A total of 78 fresh cancerous tissues and their matched adjacent non-cancerous controls (≥3 cm from the tumor margin) were obtained from patients with HCC who had undergone routine curative surgical removal of the tumor at The First Affiliated Hospital of Wenzhou Medical University (Zhejiang, China) between January 2011 and July 2012. The histological diagnoses of the HCC specimens were independently confirmed by two senior pathologists at the First Affiliated Hospital of Wenzhou Medical University (Wenzhou, China), and the relevant clinicopathological information and follow-up data were retrieved from patient hospital records. Some of the surgically resected specimens were stored in liquid nitrogen at −80°C for further reverse transcription-quantitative polymerase chain reaction (RT-qPCR) or western blot analysis. The remaining specimens were fixed in a fixation solution containing 4% paraformaldehyde for 24 h at 4°C and were embedded in paraffin blocks for immunohistochemical analysis. Ethical approval was obtained from the Ethics Committee of The First Affiliated Hospital of Wenzhou Medical University, and written informed consent was obtained from all patients. Clinical specimens were collected from 67 males and 11 females aged between 30 and 83 years (mean, 57 years). Tumor size was classified according to the maximum size of the tumor detected by magnetic resonance imaging. Tumors were staged between I and IV, according to the latest edition of Tumor-Node-Metastasis (TNM) classification system of the American Joint Committee Cancer and Union for International Cancer Control (31). Tumor differentiation was assigned using the World Health Organization classification and grading system (32,33). The detailed clinicopathological data regarding the specimens are summarized in Table I.

Total RNA was extracted from tumor tissues and adjacent non-cancerous tissues using the E.Z.N.A miRNA kit (Omega Bio-Tek, Inc., Norcross, GA, USA), according to the manufacturer's protocol. miRNA expression analysis, synthesis of first strand cDNA and RT-qPCR were performed using the All-in-One™ miR RT-qPCR detection kit (GeneCopoeia, Inc., Rockville, MD, USA), according to the manufacturer's protocol. The reverse transcriptase reaction mixture (25 µl), containing 2 µg total RNA, was applied by incubating mixtures at 37°C for 60 min, 85°C for 5 min, and 4°C for 30 min. The 20 µl PCR reaction mixture contained 1.2 µl RT product, 10 µl 2× All-in-One qPCR Master mix, 2 µl of each primer and 4.8 µl ddH₂O. The primer sequences for miR-638 (cat. no. HmiR0295) and the reference gene small nuceloular RNA, C/D box 44 (RNU44; cat. no. HmiRQP9011) were designed and purchased from GeneCopoeia Inc. Following an initial denaturation step at 95°C for 10 min, the PCR samples were run for 40 cycles of 95°C for 10 sec, 60°C for 20 sec and 72°C for 10 sec. The relative expression of miR-638 was calculated using the 2^−∆∆Cq method (34), based upon the quantification cycle (Cq) method with RNU44 small nuclear RNA molecule as an endogenous reference.

For the measurement and quantification of SOX2 mRNA, cDNA was synthesized using the Hiscript^® Q RT SuperMix for qPCR (Vazyme, Piscataway, NJ, USA). The PCR reaction (20 µl) contained 2 µl reverse transcriptase product, 10 µl AceQ^® qPCR SYBR^®-Green Master mix (Vazyme), and 0.4 µl of each primer. The PCR samples were subsequently incubated at 95°C for 3 min, followed by 49 cycles at 95°C for 15 sec, 60°C for 1 min and 50°C for 30 sec. Relative expression levels of SOX2 were calculated based on the 2^−∆∆Cq method and GAPDH was used as an internal control. All PCR analyses were performed using a CFX96^™ Real-Time PCR detection system (Bio-Rad Laboratories, Inc., Hercules, CA, USA).

The primers for SOX2 and GAPDH were as follows: SOX2 forward, 5′-CGAGATAAACATGGCAATCAAAAT-3′; and reverse, 5′-AATTCAGCAAGAAGCCTCTCCTT-3′; GAPDH forward, 5′-TGCACCACCAACTGCTTAGC-3′ and reverse, 5′-GGCATGGACTGTGGTCATGAG-3′. The primer sequences for miR-638 (cat. no. HmiR0295) and RNU44 (cat. no. HmiRQP9011) were purchased from GeneCopoeia Inc.

