Targeting RORγ inhibits the growth and metastasis of hepatocellular carcinoma

Abstract

Approximately 80%–90% of hepatocellular carcinomas (HCC) occur in a premalignant environment of fibrosis and abnormal extracellular matrix (ECM), highlighting an essential role of ECM in the tumorigenesis and progress of HCC. However, the determinants of ECM in HCC are poorly defined. Here, we show that nuclear receptor RORγ is highly expressed and amplified in HCC tumors. RORγ functions as an essential activator of the matrisome program via directly driving the expression of major ECM genes in HCC cells. Elevated RORγ increases fibronectin-1 deposition, cell-matrix adhesion, and collagen production, creating a favorable microenvironment to boost liver cancer metastasis. Moreover, RORγ antagonists effectively inhibit tumor growth and metastasis in multiple HCC xenografts and immune-intact models, and they effectively sensitize HCC tumors to sorafenib therapy in mice. Notably, elevated RORγ expression is associated with ECM remodeling and metastasis in patients with HCC. Taken together, we identify RORγ as a key player of ECM remodeling in HCC and as an attractive therapeutic target for advanced HCC.

Graphical abstract

Introduction

As the sixth most prevalent (4.7%) and the third deadliest (8.3%) malignant tumor in the world, liver cancer poses a continuous threat to human health.¹ Hepatocellular carcinoma (HCC) is the most common histological type of liver cancer and accounts for ∼90% of all cases.² The incidence of liver cancer continues to increase due to alcohol abuse, HBV and HCV infections, and the worsening of NAFLD.³ The high mortality rate of HCC is mainly due to the fact that patients are frequently diagnosed at an advanced stage with high incidence of tumor metastasis. Even among patients with early stage HCC, up to 70% suffer from tumor recurrence with intra- or extra-hepatic metastases after hepatectomy.⁴ Unfortunately, treatment options for advanced liver cancer are very limited in the clinic. Sorafenib, a multi-kinase inhibitor, has been the main choice to treat advanced HCC for a decade. However, it is often accompanied by a rapid development of drug resistance. In recent years, the clinical situation of HCC has improved with the regulatory approval of regorafenib, lenvatinib, and cabozantinib, and the breakthrough discovery of atezolizumab plus bevacizumab combination therapy.⁵ Nevertheless, advanced HCC remains largely incurable due to low response rates and therapeutic resistance, which makes the development of novel and effective therapeutics an urgent medical need.

Based on clinical statistics, almost 80%–90% of HCCs occur in the context of fibrosis and cirrhosis,⁶ predicting an essential role of fibrosis in the premalignant environment of the liver. The fibrotic environment in HCC, which can also be termed as abnormal extracellular matrix (ECM), is primarily driven by the excessive accumulation of collagen and other matrix components originating from activated hepatic stellate cells (aHSCs) and cancer-associated fibroblasts (CAFs).⁷ ECM is the non-cellular component of tissue, which is composed of collagens, glycoproteins, proteoglycans, and ECM-associated proteins, playing a crucial role in structural support and signal transduction.⁸ Even though stromal cells produce most of the bulk ECM, Tian et al. discovered that tumor cell-derived ECM proteins were most strongly correlated with poor prognosis in patients with pancreatic ductal adenocarcinoma, with no consistent correlation between stromal cell-derived ECM proteins and patient prognosis.⁹ In HCC, apart from the well-known contributions of CAFs and aHSCs to the tumor stroma, accumulated literature over recent years have revealed that the regulation of ECM remodeling by tumor cells, including matrix stiffness, cell adhesion, vasculogenic mimicry, drug delivery, and immunosuppressive microenvironment, strongly facilitates HCC tumorigenesis and metastasis.^{10,11,12,13,14} However, there is no well-established therapeutic strategy for targeting ECM. Therefore, identifying new key targets in ECM becomes the current focus.

RAR-related orphan receptor gamma (RORγ) (encoded by RORC) and RORα, RORβ constitute a family of ROR nuclear receptors.¹⁵ They regulate gene expression by directly binding to specific sequences (ROR response elements) on target genes, and perform a variety of physiological functions. RORγt, a specific isoform of RORγ, is extremely abundant in immune organs such as the thymus.¹⁶ RORγt plays an essential role in driving T helper 17 (Th17) differentiation and thus represents an attractive therapeutic target for autoimmune diseases.^17,18 Multiple inhibitors targeting RORγ are undergoing clinical trials for immune therapy. However, the mechanism and function of RORγ in other diseases such as cancer remains largely unknown. Recently, we and others found that RORγ plays crucial roles in tumorigenesis of prostate cancer,^19,20 breast cancer,²¹ and pancreatic cancer,²² respectively, via androgen receptor signaling, cholesterol biosynthesis pathway, and cancer stem cells.

Here, we found that RORγ is highly overexpressed in tumors from patients with HCC and associated with cancer progression. RORγ functions as a key determinant of ECM-associated gene transcription. RORγ antagonists effectively inhibit HCC tumor growth and metastasis. Moreover, RORγ antagonists strongly sensitize HCC tumors to sorafenib therapy. Taken together, our study uncovers an essential role of RORγ in promoting HCC tumor growth and metastasis, and provides a novel therapeutic target for HCC treatment.

