MiR-125b-5p Is Involved in Sorafenib Resistance through Ataxin-1-Mediated Epithelial-Mesenchymal Transition in Hepatocellular Carcinoma

Simple Summary The mechanism of resistance to multikinase inhibitors in hepatocellular carcinoma (HCC) remains unclear. We analyzed miRNA expression profiles in sorafenib-resistant HCC cell lines (PLC/PRF5-R1/R2) and parental cell lines (PLC/PRF5) to identify the responsible miRNAs and target genes involved in the mechanism of resistance. Four miRNAs were significantly upregulated. Among them, we found that miR-125-5p induced sorafenib resistance in HCC cells and in a mouse model. We also revealed that miR-125-5p suppressed ataxin-1 as a target gene and consequently induced Snail-mediated epithelial-mesenchymal transition (EMT) and cancer stemness. Moreover, we demonstrated that ataxin-1 expression has an impact on the prognosis of patients with HCCs. In the future, by comparing the expression status of miR-125b-5p/ataxin-1 and the effect of sorafenib in the clinical setting, it is expected that miR-125b-5p will be established as an effective drug selection marker for treatment selection in patients with HCC. Abstract The mechanism of resistance to sorafenib in hepatocellular carcinoma (HCC) remains unclear. We analyzed miRNA expression profiles in sorafenib-resistant HCC cell lines (PLC/PRF5-R1/R2) and parental cell lines (PLC/PRF5) to identify the miRNAs responsible for resistance. Drug sensitivity, migration/invasion capabilities, and epithelial-mesenchymal transition (EMT) properties were analyzed by biochemical methods. The clinical relevance of the target genes to survival in HCC patients were assessed using a public database. Four miRNAs were significantly upregulated in PLC/PRF5-R1/-R2 compared with PLC/PRF5. Among them, miR-125b-5p mimic-transfected PLC/PRF5 cells (PLC/PRF5-miR125b) and showed a significantly higher IC50 for sorafenib compared with controls, while the other miRNA mimics did not. PLC/PRF5-miR125b showed lower E-cadherin and higher Snail and vimentin expression—findings similar to those for PLC/PRF5-R2—which suggests the induction of EMT in those cells. PLC/PRF5-miR125b exhibited significantly higher migration and invasion capabilities and induced sorafenib resistance in an in vivo mouse model. Bioinformatic analysis revealed ataxin-1 as a target gene of miR-125b-5p. PLC/PRF5 cells transfected with ataxin-1 siRNA showed a significantly higher IC50, higher migration/invasion capability, higher cancer stem cell population, and an EMT phenotype. Median overall survival in the low-ataxin-1 patient group was significantly shorter than in the high-ataxin-1 group. In conclusion, miR-125b-5p suppressed ataxin-1 and consequently induced Snail-mediated EMT and stemness, leading to a poor prognosis in HCC patients.


