Microrna-2392 Functions as a Tumor Suppressor Gene and Inhibits Malignant Progression of Hepatocellular Carcinoma Via Directly Targeting JAG2

Background Dysregulation of microRNA (miRNA) expression in various cancers and their vital roles in malignant progression of cancers are well investigated. Our previous studies have analyzed miRNAs that promote malignant progression of hepatocellular carcinoma (HCC), this study aim to systematically elucidate metastasis suppressor miRNAs in HCC. miR-2392 was signicantly down-regulated in HCC. Overexpression of miR-2392 suppressed proliferation, mobility, spheroid formation and maintenance of cancer stem cells (CSC)-like characteristics in HCC cell lines, whereas down-regulation of miR-2392 led to the opposite results. CTCs from HCC patients with lower serum miR-2392 level had stronger cell spheroid formation ability. A negative correlation between the content of miR-2392 in serum and the number of CTC spheroids had been found. We identied Jagged2 (JAG2) as a direct target of miR-2392, miR-2392 inhibited the expression and nuclear accumulation of JAG2 by targeting 3’-UTR of JAG2. HCC cells were treated with LV-miR-2392 inhibitor and JAG2-siRNA to explore the mechanism of miR-2392 and JAG2 on HCC. Down-regulation of JAG2 inhibited the overexpression effects of miR-2392 in vitro and in vivo. JAG2 is highly expressed in HCC and is closely related to poor prognosis and survival of patients. Our ndings indicated a signicant role of the miR-2392/JAG2 axis in suppressing HCC cell growth and aggressiveness, miR-2392 may play a role as a tumor suppressor gene to guide the individualized precise treatment of HCC.


Introduction
Cancer incidence and mortality are rapidly growing worldwide. There were 18.1 million new cases and 9.6 million cancer deaths worldwide in 2018 [1]. The latest results showed that 1,898,160 new cancer cases and 608,570 cancer deaths are projected to occur in the United States in 2021 [2]. Primary liver cancer is con rmed by statistics to be the sixth most commonly diagnosed cancer and the fourth leading cause of cancer death, with about 841,000 new cases and 782,000 deaths annually [1]. Hepatocellular carcinoma (HCC) is the most common histological subtype of primary liver cancer (accounting for 75-85%) with high invasive, metastatic potential, and has a low survival rate [3]. Due to the insidious onset of HCC patients at early stages and lack of speci c biomarkers for early diagnosis, majority of patients had reached advanced stages at their rst diagnosis and missed the opportunity of radical resection and liver Page 3/29 transplantation. The therapeutic outcomes and quality of life in HCC patients are far from satisfaction.
For this reason, it is urgently needed to develop novel effective biomarkers for early diagnosis of HCC.
Circulating tumor cells (CTCs) are tumor cells that shed off the primary tumor, migrate to the blood vessels or lymphatic vessels around the tissues, and exist in the circulatory system [4]. At present, accumulating clinical and basic studies have shown that CTCs might directly participate in metastasis and recurrence of multiple types of malignant tumor, including prostate cancer [5], colorectal cancer [6] and metastatic breast cancer [7]. The metastasis of tumor cells in vivo and in vitro requires the coordinated regulation of tumor microenvironment. The special genetic properties of different tumor cells endow CTCs with the characteristics of survival, proliferation and invasion, and the special tumor microenvironment provides the structural and material basis for metastasis and growth of CTCs.
However, the exact regulatory mechanism of survival, migration and colonization of CTCs has not been clari ed. microRNAs (miRNAs) are a class of highly conserved, stable and abundant non-coding RNAs that could directly regulate gene expression at transcriptional or post-transcriptional levels. Finally, they affect the speci c protein to perform speci c physiological functions in vivo [8]. In recent years, studies on the relationship between miRNAs and tumors have been booming. A variety of miRNAs are identi ed to act as tumor suppressor or oncogenes, participate in the abnormal changes of biological behavior of tumor cells, and play a crucial role in tumorigenesis and malignant progression [9,10]. It is widely recognized that miRNAs mainly derived from the secretion process of animal and plant cells [11]. miRNAs could bind with lipid or lipoprotein components in cells, and secrete into the peripheral blood circulation system in the form of microcapsules. It can also enter the receptor cells through endocytosis, thus playing its speci c biological function by regulating target genes. A large number of tumor associated miRNAs have been reported to be tissue-speci c and organ speci c, and participate in the formation of speci c tumor microenvironment. In our previous studies, we found that miRNAs are involved in the formation of tumor microenvironment, and affect the development of HCC by interacting with HCC cells [12,13]. miRNA could tolerate degradation of RNase, and highly stable in the peripheral circulation system for a long time. As long as we use the conventional total RNA extraction method, we could obtain the concentration of miRNA meeting the experimental requirements, and its expression pattern is tissue-speci c. Therefore, the circulating miRNA in peripheral blood has become a target candidate for clinical diagnosis and treatment of HCC [14][15][16].
We speculated according to our previous studies that, on one hand, HCC cells in the primary tumor site and its surrounding tissues could participate in the transformation of HCC microenvironment and malignant biological phenotype of HCC cells by releasing a large number of cancer-related miRNAs, which provides favorable conditions for HCC cells to escape immune surveillance, clonal growth, and spread into blood to form CTCs; on the other hand, circulating miRNAs released into the circulation could also regulate apoptosis-resistance, survival, proliferation and formation of intrahepatic and distant metastasis of CTCs. the delta-delta Ct method to conduct quanti cation as well as calculation of the relative expression of each mRNA or miRNA. The relative expression levels of miRNA and mRNA expression were normalized to U6. Primer sequences are listed in Table S2. Each sample was carried out three times.

