Up-regulation of miR-95-3p in hepatocellular carcinoma promotes tumorigenesis by targeting p21 expression

Hepatocellular carcinoma (HCC) is one of the most common malignant cancers. To elucidate new regulatory mechanisms for heptocarcinogenesis, we investigated the regulation of p21, a cyclin-dependent kinase (CDK) inhibitor encoded by CDKN1A, in HCC. The expression level of p21 is decreased with the progression of HCC. Luciferase assays with a luciferase-p21-3′ UTR reporter and its serial deletions identified a 15-bp repressor element at the 3′-UTR of CDKN1A, which contains a binding site for miR-95-3p. Mutation of the binding site eliminated the regulatory effect of miR-95-3p on p21 expression. Posttranscriptional regulation of p21 expression by miR-95-3p is mainly on the protein level (suppression of translation). Overexpression of miR-95-3p in two different HCC cell lines, HepG2 and SMMC7721, significantly promoted cell proliferation, cell cycle progression and cell migration, whereas a miR-95-3p specific inhibitor decreased cell proliferation, cell cycle progression and cell migration. The effects of miR-95-3p on cellular functions were rescued by overexpression of p21. Overexpression of miR-95-3p promoted cell proliferation and tumor growth in HCC xenograft mouse models. Expression of miR-95-3p was significantly higher in HCC samples than in adjacent non-cancerous samples. These results demonstrate that miR-95-3p is a potential new marker for HCC and regulates hepatocarcinogenesis by directly targeting CDKN1A/p21 expression.

regulation of p21 is important for further understanding of HCC progression and for exploring new treatment and prevention for HCC.
MicroRNAs (miRNAs) are a cluster of noncoding RNA molecules that are approximately 18-25 nucleotides in length and negatively regulate the expression of downstream target genes mainly through direct interaction with the 3′ -untranslated regions (3′ -UTR) of their corresponding mRNA targets 19 . MiRNAs generally decrease target gene expression through mRNA degradation and/or translational suppression 20 . Increasing evidence in the recent years indicates that many microRNAs play critical roles in tumorigenesis and progression 21,22 . In this study, we found that CDKN1A/p21 was regulated by miR-95-3p by targeting the 3′ -UTR. We further showed that expression of miR-95-3p was up-regulated in HCC, whereas expression of p21 was decreased with progression of HCC. Overexpression of miR-95-3p led to increased tumor cell proliferation and growth in mice. Our study defines miR-95-3p as a new oncogenic miRNA involved in hepatocarcinogenesis.

Results
Expression of p21 is decreased with progression of HCC. The p21 protein is a tumor suppressor which has been reported to participate in tumor progression and proliferation 6 . In this study, we characterized the expression level of p21 in a total of 60 HCC samples and 3 adjacent non-cancerous tissue samples (NCT) using immunohistochemistry. As shown in Fig. 1, the expression level of p21 was high in the 3 adjacent NCT samples. As HCC progresses from stage II to a higher grade of stage IIIC, the expression levels of p21 were decreased (Fig. 1).
Identification of a repressor element regulating p21 expression at the 3′-UTR of CDKN1A. One potential mechanism for down-regulation of p21 in HCC samples is posttranscriptional regulation by microRNA at 3′ -UTR of the CDKN1A gene (encoding p21). To identify such a posttranscriptional mechanism for p21 regulation, we constructed a luciferase reporter with the 1,539 bp 3′ -UTR of CDKN1A sub-cloned downstream of the luciferase gene in the pMIR-REPORTER vector, resulting in the pMIR-p21-wt reporter (reporter pMIR-1 in Fig. 2a,b). Serial deletions were then made in pMIR-p21-wt, resulting in construction of 8 different deletion mutant reporters (reporter pMIR-2-9 in Fig. 2b). Luciferase assays in HepG2 cells showed that compared to the vector pMIR-REPORTER, luciferase activity of reporter pMIR-1 was reduced, suggesting that there is a repressor element in the 3′ -UTR of CDKN1A for regulation of CDKN1A/p21 expression (Fig. 2c). When the region from + 1,500 to + 1,515 at the 3′ -UTR of CDKN1A was deleted, the luciferase activity was significantly increased to the level of pMIR-REPORTER (compare reporter pMIR-9 to others, Fig. 2c). The data suggest that the 3′ -UTR repressor element regulating p21 expression is located at the 15-bp region between 1,500 bp and 1,515 bp from the stop codon TAA at the 3′ -UTR of CDKN1A. MiR-95-3p regulates expression of p21 by targeting the 3′-UTR of CDKN1A. We analyzed the 3′ -UTR repressor element regulating p21 expression for a potential microRNA binding site using miRBase (http://www.mirbase.org/). A potential binding site for Hsa-miR-95-3p was found at the 3′ -UTR repressor element (Fig. 3a).
