Hedgehog signalling mediates drug resistance through targeting TAP1 in hepatocellular carcinoma

Abstract Multidrug resistance is one of the reasons for low survival of advanced hepatocellular carcinoma (HCC). Our previous studies indicate that the hedgehog signalling is involved in hepatic carcinogenesis, metastasis and chemo‐resistance. The present study aims to uncover molecular mechanisms underlying hepatoma chemo‐resistance. TAP1 and GLI1/2 gene expression was assessed in both poorly differentiated hepatoma cells and HCC specimens. Potential GLI‐binding site in the TAP1 promoter sequence was validated by molecular assays. Approximately 75% HCC specimens exhibited an elevated expression of hedgehog GLI1 transcription factor compared with adjacent liver tissue. Both GLI1/2 and TAP1 protein levels were significantly elevated in poorly differentiated hepatoma cells. Both Huh‐7‐trans and Huh‐7‐DN displayed more karyotypic abnormalities and differential gene expression profiles than their native Huh‐7 cells. Sensitivity to Sorafenib, doxorubicin and cisplatin was remarkably improved after either GLI1 or TAP1 gene was inhibited by an RNAi approach or by a specific GLI1/2 inhibitor, GANT61. Further experiments confirmed that hedgehog transcription factor GLI1/2 binds to the TAP1 promoter, indicating that TAP1 is one of GLI1/2 target genes. In conclusion, TAP1 is under direct transcriptional control of the hedgehog signalling. Targeting hedgehog signalling confers a novel insight into alleviating drug resistance in the treatment of refractory HCC.


| INTRODUC TI ON
With an increasing incidence, hepatocellular carcinoma (HCC) ranks the third cancer-related death. 1  A previous study demonstrated that hedgehog (Hh) signalling components were detected in 50%-70% of specimens in advanced HCC, indicating that Hh signalling activation in HCC is not uncommon. 2 Our previous studies implicated Hh signalling plays an important role in enhancing hepatic carcinogenesis, epithelial-mesenchymal transition (EMT) and invasion. 3 However, available studies could not provide a clear picture of Hh molecular interplays in modulating HCC drug resistance. In our previous studies, two poorly differentiated hepatoma subpopulations, Huh-7-trans and Huh-7-DN with aberrant Hh activation, were isolated from their parental Huh-7 cells by fluorescence-activated cell sorting (FACS) with a negative CD133/ EpCAM surface marker profile. These two subpopulations displayed enhanced chemo-resistance and invasive behaviour compared with their counterparts. 4,5 Our further study demonstrated that Hh signalling transcription factor GLI2 modulates drug resistance in Huh-7-DN cells through ABCC1 (ATP-binding cassette, ABC subfamily C member 1). 5 Therefore, it is intriguing to explore whether the Hh signalling controls drug resistance in poorly differentiated hepatoma cells with a negative CD133/EpCAM surface marker profile through a particular molecular mechanism. Our working hypothesis lies in that subpopulations derived of heterogeneous hepatoma mass may possess primary resistance to chemotherapies under the hedgehog signalling control. Resistant subpopulations which are often poorly differentiated tend to escape from chemotherapies and overgrow due to a forced selection in exposure to chemotherapy. 6 Our preliminary RNA-sequencing (RNA-Seq) study revealed that another ABC member, TAP1 (ABCB2, ABC subfamily B member) highly expressed in both Huh-7-trans and Huh-7-DN subpopulations.
Therefore, the aim of the present study was to elucidate the underlying mechanism of the hedgehog signalling in governing chemosensitivity through modulating TAP1 expression in both Huh-7-trans and Huh-7-DN subpopulations, the representatives of poorly differentiated hepatoma cells. To test our hypothesis, karyotypic analysis, RNA-seq analysis, drug sensitivity assay, GLI1 and TAP1 silencing by lentiviral transduction were employed to explore crucial role of the hedgehog signalling in mediating chemo-resistance. The findings of this study deepen our understanding of the drug resistance modulation by the hedgehog signalling and provide a promising strategy for improving chemosensitivity of HCC.