Introduction
Hepatocellular carcinoma (HCC) is reportedly the fifth most commonly diagnosed malignancy and the second leading cause of cancer-related death worldwide [1]. For patients with unresectable advanced HCC, sorafenib has been the first recommended systemic therapy to demonstrate a survival benefit with an adequate safety profile [2,3]. Sorafenib is an oral tyrosine-kinase inhibitor (TKI) that targets RAF kinase, c-KIT kinase, vascular endothelial growth factor (VEGF) receptors, and platelet-derived growth factor (PDGF) receptors. The phase III SHARP trial showed a median overall survival (OS) of 10.7 months and a disease control rate (DCR) of 43% in the sorafenib treatment group of unresectable HCC patients with well-preserved liver function. The recent phase III REFLECT study demonstrated that lenvatinib was noninferior to sorafenib as a first-line systemic therapy for unresectable HCC [4]. Based on these results, several guidelines positioned sorafenib and lenvatinib as first-line treatments for unresectable HCC. However, currently there is no consensus as to which drug should be used first [5].
To improve OS in unresectable HCC, it is critical to select the appropriate therapeutic agent. Therefore, it is important to identify an appropriate biomarker to distinguish patients who are sensitive to sorafenib. Elucidation of the mechanisms underlying sorafenib resistance and evaluation of resistance factors before treatment would help in selecting an effective drug and developing individualized therapeutic strategies.
In our previous study, we established two sorafenib-resistant cell lines (PLC/PRF5-R1, PLC/PRF5-R2) from PLC/PRF5 cells and reported that high expression of ABCC3 transporter (MRP3) is one of the mechanisms of sorafenib resistance [6]. Moreover, recent studies have revealed that there are several mechanisms underlying acquired resistance to sorafenib, such as activation of PI3K/Akt [7] and JAK-STAT pathways, which are alternative pathways to the MAP kinase signaling pathway [8], activation of hypoxiainducible pathways [9], and induction of epithelial-mesenchymal transition (EMT) [10]. Although various mechanisms have been suggested, there are no predictive factors that are useful in clinical practice. All of these previous studies utilized liver biopsy tissue to evaluate candidate biomarkers-a method which is difficult to apply clinically. Thus, the development of blood biomarkers is needed for personalized medicine in HCC.
MicroRNAs (miRNA) are small non-coding RNA molecules of 20-25 nt that regulate gene expression through transcriptional repression and mRNA degradation [11][12][13][14][15]. It has been reported that miRNAs are involved in carcinogenesis, invasion/metastasis, and EMT in various types of cancers [16,17]. For example, miR-125b-5p is abnormally expressed in multiple cancers and is identified as both a tumor promoter and a tumor suppressor in different kinds of cancers [18]. In HCC, miR-125b-5p has been reported to act as a tumor suppressor, exerting inhibitory effects on EMT by small mothers against decapentaplegic (SMAD)2 and 4 [19]. In colorectal cancer, it has been reported that several small non-coding RNAs or miRNAs regulate the epithelial phenotype and EMT by inhibiting the expression of EMT regulators [20]. miRNAs have been used as targets for liquid biopsies, which allow for easy and safe sample collection. Moreover, the development of miRNA microarray assays has made it possible to evaluate the expression of >3000 miRNAs simultaneously [21]. Using this method, Lin and associates reported that miR-378a expression was downregulated in sorafenib-resistant cell lines, and that its target gene IGF1-R was overexpressed, and consequently resistance to sorafenib-induced apoptosis was acquired [22]. However, the association of miR-378a with EMT, cancer stem cell phenotype, and the mechanism of apoptosis resistance to sorafenib is unclear. Moreover, because miR-378a was shown to be downregulated in a resistant cell line, it cannot be used as a biomarker for liquid biopsy.
EMT is associated with poor patient survival because it is a key step in the development of metastasis in cancer. It has been reported that epithelial cells with EMT lose cell adhesion molecules (such as E-cadherin) and gain mesenchymal cell markers (such as vimentin and Snail), resulting in the loss of polarity and cell-to-cell contacts, enhancement of tumor cell migration and invasion, and resistance to anti-cancer drugs including sorafenib [23][24][25]. Different studies have demonstrated that EMT may be one of the mechanisms of sorafenib resistance [24,26], but the mechanism of EMT regulation remains largely unknown. Therefore, in the process of acquiring resistance to sorafenib, it has been postulated that abnormal miRNA expression is responsible for inducing EMT of HCC, resulting in the acquisition of resistance to sorafenib and enhancement of cell proliferation, metastasis, and invasion capabilities. However, the mechanism by which miRNA regulates EMT in HCC remains poorly understood.
Ataxin-1 (ATXN1) is a ubiquitous polyglutamine protein expressed primarily in the nucleus where it binds chromatin and interacts with a number of known transcriptional repressors, indicating a role in the regulation of gene expression [27]. ATXN1 loss-offunction is implicated in cancer pathogenesis. In colorectal cancer, it has been reported that ATXN1 is a putative cancer gene and expression of ATXN1 in tumor cells is downregulated compared with normal colon cells [28,29]. However, the full spectrum of ATXN1 functions is far from being fully characterized.
In this study, we investigated miRNA expression profiles in sorafenib-resistant HCC cell lines (PLC/PRF5-R1/R2) to clarify relevant miRNA expression related to sorafenib resistance in comparison with parental PLC/PRF5 cells. We ultimately found upregulation of miRNA125b-5p in resistant cells and identified ATXN1 as the target gene. We also analyzed the potential role of miRNA125b-5p and ATXN1 in EMT, invasion, migration, and stemness in HCC. Moreover, we used a public database to evaluate the clinical relevance of ATXN1 expression in patients with HCC in association with their survival.

Cell Viability Analysis
Drug sensitivity of cells to sorafenib was estimated using the WST assay as previously described [30]. In brief, cells were seeded in 96-well plates (3 × 10 3 cells/well) and incubated for 24 h at 37 • C. Sorafenib was then added to the wells at various concentrations and the plates were incubated for another 72 h at 37 • C. Cell viability was measured using a cell counting kit-8 (CCK-8) assay (Dojindo Laboratories, Kumamoto, Japan). CKK-8 solution was added to the wells and the plates were incubated for 3 h at 37 • C. Absorbance at 450 nm was determined using a Spectra Max i3x Platform (Molecular Devices, Inc., Danaher Cor-

MicroRNA Microarray Analysis
A human miRNA microarray (based on miRbase release 21.0; Agilent Technologies, Palo Alto, CA, USA) was used for measuring global miRNA expression in cell lines, as previously described [30]. In brief, total RNA was labeled with cyanine 3-cytidine bisphosphate by T4 RNA ligase and hybridized to SurePrint G3 human miRNA microarray release 21.0 using miRNA complete labeling reagent and a hybridization kit (Agilent Technologies). Subsequently, each sample was scanned by a DNA microarray scanner (G2505C; Agilent Technologies), and the fluorescence signal was extracted using feature extraction software (version 10.7.3.1). Raw intensity miRNA data were analyzed using GeneSpring GX version 12 software (Agilent Technologies, Santa Clara, CA, USA).

