MiR-30e-3p Influences Tumor Phenotype through MDM2/TP53 Axis and Predicts Sorafenib Resistance in Hepatocellular Carcinoma

Hepatocellular carcinoma (HCC) accounts for 90% of primary liver cancers representing the second leading cause of cancer mortality with increasing incidence in western countries (1). HCC is characterized by a high inter- and intraindividual tumor heterogeneity: the former ascribed to different risk factors and genomic contexts (2) and the latter to clonal evolution of cancer cells (3), both contributing to limited targeted therapy efficacy. Primary resistance and escape from antitumor strategies remain poorly understood, and no biomarker predicting response to sorafenib or other targeted treatments has been identified so far, highlighting the need for novel tools to allocate patients to the best treatment.

The deregulation of tumor-specific miRNAs with tumor-suppressor or oncogenic properties has been described (4, 5). Our group and others demonstrated the key role of deregulated miRNAs in tumor progression and metastasis (6, 7), as well as their involvement in drug resistance phenotype (8–10). Aberrant expression of miR-30 family members is a frequent event in solid tumors (11, 12), displaying a prominent connection with other deregulated miRNAs in neoplastic tissues (13). The miR-30 family contains five members encoded by six genes located on three distinct loci of the human genome (chromosomes 1, 6, and 8) giving rise to six mature miRNAs (miR-30a, -30b, -30c-1, -30c-2, -30d, -30e) with the same seed sequence. The downregulation of these miRNAs was associated with proliferation and invasion, as well as with induction of epithelial-to-mesenchymal transition (EMT), exerted by direct targeting of EMT-associated factors (14, 15) and cell adhesion molecules (16). Remarkably, besides studies supporting the tumor suppressor role of miR-30 family, other studies point to its opposite behavior (17, 18).

TP53 is one of the most frequently mutated genes in human cancers with about 30% of affected HCC cases (19). TP53 regulates miRNA expression by acting at two different levels: on the one hand, it activates their transcriptional regulation (20–22) and, on the other hand, it interferes with their maturation process by interacting with key molecules, such as Drosha and p68 RNA helicase (23). Notably, TP53 regulates the expression of miR-30 family members contributing to cell invasion and distal spreading as reported in colorectal and breast cancers (24, 25).

Despite the prominent role of miR-30 family as tumor suppressor miRNAs, there are conflicting data regarding its aberrant expression in HCC. Previous studies reported the downregulation of miR-30a and miR-30e and the upregulation of miR-30d (26–28). Although miR-30 was reported as participating in hepatobiliary development (29), its functional roles in HCC remain poorly understood. Here, we investigated the expression of miR-30 family members in two cohorts of patients with HCC, in an HCC animal model and in HCC cells, and focused on miR-30e-3p, characterizing its biological activity as well as its involvement in drug resistance and modulation of MDM2/TP53 and PTEN/AKT axes. Moreover, we investigated circulating miR-30e-3p levels in a preliminary series of patients with HCC and in preclinical models in order to evaluate its contribution as a noninvasive biomarker of treatment response.

HCC and cirrhotic tissues were obtained from two independent cohorts of patients undergoing liver surgery for HCC at the Department of Surgery and Transplant Unit of St. Orsola-Malpighi University Hospital of Bologna. The discovery cohort consists of 22 patients, whereas the validation cohort consists of 48 patients, all enrolled at the Department of Surgery of St. Orsola-Malpighi University Hospital. We collected tissue samples at surgery and stored as previously described (30). Local ethics committee of St. Orsola-Malpighi University Hospital approved this study (138/2015/O/Tess). Normal liver tissues were from patients undergoing liver surgery for traumatic lesions or hemangioma resection. The clinical characteristics of patients are detailed in Supplementary Table S1. TP53 mutations were identified as previously described (20).

A further cohort of patients with advanced HCC treated with sorafenib (Supplementary Table S2) was tested for serum miR-30e-3p levels before treatment and at 2-month follow-up assessment. Sera samples, obtained after the local committee approval (271/2012/O/Oss), were processed as previously described (31). Informed written consent has been obtained from patients enrolled in this study.

The diethyl-nitrosamine (DEN)–induced HCC rat model and the xenograft model were established and treated as previously described (9). Total RNA was extracted from frozen tissues by using TRIzol Reagent (Invitrogen) and was analyzed by microarray and qPCR. Local ethics committee approved the study protocols (14/70/12 and 23/79/14).

RNAs from 22 rat samples (12 HCCs, 8 surrounding livers, and 2 normal livers) were hybridized on Agilent rat whole-genome miRNAs microarray (#G4471A_046066 Release 19.0, Agilent Technologies). One-color gene expression was performed according to the manufacturer's procedure. Raw data are available in ArrayExpress repository (accession number E-MTAB-7624).

Technical details are described in the Supplementary section.

HCC cell lines were cultured as previously described (9). Starvation was obtained by growing cells in culture media without fetal bovine serum for 48 to 60 hours. Cells were treated with 5.0 to 7.5 μmol/L sorafenib-tosylate (Bayer) or Nutlin-3 (Sigma-Aldrich) or with 1.0 to 2.0 μg/mL of doxorubicin (Pfizer) for 48 hours. Cell transfection and proliferation, clonogenic, and sphere formation assays are detailed in Supplementary Material. Genetics Unit at S.Orsola-Malpighi Hospital performed HCC authentication, and cell identification was obtained by online short tandem repeat analysis (https://www.dsmz.de/).