| Cell culture
Hep3B hepatoma cell line was obtained from ATCC, and Huh-7 hepatoma cell line was a gift from Prof. Mark Feitelson (Temple University, PA, USA). HLE and HLF hepatoma cell lines were obtained from the Japanese Collection of Research Bioresource bank.
The above-mentioned cell lines were all authenticated using Short Tandem Repeat (STR) analysis (GENEWIZ, Inc) and tested periodically for mycoplasma by polymerase chain reaction (PCR). Huh-7-DN subpopulation was enriched with EpCAM − /CD133 − surface markers from Huh-7 cells by FACS, while Huh-7-trans cells were further selected using the transwell migration assay after FACS enrichment as we previously reported. 4,5

| Human samples
A total of 12 pairs of HCC specimens and matched pericancerous tissues were obtained from the tissue repository of the Department of Hepatobiliary Surgery, Huashan Hospital of Fudan University, Shanghai, China. All specimens were cut into pieces of 3 mm in diameter for RNA and protein extraction, or fixed in 4% paraformaldehyde for immunohistochemistry. Corresponding clinical data and pathologic diagnosis were obtained from medical record with written consent prior to surgical procedures. The human subject protocol (2018-M01) was approved by the Ethic Committee of Fudan University School of Basic Medical Sciences.

| RNA-Seq analysis
To find potential genes involved in drug resistance of Huh-7-trans and Huh-7-DN cells, total mRNA of Huh-7, Huh-7-trans and Huh-7-DN cells was extracted by Trizol ® Reagent (Invitrogen) according to the instruction manual. After verification of integrity and purity, total mRNA was randomly fractured into fragments of approximately 200 bp and transcribed to double-strand DNA with added adaptors.
Sequences of resulting DNA products were probed through Illumina Hiseq4000 by Majorbio Ltd. Differential gene expression was analysed through edgeR software using gene read count. False discovery rate (FDR) < 0.01 & log 2 |FC| ≥ 2 was considered as significantly differential gene expression. Gene ontology (GO) functional analysis was completed by Fisher's exact test using GOatools Software, and P value was adjusted by four tests (Bonferroni, Holm, Sidak and false discovery rate) to control false positive rate, when corrected P value (p_fdr) ≤0.05 was considered to be significant for GO enrichment.
Kyoto Encyclopaedia of Genes and Genomes (KEGG) pathway was analysed by Fisher's exact test using KOBAS software. The corrected P ≤ .05 was considered to be significant for KEGG enrichment.

| Quantitative reverse transcription polymerase chain reaction (qRT-PCR)
Total RNA of hepatoma cells was isolated using RNA Prep Pure Cell kit (TIANGEN), and total RNA of human HCC tissues was isolated by the TRIzol reagent (Thermo-Fisher) according to the manufacturer's instruction. Then, total RNA was transcribed to cDNA by PrimeScript RT reagent kit (TAKARA). Q uantitative RT-PCR was performed using the Power SYBR Green Master Mix (Invitrogen Inc) in the Eppendorf AG 22331 RT-PCR system. Relative mRNA expression of ABC transporters, hedgehog signalling molecules, epithelial-mesenchymal transition (EMT) and hepatocyte-specific proteins was normalized to glyceraldehyde-3-phosphate dehydrogenase (GAPDH) as a house-keeping gene control and calculated using 2 (−ΔΔCT) methods as previously described. 4 All primer pair sequences were listed in Table S1.

