CircRNA GFRA1 promotes hepatocellular carcinoma progression by modulating the miR-498/NAP1L3 axis

Circular RNAs (circRNAs) play essential roles in tumorigenesis and tumor progression. CircRNA GFRA1 (circGFRA1) was dysregulated in many cancer samples and acted as an independent marker for prediction of survivals in various cancer patients. However, the functions and molecular mechanisms of circGFRA1 in hepatocellular carcinoma (HCC) remain unclear. We collected 62 HCC tissues and normal adjacent tissues to evaluate the expression of circGFRA1 and the relationship between circGFRA1 expression and HCC patients’ survival. We carried out a list of characterization experiments to investigate the roles and underling mechanisms of circGFRA1 and miR-498 in HCC progressions. CircGFRA1 was greatly increased in HCC tissues and cells, and the over-expression of circGFRA1 was intimately related with the advanced clinical stage and poor survival of HCC patients. The expression of circGFRA1 was negatively correlated with the expression of miR-498, but a positive correlation was found between circGFRA1 and NAP1L3 expression in HCC tissues. Silencing circGFRA1 inhibited the growth and invasion of hepatocellular carcinoma. Moreover, miR-498 over-expression or NAP1L3 inhibition could abrogate the oncogene role of circGFRA1 in HCC in vivo. Our findings indicated that circGFRA1 contributed to HCC progression by modulating the miR-498/NAP1L3 axis in HCC.

and osteosarcoma 14 . However, for HCC, the dysregulation of GFRA1 remains to be elucidated. It was widely accepted that GFRA1 could act as competitive endogenous RNAs to co-regulate each other by sponging microRNAs 13,15 . Many biologists have illustrated that circRNAs could work as miRNA sponges to contribute to the regulation of cancers 13 . MiRNAs are endogenous single-stranded with ~ 23 nucleotide RNAs that have an essential role in the development of human cancers 16 . MiR-498 was previously demonstrated to have tumor-suppressive effects in regulating cancer cell progressions in ovarian cancer 17 , colorectal cancer 18 and lung cancer 19 . Considering the significant role of miR-498 in so many cancer types, we are encouraged to discover the role of miR-498 in HCC, as well as its potential interactions with circGFRA1.
The nucleosome assembly proteins (NAP) originated mammalian cells and was identified as a family of evolutionarily conserved histone chaperones 20 . NAPL3 has been illustrated to play essential roles in maintaining cell viability, especially in the formation and maintenance of the nervous system 21 . Besides, Kress et al. reported that NAPL3 promoted pre-mRNA splicing in budding yeast 22 . Motivated by its role in cellular biology, we aim to investigate the roles and inner associations among GFRA1, miR-498 and NAP1L3, as well as their functional mechanisms in HCC.

Results
CircGFRA1 was markedly over-expressed in HCC. To investigate the expression of circGFRA1 in HCC, qRT-PCR was used. Figure 1A showed that circGFRA1 was significantly over-expressed in HCC tissues compared with normal adjacent tissues (n = 62). Figure 1B demonstrated that circGFRA1 expression was significantly elevated in HCC cell lines of SMMC-7721, HepG2, PLC, SK-Hep1, Huh7, HCCLM3 and Hep3B compared with normal liver cell line. Moreover, as shown in Table 1, high circGFRA1 expression was markedly related to tumor size, intrahepatic metastasis, extrahepatic metastasis, BCLC stage and TNM stage. High circG-FRA1 levels were also correlated with poor overall survival (Fig. 1C). The results revealed the oncogenic role of circGFRA1 in HCC. Figure 2A searched the web tool Starbase and found that circGFRA1 might obtain shared binding sequences with miR-498. Figure 2B showed the qRT-PCR results and revealed that si-circGFRA1 effectively reduced circGFRA1 expression, but it did not affect its linear isoform GFRA1 mRNA (Fig. 2C). Similarly, as shown in Fig. 2C, circGFRA1 expression plasmid greatly elevated the expression of circGFRA1 and did not affect the expression of GFRA1 mRNA. In addition, Fig. 2D showed that circGFRA1 knockdown significantly promoted the expression of miR-498, but not circGFRA1 over-expression in HCCLM3 and Hep3B cells. From Fig. 2E, miR-498 significantly decreased the luciferase activity of the circGFRA1-WT, but did not change the luciferase activity of the circGFRA1-MUT. Moreover, Fig. 2F,G illustrated that miR-498 was greatly decreased in HCC tissues and cell lines compared with control. Figure 2H found a negative correlation between the expression of circGFRA1 and miR-498 in HCC tissues (r = − 0.533, P < 0.01). Figure 2I revealed that miR-498 effectively increased circGFRA1 expression, but it did not affect its linear isoform GFRA1 mRNA (Fig. 2J). Overall, it suggested that circGFRA1 might exert its functions by sponging miR-498.

