoma

Background and Purpose As a typical hypervascular tumour, hepatocellular carcinoma (HCC) is predominantly grown through angiogenesis. Geniposide is a promising anti‐inflammatory compound found in Gardenia jasminoides, but its effects on the progression of HCC remain untested. Experimental Approach The anti‐HCC effects of geniposide was investigated in cellular models and orthotopic HCC mice. Transcriptional regulation of the VEGF promoter was measured by dual‐luciferase reporter assay. The anti‐angiogenic action of geniposide was measured by tube formation assay. Both surface plasmon resonance techniques and human phospho‐kinase array analysis were utilized to validate the relationship between targets of geniposide and hepatocarcinogenesis. Key Results Geniposide exhibited significant disruption of HCC proliferation, invasion, angiogenesis and lung metastasis in orthotopic HCC mice. Geniposide inhibited secretion of VEGF by HCC and suppressed the migration of endothelial cells and the formation of intra‐tumour blood vessels, without cytotoxicity and independently of the transcription factor HIF‐1α. Direct inhibition of TLR4 by geniposide led to the shutdown of the TLR4/MyD88 pathway and STAT3/Sp1‐dependent VEGF production. However, LPS, an agonist of TLR4, restored STAT3/Sp1‐related VEGF production in geniposide‐inhibited HCC angiogenesis. Conclusion and Implications The direct inhibitory effect of geniposide on TLR4/MyD88 activation contributes to the suppression of STAT3/Sp1‐dependent VEGF overexpression in HCC angiogenesis and pulmonary metastasis. This action of geniposide was not affected by stabilization of HIF‐1α. Our study offers a novel anti‐VEGF mechanism for the inhibition of HCC.


| INTRODUCTION
Hepatocellular carcinoma (HCC), a common and lethal form of cancer, results in an annual mortality of one million people worldwide. The high death rate reflects the unpredictable manifestations of HCC in the early stage and poor prognosis in the advanced stages of this condition (Sheng, Qin, Zhang, Li, & Zhang, 2018). Notably, only 30% of HCC patients are suitable for surgical therapy, including liver transplantation (Santambrogio et al., 2016). As a typical hypervascular cancer, HCC relies on angiogenesis to grow and metastasize. Accordingly, treatment of HCC has adopted an anti-angiogenic approach, by blocking the expression of the VEGF family of ligands and receptors . The multi-kinase inhibitor sorafenib has been approved as the first-line therapy for advanced-stage HCC with its anti-angiogenic effect. However, its benefits are transient due to the worsening hepatic dysfunction and promoting regrowth (Fang et al., 2018). Thus, there is a pressing need for a novel therapeutic mechanism and new compounds for the suppression of HCC growth and metastasis.
In HCC, the production of VEGF facilitates the recruitment of endotheliocytes and enhances vascular permeability. Subsequently, angiogenesis generates an increased supply of nutrition and of metastatic HCC cells (Chesnokov et al., 2018). VEGF is a principal factor for abnormalities in vascular structure and function, contributing to HCC angiogenesis and metastasis .
Aberrant expression of VEGF can be directly involved in the pathological aspects of HCC (Deng et al., 2013). The level of VEGF is proportional to the progression of HCC, including tumour invasion, metastasis and differentiation . This pathological mechanism provides a robust theoretical basis for the antiangiogenic strategy of controlling HCC progression, by VEGF inhibition.
However, scientific investigations focusing on its anti-tumour property are scanty, especially in HCC. Regarding its reported anti-HCC studies in nearly two decades, geniposide suppresses HepG2 and Huh7 cells by regulating miR-224 via blocking the Wnt/β-catenin and Akt cascades (Yu, Wang, Tao, & Sun, 2019). Also, aflatoxin B 1induced HCC in the rat was improved by geniposide through inhibiting γ-glutamyl transpeptidase activity (Lin et al., 2000). Our previous studies have demonstrated the suppressive effect of its aglycone, genipin, on HCC through inhibiting intrahepatic metastasis and regulating the tumour-associated macrophages (Tan et al., 2016; N. Wang et al., 2012). Notably, an earlier study showed that geniposide exhibited an anti-angiogenic effect in the chick embryo chorioallantoic membrane assay (Koo et al., 2004). However, whether geniposide inhibits HCC through its anti-angiogenic action remains unknown.
In the present study, we evaluated the inhibitory effect of geniposide on the development of HCC, in vivo and in vitro, including the proliferation, invasion, metastasis and angiogenesis of HCC.