Briefly, paraffin-embedded tissue blocks were sectioned (5 µm thick) using a microtome, transferred onto tissue anti-off slides and heated at 60°C for 4 h. Tissue sections were dewaxed with dimethylbenzene and rehydrated in a descending ethanol series (100, 95, 85 and 75%). Endogenous peroxidase activity was blocked with 0.3% hydrogen peroxide for 10 min at room temperature, and the sections were subsequently boiled by microwave in 0.01 mol/l citrate antigen retrieval solution (pH 6.0) for 15 min at 100°C. Following 3 washes with phosphate-buffered saline, the sections were blocked with 10% goat serum (cat. no C0265; Beyotime Institute of Biotechnology, Haimen, China) in phosphate-buffered saline (PBS) for 15 min at 37°C, and incubated with a rabbit anti-human SOX2 polyconal antibody (dilution, 1:200; cat. no. ab97959; Abcam, Cambridge, UK) in a humidified chamber overnight at 4°C. The sections were subsequently incubated with a horseradish peroxidase (HRP)-labeled goat anti-rabbit secondary antibody (dilution, 1:100; cat. no. pv-6001; OriGene Technologies, Inc., Rockville, MD, USA) for 30 min at 37°C and washed with PBS to remove excess antibody. The sections were stained with 3,3′-Diaminobenzidine tetrahydrochloride for 1 min at room temperature and observed using an optical microscope at a magnification, ×40. Finally, the sections were counterstained with hematoxylin (OriGene Technologies, Inc.) for 3 min at room temperature, dehydrated, cleared, mounted and examined. In the present study, a PBS-only stained liver sample was used as a negative control, and human gastric cancer tissue was used as a positive control.

SOX2 expression in tumor tissues was semi-quantitatively evaluated and scored on the basis of the staining extent of positively stained cells, as described previously by Chen et al (24). Only nuclei stained brown were defined as SOX2-positive. For all samples, the staining intensity and percentage of positive tumor cells were evaluated and classified under double-blind conditions. Briefly, the staining intensities were scored as follows: 0, negative staining; 1, weak staining; 2, medium staining; or 3, strong staining. The staining percentage score (positively-stained cells/total tumor cells ×100%) was defined as follows: 0, 0; 1, 1–9; 2, 10–29; 3, 30–49; or 4, 50–100%. Consequently, a final semi-quantitative immunoreactivity score (IRS) of SOX2 staining was obtained by multiplying the staining percentage score by the staining intensity score, resulting in scores ranging between 0 and 12. For statistical analyses, tumors with a final IRS of <5 were categorized as a low protein expression group and those with a final IRS of ≥6 were categorized as a high protein expression group.

The cancerous liver tissue samples and their adjacent non-cancerous tissue samples were homogenized using a radioimmunoprecipitation acid lysis buffer kit (Beyotime Institute of Biotechnology), and the lysates were subsequently centrifuged at 13,400 × g for 15 min at 4°C. The total protein concentration was measured by using the BCA method (Beyotime Institute of Biotechnology). Protein samples (40 µg of each sample) were separated by 12% SDS-PAGE, prior to being transferred onto polyvinylidene difluoride membranes. Following blocking with 5% skimmed milk buffer for 120 min at room temperature, the membrane was incubated with anti-SOX2 (dilution, 1:2,000; cat. no. ab97957; Abcam, Cambridge, UK) and GAPDH (dilution, 1:2,000; cat. no. 5174; Cell Signaling Technology, Inc., Danvers, MA, USA) primary antibodies overnight at 4°C. Following three washes with 0.1% TBST for 5 min each time, the membranes were incubated with a HRP-conjugated goat anti-rabbit IgG secondary antibody (dilution, 1:5,000; cat. no 774P2; Cell Signaling Technology, Inc.) for 2 h at room temperature, prior to being washed again with 0.1% TBST 3 times. Finally, the protein complexes on the band were detected using a SuperSignal Western Femto kit (Pierce; Thermo Fisher Scientific, Inc., Waltham, MA, USA), and were subsequently quantified using the ChemiDoc XRS⁺ system with Image Lab™ 3.0 Software (cat. no. 1708265; Bio-Rad Laboratories, Inc.).

Data are presented as the mean ± standard deviation, and were analyzed using SPSS 20.0 software (IBM Corp., Armonk, NY, USA) and GraphPad Prism 5.01 statistical software (GraphPad Software, Inc., La Jolla, CA, USA). The Pearson χ² and Fisher's exact tests were performed in order to test the significance of observed clinicopathological variables among different groups. Student's t-test was performed to compare the miR-638 and SOX2 expression levels between different groups. Comparisons among multiple groups were performed using one-way analysis of variance and a Tukey's test. The association between miR-638 and SOX2 in the matched HCC tumor specimens was determined using Pearson's correlation analysis, and r-values represent the Pearson's correlation coefficient. The Kaplan-Meier method, followed by the log-rank test, and Cox proportional hazards regression model analysis were performed in order to plot survival curves and to calculate the hazard ratios (HRs) and the 95% confidence intervals (CIs), respectively. P<0.05 was considered to indicate a statistically significant difference.