CircGFRA1 sponged miR-498.
CircGFRA1 sequestered miR-498 and up-regulated NAP1L3 levels. Figure 3A showed the binding sites between miR-498 and NAP1L3. In Fig. 3B,C, we noticed that the mRNA and protein expression of NAP1L3 were significantly decreased in miR-498 in HCCLM3 and Hep3B. However, this effect was reversed by circG-FRA1. Figure 3D found that miR-498 over-expression inhibited the luciferase activities of the NAP1L3-WT, instead of NAP1L3-MUT. Figure 3E,F demonstrate that NAP1L3 was markedly increased in HCC tissues and cell lines compared with control. According to Fig. 3G, we observed that circGFRA1 expression was positively related with NAP1L3 expression in HCC tissues (r = 0.556, P < 0.01). The above results revealed that circGFRA1 elevated NAP1L3 expression through sponging miR-498 in HCC (" Supplementary Information").  Figure 4D also observed that miR-498 over-expression significantly reversed the circGFRA1-induced invasion rates. It was noted that circGFRA1 cotransfected with si-NAP1L3 had the same effect of co-transfection with miR-498 mimics on HCC malignant phenotypes.
CircGFRA1 over-expression promoted NAP1L3-related signaling pathway in HCC cells.

Discussions
Although liver section and transplantation are available, hepatocellular carcinoma still ranks as the third leading cause for cancer-related death worldwide, especially in Asian 23,24 . Attributed to the immense opportunities of HCC cell metastasis, and tumor recurrence, HCC patients are experiencing a significant life risk even after tremendous treatments like surgeries, chemotherapies or targeted drugs 25,26 . Increasing studies have shown that various oncogenes are related to HCC metastasis, such as lncRNA PVT1 27 , lncRNA UCA1 28 and circMTO1 29 . Therefore, it is quite important to reveal more therapeutic targets to improve the functional mechanisms and practical approaches for HCC patients.
There is growing evidence supporting that circRNAs are involved in the growth of colorectal 10 and ovarian cancers 30 , as well as HCC 31 . CircGFRA1 has been implicated in the regulation of neuronal cell survival and differentiation 32 . Studies also have indicated that GFRA1 has a role in the progression and metastasis of human cancers such as breast cancer 13  www.nature.com/scientificreports/ act as ceRNAs in triple negative breast cancer by regulating miR-34a 13 . In 2015, Liu proposed that the downregulated expression of GFRa1 promoted HCC progression though Epithelial-to-Mesenchymal Transition 33 . Although this is a preliminary report, we still found its dysregulated expression in HCC samples. In our experiments, circGFRA1 was greatly elevated in HCC tissues and cell lines. High circGFRA1 expression was markedly related to tumor size, intrahepatic metastasis, extrahepatic metastasis, BCLC stage and TNM stage as well as poor overall survival. CircGFRA1 was dramatically over-expressed in HCC. It was in consistence with previous studies that circGFRA1 can be regarded as an oncogene in HCC. It is well-known that lncRNAs could play as competitive endogenous RNAs to regulate other genes' expression by sponging microRNAs 34 . Many findings have proved that circRNAs could function as miRNA sponges 35 . We found that circGFRA1 might directly sponge miR-498. The knockdown of circGFRA1 significantly promoted the expression of miR-498, and miR-498 significantly decreased the luciferase activity of the circGFRA1-WT. It was possible that there existed a negative correlation between the expression of circGFRA1 and miR-498 in HCC tissues. Our data further confirmed that circGFRA1 might exert its functions by sponging miR-498.
Some evidence indicates that NAP1L3 play essential roles in maintaining cell viability 36 . We noticed that NAP1L3 was markedly increased in HCC tissues and cell lines. The mRNA and protein expression of NAP1L3 were significantly decreased in miR-498 in HCC cell lines, and circGFRA1 expression was positively correlated with NAP1L3 expression in HCC tissues. CircGFRA1 elevated oncogene NAP1L3 expression by sponging miR-498 in HCC. In addition, we also found that miR-498 over-expression or NAP1L3 silencing effectively reversed circGFRA1-induced HCC progression. For the first time, we established the fact that circGFRA1 elevated NAP1L3 expression by acting as a sponge of miR-498 in HCC.
It's reported that the knockdown of circGFRA1 inhibited proliferation and promoted apoptosis of triplenegative breast cancer cells 13 . From the western blotting results, the expression of proliferation indicators of c-Myc and cyclin D1 and invasion indicators of MMP-2 and MMP-9 were greatly elevated in HCC LM3 and Hep3B cells transfected with circGFRA1 expression plasmid. However, miR-489 mimics or si-NAP1L3 attenuated this   37 . NAP1L1 is an essential participant to the aggressive clinic pathologic features of HCC. We constructed the xenograft model based on nude mice and conducted in vivo experiments and compassion. We found that circGFRA1 over-expression markedly elevated the tumor growth rate and tumor volume, accompanied by the up-regulation of NAP1L3. However, the promoted effects of circGFRA1 were relieved by miR-498 agomir and si-NAP1L3 in HCC cells. As far as we know, we are the first to propose that circGFRA1 could effectively promote HCC progression by regulating the miR-498/NAP1L3 axis.