Results

RORγ is overexpressed in liver cancer tumors and predicts poor prognosis

To investigate the potential functions of RORγ (encoded by RORC) in cancer progression, we first examined the expression levels of RORγ in various cancer types. Analysis of TCGA database showed that the mRNA expression level of RORγ is the highest in liver cancer (Figure S1A). Besides, gene amplification and high level of RORγ occurs in 20% of liver tumors (Figure 1A), far exceeding that of other family members RORα and RORβ. And RORγ protein levels were significantly higher among liver tumors than normal tissues (Figure 1B). To better define the expression characteristics of RORγ in patients with HCC, we performed single-cell transcriptomes analysis of HCC patients with different tumor node metastasis stages (GSE149614). As HCC originates from hepatic epithelial cells, we classified all single cells into distinct clusters including epithelial cells according to their genomic characteristics (Figures 1C and S1B). Among various cell types, RORγ was highly expressed in epithelial cells and T cells (Figures 1D and S1C). We found that RORγ was markedly upregulated in tumors compared with normal tissues, which was primarily attributed to epithelial cells (Figures 1E, 1F, and S1D). To verify the correlation between RORγ and HCC progression, we performed pseudo-time analysis of transcriptional changes in RORC with various cells derived from four relevant tissues (non-tumor liver, primary tumor, metastatic lymph node, and portal vein tumor thrombus). This resulted in a cell trajectory with two fates and three monocle states (Figure 1G). The malignancy of the cells deepened over pseudo-time with stage 1–3 (Figure 1H). In parallel, a gradual increase in the expression of RORC also occurred. In line with this, immunohistochemical assays of HCC patients also verified that RORγ expression level was increased with the aggravation of tumor stages (Figure 1I). Moreover, analysis of published tumor datasets showed that higher tumor RORγ mRNA level is significantly associated with poor survival of patients with liver cancer (Figures 1J and S1E). Together, these results suggest that RORγ is overexpressed in liver cancer and associated with tumor stages.

Figure 1.

RORγ is overexpressed in liver cancer and predicts poor prognosis

(A) Gene amplification profile of ROR family in HCC obtained by cBioPortal platform. (B) The protein expression of RORγ was higher among liver tumors than normal tissues in the CPTAC database. (C–H) Single-cell sequencing analyses of 10 HCC patients with different degrees of metastasis (GSE149614). The single cells were classified into distinct clusters according to their genomic characteristics (C). The mRNA levels of RORγ in various cell types in all tissue samples (D) and or normal and tumor samples (E). The expression level of RORγ in epithelial cells in tumors was higher than that in normal tissues (F). Pseudo-time analysis of transcriptional changes in RORC with various cells in four relevant tissues (N, non-tumor liver; T, primary tumor; L, metastatic lymph node; P, portal vein tumor thrombus) (G and H). (I) IHC staining of RORγ expression in liver tumors of patients with different stages. (J) Kaplan-Meier survival plots of RORγ. HCC Patients were grouped by the trichotomization of the RORγ expression (Q1 vs. Q4, lower quartile vs. upper quartile, GSE14520). Scale bars, 200 μm. All data shown as the mean ± SD, ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001; n = 3, Student’s t test.

RORγ is required for HCC cell survival both in vitro and in vivo

To further validate the significance of RORγ on the survival of liver cancer, we used siRNA and shRNA approaches to specially silence RORγ expression in HCC cell lines HepG2 and Huh7. The knockdown of RORC resulted in poor survival of HCC cells as measured by growth and colony formation assay (Figures 2A, 2C, S1A, and S1B). RORγ knockout also potently inhibited HCC cell growth (Figure S2C). In addition, RORγ knockdown significantly induced apoptosis as indicated by an increased level of apoptotic protein markers, cleaved PARP1 and cleaved caspase-7 (Figure 2E). Meanwhile, we tested the effects of two RORγ antagonists GSK805 and XY018 in HCC cell lines. Consistent with the effects of RORγ knockdown, both antagonists potently inhibited the proliferation and induced apoptosis in HCC cells (Figures 2B–2D, 2F, S2D, and S2E). Since 3D organoids may better preserve the properties of human tumors, we constructed a human-derived organoid (PDO) model and culture system for HCC. Representative pictures and cell viability results indicated that both antagonists also notably inhibited the growth of HCC PDX organoids (Figures 2G and 2H). To investigate whether the RORγ antagonists have the same inhibitory effect on tumor growth in vivo, we generated xenograft tumor models by implanting MHCC-97H cells in male athymic nude mice. GSK805 (10 or 20 mg/kg) or vehicle was intraperitoneally injected into tumor-bearing mice. The results showed that GSK805 could effectively inhibit the growth of subcutaneous tumors and did not cause any apparent toxicity to major organs or loss of body weight in mice (Figures 2I–2K, S2F, and S2G). Immunohistochemical staining of the tumor sections showed that GSK805 considerably repressed cancer cell proliferation (Ki67) and induced apoptosis (cleaved caspase-3) (Figure 2L), consistent with in vitro observations. Together, these results suggest that RORγ is a crucial determinant of HCC cell survival both in vitro and in vivo.

Figure 2.

RORγ is required for HCC cell survival both in vitro and in vivo

(A–F) HCC cells were transfected/infected with control or RORC siRNA/shRNA, or treated with DMSO or GSK805/XY018 as indicated. At certain time points, live cells (A and B) or colonies (C and D) were counted, and total cells were collected for western blot assay (E and F). (G and H) PDX-derived organoids were treated with DMSO or GSK805/XY018 (10 μM) for 4 days. Representative photographs were taken after fluorescent staining (G) and cell viability in PDO was assayed by Cell-Titer-Glo (H). Scale bars, 200 μm. (I–L) Male nude mice bearing MHCC-97H subcutaneous xenografts received GSK805 (i.p., 10 and 20 mg/kg) or vehicle 5 days per week for 20 days. Tumor volume was measured every 4 days (n = 6) (J). Representative photos of the tumor were snapped (I) and the tumor weight was measured (K) on the last day. IHC staining of Ki67 and cleaved caspase-3 in the randomly selected tumor tissues (L). Scale bar, 200 μm. All data shown as the mean ± SD, ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001; n = 3, Student’s t test.