Note that 3′ untranslated regions (3′UTR) of MDM2, CDKN1B/p27, PTEN, and EpCAM mRNAs were amplified by PCR as reported in Supplementary Table S2. Mutagenesis of miR-30e-3p seed sequence was performed by using QuikChange II Site-Directed Mutagenesis Kit (Agilent Technologies) following the manufacturer's instruction (Supplementary Table S3). Sanger sequencing verified mutated sequences. Dual-luciferase reporter and p53 activity assays were performed as previously described (32).

TaqMan MicroRNA Assays (Applied Biosystems) were used to evaluate miRNA expression, as previously described (30). RNU6B was used as housekeeping gene. qPCR was used for gene expression analysis. β-Actin and GAPDH housekeeping genes were considered for gene normalization. qPCR experiments were run in triplicate. Primers and conditions are detailed in Supplementary Table S4.

Western blot was used to analyze protein extracts (30 μg) from cell and tissues with antibodies reported in Supplementary Table S5. ChemiDocTM XRS+ (Image LabTM Software, Bio-Rad) was employed to quantify digital images of X-ray films. Western blot analysis was performed in duplicate.

HepG2 cells were subjected to chromatin immunoprecipitation (ChIP) with a polyclonal p53 antibody (Novocastra) as previously described (21). Primers and conditions are reported in Supplementary Table S4. Amplicon design has been detailed in Supplementary Material. Electrophoretic mobility shift assay (EMSA) was performed with nuclear extract (NE) from Nutlin-3–treated HepG2 cells by using LightShift Chemiluminescent EMSA Kit (Thermo Scientific) as previously described (21). Probe sequences are reported in Supplementary Table S6.

Annexin-V assay was performed in duplicate by flow cytometry (FACSaria I, BD) as previously reported (33). Annexin-V assay in miR-30e-3p–overexpressing HepG2 cells was assessed on Cytoflex S (Beckam Coulter). Immunophenotype analysis of EpCAM was performed by using CD326 monoclonal antibody (MH99)-Alexa Fluor 488 (eBioscience).

Real-time cell invasion, performed on xCELLigence DP instrument (ACEA), and wound-healing assays were executed as previously described (34).

Caspase pathway activation was evaluated by Caspase-Glo 3/7 assay (Promega) according to the manufacturer's instructions. Each sample was performed in quadruplicate in two independent experiments.

Isolation of circulating miRNAs from exosomes, cell culture supernatant, and serum was executed as previously reported (9, 31).

Differences between two or more groups were analyzed using unpaired Student t test or ANOVA. Tukey post hoc test was used for comparisons among groups after ANOVA analysis. Pearson's correlation coefficient was used to investigate relationships between two variables. Time to recurrence curve based on miR-30e-3p levels was computed by Kaplan–Meier product-limit method and compared using a log-rank test. Paired t test was used to evaluate the relationship between circulating miR-30e-3p and response to sorafenib in patients with HCC analyzed both before and on treatment. Reported P values were two-sided and considered significant when lower than 0.05. Statistical calculations were executed using SPSS version 20.0 (SPSS inc). *, P < 0.05; **, P < 0.01; ***, P < 0.001; and ****, P < 0.0001.

We previously reported a deregulation of miR-30 family members in histologically aggressive HCCs (35). Here, we confirmed in a first patient cohort, a decrease of miR-30e-3p in 73% of HCCs in comparison with matched cirrhotic livers (t test; P = 0.0002), showing a mean downregulation of 1.7-fold (Fig. 1A). Similarly, a decrease of miR-30e-5p and miR-30a-3p was detected in 64% of tumors with an average 1.6-fold change (t test; P = 0.003 and P = 0.008, respectively; Fig. 1B and C). A positive correlation between these miR-30 family members was detected in HCC and cirrhotic human samples (Supplementary Fig. S1A–S1F), suggesting common regulation mechanisms for these miR-30 family members.

We next tested the DEN-induced HCC rat model (9) by microarray analysis on 12 HCCs and 10 noncancerous livers. The downregulation of five miRNAs belonging to miR-30 family was observed in tumors compared with surrounding livers (Supplementary Fig. S2; Supplementary Table S7). In rat HCC, miR-30e-3p decreased in 64% of tumors, showing a 1.5-fold change (t test; P = 0.034), mirroring findings from human HCCs (Fig. 1D). MiR-30e-5p and miR-30a-3p were downregulated in 59% and 32% of rat HCCs, showing no difference between tumor and nontumor samples (Fig. 1E and F). The discrepancy between human and rat HCCs might be ascribed to the lack of cirrhotic microenvironment in the rat model.

In the validation cohort, we confirmed miR-30e-3p, miR-30e-5p, and miR-30a-3p downregulation in HCCs with respect to both matched cirrhosis (Tukey post hoc test; P = 0.012, P = 0.013, and P = 0.025, respectively) and normal livers (Tukey post hoc test; P = 0.046, P = 0.001, and P = 0.044, respectively; Fig. 1G–I). Here, we focused on miR-30e-3p, which is the most downregulated miRNA in human and rat HCCs.