Conclusion
These data suggest that circGFRA1 contributed to HCC progression by modulating the miR-498/NAP1L3 axis. Our findings may provide a potential therapeutic target for HCC.

Methods
Patients and tissue specimens. 62 HCC tissues and healthy tissues were taken from Henan Provincial People's Hospital. The experimental protocols were approved by the ethics committee of Henan Provincial People's Hospital (No. HPPH201301HCC6#3), and has been performed in accordance with the Declaration of Helsinki. All patients signed the written informed consent. Patient inclusive criteria were in the following. They were pathologically diagnosed as HCC by two senior pathologists without adjunctive treatment before curative hepatectomy from 2013 to 2014. Patients were excluded if they had cholangiocarcinoma or other malignancy and incomplete clinical or prognostic data. All samples were collected in 15 min after removal from the body and immediately frozen in liquid nitrogen and stored at − 80 °C.

QRT-PCR.
Total RNA was isolated from 10 6 cells via TRIzol kit (Invitrogen, USA). All RNA samples were digested with DNase I at 37 °C for 1 h to remove genomic DNA. QRT-PCRs were carried out by SYBR Premix Ex Taq II kit (Takara, Japan). GAPDH and U6 were regarded as controls. Three replicates were performed for each reaction. Gene expression levels were calculated using 2 −ΔΔCT method. The primer sequences were listed in the below.    TGC TTC GCG AC-3′  NAP1L3: reverse: 5′-TAC TTG CGC CGA AGT TGG C-3′  GAPDH: forward: 5′-ATG CGA CCC ACG GGA GAA T-3′  GAPDH: reverse: 5′-AAA AAG GCT GCT TGT TGG AC-3′  U6: forward: 5′-GCA GGG CTG TGA TCT GTC GAC-3′  U6: reverse: 5′-CCC CGA CAC CCC GGA TTA TTC-3′ Animal studies. week-old female BALB/c nude mice (8 per group to provide a power of 90% for a significance level of 0.05 with a two-tailed t-test.) were classified to 2 teams. All the mice were subcutaneously administered with 10 6 Hep3B cells for a xenograft model. At the 9th day, circGFRA1 over-expression plasmid, miR-498 agomir, si-NAP1L3 and negative control were intratumorally administered. After thirty days, we euthanized the mice. Tumor weights and volumes were measured and calculated by length × width 2  Transwell assay. 24-well transwell chamber, without or with Matrigel (Corning, USA), was employed to detect hepatocellular carcinoma cell invasions. 5 × 10 4 cells in non-serum culture medium were transferred to the upper chamber, and the lower chamber was filled with a culture medium with 20% FBS. The upper chamber was coated with Matrigel (BD Biosciences). Crystal violet (0.1%) was utilized to stain cells, and cells were observed under IX71 inverted microscope (Olympus, Tokyo, Japan). After one day, the migrated or invaded cells were fixed, stained and analyzed by using microscopy.

Immunohistochemistry (IHC).
We fixed the tissues in 4% formalin, embedded them in paraffin and sectioned them (4 µm thickness). They were baked at 60 °C for 2 h and incubated with xylene for de-paraffinization and gradient ethanol. We carried out antigen retrieval and used 3% hydrogen peroxide to block endogenous peroxidase for twenty min. The sections were cured by goat serum to avoid non-specific staining. Then, they were treated with anti-NAP1L3 (1:1000, #ab158953, Abcam, UK) overnight at 4 °C. Primary Antibody Enhancer was used and incubated at room temperature for 20 min. Then, HRP Polymer (enzyme labeled second antibody) was added at room temperature for 30 min. At last, DAB was used to evaluate the results, and positive immune staining was measured by the proportion of positive cells.

Statistical analysis.
Data analyzing was carried out by GraphPad software 7.0 and SPSS19.0. All experiments were carried out in triplicate, and results represent the average of 3 independent experiments. Data were expressed as the mean value ± standard deviation. Student's t-test (2 groups) and one-way ANOVA (> 2 groups) were utilized to analyze the differences. We employed Kaplan-Meier plots to evaluate the survival rates. Correlations were made by Pearson correlation. P < 0.05 was regarded as statistically significant.