Transcriptome analysis
Total RNA was extracted from three pairs of fresh human HCC tissues (tumor and para-tumorous tissue) by Trizol reagent (Invitrogen) separately. The RNA quality was checked by Bioanalyzer 2200 (Aligent) and kept at -80 °C. The RNA with RIN (RNA integrity number) > 8.0 is acceptable for cDNA library construction. RNA-seq and bioinformatic data analysis were performed by Shanghai Novelbio Ltd.

RNA-seq mapping
RNA-Seq reads were ltered and mapped to Human genome (GRCh38, NCBI) utilizing HISAT2. HTSeq was used to calculate the gene count. miRNA-Seq reads were ltered and mapped to Human miRNA database (miRBase v21.0) and Human genome (GRCh38, NCBI) to achieve the miRNA expression. Differentially expressed genes analysis was applied utilizing DESeq2. The RNA-seq data was deposited in GEO (Accession code: GSE146719). The target gene of miR-2392 was predicted using the TargetScan database (http://www.targetscan.org/vert_71/). The expression of JAG2 and Notch3 in HCC samples of the Cancer Genome Atlas (TCGA) database was searched using the UALCAN database (http://ualcan.path.uab.edu/analysis.html).

Luciferase activity assay
The wild-type (WT) sequence containing the predicted target sites of miR-2392 in the 3'UTR of JAG2 mRNA. We mutated the target sites from CCAUCCU to AACGUUG. The wild-type (WT) or mutant (MUT) sequences of JAG2 was inserted into the pGL3 Basic vector (Promega, Madison, WI). Huh7 cells were seeded in 24-well plates at a density of 5000 cells per well. After 24 h, the cells were transiently transfected with a mixture of 0.1 μg of luciferase reporter plasmid containing 3'-UTR segment of JAG2, 0.02 μg of Renilla luciferase reporter plasmid, and 0.4 μg miRNAs by using X-tremegene HP transfection reagent (Roche, Germany) as described by the manufacturer. Luciferase activity was detected 48 h after transfection by using the Dual Luciferase Reporter Assay System as described by the manufacturer (Promega, Madison, WI, USA). All samples were independently repeated three times.
Establishment of stable miR-2392 knockdown and overexpressing cells miR-2392 was stably knocked down (LV-hsa-miR-2392 inhibitor) and overexpressing (LV-hsa-miR-2392 mimics) in HCC cell lines by means of lentiviral approach which was designed and produced by GeneCopoeia (Rockville, MD). LV-hsa-miR-2392 NC was used as non-target control. The information of sequences is listed in Table S3.