To further validate the finding that miR-95-3p regulates the expression of p21, we examined the protein expression level of p21 using Western blot analysis with HepG2 and SMMC7721 cells transfected with miR-95-3p mimics vs. Ncontrol, a miR-95-3p specific inhibitor, and a negative control miRNA inhibitor (NC inhibitor). As expected, compared to Ncontrol mimics, miR-95-3p mimics significantly reduced the protein expression level of p21 (Figs 4a and 5a). On the contrary, the miR-95-3p inhibitor significantly increased the protein expression level of p21 in both HepG2 and SMMC7721 cells compared to the NC inhibitor (Figs 4b and 5b).
The expression level of the CDKN1A mRNA was also assessed using real time RT-PCR analysis. The real-time RT-PCR analysis showed that the CDKN1A mRNA expression level was slightly decreased in HepG2 and SMMC7721 cells transfected with miR-95-3p mimics compared to Ncontrol mimics (Supplementary Fig. S1a and Supplementary Fig. S2a), and slightly increased by the miR-95-3p inhibitor compared to the NC inhibitor (Supplementary Fig. S1b and Supplementary Fig. S2b). However, the differences were not statistically significant (P > 0.05). Taken together, these data suggest that miR-95-3p down-regulates p21 expression mainly at the protein or translational level by directly targeting the 3′ -UTR.

MiR-95-3p promotes tumor cell proliferation and migration. Considering the finding that miR-95-3p
negatively regulates expression of tumor suppressor p21, we hypothesized that miR-95-3p could promote tumor cell proliferation and migration as well as tumor growth. To determine the impact of miR-95-3p on HCC cell proliferation, we transfected HepG2 and SMMC7721 cells with miR-95-3p mimics vs. Ncontrol, and the miR-95-3p specific inhibitor vs. the NC inhibitor. Cell proliferation assays with the CCK-8 kit revealed that overexpression of miR-95-3p significantly promoted cell proliferation of HepG2 and SMMC7721 cells (Figs 4c and 5c). The miR-95-3p inhibitor significantly reduced cell proliferation of HepG2 and SMMC7721 cells (Figs 4c and 5c). To examine whether the observed effects of miR-95-3p are due to down-regulation of p21, we co-transfected HepG2 and SMMC7721 cells with a p21 expression plasmid together with miR-95-3p mimics. Overexpression of p21 significantly abrogated the effects of miR-95-3p mimics on cell proliferation, cell division and migration in both HepG2 and SMMC7721 cells (Figs 4c,d,e and 5c,d,e). Taken together, these results suggest that miR-95-3p promotes HCC tumor cell proliferation, cell cycle progression and cell migration by directly targeting p21. MiR-95-3p promotes tumor growth in mice. We used mouse Hepa1-6 cells to create a xenograft tumor model in mice for HCC. AgomiR-95-3P or AgomiR-NC was transfected into Hepa1-6 cells. Transfected cells were injected subcutaneously into the back of the C57BL/6 mice and tumor growth was monitored. At the end of the study, tumors were excised, weighed and photographed (Fig. 6a). The tumor size from the AgomiR-95-3p group was larger than those from the AgomiR-NC group (Fig. 6a,b). The tumor weight from the AgomiR-95-3p group was significantly heavier than that from the AgomiR-NC group (Fig. 6c). The tumor growth curves showed that the tumors from the AgomiR-95-3p group grew faster than those from the AgomiR-NC control group (Fig. 6d). Real time qPCR analysis showed that the expression level of miR-95-3p was significantly higher in tumors from the AgomiR-95-3p group than those from the AgomiR-NC group (Fig. 6e). Together, these data suggest that overexpression of miR-95-3p promoted tumor growth in vivo.