What is already known
• Progression of hepatocellular carcinoma is known to involve VEGF-induced angiogenesis which is dependent on HIF-1α.
• Geniposide exerts many anti-inflammatory effects, but its effects on angiogenesis in hepatocellular carcinoma are untested.

What this study adds
• A novel anti-VEGF mechanism, involving the TLR4/ MyD88 pathway, to inhibit progression of hepatocellular carcinoma .

What is the clinical significance
• Novel therapeutic targets for VEGF inhibition in hepatocellular carcinoma may improve pharmacological management of cancer.
• Geniposide, a potential and affordable anti-VEGF compound, shows promise as clinical treatment for hepatocellular carcinoma.
2.2 | Orthotopic HCC implantation in athymic nude mouse All animal care and experimental protocols were approved by the Committee on the Use of Live Animals in Teaching and Research, The University of Hong . Animal studies are reported in compliance with the ARRIVE guidelines (Kilkenny, Browne, Cuthill, Emerson, & Altman, 2010) and with the recommendations made by the British Journal of Pharmacology.
A model of orthotopic HCC-implanted mice was established as described in our previous report (Wang, Han, et al., 2017). In brief, a solid HCC tumour was formed within 2 weeks by a subcutaneous injection of 5 × 10 6 luciferase-labelled MHCC-97L cells into the lateral abdomen of nude mice. Afterwards, cluster-shaped HCCs were cut into 1-mm 3 pieces followed by transplantion into the left lobe of a nude mouse liver. Seven days after implantation, mice were given geniposide (Neautus, China; 30 mgÁkg −1 for 2 days, intragastrically) and/or a single i.p. injection of LPS (3 mgÁkg −1 ; Sigma-Aldrich, USA) at the beginning of a 5-week treatment. LPS was utilized as an agonist of toll-like receptor 4 (TLR4; Pizzuto et al., 2019). Growth of the HCC was monitored, once a week, by luciferase imaging analysis, using the IVIS Spectrum system (PerkinElmer, USA). All the mice were killed via an overdose injection of pentobarbitone (200 mgÁkg −1 ) at the end of the experiment.

| Histology and immunofluorescence
For the histological study, paraffin-embedded sections (5 μm) were collected on the surface of the slides followed by de-waxing in xylene and rehydrated by alcohol with gradient concentration from 100% to 70% (10% per interval). After the processing of rehydration, slides were incubated with haematoxylin and 0.25% eosin for 5 and 1 min, respectively (Sigma-Aldrich, USA). Afterwards, slides were immersed in Canada balsam (Sigma-Aldrich, USA) followed by the observation of haematoxylin and eosin-stained HCC morphology with a BX43 light microscope (Olympus, Japan; n = 5 per group).
The immuno-related procedures used comply with the recommendations made by the British Journal of Pharmacology. For the immunofluorescent analysis, staining in tissue is slightly different from that in the cell. Both de-waxing and antigen retrieval steps were applied in the tissue staining alone. Other staining steps are identical.
For details, the de-waxed sections (n = 5 per group) were immersed in 10-mM citrate buffer (Sigma-Aldrich, USA) for antigen retrieval, followed by incubating with 10% goat serum for 30 min for blocking procedure. Then, sections were incubated for 12 h at 4 C with anti-

| Bioinformatics study
Based on the analysis of human phospho-kinase array, the underlying mechanism of geniposide-induced HCC inhibition was predicted by bioinformatics analysis, as in our previous studies (Wang, Tan, Li, & Feng, 2017;Zhang, Wang, et al., 2018). In brief, differentially expressed genes were imported into Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis followed by the identification of the contributive targets and generich terms according to the enrichment analysis on the platform of DAVID (available at https://david.ncifcrf.gov). In particular, GO analysis for target mining is composed of three ontologies, including biological process, molecular functions and cellular components.
GlueGo-based network pharmacology was constructed by Cytoscape (access at http://www.cytoscape.org/). To enhance the functionality of target prediction, cluster tree analysis was applied by SPSS Statistics (SPSS Inc., USA) and RStudio (R Inc., USA). The correlation degree of -log 10 (P value) from each identified geniposide-regulated pathway was imported into SPSS for transforming all variables to Z scores by choosing the method of squared Euclidean distance, which aimed to yield equal metrics and weighting value. Next, the hierarchical cluster membership was defined by SPSS-dependent weighted cluster analysis, followed by further identifying homogeneous biological terms in each cluster.
Finally, the R programming visualization was performed by RStudio to circle-hierarchically show the outcomes.
Ubc.GFP (lentiviral expression of constitutively active STAT3). In detail, the contributors of these specific plasmids were listed as fol-