RT-PCR Analysis
Quantitative real-time PCR was performed as described previously [30]. Total RNA of cell lines, including miRNAs, was extracted using a Qiagen miRNeasy Mini kit (Qiagen, Hilden, Germany). RNA was then reverse transcribed into cDNA using a High Capacity RNA-to-cDNA kit (Applied Biosystems, Waltham, MA, USA). Quantitative real-time PCR was then carried out in 96-well plates in a StepOnePlus Real-Time PCR System (Applied Biosystems) using TaqMan Universal PCR Master Mix (Applied Biosystems) to monitor the PCR amplification. The real-time PCR mixtures using TaqMan Universal PCR Master Mix consisted of 5.0 µL of TaqMan Gene Expression Master Mix (2×) and 0.5 µL TaqMan Gene Expression Assay (20×) ( Table S1). The following two-step cycling program was used for PCR reactions: 50 • C for 2 min and 95 • C for 10 min, and 40 cycles of (95 • C, 15 s; and 60 • C, 1 min). All samples were amplified in triplicate and relative quantification of the expression level of each gene was calculated. GAPDH was used as the endogenous reference gene.

Overexpression and Knockdown Experiments
The miRNA mimics and random miRNA used as a control were purchased from Thermo Fisher Scientific (Waltham, MA, USA) (Table S2). Each miRNA was transfected into cells at a concentration of 10 nM, as we previously described [31]. Sorafenib was added to the cells after 24 h, followed by incubation for 72 h. Cell viability was then determined by WST assay. ATXN1 expression plasmid or negative control vector (vector NC) (Table S2) was transiently transfected to the cells using Lipofectamine 3000 (Invitrogen, Carlsbad, CA, USA) according to the manufacturer's instructions. For knockdown experiments, cells were transfected with 10 nM of ATXN1 specific small interfering RNA (siRNA) or random siRNA as a control (Table S2). Sorafenib was added to the cells after 24 h, followed by incubation for 72 h. Cell viability was then determined by WST assay.

In Silico Identification of miRNA Target Genes
Target genes were bioinformatically predicted based on the miRNA seed sequence by using miRDB [32], an online database for miRNA target prediction and functional annotations [33], and Target Scan [34].

miRNA Luciferase Reporter Assay
Double-stranded oligonucleotides for the 3 -UTR of ATXN1 harboring miR-125b-5p binding sites 1 and 2, respectively, were prepared by heating equal amounts of complementary single strands at 95 • C for 15 min and gradually cooling to room temperature (Table S3). The DNA fragments were subcloned into psiCHECK-2 vector (Promega, Madison, WA, USA) using XhoI and NotI restriction sites. The plasmids with two mutant types for each miR-125b-5p binding site were also prepared by replacing six base pairs at the 3 -UTR of the seed sequence.

Promoter Assay
The 5 -flank of the human ATXN1 gene was cloned into the pGL4.21-basic luciferase reporter vector (Promega, Madison, WI, USA). In brief, the first PCR was performed using human genomic DNA as a template. The ATXN1 proximal promoter region between 834 bp upstream and 151 bp downstream from the transcriptional start site was amplified using primer sets listed in Table S3. The amplified products were subcloned into the pGL4.21-basic vector using NheI and HindIII restriction sites. HEK293T cells (8.0 × 10 4 ) were cultured on 24-well plates. After 24 h incubation, miR-125b-5p (10 nM) or control mimic miRNA (10 nM) was transfected. Twelve hours after the transfection of mimic miRNAs, pGL-4.21 luciferase constructs with or without the promoter region (100 ng) were cotransfected with pGL4.74 vector (50 ng) using X-tremeGENE HP DNA transfection reagent (Roche, Basel, Switzerland). Twenty-four hours after the transfection, cells were harvested, and the firefly and Renilla luciferase activities were measured using the Dual-Luciferase Reporter Assay System (Promega, Madison, WI, USA).