Western blotting
Protein from fresh HCC samples were obtained with RIPA lysis buffer added with protease inhibitors. After quanti cation with bicinchoninic acid (BCA) assay (Beyotime, Jiangsu, China), we separated each protein through 10% SDS-PAGE and then moved them onto PVDF membranes (Millipore, USA). Then, samples were blocked with 5% nonfat milk. After incubation with primary antibodies against GAPDH and JAG2 (1:200, ab109627, Abcam, UK) and secondary antibodies, protein levels were detected with Image J (GE Healthcare Life Sciences). Each sample was analyzed three times.
The total cellular protein and the nuclear protein were extracted according to instructions of nuclear and cytoplasmic extraction reagents kit (Beyotime, Jiangsu, China). The protein concentrations were

Cell proliferation assay
Cell proliferation assay was conducted with Counting Kit-8 (CCK-8) (Beyotime, Jiangsu, China). 5000 cells (transfected with miR-NC, miR-2392 mimic, miR-2392 inhibitor) were planted into 96-well plates for 24h, 48h and 72h respectively. Then 10% CCK-8 solution was added. The absorbance of each sample was assessed by a microplate reader set at 450 nM. Each sample was performed for three times.

Cell migration and invasion assay
The methods of cell migration and invasion assays were constructed as previously described. Con uent cell monolayers which transfected with miR-2392 mimic, inhibitor and miR-NC in 6-well plates were wounded by manual scraping with a sterile pipette tip. The imagines were analyzed by Image J (NIH). Cell invasion assay was performed on Matrigel-coated transwell chambers (Corning, NY, USA) with the treatment of miR-2392 mimic, inhibitor and miR-NC.
Colony formation assay Cells (2000 cells/well) which transfected with miR-NC, miR-2392 mimic and miR-2392 inhibitor were seeded in each well of a six-well cell culture plate. After 2 weeks, the plates were xed in 4% paraformaldehyde and stained with 1% crystal violet. The colony numbers were counted to assess cell proliferation. The assays were performed in three independent experiments.

Immunohistochemistry
Tissue microarrays consisting of 180 cases of HCC tissues and matched para-cancerous tissues were constructed. Tissue sections were depara nized, rehydrated, and treated with 3% H 2 O 2 for 15 min to inhibit endogenous peroxidase activity. Following heat-induced epitope retrieval in 10 mM citrate buffer (pH 6.0) in a microwave for 15 min, the slides were incubated at 4 °C overnight with a pre-diluted primary antibody (JAG2, 1:200, ab109627; Notch3, 1:200, ab23426, Abcam, UK). After incubation with a secondary antibody, the signal was developed with 3,3'-diaminobenzidine tetrachloride.
The staining scores were evaluated by two individuals in a blinded fashion. A quick scoring system from 0 to 12 that combined the intensity and percentage of the positive signal was used. Brie y, a signal of 0 indicated no staining, 1 indicated weak staining, 2 indicated intermediate staining and 3 indicated strong staining. Percentage scores were assigned as follows: 0 corresponded to 0%, 1 to 1-25%, 2 to 26-50%, 3 to 51-75%, and 4 to > 75%. The median value of total staining scores was identi ed as the optimal cut-off value. If the evaluated score was lower than the median, the indicator expression of in those HCC samples was classi ed as low; otherwise, it was classi ed as high [17,18].
CTC enrichment CTC enrichment of whole blood samples was conducted according to the method previously described. Brie y, following density gradient centrifugation (Ficoll-Paque PLUS,GE Healthcare, WI, USA), CTCs were enriched by extracting CD45-expressing leukocytes with magnetically beads labeled anti-CD45 monoclonal antibody (130-045-801, Miltenyi Biotec, Germany) according to the instructions. The remaining cells were cyto-centrifuged on polylysine-coated slides for identi cation and enumeration of CTCs, or resuspended in 150 μL Dulbecco modi ed Eagle medium (DMEM) (high glucose) (Thermo Fisher Scienti c, MA, USA) for the in vitro three-dimensional (3D) culture of CTCs.

Three-dimensional (3D) culturein vitro
The enriched CTCs were cultured using a 3D cell culture assay. Brie y, Matrigel (BD Biosciences, San Jose, CA, USA) was thawed according to the manufacturer's instructions and mixed equally with the CTCcontaining DMEM. When HCC cell lines are used for 3D culture, 1×10 3 cells suspended in DMEM were mixed with Matrigel. The prepared mixture was then incubated in a 24-well plate for 30 min at 37°C. Subsequently, 500 μL DMEM supplemented with 10% fetal bovine serum (FBS) were added to the plate, and the media was supplied every two days during culture. Spheroid formation was observed under a microscope every day and counted on day 7. A spheroid was de ned as 3D cell structure > 100 μm in diameter that was determined using images taken on a microscope.
In vivo tumor growth and metastasis model BALB/c male nude mice 4-6 weeks old were purchased from Shanghai Laboratory Animal Center, Chinese Academy of Sciences (Shanghai, China). Mice were raised under pathogen-free conditions. All in vivo experiments were done according to approved protocols from the Institutional Animal Care and Use Committees, according to national and institutional guidelines. Brie y, to establish in vivo lung metastasis model, 2 × 10 6 viable cells (LV-hsa-miR-2392 inhibitor or LV-hsa-miR-2392 NC) were injected into the tail veins of mice. One week after inoculation, si-h-JAG2-3 (2'OMe + 5'Chol modi ed) was injected into the tail vein three times a week (10nmol/time/mouse) for treatment. After monitoring for 6 weeks, mice were sacri ced, and their lungs were dissected, xed with phosphate-buffered neutral formalin, and prepared for standard histological examination.