| Real-time quantitative PCR
Total RNA from HCC cells was extracted and purified by using TRIzol (Takara, Japan; n = 5 per group). PrimeScript RT master mix (Takara, Japan) was used to reverse transcription. Quantitative PCR was conducted by using SYBR Green Probe (Takara, Japan) and

| Flow cytometry
The assays for programmed cell apoptosis and necrosis were per-  At the end of the experiment, the enzyme/substrate reaction was terminated by the stop buffer. VEGF concentration was determined at 450-nm absorbance (n = 5 per group). Bands were visualized under the chemiluminescence system (Bio-Rad, USA).

| Migration assay
A co-culture system was established to assess the cross-biological activity of HCC-induced HUVEC migration. HUVECs were inoculated in the donor chamber, whereas PLC/PRF/5 cells were seeded in the receiving chamber. Both cells were maintained in the serum-deprived medium condition for 6 h. After then, HUVECs and geniposidetreated PLC/PRF/5 were cultured in the complete medium under either normoxic or hypoxic condition at 37 C for 24 h. Next, the remaining cells located in the donor insert were removed by a cotton ball, while cells at the receiving chamber were fixed by 4% paraformaldehyde (PFA) for 2 h followed by staining with crystal violet. The number of migrated HUVECs was quantified by ImageJ (NIH, USA) in five random fields of each well. Representative images were captured via a microscope (EVOS, USA).

| Endothelial cell tube formation assay
Each well of a 24-well plate was coated with 300-μl basement membrane matrix with reduced growth factor (Invitrogen, USA) for 30 min at 37 C prior to seeding with HUVECs with complete (EGM-2) or conditioned medium. PLC/PRF/5-culture medium (DMEM for 24 h) with or without either LPS (100 ngÁml −1 ) or recombinant VEGF (re-VEGF) treatment (20 ngÁml −1 , 293-VE, R&D Systems) were prepared as the conditional media according to the published paper (Bhattacharya et al., 2016). After setting the matrix, HUVECs (JCRB Cell Bank, Japan) were seeded in each well (approximately 4 × 10 4 cells per well) and incubated in various conditional media for 6 h. Then, each well was washed with PBS followed by adding 4% PFA for 1 h at room temperature. After removing 4% PFA, the figures of tube formation were captured by a microscope (EVOS, USA). Quantification of tubular networks, including total branching length and number of tubes, was performed by ImageJ with Angiogenesis Analyzer plugin (NIH, USA) in six independent wells.

| Dual-luciferase assay
The determination of the activity of the VEGF promoter reporter

| Proteome profiler human phospho-kinase array
Total proteins (100 μg) were extracted from PLC/PRF/5 cells with or without 24-h geniposide treatment in the normoxic condition. The lysed proteins were transferred to the human phospho-kinase array (R&D Systems, USA). Forty-three phospho-site-specific antibodies were spotted onto the antibody array. Measurement of signal detection was conducted in line with instructions from the manufacturer.
Signal intensities of spots on the array were analysed via ImageJ (NIH, USA).