| Immunohistochemical staining
Hepatocellular carcinoma specimens and paired pericancerous tissue were fixed in 4% formalin, sectioned and embedded in paraffin for immunohistochemistry (IHC). Sections were deparafinized in xylene and rehydrated by ethanol at different concentration.
Slides were immersed in 3% H 2 O 2 to block endogenous peroxidase activity, followed by heating sodium citrate buffer at pH 6.0 with 0.05% Tween 20 for antigen retrieval. Non-specific protein binding was blocked by 3% bovine serum albumin (BSA), and sections were incubated at 4°C with primary TAP1 antibody. After overnight incubation, sections were washed three times for 5 minutes each with 1X TBST and further incubated at room temperature with HRP-conjugated anti-rabbit IgG antibody for 1 hour. The sections were visualized with colour development by DAB substrate (ZSGB-Bio) and counterstained with haematoxylin. Images were taken by inverted microscope (Nikon).

| Western blot analysis
Total protein of hepatoma cells was extracted by RIPA buffer as we previously described, 7 and total protein of human HCC specimens was extracted using Minute™ total protein extraction kit (Invent Biotechnologies) according to the manufacturer's protocol. Equal amount of protein (30 μg) was loaded and separated by a 4%-12% gradient Bis-Tris gel (Tanon Science & Technology Co., Ltd.), then transferred to a polyvinylidene fluoride (PVDF) membrane. After being blocked by 5% non-fat milk and incubated with primary antibodies against TAP1, GLI1/2, GAPDH or β-actin overnight at 4°C, blotted membranes were incubated with horseradish peroxidase (HRP)-conjugated anti-mouse or rabbit IgG antibody for 1 hour at room temperature. Protein bands were imaged by High-sig Enhanced chemiluminescent system (Tanon 5200) as previously reported. 5 The primary antibodies used were listed in Table S2.

| Construction of a reporter system and dualluciferase assay
A potential GLI-binding site on the TAP1 promoter (5′-TAGACGACCCCCCTCAGAAA-3′, NG_011759.1:4498-4517) was identified using the program MatInspector. Two fragments (TAP1_a and TAP1_b) containing the potential GLI-binding sequence in the TAP1 promoter, and one fragment (TAP1_c) without the binding site were PCR-amplified and cloned into pGL4.23 vector (Promega). Predicted binding sequence and primers for three TAP1 promoter constructs were listed in Table S4. Three TAP1 promoter constructs were cotransfected with pRL-TK renilla vector into HLE cells by Lipofectamine 3000 (Invitrogen). Forty-eight hours after transfection, promoter activity was determined by a dual-luciferase reporter assay system (Promega) as previously described, 5 where the renilla luciferase was used to normalize the transfection efficiency.

| Chromatin immunoprecipitation (ChIP) assay
The GLI-binding site of the TAP1 promoter was verified by a ChIP assay using Epiquik chromatin immunoprecipitation kit (Epigentek). In brief, DNA was sheared into fragments of 250-1000 bp by a sonicator (SCIENTZ) after cells were collected, cross-linked and lysed. Monoclonal GLI1 antibody and polyclonal GLI2 antibody were used for immunoprecipitation.
Primer sequences used to PCR-amplify DNA in the precipitated protein were sense 5′-TGTGATGAGTTGGT-3′ and anti-sense 5′-CGGAGAAGTGAATG-3′ and resulted in a product size of 368 bp.

| Statistical analysis
All the experiments were repeated at least three times independently, and data of continuous variables were represented as mean ± SD.
The difference between two groups was measured using a twosided t test or one-way ANOVA when analysing data is in more than two groups, and LSD or Bonferroni tests were further employed for multiple comparisons in given two groups. If the assumptions of the normal distribution were violated, Wilcoxon rank-sum test was used to analyze the difference. All tests were performed by IBM SPSS Statistics 20 and Graphpad Prism 6, and P ≤ .05 was considered to be statistically significant.

