New tumor suppressor microRNAs target glypican-3 in human liver cancer

Glypican-3 (GPC3) is an oncogene, frequently upregulated in liver malignancies such as hepatocellular carcinoma (HCC) and hepatoblastoma and constitutes a potential molecular target for therapy in liver cancer. Using a functional screening system, we identified 10 new microRNAs controlling GPC3 expression in malignant liver cells, five of them e.g. miR-4510, miR-203a-3p, miR-548aa, miR-376b-3p and miR-548v reduce GPC3 expression. These 5 microRNAs were significantly downregulated in tumoral compared to non-tumoral liver and inhibited tumor cell proliferation. Interestingly, miR-4510 inversely correlated with GPC3 mRNA and protein in HCC samples. This microRNA also induced apoptosis of hepatoma cells and blocked tumor growth in vivo in the chick chorioallantoic membrane model. We further show that the tumor suppressive effect of miR-4510 is mediated through direct targeting of GPC3 mRNA and inactivation of Wnt/β-catenin transcriptional activity and signaling pathway. Moreover, miR-4510 up-regulated the expression of several tumor suppressor genes while reducing the expression of other pro-oncogenes. In summary, we uncovered several new microRNAs targeting the oncogenic functions of GPC3. We provided strong molecular, cellular and in vivo evidences for the tumor suppressive activities of miR-4510 bringing to the fore the potential value of this microRNA in HCC therapy.


INTRODUCTION
MicroRNAs (miRNAs) actively participate in gene regulation in hepatocellular carcinoma (HCC), a primary cancer of the liver affecting adults [1], and in hepatoblastoma (HBL), a rare childhood neoplasm of the liver [2]. HCC is the 5 th most common cancer worldwide and the 2 nd -leading cause of death from cancer, due to its aggressiveness [1]. HCC is a heterogeneous malignancy that usually develops silently on a pre-existing diseased liver with severe fibrosis or cirrhosis [1]. It is therefore often diagnosed at advanced stages resulting in an overall 5-year survival of 20% [3]. Consequently, only one-third of patients can benefit from middle-term curative regimens and Sorafenib is the only validated drug option in patients with more advanced disease [1]. In HBL the introduction of pre-operative chemotherapy and efficient surgical practices have improved disease outcome. However, the prognosis of patients suffering Research Paper from high-risk tumors is still poor [2]. In this context new effective therapeutic options are urgently needed for the treatment of advanced liver tumors in adults and children.
MiRNAs are small non-coding RNAs capable of modulating gene expression at the post-transcriptional level [14] by interacting with specific sites, usually located in the 3' untranslated region (UTR) [11,15,16]. As regulators of most cellular functions, miRNAs are intricately involved in human diseases including cancer [17,18]. Although the effect mediated by some miRNAs on any particular target is modest, the simultaneous regulation of a broad array of target genes by one miRNA can lead to a profound genetic reprogramming and cell-phenotypic changes [17,18]. Due to their functional redundancy (e.g. multigene targeting) and regulatory properties miRNAs can be used as drugs with promising therapeutic applications in cancer. For instance MRX34, a synthetic encapsulated form of miR-34a-5p, is currently tested for the treatment of primary adult liver cancer [17,18]. Therefore, miRNA-based therapy might be a solution for the improvement of life expectancy in patients with advanced HCC or high-risk HBL.
Using a functional screening system, the Dual Fluorescence (DF)-FunREG system [11,19], we uncovered new miRNAs inhibiting GPC3 expression in HCC cells. Then, we measured expression of these miRNAs in tumoral and non-tumoral liver samples and studied their propensity to act as tumor suppressors in vitro using cell-based methods. Finally, the tumor suppressive effect of the most promising GPC3-regulating miRNA was further investigated using molecular and functional tools, as well as the chick chorioallantoic membrane model.

