Screening differential circular RNA expression profiles reveals hsa_circ_0004018 is associated with hepatocellular carcinoma

Circular RNAs (circRNAs) have been emerged as an indispensable part of endogenous RNA network. However, the expression significance of circRNAs in hepatocellular carcinoma (HCC) is rarely revealed. The aim of this study was to determine the circRNA expression profile in HCC, and to investigate their clinical significances and relevant mechanisms for cancer progression. The global circRNA expression profile in HCC was measured by circRNA microarray. Levels of one representative circRNAs, hsa_circ_0004018, were confirmed by real-time reverse transcription-polymerase chain reaction. The expression levels of hsa_circ_0004018 in HCC were significantly lower compared with para-tumorous tissue (P<0.001). Our data further showed that lower expression of hsa_circ_0004018 was correlated with serum alpha-fetoprotein (AFP) level, tumor diameters, differentiation, Barcelona Clinic Liver Cancer stage and Tumor-node-metastasis stage. More importantly, we detected liver tissues from chronic hepatitis, cirrhosis and HCC patients; and found that hsa_circ_0004018 harbored HCC-stage-specific expression features in diverse chronic liver diseases (P<0.001). The area under receiver operating characteristic curve was up to 0.848 (95% CI=0.803–0.894, P<0.001). The sensitivity and specificity were 0.716 and 0.815, respectively. Finally, hsa_circ_0004018 might be involved in cancer-related pathways via interactions with miRNAs.


INTRODUCTION
Globally, hepatocellular carcinoma (HCC) is the most common type of hepatic malignancies, accounting for approximately 90% of primary liver cancer. It ranks as the second most significant cause of cancer-related deaths in men, 50% of the cases and deaths occurred in China [1]. It is disappointing that most HCC patients were diagnosed at advanced stages with metastasis, missing the best opportunity for curative therapy, such as resection, transplantation or ablation [2]. The early diagnosis is urgent for prognosis of HCC [3]. The risk of HCC increases with liver fibrosis stages [4]. Many studies revealed that the intimate relationship between cirrhosis and HCC [5][6][7][8]. In general, anyone with cirrhosis should be screened for HCC [9].
In recent years, circular RNAs (circRNAs) have emerged as a new star in noncoding RNA (ncRNA) world, representing a class of endogenous RNAs existing in mammalian cells and featuring stable structure and high cell-type-specific, tissue-specific and developmentalspecific expression [10]. By interacting with microRNAs (miRNAs) or other molecules, circRNAs regulate gene expression at the transcriptional or post-transcriptional level [10,11]. Compared to linear RNAs, circRNAs have the outstanding feature of non-canonical splicing without a free 3' and 5'end, which enables them to resist RNA exonucleases [10][11][12]. As a result, they might be suitable www.impactjournals.com/oncotarget/ Oncotarget, 2017, Vol. 8, (No. 35), pp: 58405-58416 Research Paper www.impactjournals.com/oncotarget as potential biomarkers and targets for novel therapeutic approaches for human diseases. Recently, researchers have found that circRNAs are linked to several cancers, such as gastric cancer, colorectal cancer, HCC, pancreatic ductal adenocarcinoma and ovarian cancer [13][14][15][16].
Since the global circRNA expression profile in HCC is not fully uncovered, in the present study, we explored the circRNA expression profile in HCC. We identified 527 differentially expressed circRNAs (including 174 upregulated and 353 downregulated genes) in HCC tissues compared with para-tumorous tissues. And then, we focused on hsa_circ_0004018, one of the most downregulated circRNAs in microarray detection (fold change=294.86, P value=4.7164E-07). It is transcribed from SMYD4 (SET and MYND domain containing 4) on chromosome 17.