| Differential gene expression profiles between Huh-7-trans, Huh-7-DN and Huh-7 cells
To clarify the correlation between the karyotypic abnormality and gene expression profile, RNA-seq and KEGG pathway analysis were used to explore transcriptome of Huh-7-trans and Huh-7-DN cells.
The similarity in gene expression between Huh-7-trans or Huh-7-DN cells and parental Huh-7 cells was 91.4% and 91.6% (Figure 2A)   Figure 4A,B). Consistent with qRT-PCR and immunohistochemical staining, TAP1 protein levels significantly increased in 3 out 12 of HCC specimens accompanied with an increase in GLI1 protein ( Figure 4C). In summary, hedgehog signalling activation is common in HCC tissues; whereas TAP1 is heterogeneously expressed. Both GLI1 and TAP1 tend to increase in HCC specimens with normal serum AFP value.

| GLI1 mediates TAP1 expression in poorly differentiated hepatoma cells
Given that enhanced hedgehog signalling activation and elevated Along with the Hh signalling activation, TAP1 protein levels in these hepatoma cell types were significantly higher than Huh-7 cells ( Figure 5A-G). These findings suggested that the Hh signalling activation may contribute to TAP1 expression in poorly differentiated hepatoma cells.

| GLI1 transactivates TAP1 expression in poorly differentiated hepatoma cells
Having shown that GLI1 may be an upper stream regulator of TAP1, we further determine whether TAP1 is a direct transcriptional target of GLI1. Hedgehog transcription factors GLI1/2 bind to a conserved sequence 5′-GACCACCCA-3′ to initiate the transcription of target genes, and a potential GLI-binding sequence 5′-GACCCCCCT-3′ (NG_011759.1:4503-4511) located at 500 bp proximal to the F I G U R E 2 Transcriptome heat map of Huh-7-trans and Huh-7-DN populations. Two aliquots of Huh-7, Huh-7-trans and Huh-7-DN cells were used for RNA extraction in each cell type. B, Heat map of representative differential expression of genes in drug resistance, hedgehog signalling, stemness, cancer progression and hepatocyte-specific genes from RNA-sequencing data. A, Venn diagrams of significant differentially expressed genes between Huh-7-trans and Huh-7-DN vs Huh-7 cells. FDR < 0.01 & log2|FC| ≥ 2 was considered as significantly differential gene expression. Representative of differential gene expression levels of hepatocyte-specific genes (C-F), EMT-associated transcription factors (G-J) and cancer progressive genes (K-N) was verified by quantitative RT-PCR using Huh-7 cells as controls (n = 3). *P < .05; **P < .01 compared with controls  Figure 7A). Three luciferase-reporter constructs containing predicted sequences or lacking the potential sequence were transfected into HLE cells which displayed highest transfection efficiency in three hepatoma cell types, and pRL-TK renilla luciferase vector was cotransfected for the normalization of transfection efficiency. Normalized luciferase ratio of pGL4.23-TAP1_a/b constructs was increased more than 18-fold compared with pGL4.23-TAP1_c construct 48 hours after transfection ( Figure 7B). These results demonstrate that pGL4.23-TAP1_a/b does contain the binding sequence in the TAP1 promoter. Furthermore, ChIP assay validated that after immune precipitation with antibodies against GLI1 or GLI2, the potential GLI-binding sequence within the TAP1 promoter could be amplified and visualized in gel of PCR products from the precipitated DNA, and it appears to be that the GLI1/2 transcription factors elicit TAP1 transcription when they bind to the putative sequence in poorly differentiated hepatoma cells ( Figure 7C). From surgically resected HCC specimens, it is clear that Hh signalling was activated, since GLI1 was up-regulated in approximately 75% in HCC specimens, indicating that the Hh signalling might play a pivotal role in oncogenesis or progression. In our previous study, we found that ABCC1 and Hh transcription factor GLI2 were up-regulated in Huh-7-DN cells, which displayed profound resistance to chemotherapy drugs, indicating GLI2 could regulate ABCC1 to increase drug resistance. 5 However, Huh-7-trans cells were also found to be drug-resistant in addition to their highly metastatic property.