RORγ inhibition shifts the HCC matrix

Given that the mechanism of RORγ in HCC is poorly understood, we performed bulk RNA-seq of HepG2 cells treated with GSK805 to identify the transcription programs that were affected by RORγ inhibition (Figure 3A). Hallmark gene set analysis and canonical pathway analysis revealed majority of the downregulated genes were associated with matrisome, hypoxia, and cholesterol homeostasis signaling, and the upregulated genes were involved in RNA polymerase II and TP53 transcription pathways (Figure 3B). It is quite evident that the matrisome pathway was the most highly enriched among all the pathways. Further corroboration by GSEA also manifested that the matrisome pathway was significantly suppressed by GSK805 in HepG2 (Figure 3C). Considering that the abundant fibrotic matrix environment is crucial for the oncogenesis and progression of HCC, we speculate that RORγ antagonists may attenuate the development of HCC via inhibiting matrisome signaling. To verify the outcomes of RNA-seq, we conducted qRT-PCR in HepG2 treated with GSK805/XY018 or RORC siRNA. The results showed a marked decrease in the expression of genes across some matrix components including COL9A3 (collagen type IX alpha 3 chain), FN1 (fibronectin 1), and SPARC (secreted protein acidic and rich in cysteine) with RORγ inhibition, as well as TGF-β1 (transforming growth factor β 1), MMP15 (matrix metalloproteinase 15), LOX (lysyl oxidase), PLOD2 (procollagen-lysine, 2-oxoglutarate 5-dioxygenase 2), and S100A4 (S100 calcium-binding protein A4; Figures 3D and 3E). While RORγ overexpression significantly increased the expression of major ECM-related genes including LOX, PLOD2, MMP15, TGFB1, and SPARC in liver cancer cells (Figure S3A). Further examination by western blot showed consistent results (Figures 3F and S3B‒S3D). To understand whether RORγ directly regulates the transcription of ECM-related genes, we question our previous RORγ chromatin immunoprecipitation (ChIP-seq) data performed in cancer cells and found that RORγ specifically occupied the promoter or enhancer region of LOX, COL9A3, TGF-β1, and MMP15 (Figure 3G). Our subsequent ChIP-qPCR experiment validated that RORγ indeed binds to LOX, COL9A3, TGF-β1, and MMP15 target loci and the RORγ occupancy at their promoter site was significantly decreased in cells treated with either GSK805 or siRNA knockdown (Figures 3H and S3E). We also performed ChIP-qPCR of H3K27ac, which is a marker of transcriptional activation.²³ The results which were in accordance with RORγ ChIP-qPCR revealed that the binding of H3K27ac to the promoter of ECM-related genes was strongly reduced by RORγ inhibition (Figures 3I and S3E). RORγ activates gene transcription through association with nuclear receptor coactivator (SRC) family members. Previous studies suggest that SRC-1 promote HCC progression.^24,25 Our coIP assay showed that RORγ was associated with SRC-1 in HCC cells, suggesting natural association between these two proteins (Figure S3F). Furthermore, RORγ antagonists strongly reduced SRC-1 occupancy at RORγ binding sites (Figure S3G). Knockdown of SRC-1 decreased the expression of ECM-related genes (Figure S3H). Collectively, these results indicate that RORγ markedly alter the matrisome pathway and directly regulate ECM-associated genes transcription.

Figure 3.

RORγ inhibition shifts the HCC matrix

(A–C) Analysis results of RNA-seq in HepG2 cells treated with DMSO or GSK805 (10 μM) for 48 h. Volcano plot reflecting gene expression alterations after GSK805 treatment (A). Hallmark and canonical enrichment analysis of downregulated (green) or upregulated (purple) genes among GSK805-treated cells determined representative pathways (B). GSEA analysis of the matrisome pathway (C). (D and E) qRT-PCR assay of critical genes in the matrisome pathway among cells with GSK805/XY018 (10 μM) treated (D) or RORC siRNA transfected (E). (F) Western blot of indicated ECM proteins in HCC cells treated with GSK805/XY018 (10 μM) for 48 h. (G) ChIP-seq signal visualization of RORγ-binding events on promoters or enhancers of LOX, COL9A3, TGFB1, and MMP15 genes in breast cancer cells as reported previously. (H and I) ChIP-qPCR assay of relative enrichment of RORγ (H) or H3K27ac (I) at the promoter of indicated genes in HepG2 treated with GSK805 (10 μM) for 48 h. Fold change refers to the indicated enrichment on this gene under the interference of GSK805 vs. the IgG enrichment in cells treated with vehicle control set as 1. All data shown as the mean ± SD, ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001; n = 3, Student’s t test.

RORγ is a major regulator of ECM remodeling and cell adhesion

Having observed that the pharmacological or genetic suppression of RORγ among HCC cells resulted in decreased expression of major ECM-related genes, we next investigated whether RORγ could remodel the ECM. We first examined the level of collagen, the most abundant component of the ECM. Collagen labeling assay showed that intracellular collagen levels were significantly reduced after either RORγ antagonist treatment (Figures 4A and 4B). Moreover, a significant decrease in extracellular collagen crosslinking and thickness occurred in subcutaneous tumors treated with GSK805 (Figure 4C). In line with collagen assay results, the abundance of FN1 was considerably lower among GSK805 tumors in relation to vehicle (Figures 4D and 4E). Given that RORγ inhibitors reduced ECM macromolecules (collagen and FN1) deposition, we further asked whether RORγ would interfere with cell-matrix adhesion. As expected, we observed a reduction in cell attachment among GSK805-treated cells (Figure 4F). Cell adhesion is a process led by cell adhesion molecules, which form a bridge between the ECM and the actin cytoskeleton, thereby mediating strong adhesion and bi-directional signaling.²⁶ We speculated that the reduction in cell adhesion capacity may not only be due to ECM remodeling but also related to cell adhesion molecules or cytoskeletal alterations caused by RORγ. Results showed that RORγ inhibitors significantly reduced the mRNA levels of a variety of cell adhesion molecules, such as intercellular adhesion molecule family members ICAM2 and ICAM4, integrin family members ITGA7 and ITGB2, and Ca²⁺-dependent cell adhesion molecule family members CDH16, PCDHA7, PCDHB16, and JAM3 (junctional adhesion molecule 3) (Figure 4G). Consistently, the level of actin cytoskeleton labeled by phalloidin was also visibly decreased in RORγ inhibitor-treated cells (Figures 4H and 4I). Intriguingly, the actin cytoskeleton expression was enhanced with increasing levels of RORγ (Figures 4J and 4K). Overall, these results suggest that RORγ mediates alterations in ECM deposition, crosslinking and cell adhesion potential, establishing a favorable microenvironment to facilitate HCC metastasis.