Because TP53 is mutated in about 30% of HCCs and it establishes complex regulatory networks with miRNAs (1, 36), we investigated the relationship between TP53 status and miR-30e-3p in HCC. Lower miR-30e-3p levels were observed in TP53-mutated HCCs, whereas no difference was found for primary miR-30e (pri-miR-30e) levels (Fig. 1J and K). Remarkably, a negative correlation between primary and mature miR-30e levels was found in HCCs (Pearson correlation; R = -0.32; Supplementary Fig. S3A), suggesting a role for posttranscriptional mechanisms in miR-30e biogenesis regulation. Interestingly, after splitting HCCs according to TP53 status, a negative correlation was confirmed between primary and mature miRNA levels in TP53-mutated cases, whereas a trend toward a positive correlation was found in TP53 WT cases (Supplementary Fig. S3B and S3C). These findings suggest an impairment of miR-30e-3p maturation by mutant p53 isoforms, which were previously reported to interfere with the functional assembly of Drosha/p68 processing complex (23).

To start dissecting p53 involvement in miR-30e biogenesis in HCC, we modulated its expression in TP53 WT HepG2 cells by using different strategies. First, we investigated the influence of p53 silencing on miR-30e primary transcript and mature isoforms. As showed in Fig. 2A, a decrease of primary and mature miR-30e-3p and miR-30e-5p levels was detected in p53-silenced cells. MiR-34a was used as a positive control because it represents the first identified p53-target miRNA (37), whereas p53 silencing was verified by Western blot and qPCR analyses of target genes (Fig. 2A). Secondly, we investigated the effect of p53 overexpression in HepG2 cells following transfection of a p53-overexpressing vector or treatment with Nutlin-3, a MDM2 inhibitor. In line, an increase of miR-30e primary transcript and mature miRNA isoforms was detected in both experimental settings (Fig. 2B and C). Notably, Nutlin-3 administration led to a stronger p53 transcriptional activation with respect to p53-overexpressing vector, as confirmed by the extent of variation of CDKN1A/p21 and MDM2 mRNAs. In turn, this resulted in higher pri-miR-30e levels, but not mature isoforms (Fig. 2B and C), letting us to speculate that mdm2/p53 axis might have an additional role in miR-30e biogenesis. To investigate p53 involvement in miRNA maturation, we quantified primary and mature miR-30e levels in p53-deleted (CRISPR/Cas9 technology) HepG2 cells and p53-null Hep3B cells following Nutlin-3 treatment. In line with p53 absence, no increase of miR-30e primary transcript was observed, whereas increased mature miRNA isoforms were detected in both treated cell lines (Fig. 2D and E), confirming p53 role in miRNA processing impairment during Nutlin-3 administration. Because Dicer1 promoter contains several p63/p53 responsive elements (38), we investigated Dicer1 expression in HepG2 cells following p53 overexpression or Nutlin-3 treatment in both p53 WT and p53-deleted cells. A downregulation of Dicer1 was observed in Nutlin-3–treated HepG2 cells, but not in p53-overexpressing cells, which is in line with miRNA maturation impairment detected after Nutlin-3 administration in p53 WT cells (Supplementary Fig. S3D and S3E). On the contrary, higher Dicer1 levels were detected in p53-deleted HepG2 and p53-null Hep3B cells (Supplementary Fig. S3F and S3G), suggesting that mdm2 inhibition by Nutlin-3 influences miR-30e processing by regulating Dicer1 transcription in a p53-dependent and -independent manner.

To go deeper into p53-mediated mechanisms regulating miR-30e expression, we overexpressed WT and mutant (truncated, dominant negative) TP53 isoforms in p53-deleted Hep3B and p53 WT HepG2 cells, respectively. As displayed in Fig. 2F, an increase of primary and mature miR-30e isoforms was observed in TP53-overexpressing Hep3B cells confirming its role in miRNA transcription. Interestingly, mutant p53 expression in HepG2 cells decreased miR-30e transcription and, consequently, mature isoforms (Fig. 2G), suggesting a negative role for p53 mutations in miR-30e regulation. These findings demonstrate that TP53 WT isoform is necessary to induce miR-30e transcription in HCC cells and that p53/mdm2 axis is involved in a multistep regulation of miR-30e biogenesis.

To prove p53 binding to hypothetical consensus elements in proximity of miR-30e precursor region, a ChIP experiment was conducted in control and Nutlin-3–treated HepG2 cells. Six amplicons containing TP53 hypothetical binding sites were designed in a DNA region spanning from −4000 to +1000 nucleotides, considering +1 the first nucleotide of miR-30e precursor (Supplementary Fig. S3H). This region was chosen based on described promoter distance for intronic miRNA (39) and due to the presence of histone modifications (Supplementary Fig. S3I). The ChIP analysis showed an enrichment of three out of six p53 hypothetical binding sites, namely BS1, BS2, and BS6, in Nutlin-3–treated cells only (Supplementary Fig. S3J), demonstrating that p53 activation strengthens its binding to miR-30e-3p genomic region favoring pri-miR-30e regulation. EMSA assay further confirmed p53 binding to hypothetical consensus sites identified in the ChIP experiment, showing a specific shift for BS1, BS2, and BS6 probes in presence of NEs from Nutlin-3–treated HepG2 cells (Supplementary Fig. S3K).