| Surface plasmon resonance-based binding analysis
The affinity determination of geniposide binding to TLR4 was measured by Biacore X100-based surface plasmon resonance (SPR) technology with the BIA evaluation system (GE Healthcare, Sweden). SPR assay is performed by an extremely sensitive biosensor that offers rapid screening of the alterations of refractive index on a molecular-immobilized chip in response to another small molecule in a real-time and label-free manner. It is a powerful quantitative tool to monitor biomolecular interactions and provide specific kinetic and affinity determination, including the dissociation constant (K D ). In the Biacore X100 system, recombinant human TLR4 (R&D Systems, USA) is covalently immobilized on the surface of a CM5 chip by a coupling buffer (10-mM sodium acetate, pH = 5) from the Amine Coupling Kit (GE Healthcare, Sweden). The coupling level was 1,300 RU. Then, a running buffer (PBS buffer mixed with 0.05% Triton X-100, pH = 7.4) was prepared prior to the dilution of geniposide. Afterwards, TLR4-immobilized chip was treated with a running buffer containing the geniposide with different concentrations (0-12.8 μM) followed by the regeneration scouting by 10-mM glycine-HCl (pH = 1.7). The injection and dissociation time was set to 3 and 15 min, respectively. The binding affinity of geniposide to TLR4 was analysed by BIAevaluation (GE Healthcare, Sweden).

| Data and statistical analysis
The data and statistical analysis comply with the recommendations of the British Journal of Pharmacology on experimental design and analysis in pharmacology. All the studies were designed to generate groups of equal sample size, using randomization and blinded analysis.
Statistical analysis was undertaken only for studies where the sample size of each group was at least n = 5. The declared group size was the number of independent values, and the statistical analysis was done using these independent values by SPSS 21.0 (SPSS Inc., USA).

| Geniposide suppresses in vivo primary and metastatic tumour growth in a mouse model of orthotopically implanted HCC
We, first of all, identified the anti-HCC potential of geniposide using an orthotopic HCC implantation model (Figure 1a). The model was established by implanting small cubes of subcutaneous-grown HCC onto the right lobe of the liver in nude mice, as indicated in our previous study (Tan et al., 2015). Bioluminescence emitted from luciferasetagged MHCC-97L represented the tumour size, which was weekly