Overview of circRNA profiles in HCC tissues
Our microarray analysis revealed the circRNA expression profiles in five paired human HCC and paratumorous tissues (GEO No. 94508: https://www.ncbi.nlm. nih.gov/geo/query/acc.cgi?acc=GSE94508). The box plot is a direct way to rapidly visualize the distributions of a dataset for the circRNAs profiles. After normalization, the distributions of log2 ratios among ten samples are nearly the same ( Figure 1A). Differentially expressed circRNAs with statistical significance (fold changes ≥2.0 and P <0.05) between groups were identified by volcano plot filtering ( Figure 1B). The circRNA expression patterns between HCC tissues and para-tumorous tissues were found to be significantly different ( Figure 1C). A total of 527 circRNAs, whose expression change were more than twofold, were found. Among them, 174 and 353 were up and down expressed in tumor tissues, respectively. The number of down-expressed circRNAs was bigger than that of up-expressed circRNAs. We selected hsa_circ_0004018 for further study.

Correlations between hsa_circ_0004018 expression levels and clinicopathological parameters in HCC patients
Then, we analyzed the relationship between the expression levels of hsa_circ_0004018 and clinicopathological factors of patients with HCC. As Table 1 indicates, hsa_circ_0004018 level was correlated with serum alpha-fetoprotein (AFP) level, tumor diameters, differentiation, Barcelona Clinic Liver Cancer (BCLC) stage and Tumor-node-metastasis (TNM) stage.

Dynamic changes of hsa_circ_0004018 levels in groups from chronic hepatitis to cirrhosis to HCC
To confirm the results of hsa_circ_0004018 in HCC patients ( Figure 2B), we selected cirrhosis and the chronic hepatitis biopsy tissues. Intriguingly, we revealed that the levels of hsa_circ_0004018 in HCC tissues were significantly lower than those of cirrhosis (F=4) (P<0.001); and its levels in cirrhosis tissues were significantly lower than those in chronic hepatitis tissues (F=0-3) (P<0.001). Hsa_circ_0004018 expression levels exhibited HCC-stage-specific characteristics ( Figure 2D).

Potential diagnostic values of hsa_circ_0004018
We then used the ROC curve to investigate the diagnostic value of hsa_circ_0004018 in distinguishing HCC tissues from para-tumorous and chronic hepatitis liver tissues. We found that the area under the ROC curve (AUC) was 0.848 (95% CI=0.803-0.894, P<0.001, Figure 3). The sensitivity and specificity were 0.716 and 0.815, respectively. The Youden index was 0.531.

Prediction of circRNA-miRNA-mRNA associations
In order to explore the molecular mechanism of hsa_circ_0004018, the hsa_circ_0004018-miRNA-mRNA axis in cancer-related pathways was predicted. DIANA-miRPath determined all the candidate miRNAs that are www.impactjournals.com/oncotarget involved in possible pathways by p-value cutoff at 0.05. As Figure 4 shown, hsa_circ_0004018 could harbor five miRNAs seed sequence matching. Then, DIANA-miRPath analysis revealed that both hsa-miR-30e-5p and hsa-miR-626 were associated with cancer-related pathways ( Figure 5A). After that, through TarBase7.0, MYC are hsa-miR-30e-5p and hsa-miR-626 validated common target genes. A network of hsa_circ_0004018-hsa-miR-30e-5p/ hsa-miR-626-mRNA interaction was delineated using Cytoscape ( Figure 5B); and the Venn diagram revealed the gene intersection ( Figure 5C). These findings suggest that hsa_circ_0004018-miR-30e-5p/miR-626-MYC axes might play a part in HCC carcinogenesis and metastasis.

