Article Text
Abstract
Objective Hepatocellular carcinoma (HCC) is the most prevalent primary tumour of the liver. About a third of these tumours presents activating mutations of the β-catenin gene. The molecular pathogenesis of HCC has been elucidated, but mortality remains high, and new therapeutic approaches, including treatments based on microRNAs, are required. We aimed to identify candidate microRNAs, regulated by β-catenin, potentially involved in liver tumorigenesis.
Design We used a mouse model, in which β-catenin signalling was overactivated exclusively in the liver by the tamoxifen-inducible and Cre-Lox-mediated inactivation of the Apc gene. This model develops tumours with properties similar to human HCC.
Results We found that miR-34a was regulated by β-catenin, and significantly induced by the overactivation of β-catenin signalling in mouse tumours and in patients with HCC. An inhibitor of miR-34a (locked nucleic acid, LNA-34a) exerted antiproliferative activity in primary cultures of hepatocyte. This inhibition of proliferation was associated with a decrease in cyclin D1 levels, orchestrated principally by HNF-4α, a target of miR-34a considered to act as a tumour suppressor in the liver. In vivo, LNA-34a approximately halved progression rates for tumours displaying β-catenin activation together with an activation of caspases 2 and 3.
Conclusions This work demonstrates the key oncogenic role of miR-34a in liver tumours with β-catenin gene mutations. We suggest that patients diagnosed with HCC with β-catenin mutations could be treated with an inhibitor of miR-34a. The potential value of this strategy lies in the modulation of the tumour suppressor HNF-4α, which targets cyclin D1, and the induction of a proapoptotic programme.
- CANCER
- CELL SIGNALLING
- HEPATOCELLULAR CARCINOMA
- LIVER
- MOLECULAR ONCOLOGY
Statistics from Altmetric.com
Significance of this study
What is already known on this subject?
Hepatocellular carcinoma (HCC) is the most common form of primary liver cancer and the third leading cause of death by cancer worldwide, due to the lack of effective treatments. We, and others, have shown that mutations of the CTNNB1 gene, encoding β-catenin, are observed in 15%–40% of HCCs. HCCs with β-catenin mutation constitute one of the two subclasses of HCC, and are thought to follow a particular pattern of pathogenesis.
What are the new findings?
Using a transgenic mouse model developing tumours with β-catenin activation, we provide the first evidence of an oncogenic role of miR-34a in liver tumours with β-catenin mutations. An inhibitor against miR-34a slows tumour progression, and inhibits the development of new tumours in this model.
How might it impact on clinical practice in the foreseeable future?
These findings call into question the preclinical trial designed by miRNA therapeutics, using miR-34a mimics to combat liver tumour growth. Instead, we provide here proof-of-concept that a locked nucleic acid-based inhibitor of miR-34a may be useful for treatment exclusively in HCCs with β-catenin mutation.
Introduction
Liver cancer is the third leading cause of cancer-related death worldwide. Hepatocellular carcinoma (HCC) is the most frequent form of primary liver cancer, which mostly affects men with cirrhosis. HCC frequently appears in a context of chronic liver disease caused by infection with the hepatitis B and C viruses (HBV and HCV, respectively), alcohol consumption, obesity or genotoxic exposure (ref. 1 for a review). We, and others, have shown that mutations of the CTNNB1 gene, encoding β-catenin, are observed in 15%–40% of HCCs.2 ,3 HCCs with β-catenin mutation constitute one of the two subclasses of HCC4 ,5 and are thought to follow a particular pattern of pathogenesis, as most are well differentiated, cholestatic, and with a better prognosis.6 Several years ago, we engineered a mouse model in which β-catenin signalling was overactivated exclusively in the liver by the tamoxifen-inducible and Cre-Lox-mediated inactivation of the Apc (ApcKO) gene encoding a tumour suppressor, which is an inhibitor of the Wnt/β-catenin pathway, and the liver ‘zonation-keeper’.7 ,8 In this model, the absence of Apc in single hepatocytes results in the development of β-catenin-activated liver tumours with properties similar to those observed in human patients.8 If Apc is deleted in all hepatocytes, the β-catenin signalling in hepatocytes leads to the induction of a transcriptional programme very similar to that of human HCC with β-catenin mutation.8 ,9