Statistical analysis
Data was reported as mean ± SD. Data analysis was conducted by IBM SPSS Statistics Version 19.0 and Graphpad Prism 5. Analysis of Variance (ANOVA) was used for comparison between groups. The gray value was analyzed by Image-Pro Plus 6.0 software. Survival analysis was estimated by the Kaplan-Meier survival method. Variables with a P value of < 0.05 in the univariate analyses were further tested using multivariate Cox proportional hazards model. Results were determined to be statistically signi cant when < 0.05.

Results
miRNA and mRNA expression pro ling in HCC tissues miRNAs could exist for a long time in the peripheral circulatory system which endows them with the potential to be ideal biomarkers for human diseases. In the present study, to explore novel biomarkers of HCC, we rst explored the expression pro les of mRNAs and miRNAs in three pairs of HCC patients (tumor tissues and non-tumorous tissues) by high-throughput RNA sequencing. A total of 1,024 mRNAs were found to be signi cantly dysregulated (636 up-regulated and 388 down-regulated) (Supplementary Figure 1A and 1B), 50 miRNAs were found to be signi cantly dysregulated (27 up-regulated and 23 downregulated) in the HCC tissues compared with the non-tumorous tissues ( Figure 1A and 1B). Signi cant differences in mRNA and miRNAs between HCC tissues and non-tumorous tissues were classi ed as fold change ≥ 2 or fold change ≤ 0.5, FDR < 0.05, as determined by having to satisfy both of the above conditions. Two of the up-regulated miRNAs (miR-206 and miR-6812-5p) and two of the down-regulated miRNAs (miR-2392 and miR-96-5p) were selected for validation by qRT-PCR.
Validation of differentially expressed miRNAs by qRT-PCR To validate the result, two of the down-regulated miRNAs (miR-2392 and miR-6812-5p) and two of the upregulated miRNAs (miR-206 and miR-96-5p) were selected for validation by qRT-PCR. We analyzed miRNAs expression in paired HCC tissues and corresponding non-tumorous tissues, and found that the expression level of miR-2392 was markedly lower in HCC tissues, which was in consistence with the detection of miR-2392 in serum of HCC patients ( Figure 1C and 1D). The expression of the other three miRNAs in paired HCC and para-cancerous tissues was consistent with the sequencing data, but the serum content was different (Supplementary Figure 1C and 1D). The results showed that the expression pro le of miR-2392 was consistent with those from the sequencing results. Collectively, these data suggest the accuracy and reproducibility of the RNA sequencing ndings.
Bioinformatics analysis of the potential role of miR-2392 in HCC Gene ontology analysis was performed to facilitate elucidating the biological implications of differentially expressed miRNA. Pathway analysis was used to nd out the signi cant pathway of the differential genes according to KEGG database. Go analysis and pathway analysis results revealed that the changes of Notch signaling pathway is one of the most signi cant in all signaling pathways ( Figure  1E). In order to further explore the potential functions of miR-2392 in HCC patients, we performed a miRNA-mRNA-pathway network analysis from our RNA-seq data were selected ( Figure 1F). We found that there were several potential target mRNAs of miR-2392 including JAG2 (a key regulator of Notch signaling pathway), NGFR, IGF2, IL7R, etc. Protein-protein interaction network (PPI network) results predicted that JAG2 and Notch3 present the highest score ( Figure 1G). These results suggested that JAG2/Notch3 signaling pathway may be involved in the biological regulation of miR-2392 in HCC.