Selection of fourteen GPC3-regulating miRNAs by functional screening
To identify novel miRNAs negatively regulating GPC3 through its two UTRs, we performed a new in vitro functional screening of a library of 1712 individual miRNAs using the HCC-derived Huh7 cell line and an updated version of the DF-FunREG system [11]. Twenty out of 1712 miRNAs were retained according to a decrease of -0.5 or more of GFP/Tomato ratio fold change ( Figure  1A and Supplementary Table 1). Such threshold is indicative of a significant decrease resulting from the specific action of these miRNA on the UTRs of GPC3 bordering a gene reporter (GFP) in comparison to a reference gene (Tomato). Three out of these 20 miRNAs e.g. miR-96-5p, miR-1271-5p and miR-1973 were excluded from further analyses because they were subject of early investigations by our team [11,13]. The remaining 17 candidate miRNAs were transfected into Tomato-positive Huh7 cells co-expressing the "GFP" transgene lacking the GPC3 UTRs to eliminate false positive miRNAs or either the "GFP-5'UTR-GPC3" or the "GFP-GPC3-3'UTR" transgene to determine their UTR preference. These control experiments allowed us to exclude three false positive miRNAs (Supplementary Figure 2) and to keep only 14 miRNAs ( Figure 1B-1D). As expected, most of them act through the 3'-UTR ( Figure 1B) but some exert a moderate effect either through GPC3 5'-UTR [16] or both UTRs [20] ( Figure 1C). Contrarily to miRNA paralogs miR-96-5p and miR-1271-5p [11,13], the previously described GPC3-regulating miRNAs, miR-219-5p and miR-520c-3p [21,22] were below the cut-off (Supplementary Table 1) and thus not selected in our screening. The final outcome of selection by the DF-FunREG screening is a total of 14 new and potent GPC3-regulator miRNAs ( Figure 1D).

The five HCC down-regulated miRNAs exert antitumor effects
We next investigated the effect of the five downregulated miRNAs on tumor cells in vitro. All miRNAs miR-548aa, miR-376b-3p and miR-548v was measured by real-time quantitative RT-PCR in 19 NTL and 98 HCC. Data are presented as box and whiskers with minimal and maximal values (two-tailed unpaired t test). (B) The relative expression of miR-4510, miR-203a-3p, miR-548aa, miR-376b-3p and miR-548v was measured by real-time quantitative RT-PCR in 19 pairs of HCC and adjacent NTL. Results are presented as HCC/NTL expression ratios. The median is shown as a full line and the reference ratio value "1" is shown as a dotted line. The statistical analyses were done with the two-tailed Wilcoxon matched-pairs signed ranked test. *: p < 0.05, **: p < 0.01, ***: p < 0.001. significantly inhibited Huh7 cell growth and proliferation ( Figure 4A-4B). However, compared to miR-4510 alone, different combinations of the miRNAs did not further potentiate growth inhibition (data not shown). The inhibitory effect of each miRNA on Huh7 cell proliferation could be abolished by the simultaneous use of corresponding anti-miRs (Supplementary Figure  4). The specific blocking of four out of five endogenous miRNAs led to a slight increase in Huh7 cell proliferation (Supplementary Figure 4). We further studied the effect of these five miRNAs on different phases of cell division. All five miRNAs increased the percentage of cells in G0/ G1 phase and reduced the percentage of cells in S phase ( Figure 4C). Finally, we investigated the ability of these miRNAs to induce apoptosis using two complementary assays. Annexin V/7AAD staining and caspase 3/7 activity revealed that only miR-4510 induced Huh7 apoptosis ( Figure 4D and Supplementary Figure 5). Altogether these results demonstrate that miR-4510, miR-203a-3p, miR-548aa, miR-376b-3p and miR-548v are potent inhibitors of HCC cell proliferation.