Figure 4.

RORγ is a major regulator of ECM remodeling and cell adhesion

(A and B) Representative images (A) and quantification (B) of collagen detected using CNA35-mCherry probe in Huh7 treated with GSK805/XY108 (5 μM) for 48 h. Scale bars, 20 μm. (C) The collagen level in mouse tumor tissues was examined by Masson staining. Scale bars, 200 μm. (D and E) Representative photos (D) and quantification (E) of FN1 immunostaining in MHCC-97H subcutaneous tumor. Scale bar, 50 μm. (F) Adhesion assay of HepG2 and HCC-LM3 treated with the indicated concentrations of GSK805. The adherent rate was calculated using control wells as 100%. (G) qRT-PCR analysis of vital adhesion factors in HepG2. (H–K) Phalloidin staining (H and J) and quantification (J and K) in HepG2 treated with GSK805/XY108 (5 μM) (H and J) or RORγ overexpressed-HepG2 cells (J and K). Scale bar, 20 μm. All data shown as the mean ± SD, ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001; n = 3, Student’s t test.

RORγ antagonists block HCC metastasis

It is broadly recognized that the linearized and thicker collagen could increase the migration rate of tumor cells and thus facilitate tumor metastasis.^27,28,29 As RORγ antagonists could inhibit collagen crosslinking and deposition, the question arises whether they could block HCC cell metastasis. As wound healing assay and transwell experiments demonstrated that RORγ antagonists effectively curbed HCC cell migration and invasion in vitro (Figures 5A, 5B, and S4A). As for in vivo tumorigenesis, we constructed HCC intraperitoneal or pulmonary metastasis models by orthotopic liver implantation or tail vein injection of mouse liver cancer cells H22. Bioluminescence imaging showed that GSK805 significantly inhibited tumor peritoneal metastasis at day 8 and malignant ascites formation at day 15 after orthotopic liver injection H22 cells (Figures 5C and S4B). Moreover, GSK805 could also markedly inhibits the progression of H22 tumor in liver and lung metastasis sites (Figures 5D–5G and S4C). Apparently, fewer intrapulmonary metastatic tumor nodules occurred in the GSK805-treated mouse compared with vehicle (Figure 5G). To determine whether the inhibition of HCC metastasis by RORγ antagonists was relevant to the tumor matrix, the protein levels of RORγ and ECM core components (collagen, FN1) in mice orthotopic tumors and pulmonary metastatic tumor were analyzed. The staining results showed that both RORγ, collagen and FN1 were obviously increased in pulmonary metastatic HCC tissue, and all of them were significantly reduced in tumors of GSK805-treated mice (Figure 5H).

Figure 5.

RORγ antagonists block HCC metastasis primarily by repressing ECM deposition

(A and B) Representative photos and quantitative analysis of wound healing experiment (A) and transwell invasion assay (B). HCC cells were treated with DMSO or GSK805/XY018 as indicated for 48 h. (C) H22-EGFP cells were injected into the left liver lobes of BALB/c mice, and then received GSK805 (i.p., 5 mg/kg) or vehicle 5 days per week for 15 days (n = 7). Tumor growth was monitored by bioluminescence imaging. (D–H) Well-vitalized H22-EGFP cells were injected into the tail vein of BALB/c mice to establish pulmonary metastasis models (n = 7). Administration treatment was the same as (D). Bioluminescent images of in vivo tumor growth status (E, left) and representative photos of tumors in lung and liver sites on the last day (E, right). Luciferase fluorescence signal intensity of lungs from mice treated with vehicle or GSK805 (F). H&E staining and IHC assay of lung and liver with tumors (G and H). Scale bars, 200 μm. All data shown as the mean ± SD, ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001; n = 3, Student’s t test.

To determine whether the above results were the same in patients with HCC, levels of RORγ, collagen and FN1 were assayed in primary and metastatic tumors of HCC patients. Compared with the primary HCC, the expression level of RORγ was elevated significantly in metastatic HCC (Figures 6A and 6B). Consistently, collagen crosslinking and FN1 deposition were also increased in metastatic HCC. Pearson correlation analysis indicated a positive correlation between the levels of RORγ and collagen or FN1 (Figure 6C). Moreover, data analysis of a public single-cell RNA sequencing (RNA-seq) result further validated the correlation between RORγ and ECM genes in primary and lymphatic metastatic tumors of HCC patients (Figures 6D, 6E, S4D, and S4E). RORγ expression were elevated with the malignant differentiation of cells and exhibited the highest levels in lymphatic metastatic tumors (Figures 6F–6H). In all, the above results suggested that RORγ may facilitate HCC metastasis by regulating ECM remodeling.

Figure 6.

RORγ correlates with ECM core proteins in primary and metastatic HCC

(A and B) Representative images (A) and quantitative analysis (B) of immunofluorescence staining in HCC patient tumors. Primary HCC, n = 20; metastatic HCC, n = 11. Scale bar, 50 μm. (C) Pearson’s correlation analysis of RORγ with collagen or FN1 in tumors from HCC patients. (D–H) Single-cell sequencing analyses of one HCC patient with lymph metastasis (GSE149614). Cell co-localization expressions of RORγ and ECM gene epithelial cells derived from normal liver, primary, and lymph metastatic tumor tissues (D and E). The cytoTRACE analysis (F and G) and the expression level of RORγ (H) of epithelial cells in three types of tissues. All data shown as the mean ± SD, ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001; n = 3, Student’s t test.