Western Blot Analysis
Expression of proteins was analyzed by western blot analysis, as described previously [35]. Briefly, cells were washed with phosphate buffered saline (PBS) and lysed in RIPA buffer containing protease inhibitors (Sigma-Aldrich, Tokyo, Japan). Cell lysates were analyzed for protein content, resolved by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE), and transferred to polyvinylidene fluoride membranes using a semi-dry transfer apparatus (Bio-Rad, Hercules, CA, USA). Blots were blocked with 5% fat-free dry milk in Tris-buffered saline with Tween 20 (TBS-T) for 1 h and then incubated overnight with rabbit anti-human E-cadherin monoclonal antibody, rabbit anti-vimentin monoclonal antibody, rabbit anti-human Snail monoclonal antibody (Cell Signaling Technology, Tokyo, Japan), mouse anti-ATXN1 monoclonal antibody (Santa Cruz Biotechnology, Dallas, TX, USA), and mouse anti-β-actin monoclonal antibody (Sigma-Aldrich, Tokyo, Japan) as primary antibodies. The membranes were washed with TBS-T and incubated with secondary horseradish conjugated sheep anti-mouse antibody or donkey anti-rabbit antibody (GE Healthcare UK Limited, Buckinghamshire, UK). The proteins were visualized by standard procedures including an ECL detection system (GE Healthcare UK Limited). To ensure equal protein loading, the same blot was developed for β-actin (Sigma-Aldrich, Tokyo, Japan) as a loading control. The expression levels of each protein were quantified by densitometric analysis using Image Lab Software (Bio-Rad, Hercules, CA, USA).

Wound Healing Assay
Cells (1.0 × 10 6 ) for the wound healing assay were seeded in 6-well plates. When the cell confluence reached about 80%, scratch wounds were made by scraping the cell layer across each culture plate using the tip of a 200 µL pipette. After wounding, the debris was removed by washing the cells with PBS. Wounded cultures were incubated in the culture medium containing 10% FBS with miRNA mimic or siRNA. Wound healing was imaged every 24 h for 72 h using a BZ-X710 fluorescence microscope (Keyence, Osaka, Japan). Wound healing ability was determined by measuring the mean migration distance, which was calculated by dividing the wound repair area by the width. Experiments were carried out three times.

Cell Invasion Assay
Cell invasion assays were performed as described previously [31]. In brief, cells transfected with miRNA mimic or siRNA were seeded at 5 × 10 5 cells/well in the upper chamber of a CytoSelect 24-Well Cell Invasion Assay kit (Cell Biolabs, San Diego, CA, USA). The number of invading cells was counted in three randomly selected views under a microscope.

Flow Cytometry Analysis
CD44 and CD133 are well-known cancer stem cell (CSC) surface markers commonly used in HCC [36]. Cells were tested for CD44 and CD133 by incubating them with mouse anti-human CD44 monoclonal antibody (APC) (17044181, Thermo Fisher Scientific) and mouse anti-human CD133 monoclonal antibody (PE) (372803, BioLegend, San Diego, CA, USA). After washing the cells two times with PBS containing 1% BSA and 1% sodium azide, the expression of CD44 and CD133 was assessed using a BD FACS Verse flow cytometer (BD Biosciences, San Jose, CA, USA). Annexin V and propidium iodide (PI) were used for the detection of apoptosis.

Tumor Xenograft Experiments
Female 6-week-old BALB/c nu/nu mice (CLEA Japan Inc., Tokyo, Japan) were used for all studies. PLC/PRF5 cells were infected with MISSION ® Lenti microRNA expressing miR125b-5p or negative control virus (Table S2) according to the manufacturer's instructions. Tumor cells (0.5 × 10 7 ) were then inoculated into the flanks of mice. Sorafenib or vehicle control was administered orally, once a day, for 21 days at dose levels of 30 mg/kg, as reported previously [37]. Treatment began when tumors reached a volume of 200 mm 3 . Tumor size was calculated using the equation (l × w 2 )/2, where l and w refer to the larger and smaller dimensions collected at each measurement. Tumor dimensions were recorded three times a week starting on the first day of treatment. All animal experiments were carried out according to the Guidelines for Animal Experiments at Tokushima University.

ATXN1 Expression from RNA-Chip Data
The profiling expression of ATXN1 in HCC tissues from microarray studies was acquired from the Genomic Data Commons data portal [38]. Three studies (TCGA-LIHC [39], GSE76427 [40], and GSE10141 [41]) were available for analyzing both ATXN1 mRNA expression and useable survival data. However, since GSE76427 and GSE10141 contained small study populations (24 and 80 patients, respectively), only data from TCGA-LIHC (377 patients) were used. Fragments Per Kilobase Million with Upper Quantile (FPKM-UQ) values normalized by RNA-seq counts were used to perform survival analysis. The cutoff value was set at 50,000 FPKM-UQ based on the median ATXN1 mRNA expression; cases with ≥50,000 FPKM-UQ were designated as the high expression group, and cases with lower FPKM-UQ (<50,000) as the low expression group. OS in the high-versus low-ATXN1 expression group was compared using the Kaplan-Meier method and log-rank test.