The molecular pathogenesis of HCC has been elucidated,10 but the incidence and mortality of HCC remain high, despite the use of the tyrosine kinase inhibitor sorafenib. Different therapeutic strategies could, therefore, be used to target tumours in accordance with their pathogenic pathways. MicroRNAs (miRNAs) have recently emerged as potent diagnostic and prognostic biomarkers, and numbers of reports devoted to miRNAs have shown that these molecules act as oncogenes or tumour suppressors.11 The global analysis of the miRNAs expressed in liver tumours highlighted a miRNA signature common to all types of HCC.12–14 A loss of miR-122,13 miR-29,15 and miR-12416 has been described. Inversely, some oncogenic miRNAs have been identified: miR-21,17 miR-221/miR-222,18 and miR-224.19 Despite the number of global analyses, the role of miR-34a in liver tumorigenesis remains unclear. Even if this miRNA is currently described as a tumour suppressor in various cancers,20 two studies suggested that miR-34a levels decrease in HCCs, mainly HBV+,21 ,22 while other studies reported an induction of miR-34a, particularly in HCC with β-catenin mutations.17 ,18
We aimed to identify candidate miRNAs, regulated by β-catenin, potentially involved in liver tumorigenesis. We observed that miR-34a was overexpressed in response to β-catenin overactivation; the levels of miR-34a being inversely correlated with those of HNF-4α (hepatocyte nuclear receptor α), another key factor for hepatocyte differentiation which we have shown to antagonise β-catenin function.9 MiR-34a induction was observed specifically in mouse and human liver tumours with β-catenin mutations, and miR-34a was found to play an oncogenic role in the liver. These findings call into question the systematic use of miR-34a mimics to combat liver tumour growth (http://www.mirnarx.com/pipeline/mirna-MRX34.html). Instead, we provide here proof-of-concept that a locked nucleic acid (LNA)-based inhibitor of miR-34a may be useful for treatment exclusively in HCCs with β-catenin mutation.
Methods
Animals and reagents
Transgenic mice, liver sampling and immunostaining have been described elsewhere7 ,9 ,23 and have also been detailed in the supplementary materials.
Hepatocyte isolation
To directly lyse hepatocytes after collagenase perfusion, that is, for RNA or protein extraction, either transgenic or wild-type (wt) livers were perfused 6 days after tamoxifen injection (1.5 mg) as previously described.9 ,23 Non-parenchymal cells were excluded in the supernatant, and hepatocytes preparations were used for subsequent experiments. Periportal (PP) and pericentral (PC) subpopulations of wild-type (wt) hepatocytes were isolated by perfusion with digitonin and collagenase, as previously reported.24
Cell culture
For primary culture of hepatocytes, livers were perfused 3 days after tamoxifen injection to Apclox/lox mice, the cells were dispersed in William's medium (supplemented with 10% fetal bovine serum, penicillin-streptomycin, fungizone, 25 nM dexamethasone, insulin 4 μg/mL, and 1% Bovine Serum Albumin), and plated in wells coated with rat-tail collagen I (Gibco).
Transfection assays
The cells were transfected 4 h after plating, as described in the supplementary materials.
xCELLigence assay
Hepatocyte adhesion and cell cycle progression were measured with the xCELLigence system (Roche). Cells were transfected with 100 nM LNA or mimic 4 h after plating, as described above (see supplementary materials). The impedance was then measured for 48 h.
Western blotting
The membrane with 50 μg proteins was probed overnight with a primary antibody against the protein of interest (see online supplementary materials), and then incubated with a HRP-conjugated secondary antibody (Cell Signaling) for detection with the Enhanced ChemiLuminescence system (Biorad).
In situ hybridisation
Sections were deparaffinised and treated with 0.1 mg/mL proteinase K, then hybridised for 1 h at 55°C with 100 nmol digoxigenin-labeled-LNA-scramble or LNA-34a (Exiqon). Signals were detected with anti-digoxigenin (Roche) and NitroBlueTetrazolium/5Bromo-4Chloro-3Indolyl phosphate (Roche).
Samples processing for miRNA-seq
Total RNA was extracted from purified hepatocytes six days after tamoxifen injection with Trizol reagent (Life Technologies). For each set of conditions, we analysed four distinct hepatocyte preparations by deep sequencing on an Illumina platform (Fasteris, Switzerland). The reads generated by the HiSeq were analysed with miRDeep2 and miRAnalyzer software (see online supplementary materials).
