Opposing mechanisms by which miRNAs mediate distinct Nrf1 and Nrf2 regulation of epithelial–mesenchymal transition in hepatocellular carcinoma

ABSTRACT

Accumulation of various genetics and epigenetics alterations are accepted to result in the initiation and progression of hepatocellular carcinoma (HCC), and its high metastasis is viewed as a critical bottleneck leading to its treatment failure. Amongst them, the microRNAs arising from the lack of the antioxidant transcription factor Nrf2 lead to cancer metastasis. However, much less is known about the regulation of microRNAs by Nrf1, even though it acts as an essential determinon of cell homoeostasis by governing the transcriptional expression of those driver genes contributing to the EMT involved in its metastasis. In this study, distinct EMT phenotypes resulted from specific knockouts of Nrf1 and Nrf2 in HepG2 cells, as accompanied by their differential migratory and invasive capabilities. The Nrf1α^–/–-leading EMT results from a significant decrease in the epithelial CDH1 expression, plus another increased expression of the mesenchymal CDH2. Such distinct phenotypes of Nrf1α^–/– from Nrf2^–/– cell lines were also attributable to differential regulation of two key microRNAs, i.e. miR-3187-3p and miR-1247-5p. Further experiments also unravelled that Nrf1 activates the miR-3187-3p expression, directly targeting for the inhibition of SNAI1, leading to CDH1 activation but with CDH2 inhibition insomuch as to prevent the process of EMT. By contrast, Nrf2 inhibits the miR-1247-5p expression, relieving its inhibitory effect on MMP15 and MMP17 to promote the EMT. Collectively, these results demonstrate that the EMT of HCC is likely prevented by Nrf1 via the miR-3187-3p signalling to SNAI1-CDH1/2 axis, but conversely promoted by Nrf2 through the miR-1247-5p-MMP15/17 signalling axis.

1. Introduction

According to the statistic evaluation by the International Agency for Research on Cancer (IARC) of the World Health Organization (WHO), the morbidity of liver cancer is ranked world-widely in the sixth place of all new cancer cases, with its mortality rate being reached to the third [1]. Only in China, the number of newly increased cases with liver cancer has highly reached 410,000, at the fifth ranked in the incidence of all various types of cancer. The number of deaths from liver cancer has reached 390,000; that is, its mortality rate is at the second highest in China. Such refractory liver cancer has heretofore been recognized as one of the main factors endangering human health [1,2] because the detailed mechanisms underlying its malignant progression and metastasis remain elusive to date.

As one key issue in liver cancer treatment, the high invasiveness and metastasis of tumour cells is widely accepted to be attributable to the increasing epithelial–mesenchymal transition (EMT) in the process, albeit the EMT has been, for many years, described mainly in the early embryonic development stage [3]. Specially, the occurrence and progression of liver cancer are modulated predominantly by EMT-related transcription factors, including canonical SNAI11, SNAI2, ZEB1/2, Twist, and other non-canonical factors such as WNT1/β-catenin, Prrx1, c-Myc and TTF1 [4–8]. Besides, the EMT process is also activated by a variety of signalling pathways such as TGF-β, MAPK, PI3K, Akt, and PTEN, controlling the expression of those downstream genes responsible for tumour metastasis, but depending on the inhibitory effect of the above transcription factors within different biological backgrounds [9]. The regulatory pattern of those non-targeted transcription factors has been gradually reported by regulating EMT markers CDH1, CDH2 and VIM, as well as other non-labelled proteins [10–12]. During EMT, inhibition of the E-cadherin (CDH1, its expression resulting in typically polygonal, cobblestone-like epithelial shapes) enables it to acquire another spindle mesenchymal morphology, along with the expression of mesenchymal-related markers, e.g. N-cadherin (CDH2), vimentin and fibronectin [13]. In the meantime, cell junctions (tight, adhesion, gap junctions and desmosomes) were gradually disintegrated, such that cell polarity was lost, actin expression was increased, pseudopodia were formed, the cell adhesion was promoted, and the expression of both CDH2 and integrin was enhanced, but the epithelial cell adhesion was reduced. This is accompanied by differential expression patterns of matrix metallopeptideases (MMPs, which is a family of both zinc- and calcium-dependent proteolytic enzymes to degrade almost all components of the extracellular matrix (ECM)) [14].

