The human liver cancer cell lines Li-7, Hep3B, HepG2, HLE, HLF, Alex and Huh7 were obtained from the Japanese Cancer Research Resources Bank and transported to our laboratory. The cell lines were authenticated by the cell bank using short tandem repeat PCR. Mycoplasma testing has been done for cell lines used in our experiments. Cells were grown in minimum essential medium (MEM; Gibco-Invitrogen) supplemented with 10% fetal bovine serum (FBS, 533–69545; Wako) and penicillin-streptomycin (100 mg/l; Invitrogen) at 37°C in a humidified atmosphere containing 5% CO₂.

The cell proliferation assays were conducted using CCK-8 according to the manufacturer's instructions. Each cell line (0.5×10⁴) was seeded on 96-well plates and cultured in 100 µl MEM supplemented with 10% FBS. After 24 h, seeded cells were treated with 0, 1, 3, 5, or 10 µM sorafenib that was added to the culture medium. At the indicated time points, the medium was exchanged for 100 µl of MEM with CCK-8 reagent (10 µl CCK-8 and 90 µl MEM). The absorbance of each well was measured at a wavelength of 450 nm using an auto-microplate reader.

We obtained serum samples from 11 patients with advanced HCC who underwent sorafenib therapy at the Kagawa University hospital from 2012 to 2015. All samples were drawn before sorafenib administration (Table I). This study was conducted in accordance with the ethical principles of the Declaration of Helsinki and approved by The Institutional Review Board (IRB) of Kagawa University, Faculty of Medicine (Heisei-22-063). Informed consent to use the clinical data and samples for the present study was obtained from the patients or their relatives. For patients who died and had no relatives listed in their clinical records, we provided opt-out methods for the relatives of the dead participants by publishing a summary of this study on the university website. Ethics approval was obtained from The Ethics Committee of Kagawa University Faculty of Medicine.

Total miRNA was extracted from liver cancer cell lines using the Qiagen miRNeasy kit (Qiagen K.K.) according to the instructions provided by the manufacturer. Each serum sample was processed for total RNA extraction with the miRNeasy Mini Kit (Qiagen) according to the manufacturer's instructions. The RNA sample from both sets typically showed A_260/280 ratios between 1.9 and 2.1 measured with an Agilent 2100 Bioanalyzer (Agilent Technologies).

After RNA quantification with an RNA 6000 Nano kit (Agilent Technologies), the samples were labeled using a miRCURY Hy3/Hy5 Power labeling kit and hybridized on a human miRNA Oligo chip10, version 14.0 (Toray Industries). Scanning was conducted with the 3D-Gene Scanner 3000 (Toray Industries). The 3D-Gene extraction version 1.2 software (Toray Industries) was used to read the raw intensity of the image. The raw data were analyzed with GeneSpringGX v 10.0 (Agilent Technologies) to determine the change in miRNA expression. The samples were frozen at −80°C within 4 h of collection and thawed just before analysis.

Total miRNA was extracted from liver cancer cells using the QIAGEN miRNeasy kit (Qiagen K.K.) for miRNA quantification according to the instructions provided by the manufacturer. Total RNA was reverse-transcribed using TaqMan MicroRNA Reverse Transcription kit (Life Technologies Japan). RNU6B was used as an internal control for relative quantification of hsa-miR-30d.

Circulating miRNA was extracted from 200 µl of serum samples using the Qiagen miRNeasy serum-plasma kit (Qiagen K.K.) according to the instructions provided by the manufacturer. RNA was reverse-transcribed using the TaqMan MicroRNA Reverse Transcription kit (Life Technologies Japan) following the manufacturer's instructions. Caenorhabditis elegans miR-39 (cel-miR-39) was spiked in each sample as a control for the extraction and amplification steps. Serum miRNA was amplified using primers and probes provided by Applied Biosystems by the TaqMan MicroRNA assay, according to the instructions provided by the manufacturer. The relative expression of serum miRNA was calculated using the comparative cycle quantification (Cq) method (2^−ΔΔCq) (18) with spiked cel-miR-39 as a normalized internal control.

