In this study, a tissue microarray was used containing 138 paired paraffin-embedded ICC tissue samples and corresponding peritumoral tissues samples obtained from patients who had undergone hepatic resection at Zhongshan Hospital, Fudan University from 2007 to 2012. The use of these tissue specimens was approved by the Zhongshan Hospital Research Ethics Committee and written consent was obtained from the patients. The detailed clinicopathological characteristics of the patients are presented in Table I.

Three ICC cell lines, QBC939, HCCC-9810 and RBE, were obtained from the Institute of Biochemistry and Cell Biology of the Chinese Academy of Sciences and maintained in RPMI-1640 medium supplemented with 10% fetal bovine serum, penicillin (100 U/ml) and streptomycin (100 µg/ml) (all Gibco; Thermo Fisher Scientific) under 95% air and 5% CO₂ at 37°C.

A tissue microarray that included 138 ICC tissue samples with corresponding adjacent tissue samples was constructed by Shanghai Outdo Biotech Co., Ltd. (13). Briefly, 4-µm sections were obtained from the TMA blocks using a microtome Shanghai Outdo Biotech Co., Ltd. The 4-µm sections on the TMAs were dried at 60°C for 4 h, de-paraffinized in xylene I for 10 min, de-paraffinized in xylene II for another 10 min, and then rehydrated in graded ethanol (100% ethanol for 5 min, 90% ethanol for 5 min, 80% ethanol for 5 min and 70% ethanol for 5 min). Ethanol was subsequently removed by washing with phosphate-buffered saline (PBS) for 15 min. For immunohistochemistry, the sections were incubated in 3% H₂O₂ at room temperature for 20 min to block endogenous peroxidases. The TMAs were then microwaved in EDTA [1:500; pH 8.0; Absin (Shanghai) Biotechnology Co., Ltd.] for 15 min for antigen retrieval and subsequently incubated with 1% bovine serum albumin [1:500; pH 8.0; Absin (Shanghai) Biotechnology Co., Ltd.]. Subsequently, TMAs were incubated with the following rabbit anti-human primary antibodies for 12 h at 4°C: Anti-GLS1 (1:200; cat. no. ab93434; Abcam), anti-E-cadherin (1:200, cat. no. 24E10; Cell Signaling Technology, Inc.) and anti-Vimentin (1:150; cat. no. D21H3; Cell Signaling Technology, Inc.). Subsequently, TMAs were incubated with horseradish peroxidase-labeled secondary antibodies [1:1,000; cat. no. abs957; Absin (Shanghai) Biotechnology Co., Ltd.] for 2 h at 37°C. DAB [1:500; Absin (Shanghai) Biotechnology Co., Ltd.] was used as a detection reagent and was incubated with the TMAs for 1.5 min, after which, the samples were counterstained with hematoxylin for 2 min at room temperature, and were subsequently dehydrated in a gradient series of ethanol. Images of the sections were subsequently captured under a light microscope (Olympus BX-51; Olympus Corporation). All specimens were reviewed and scored by investigators blinded to the clinical characteristics of the patients. The expression of GLS1, Vimentin and E-cadherin was scored as follows: 0, no staining; 1, weak staining; 2, moderate staining; and 3, strong staining. The percentage of positively stained cells was scored as follows: 0, 0–10%; 1, 1–25%; 2, 26–50%; 3, 51–75%; and 4, >75%. A final score of 0–2 was considered as negative expression, whereas a score of 3–12 was considered as positive expression.

For H&E staining, sections were stained with hematoxylin solution (0.2%) for 4 min, followed by eosin solution (0.5%) for 90 sec at room temperature.

Cell and tissue proteins were extracted using radioimmunoprecipitation assay buffer (cat. no. P0013C; Beyotime Institute of Biotechnology), and the protein concentration was measured using an enhanced Bicinchoninic Acid Protein Assay kit (cat. no. P0010; Beyotime Institute of Biotechnology). Proteins were loaded at 20 µg/lane, separated by 10% SDS-PAGE (cat. no. P0012A; Beyotime Institute of Biotechnology) and were transferred to polyvinylidene fluoride membranes (EMD Millipore) for western blot analysis. Subsequently, the membranes were blocked with TBS containing 0.1% Tween 20 and 5% non-fat milk for 2 h at room temperature, and were subsequently incubated with the following rabbit primary antibodies at 4°C for 12 h: Anti-GLS1 (1:1,000; cat. no. ab93434; Abcam), anti-E-cadherin (1:1,000; cat. no. 24E10; Cell Signaling Technology, Inc.), anti-Vimentin (1:500; cat. no. D21H3; Cell Signaling Technology, Inc.) and anti-GAPDH (1:1,000; cat. no. D16H11; Cell Signaling Technology, Inc.). The membranes were then rinsed and incubated with secondary antibody (1:5,000; cat. no. A0208; Beyotime Institute of Biotechnology) at room temperature for 2 h. Densitometric analysis using an enhanced chemiluminescence system (EMD Millipore) and ImageJ software (version 1.49; National Institutes of Health) was performed to detect protein expression.