Intriguingly, the abnormal expression of microRNAs (miRNAs) in tumour cells has clearly been shown to promote the occurrence, development, migration and invasion of cancer cells in many ways compared with normal tissues [4,10,15,16]. That is, the abnormal miRNA expression is accompanied by the development of carcinogenic transformation, i.e. EMT, with the high metastasis and invasion ability [17]. Of note, the EMT of tumour cells were reportedly regulated by a big class of miRNAs, including miR29b, miR30a, miR30c, miR137, miR34a, and miR15, through transcription factor SNAI1 [18–23], exerting distinct biological effects of tumour suppressors or promoters in different types of cancers [24]. Besides, the EMT of tumour cells is clearly regulated by nuclear factor erythroid 2-related factor 2 (Nrf2, encoded by Nfe2l2) through distinct miRNA pathways. However, much less is known about how Nrf2, together with nuclear factor erythroid 2-related factor 1 (Nrf1, encoded by Nfe2l1), regulates those critical miRNAs for differentially controlling the development of EMT in liver cancer cells.

In mammalians, Nrf1 and Nrf2 are two principal members of the cap’n’collar (CNC) basic region-leucine zipper (bZIP) transcription factor family that governs cellular (redox, energy and metabolism) homoeostasis and organ integrity [25]. Indeed, prior studies have clearly shown that Nrf1 and Nrf2 interact with each other in the regulation of redox, glucose, and lipid (cholesterol) metabolisms and malignant proliferation of hepatocellular carcinoma [26]. However, it has been only reported that Nrf2 activation can promote the occurrence and development of EMT, and the persistent activation of Nrf2-mediated miRNA regulation of EMT in cancer cells has been also widely reported [27]. In the present study, we discovered that specific knockout of Nrf1 and Nrf2 in HepG2 cells results in distinct EMT phenotypes, as accompanied by differential migratory and invasive capabilities. Such distinct phenotypes of Nrf1α^–/– from Nrf2^–/– cell lines were also attributable to differential or even opposing regulation of two key miRNAs, i.e. miR-3187-3p and miR-1247-5p. Further experimental evidence has also been provided revealing that Nrf1α activates the expression of miR-3187-3p, directly targeting for inhibition of SNAI1, thereby leading to CDH1 activation, but CDH2 inhibition, insomuch as to prevent the EMT process. By contrast, Nrf2 inhibits miR-1247-5p, relieving its inhibitory effects on MMP15 and MMP17 to promote EMT. Collectively, these findings demonstrate that the EMT of liver cancer cells is much likely prevented by Nrf1α (as a major full-length isoform of Nrf1) via miR-3187-3p signalling to the SNAI1-CDH1/2 axis but rather promoted by Nrf2 via the miR-1247-5p signalling to the MMP15/17 axis.

2. Materials and methods

2.1. Cell lines, culture, and transfection

The Nrf1α^–/– cells used in this study were constructed by TALENs-mediated genome editing of HepG2 cells, while Nrf2^–/– cells were constructed by CRISPR/Cas9-editing system through HepG2 cells [28]. These two cell lines were saved in DMEM media containing 5 mM glutamine, 10% (v/v) of foetal bovine serum (FBS) and 100 units/mL of penicillin and streptomycin at 37°C in a 5% CO₂ incubator. The experimental cells were transfected with indicated plasmids-contained Lipofectamine® 3000 reagent and then cultured for 8 h in Opti-MEM (Gibco, Walsam, MA, USA). These cells were allowed for 24-h recovery from transfection in a fresh complete medium before the following experiments were performed.