Notably, a positive correlation (Pearson correlation; R = 0.71) exists between pri-miR-30e and its host gene, nuclear transcription factor Y subunit gamma (NFYC, NM_014223) in HCCs, whereas a negative correlation (Pearson correlation; R = -0.76) was found between mature miR-30e-3p and NFYC, outlining the relevance of maturation process in the regulation of miR-30e levels (Supplementary Fig. S3L and S3M).

To investigate the interplay between miR-30e-3p and TP53, we interrogated bioinformatics tools and identified the principal p53-negative regulator, MDM2 (NM_002392), as a hypothetic miR-30e-3p target gene containing three complementary binding sites (Fig. 3A). We quantified miR-30e-3p levels in seven HCC cell lines and compared its expression with mean values observed in HCCs, cirrhosis, and normal livers (Fig. 3B). In order to avoid a confounding effect due TP53/MDM2 autoregulatory loop, we performed a functional analysis in HCC cell lines with mutated or null p53 isoform. Huh-7, Hep3B, and SNU475 cells were chosen based on their low/intermediate miR-30e-3p basal levels, whereas SNU449 cells were selected because of their high basal levels (Fig. 3B). MiR-30e-3p overexpression in Huh-7, Hep3B, and SNU475 cells decreased mdm2 mRNA and protein levels, whereas miRNA silencing in SNU449 cells increased its mRNA and protein expression (Fig. 3C–F). Subsequently, we explored MDM2 targeting by miR-30e-3p in HepG2 cells. We observed increased mRNA and protein levels following transient miRNA silencing in p53 WT and silenced (1.3- and 1.5-fold, respectively) HepG2 cells (Supplementary Fig. S4A and S4B). To assess if miR-30e-3p regulation of MDM2 is long lasting, we evaluated its expression in miR-30e-3p stably silenced (MZIP-30e-3p) HepG2 cells and detected increased protein levels (Supplementary Fig. S4C), highlighting a prolonged posttranscriptional regulation. Finally, we verified miRNA/mRNA interaction by performing a reporter assay in HepG2 cells. MiR-30e-3p cotransfection decreased luciferase activity of WT MDM2-3′UTR vectors, but not that of mutant ones, proving miRNA direct interaction with its complementary sequence in MDM2 mRNA (Fig. 3G). To verify the existence of a possible miR-30e/MDM2/TP53 feedback loop, we measured p53 transcriptional activity and expression in WT and mutant TP53-bearing cells following miR-30e-3p overexpression. An increase of luciferase signal and p53 protein expression was registered in miR-30e-3p–transfected HepG2 cells; whereas no change was observed in Huh-7 cells harboring an inactive p53 isoform (Fig. 3H; Supplementary Fig. S4D). These data demonstrated that, in p53 WT contexts, miR-30e-3p increases p53 expression and activity by targeting mdm2, leading to a feedforward loop sustained by p53-dependent transcription of miR-30e-3p itself. Conversely, in TP53-mutated backgrounds this positive loop is hampered, suggesting that other targets might mediate miR-30e-3p functions in these contexts.

The regulation of cell proliferation and invasion by miR-30 family members has been reported in tumors (40, 41); nevertheless, miR-30e-3p, the most downregulated miR-30 member in our patient cohorts and preclinical model, was not previously studied. Besides MDM2, a bioinformatics analysis highlighted p27 and PTEN among miR-30e-3p hypothetic targets (Supplementary Fig. S5A). Due to their key role in hepatocarcinogenesis, it is conceivable that miR-30e-3p establishes multitarget networks driving HCC phenotype. qPCR and Western blot analyses in transfected HCC cells proved p27 and PTEN inhibition by miR-30e-3p at mRNA and protein levels (Supplementary Fig. S5B). A reporter assay showed decreased luciferase activity of p27 and PTEN 3′UTR WT vectors upon miR-30e-3p overexpression, demonstrating that miR-30e-3p directly regulates these targets through mRNA degradation (Supplementary Fig. S5C).