RORγ antagonists sensitize HCC tumors to sorafenib therapy

Beyond changes in ECM deposition, ECM-associated angiogenesis inhibited by RORγ antagonists may also be a suppressor of HCC progression. In addition to the major components (FN1) and crosslinking proteins (LOX, SPARC) of the vascular basement membrane, other ECM-associated proteins and growth factors were also contributors to tumor angiogenesis, such as MMP3 (matrix metalloproteinase 3), MMP19 (matrix metalloproteinase 19), SERPINE1 (serine protease inhibitor clade E member 1, PAI-1), SERPINH1 (serpin protease inhibitor clade H member 1), IGFBP1 (insulin-like growth factor binding protein 1), IGFBP2 (insulin-like growth factor binding protein 2), CCN1 (cellular communication network factor 1), CCN2 (cellular communication network factor 2), CCL20 (C-C motif chemokine ligand 20), TGF-β1, and S100A4. The qRT-PCR assay demonstrated that the genes encoding all of these proteins were significantly repressed by RORγ antagonists (Figures 3D and 7A).^30,31,32,33 Besides, the protein level of CD31 (an angiogenesis biomarker) was clearly decreased in primary and metastatic liver tumors treated with GSK805 (Figure 7B). In addition, matrix stiffness, which is intimately related to collagen, could also impact endothelial cell behavior during angiogenesis.^34,35 So, we believe that the positive regulation of HCC development and metastasis by RORγ may be partially through the induction of ECM-associated angiogenesis.

Figure 7.

RORγ antagonists sensitize HCC tumors to sorafenib therapy

(A) qRT-PCR analysis of angiogenesis-related ECM genes in HepG2. (B) IHC staining of CD31 in primary and metastatic liver tumors. Scale bars, 100 μm. (C–G) Normal cells and sorafenib-resistant cells were treated with GSK805 (5 μM) alone, or in combination with sorafenib (1 μM) for certain times. Live cells (C) or colonies (D) were counted. The migration (E), invasion (F), and adhesion (G) assays were statistically analyzed. (H–K) Nude mice bearing MHCC-97H xenografts received GSK805 (i.p., 5 mg/kg) alone, or in combination with sorafenib (i.p., 10 mg/kg) 5 days per week for 24 days (n = 7). Tumor volume was monitored (H) and tumor weight was measured (J) on the last day. IHC staining of Ki67 and cleaved caspase-3, FN1, and collagen expression detected by Masson staining in tumors (K). Scale bar, 200 μm. (L) Schematic diagram of RORγ antagonists regulating ECM signal and exerting inhibitory effects on HCC. All data shown as the mean ± SD, ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001; n = 3, Student’s t test.

Sorafenib is a multi-kinase inhibitor that exhibited anti-angiogenesis effects by targeting vascular endothelial growth factor receptor 2 (VEGFR 2), platelet-derived growth factor receptor (PDGFR-β), and hepatocyte factor receptor (c-KIT).³⁶ RORγ antagonists appeared to be promising adjuncts to sorafenib therapy for HCC as they targeted distinct pro-angiogenic proteins thereby exerting a broader anti-angiogenic effect and inhibiting compensatory angiogenic rebound. Results showed that sorafenib plus GSK805 treatment markedly inhibited the proliferation, migration, invasion, and adhesion of normal HCC cells and sorafenib-resistant HCC cells and exhibited more efficacy than monotherapy (Figures 7C–7G and S5A‒S5C). Correspondingly, a similar inhibition of sorafenib combined with GSK805 treatment in MHCC-97H subcutaneous tumor progression was observed (Figures 7H–7J, S5D, and S5E). Staining assays showed that the combination therapy considerably inhibited cancer cell proliferation (Ki67), the expression of FN1 and collagen deposition (Masson), and induced apoptosis (cleaved caspase-3) (Figure 7K). In sum, RORγ antagonists inhibited ECM-associated angiogenesis and complementarily sensitized HCC to sorafenib therapy.

Discussion

The function of RORγ in control of tumor growth has only recently been recognized. Our work presented here not only demonstrates that RORγ is a key determinant in ECM remodeling but also provides a unique opportunity for an effective therapeutic intervention for HCC. We found that RORγ is highly overexpressed and amplified in HCC tumors. RORγ functions as an essential activator of the matrix program via directly driving the expression of major ECM genes in HCC cells. Through upregulation of FN1 deposition, cell-matrix adhesion and collagen production, RORγ may facilitate HCC progression. Furthermore, RORγ antagonists effectively inhibit tumor growth and metastasis. Combined with the druggability of RORγ, this study establishes RORγ as a previously unsuspected key player of ECM remodeling and an attractive target for HCC.

In HCC, ECM derived from tumor cells plays a critical role in caner progression through the modulation of matrix stiffness, cell adhesion, vasculogenic mimicry, immunosuppressive microenvironment, among others.³⁷ Current strategies targeting ECM focus on the major components of the ECM, such as collagen, fibronectin, and hyaluronic acid. However, targeting only one constituent may cause related adverse effects that in turn affect the effectiveness of cancer therapy. For instance, the degradation of collagen may increase the secretion of growth factors that could stimulate tumor development.³⁸ Our work indicates that RORγ could regulate the production, degradation and crosslinking process of collagen by modulating the expression of collagen, PLOD2, MMPs, and LOX. RORγ antagonists could markedly reduce the thickness and parallel alignment of collagen in tumor environment. RORγ can also regulate the expression of FN1, another ECM major component. Together with the regulation of the growth factor TGF-β and other ECM proteins, RORγ achieves multifaceted effects on the ECM remodeling process in HCC. In addition, RORγ also regulates several key segments of the cell-matrix adhesion process including intracellular cytoskeleton, ECM, and adhesion factors on the cell membrane. Multiple ECM mediators, including those we identified here as RORγ target genes such as LOX, PLOD2, and FN1, play a crucial role in promoting tumor growth, invasion, metastasis, and therapy resistance in various cancers, including HCC.^{39,40,41,42,43,44,45} Peng et al. demonstrated that inhibition of cell-matrix adhesion leads to reduced infiltration of liver tumor cells into the alveoli.¹¹ Likewise, we found that RORγ inhibition can block HCC cells from metastasizing to the abdomen and lungs. Therefore, we speculate that RORγ may modulate the interactions between liver cancer cells and the matrix environment to promote HCC growth and metastasis. However, further studies are needed to reveal the exact contributions of ECM remodeling by RORγ in its overall function in driving HCC.