2.2. Expression constructs for Nrf1, Nrf2 and indicated reporters

Two expression constructs for human Nrf1 and Nrf2 were made by inserting their full-length cDNA sequences, respectively, into the KpnI/XbaI site of pcDNA3.1/V5His B. The seed sequences of miR-3187-3p and miR-1247-5p were, respectively, constructed into the AgeI/EcoRI site of PLKO.1-TRC cloning vector. Those oligos for indicated 3×ARE sites of miR-3187-3p (i.e. 1# to 5#) and miR-1247-5p (i.e. 1# to7#) were synthesized and ligated into the KnpI/XhoI site of PGL3-Promoter vector. The resulting ARE-Luc reporters were hence created by inserting the consensus ARE-adjoining sequences from indicated gene promoters. Lastly, the 3’-UTR sequences of SNAI1, MMP15, MMP17, and MMP25 were also subcloned into the XhoI/NotI site of psiCHECK-2 plasmid, respectively. All oligo sequences were listed in Table S1.

2.3. Real-time qPCR analysis of mRNA expression levels

Total RNAs were extracted from experimental cells by using an RNA extraction kit (TIANGEN, Beijing, China), and then approximately 2.0 ~ 2.5 μg of total RNAs were added in a reverse-transcriptase reaction to generate the first strand of cDNA (by using the Revert Aid First Strand Synthesis Kit, Thermo, Waltham, MA, USA). The synthesized cDNA served as the template for quantitative PCR (qPCR) in the GoTaq®qPCR Master Mix (Promega, Madison, WI, USA). Subsequently, the mRNA expression levels were measured by RT-qPCR with indicated pairs of primers (as listed in Table S2). The mRNA expression level of β-actin served as an optimal internal standard control and all relative mRNA expression abundances of other genes were hence presented as fold changes. Of note, the expression levels of miRNA need to involve specific reverse transcription primers to ensure its specificity (as shown in Table S2).

2.4. Western blotting analysis of protein expression abundances

Experimental cells were harvested in a denatured lysis buffer (0.5% SDS, 0.04 mol/L DTT, pH 7.5, containing 1 tablet of complete protease inhibitor EASYpacks in 10 ml of this buffer). The total lysates were further denatured by boiling at 100°C for 10 ~ 15 min, sonicated sufficiently, and diluted with 3 × loading buffer (187.5 mmol/L Tris-HCl, pH 6.8, 6% SDS, 30% Glycerol, 150 mmol/L DTT, 0.3% Bromphenol Blue), before being re-boiled at 100°C for 5 min. Thereafter, equal amounts of protein extracts were subjected to separation by SDS-PAGE, and then transferred to polyvinylidene fluoride (PVDF) membranes (Millipore, Billerica, MA, USA) before being visualized by Western blotting with distinct antibodies (as listed in Table S3). β-actin served as an internal control to verify equal amounts of proteins loaded in each of the electrophoretic wells.

2.5. Luciferase reporter assay

Equal numbers (1.0 × 10⁵) of HepG2 cells were allowed for growth in each well of 12-well plates. After reaching 75% ~85% confluence, the cells were co-transfected for 8 h with an indicated luciferase plasmid alone or together with one of the indicated expression constructs mixed with the Lipofectamine®3000 agent in Opti-MEM (Gibco, Waltham, MA, USA), in which the pRL-TK reporter served as an internal control for transfection efficiency. After being recovered for 24 h in a fresh complete medium, the cells were lysed and then subjected to the dual-reporter assay (Promega, Madison, WI, USA). The ARE-driven luciferase reporter activity was calculated by normalization to the internal Renilla activity, whilst specific target gene-mediated luciferase reporter activity was evaluated by further normalization to background activity measured from an empty expression vector being transfected with ARE-driven luciferase plus pRL-TK reporters. The resulting data are graphically shown as Mean ± S.D. of at least three independent experiments performed in triplicates.