Chromatin sonication and immunoprecipitation
The protocol for Chromatin ImmunoPrecipitation (ChIP) was adapted from that of Nelson,25 as previously described9 (see online supplementary materials).
Human samples
In this study, we included 44 patients treated for liver cancer at Cochin hospital (see online supplementary table S3). All tumour samples were frozen after surgery.
RNA extraction and RT-qPCR
Total RNA was extracted from purified hepatocytes with Trizol reagent (Life Technologies). Levels of miRNA were determined on 2 ng total mRNA, with a specific Taqman miRNA assay (Applied Biosystems). Levels of cyclin D1 and HNF-4α were determined on 100 ng total mRNA relative to 18S RNA (see online supplementary materials).
Statistical analysis
We assessed the significance of differences between two groups of samples in Wilcoxon tests if n<30, and Student's t tests if n>30. Analysis of variance was used to compare three groups of samples. p<0.05 Was considered statistically significant.
Results
MiR-34a is induced following β-catenin activation
We searched for candidate miRNAs regulated by β-catenin in hepatocytes by deep-sequencing the miRNAs expressed in the pretumoral Apc deletion (ApcKO) model in which there is pan-lobular β-catenin overactivation, comparing the results with those for wt hepatocytes. A statistical analysis showed that 123 miRNAs discriminated between ApcKO and wt hepatocytes (see online supplementary tables S1 and S2), and that these miRNAs mostly targeted the Wnt/β-catenin pathway (see online supplementary figure S1). We confirmed that miR-375 levels were significantly decreased by β-catenin activation (see online supplementary table S2), as reported in HCCs with β-catenin mutation.13 We also identified 54 miRNAs generated from the imprinted Dlk1/Gtl2 locus as positive targets of β-catenin (see online supplementary table S1). We discarded this locus for further studies as this is not a relevant target of β-catenin in human HCCs (see online supplementary materials, tables S1, S3 and figure S2). We thus focused on miR-34a which was induced by a factor of about 3 after β-catenin activation, with respect to wt hepatocytes in the miRNA-seq analysis (see online supplementary table S1) and in qPCR (figure 1A). This miRNA displayed the most significant induction following β-catenin activation after the miRNAs from the Dlk1/Gtl2 locus, and its overexpression in HCC with β-catenin mutation has already been reported.18 We collected 24 human tumours with a mutation of CTNNB1 leading to β-catenin activation, and 20 tumours without CTNNB1 mutation. This cohort covered the major aetiologies observed for HCC (HBV, HCV, alcohol abuse) (see online supplementary table S3). In our cohort of patients, miR-34a levels were higher in tumours than in normal tissue (by a factor of 2 in the absence of CTNNB1 mutation, and 5 in the presence of CTNNB1 mutation, figure 1B); this induction was significantly stronger in tumours with β-catenin mutations (p<0.005). The overexpression of miR-34a in tumours with mutations of β-catenin was correlated with a context of HCV infection or no particular risk factors (figure 1C). The sample size was too small to obtain a significant difference in miR-34a overexpression as a function of alcohol abuse status (figure 1C). The overexpression of miR-34a was confirmed by in situ hybridisation in another panel of tumours with β-catenin mutation (figure 1D). Thus, miR-34a is overexpressed in human HCCs with CTNNB1 mutations, and could constitute a major mediator of the oncogenic signal initiated by β-catenin activation in hepatocytes.
β-catenin directly regulates miR-34a expression
For confirmation of miR-34a regulation by β-catenin, we transfected isolated wt and ApcKO hepatocytes with a siRNA against β-catenin. β-catenin silencing decreased miR-34a levels by about 47% for wt hepatocytes and 33% for ApcKO hepatocytes (figure 2A). We thus investigated whether β-catenin could directly control miR-34a expression by ChIP experiments, with an antibody against TCF-4 (T-cell factor 4), the main LEF/TCF factor associated with β-catenin in liver,9 and an antibody against β-catenin. We measured the binding of TCF-4 and β-catenin to sites upstream from miR-34a identified as a TCF-4 binding site in our ChIP-seq data.9 β-catenin activation led to a significant increase in the binding of TCF-4 (figure 2B), and β-catenin (figure 2C) to a site 21 kb upstream from miR-34a. β-catenin binding was also significantly increased following Apc depletion on a more proximal TCF-4 site (see online supplementary figure S3). Overall, our data suggest that an excess of β-catenin could significantly increase miR-34a transcription in hepatocytes.