We used immunohistochemical staining with an anti-Ki67 antibody to examine the density of positive Ki67 proliferating cells. An anti-p21 antibody was also used to examine the expression level of p21 in mouse tumor sections. The density of positive Ki67 cells was much higher in tumors from the AgomiR-95-3p group than those from the AgomiR-NC group (Fig. 7a). Immunohistochemical analysis showed that the expression level of p21 was much lower in tumors from the AgomiR-95-3p group than those from the AgomiR-NC group (Fig. 7a). Western blot analysis also showed that p21 expression was lower in the AgomiR-95-3p group than in the AgomiR-NC group (Fig. 7b,c). These data provide in vivo evidence that miR-95-3p down-regulates the expression of p21, which leads to increased cell proliferation and tumor growth.
MiR-95-3p expression was increased in HCC tissues. We used semi-quantitative RT-PCR analysis to measure the relative expression level of miR-95-3p in 10 pairs of HCC samples and their respective adjacent non-cancerous samples. In each pair, the expression level of miR-95-3p was consistently higher in HCC samples than in adjacent non-cancerous samples (Fig. 8a,b,d). Together, the expression level of miR-95-3p was significantly higher in HCC samples than in adjacent non-cancerous samples (P < 0.01; Fig. 8c,e).

Discussion
In this study, we demonstrated that the expression level of miR-95-3p was consistently up-regulated in HCC tissues compared with adjacent non-cancerous tissues in all 10 groups of samples examined (Fig. 8). Moreover, the expression levels of miR-95-3p in HCC tissues were higher than that from all non-tumorous tissues (Fig. 8). Because the sample size is small, future studies with large sample sizes are needed to further validate this interesting finding. If confirmed, miR-95-3p may serve as a potential marker for diagnosis of HCC.
We found that overexpression of miR-95-3p significantly promoted HepG2 and SMMC7721 cell proliferation and migration in cultured cells (Figs 4 and 5) and Hepa1-6 tumor cell proliferation and tumor growth in mice (Figs 6 and 7). Therefore, it is highly likely that up-regulation of miR-95-3p in HCC tissues is causative to hepatocarcinogenesis and tumor growth. Mechanistically, we showed that miR-95-3p causes hepatocarcinogenesis by posttranscriptional suppression of p21 expression by binding to the 3′ -UTR. Several pieces of evidence strongly supports the conclusion. First, we found that the expression level of p21 is reduced with the progression of HCC. When HCC reached stageIIIC, there is little expression of p21 (Fig. 1). Second, we constructed several luciferase reporters which contain different regions of 3′ -UTR of CDKN1A cloned downstream of the luciferase gene. Luciferase assays showed that there is a potential repressor element in the 3′ -UTR of CDKN1A, which contains a miRNA binding site for miR-95-3p (Fig. 2). Overexpression of miR-95-3p mimics decreased the expression of p21, and the effect was eliminated by mutation of the binding site for miR-95-3p (Figs 3, 4 and 5). Moreover, a miR-95-3p specific inhibitor increased the expression of p21 (Figs 4 and 5). Third, overexpression of miR-95-3p in a mouse hepatoma xenograft model decreased expression of p21 (Fig. 7). Together, we conclude that CDKN1A encoding p21 is a downstream gene regulated by miR-95-3p. Similar to our finding of up-regulation of miR-95-3p in HCC tissues and promotion of tumorigenesis by overexpression of miR-95-3p, three other reports revealed involvement of miR-95-3p in other types of tumors. The expression level of miR-95-3p was reported to be up-regulated in glioma tissues and down-regulation of miR-95-3p inhibited proliferation and invasion and promoted apoptosis of glioma cells by targeting CELF2 encoding CUGBP-and ETR-3-like family 2 23 . The expression level of miR-95-3p was also up-regulated in human prostate and breast cancer