| Geniposide suppressed VEGF transcription by a HIF-1α-independent mechanism
Regarding the outcomes of the luciferase reporter assay, geniposide The correlation degree is represented as −log10 P value. In the right panel, the annotations of GO terms (A-O) are distributed into three principal embranchments in the cluster tree ranging from strong (red portion) to weak correlation (green portion). Note that geniposide significantly attenuates VEGF transcription by HIF-1α-independent mechanism. The anti-VEGF action of geniposide is potentially related to the regulation of TLR4 and STAT3 signalling pathways in hepatocellular carcinoma cells. All data are indicated as mean ± SD of five independent experiments with at least three replicates. * P < 0.05 significantly different from the control group; n.s. No significant difference versus control clustered in the heatmap (Figure 3h). To gain further details of the geniposide-regulated molecules involved in the suppression of HCC, integrative bioinformatics analysis was applied for target validation, as described in our recent study  be the candidate nuclear target. As an essential transcription factor located in the nucleus and known to regulate cell survival and angiogenesis, Sp1 is frequently overactivated during tumourigenesis (Liu, Du, Hu, Zhao, & Xia, 2018). The crosstalk between two transcription factors, Sp1 and STAT3, is essential for tumour angiogenesis and metastasis (Huang & Xie, 2012). Surprisingly, the transcription binding sites of Sp1 and STAT are involved in the luciferase-tagged region of the VEGF promoter (−1 to −1,000 bp; Figure 4a; Pages & Pouyssegur, 2005). Therefore, geniposide-induced inactivation of theVEGF promoter may result from the disruption of the binding potential of Sp1 and STAT3 to its transcriptional regions in the VEGF promoter.
In order to demonstrate, experimentally, the target predictions from the bioinformatics study, both immunoblotting and immunofluorescence assays were performed. Suppression of protein levels of 3.5 | Antagonization of the TLR4/MyD88 pathway by TLR4-geniposide interaction suppresses VEGF expression in HCC cells Our bioinformatics study and published reports demonstrated that the TLR4/MyD88 signalling pathway plays a vital role in tumour suppression (Kang, Su, Sun, & Zhang, 2018;Murad, 2014). We, first of all, examined the interaction between geniposide and its predicted anti-HCC targets by in silico molecular docking. Coincidentally, TLR4 exhibited the highest binding potency with geniposide (fill fitness: −3,318.93 kcalÁmol −1 ; 4G: −7.75 kcalÁmol −1 ) by six residues, including ARG53, ARG55, ARG90, ARG132, LYS89 and LYS128 (Figure 5a).
Furthermore, SPR technology was applied to determine experimentally the binding ability of geniposide with TLR4 ( Figure 5a). The Biacore system-based SPR analysis illustrated that geniposide could specifically bind to TLR4 (K D = 6.716e Meanwhile, the exploratory interaction of MD2, also known as lymphocyte antigen 96, with TLR4 was determined by coimmunoprecipitation assay (Figure 5b), which revealed that the TLR4- Quantification of the mitotic index is shown in the right panel. (f) HCC-derived in vivo protein expression of TLR4, Sp1, p-STAT3 and p65 in the HCC mouse model with diversified treatments. (g) Serum VEGF expression in orthotopic HCC mice with three independent treatments, including vehicle, geniposide (30 mgÁkg −1 for 2 days) and geniposide with LPS (3 mgÁkg −1 per single injection), respectively. (h) Representative immunofluorescence graphs of hepatic tumour tissues from HCC-bearing mice with various treatments. Inset images in the right corner are the respectively enlarged fields with merged signals, showing that either CD31 (red) or MyD88 (green) is overlapped with DAPI staining (blue). Note that targeting HCC angiogenesis by geniposide (30 mgÁkg −1 for 2 days) is associated with the direct shutdown of the TLR4/MyD88-dependent pathway, which was reversed by the additional administration of LPS (3 mgÁkg −1 in a single injection). These results demonstrated that geniposide can significantly repress HCC proliferation, angiogenesis and pulmonary metastasis by down-regulating the TLR4/MyD88 signalling pathway. All data indicated are means ± SD of five independent experiments with at least three replicates. * P < 0.05 significantly different from the control group; # P < 0.05 significantly different from the geniposide group. Scale bar, 30 μm for all images transcripts and secretion was neutralized by the addition of LPS (Figure 5f,g). In further experiments, the anti-angiogenesis effect of geniposide in HCC was confirmed by endothelial cell tube formation assays (Figure 5h). The skeletonized tube-like structures were abundant in cultures of HUVECs incubated in PLC/PRF/5 cell culture medium after geniposide treatment rather than that in geniposide-LPS or geniposide-re-VEGF (20 ngÁml −1 ) co-treatment ( Figure 5h).
Besides, the migration of HUVECs in PLC/PRF/5-cultured medium was significantly reduced by geniposide administration, but its benefits were abolished when co-treated with 20 ngÁml −1 of recombinant VEGF (Figure 5i). These findings suggested that geniposide suppressed the TLR4/MyD88 pathway leading to the inhibition of STAT3/Sp1-dependent VEGF production in HCC cells.