2.6. The transwell-based migration and invasion assays

The transwell-based cell migration and invasion were assayed in the modified Boyden chambers (Transwell, Corning Inc. Lowell, MA, USA). Equal numbers of cells were allowed for growth in each well of 12-well plates. After reaching 70–80% confluence, they were starved for 12 h in a serum-free medium. The experimental cells (1 × 10⁵) were suspended in a 0.2-ml medium containing free FBS and seeded in the upper chamber of each transwell. The cell-seeded transwells were placed in each well of 24-well plates containing 1 ml of complete medium (i.e. the lower chamber) and cultured for 24 h in the incubator at 37°C with 5% CO₂. The remaining cells in the upper chamber were removed whilst the cells attached to the lower surface of the transwell membranes were fixed with 4% paraformaldehyde (AR10669, BOSTER) and stained with 1% crystal violet reagent (Sigma) before being counted.

2.7. Cell scratch assay

The cells were digested with trypsin and counted after suspension so that 3 × 10⁵ cells were inoculated in per well of 6-well plate (at the bottom of which 4–6 horizontally dividing lines per well were already sculptured with a maker pen). When the cells reached more than 95% confluence, additional 3–5 lines were drawn along the vertical direction of the pre-maker line in the hole. The cells were washed with PBS and replaced with 2 mL serum-free DMEM. The initial images of the scratch were acquired under an inverted microscope, and the cells continued to be cultured for indicated lengths of time at 5% CO₂ and 37°C. Thereafter, the scratched images were recovered to distinct extents, all of which were, in the meantime, collected by microscopy at 24, 48, 72, and 96 h in serum-free DMEM media.

2.8. Lentiviral packaging for overexpression of indicated miRnas

The indicated miRNA genes were, respectively, subcloned into the Plko.1 TRC vector and subjected to sequencing verification. The miRNA-expressing Plko.1 TRC, psPAX2 and pMD2G plasmids were combined in a 4:3:2 ratio (total mass of 24 μg DNA) and introduced into 293T cells via transfection. Following the digestion of such HEK 293T cells, 7 × 10⁵ cells were inoculated into a 10-cm Petri dish, subsequently with 8 mL of a high glucose medium added. The cells were cultured overnight until the bottom of the dish was covered by 70% of the cell confluence. At this point, the medium was replaced with 4 mL of another serum-reduced medium (Opti-MEM). Following a 30-minute incubation period with the transfection reagent, the plasmid transfection was performed. Following a transfection period of 8 h, the transfection medium was replaced with another standard growth medium to allow cells for 48-h recovery, and thereafter, the medium was supplemented in accordance with the rate of cell growth. The supernatants were collected at 48, 72 and 96 h post-transfection and stored at 4°C. The concentrated lentivirus was diluted appropriately and 50 μL of the virus suspension added to each well of the experimental cells. Following a 24-h infection period, the lentiviral resistant cells were selected with 25 mg/ml of puromycin. Subsequently, the monoclonal miRNA-expressing cells were further selected, and analysed by Western blotting and RT-qPCR. The experimental cell lines utilized in this study are listed in Table S4.

2.9. Clone formation assay

About 500, 1000 or 1500 experimental cells were seeded in each well of six-well plates and then cultured in 5% CO₂ at 37°C for 2 weeks. During this period, 1 mL of fresh complete medium containing 10% serum was added every 3 days. Finally, the cells were fixed, stained, and photographed under an inverted microscope before being counted the number of cell clones.

2.10. Flow cytometry analysis of cell cycle and apoptosis

After experimental cells (3 × 10⁵) were allowed for growth in each well of 6-well plates, they were then pelleted by centrifuging at 1000×g for 5 min and washed with PBS for three times, before being incubated for 15 min with 5 μL of Annexin V-FITC and 10 μL of propidium iodide (PI) in 195 μL of the binding buffer prior to flow cytometry analysis. The resulting data were analysed by the FlowJo 7.6.1 software (FlowJo, Ashland, OR, USA) and shown graphically. For cell cycle analysis by flow cytometry, the cells growing in the logarithmic phase were digested, centrifuged at 1000 rpm for 5 min, washed 2–3 times with PBS after the supernatants were removed. The cell pellets were resuspended in 300 uL pre-cooled PBS, followed by slow addition of 700 uL anhydrous ethanol while slightly oscillating and mixing, before being stored at 4°C overnight. On the next day, the cells were centrifuged at 1000 rpm for 5 min at 4°C. The cell pellets were re-suspended in 100 uL of binding Buffer and incubated for 15 min with 5 μl of PI-Staining Solution and 5 μL Annexin V-FITC (in the dark) at room temperature before being subjected to flow cytometry analysis of the amounts of DNA for different periods of experimental cells.