An essential consideration is that hypoxic signaling and cholesterol homeostasis pathway are also significantly enriched in RNA-seq analysis. It appears that the overall effects of RORγ on HCC is unlikely to be limited to ECM rebuilding. The hypoxic microenvironment, which arises in a variety of rapidly proliferating tumors including HCC, is closely related with ECM. Dang et al. pinpointed that HIF-1α, a pivotal factor in the hypoxia-inducible factor (HIF) family, could activate RORγ transcription and recruit p300 to form a HIF-1α-p300-RORγ complex to co-regulate downstream gene transcription.⁴⁶ Besides, HIFs can affect the post-translational modification, crosslinking, and degradation process of collagen through regulating the expression of MMPs and several collagen-modifying enzymes including PLOD1, PLOD2, and LOX family.^27,47 Above all, we suggest that the hypoxic environment of HCC may be engaged in the regulation of ECM remodeling by RORγ. As for cholesterol homeostasis, in our previous work, RORγ serves as an activator of the cholesterol biosynthesis process in TNBC by dominating the sterol regulatory element binding protein 2 (SREBP2).²¹ Besides, several studies have demonstrated that cholesterol biosynthetic processes support HCC tumorigenesis, growth, and drug resistance, which are associated with aberrant lipogenesis, IL-17, and caspase-3 signaling mediated by SREBP2.^48,49,50 And some others discovered that malignant breast cancer cells depend on cholesterol for matrix degradation.⁵¹ We speculate that the positive contributions of RORγ to HCC may act in part through a cholesterol biosynthetic pathway linked to SREBP2 or ECM remodeling.

In addition to the liver tumor cells, RORγ may play a role in other aspects of the tumor microenvironment. In HCC, aHSCs and CAFs contribute to the bulk of the fibrotic matrix environment. We demonstrated that RORγ antagonists could inhibit the secretion of TGF-β and other cytokines from HCC cells, which may have an impact on other cells in the tumor microenvironment. As Guido and co-workers found that tumor-derived TGF-β promotes the migration of fibroblasts to the tumor microenvironment and facilitates their conversion into CAFs.^52,53 TGF-β could also activate hepatic stellate cells to secrete abundant matrix proteins. Beyond the possible indirect inhibition of CAFs and HSCs, RORγ antagonists may directly suppress HSCs activation by inhibiting the expression of collagen and fibrotic factor CCN2 in HSCs,³⁸ as it does in tumor cells. However, these conjectures need to be validated by subsequent experiments. Furthermore, RORγt, a thymus-specific isoform of RORγ, plays a critical role in driving the differentiation and expansion of type 17 T cells, including T helper type 17 (Th17) and Tc17 cells, and activates them to secrete effector cytokines such as IL-17A, which is thought to be an important factor in alcohol-induced liver cancer.⁴⁹ Overall, inhibition of RORγ can restrain the HCC development in a multifaceted combination. Since multiple RORγ antagonists are undergoing clinical trials for autoimmune diseases, our work warrants further exploration of these drugs in their applications against liver cancer.

Materials and Methods

Cell culture

HepG2 and HEK293T cells were from American Type Culture Collection. Huh7 was obtained from the JCRB cell bank (Tokyo, Japan). Human highly metastatic HCC cell lines MHCC-97H and HCC-LM3 were from the Shanghai Cell Bank of the Chinese Academy of Sciences (Shanghai, China). Murine HCC cell line H22 was provided by Procell Life Science & Technology (Wuhan, China). All cells were cultured with DMEM medium at 37°C in 5% CO₂ incubators except H22, which was cultured in RPMI-1640 medium. All culture media were supplemented with 10% FBS (Gibco), 100 U/mL penicillin, and 100 μg/mL streptomycin (Gibco).

Chemicals and antibodies

GSK805 was obtained from WuXi AppTec (Shanghai, China) and XY018 was synthesized by Yuanxiang Wang Lab (Sun Yat-sen University). Sorafenib (T0093L) was purchased from TargetMol (Shanghai, China). Antibodies against the below proteins were applied with origin and dilution ratio specified: RORγ (1:1,000, Ebioscience, no. 14–6988–82; immunohistochemistry [IHC], 1:50), cleaved caspase-3 (Cell Signaling Technology, no. 9664;, IHC, 1:100), cleaved caspase-7 (1:1,000, Cell Signaling Technology, no. 9491), PARP (1:1,000, Cell Signaling Technology, no. 9532), GAPDH (1:1,000, Cell Signaling Technology, no. 8884), TGF-β (1:1,000, Cell Signaling Technology, no. 3711), PLOD2 (1:1,000, Proteintech, no. 66342-1-Ig), FN1 (1:1,000, Proteintech, no. 15613-1-AP; immunofluorescence [IF], 1:100), COL9A3 (1:1,000, NOVUS, NBP1-68937), LOX (1:500, Santa cruz, sc-373995), V5 (1:2,000, Abcam, ab27671), and Ki67 (ZSGB-BIO, ZM-0166; IHC). H3K27ac (Abcam, ab4729) was used for ChIP-qPCR.