tissues or after ionizing radiation. Overexpression of miR-95-3p promoted radiation resistance and cell proliferation following ionizing radiation and increased tumor cell invasiveness as well as tumor growth by targeting the sphingolipid phosphatase SGPP1 24 . The expression level of miR-95-3p was up-regulated in human non-small cell lung cancer tissues and overexpression of miR-95-3p increased tumor growth in xenograft mouse models by targeting SNX1 encoding sorting nexin1 25 . Our study identified the CDKN1A gene encoding p21 as a new target gene for miR-95-3p. In contrast, overexpression of miR-95-3p was found to inhibit brain metastasis of lung adenocarcinoma by suppressing expression of cyclin D1 26 . Together, these studies indicate that miR-95-3p plays important roles in tumorigenesis in different types of cancer by targeting different downstream genes.
There are several limitations with the present study. (1) We have shown that down-regulation of p21 appears to be responsible for the effect of miR-95-3p on cell proliferation, cell cycle progression and cell migration in two independent HCC cell lines because co-expression of p21 rescued the effects of miR-95-3p (Figs 4 and 5). However, as discussed above, miR-95-3p also regulates other target genes such as CELF2, SGPP1, SNX1 and many other unidentified target genes, future studies are needed to investigate the roles of other miR-95-3p target genes in the pathogenesis of HCC. (2) There are other microRNAs that can also regulate the expression of p21. It should be interesting to investigate the roles of other p21-regulating miRNAs in the pathogenesis of HCC.
(3) Overexpression of microRNAs may have off-target effects, and caution should be used in interpreting the data, although complementary studies of both miR-95-3p mimics and a miR-95-3p specific inhibitor strengthened the conclusions. (4) The 15-bp 3′ -UTR region missing in pMIR-9 (Fig. 2) also contains a less-well matched seed sequence for miR-545-5p (6-nucleotide match). It may be interesting to examine whether miR-545-5p also regulates expression of CDKN1A in the future. (5) The sample sizes of three normal healthy liver specimens for the immunohistochemistry of p21 (Fig. 1) and 10 pairs of HCC tissue samples and adjacent non-cancerous samples In summary, we have found that miR-95-3p is up-regulated in HCC tissues compared with adjacent non-cancerous tissues. Overexpression of miR-95-3p promoted tumor cell proliferation and migration in cultured cells and tumor growth in xenograft mouse models through negative posttranscriptional regulation of p21 by directly targeting the 3′ -UTR. This study establishes miR-95-3p as a potential biomarker for diagnosis of HCC and as a new therapeutic target for treatment and prevention of HCC.

Methods
Cell lines and human tumor samples. Three HCC cell lines, including HepG2, SMMC7721, and Hepa1-6, were cultured in DMEM media supplemented with 10% fetal bovine serum (FBS) (Gibco Life Technologies, Gaitherburg, MD, USA) under 5% CO 2 and at 37 °C. We screened HepG2 and SMMC7721 cells for mutations in genes encoding p53, p21 and MDM2, but no mutation was found.
HCC tissues and matched human adjacent non-cancerous liver tissue samples (NCT) were collected from patients undergoing surgical resection of tumors in Affiliated Hospitals of Huazhong University of Science and Technology. Clinical stages were classified according to the International Union against Cancer TNM classification system 27 . The demographical features of age and sex and clinical stages of HCC patients were listed in Supplementary Tables S1 and S2. This study was approved by the Ethics Committees on human subject research of Huazhong University of Science and Technology and local institutions and written informed consent was obtained from all study subjects. This study conformed to guidelines set forth by the Declaration of Helsinki.