2.11. Statistical analysis

Statistical significance of changes in the reporter gene activity and/or other gene expression in indicated cell lines (Table S4) was determined using either the Student’s t-test or Multiple Analysis of Variations (MANOVA). The resulting data are shown as a fold change (Mean ± S.D.) relative to control values, which represents at least three independent experiments that were each performed in triplicates (n = 3 X 3).

3. Results

3.1. Distinctive effects of Nrf1- and Nrf2-deficiencies dictating their cellular phenotypes related to EMT

To elucidate distinct impacts of Nrf1 and Nrf2 on the EMT phenotype, herein we have observed that cellular morphological plasticity exhibited characteristics of EMT, especially upon knockout of Nrf1α from HepG2 cells, as evidenced by altered morphogenesis from oval cell contours to spindle-shaped fusiform (Figure 1(A), a1, middle). This phenomenon closely resembled that yielded by the loss of cell–cell adhesion during the EMT transformation, leading to detachment from the basement membrane and subsequent long-distance metastasis and invasion [2,29]. By sharp contrast, no significant differences in the cell morphology were observed following knockout of Nrf2, when compared with that of wild-type cells (Figure 1(A), a1, right vs left panels).

Figure 1.

Distinct changes in the cellular morphology and relevant gene expression caused by loss of Nrf1α^–/– or Nrf2^–/–.

A. Distinctions in the EMT-relevant morphological plasticity of between WT, Nrf1α^–/–, and Nrf2^–/– cell lines. These were evaluated by their distinctive representative images obtained from routine microscopy (a1), the transwell assays for their basemembrane permeability (a2), and another scratch assay for their migratory capacity (a3), respectively.

B. The permeability of indicated cells throughout transwell basemembranes was subject to quantitative analysis. The results are shown graphically as Mean ± S.D. (n = 3) with significant increases ($$, p < 0.01) and significant decreases (*, p < 0.05; **, p < 0.01). All these data were determined from at least three independent experiments each performed in triplicates.

C. The migratory capacity evaluated by indicated cell scratch assay was subjected to quantitative analysis. The resulting data are shown graphically as Mean ± S.D., with significant increases ($$, p < 0.01) and significant decreases (*, p < 0.05; **, p < 0.01). All these were determined from at least three independent experiments each performed in triplicates (n = 3 × 3).

D. Western blotting analysis of indicated protein expression abundances in WT, Nrf1α^–/–, and Nrf2^–/– cell lines with distinct antibodies. The intensity of all immunoblots was quantified and normalized by that of β-actin, before being shown as fold changes (on the bottom of indicated blots); the data are representative of three independent experiments (n = 3) with significant increases ($, p < 0.05; $$, p < 0.01) and significant decreases (*, p < 0.05; **, p < 0.01) compared to their corresponding control values.

E. Real-time qPCR analysis of indicated gene expression at their mRNA levels in WT, Nrf1α^–/–, and Nrf2^–/– cell lines. The resulting data are shown as fold changes (Mean ± S.D.), with significant increases ($$, p < 0.01), which were statistically determined from at least three independent experiments performed each in triplicates (n = 3 × 3).

F. A schematic representation of the hierarchical interaction networks between the EMT-related transcription factors and target genes, which were drawn by using the Cytoscape software.