MiR-34a zonal redistribution after β-catenin overactivation
We then assessed the hepatic distribution of miR-34a by in situ hybridisation. Exceptions, such as cyclin D1,23 are known, but the direct targets of β-catenin are generally expressed in PC hepatocytes, in which Wnt/β-catenin signalling is activated.7 Counterintuitively, in normal livers, miR-34a was found to be preferentially expressed in hepatocytes around the portal vein and, more particularly, in bile duct cells (figure 3A). For confirmation of this prevalence of miR-34a in the PP subpopulation, we isolated PP hepatocytes and PC hepatocytes separately by injecting digitonin into the central vein and the portal vein, respectively. This differential perfusion also showed that PP samples were enriched in miR-34a (figure 3B). The results were non-significant, since using this technique, biliary cells are mainly lost during washing steps. Additionally, β-catenin overactivation led to a loss of miR-34a zonation, and to miR-34a expression in all hepatocytes (figure 3A). In conclusion, miR-34a appears to be expressed in hepatocytes throughout the liver only when β-catenin is overactivated.
We then searched the DIANA database (http://diana.imis.athena-innovation.gr/DianaTools/index.php?r=microT_CDS/index), a database with experimentally validated targets for miRNAs, for potential targets of miR-34a. A number of potent targets have been identified (see online supplementary table S4), but we focused on three interesting proteins: LEF-1, a transcription factor associated with β-catenin, caspase 2, recently identified as a direct target of miR-34a in the liver,26 and HNF-4α, which we recently showed to be an antagonist of β-catenin in the liver.9 These opposing activities result in a reciprocal physiological distribution in the liver, with miR-34a found in biliary cells (figure 3A) and HNF-4α in hepatocytes, consistent with the hypothesis that miR-34a could repress HNF-4α expression in biliary cells.27
MiR-34a inhibition improves HNF-4α transactivation capacity in hepatocytes displaying β-catenin-activation
We confirmed by western blotting that the levels of HNF-4α, caspase 2, and LEF-1 proteins (figure 3C) were inversely correlated with miR-34a levels in ApcKO mice. Similarly, the overexpression of miR-34a in human HCCs with CTNNB1 mutations was also associated with weaker immunostaining for HNF-4α in these tumours (see online supplementary figure S4). In a previous study, we showed that β-catenin impaired the transcriptional activity of HNF-4α on a luciferase reporter gene under the control of an HNF-4α-response element (FOP for Far-from-OPtimal Tcf-binding site).9 Having shown that β-catenin induces miR-34a, which could, in turn, regulate HNF-4α levels, we investigated the involvement of miR-34a in the antagonism between HNF-4α and β-catenin in ApcKO hepatocytes. An inhibitor of miR-34a based on LNA technology (LNA-34a), which consists of the complementary sequence for miR-34a with a methylene group in the ribose, significantly activated the FOP reporter (figure 3D). A mimic of miR-34a had no effect, probably due to the maximal miR-34a overexpression in ApcKO hepatocytes (figure 3C). In conclusion, miR-34a represses HNF-4α transcriptional activity, and appears to be involved in the β-catenin/HNF-4α antagonism in our model of β-catenin overactivation.
MiR-34a is induced in tumours displaying β-catenin activation
We then assessed the level of miR-34a in liver tumours mutated for β-catenin in mice. We compared 35 tumours with seven normal tissues obtained from wt mice that had received tamoxifen injections. MiR-34a levels in these tumours were twice those in normal tissues (figure 4A). HNF-4α mRNA levels in these tumours were also half those in normal tissues (figure 4B), confirming its tumour suppressor role in the liver.28 ,29 Interestingly, the tumours with the highest miR-34a levels also had the lowest HNF-4α levels (figure 4C, D). The deregulation of miR-34a in these tumours was independent of p53 mRNA levels (see online supplementary figure S5), despite miR-34a being a target of p53.30 Finally, we confirmed that overexpression of miR-34a in tumours with β-catenin overactivation by in situ hybridisation was associated to a loss of HNF-4α (figure 4E).