We subsequently explored the effect of miR-30e-3p on HCC cell proliferation by dissecting the contribution of mdm2/p53 axis and PTEN/AKT pathway. In order to investigate the former, we performed the proliferation assay in TP53 WT and p53-silenced HepG2 cells. Cell growth inhibition (1.2-fold) occurred upon miR-30e-3p enforced expression in TP53 WT cells, whereas an increase of cell proliferation (1.4-fold) was detected in TP53-silenced HepG2 cells (Fig. 4A and B), suggesting p53 as a pivotal factor modulating miR-30e-3p effect. The relevance of PTEN targeting on cell proliferation was next assessed in p53-silenced cells by evaluating AKT phosphorylation upon miR-30e-3p enforced expression. An increase of phospho-AKT occurred in both TP53 WT and p53-silenced HepG2 cells highlighting AKT, beside p53, as pivotal factors mediating miR-30e-3p effects. We next assayed p53-mutated Huh-7 and p53-deleted Hep3B cells displaying high constitutive pten levels (20). MiR-30e-3p overexpression increased proliferation of Huh-7 and Hep3B cells (1.9- and 2.1-fold, respectively). Enhanced AKT phosphorylation confirmed PTEN/AKT targeting as a trigger for proliferation in these nonfunctional p53 backgrounds (Fig. 4C and D). To check whether p27 contributes to cell proliferation, we chose p53-mutated SNU475 cells displaying null pten levels (20). MiR-30e-3p increased proliferation of SNU475 cells without changing phospho-AKT levels, but strongly decreasing p27 (Fig. 4E). These findings support the crucial role of mdm2/p53, p27, and AKT in mediating miR-30e-3p regulation of cell proliferation. We hence analyzed cell-cycle regulators p21 and p27 in miR-30e-3p–overexpressing HCC cells. P21 modulation reflected p53 status, showing an increase in TP53 WT HepG2 cells only, whereas it did not change in TP53-mutated cell lines (Fig. 4A–E). Conversely, p27 downregulation was observed in all cell lines, confirming its targeting by miR-30e-3p (Fig. 4A–E). Opposite proliferative changes induced by miR-30e-3p in HCC cell lines can be explained by its multitarget activity. Indeed, the inhibition of mdm2 prevents cell-cycle acceleration through p53/p21 signaling in TP53 WT contexts only. On the contrary, p27 and PTEN inhibition prevails in TP53-mutated/deleted backgrounds, leading to increased proliferation.

We next explored miR-30e-3p ability to modulate invasive properties of HepG2 and SNU449 cells. These cell lines were chosen due to high miR-30e-3p constitutive levels that allowed the investigation of miR-30e-3p stable silencing (Supplementary Fig. S5D) in a p53 WT (HepG2) and p53-mutated context (SNU449). No change was observed in miR-30e-3p–silenced HepG2 cells, whereas decreased invasive and migratory potentials were detected in miRNA-silenced SNU449 cells (Student t test; P < 0.0001; Fig. 4F; Supplementary Fig. S5E). Despite decreased phospho-AKT levels in both cell lines, their different invasion capability highlights cell context as pivotal in driving miR-30e-3p influence on cell aggressiveness.

We investigated miR-30e-3p–driven changes on stemness in HCC and observed a negative correlation between miR-30e-3p and both AFP and EpCAM mRNAs in patients with HCC (Pearson correlation; R = −0.45 and -0.36, respectively; Fig. 5A) and cell lines (Pearson correlation; R = −0.62; Supplementary Fig. S6A). Analyzing HCCs according to TP53 status, an even stronger correlation was observed in TP53-mutated patients (Pearson correlation; R = -0.78; Supplementary Fig. S6B), suggesting a stronger activity of miR-30e-3p in the absence of miR-30e-3p/mdm2/p53 regulatory loop. An opposite behavior was detected for pri-miR-30e that positively correlated with EpCAM mRNA (Pearson correlation; R = 0.39; Supplementary Fig. S6C). Accordingly, a change of EpCAM mRNA and immunophenotype was observed following miR-30e-3p modulation in HCC cells (Fig. 5B; Supplementary Fig. S6D). To investigate how miR-30e-3p influences stem cell properties, we interrogated bioinformatics algorithms and identified two hypothetical miRNA-binding sites in EpCAM 3′UTR (Supplementary Fig. S6E). A decreased luciferase activity was observed for WT EpCAM-3′UTR vector in presence of miR-30e-3p overexpression (Student t test; P = 0.00045), demonstrating a direct regulation of this target in HCC (Fig. 5C). We next performed clonogenic and sphere formation assays in miR-30e-3p stably silenced (MZIP-30e-3p) p53 WT HepG2 and p53-mutated SNU449 cells (Supplementary Figs. S5E and S6F), chosen on the basis of their different stemness features (Supplementary Fig. S6G). The clonogenic assay displayed an increase of colony number (t test; P = 0.0005) and OD 595 (t test; P = 0.001) in MZIP-30e-3p HepG2 cells, whereas an opposite behavior was observed in SNU449 cells displaying a decreased colony number (t test; P < 0.0001) and OD 595 (t test; P = 0.025) in miR-30e-3p–silenced cells (Fig. 5D and E). Finally, we performed colony-forming unit and sphere formation assays in miR-30e-3p stably overexpressing Huh-7 cells exhibiting high stem cell characteristics and a mutated p53 isoform (Supplementary Fig. S6F and S6G). A decrease of colony and sphere number (t test; P = 0.0033 and P = 0.010, respectively) was observed in miR-30e-3p-Huh-7 cells (Fig. 5F and G). Similarly, miR-30e-3p silencing determined an increased sphere formation (Student t test; P < 0.0001) in HepG2 cells (Fig. 5G; Supplementary Fig. S6H). Due to low expression of stem cell characteristics, SNU449 cells displayed no capability to sphere formation (Supplementary Fig. S6I). An increase of AFP and EpCAM levels was observed in MZIP-30e-3p HepG2-derived spheres, whereas their decrease was detected in pMXs-miR-30e-3p Huh-7-derived spheres (Fig. 5H), confirming findings observed in HCC specimens and cell lines.

In primary HCCs, low miR-30e-3p expression associated with microvascular invasion (Student t test; P = 0.012) and increased recurrence rate (log-rank test; P = 0.027; Fig. 5I and J). These findings demonstrated that miR-30e-3p downregulation participates to boost stemness properties in HCC cells characterized by a stem-like phenotype.