MiR-4510 is a powerful antitumoral agent in liver cancer and acts through GPC3 3'-UTR targeting
We next investigated in more details expression of miR-4510 and GPC3 mRNA in liver tumor samples. We noticed that miR-4510 was constantly decreased in HCC ( Figure 3B) and inversely correlated with GPC3 mRNA ( Figure 5A) and protein ( Figure 5B) in HCC samples, making a strong and direct connection between the reduction of this miRNA and GPC3 overexpression. No inverse correlation was observed when we compared expression of other miRNAs (miR-203a-3p, miR-548aa, miR-376b-3p and miR-548v) and the amount of GPC3 mRNA or protein in patient samples (Supplementary Figure 6A-6B). MiR-4510 decrease was independent of HCC subgroup clustering ( Figure 5C) [23] and was also observed in HBL tumors ( Figure 5D). Thus a decrease of this miRNA constitutes a good indicator of liver tumorigenesis. Because miR-4510 efficiently impaired cell proliferation and induced apoptosis of Huh7 cells (Figure 4), we compared its effects on hepatoma cells with that of miR-34a-5p, a miRNA currently tested in clinic (MRX34) [1,18,24]. MiR-4510 inhibited the growth of HCC-derived Huh7 and Hep3B cells and of HBL-derived Huh6 cells (which carry a missense G34V mutation in β-catenin/CTNNB1 gene) and it was significantly more effective than miR-34a-5p in Hep3B cells ( Figure 6A). Moreover, it induced apoptosis in the three hepatoma cell lines tested while miR-34a-5p had no pro-apoptotic effect ( Figure 6B). Finally, the proliferative capacity of miR-4510-transfected Huh7 cells was partly rescued by the ectopic expression of a GPC3 transgene devoid of its 5'and 3'-UTRs (pL-hGPC3, Figure 6C). Finally, according to the screening data ( Figure 1B) and in silico predictions, a target site of miR-4510 was located in GPC3 3'-UTR at position 308-315 ( Figure 7A). To assess the relevance of this site, two point mutations (r.311C>G and r.313C>G) Results are presented as HBL/NTL expression ratios. The median is shown as a full line and the reference ratio value "1" is shown as a dotted line. Two-tailed Wilcoxon matched-pairs signed ranked test. **: p < 0.01, ***: p < 0.001. www.impactjournals.com/oncotarget were inserted in this 3'-UTR sequence. Then FunREG analysis was performed [25,26]. As expected, miR-4510 decreased the fluorescence of GFP with a wild-type GPC3 3'-UTR ( Figure 7B) demonstrating its ability to regulate GPC3 at a post-transcriptional level. Interestingly, this decrease was partly abrogated when the 3'UTR contained the mutated sequence ( Figure 7B) showing that miR-4510 physically interacts with GPC3 3'-UTR at this position. Altogether these data demonstrate that miR-4510 is a powerful antitumoral agent in liver cancer and acts in part through GPC3 downregulation.

MiR-4510 inhibits Wnt/β-catenin signaling pathway
Because miR-4510 inhibits liver cancer cell proliferation through GPC3, which is involved in Wnt/βcatenin pathway activation [5,6], we investigated miR-4510 effect on this pathway. MiR-4510 inhibited the transcriptional activity of β-catenin in Huh7 and Huh6 cells without affecting β-catenin expression ( Figure 7C and Supplementary Figure 7A) or its subcellular localization (Supplementary Figure 8). The inhibition of Wnt pathway by miR-4510 in Huh7 cells was accompanied by the decrease of associated genes including two transcriptional cofactors of β-catenin, such as LEF1 and TCF7L1 [27,28], direct targets of Wnt/β-catenin signaling (IRS1, BMP4, FGF9 and LEF1), and various genes associated with poor-prognosis HCC such as BMP4, PDGFRA and SIX1 ( Figure 7D). On the other hand, miR-4510 treatment led to an increase in several important regulators of β-catenin functions and tumor suppressors such as AXIN2, E-cadherin (CDH1), LRP1 and WNT5A ( Figure 7D). The expression of several pro-oncogenic genes in liver (GJA1, IRS1, PDGFRA and SIX1) was also decreased in miR-4510-transfected Huh7 cells ( Figure 7D). Surprisingly, we did not observe any effect of miR-4510 on MYC and Cyclin D1 mRNA in Huh7 but this might be due to the wild-type status of β-catenin or the possible influence of other signaling pathways such as p53 in these cells [1,28]. In contrast, miR-4510 decreased the amount of Cyclin D1 protein in Huh6 cells (Supplementary Figure  7B). Altogether these data showed that miR-4510 inhibits Wnt/β-catenin pathway.