Next, the membrane permeability of examined cells was investigated by using transwell experiments. The resulting observations revealed that knockout of Nrf1α led to a notable increase in the cell membrane permeability (Figure 1(A), a2, middle, and Figure 1(B)), whereas no discernible differences were observed by comparison of Nrf2^–/– cells with wild-type (Figure 1(A), a2, right vs left panels). Subsequently, the migration capacity of examined cell types was also evaluated by a cell scratch assay. The experimental observations demonstrated that the migration distance of Nrf1α^–/– cells was significantly greater than that of Nrf2^–/– cells, and even much greater than that of the control cells, remarkably occurring at 48 h (Figure 1(A), a3). Of note, the migration of Nrf1α^–/– cells was incrementing as the time increases, and hence Nrf1α^–/– cells exhibited a greater capacity than those of both WT and Nrf2^–/– cell lines migrated at varying distances (Figure 1(C)).

Subsequently, both protein abundance and mRNA expression levels of those EMT-related markers were examined by Western blotting and RT-qPCR, respectively. The results unravelled that a significant reduction of CDH1, a classic marker associated with the epithelial characteristics, was determined in Nrf1α^–/– cells (Figure 1(D,F)). Conversely, varying extents of increases in all other examined protein and mRNA expression of FN1, ITGB4, CDH2, VIM, SNAI1, and SNAI2 (all of which are linked to mesenchymal genes) were exhibited in Nrf1α^–/– cells (Figure 1(D,E)). Notably, the average expression abundance of FN1 at protein and mRNA levels reached to certain levels of nearly 60 times higher than that observed in WT cells. By contrast, knockout of Nrf2^–/– resulted in only marginal or even no changes in the above-examined proteins and their mRNA expression levels (Figure 1(D,E)). However, the expression of MMP9 was almost unaffected by specific loss of either Nrf1α^–/– or Nrf2^–/–, albeit down-regulated expression of MMP17 by each loss of Nrf1α^–/– or Nrf2^–/– (Figure 1(D,E)).

Collectively, these results demonstrate distinct effects of Nrf1α^–/– from Nrf2^–/– on its deficient cell morphogenesis that are contributable to the EMT-relevant phenotypes. Moreover, according to the regulatory relationship of the above-examined EMT molecules, including EMT markers and relevant transcription factors, along with their effects on targeting gene expression profiling, particularly arising by the loss of Nrf1α^–/– or Nrf2^–/–, such an interaction diagram between the EMT-related transcription factors and their target genes was presented herein (Figure 1(F)).

3.2. Specific loss of Nrf1α^–/– or Nrf2^–/– leads to differential miRNA expression profiling critically required for EMT

To determine whether such distinct phenotypes of between both Nrf1α^–/– and Nrf2^–/– cell lines are dictated or modulated by differential expression profiling of those putative miRNAs, we here performed small RNA sequencing of indicated cell lines, each with two replicates. As clearly shown in Figure S1A, those samples were controlled with a rather high quality, aside from almost negligible small differences. The resulting data were subject to bioinformatics prediction by miRanda and TargetScan, revealing the target genes of 2154 miRNAs regulated in Nrf1α^–/– and Nrf2^–/– cell lines, as a result with a high reliability of such target prediction being determined by the Venn statistics (Figure S1B). In addition, some target gene intersections between miRanda and TargetScan exist (as illustrated in Figure S1C).