miR-2392 was down-regulated in HCC cell lines which promoted the proliferation, migration, invasion, clonal formation and spheroid formation of HCC cells in vitro By qRT-PCR, the expression of miR-2392 was down-regulated in HCC cells compared to normal liver cells ( Figure 1H). We chose Huh7 (relatively low endogenous expression) and LM3 (relatively high endogenous expression) as cell model for subsequent experiments according to the expression pro les of miR-2392 in HCC cells. To investigate the potential role of miR-2392 in HCC cells, we used miR-2392 mimic and inhibitor to transfect two HCC cell lines (Huh7 and LM3). The optimal concentration of miR-2392 mimic and inhibitor has been selected and veri ed (Supplementary Figure 2).The results suggested that downregulation of miR-2392 promoted the proliferation, invasion and neoplasia ability of LM3. On the contrary, up-regulation of miR-2392 inhibited the proliferation, invasion and neoplasia ability of Huh7 (Figure 2A-2D). Previous study indicated that miRNA could maintain CSC-like phenotype in tumor cells. A tumor sphere formation assay was used to explore the effect of miR-2392 on CSC self-renewal properties in HCC. Our results revealed that miR-2392 inhibitor-transfected HCC cells (LM3) formed more and larger spheroids than miR-2392 NC group and blank group. On the contrary, the miR-2392 mimic-transfected HCC cells (Huh7) formed less and smaller spheroids than miR-2392 NC group and blank group ( Figure  3A). The protein expression level of stem cell related genes, EpCAM and CD133 were further veri ed by cellular immunostaining. The present data indicated that EpCAM could be observed in almost all spheroids, but down-regulation of miR-2392 promotes the expression of CD133 in LM3 cells. Transfection of miR-2392 mimic decreased the expression level of CD133 in Huh cells ( Figure 3B). Overall, these data suggested that down-regulation of miR-2392 might be closely related to the malignant phenotype (proliferation and motility) and maintenance of CSC-like characteristics of HCC cells in vitro.
Low level of miR-2392 promotes CSC-like properties of HCC CTCs in vitro According to the concentration of miR-2392 in sera of HCC patients, all patients could be divided into two groups: miR-2392 high group (mean concentration ≥ 100,000) and miR-2392 low group (mean concentration < 100,000). Three dimensional culture of CTCs isolated from HCC patients in vitro for 14 days. We analyzed the correlation between the number of CTC spheroids and the content of miR-2392 in sera of HCC patients. The results found that the CTCs from HCC patients with lower serum miR-2392 level had stronger cell spheroid formation ability. In addition, a negative correlation between the content of miR-2392 in serum and the number of CTC spheroids was observed (P = 0.0002, R 2 = -0.2539) ( Figure  3C). The results of cellular immuno uorescence also con rmed that low level of miR-2392 promotes the CSC-like properties of HCC CTCs in vitro ( Figure 3D).