Statistical Analysis
Data are presented as mean ± SD. The statistical significance of the difference between the values of the control and treatment groups was evaluated by either Student's t-test or ANOVA, followed by Dunnett's test using Prism version 5 (GraphPad Software, San Diego, CA, USA). Values of p < 0.05 were considered statistically significant.

miR-125b-5p Induces EMT in PLC/PRF5 Cells
Since drug resistance is one of the major phenotypes associated with EMT, we investigated expression of EMT-related proteins in PLC/PRF5-miR125b cells and sorafenibresistant cells (PLC/PRF5-R2-miNC) in comparison with PLC/PRF5-miNC by western blot analysis. The expression of E-cadherin in PLC/PRF5-R2-miNC cells was obviously lower compared with PLC/PRF5-miNC cells, whereas expression of Snail and vimentin in PLC/PRF5-R2-miNC cells was obviously higher compared with PLC/PRF5-miNC cells. In PLC/PRF5-miR125b cells, the expression of E-cadherin was slightly lower and expression of Snail and vimentin was higher compared with PLC/PRF5-miNC cells (Figure 2c). Thus, the expression of EMT-related proteins in PLC/PRF5-miR125b cells was changed toward that of PLC/PRF5-R2 cells.
In the microscopic analysis of the morphology, PLC/PRF5-R2 cells showed a fibroblastlike shape, which was morphologically distinct from the round or oval shape of PLC/PRF5-miNC cells-findings characteristic of epithelial cells-suggesting that PLC/PRF5-R2 cells acquired a mesenchymal phenotype. The morphology of PLC/PRF5-miR125b cells was distinct from that of PLC/PRF5-miNC cells, but similar to PLC/PRF5-R2 cells (Figure 2d). Taken together, these results indicate that miR-125b-5p has EMT-inducing ability in PLC/PRL5 cells, leading to acquisition of sorafenib resistance.

Predicted Target Genes of miR-125b-5p
To elucidate the underlying regulatory mechanism of miR-125b-5p on EMT, we applied miRDB to screen potential targets of miR-125b-5p and selected the top 50 genes sorted by the target scores (Table S6). Among them, we chose the genes that negatively controlled EMT and were also downregulated in PLC/PRF5-R1/R2 cells compared with PLC/PRF5 cells. As a result, we identified the ATXN1 gene as a potential target of miR-125b-5p. Target Scan predicted two nucleotide sequences (position 2439-2445, 4097-4104) in the 3 -UTR of the ATXN1 gene matching the miR-125b-5p sequence (Figure 3a). To verify the potential binding sequences of miR-125b-5p, we prepared the reporter plasmids (psiCHECK2) that expressed chimeric RNAs containing the sequence for Renilla and 3 -UTR sequence of ATXN1 and evaluated the effects of miRNA on the chimeric RNAs using the dual luciferase assay system. The overexpression of miR-125b-5p significantly reduced the luciferase activity of Renilla_ATXN1-3 -UTR (Figure 3b). Two mutant vectors (MT1 and MT2), which harbored seven-point mutations within each binding site for miR-125b-5p (Figure 3a), did not show a change in luciferase activity (Figure 3b). To reveal whether miR-125b-5p regulates the transcription of ATXN1, we assessed the promoter activity of ATXN1 using a dual-luciferase reporter assay. Overexpression of miR-125b-5p resulted in no decrease in luciferase activity (Figure 3c). These results suggested that miR-125b-5p does not regulate the transcription of ATXN1, but is involved in translational regulation including RNA degradation.  The ATXN1 mRNA level in PLC/PRF5-miR125b determined by RT-PCR was decreased to 34% of control cells (p < 0.01; Figure 3d). The protein expression of ATXN1 in PLC/PRF5-miR125b cells was lower than in PLC/PRF5-miNC (Figure 3e).
Since ATXN1 is known to suppress the transcription of Snail [42], we transfected ATXN1 siRNA (siATXN1) or negative control siRNA (siNC) in PLC/PRF5 cells (PLC/PRF5-siATXN1 and PLC/PRF5-siNC) and examined changes in Snail expression. The ATXN1 mRNA level in PLC/PRF5-siATXN1 cells was suppressed to 16% of the level in PLC-PRF5-siNC cells ( Figure S4). The mRNA level of Snail in PLC/PRF5-siATXN1 cells was significantly higher compared with PLC/PRF5-siNC cells (p < 0.01; Figure 3f). These results indicate that ATXN1 is the target gene of miR-125b-5p and that knocking down ATXN1 directly upregulates Snail, possibly leading to promotion of EMT and acquisition of drug resistance (Figure 3g).