Cell transfection and lentivirus transduction

For siRNA knockdown, cells were transfected with control or RORC siRNA using DharmaFECT (Dharmacon) according to the manufacturer’s specification. For shRNA lentivirus transduction, RORC and control shRNA lentivirus were produced from HEK293T cells by co-transfected with packaging plasmids. And then lentivirus infection was conducted using Lipofectamine 2000 (Invitrogen) as described previously.¹⁹ sgRNA was conducted as described previously.²¹

The siRNA and shRNA sequences targeting RORC are listed below.

siRORC#1: CGAGGATGAGATTGCCCTCTA;

siRORC#2: GCCCTCATATTCCAACAACTT;

shRORC: Forward oligo, 5′-3′ CCGGGCCCTCATATTCCAACAACTTCTCGAGA.

AGTTGTTGGAATATGAGGGCTTTTTG, Reverse oligo, 5′-3′ AATTCAAAAAG.

CCCTCATATTCCAACAACTTCTCGAGAAGTTGTTGGAATATGAGGGC. The sgRNA sequences are as follows: control, GGGCGAGGAGCTGTTCACCG; sgRNA, GATACCCTCACCTACACCTT.

Cell growth and colony formation

For cell growth, 1 × 10⁵ HCC cells per well were seeded in six-well plates. After 3 or 5 days of siRNA transfection or 4 days of drug treatment, all viable cells were counted. For colony formation, 800 cells per well were seeded in six-well plates, treated with different concentrations of compounds, and cultured for 12–15 days. Cells were then washed with PBS and fixed in 4% paraformaldehyde. After crystal violet staining and ddH₂O rinsing, cell colonies were scanned and counted.

Western blotting and qRT-PCR

Proteins from cell lysates were incubated with primary antibody and corresponding secondary antibody and detected by Bio-Rad instrument through chemiluminescence.

For RT-PCR assay, total RNA was extracted from cells first and reverse transcribed into cDNA. The cDNA was mixed with primers and SYBR as directed and then detected by BIO-RAD CFX96 (Bio-Rad). After collecting the fluorescence values, a melting curve analysis was analyzed. The primers for qRT-PCR are listed in Table S1.

Patient-derived organoid culture

The organoid was derived from PDX xenograft in nude mice. Pink tumor masses were dissected and finely chopped, then transferred into a centrifuge tube containing serum-free Ad-DMEM/F-12 medium (Gibco) and 1 mg/mL collagenase IV (Sigma) digested for 40 min at 37°C, 220 rpm. Isolated organoids were mixed with Matrigel (BD Biosciences) and seeded 5 μL per well in a preheated 96-well plate. After incubation for 30 min, a supplemented medium of 100 μL per well was carefully added, and organoids were cultured in a 37°C humidified atmosphere under 5% CO₂. The culture medium contains N-acetylcysteine (1.25 mM), FGF 7 (5 ng/mL), B27 supplement (1×), nicotinamide (5 mM), SB202190 (500 nM), EGF (ng/mL), HEPES (10 mM), FGF 10 (20 ng/mL), A83–01 (500 nM), R-Spondin 1 (250 ng/mL), neuregulin 1 (5 nM), Y-27632 (5 mM), primocin (50 mg/mL), and phenol red-free DMEM/F-12 with glutamine/streptomycin/penicillin (100 mg/mL). After 4 days, organoids were treated with GSK805 (10 μM) or XY018 (10 μM) for another 4 days. Representative photos were obtained using a fluorescence microscope after incubation with live/dead reagents (US Everbright, China), and then cell viability was measured with Cell-Titer Glo reagents (Promega,).

ChIP assay

The ChIP assay was performed as described previously.⁵⁴ In brief, HepG2 and HepG2-RORγ cells were treated with GSK805 (10 μM) for 48 h first. After discarding the media, 1% paraformaldehyde was added to crosslink protein-DNA interactions, followed by quenching with glycine for 5 min. Cells were collected, centrifuged, and resuspended with lysis buffer to lyse cells and isolate nuclei. After the washing steps, nuclei were resuspended in the shearing buffer to shear chromatin in an AFA tube with Covaris E220 according to the manufacturer’s specification. Chromatin fragments were diluted with IP dilution buffer and precipitated by Protein G beads and indicated antibodies. After reverse crosslinking and DNA fragment purifying, real-time PCR analysis was performed. The primers for ChIP-qPCR are listed in Table S2.

IHC, IF, and Masson staining

The IHC assay was performed as described previously.⁵⁵ Paraffin-embedded tissue slides were dewaxed and hydrated by xylene and gradient ethanol, then heated with PH6.0 citric acid buffer (Servicebio, China) or PH8.0 EDTA buffer (ZSGB-BIO, China) for antigen retrieval at 100°C for 20 min. Then tissue sections were endogenously blocked with 3% H₂O₂ for 10 min and normal goat serum for 30 min. After incubating the primary antibody overnight at 4°C, the HRP-conjugated secondary antibody (IHC) of the corresponding species or anti-rabbit IgG Fab2 Alexa Fluor (R) 488 (IF, 1:500 for FN1) was incubated at room temperature for 1 h. For the IHC assay, protein expression was detected by a DAB kit (Beyotime, China), followed by counterstaining with hematoxylin (ZSGB-BIO, China). For the IF assay, tissue sections were mounted with DAPI (Beyotime). For Masson staining, slides were prepared with a Masson’s Trichrome Stain Kit (Solarbio, China). Images were captured by a microscope after dehydration, neutral gum capping, and drying tissue sections.

Phalloidin staining and collagen quantification by CNA35-mCherry

Huh7 cells were cultured for 48 h in the presence of GSK805 (5 μM) or XY018 (5 μM). Cells were fixed with 4% paraformaldehyde for 15 min, then permeabilized with 0.3% Triton X-100 for 15 min and blocked by 5% normal goat serum for 1 h. For phalloidin staining, iFluor 488 phalloidin (1:1,000, Yeasen, China) was added to label the cytoskeleton for 1 h. For collagen staining, cells were incubated with pre-filtered collagen binding protein CNA35-mCherry (1 μM)⁵⁶ for 48 h at 37°C in 5% CO₂ incubators before fixation. Finally, cells were counterstained with DAPI to label the nucleus prior to microscopic observation in both assays. As for tumor staining, tissue sections were dewaxed, hydrated, and antigenically retrieved as in the IHC assay, followed by permeating and blocking. Subsequently, tissue slides were incubated with CNA35-mCherry (1 μM) overnight at 37°C and mounted with DAPI. Images were taken in a Zeiss LSM780 confocal microscope and regions of labeled collagen were quantified by ImageJ software.