JAG2 is a direct target of miR-2392
Previous RNA-seq data and bioinformatics analysis suggested that JAG2 may be the downstream target of miR-2392. We predicted miR-2392 potential regulatory target genes by Targetscan (http://www.targetscan.org/vert_71/). The result revealed that miR-2392 has a direct binding site with JAG2 ( Figure 4A). To verify JAG2 was a downstream target of miR-2392, we carried out a luciferase reporter assay to corroborate the direct interaction between JAG2 and miR-2392. WT or MUT 3'-UTR target sequences were cloned into a luciferase reporter vector ( Figure 4B). Our results showed that miR-2392 inhibited the luciferase activity of the WT 3'-UTR of JAG2 ( Figure 4C). These results con rmed that JAG2 was a direct target of miR-2392.
miR-2392 regulates the transcription of downstream genes by inhibiting the nuclear translocation of JAG2 protein Cellular immuno uorescence assay was performed to measure the expression and localization of JAG2 protein in HCC cells treated with miR-2392. The results indicated that miR-2392 mimic induced a decrease in the expression of JAG2, which was accompanied by a less nuclear accumulation of JAG2. By contrast, pretreatment with miR-2392 inhibitor led to an increase in the expression of JAG2, which was accompanied by a greater amount of nuclear accumulation of JAG2 ( Figure 4D). We extracted proteins from the cytoplasm and nucleus of HCC cells which pretreated with miR-2392 mimic and inhibitor, respectively. The results of western blotting con rmed that miR-2392 mimic decreased the expression and nuclear translocation of JAG2. Conversely, miR-2392 inhibitor signi cantly increased the expression and nuclear translocation of JAG2 in HCC cells ( Figure 4E).
Notch signaling pathway target genes vary in different tissue and different cells, such as HES and HEY families, MMP2, MMP9, cyclin D1, Her2, c-Myc, p21 and cell apoptosis-associated genes. Notch precisely regulates cell differentiation, proliferation and tumorigenesis might by modulating these genes expression. Our results showed that down-regualtion of miR-2392 increased MMP2, MMP9, PCNA and Vimentin expression. To the opposite, up-regualtion of miR-2392 decreased MMP2, MMP9, PCNA and Vimentin expression. Therefore, we speculate that low level of miR-2392 enhanced the proliferation ability of HCC cells and accelerated the degradation of extracellular matrix (ECM), which will facilitate the invasion and metastasis of HCC cells. Collectively, these data provide strong biochemical evidence that low level of miR-2392 signi cantly increased the expression and nuclear translocation of JAG2 which activates Notch signaling pathway to facilitate HCC growth and metastasis.
Relationship between the mRNA levels of JAG2/Notch3 and the clinicopathological parameters of patients with HCC Using the UALCAN (Data from TCGA database) (http://ualcan.path.uab.edu), we compared the mRNA expression of JAG2 and Notch3 between HCC tissues and normal liver tissues (n = 371). The results indicated that the expression levels of JAG2 and Notch3 were higher in HCC tissues than in normal tissues ( Figure 5A; Supplementary Figure 3A).
Next, we performed immunohistochemistry (IHC) to test JAG2 and Notch3 protein expression in HCC tissues and their counterparts. We found that JAG2 and Notch3 proteins were more highly expressed in HCC tissues than in the normal tissues ( Figure 5B and Supplementary Figure 3B). Furthermore, western blot assay was carried out to detect the expression of JAG2 and Notch3 in extracted protein from fresh HCC tissues (n = 32) ( Figure 5C, 5D; Supplementary Figure 3C). The results suggested that the expression levels of JAG2 and Notch3 was higher in HCC tissues than in non-tumorous tissues.
According to the expression pattern and evaluation of JAG2 in HCC tissues, the average score is 6 ( Figure  5B). 131 HCC samples which score was higher than the average score was classi ed as high expression group. Otherwise, 49 HCC samples which score was lower than the average score was classi ed as low expression group. As summarized in Table 1, the correlation between JAG2 protein expression and various clinicopathological characteristics was investigated, as well as follow-up data, were collected and recorded. The median follow-up period was 37.1 months (range, 2-60 months). Subsequent association analysis determined that high expression levels of JAG2 in HCC tissues signi cantly correlated with vascular invasion (P = 0.008) and TNM stage (P = 0.034). The results of the present study indicate that abnormal expression levels of JAG2 may be associated with the recurrence and metastasis of HCC. As shown in Figure 5E, HCC with high expression of JAG2 was associated with a signi cantly lower overall survival rate (P = 0.0359) and recurrence-free survival (P = 0.0262) than was HCC with low expression of JAG2.
Based on the univariate analysis, vascular invasion, TNM stage and JAG2 expression were selected to be further tested for multivariate analysis (Table 2 and 3). The results indicated that JAG2 was one of the independent favorable factors for overall survival and recurrence, as were advanced TNM stage, and vascular invasion.