MiR-4510 inhibits tumor growth and induces apoptosis of HCC cells in vivo
We next evaluated the antitumor activity of miR-4510 in vivo using the CAM model. Critical biological features of human tumor progression such as cell proliferation, angiogenesis, normal tissue invasion,   and tumor cell-host interactions have been previously successfully reproduced in this model [29,30]. Moreover, the CAM model is useful for testing small non-coding RNA-mediated gene knockdown on tumor growth and angiogenesis [31]. We transfected Huh7 cells with miR-4510 or Ctrl and validated the inhibition of GPC3 by miR-4510 (Supplementary Figure 9A-9B). We deposited these Huh7 cells on the CAM and monitored tumor growth on days 3 and 6 ( Figure 8A). At day 3 no obvious macroscopic differences were observed between Ctrl and miR-4510 in tumor appearance or size ( Figure  8B, upper panels). Staining of tissue cross-sections with Hematoxylin and Eosin did not reveal any difference either ( Figure 8B-8C, H&E panels). However, there was a significantly lower number of tumors with bleeding in presence of miR-4510 indicating a reduction of Huh7 cells aggressiveness ( Figure 8D, left panel). GPC3 protein expression was also decreased in tumors transfected with miR-4510 compared to control (Supplementary Figure  9C). As expected, miR-4510 levels were increased while GPC3 mRNA was unchanged (Supplementary Figure  9D-9E). At day 6, miR-4510 levels remained high and GPC3 mRNA expression significantly decreased (Supplementary Figure 9D-9E). At this stage, the growth of miR-4510 tumors was noticeably impeded compared to control tumors ( Figure 8B-8C, upper panels), as assessed by a disappearance of bloody and coagulation areas ( Figure 8B, upper panels) and large vessels in tumoral tissue ( Figure 8B, lower panels and 8C, upper panels). As shown in Figure 8D, right panel, 80% of Ctrl tumors and only 30 % of miR-4510 tumors were characterized by bleeding at Day 6, further indicating a reduction of the aggressiveness of miR-4510-transfected HCC cells during tumor development. Next, we characterized the effects of miR-4510 on tumor cell proliferation and apoptosis, respectively by staining of Ki67 and cleaved Caspase 3. The decrease of the proliferative marker Ki67 in miR-4510 transfected-tumors was obvious both at day 3 and day 6 of tumor growth ( Figure 8C, middle panels), indicating that miR-4510 was capable of inhibiting the proliferation of HCC. As for apoptosis analysis, cleaved Caspase-3 staining was not detectable at day 3 and 6 in control tumors. However, tumors transfected with miR-4510 were highly positive for cleaved Caspase-3 at day 6 suggesting that miR-4510 induces HCC cell apoptosis at later stages of tumor development ( Figure 8C, lower panels). Altogether these results showed that miR-4510 induces apoptosis and inhibits the growth and angiogenesis of HCC tumors in vivo.