Downregulation of ATXN1 Induces EMT and Sorafenib Resistance
To examine the role of ATXN1 in EMT and drug resistance, we next examined EMT-related protein expression, morphology, and sensitivity to sorafenib in PLC/PRF5-siATXN1 cells compared with those in PLC/PRF5-siNC and PLC/PRF5-R2 cells transfected with siRNA negative control (PLC/PRF5-R2-siNC). The protein expression of ATXN1 in PLC/PRF5-siATXN1 and PLC/PRF5-R2-siNC was substantially lower than that of control cells. The expression of Snail and vimentin in PLC/PRF5-siATXN1 and PLC/PRF5-R2-siNC cells was obviously higher compared with PLC/PRF5-siNC cells, whereas the expression of E-cadherin in those cells was lower compared with PLC/PRF5-siNC cells (Figure 4a). The morphology of PLC/PRF5-siATXN1 cells was distinct from that of PLC/PRF5-siNC cells; the latter were adhesively contacted with each other and showed an oval or round shape, but the former lost cell-cell adhesion and had a non-round shape-findings similar to those for PLC/PRF5-miR125b cells (Figure 4b). These results suggest that knockdown of the ATXN1 gene induced EMT in PLC/PRF5 cells.

Overexpression of ATXN1 Confer Drug Resistance and EMT in HCC Cells
To verify the function of ATXN1, PLC/PRF-R2 cells and HLF cells, which possess a more mesenchymal phenotype [45], were transfected with ATXN1 expressing the plasmid vector or negative control vector (PLC/PRF5-R2-ATXN1, PLC/PRF5-R2-vector NC, HLF-ATXN1, and HLF-vector NC, respectively). The IC50 of sorafenib against PLC/PRF5-R2-ATXN1 cells was significantly lower compared with PLC/PRF5-R2-vector NC (14.9 μM (95%Cl: 9.66-17.03 μM) vs. 16.53 μM (95%Cl: 14.99-17.96 μM); p < 0.05) (Figure 7a). Moreover, the IC50 of sorafenib against HLF-ATXN1 cells was significantly lower compared with HLF-vector NC (7.81 μM (95%Cl: 5.99-8.80 μM) vs. 10.74 μM (95%Cl: 9.69-11.89 μM), p < 0.05) (Figure 7b). The expression of Snail and vimentin in those cells was lower, and the expression of E-cadherin was higher than in control cells (Figure 7c,d). The morphology of PLC/PRF5-R2-ATXN1 and HLF-ATXN1 cells were more epithelial-like, adhesively contacted with each other, and showed an oval or round shape ( Figure S6b,c). These results strongly suggest that ATXN1 abrogated sorafenib resistance and EMT in PLC/PRF5-R2 and HLF cells. . They were exposed to sorafenib, and cell viability was assessed by WST assay. (c,d) Expression of ATXN1 and EMT markers (E-cadherin, Snail, vimentin) were examined by western blot analysis. Densitometry analysis indicates relative protein levels from one representative of three independent experiments. Numbers below the bands represent protein expression normalized to β-actin. . They were exposed to sorafenib, and cell viability was assessed by WST assay. (c,d) Expression of ATXN1 and EMT markers (E-cadherin, Snail, vimentin) were examined by western blot analysis. Densitometry analysis indicates relative protein levels from one representative of three independent experiments. Numbers below the bands represent protein expression normalized to β-actin.

Overexpression of miR-125b-5p Confer Sorafenib Resistance In Vivo
To investigate whether miR-125b-5p could promote sorafenib resistance in vivo, we established a xenograft mouse model. After establishing PLC/PRF5 cells stably expressing miR-125b-5p using lentiviral vectors, cells were inoculated subcutaneously in the dorsal flank of nude mice. After 21 sorafenib treatments (30 mg/kg/day), the volume of the experimental group was significantly larger than that in controls (n = 6, 773 ± 145 mm 3 vs. 427 ± 84 mm 3 , p < 0.05). These results suggested that overexpression of miR-125b-5p increased the resistance to sorafenib in vivo (Figure 8a,b).

Overexpression of miR-125b-5p Confer Sorafenib Resistance In Vivo
To investigate whether miR-125b-5p could promote sorafenib resistance in vivo, we established a xenograft mouse model. After establishing PLC/PRF5 cells stably expressing miR-125b-5p using lentiviral vectors, cells were inoculated subcutaneously in the dorsal flank of nude mice. After 21 sorafenib treatments (30 mg/kg/day), the volume of the experimental group was significantly larger than that in controls (n = 6, 773 ± 145 mm 3 vs. 427 ± 84 mm 3 , p < 0.05). These results suggested that overexpression of miR-125b-5p increased the resistance to sorafenib in vivo (Figure 8a,b).

ATXN1 Expression was Significantly Associated with Survival in Patients with Advanced HCC
To investigate the clinical relevance of ATXN1 expression in patients with advanced HCC, we performed a Kaplan-Meier analysis to determine the association between ATXN1 expression and OS using the TCGA-LIHC dataset. Patients with Stage III and IV HCC (n = 91) were chosen and 90 patients with ATXN1 data available were divided into low-ATXN1 and high-ATXN1 groups (Figure 9a). The low-ATXN1 group exhibited significantly shorter survival than the high-ATXN1 group (p = 0.0153; Figure 9b). The median OS in the low-ATXN1 group was much shorter than in the high-ATXN1 group (18.35 vs. 39.78 months, hazard ratio 0.49, 95%Cl: 0.28-0.87; p < 0.05).