To decipher the certain correlation of miRNAs with significant differences between distinct examined groups, we performed miRNA association analysis of their significant differences. The results revealed that only 16 of significantly differentially expressed 317 miRNAs were overlapped in the three groups (Figure S1D). Further bioinformatics analysis of overlapped genes unravelled those putative Nrf1/2-regulated ARE sites existing within the promoter regions of the aforementioned 16 miRNAs (as enlisted in Figure S2). In order to gain further insights into the functional distribution characteristics of differentially expressed miRNAs, we conducted a GO enrichment analysis of differential miRNAs-targeting genes in Nrf1α^–/– or Nrf2^–/– cell lines compared with WT cells. The results showed that differential miRNAs exhibited significant gene differences in their biological processes, cellular components and molecular functions, such as biological adhesion and cell colonization (Figure S3). Concurrently, another pathway enrichment analysis of those miRNAs-target genes was also conducted using the KEGG database. Such KEGG enrichment analysis of Nrf1α^–/– vs WT cell lines revealed significant differences in key target genes involved in cell growth and death, transcription and translation, signal transduction, and also tumour metabolism (Figure S4A). Similarly, another pathway enrichment of Nrf1α^–/– vs WT cell lines indicated significant discrepancies in signalling pathways including intercellular receptors, liver cancer, and endoplasmic reticulum (ER) protein processing (Figure S4B) between the two cell lines. These indicators aforementioned are evidently associated with the EMT process of liver cancer cells. Further comparison of the KEGG analysis of Nrf2^–/– vs. WT cell lines revealed significant discrepancies in amino acid metabolism, lipid synthesis and signal transduction (Figure S4C), while pathway enrichment analysis also unveiled that their primary discrepancies of their target genes were observed in those pathways associated with breast cancer, human cytomegalovirus, and bile secretion (Figure S4D).

Upon further comparison between Nrf1α^–/– and Nrf2^–/– cell lines, 232 of significant differentially expressed miRNAs were determined, of which 188 were upregulated whilst additional 44 were downregulated (Figure 2(A), left two columns). However, the majority of miRNAs exhibited more evident downregulation by the loss of Nrf1α^–/– versus WT controls (Figure 2(A), middle two columns) but the loss of Nrf2^–/– caused significant upregulation of those major miRNAs (Figure 2(A), right two columns). Such marked differences in distinct miRNA expression were also further presented by two volcano plots created from Nrf1α^–/– cells (Figure 2(B)) and Nrf2^–/– cells (Figure 2(C)), when compared with WT cells

Figure 2.

Selection of key miRNAs essentially responsible for the EMT.

A. Differentially up- or down-expressed miRNAs were selected by small RNA sequencing of three examined cell lines.

B,C. Two volcano plots of significant differential miRNA expression genes were created by comparison of Nrf1α^–/– cells (B) or Nrf2^–/– cells (C) with WT cells. Some miRNAs were marked out.

D. A heatmap was made up of 16 candidate miRNAs selected from Nrf1α^–/– and Nrf2^–/– cell lines.

E. Two interactive miRNA networks regulated, respectively, by Nrf1 or Nrf2, along with their cognate targets, were drawn by Cytoscape. The inner ring of the colour line in the figure represents the most conventional EMT marker gene and related major transcription factors and the lines of the outer ring represent EMT-related genes.

F. Relative expression levels of 16 candidate miRNAs were determined by their transcriptome sequencing. The data are shown graphically as fold changes (Mean ± S.D.) with significant increases ($, p < 0.05; $$, p < 0.01) and significant decreases (*, p < 0.05; **, p < 0.01), all of which were determined from three independent experiments each performed in mixed triplicates (n = 3 × 3).

G. Relative expression abundances of the above-mentioned 16 candidate miRNAs were validated by real-time qPCR. The resulting data are shown graphically as fold changes (Mean ± S.D., n = 3 × 3) with significant increases ($, p < 0.05; $$, p < 0.01) and significant decreases (*, p < 0.05; **, p < 0.01), all of which were determined from three independent experiments each performed in triplicates.

H. Left panel shows real-time qPCR analysis of relative miRNA expression levels of experimental WT cells that had been transfected with a mimic miRNA-3187-3p (i.e. miR-3187-3p) or another negative control (NC). Right panel shows relative expression levels of putative miR-3187-3p-targeting downstream genes involved in EMT. The resulting data are shown graphically as fold changes (Mean ± S.D.) with significant increases ($$, p < 0.01) and significant decreases (*, p < 0.05; **, p < 0.01), which were determined from three independent experiments each performed in triplicates (n = 3 × 3).

I. Similarly, real-time qPCR analysis of miR-1247-5p mimic (in parallel with negative control, i.e. NC) and its downstream target gene expression was also carried out as described above.