DISCUSSION
In this work we identified new miRNAs regulating GPC3. Among them, 5 (namely miR-203a-3p miR-4510, miR-548aa, miR-376b-3p and miR-548v) inhibited GPC3 expression in HCC cells and acted as tumor suppressors in liver by inhibiting HCC cell growth and proliferation. So far, only miR-203a-3p has been described as a mediator of liver carcinogenesis [32]. Therefore, miR-4510, miR-548aa, miR-376b-3p and miR-548v are new actors in liver cancer. Remarkably, these 5 miRNAs were significantly decreased in HCC tumors (Figure 3). These results suggest that the upregulation of GPC3 in liver cancer originates in part from the simultaneous downregulation of multiple specific miRNAs as already described for other cancers [24,33,34]. Thus this might be a more general mechanism developed by cancer cells to maintain oncogenicity as discussed previously [33,35,36]. Other mechanisms responsible for the oncogenic overexpression of GPC3 in HCC have been shown before, including transcriptional activation by Myc [37] and increase in GPC3 copy number [38]. Therefore, miRNA downregulation might be one of the molecular processes set up by HCC cells to maintain high level of GPC3 and activate oncogenic downstream signaling including the Wnt pathway [1].
Several microRNAs negatively regulating GPC3 have been described in liver cancer cells [11,13,16,21,22]. Surprisingly, while we confirmed the inhibition of GPC3 by miR-219-5p (Figure 2A), this miRNA and miR-520c-3p [21,22] were not retained in our screen using GPC3 5'-and 3'-UTRs as baits (Supplementary Table  1). These data suggest that those two miRNAs may target GPC3 mRNA indirectly or through its coding region. Only miR-1271-5p and miR-219-5p are decreased in HCC samples [11,22] and we previously reported that GPC3 mRNA inversely correlates with miR-1271-5p in HBVpositive HCC subgroup [11]. Here, we report the decrease of 5 new GPC3-regulating miRNAs in two independent cohorts of liver cancer tissues. While we cannot exclude the possibility that different miRNAs (including miR-219-5p and miR-1271-5p) play a role in HCC-associated GPC3 overexpression, our data suggest that miR-4510 is of particular importance in this cancer as it strongly and inversely correlated with GPC3 mRNA and protein levels in HCC tumors ( Figure 5A-5B). MiR-4510 may also play a tumor suppressive role in colorectal adenocarcinoma in which it is decreased [39]. Therefore, we speculate that miR-4510 plays a major role in the deregulation of GPC3 in liver cancer.
Among the 10 new miRNAs identified, miR-4510 was the most potent inhibitor of GPC3 expression. It induced proliferation arrest and apoptosis in all tested cell lines more efficiently than miR-34a-5p, the first miRNA-mimic tested in clinic. Our data show that the tumor suppressor effect of miR-4510 is related to the reduction of GPC3 but also to Wnt signaling inactivation. Importantly, the inhibition of Wnt pathway mediated by miR-4510 was independent of β-catenin expression and localization. Rather, miR-4510 inactivates Wnt/β-catenin signaling not only at its initiation step through GPC3 but also at a later stage by silencing LEF1 and TCF7L1 ( Figure 7D) [28]. Moreover, it decreases expression of several oncogenic genes under the control of Wnt/βcatenin pathway including GJA1, IRS1, PDGFRA and SIX1 [40][41][42]. We speculate that miR-4510-transfected cells reactivate the inhibited Wnt pathway by upregulating the pro-inflammatory factor IL6 and several Wnt cofactors such as CDH1/E-cadherin, LRP1 and WNT5A ( Figure  7D). Thus, miR-4510 directly and indirectly inhibits several oncogenes in liver tissues and might constitute an important tumor suppressor in liver that antagonizes key tumoral processes (e.g. uncontrolled proliferation and cell survival) and inactivates signaling cascades (e.g. Wnt pathway). Further experiments aiming at identifying the other targets of miR-4510 in HCC cells should shed some light on its precise role in liver carcinogenesis.
There is a growing therapeutic interest in the use of miRNAs in clinical oncology because of their small size, easy manufacturing and pleiotropic effect [1,18,43]. MiR-34a-5p, which constitutes a miRNA mimic encapsulated in ionizable liposomes (MRX34), is being tested since spring 2013 in a multicentric clinical trial [24]. Another miRNA, miR-16, is in clinical trials since early 2015 for mesothelioma and lung cancer [43]. Given the progress of siRNA-and miRNA-replacement therapy in cancer [1,18,43], our data provides strong molecular, cellular and in vivo evidences supportive of miR-4510-based drugs as an option for the treatment of patients with HCC or HBL. The tumor growth inhibition mediated by miR-4510 in vivo further supports its relevance as a candidate for miRNA-replacement therapy for the treatment of patients with advanced liver cancer.
In conclusion, our report further demonstrates the active role of miRNA deregulations in oncogenic processes and brings new information about the complex miRNA:GPC3 relationships occurring in liver tumors. We therefore expect than the data presented in this report will favor the development of new therapeutic solutions for the treatment of patient with HCC or HBL.
To construct pGEM-T-hGPC3 plasmid, the GPC3 Open Reading Frame (ORF) was amplified by PCR using the primers 5'-ATTCTCTAGAGAATTCGGATCCATGGC CGGGACCGTGCGC-3' and 5'-CTCACTCTAGAGC GGCCGCTCAGTGCACCAGGAAG-3' and the pEF-BOS plasmid containing the human GPC3 cDNA as template, which was kindly provided by Jorge Filmus [44]. After adenylation, the PCR fragment was subcloned into the pGEM®-T vector (Promega). The sequences in all constructions were verified by DNA sequencing. The lentiviral pL-hGPC3 was constructed by subcloning the human GPC3 ORF from the pGEM-T-hGPC3 plasmid into the BamHI-XbaI-digested pL-GFP plasmid.

Cell lines
The HCC-derived Huh7 and Hep3B cell lines were grown as described before [11,13,26]. The HBL-derived Huh6 cells were grown in DMEM 1g/L (Invitrogen, Carlsbad, California, USA) containing 10% fetal bovine serum (FBS) and penicillin/streptomycin (1000 units/mL). Cell lines were yearly identified using STR profiling (ATCC-LGC Standard) and regularly tested for mycoplasma-free infection. Stable Huh7 cell lines coexpressing Tomato and GFP transgenes (with the GPC3 5'-UTR, 3'-UTR, both or neither) were developed by lentiviral transduction using a multiplicity of infection of 1 and cell sorting. Production and titration of infectious lentiviral particles, cell transduction, as well as biosafety considerations and procedures have been described previously [11,13,26].