With regards to tumor size, 51 cases (65.4%) were classified as small (<5 cm in maximum diameter) and 27 cases (34.6%) were classified as large (≥5 cm in maximum diameter). With respect to tumor grades, 31 cases (39.7%) were categorized as well-differentiated, 31 cases (39.7%) as moderately-differentiated, and 16 cases (20.6%) as poorly-differentiated (Table I). In terms of clinical TNM stage, 48 cases (61.5%) exhibited TNM stage I–II disease and 30 cases (38.5%) exhibited TNM stage III–IV disease. With regards to tumor vascular invasion, 21 cases (26.9%) exhibited portal vascular invasion and 57 cases (73.1%) exhibited no vascular invasion (Table I). There were 61 cases (78.2%) of hepatitis B (HBV) positivity and 9 cases (11.5%) of distant metastasis. The overall survival period was defined as the time period between the date of surgical resection of HCC to the last follow-up or patient mortality. Mortalities caused by other events were considered to be censored (Table I).

To assess miR-638 and SOX2 expression in HCC tumor tissues and paired adjacent normal tissues, RT-qPCR was performed. In the present study, it was demonstrated that miR-638 expression in tumor tissues was significantly lower than that in paired pericancerous healthy tissues (Fig. 1A; P<0.001). However, SOX2 mRNA expression was significantly upregulated (Fig. 1B; P<0.001). Furthermore, miR-638 expression in HCC tissues were negatively correlated with that of SOX2 mRNA (r=−0.675; Fig. 1C; P<0.001). The median expression of miR-638 and SOX2 were used as cut-off points to divide 78 samples into two groups. For statistical analysis, samples with miR-638 expression levels equal to or above the cut-off point were categorized as the high expression group (n=39) and samples with miR-638 expression levels below the cut-off point were categorized as the low expression group (n=39).

In order to study the association between miR-638 and SOX2 protein in HCC, SOX2 protein levels were measured in 78 pairs of tumor tissues and their adjacent non-cancerous tissues using western blot and immunohistochemical analyses. Western blot analysis demonstrated that SOX2 protein levels were significantly increased in 15 paired HCC samples, compared with adjacent non-cancerous samples (Fig. 2A; P<0.001). Immunohistochemical staining results revealed that SOX2-positive staining was only evident in the nuclei of tumor cells but not in peritumoral tissues. The representative images of SOX2 immunostaining in HCC tissues are presented in Fig. 2B. Of the 78 samples, 45 (57.7%) displayed high expression of SOX2 in tumor tissues compared with 21 (26.9%) paracancerous tissues samples exhibiting high expression (χ²=15.127, Table II; P<0.001). Furthermore, in order to statistically analyze the correlation between miR-638 and SOX2, a cross analysis was performed and is presented in Table III (χ²=15.551; P<0.0001), and the results revealed that the semi-quantitative immunoreactivity scores (IRS) of SOX2 staining in HCC tissues were negatively associated with miR-638 expression level (one-way analysis of variance, P<0.05; r=−0.478, P<0.001; Fig. 2C). The aforementioned results indicated that increased SOX2 expression in HCC may be due to miR-638 underexpression, suggesting a potential functional link between these two molecules.

To observe whether miR-638 and SOX2 expression levels were correlated with clinicopathological features of HCC, clinicopathological analysis was performed (Table I). The data indicated that miR-638 and SOX2 were not significantly correlated with age, sex, tumor size, tumor number, serum a-fetoprotein (AFP) level, tumor differentiation (all P>0.05). By contrast, a significant association was observed between miR-638 or SOX2 expression and TNM staging (P=0.001 and 0.002, respectively). It was also demonstrated that patients with HCC with lower miR-638 expression and higher SOX2 expression were more likely to be at a higher portal vascular rate (P=0.005 and 0.012, respectively). The results from the present study suggested that miR-638 and SOX2 expression may serve a vital role in HCC progression and may have potential as novel prognosis biomarkers for HCC.

In order to further evaluate whether there was an association between miR-638 or SOX2 expression and the overall prognosis of patients with HCC, the Kaplan-Meier method and Cox regression analysis were performed. Kaplan-Meier survival analysis suggested that miR-638 expression was significantly associated with a higher overall survival rate [median survival time (determined from the date of operation to the last visit or death), 50.20 vs. 19.10 months; Fig. 3A; P<0.001]. Consistently, the overall survival rate of patients with exhibiting SOX2-positive expression was significantly lower than in those with SOX2-negative expression (median survival time, 19.80 vs. 50.75 months; Fig. 3B; P<0.001). Furthermore, Univariate Cox proportional hazards regression model analysis demonstrated that tumor size, tumor number, HBV status, TNM stage, portal vascular invasion, distant metastasis, and miR-638 and SOX2 expression were statistically significant risk factors for the overall survival of patients with HCC (Table IV). However, age, sex, cirrhosis, serum AFP level and tumor differentiation had no prognostic value. Multivariate analysis, stratified for the known prognostic variables, revealed that portal vascular invasion (P=0.029), expression of SOX2 (P=0.013) and expression of miR-638 (P=0.008) were independent prognostic factors of HCC. These observations suggested that miR-638 and SOX2 may serve as novel prognostic biomarkers for patients diagnosed with HCC.