ATXN1 Expression was Significantly Associated with Survival in Patients with Advanced HCC
To investigate the clinical relevance of ATXN1 expression in patients with advanced HCC, we performed a Kaplan-Meier analysis to determine the association between ATXN1 expression and OS using the TCGA-LIHC dataset. Patients with Stage III and IV HCC (n = 91) were chosen and 90 patients with ATXN1 data available were divided into low-ATXN1 and high-ATXN1 groups (Figure 9a). The low-ATXN1 group exhibited significantly shorter survival than the high-ATXN1 group (p = 0.0153; Figure 9b). The median OS in the low-ATXN1 group was much shorter than in the high-ATXN1 group (18.35 vs. 39.78 months, hazard ratio 0.49, 95%Cl: 0.28-0.87; p < 0.05).

Overexpression of miR-125b-5p Confer Sorafenib Resistance In Vivo
To investigate whether miR-125b-5p could promote sorafenib resistance in vivo, we established a xenograft mouse model. After establishing PLC/PRF5 cells stably expressing miR-125b-5p using lentiviral vectors, cells were inoculated subcutaneously in the dorsal flank of nude mice. After 21 sorafenib treatments (30 mg/kg/day), the volume of the experimental group was significantly larger than that in controls (n = 6, 773 ± 145 mm 3 vs. 427 ± 84 mm 3 , p < 0.05). These results suggested that overexpression of miR-125b-5p increased the resistance to sorafenib in vivo (Figure 8a,b). Sorafenib or vehicle control was administered orally, once a day, for 21 days at a dose of 30 mg/kg. There was no lethality in any group. (a) Images of the tumors from all mice in each group (n = 6). (b) Tumor dimensions were recorded three times a week starting with the first day of treatment (n = 6). Data represent mean ± SD. * p < 0.05 compared with the control group.

ATXN1 Expression was Significantly Associated with Survival in Patients with Advanced HCC
To investigate the clinical relevance of ATXN1 expression in patients with advanced HCC, we performed a Kaplan-Meier analysis to determine the association between ATXN1 expression and OS using the TCGA-LIHC dataset. Patients with Stage III and IV HCC (n = 91) were chosen and 90 patients with ATXN1 data available were divided into low-ATXN1 and high-ATXN1 groups (Figure 9a). The low-ATXN1 group exhibited significantly shorter survival than the high-ATXN1 group (p = 0.0153; Figure 9b). The median OS in the low-ATXN1 group was much shorter than in the high-ATXN1 group (18.35 vs. 39.78 months, hazard ratio 0.49, 95%Cl: 0.28-0.87; p < 0.05). Figure 9. ATXN1 expression was significantly associated with the overall survival of patients with advanced HCC. (a) A flowchart of cohort selection from the public database. Among the TCGA-LIHC dataset, patients with Stage III or IV (n = 91) were selected, and one patient was excluded due to lack of array data. Ninety patients were divided into two groups according to high or low expression of ATXN1. (b) Kaplan-Meier survival curves for overall survival. HR, hazard ratio; CI, confidence interval. The p-value was calculated by log-rank test. Among the TCGA-LIHC dataset, patients with Stage III or IV (n = 91) were selected, and one patient was excluded due to lack of array data. Ninety patients were divided into two groups according to high or low expression of ATXN1. (b) Kaplan-Meier survival curves for overall survival. HR, hazard ratio; CI, confidence interval. The p-value was calculated by log-rank test.