Liver samples and clinical data
Liver tissues were treated and samples were clinically, histologically, and genetically characterized as previously described [11]. All samples were from patients recruited in accordance with French law and institutional ethical guidelines. Two sets of liver samples (set 1: 98 HCC and 19 NTL samples, Supplementary

DF-FunREG screening, FunREG analyses and real-time quantitative PCR and RT-PCR
FunREG and DF-FunREG analyses were performed three days after transfection as previously described [11,13,26] with few modifications including fluorescence measurements using the Envision multiplate reader (Perkin Elmer). Real-time quantitative PCR and RT-PCR procedures and GPC3 primers were as described elsewhere [13]. Taqman microRNA assays (Applied Biosystems) were used to quantify the relative expression levels of mature miRNAs in the first set of 117 liver samples. MiScript Sybergreen assays (Qiagen) were used to quantify the absolute expression of mature miRNAs in the second set of 38 liver paired samples.
For flow cytometry analyses, Huh7 cells were washed in PBS, detached with PBS/EDTA, collected and incubated with the fluorescent anti-GPC3 or control antibody. Expression of the membrane GPC3 protein was analyzed by FACS. Cells incubated with the IgG2a-APC isotype control were used as negative control to gate the eGFP-positive cell populations and to measure the basal Mean Fluorescence Intensity of the whole cell population.

Immunofluorescence assays
For immunostainings, 30,000 cells were transfected with control and miR-4510 and seeded on coverslips in 12-well plates in a volume of 1 ml. Three days later, cells were fixed with 4% PFA for 30 min and permeabilized with 0.1% Triton X-100 in PBS. Blocking was performed with 5% BSA in PBS for 30min. Primary antibody of mouse-anti-beta-catenin (C-14, BD Transduction Laboratories) was diluted to 1:200 in 5% BSA/PBS solution and incubated for 45 min. The secondary antibody Donkey anti-mouse IgG (H+L) 488 (Interchim Fluoprobes, #FP-SA4110) was diluted to 1:200 in blocking solution in presence of Hoechst dye (1:20,000) and incubated for 45min. Coverslips were washed and mounted on slides with Fluoromount G (Interchim). Slides were analyzed using confocal fluorescence microscopy (Leica SP5).

Cell growth and proliferation assays
Cell growth was measured with the In vitro Toxicology assay kit (Sigma) as described previously [11], except that absorbance at 492 nm was measured using the CLARIOstar multiplate reader (BMG Labtech). For cell counting experiments, 200 000 cells were transfected and seeded into 6-well plates in a volume of 2.5 ml. Three days later total cells were counted using a Malassez cell. For the rescue experiments Huh7 cell lines ectopically expressing a GPC3 transgene lacking the UTRs (pL-hGPC3) or a control empty transgene (pTRIP-0 plasmid) were established by lentiviral transduction. Then cells were transfected by miR-4510 or Ctrl. Three days later cells were counted and GPC3 protein expression was analyzed by Western blotting.

Cell cycle assay
Cell cycle was studied with the APC BrdU flow kit from BD Pharmingen according to manufacturer's instruction. Briefly, 200 000 cells were transfected and seeded into 6-well plates in a volume of 2.5 ml. Three days later BrdU was added in each well and incorporated into newly synthesized DNA by cells entering and progressing through the S phase of the cell cycle. The incorporated BrdU was detected using an APC anti-BrdU fluorescent antibody and the levels of cell-associated BrdU were then measured by FACS.

Cell death and caspase assays
Prior to apoptosis detection 200 000 cells were transfected and seeded into 6-well plates in a volume of 2.5 ml. Three days later total cells were collected and apoptosis was analyzed using the Annexin V-PE/7-Amino-Actinomycin (AAD) apoptosis detection kit (BD Pharmingen). Viable cells with intact membranes exclude 7-ADD and are PE Annexin V negative. Fluorescence generated by the cell-bound Annexin V-PE, which measures the percentage of early apoptotic cells, and the 7AAD, which measures the percentage of late apoptotic cells, were analyzed by FACS. Total apoptosis was calculated by adding the percentage of late apoptotic cells (Annexin V-PE High / 7ADD High ) and the percentage of early apoptotic cells (Annexin V-PE High / 7ADD Low ). Activities of Caspases 3 and 7 were measured using the Luminescent Caspase-Glo 3/7 assay from Promega as described previously [11], except that luminescence was measured using the CLARIOstar multiplate reader (BMG Labtech).