MiR-34a inhibition by a LNA
These data highlight the unique oncogenic role of miR-34a in a context of β-catenin overactivation. This led us to try to block miR-34a activity with a LNA, which has proved promising in human liver disease.31 We assessed the efficacy of this approach in cultures of hepatocytes isolated from wt and ApcKO mice. No change in the morphology of hepatocytes transfected with LNAs was observed on microscopy 72 h after transfection (data not shown). The efficacy of LNA-34a was assessed by cotransfection with a luciferase reporter construct with a miR-34a binding site in its 3′UnTranslated Region (UTR). As expected, following cotransfection with LNA-34a, miR-34a activity was impaired, resulting in an increase in luciferase activity (about three times higher than that obtained after cotransfection with LNA-scramble, figure 5A). The LNA efficacy was also confirmed by western blotting for HNF-4α: HNF-4α levels were higher in the presence of LNA-34a (figure 5B).
LNA-34a impairs hepatocyte proliferation after β-catenin overactivation
We then investigated the effect of LNA-34a on the proliferation of hepatocytes. We used the xCELLigence system which measures cell impedance in real time through the use of gold microelectrodes placed under each well. WT hepatocytes underwent one round of cell division, due to the proliferative effect of tamoxifen and collagenase perfusion (se online supplementary figure S6A). WT hepatocytes had a lower cell proliferation index than ApcKO hepatocytes (see online supplementary figure S6A), confirming that an absence of Apc leads to hepatocyte hyperproliferation.8 Transfection with LNA-34a significantly impaired ApcKO hepatocyte proliferation (35% inhibition), but had no effect on wt hepatocytes (figure 5C). Similar results were obtained with the Cell Titer assay, measuring the reduction of tetrazolium in formazan, a process dependent on NADPH (data not shown). The activity of lactate deshydrogenase in cell supernatant was not altered following LNA-34a transfection, suggesting an absence of LNA-34a toxicity (see online supplementary figure S6B). This suggests that LNA-34a has a specific antiproliferative effect on hepatocytes displaying β-catenin overactivation. Additionally, the LNA-34a transfected over 3 days successively increased the amount of caspase 2 propeptide, as well as caspase 2 L, its proapototic isoform, and also decreased the level of full-length PARP-1, suggesting its increased cleavage (figure 5D). This was not observed for wt hepatocytes (see online supplementary figure S6C). In vivo, the LNA-34a injected in mice only increased caspase 2 L level in ApcKO mice (see online supplementary figure S6D). In conclusion, LNA-34a promotes a proapoptotic signal in the ApcKO hepatocytes.
Cyclin D1 has been identified as a target of miR-34a.32 We therefore assessed cyclin D1 levels after transfection with LNA-34a. Contrary to expectations, cyclin D1 protein and mRNA levels at a lower extent were decreased in the presence of LNA-34a (figure 5E). However, this is consistent with the antiproliferative effects of LNA-34a on ApcKO hepatocytes (figure 5C). The loss of cyclin D1 may be a consequence of the increase in HNF-4α amount in the presence of LNA-34a (figure 5B), as HNF-4α has been shown to repress cyclin D133 (figure 6E). In agreement with this hypothesis, no change in cyclin D1 protein and mRNA levels was observed when ApcKO hepatocytes also lacking hnf-4α, were transfected with LNA-34a (figure 6A). LNA-34a had also a weaker antiproliferative effect on these hepatocytes isolated from a transgenic mouse model, in which both Apc and hnf4a had been deleted by Cre-Lox strategy (figure 6B). The higher proliferation rate observed for hnf4/ApcKO hepatocytes in the presence of LNA-34a was associated with a minor induction of caspase 2 L level, even if caspase 2 propeptide was efficiently increased (figure 6C); PARP amount was also not altered in this case (figure 6C). Finally, we realised ChIP experiments with an antibody against HNF-4α in ApcKO hepatocytes, and analysed the binding of HNF-4α in the presence of LNA-34a on two binding sites identified in published ChIP-seq experiments done in HepG2 cells and in murine livers (see online supplementary figure S7). We showed that the binding of HNF-4α to cyclin D1 promoter tended to increase following exposure to LNA-34a, in particular for peak 1 (figure 6D). This finding corroborates that HNF-4α mediates the decrease in cyclin D1 levels in response to LNA-34a. In conclusion, following its overactivation, β-catenin binds to the miR-34a promoter, thereby increasing its expression. MiR-34a, in turn, decreases the amount of HNF-4α leading to increases in cyclin D1 transcription and hepatocyte proliferation (figure 6E).