DISCUSSION
CircRNAs have been recently found to be pervasively transcribed in the genome, and to be associated with human diseases. Studies have revealed that circRNAs express in a complex tissue-specific, cell- type-specific or developmental-stage specific manner [10,17]. In present study, we investigated circRNAs expression profile of HCC using the microarray analysis. We found that the circRNAs expression levels in HCC samples were different from paracancerous tissues. The microarray expression profiles exhibited that 174 upregulated circRNAs and 353 down-regulated circRNAs were significantly differentially expressed in five paired HCC samples; and showed hsa_circ_0004018's expressive feature in HCC for the first time. Compared with matched non-tumorous tissues, the expression level of hsa_circ_0004018 in HCC patients significantly decreased ( Figure 2C). In addition, the levels of has_ circ_0003570 in HCC cell lines were lower than those in L02. To get a closer insight into the clinical process of HCC, we compared the differences among liver tissues from patients with chronic hepatitis, cirrhosis, and HCC. Chronic hepatitis is well known as a dynamic disease which experience variable periods of immune activity versus quiescence that lead to the development of cirrhosis  and liver cancer in a part of patients [18]. Liver fibrosis is a reversible wound-healing response to any etiology of chronic hepatic injuries. The risk of liver-related mortality increases exponentially with the increase in fibrosis stage [19,20]. Our results showed hsa_circ_0004018 possessed a HCC-stage-specific expression feature ( Figure 2D). Furthermore, lower expression of hsa_circ_0004018 was associated with serum AFP level, tumor diameters, differentiation, BCLC stage and TNM stage (Table 1). Hsa_circ_0004018 is transcribed from SMYD4, which is a potential tumor suppressor and plays critical roles in carcinogenesis, the development of heart and skeletal muscle, and hematopoiesis [21][22][23].
Recently, several laboratories have revealed thousands of circRNAs [10,17,24,25]. In comparison to their cognate linear isoforms, some circRNAs' expression levels are even above 10-fold. Most importantly, because of the fact that circRNAs are free of ends, they are resistant to digestion with conventional RNA degradation pathways [25,26]. They are appealed as biomarkers for disease diagnosis and monitoring. For example, Li et al. found that hsa_circ_002059 could be used as a biomarker in the  [28]. Qin et al revealed that hsa_circ_0001649 is potential novel biomarker for HCC [29]. AFP, most used serum test for HCC surveillance and diagnosis, has been challenged in recent years [30]. A lot of clinical research showed diagnostic sensitivity is around 60% even with 10-20μg/L as AFP cutoff value [31]. Our results showed hsa_circ_0004018 is a valuable biomarker for HCC monitoring, the sensitivity and specificity of which were 0.716 and 0.815, respectively. And most importantly, the sensitivity of hsa_circ_0004018 is superior to that of AFP.
In conclusion, as one of circRNAs, hsa_ circ_0004018 was lowly expressed in HCC. At the same time, hsa_circ_0004018 showed HCC-stage-expressive characteristics from chronic hepatitis to cirrhosis and to HCC. These indicate that hsa_circ_0004018 not only might be a potential biomarker for the diagnosis of HCC, but also play a role in the carcinogenesis and metastasis of HCC. And further detailed mechanism studies underlying hsa_circ_0004018 are being carried out in our laboratory.

Patients and specimens
The total of 102 HCC patients, who underwent surgeries at three medical centers (Ningbo No. 2 Hospital, Ningbo Lihuili Hospital and Ningbo Yinzhou Peoples' Hospital, China) from March 2013 to December 2016, were included in this study. The para-tumorous tissues were obtained from 1 cm away from the edge of the HCC; and there were no obvious tumor cells. The diagnosis of HCC was confirmed by histological examination. Staging was determined by the BCLC staging system [30] and American Joint Committee on Cancer criteria [46]. Patients with HCC who had prior treatment of their tumor or history of other solid tumors were excluded in this study.
Other liver tissues were collected from 55 cases of chronic hepatitis patients from September 2013 to December 2016 in Ningbo No. 2 Hospital through liver biopsy under guided ultrasound. Fibrosis stage was assessed by the METAVIR scoring system (grades the stage of histological activity on a four-point scale: no activity=A0, mild activity=A1, moderate activity=A2, and severe activity=A3; grades the stage of fibrosis on a fivepoint scale: no fibrosis=F0, portal fibrosis without septa= F1, portal fibrosis with rare septa=F2, numerous septa without cirrhosis=F3, cirrhosis=F4) [47]. In all adjacent non-tumorous tissues, 74 cases of para-tumorous liver tissues were pathological diagnosis; among them, 60 cases showed LC (F=4) and 14 cases showed CH (F=0-3). Among 55 cases of chronic liver tissues by ultrasound guided, 3 cases showed LC (F=4) and 52 cases showed CH (F=0-3). Cirrhosis tissues group consisted of 60 para-tumorous liver tissues and 3 liver tissues of chronic hepatitis patients if METAVIR scoring was F4. Chronic hepatitis group consisted of 14 para-tumorous liver tissues and 52 liver tissues of chronic hepatitis patients METAVIR scoring was F0-3 with any A grade.
After being obtained, tissue samples were immediately soaked in RNA fixer Reagent (Bioteke, Beijing, China) and stored at −80 °C until used. Histology was independently assessed by two experienced pathologists who were blinded to the clinical data. This study was approved by the Human Research Ethics Committee from Ningbo University. Informed consent was obtained from all patients.