Due to the relevance of miR-30e-3p targets in stressful conditions and drug response, we evaluated apoptotic cell death in HepG2 and Huh-7 cells following starvation and doxorubicin or sorafenib treatment. Firstly, we considered TP53 WT HepG2 cells and observed that miR-30e-3p overexpression had no effect in basal conditions (Fig. 6A). On the contrary, it increased early apoptosis (2.3-fold) following serum deprivation and both early and late apoptosis (1.3 and 7.8-fold, respectively) in doxorubicin-treated cells (Fig. 6B and C), which is consistent with p53 activation in these conditions (Supplementary Fig. S7A). Western blot and caspase assay confirmed increased p53 expression, apoptotic markers, and caspase-3/7 activity upon miR-30e-3p enforced expression in serum-deprived and doxorubicin-treated HepG2 cells (Fig. 6C–E).

Similarly, sorafenib increased early apoptosis (3.1-fold increase) in miR-30e-3p–transfected HepG2 cells (Fig. 6F–H). Because both sorafenib (Supplementary Fig. S7A) and miR-30e-3p induce p53 expression (Fig. 6G), we wondered if p53 might account for increased apoptotic cell death in miR-30e-3p–overexpressing cells. To prove a p53-dependent response, we performed the same assays in p53-silenced HepG2 cells. In this setting, miR-30e-3p overexpression decreased (4.2-fold) early apoptotic cells (Fig. 6I–K), demonstrating the central role of p53 in triggering miR-30e-3p–mediated apoptosis. In agreement, miR-30e-3p overexpression in TP53-deleted (CRISPR/Cas9 technology) HepG2 cells determined a resistance to both starvation and sorafenib (Supplementary Fig. S7B–S7E), confirming p53 as a determinant of miR-30e-3p duality in HCC. To demonstrate the relevance of PTEN/AKT axis in the absence of p53, we coupled miR-30e-3p inhibition with PTEN silencing in p53-deleted HepG2 cells. Annexin-V assay after sorafenib administration showed that PTEN inhibition decreases early apoptosis (1.5-fold) in miR-30e-3p–silenced/p53-knockout HepG2 cells confirming PTEN/AKT pathway as a driver of sorafenib resistance in the absence of p53 (Supplementary Fig. S7F and S7G). Finally, because sorafenib triggers p27 in HepG2 cells (Supplementary Fig. S7H), we wondered whether p27 targeting by miR-30e-3p might influence sorafenib response. Strikingly, when both TP53 and PTEN were inhibited, miR-30e-3p overexpression decreased early apoptotic cell population (1.4-fold) in sorafenib-treated HepG2 cells, suggesting p27 participation to miR-30e-3p–mediated sorafenib resistance in p53-depleted cells (Supplementary Fig. S7I). These findings confirm the importance of PTEN and p27 in driving miR-30e-3p–associated phenotype in TP53-mutated backgrounds. In line, miR-30e-3p overexpression in TP53-mutated Huh-7 cells determined a resistance to starvation, doxorubicin, and sorafenib administration (Supplementary Fig. S8A–S8H) and high miR-30e-3p levels associated with increased tumor cell proliferation in sorafenib-treated Huh-7–derived xenograft mice (Supplementary Fig. S8I and S8J), confirming its oncogenic role in p53-mutated cells.

In sorafenib-treated DEN-HCC rats, a downregulation of miR-30e-3p expression was found in 73% of nodules (Student t test; P = 0.02; Fig. 7A). Sorafenib-resistant tumors showed lower miR-30e-3p tissue levels (Student t test; P = 0.004; Fig. 7B), suggesting miR-30e-3p downregulation as an adverse event with respect to sorafenib efficacy in this model. In agreement, an inverse correlation between miR-30e-3p and tumor size (Pearson correlation; R = −0.69; Fig. 7C) and a positive correlation between miR-30e-3p and apoptotic markers (Pearson correlation; R = 0.70, 0.71 and 0.49) were identified in treated rat HCCs (Supplementary Fig. S9A–S9C). Finally, the positive correlation between miR-30e-3p and p21 levels in tumor tissues (Pearson correlation; R = 0.52; Supplementary Fig. S9D) from TP53 WT rat DEN-HCCs (42) led us to speculate that miR-30e-3p/mdm2/p53 axis might take part in sorafenib sensitization also in rat tumors.

Circulating miRNAs were proposed as biomarkers of treatment response (43) and tyrosine kinase inhibitors, such as gefitinib, promote exosomal secretion (44). Thus, we assayed if sorafenib could modulate miR-30e-3p extracellular levels. An exosome-mediated miR-30e-3p extrusion was confirmed in all HCC cell lines treated by sorafenib, except for SNU449 cells. Sorafenib decreased intracellular miR-30e-3p levels in sensitive cell lines (HepG2, Hep3B, and Huh-7), whereas it increased miRNA levels in the resistant (SNU475) one (Fig. 7D–F). Aiming at identifying molecular mechanisms underneath miR-30e-3p extrusion triggered by sorafenib, we focused on p53, a known promoter of exosomal secretion (45). We assayed p53 activity following sorafenib treatment in WT TP53 HepG2 cells, in SNU475 cells that harbor a double TP53 heterozygous mutation but maintain a residual TP53 transcriptional activity and in SNU449 cells with a mutated and transcriptionally inactive TP53 isoform (Fig. 3B). Sorafenib increased TP53 activity in both HepG2 and SNU475 cells, but not in SNU449 cells (Fig. 7G). We next performed TP53 silencing and deletion in HepG2 cells and observed that p53 absence prevented miR-30e-3p extracellular and exosomal rising (Fig. 7H; Supplementary Fig. S9E), confirming that p53 contributes to sorafenib-mediated miR-30e-3p exosomal secretion by HCC cells.