MiR-34a inhibition reduces tumour progression in ApcKO mice
We thus developed a tumour-suppressive strategy based on miR-34a inhibition, which could constitute the first targeted therapy for liver tumours displaying β-catenin activation. We assessed the impact of LNA-34a on liver tumour progression in an ApcKO context. Once the tumours became detectable on ultrasound scans, we injected LNA-34a (10 mg/kg) once weekly, and followed tumour development twice monthly by 2D-ultrasound (see online supplementary figure S8). We found that LNA-34a slowed the progression of tumours displaying β-catenin activation to rates about half those for untreated tumours (see online supplementary figure 7A and S8). This treatment also inhibited the development of new tumours, with only one new tumour, on average, occurring on LNA-34a treatment, versus four on LNA-scramble treatment (figure 7B). As expected, HNF-4α levels increased and cyclin D1 levels decreased in tumours treated with LNA-34a (see online supplementary figure S9A). Even if cyclin D1 level was impaired, Ki-67 labelling was similar in tumours treated with LNA-34a, and in those treated with the scramble LNA (see online supplementary figure S9B). Similar results were obtained with western blots for phospho-Histone H3 and cyclin A labelling (data not shown). However, LNA-34a promoted caspases 2 and 3 activation in tumours, in both western blot and immunohistochemistry experiments, confirming its proapoptotic effect (figure 7C,E), with minor effects on non-tumoural liver (figure 7D). Thus, treatment based on the use of LNA-34a could reduce tumour development through an increase in HNF-4α levels, leading to a loss of cyclin D1, and through a proapoptotic programme associated with the induction of caspase 2 and caspase 3 cleavage.
Discussion
The miRNA miR-34a has been widely described as a tumour suppressor. Indeed, its expression is frequently impaired in various cancers following a hypermethylation of its promoter20 including breast and prostate cancers.34 ,35 A preclinical trial is currently underway with a mimic of miR-34a encapsulated into liposomes in various murine models of solid tumours (miRNATherapeutics, http://www.mirnarx.com/pipeline/mirna-MRX34.html). However, the role of miR-34a in liver tumorigenesis remains unclear. One study suggested that miR-34a level decreases in HCC22 and is associated with migration and invasion,21 but two other studies reported the induction of miR-34a in HCC samples,17 ,18 in HCV+ sera,36 and in HCC cells in response to alcohol.2 ,6 Our findings suggest that miR-34a is an oncogenic mediator in tumours displaying β-catenin activation both in mice and humans, confirming the data reported by Pineau et al.18 Our results could reconcile the data of Lou et al and Li et al, which suggest that miR-34a was lost in two-thirds of the HCCs studied.21 ,22 These samples corresponded mostly to HBV+ tumours, a subgroup with a lower frequency of CTNNB1 mutations.10 The other third of patients in their cohort, with HCV infection, displayed an increase in miR-34a expression, and this group is likely to be enriched in CTNNB1 mutations, according to transcriptomic classifications.4 ,5 In our cohort, tumours without β-catenin mutations displayed heterogeneity in terms of miR-34a amount. A greater understanding of this variability would be very useful.
In our study, we used a LNA approach against miR-34a to affect the cell viability of hepatocytes isolated from ApcKO mice (figure 5C). The reduction of the proliferation rate was mainly associated with a decrease in cyclin D1 protein and mRNA levels (figure 5E), and with caspase 2 activation (figure 5D). The decrease in cyclin D1 protein levels was actually due to HNF-4α, because cyclin D1 levels were not altered if hnf4a was invalidated in the ApcKO model (figure 6A). Additionally, in the absence of HNF-4α, the cell proliferation of hepatocytes depleted for Apc was weakly impaired (figure 6B) in the presence of LNA-34a. LNA-34a increased HNF-4α protein level (figure 5B), and also its transcriptional activity on a luciferase construct under the control of a HNF-4 responsive element (figure 3D), and on the cyclin D1 promoter (figure 6D). Thus, HNF-4α, a tumour suppressor in the liver,28 appears to orchestrate some of the effects of miR-34a in the ApcKO model (figure 6E). More importantly, miR-34a emerges as a crucial component of the oncogenic signal induced by β-catenin in hepatocytes.