RNA extraction
Total RNA was extracted by TRIzol reagent (Invitrogen, Karlsruhe, Germany) in accordance with the manufacturer's instructions. Concentration and purity of total RNA samples were measured by the Smart Spec Plus spectrophotometer (Bio-Rad, Hercules, CA). If the ratio of A260/A280 was 1.8-2.0, RNA was used for further experiments.

Microarray data analysis
The microarray detection was performed by KangChen Bio-tech (Shanghai, China) under the guidance of the experiment workflow [15]. The circRNA/microRNA (miRNA) interaction was predicted using Arraystar's home-made miRNA target prediction software based on TargetScan and miRanda [15].

Reverse transcription
The cDNA was generated using the GoScript Reverse Transcription (RT) System (Promega, Madison, WI) following the manufacturer's instructions. Briefly, 2μg total RNA, 1μl random primer, 1μl oligo(dT)15 primer, 2μl MgCl 2 , 4μl GoScript 5×reaction buffer, 1μl nucleotide mix, 0.5μl recombinant RNasin ribonuclease, and 1μl GoScript reverse transcriptase were added in the system and then incubated at 42 °C for 1h. RT reaction and notemplate control were run at the same time.

Quantitative real-time PCR
Quantitative polymerase chain reaction (qPCR) was performed using the GoTaq qPCR Master www.impactjournals.com/oncotarget Mix (Promega) on an Mx3005P real-time PCR System (Stratagene, La Jolla, CA) in the light of the protocol. Outward facing Primers were designed with Primer3 (http://www-genome.wi.mit.edu) and synthesized by Sangon Biotech (Shanghai, China). Their sequences were as follows: for hsa_circ_0004018 (target gene) 5'-TCAACCTTTTGCCCCACACT-3' and 5'-AAGACACGTCTGTGTGTTGT-3'; and for glyceraldehyde 3-phosphate dehydrogenase (GAPDH, reference gene), 5'-TCGACAGTCAGCCGCATCTTC TTT-3' and 5'-ACCAAATCCGTTGACTCCGACCTT-3'. Real-time PCR was done in triplicate. The amplification specific was confirmed by melting curve analysis. The data from qRT-PCR were analyzed by the ΔCt method and the 2 −ΔΔCt method. All results are expressed as the means±SD. All of assays were performed in a blinded fashion.

Electrophoresis and sequencing of qRT-PCR products
To prove the qRT-PCR products, 1.5% agarose gel electrophoresis was used. Afterwards, based on the manufacturer's instructions, the qRT-PCR product of hsa_ circ_0004018 was first purified by using a UNIQ-10 PCR Product Purification Kit, and cloned into the pUCm-T vector (Sangon Biotech). After that, DNA sequencing was performed by Sangon Biotech Co., Ltd.

Prediction for circRNA-miRNA-mRNA pathways
The miRNA pathway was carried out based on DIANA-miRPath [48]. All of miRNA gene targets are experimentally validated (derived from TarBase 7.0) [49]. The graph of the circRNA/miRNA network was drawn with the help of Cytoscape 2.8.2. P<0.05 was used as the criterion for statistical significance.

Statistical analysis
All statistical analysis in this study were performed by the Statistical Product and Service Solutions (SPSS) 16.0 software package (IBM, Chicago, IL) and GraphPad Prism 6.0 (GraphPad Software, La Jolla, CA). Paired t test, independent t test and one way analysis of variance (ANOVA) were used in this study correctly. A receiver operating characteristic (ROC) curve was established to value the diagnostic power. P value of 0.05 or less was considered statistically significant.