Wnt transcriptional activity and associated reagents
Wnt transcriptional activity was assessed using the TOPflash/FOPflash assay. Firstly 200 000 Huh7 cells were transfected with Ctrl or miR-4510 and seeded into 6-well plates in a volume of 2.5 ml. Cells were collected two days later, 10 000 cells were seeded into 96-well plates in a volume of 100 μL and transfected with the control plasmid pRL-TK-Renilla (Promega) and either the TOPflash or FOPflash plasmids kindly provided by Hans Clevers [45]. Cells were lysed 24 hours later and luciferase activity was measured using the Dual-Luciferase ® Reporter Assay System (Promega) according to manufacturer's instructions. The expression of 84 genes related to WNTmediated signal transduction was estimated using Human Wnt Signaling Targets RT² Profiler PCR Array (Qiagen) in Huh7 cells 72hr after transfection.

Chick CAM assays
Animal procedures were carried out in agreement with the European (directive 2010/63/UE) and French (decree 2013-118) guidelines. Embryos were received at the stage of segmentation and then incubated at 37.4°C at 70% humidity. At day three of development, the eggshell was opened on the top and the opening sealed with medical-grade Durapore tape. When embryos were at day 9 of development, Huh7 cells were transfected with either control or miR-4510 as described above. The next day cells were washed in PBS and 2 million Huh7 cells were deposited on the CAM, in the center of a Thermanox plastic ring. Photographs of the tumor growth were taken every day until day sixteen using a stereomicroscope (SMZ745T) and a camera (DS-Fi2, Nikon) and then analyzed with the NSI Element D software. Three and 6 days after deposition of tumor cells CAMs were fixed and processed for histology. Tumor-containing CAM were cut in 4 μm-thick sections and stained with Eosin-hematoxylin or rabbit polyclonal anti-cleaved Caspase-3 antibody (AF835; 1:200) (R&D systems), mouse monoclonal anti-GPC3 (C-1G12; 1:50) antibody (Zytomed) or a mouse monoclonal anti-Ki-67 (MIB-1; 1:75) antibody (Dako). Finally, tumor-CAM sections were scanned using a Hamamatsu Nanozoomer 2.0HT (Bordeaux Imaging Center, Bordeaux University).

Bioinformatic tools
Different algorithms of prediction were used to investigate target:miRNA interactions including TargetScan, miRDB, TargetMiner, miRanda, RNA Hybrid, PICTAR5, DIANAmt and Diana lab. The list of miR-4510-target genes generated from miRDB was imported to Ingenuity Pathways Analysis (IPA) to investigate the cellular functions and molecular pathways of miR-4510 target genes.

Statistical analyses
Statistical analyses were performed using GraphPad Prism 6.0 software. Analyses with the Mann-Whitney and ANOVA tests were done as described elsewhere [11,13]. When experiment contained two paired groups, the two-tailed Wilcoxon matched-pairs signed ranked test was used. When experiment contained two groups of categorical variable (e.g. bleeding versus no bleeding), the two-sided Fisher's exact test was used. The p-value is indicated at the bottom of each figure legend. The ANOVA test was followed by the Dunnett's multiple-comparison post-test when all data were compared to control or by the Tukey's multiple-comparison post-test when all data were compared. In each figure the number of independent experiments (n) and the ANOVA p-value is indicated in brackets. Results were considered significant when p < 0.05. For all data in figures, *: p < 0.05, **: p < 0.01, ***: p < 0.001.

Author contributions
FS and VT constructed plasmids. FC, EI, SL, JC, AG and FS performed the cellular and molecular analyses. FC, AG, AD and NDS performed in vivo and histological analyses. FC, KBH and CFG performed bioinformatic and biostatistic analyses. MH and CFG supervised the work. FC, EI, KBH, MH and CFG wrote the manuscript. CFG obtained the different grants supporting this work.
Gelabale for their technical and analytical support, and Aksam Merched for his critical reading of the manuscript.