Remarkably, an inverse correlation between tissue and serum miR-30e-3p levels was detected in both DEN-HCC rats and xenograft mice subjected to sorafenib treatment (Pearson correlation; R = −0.58 and R = −0.88, respectively; Supplementary Fig. S9F and S9G). In addition, a negative correlation between miR-30e-3p circulating levels and apoptotic molecules was found in sorafenib-treated HCC rats (Pearson correlation; R = −0.82, −0.65 and −0.51, respectively; Supplementary Fig. S9H–S9J) and higher circulating miR-30e-3p levels were detected in sorafenib-resistant group (Student t test; P = 0.05; Fig. 7I). Taken together, these correlative findings suggested miR-30e-3p as a potential biomarker of treatment response.

We thus tested circulating miR-30e-3p levels in patients with advanced HCC (Supplementary Table S2), assaying sera samples collected before (46 patients, 25 responders, and 21 nonresponders at 2 months) and on sorafenib treatment (49 patients, 23 responders, and 26 nonresponders at the 4 months follow-up assessment). Circulating miR-30e-3p levels did not differ between responders and nonresponders (mean circulating miR-30e-3p: 2.24 in R vs. 1.96 in NR; unpaired t test, P = 0.3, Fig. 7J) ruling out any predictor role of primary response. Conversely, higher circulating miR-30e-3p was found at the 2 months follow-up in patients experiencing escape to treatment at the 4 month's TC (circulating miR-30e-3p at 2 months: 2.13 in responders vs. 3.18 in nonresponders, unpaired t test, P < 0.0001; Fig. 7K), suggesting its early elevation in acquired drug resistance. In a subgroup of 28 patients (13 responders and 15 nonresponders at the 4 months follow-up), variations of circulating miR-30e-3p were tested over time (basal vs. 2 months follow-up). Even though the cohort is small, preventing any conclusion, an increase in circulating miR-30e-3p levels at the 2-month assessment was observed in patients developing escape at the subsequent CT scan (Supplementary Fig. S9K and S9L). Notably, circulating miR-30e-5p levels did not predict sorafenib response neither before nor on treatment (Supplementary Fig. S9M and S9N), further highlighting the specificity of miR-30e-3p variations in acquired resistance to sorafenib. In conclusion, even though these data are very preliminary and circumstantial, the early rising of circulating miR-30e-3p in patients developing sorafenib resistance makes this miRNA appealing for further exploitation, in a field where noninvasive biomarkers are needed.

MiRNAs act as tumor suppressors or oncogenes depending on targets' core, driver gene mutations, and tumor microenvironment. Here, we report the downregulation of miR-30e-3p in human and rat HCCs and its association with poor prognosis, proliferation, stemness phenotype, and drug resistance. The modulation of different biological functions has to be ascribed to the simultaneous targeting of both oncogenes and tumor suppressor genes.

Previous studies reported miR-30 as a “dual” miRNA, with opposite functions in non–small cell lung cancer and pancreatic cancer (15, 18). Here, we describe the “dual” role of miR-30e-3p within the same tumor type, highlighting opposite effects based on different molecular backgrounds and in particular to TP53 status. On the one hand, miR-30e-3p decreased cell proliferation and induced sorafenib sensitization in TP53 WT HepG2 cells, behaving as a tumor suppressor miRNA. On the other hand, it induced proliferation, and drug resistance in TP53-mutated/deleted HCC cells, behaving as an oncomiR. Indeed, both tumor suppressors (CDKN1B/p27 and PTEN) and oncogenes (MDM2, EpCAM) were demonstrated to be miR-30e-3p targets. Specifically, miR-30e-3p exerts tumor suppressor functions in TP53 WT contexts where mdm2 targeting is prevalent, leading to p53 pathway activation. On the contrary, tumor-promoting effects take place in nonfunctional p53 contexts, where p27 and PTEN targeting becomes prevalent. In particular, p27 drives the proliferating phenotype; EpCAM modulates staminal properties, whereas PTEN contributes to drug resistance by triggering AKT activation (46). Similarly, we previously reported a dual behavior for miR-221 with respect to p53 status, highlighting the importance of molecular characterization when antagomiR strategies are considered (21). Indeed, both miR-221 and miR-30e-3p modulate the TP53 axis through mdm2 direct regulation explaining, at least in part, their opposite activity in different TP53 contexts.