Primers used in this study for plasmid construction and mutagenesis were listed in Supplementary Table S3.
Measurement of miRNA expression using quantitative RT-PCR (qRT-PCR) analysis. We quantified the relative expression level of miR-95-3p using stem-loop real time PCR as previously described 35,36 . Real-time RT-PCR analysis. Total cell or tissue RNA samples were extracted and reverse-transcribed to cDNA by the RevertAid First Strand cDNA synthesis kit (Fermentas) using random primers (Promega, Madison, WI, USA). The expression level of CDKN1A mRNA was quantified using a FastStart Universal SYBR Green Master kit (Roche Applied Science, Mannheim, Germany) as described by us previously [31][32][33]37,38 . The endogenous control was ACTB (encoding β -actin). Primers for real-time RT-PCR analysis were listed in Supplementary Table S3. Data were analyzed using the 2 −△△Ct method as described 36 .
Western blot analysis. Western blot analysis was carried out as described by us previously 31 In vivo tumor growth assays were performed as previously described 39,40 . In brief, 150 nM of AgomiR-95-3p or AgomiR-NC was transfected into mouse Hepa1-6 cells using Lipofectamin RNAiMAX (Gibco Life Technologies, Gaithersburg, MD, USA). After 24 h of transfection, cells were harvested, washed with cold PBS and suspended at a concentration of 5 × 10 6 cells/ml in PBS. C57BL/6 mice were divided into 2 groups (n = 7). These mice were subcutaneously injected with 5 × 10 5 Hepa1-6 cells (100 μ l) on the back. Tumor growth was examined every 2 days beginning at day 4. Tumor length (L) and width (W) were measured and the formula of V = (L × W 2 ) × 0.5 was used to calculate the tumor volume (V).
At the end of the experiment at day 22, the tumors were excised from mice and weighed.
Immunohistochemical staining. Tumors excised from mice were fixed in 4% paraformaldehyde and used for immunohistochemical staining. Immunohistochemical staining was performed as described 31,33,41 . Briefly, 4.5 μ m-tumor sections were immunostained with an antibody against p21 (1:100 dilution, Proteintech, Wuhan, China) and an antibody against Ki-67 (1:100 dilution, Proteintech, Wuhan, China). DAB PI was used for staining of nuclei. Hematoxylin-eosin staining was used to evaluate the morphology of the tumor sections as described by us 31 . Images were captured under a microscope.
Cell cycle assay. Cell cycle analysis was performed as described 31 with cells transfected with miR-95-3p mimics, negative control miRNA mimics (Ncontrol), a miR-95-3p inhibitor or a negative control inhibitor (NC inhibitor). Cells were stained with propidium iodide (PI). Cell cycle analysis was performed using a Beckman Coulter Cytomics FC 500 flow cytometry and CXP software (Beckman Coulter).
Cell proliferation and migration assays. Cells were seeded onto 96-well plates, and assayed for proliferation at 48 h using a CCK-8 kit (Dojindo Laboratories, Kumamoto, Japan) according to the manufacturer's instruction. Cell proliferation was analyzed by measurement of absorbance at 450 nm using a microplate reader as described 42 .
For cell migration, we used a wound assay as described 34 . Cells (5 × 10 5 ) were seeded onto six-well plates and cultured under standard conditions. When the cell density reached confluence, a wound was made by scraping the cell monolayer with a 200 μ l pipette tip. Cell migration was determined by measuring the movement of cells into the scraped area. The process of wound closure was monitored and photographed 12 hours after wounding under a microscope.
Bioinformatic and statistical analyses. We used the miRBase (http://www.mirbase.org/) database to predict the putative binding sites of miRNAs.
All quantitative data for statistical analyses were from at least three independent experiments. The data were presented as means ± SD (standard deviation). Statistical analysis was performed using a Student's t test. A P value of < 0.05 was considered to be statistically significant.