Discussion
In this study, we used miRNA microarray analysis to identify 125b-5p as a pivotal miRNA underlying sorafenib resistance in HCC cells. Using in vitro and in vivo approaches, we found that miR-125b-5p directly inhibited ATXN1 gene expression and consequently induced EMT through enhanced Snail expression, leading to resistance to sorafenib. We also found that miR-125b-5p and downregulation of ATXN1 enhanced migration/invasion capability and stemness of HCC cells, and that overexpression of ATXN1 reverses EMT and sorafenib resistance. Furthermore, patients with HCC in the low-ATXN1 expression group exhibited a significantly shorter survival time than those in the high-ATXN1 group, as indicated by Kaplan-Meier analysis. This is the first report to show a pivotal role of the miR-125b-5p-ATXN1 axis in the process of EMT, including acquisition of drug resistance, in HCC.
Our results showing that overexpression of miR-125b-5p was closely associated with EMT and drug resistance in sorafenib-resistant HCC cells are consistent with previous studies in other types of cancers. Lu and associates reported that miR-125b-5p augmented cetuximab resistance through activation of the Wnt signaling pathway in colorectal cancer [46]. Similarly, Yu and associates reported that miR-125b-5p enhanced 5-FU resistance in colorectal cancer [47]. In contrast, Zhang and associates reported that decreased expression of miR-125b-5p promoted EMT and conferred resistance to paclitaxel in non-small cell lung cancer [48]. These conflicting results have been explained by differences in the target genes among different types of cancer cells. In addition, Jun-Nian and associates reported that miR-125b-5p exerts inhibitory effects on EMT and EMT-associated traits in HCC by inhibiting SMAD2 and SMAD4 mRNA expression [19]. However, in our analysis of PLC/PRF5-R2 cells, SMAD2 mRNA expression was conversely upregulated, and SMAD4 was unchanged compared with PLC/PRF5 cells despite higher expression of miR-125b-5p ( Figure S7). Importantly, we were able to show that the miR-125b-5p-ATXN1 axis is associated with sorafenib resistance as well as EMT in various HCC cell lines, including PLC/PRF5 and Hep3B which were infected with HBV and JHH-6, which was derived from an HCV-positive patient.
We identified the ATXN1 gene as a target gene of miR-125b-5p in this study. The role of ATXN1 in cancers has long remained unclear. However, Kang and associates reported that ATXN1 inhibited Snail expression, which is a transcription factor for E-cadherin, leading to EMT in cervical cancer [43]. In other words, reduced ATXN1 expression caused EMT in cervical cancer cells. Our data indicated that miR-125b-5p inhibited ATXN1 expression and, therefore, induced Snail-mediated EMT, resulting in augmented drug resistance, enhanced invasion/migration capability, and a stem cell phenotype. Furthermore, overexpression of ATXN1 reversed EMT and sorafenib resistance. Thus, it is suggested that ATXN1 expressed in cancer cells protects the cells from EMT and subsequent malignant progression. This is also supported by our clinical data showing that the OS time in the low-ATXN1 HCC patient group was shorter than in the high-ATXN1 HCC patient group.
Various mechanisms for acquiring resistance to sorafenib in HCC have been reported to date. Activation of PI3K/Akt [7] and JAK-STAT pathways [8] due to crosstalk with RAS/RAF/MEK pathways, activation of hypoxia-inducible pathways [49], and EMT [24] have been implicated. Among these, EMT, which we observed in this study, is closely associated with the malignant transformation of cancer, and it is reportedly associated with acquisition of invasive capacity, drug resistance, and increased cell proliferation [50][51][52]. Thus, EMT can be assumed to be one of the major mechanisms of resistance to sorafenib. Previously, we reported that the ABC transporter MRP3 is highly expressed in the sorafenibresistant cell line used in this study [6]. It has been reported that epithelial cells, which undergo EMT, overexpress not only mesenchymal biomarkers but also CSC markers [53]. Furthermore, CSCs have been reported to acquire drug resistance through increased expression of ABC transporters [52,54]. In fact, we found that the transfection of miR-125b-5p into PLC/PRF5 cells caused MRP3 gene upregulation ( Figure S8). Thus, it was assumed that the increased expression of MRP3 in sorafenib-resistant (PLC/PRF5-R1, PLC/PRF-R2) cells could be mediated by EMT-associated CSCs.
Systemic chemotherapy is recommended for HCC patients with Barcelona Clinic Liver Cancer (BCLC) Stage C. In addition to sorafenib, lenvatinib and a combination of atezolizumab and bevacizumab, are currently recommended as first-line therapy. Regorafenib, cabozantinib, ramucirumab, nivolumab, ipilimumab, and pembrolizumab are recommended as second-line therapy. In contrast, it is recommended that patients with BCLC stage B undergo transcatheter arterial chemoembolization (TACE) as standard care. Although it has been recently reported that TKIs are preferred over TACE to maintain hepatic functional reserve [55], guidelines for choosing among these drugs are lacking. Since the miR-125b-5p-ATXN1 axis has been implicated in sorafenib resistance and EMT, it may be useful as a biomarker for drug selection in systemic therapy and the choice between TACE and TKI therapy as an initial treatment.
To date, there are no available reports investigating the effect of miR-125b-5p on the efficacy of sorafenib in clinical practice. Therefore, our finding needs to be validated clinically, ideally in prospective trials, by comparing the expression status of miR-125b-5p and the effect of sorafenib in patient samples to see if our finding can be validated in real life and determine that it is not simply a cell-line phenomenon, which could lead to miR-125b-5p being established as an effective drug selection marker for patients with HCC.

Conclusions
In conclusion, the present study clarified the mechanism of sorafenib resistance in HCC cells, in which miR-125b-5p suppresses ATXN1 and induces Snail-mediated EMT and CSCs. Moreover, we demonstrated that ATXN1 expression has an impact on the prognosis of patients with HCC.