As established for miR-30a and miR-30e-5p (24, 25), here we proved that p53 activates miR-30e transcription. Nevertheless, we observed a negative correlation between primary and mature miR-30e-3p levels, envisaging regulatory events on miRNA biogenesis. Strikingly, mature but not primary miRNA profiles clustered normal and tumor samples into separate nodes, suggesting maturation processes as more relevant than altered transcription for aberrant miRNA expression (47). Gain of function of mutated p53 isoforms hampers miRNA processing machinery, leading to miRNA downregulation due to interference with both p68 and p72/82 RNA helicases, impairing Drosha microprocessor activity (23, 48). Moreover, R273H-mutated p53 inhibits miR-30a transcription contributing to its aberrant expression in breast cancer (49). We observed an association between p53 mutations and miR-30e-3p downregulation, as well as the inhibition of miRNA maturation following Nutlin-3 treatment, suggesting the enrollment of other mdm2 targets, such as p53 family members (p63 and p73), as possible regulators of miRNA biogenesis. In this regard, an elegant study by Su and colleagues demonstrated that Dicer1 is a transcriptional target of TAp63 but not of p53 (50). Our descriptive findings on p53 involvement in miRNA biogenesis led us to speculate that p53 accumulation, as in the case of MDM2 permanent inhibition by Nutlin-3, might be responsible for the transcriptional inhibition of Dicer1 through saturation of p63/p53-REs, preventing p63-mediated transcription. Dicer1 disruption in knockout mice promoted hepatocarcinogenesis and led to the deregulation of several miRNAs, including all miR-30 family members, suggesting a strict regulation of miR-30 maturation by Dicer1 in HCC (51).

Rat HCCs, characterized by a TP53 WT background (42), showed an inverse correlation between miR-30e-3p and tumor size, as well as reduced miR-30e-3p tissue levels in sorafenib-resistant tumors, suggesting miR-30e-3p downregulation as a contributor to sorafenib resistance. New molecules, beside sorafenib, were recently registered for advanced HCC (52), however biomarkers predicting treatment outcome are lacking. In the last two decades, miRNAs have emerged as attractive diagnostic, prognostic, and therapeutic tools (53, 54) and reliable markers (43). Decreased miR-30e and miR-223 serum levels were reported in patients with HCC irrespective of tumor etiology (55), whereas miR-30e–enriched vesicles decreased mesenchymal properties in cholangiocarcinoma cells (56).

Our and other groups reported sorafenib as a trigger for TP53 activation and demonstrated TP53 as a key player for exosomal secretion in stressful conditions (45) even though mechanisms governing exosomal secretion remain poorly understood. Here, we confirmed the contribution of the p53 pathway to miR-30e-3p exosomal secretion in sorafenib-treated HCC cell lines. In addition, as reported for the tyrosine kinase inhibitor gefitinib, sorafenib triggers exosome secretion in HCC cells (44). Increased miR-30e-3p circulating levels were found in sorafenib-resistant rat and human HCCs. In particular, higher miR-30e-3p circulating levels, deriving from exosomal secretion, were associated with subsequent escape to sorafenib. This does not prove a causative effect and may simply represent a bystander event; however, miR-30e-3p role as a noninvasive predictor of treatment escape can be envisaged and should be further investigated in larger patient cohorts.

F. Piscaglia reports receiving honoraria from the speakers' bureau of Bayer, EISAI, BMS, IPSEN, Astrazeneca, Roche, Guerbet, Tiziana Life Sciences, and Alkermes. L. Bolondi reports receiving honoraria from the speakers' bureau of Bayer, BMS, Eisai, Alfa Sigma, Guerbet, Sirtex, Eli-Lilly, and Meda-Pharma. No potential conflicts of interest were disclosed by the other authors.

Conception and design: L. Gramantieri, D. Pollutri, F. Fornari

Development of methodology: D. Pollutri, C. Giovannini, S. Marinelli, M. Negrini, F. Fornari

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L. Gramantieri, D. Pollutri, M. Gagliardi, S. Quarta, A. Casadei-Gardini, E. Callegari, S. De Carolis, S. Marinelli, F. Benevento, F. Vasuri, M. Ravaioli, M. Cescon, M. Negrini, F. Fornari

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): L. Gramantieri, D. Pollutri, M. Gagliardi, M. Ferracin, E. Callegari, S. De Carolis, F. Vasuri, F. Fornari

Writing, review, and/or revision of the manuscript: L. Gramantieri, S. Quarta, M. Ferracin, F. Piscaglia, F. Fornari

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D. Pollutri, F. Benevento, M. Ravaioli

Study supervision: L. Gramantieri, L. Bolondi, F. Fornari

We thank Dr. Vilma Mantovani and Dr. Marinella Cenci of Genetics Unit at S.Orsola-Malpighi University Hospital for cell line authentication. We thank Dr. Clarissa Patrizi of Modena and Reggio Emilia University for her precious help with CRISPR technology.

Financial support was provided by the following: (i) Programma di Ricerca Regione-Università, Regione Emilia-Romagna, Bando "Alessandro Liberati" to F. Fornari. "Identification of innovative microRNAs-based biomarkers and anti-cancer strategies for the treatment of Hepatocellular carcinoma.” (ii) Programma di Ricerca Regione-Università, Regione Emilia-Romagna, Bando "Ricerca Innovativa" to L. Gramantieri and L. Bolondi. "Innovative approaches to the diagnosis and pharmacogenetic-based therapies of primary hepatic tumors, peripheral B and T-cell lymphomas and lymphoblastic leukaemias.”

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.