miR-2392 suppresses tumor metastasis of HCC cells in vivo and in vitro
To further validate inhibiting roles of miR-2392 in HCC progression, we established an experimental in vivo metastasis assay by tail injection of stably knockeddown-miR-2392 LM3 cells (LV-hsa-miR-2392 inhibitor) which had been validated by cellular immunostaining (Supplementary Figure 4). Effective JAG2-siRNA had been screened in vitro before (Supplementary Figure 5). The experimental groups were as follows: (1)

Discussion
HCC metastasis and recurrence is the main cause of HCC-related death. However, its mechanism remains poorly understood. Further study on the key factors of tumor microenvironment (TME) may be helpful. In an attempt to identify TME-related factors (such as miRNAs) that are speci c for HCC progression, we recently conducted a high-throughput RNA sequencing by comparing primary tumors with their matched non-tumorous tissues.
In this study, we found a group of cancer-related miRNAs (miR-2392, miR-6812, miR-96-5p, miR-206) by high-throughput RNA sequencing of freshly collected HCC tissues and paired non-tumorous tissues. The results of subsequent validation in serum and tissue samples of HCC patients showed that the expression level of miR-2392 was signi cantly lower than that of non-tumorous tissues which was most consistent with the high-throughput RNA sequencing results. We found that miR-2392 could signi cantly inhibit the ability of proliferation, migration, invasion and spheroidization of HCC cells in vitro. miR-2392 might act as a tumor suppressor in HCC.
Circulating tumor cells (CTCs) disseminated in circulation system are liable to suffer apoptosis by blood shear stress and anoikis. Therefore, the surviving CTCs maybe the key factor of recurrence and metastasis. Likely, the tumor microenvironment could provide clues for elucidating the survival mechanism of CTCs. The ampli cation and culture of CTCs in vitro has been a technical bottleneck for a long time due to the extremely rare number of CTCs. We successfully expand CTCs by using the threedimensional cell culture system through continuous attempts. Correlation analysis of the number of CTC spheroids and the content of miR-2392 in serum of HCC patients. Our present results indicated that the CTCs from HCC patients with lower serum miR-2392 level had stronger cell spheroid formation ability.
Furthermore, a negative correlation between the content of miR-2392 in serum and the number of CTC spheroids was observed (P = 0.0002, R 2 = -0.2539). These results suggested that miR-2392 might be one of the crucial factors affecting the survival of CTCs in vitro.
Next, we focused on the exploration for miR-2392 targeted genes and related signaling pathways. Bioinformatics analysis of the potential role of miR-2392 in HCC based on the high-throughput RNA sequencing data. Go analysis and pathway analysis results revealed that the changes of Notch signaling pathway is one of the most signi cant in all signaling pathways. miRNA-mRNA-pathway network analysis and protein-protein interaction network (PPI network) results predicted that JAG2 and Notch3 participate in miR-2392 signaling regulatory network.
The Notch pathway directly couples events at the cell membrane with the regulation of transcription, which includes 4 Notch receptors (Notch1-4) and 5 Notch ligands (Delta-like ligands (Dll1, Dll3, and Dll4) and 2 serrate-like ligands (Jagged 1 and Jagged 2)) [19]. Activation of Jagged2 (JAG2) is frequently observed in several tumors and involved in the malignant biological behavior of tumor cells [20,21]. However, the mechanism that causes JAG2 activation in HCC remains to be unraveled. First of all, we compared the mRNA expression of JAG2 and Notch3 between HCC tissues and normal liver tissues (n = 371) by using the UALCAN (Data from TCGA database) (http://ualcan.path.uab.edu). The results indicated that the expression levels of JAG2 and Notch3 were higher in HCC tissues than in normal tissues. IHC and western blot assay were used to measure the expression pattern of JAG2 in HCC tissues.
Our present results revealed that JAG2 was higher in HCC tissues than in non-tumorous tissues which was one of the independent favorable factors for overall survival and recurrence, as were advanced TNM stage, and vascular invasion. Furthermore, we predicted miR-2392 potential regulatory target genes by Targetscan (http://www.targetscan.org/vert_71/). The result revealed that miR-2392 has a direct binding site with JAG2. We carried out a luciferase reporter assay to verify the direct interaction between JAG2 and miR-2392. The results con rmed that miR-2392 inhibited the luciferase activity by combining 3'-UTR of JAG2. The results of cellular immuno uorescence and western blot assay also con rmed that down- proliferation and tumorigenesis [22,23]. We further found that inhibition of miR-2392 increased the expression of MMP2, MMP9, PCNA and Vimentin which could enhance the proliferation ability of HCC cells, accelerate the degradation of extracellular matrix (ECM), eventually facilitate the invasion and metastasis of HCC cells. Additionally, the potential mechanism of MMP2, MMP9, PCNA and Vimentin regulated by miR-2392 will be further analyzed. According to our present results, we speculate that the decrease of miR-2392 expression and concentration in HCC leads to the increase of JAG2 protein expression and nuclear accumulation, which leads to transcriptional changes of a series metastasis related genes, and ultimately promotes HCC cell metastasis.

Conclusions
Accordingly, we identi ed miR-2392 functions as a tumor suppressor gene which is down-regulated in HCC, and correlates with proliferation, metastasis, spheroid formation and maintenance of CSC-like characteristics of HCC cells via directly targeting JAG2. These ndings provide a new mechanism for understanding HCC metastasis and miR-2392 may be a potential candidate in the individualized precise treatment of HCC.