Replicate data were consolidated for each sample group as follows; differences between the effective and non-effective groups i) in vivo and ii) in vitro. The data were organized using the hierarchical clustering and ANOVA functions in the GeneSpring software. Hierarchical clustering was performed using the clustering function (condition tree) and Euclidean correlation as a distance metric. Two-way ANOVA with Bonferroni's correction and asymptotic P-value (<0.05) computation without any error correction on the samples were conducted to search for the miRNAs with the most prominent variation across the different groups. All analyzed data were scaled by global normalization. The statistical significance of differentially-expressed miRNAs was analyzed by Mann-Whitney U test. All analyses were conducted using computer-assisted JMP8.0 (SAS Institute). Paired analysis between the groups was conducted using the Student's t-test and Fisher's exact test. P<0.01 (Cluster analysis in vitro) and P<0.05 (Cluster analysis in vivo) were considered to indicate a significant difference between groups.

We examined the in vitro effects of sorafenib on seven human liver cancer cell lines, Li-7, Hep3B, HepG2, HLE, HLF, Alex and Huh7 by following the course of proliferation of each of them for three days after sorafenib addition and evaluating the direct relationship between the decrease of cell viability and the inhibition of cell proliferation. Cells were cultured and with 0, 1, 3 or 10 µM sorafenib was added to the medium after 24 h. As shown in Fig. 1, sorafenib strongly inhibited the proliferation of Li-7, Hep3B, and HepG2 cells in a dose- and time-dependent manner (sorafenib-effective group), but not that of HLE, HLF, and ALEX cells (sorafenib non-effective group). Huh7 cell growth was partially inhibited by sorafenib (Fig. 1), so it was not included in either group. These results demonstrated that sorafenib inhibits the proliferation of certain human liver cancer cell lines.

A miRNA array was performed in all liver cancer cell lines before sorafenib administration, and the results of the sorafenib-effective group were compared to those of the non-effective group. We examined the expression patterns of 2555 miRNAs extracted from the cell lines. The unsupervised hierarchical clustering analysis, using Pearson's correlation, showed that the sorafenib-effective group formed a cluster separated from that of the non-effective group (Fig. 2). Furthermore, 89 miRNAs were significantly differentially expressed between the groups (Fig. 2).

We included the serum samples of 11 patients (9 males and 2 females with a median age of 69 years, ranging between 53 and 79 years) in the miRNA analysis. All patients received sorafenib therapy. The parameter association between the effective and non-effective groups is summarized in Table I. The analysis of age, sex, etiology, tumor stage, Child-Pugh classification, vessel invasion, reason of sorafenib administration, initial sorafenib dose, and primary or recurrent cancer were conducted using various categories of the Student's t-test and Fisher's exact test. No statistically significant differences were observed between the effective and non-effective groups regarding patient background.

We also performed miRNA array using serum samples from patients with HCC and compared the results of the sorafenib-effective and non-effective groups (Fig. 3). The unsupervised hierarchical clustering analysis, using Pearson's correlation, showed that the effective group formed a cluster separated from that of the non-effective group. Additionally, 10 miRNAs had significantly different expression patterns in the sorafenib-effective group when compared to the non-effective group.

We detected 3 miRNAs that were significantly changed between the sorafenib-effective and non-effective groups jointly in the in vitro and in vivo experiments (Table II). The hsa-miR-296-5p was up-regulated in the effective groups of both the liver cancer cell lines and the serum samples of HCC patients. The hsa-miR-6729-5p was down-regulated in the effective groups of both sets of samples. In contrast, hsa-miR-30d was down-regulated in the effective group of the liver cancer cell lines but up-regulated in the serum samples (Table II). Among 3 miRNAs, hsa-miR-30d might not be leaked, but actively secreted from the cancer cell.

To determine if hsa-miR-30d was secreted from the liver cancer cell to the extracellular fluid, we compared its expression levels in the serum of patients with HCC before sorafenib administration in the sorafenib-effective and non-effective groups using real-time quantitative (RTq-PCR). hsa-miR-30d expression was up-regulated in the serum and exosomes of HCC patients of the effective group when compared to that of the non-effective group (*P=0.0065 and ^#P=0.0464, respectively, Fig. 4A and B). We also quantified hsa-miR-30d expression in the medium of liver cancer cells by RTq-PCR and found that HCC cells of the effective group up-regulated hsa-miR-30d when compared to the non-effective group (^$P=0.0021, Fig. 4C).