siRNAs and a pcDNA plasmid that can regulate the human GLS1 gene were obtained from Shanghai Genomeditech Co. Two siRNAs were designed to silence GLS1, and the cells were randomly divided into the siRNA1-transfected cell group; siRNA2-transfected cell group; the negative control (NC) group, which was transfected with non-targeting siRNA; and the mock group, which consisted of untransfected cells. A pc-DNA3.1 plasmid was designed to induce overexpression of GLS1, and the cells were divided into the pc-DNA3.1-GLS1-transfected cell group; the NC group, which was transfected with an empty pc-DNA3.1 vector; and the mock group, which consisted of untransfected cells. The siRNA sequences were as follows: siRNA1, upstream 5′-CCAGGUUGAAAGAGUGUAUTT-3′, downstream 5′-AUACACUCUUUCAACCUGGTT-3′; siRNA2, upstream 5′-CCCUGAAGCAGUUCGAAAUTT-3′, downstream 5′AUUUCGAACUGCUUCAGGGTT-3′; NC, upstream 5′-UUCUCCGAACGUGUCACGUTT-3′, downstream 5′-ACGUGACACGUUCGGAGAATT-3′. Briefly, cells were seeded at a density of 1×10⁶ cells/well in 6-well plates and were incubated until 70% confluence was reached. siRNAs and pc-DNA3.1 plasmids were transfected into the QBC939 and RBE cells using Lipofectamine^® 2000 (Invitrogen; Thermo Fisher Scientific, Inc.) according to the manufacturer's instructions. After 72 h, the cells were harvested for total protein extraction and examined by western blot analysis.

Cell migration was assessed using a wound-healing assay. ICC cells were plated in a 6-well plate and were incubated until they reached 100% confluence. Following serum starvation for 1 day, cells were wounded with a 200-µl plastic tip. Cells were washed three times with sterile PBS to remove the floating cells and were then incubated for 24 h at 37°C under 95% air and 5% CO₂. The same wound areas were observed and images were captured under a light microscope (Olympus IX71; Olympus Corporation) at 0 and 24 h. The migratory abilities were quantified by measuring the percentage of the migration area of cells in the scratched regions, as follows: 0 h scratch area-24 h scratch area/0 h scratch area ×100%.

A Transwell assay was conducted to assess the invasive ability of cells in response to GLS1 overexpression or knockdown. The upper surface of the Transwell filter used in the assay was coated with Matrigel. Cells (1×10⁵) suspended in 150 µl serum-free medium were added to the Transwell chamber (cat. no. 3413; Corning, Inc.), which was placed into a 24-well plate containing complete medium. After 24 h of incubation at 37°C, the filter was extracted and cells on the upper surface of the filter were removed with cotton swabs. Cells on the underside of the Transwell filter were fixed with 4% paraformaldehyde for 25 min and stained with 0.1% crystal violet for 15 min at 37°C, after which, images were captured (Olympus IX71; Olympus Corporation) and the number of cells was quantified.

Statistical analyses were performed using SPSS 23.0 (IBM Corp.) and GraphPad Prism7 (GraphPad Software, Inc.) software. The Student's t-test was used to compare differences between groups. One-way analysis of variance (ANOVA) was used to compare differences among groups. Spearman's correlation analysis was used for the correlation analysis and Fisher's exact test was used to determine the association of GLS1 with ICC characteristics. OS and the cumulative recurrence rate were determined using Kaplan-Meier survival curves and the log-rank test. Independent prognostic factors were evaluated by with Cox proportional hazards model. A P-value <0.05 was considered to indicate a statistically significant difference.