| DISCUSSION
Using the orthotopic mouse model, geniposide was shown to suppress growth of both primary and lung metastases of HCC. Interestingly, the in vitro experiments indicated that the anti-HCC effects of geniposide in HCC mice were expressed without cytotoxicity. As VEGF is one of the critical factors that facilitate tumour angiogenesis, the inhibition of VEGF production, induced by geniposide, blocked the downstream effector signalling of VEGF and led to decrease in F I G U R E 7 Diagram of the proposed mechanisms underlying the suppressive effects of geniposide on angiogenesis in HCC HCC growth. As an anti-VEGF strategy, the VEGF monoclonal antibody is clinically effective in cancer treatment. Our study offers a potential and affordable anti-VEGF, low MW, compound as an alternative approach to the treatment of HCC.
Although HIF-1α-triggered VEGF activation is a principal mechanism leading to angiogenesis in HCC, the inhibition of VEGF production in geniposide-treated HCC was not accompanied by changes in HIF-1α. This finding suggested that VEGF expression can be regulated independently of HIF-1α. Notably, previous evidence indicated that, in the context of HCC growth, the Sp1-related pathways can lead to the overexpression of VEGF, without altering HIF-1α activity, because significant VEGF production has been shown in siHIF-1α-transfected Hep3B cells (an HCC cell line) under normoxic or hypoxic condition and neutralized after Sp1 silencing (Choi, Park, Song, & Choi, 2011).
Both Sp1 and STAT3 are important transcription factors in the nucleus (Sp1) and cytoplasm (STAT3), respectively (Huang & Xie, 2012). In this study, we noted that geniposide-induced inhibition of the Sp1/STAT3 pathway could suppress VEGF in HCC cells, HIF-1α-independently of HIF-1α.. Re-expression of Sp1 or STAT3, or both, significantly reactivated the expression of VEGF in geniposide-treated HCC cells. Both factors can cooperatively promote VEGF-dependent tumour angiogenesis (Santra, Santra, Zhang, & Chopp, 2008). As a zinc finger transcription factor, Sp1 is an inducer of tumour angiogenesis by directly binding to the VEGF promoter (Wu et al., 2013). Moreover, the transcriptional activity of the VEGF promoter in glioblastoma cells can be synergistically activated by STAT3 and Sp1 (Loeffler, Fayard, Weis, & Weissenberger, 2005). Simultaneous inhibition of Sp1 and STAT3 was reported to arrest the invasive and metastatic ability of HCC by down-regulating VEGF expression (Zou et al., 2015). The data from our study confirmed the critical role of Sp1/STAT3 in VEGFmediated angiogenesis in HCC.
TLR4, a member of the toll-like receptor family, is involved in the responses of the innate and adaptive immune systems (Arias et al., 2018). However, TLR4 is not confined to immune cells but is found in numerous cancer cells and evokes pathological processes, including angiogenesis-driven HCC promotion (Zhe et al., 2016). LPS can particularly bind to TLR4 and recruit the adaptor molecule MyD88 to initiate TLR4/MyD88 signal transduction, which consequently activates a series of downstream intracellular adaptors, kinases and transcription factors (Lu, Xu, Chen, Zhou, & Lin, 2017). As an agonist of TLR4, administration of LPS (3 mgÁkg −1 per single injection) in HCC mice only aimed to reverse the geniposide-induced inhibition of TLR4, instead of endotoxin-triggered HCC progression in this study (Kitamura et al., 2002). Previous studies demonstrated that TLR4/ MyD88 activation might up-regulate STAT3, which accelerates proliferation, metastasis and multidrug resistance of HCC (Kang et al., 2018;F. Wang et al., 2019). Additionally, the activation of TLR4/MyD88 results in Sp1 accumulation in cancer progression (Dong et al., 2018). In our study, targeting TLR4/MyD88 could regulate both STAT3 and Sp1 in HCC growth, which could be blocked by geniposide treatment. in vitro and in vivo reactivation of the TLR4/ MyD88 pathway by LPS in geniposide-treated HCC did restore the tumour angiogenesis along with an increase of STAT3/Sp1-dependent VEGF production, suggesting that modulation of STAT3/Sp1 in the TLR4/MyD88 pathway was critical to VEGF-dependent HCC angiogenesis ( Figure 7). Our study offers a novel mechanism to achieve VEGF inhibition in the treatment of HCC.
In conclusion, our study has demonstrated in vivo and in vitro activity, and the underlying mechanism, of geniposide against HCC.
Geniposide exhibited potent inhibition of the proliferation, invasion, angiogenesis and lung metastasis of HCC. Suppression of HCC induced by geniposide was not associated with direct cytotoxicity in tumour cells but was related to a decrease of VEGF expression and HCC angiogenesis. Geniposide significantly blocked the transcription and production of VEGF from HCC cells. The inhibitory effect of geniposide on VEGF expression was independent of regulation by the HIF-1α-related pathway, but relied on suppression of Sp1 and STAT3 in HCC cells. Re-expression of Sp1 and STAT3 could restore VEGF production in geniposide-treated HCC cells. Further analysis showed that geniposide bound to TLR4 and repressed the TLR4/ MyD88-associated Sp1/STAT3 transcription activity. Re-activation of the TLR4/MyD88 pathway by LPS, in vitro, restored the Sp1/STAT3 activities and VEGF production. Thus, geniposide may be a promising candidate for clinical use for HCC. Taken together, our study demonstrated that geniposide could be an inhibitor of angiogenesis in HCCby directly targeting the TLR4/MyD88-regulated STAT3/Sp1 pathway followed by the suppression of VEGF transcription in a HIF-1α independent manner, which provides a novel mechanism for VEGF inhibition in HCC management.
research as stated in the BJP guidelines for Design & Analysis, Immunoblotting and Immunochemistry, and Animal Experimentation, and as recommended by funding agencies, publishers and other organizations engaged with supporting research.