In normal livers, miR-34a was particularly expressed in bile duct cells (figure 3A, B). This specific pattern of expression may hinder the accumulation of HNF-4α protein in the biliary cells, which is required for bile cell lineage patterning27 probably independently of β-catenin, which is not activated in normal bile duct cells.37 The pattern of miR-34a expression is not consistent with the usual model in which β-catenin targets are located near the central vein. Cyclin D1 is one exception to this model; its pattern of expression is not restricted to the PC zone.23 Interestingly, the region bound to TCF-4/β-catenin located in miR-34a promoter contains no classical Wnt-responsive element, suggesting that the control on miR-34a is complex (figure 2B). This complexity is lost when β-catenin is overactivated, as in HCCs.
The key finding of this study was the decrease in liver tumour growth afforded by the delivery of LNA-34a (see online supplementary figures S7A and S8). Only a few studies to date have demonstrated the potency of miRNA-based approaches to cancer treatment. However, the great advantage of this approach for HCC is that these molecules are preferentially delivered to the liver.38 The first authorisation for clinical trials with miRNA-based therapies concerns the LNA against miR-122 (miravirsen) for HCV treatment.31 The use of these modulators poses a number of challenges, particularly for their specificity. Our therapeutic approach, involving the inhibition of miR-34a, displays no toxicity (see online supplementary figure S6B), and has the great advantage of restoring the tumour suppressor HNF-4α, targeting cyclin D1 and the metabolic pathways regulated by HNF-4α (see online supplementary figure S9). Additionally, this approach may also alter the anti-inflammatory loop orchestrated by HNF-4α, controlling interleukin-6 production by hepatocytes.39 ,40 More importantly, the LNA-34a promotes a proapoptotic signal orchestrated by caspases 2 and 3 (figure 7C–E). In conclusion, our data reveal that miR-34a inhibition mainly induces a pro-apototic programme centred onto caspase 2, and could affect cell proliferation through cyclin D1 inhibition. However, the fact that Ki-67 and cyclin A labellings were not affected by LNA-34a treatment, suggests that miR-34a effects on cell proliferation are more complex.
In conclusion, despite the current preclinical trials being run by miRNA therapeutic, our findings suggest that the precise differential diagnosis of HCC, including an assessment of the mutational status of β-catenin, for which miR-34a inhibitors would be more appropriate than miR-34a mimics, is vital. Using this approach, we hope to affect the tumour itself, and also its environment, through the first effective targeted therapy for one-third of HCCs worldwide.
Acknowledgments
We thank Dr F Gonzalez for the hnf4-/- mice. We are grateful to Dr P Bossard and M Caüzac for LDH protocol, and to Dr A-F Burnol for the gift of PARP-1 antibody.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online supplement
- Data supplement 2 - Online table S4
Footnotes
Contributors AG: study concept and design and drafting of the manuscript; CS, LB and CG: technical support; CM and GR: acquisition of ultrasound data; FT and PN: analysis of miRNA-seq data; BT and CC: gift and genotyping of patient tumours; CP and SC: funding, study supervision and critical revision of the manuscript.
Funding This work was supported by the Ligue Nationale Contre le Cancer (Labellisation Program 2011-2013 and 2014–2016), the Agence Nationale de la Recherche (WNT-METABOLIV. 2010–2014), the Institut National du Cancer (Epigenetics and Liver Cancer. 2014–2016). The Association Française pour l'Etude du Foie also funded experimental work and fellowships for AG. This work was supported by the SESAME funding program from Region Ile de France.
Competing interests None.
Patient consent Obtained.
Ethics approval All procedures were carried out according to the French guidelines for the care and use of experimental animals. All animal studies were approved by the regional veterinary services of the Paris police headquarters (agreement no. 75-1306) and by the Comité d'Ethique pour l'Expérimentation Animale Paris Descartes and its registered number is CEEA34.SC.047.12. For patients with HCC, samples were obtained with written informed consent, and the study protocol was approved by the ethics committee of the Cochin Hospital.
Provenance and peer review Not commissioned; externally peer reviewed.