First, data from the Oncomine database were used to analyze GLS1 transcripts in 3 types of digestive system tumors. As shown in Fig. 1A, the mRNA levels of GLS1 in the tumor tissues samples were higher than those in the paired peritumoral tissues samples, including in the liver cancer, colorectal cancer and gastric cancer samples (14–16). However, the expression of GLS1 in ICC tissues remains unclear. Subsequently, the GLS1 protein levels in the ICC tissue samples and matched peritumoral specimens were examined by western blot analysis and immunohistochemistry. The results revealed that the GLS1 protein levels were higher in the ICC tumor tissue samples than in the adjacent normal tissue samples (P<0.001; Fig. 1B and D) and found that GLS1 was mainly expressed in the cytoplasm (Fig. 1C).

To further examine the role of GLS1 in ICC cell invasion and migration, the GLS1 protein levels were examined in a panel of ICC cell lines (QBC939, HCCC-9810 and RBE) by western blot analysis (Fig. 2A). The data indicated that the QBC939 cells exhibited the highest expression of GLS1 and that the RBE cells exhibited the lowest expression level of GLS1 among all the ICC cell lines examined. Therefore, the QBC939 cell line and RBE cell line were used in the subsequent experiments. The results of western blot analysis revealed that GLS1 expression was markedly knocked down in the GLS1-siRNA-2-transfected cell lines (Fig. 2B). Therefore, GLS1-siRNA-2 was used in the subsequent experiments. GLS1 was also overexpressed in the RBE cells (Fig. 1C). Subsequently, a Transwell assay revealed that the GLS1-siRNA-2-transfected cells exhibited a significantly lower rate of invasion than the GLS1-NC cells (Fig. 2D). A wound-healing assay also indicated that the downregulation of GLS1 decreased migration of the GLS1-siRNA-2 ICC cells compared with that of the GLS1-NC cells at 24 h (Fig. 2E). However, compared with the NC cells, the RBE cells with an upregulated GLS1 expression exhibited increased invasion and metastasis (Fig. 2D and E). These data indicate that GLS1 positively regulates the migratory and invasive abilities of ICC cells.

Recently, a previous study demonstrated that GLS1 reduces cell-cell contact and increases cell motility by inducing EMT in lung cancer cells (12). Thus, in this study, the expression of two EMT-related markers, Vimentin and E-cadherin, was determined by western blot analysis. This experiment indicated that the expression of Vimentin was significantly reduced, whereas the levels of E-cadherin were significantly increased following the knockdown of GLS1 expression in QBC939 cells (P<0.001; Fig. 3A). However, the overexpression of GLS1 in the RBE cells yielded opposite results (P<0.001; Fig. 3B). Subsequently, using an ICC tissue array, immunohistochemistry was performed to simultaneously analyze GLS1, Vimentin and E-cadherin expression (Fig. 3C). The results indicated that in the human ICC tissue, the GLS1 levels were positively associated with Vimentin expression and negatively associated with E-cadherin expression. In addition, Spearman's correlation analysis also revealed that the GLS1 levels positively correlated with Vimentin expression (r²=0.3175; P<0.001) and negatively correlated with E-cadherin expression (r²=−0.1061; P<0.001; Fig. 3D). These data indicate that the expression of GLS1 is essential for the EMT process and for the progression of ICC.

The samples were classified into 2 groups, a GLS1^high (++, moderate; +++, strong) group and a GLS1^low (−, absent; +, weak) group, according to the mean value of the expression of GLS1 in the tumor tissue samples (Fig. 4A-D). The GLS1^high group accounted for 54.3% (n=75) of the samples, and the GLS1^low group accounted for 45.7% (n=63). We found that the overexpression of GLS1 was associated with malignant phenotypic features, such as lymphatic metastasis (P=0.029) and poor tumor differentiation (P=0.001) (Table I). By contrast, other clinicopathological characteristics including age, microvascular invasion, tumor size and number were not associated with GLS1 expression. The GLS1^high expression group was also found to be associated with a worse OS time than the GLS1^low expression group (P<0.001; Fig. 4E). The 2- and 5-year OS rates in the GLS1^high group were significantly lower than those in the GLS1^low group (21.1 vs. 53.3% and 14.5 vs. 35.5%, respectively). The 2- and 5-year cumulative recurrence rates were also markedly higher in the GLS1^high group compared with the GLS1^low group (81.5 vs. 55.5% and 88.1 vs. 66.1%, respectively; Fig. 4E). Univariate analysis revealed that GLS1 expression, tumor size, tumor number, embolus and lymphatic metastasis were significantly associated with OS and cumulative recurrence rate in patients with ICC (Table II). In addition, multivariate analysis revealed that GLS1 expression was an independent predictor of OS (P<0.001) and cumulative recurrence (P<0.001) (Table II).