| D ISCUSS I ON
In order to narrow down which genes are fundamental for drug resistance in advanced HCC, we screened genes belonging to ABC F I G U R E 5 Hedgehog signalling activation and TAP1 expression in hepatoma cells. Relative mRNA expression of GLI1 (A), GLI2 (B) and TAP1 (C) in Hep3B, Huh-7, Huh-7-trans, Huh-7-DN, HLE and HLF cell lines was determined by quantitative RT-PCR. Western blot analysis of GLI1, GLI2 and TAP1 in hepatoma cell lines and relative expression of GLI1 (D), GLI2 (E) and TAP1 (F) after normalization for ACTIN as a loading control. GLI1 monoclonal antibody (Cell Signaling Technology) was used for blotting. G, n = 3 *P < .05; **P < .01. H, GLI1 and TAP1 protein levels in GLI1-KD Huh-7-trans, Huh-7-DN and HLE cells by Western blot analysis. Desitometric ratios of GLI1 and TAP1 over β-actin were compared with those transduced with lentivirus harbouring scrambled shRNA controls (set to 1.0). Relative mRNA expression of GLI1 (I) and TAP1 (J) in GLI1-KD Huh-7-trans, Huh-7-DN and HLE cells compared with those transduced with lentivirus harbouring scrambled shRNA controls was determined by quantitative RT-PCR. Relative mRNA expression of TAP1 (K) in TAP1-KD Huh-7-trans, Huh-7-DN and HLE cells compared with those transduced with lentivirus harbouring scrambled shRNA controls were determined by quantitative RT-PCR. GLI1-KD cells: GLI1 was knocked-down through lentiviral transduction; TAP1-KD cells: TAP1 was knocked-down through lentiviral transduction; Scrambled: scrambled shRNA control was transduced though lentiviral transduction. n = 3, *P < .05 compared to scrambled shRNA controls transporters, which were reported to function as efflux molecules of anticancer drugs 19  sensitized Huh-7-DN cells to doxorubicin toxicity, indicating that F I G U R E 6 Chemosensitivity of Huh-7-trans and DN cells exposed to sorafenib, doxorubicin or cisplatin after blocking Hh signalling. Chemosensitivity of TAP1-KD and GLI1-KD of Huh-7-DN cells exposed to Sorafenib (A), doxorubicin (B) or cisplatin (C) and viability of TAP1-KD and GLI1-KD of Huh-7-trans cells exposed to cisplatin (D) was determined by MTT assay. n = 3, *P < .05; **P < .01; GLI1-KD cells compared with scrambled shRNA controls. #P < .05; ##P < .01; TAP1-KD cells compared with scrambled shRNA controls. Chemosensitivity of Huh-7-trans (E), Huh-7-DN (F) and HLE cells (G) exposed to Sorafenib with/ without 5 μmol/L GANT61. n = 3, *P < .05; **P < .01 compared with Sorafenib monotherapy. H, GLI1 and TAP1 expression in HLE cells treated with/ without Sorafenib (16 μmol/L) or GANT61 (5 μmol/L) by Western blot analysis. GLI1 monoclonal antibody (Cell Signaling Technology) was used for blotting. Desitometric ratios of GLI1 and TAP1 over β-actin were compared with HLE cells without any treatment (set to 1.0). n = 3 TAP1 may be a critical molecule in the mediation of drug resistance in both Huh-7-trans and Huh-7-DN cell types. In this context, both hedgehog signalling pathway and TAP1 could be valuable molecular targets for improving the outcome of chemosensitivity of a poorly differentiated type of HCC, which features with often both CD133and EpCAM-negative, exhibits marked EMT properties and possesses a devastating prognosis. 3,4 In conclusion, both Huh-7-trans and Huh-7-DN are two distinct subpopulations sharing many cytogenetic and molecular similarities with parental Huh-7 cells, and worsening abnormality in karyotypic features and differential gene expression profiles are highly respon- Jian Wu https://orcid.org/0000-0001-9933-7364