Adhesion assay

HepG2 and HCC-LM3 cells treated with GSK805 or XY018 with the indicated concentrations were collected, washed, and then seeded in 96-well plates at 3 × 10⁴ per well in triplicate. After 90 min, the media were discarded, and cells were washed twice with clean medium to remove the non-adhered cells. The number of attached cells was detected by CCK-8 kit (Yeasen, China). Cells were incubated with the CCK-8 reagents at 37°C for 1–2 h. Subsequently, the absorbance (OD) at 450 nm was measured by a microplate reader (Promega). The cell adhesion capacity was calculated as the Adherent ratio (%) = [OD (Compound) – OD (Blank)]/[OD (Vehicle) – OD (Blank)] × 100%.

Migration and invasion assays

The migration of HCC cells was measured by wound healing assay. When the cell density in six-well plates reached 80%, pipette tips were used to scratch the cell monolayer. Subsequently, the culture media were replaced and RORγ antagonists were added. After 48 h, wound images were captured with a microscope, and wound closure rates were quantified by ImageJ software.

For invasion assay, the upper chamber of the transwell chamber (Costar) was pre-coated with 100 μL Matrigel and kept for 1 h at 37°C. Then HCC cells (4 × 10⁴) treated with GSK805 or XY018 were collected and seeded into the upper chamber with 1% FBS media. The bottom chamber was filled with 500 μL DMEM containing 20% FBS. Forty-eight hours later, the upper inserts were fixed with 4% paraformaldehyde and then stained with crystal violet dilution. Images of five random fields were captured under a microscope and the number of invaded cells was counted.

HCC xenograft tumor models and treatments

All procedures were approved by the Animal Experimentation Ethics Committee of Sun Yat-sen University, complying with relevant ethical guidelines (approval nos. SYSU-IACUC-2022-001118 and SYSU-IACUC-2023-000336).

For the subcutaneous xenograft model, 4-week-old male BALB/c-nu/nu mice were used. MHCC-97H cells (3 × 10⁶) were suspended in 100 μL cold PBS and subperiosteally inoculated into both sides flank of the mice. When the tumor size was approximately 50 mm³, mice were randomized into groups and treated with the indicated concentrations of GSK805 or sorafenib intraperitoneally five times weekly. Tumor size was monitored by caliper twice per week and tumor volume was calculated with the equation: π/6 × (length × width²).

For the orthotopic liver implantation model, 6-week-old male BALB/c mice were used. H22 cells (5 × 10⁵) expressing pLenti-Firefly Luciferase-EGFP (H22-EGFP cells) were suspended in 25 μL cold PBS and injected into the left liver lobes. The day after surgery, mice were randomly grouped and treated with GSK805 (5 mg/kg) intraperitoneally five times weekly. The in vivo tumor growth was detected by bioluminescence imaging using an IVIS Spectrum (PerkinElmer).

For the pulmonary metastasis model, 6-week-old male BALB/c mice were injected intraperitoneally with 7 × 10⁶ H22-EGFP cells resuspended in 100 μL cold PBS. One week later, ascites formed and the mice were sacrificed. Under an aseptic environment, mouse ascitic fluid cells were extracted and washed twice with cold PBS. Ascitic fluid cells (1.5 × 10⁶) were suspended in 100 μL cold PBS and injected into the tail vein of BALB/c mice. Tthe next day, mice were randomly grouped and treated with GSK805 (5 mg/kg) intraperitoneally five times weekly. The tumor growth was monitored by bioluminescence imaging as above.

At the end of the experiment, mice were sacrificed and tumors were collected for subsequent assays.

Patients and specimens

The liver tumor of patients was obtained from NanFang Hospital of Southern Medical University.⁵⁵ The paraffin-embedded specimens of primary and metastatic liver tumors were from The First People’s Hospital of Foshan. All specimens were confirmed by pathological examination and approved by the institutional review board of the hospital.

Total RNA-seq

HepG2 cells were exposed to GSK805 (10 M) for 48 h before total RNA was collected. The RNA samples were given to BGI Tech (China), which used an MGISEQ2000 SE50 machine to create the sequence libraries. Gene set enrichment analysis (GSEA) (v.4.1.0) was used to score genes based on the shrunken limma log2 fold changes. Enrichment analysis of Hallmark and canonical pathways for genes with expression change ≥1.5-fold (increase or decrease) was performed with GSEA v.4.1.0, and p adjust (FDR) < 0.05 was regarded as statistically significant.

Bioinformatics analysis

The single-cell RNA-seq data (GSE149614) of HCC were downloaded from the Gene Expression Omnibus (GEO) database. Analysis of single-cell RNA-seq data was performed using R (v.3.4) with the edgeR package. The analysis of RORC mRNA levels in various tumors was from the public database cBioPortal for Cancer Genomics. The data on RORC protein expression in HCC were obtained from the Office of Cancer Clinical Proteomics Research (CPTAC) database. Volcano plot analysis was conducted by Sangerbox 3.0,⁵⁷ an online bioinformatics analysis platform.

Statistical analysis

Results were presented as mean ± SD by GraphPad Prism 8.0 from three independent experiments. Two-tailed Student’s t test was used to compare the difference of means between indicated groups. A p value <0.05 was considered to be statistically significant.

Data and code availability

All associated data supporting this study are available upon request from the corresponding authors. Gene expression data from RNA-seq analysis were deposited in the NCBI GEO database (GSE234660).