By predicting target genes based on EMT-related markers, eight candidate miRNAs screened from Nrf1α^–/– cells were miR3184-3p, miR199a-5p, miR-3187-3p, miR-497-5p, miR-629-3p, miR-1260b, miR-139-3p, miR-1286. Besides miR-3184-3p, additional eight miRNAs selected from Nrf2^–/– cells were miR-675-5p, miR-3165, miR-3065-3p, miR-532-3p, miR-346, miR-1226-3p, miR-1247-5p, and miR-210-5p. Abundances of such selected miRNA-expressing genes in either Nrf1α^–/– or Nrf2^–/– versus WT cell lines were shown in a heatmap (Figure 2(D)). Two interaction networking diagrams of their regulatory relationships and targeting effects were here inferable to be mediated by Nrf1 and/or Nrf2 (Figure 2(E)). Distinct expression levels of the above-selected 16 candidate miRNAs were determined by transcriptome sequencing (Figure 2(F)), and also further validated by real-time qPCR (Figure 2G). On this base in combination of those putative Nrf1/2-binding ARE motif analysis of those miRNA promoter regions (Figure S2) with a considerable number of their relevant literature [30–35]. Amongst them, miR-3187-3p and miR-1247-5p were herein selected as two key molecules for further study of their specific biological functions regulated by Nrf1 and/or Nrf2.

To verify the biological effects of miR-3187-3p and miR-1247-5p, WT cells were transfected with each of their mimics alongside with a negative control (NC) and then subjected to real-time qPCR analysis of their miRNA expression levels, as well as the mRNA expression abundances of these miRNA-binding downstream target genes (as indicated in Figure S5), so as to explore their inhibitory effects. As shown in Figure 2(H) (left panel), the miR-3187-3p expression was significantly upregulated by its mimic to nearly 60 folds than its corresponding control. By contrast, the expression level of miR-1247-5p was substantially increased by its mimic so as to reach nearly 6000 folds compared to the NC level (Figure 2(I), left panel). Interestingly, the miR-3187-3p mimic also significantly inhibited MMP19, FN1, SNAI1, TWIST1 and TP53 (all proved to be involved in the putative EMT) (Figure 2(H), right panel). Besides, MMP15 and MMP17 were also substantially inhibited by the miR-1247-5p mimic, but with almost no effects on MMP25 (Figure 2(I), right panel). From these results, it is inferable that SNAI1, MMP15, MMP17 and MMP25 are much likely to act as key genes involved in the EMT process, and thus they were further studied as EMT-related markers in the following experiments. In addition, this notion is also supported by the survival curve of these EMT-marker proteins in clinical data (Figure S6, obtained from http://kmplot.com/analysis/index website).

3.3. Differentially targeting of miR-3187-3p and miR-1247-5p to SNAI1, MMP15, and MMP17 involved in EMT.

To further verify direct roles of miR-3187-3p and miR-1247-5p in regulating the mRNA expression levels of SNAI1, MMP15, MMP17 and MMP25, their specific site-directed mutants on the 3’-UTRs (Figure 3(A)) were constructed and subsequently co-transfected with each mimic of miR-3187-3p or miR-1247-5p into HepG2 cells. The results showed that such miRNA mimic-leading decreases in their cognate target 3’-UTRs-monitored luciferase activity were almost completely blocked by co-transfecting with each corresponding mutants (Figure 3(B-D)). However, no significant changes in the MMP25-regulated reporter expression were observed, no matter which reporter constructs driven by its wild type or mutant 3’-UTRs had been co-transfected with miR-1247-5p (Figure 3(E)).

Figure 3.

Differential targets of miR-3187-3p and miR-1247-5p to SNAI1, and MMP15, MMP17, with distinct effects.

A. Schematic representation of indicated targets of miR-3187-3p and miR-1247-5p, together with their mutants, with distinct effects measured by double luciferase reporter constructs used for the following experiments.

B to E. HepG2 cells were co-transfected with miR-3187-3p, miR-1247-5p or their negative controls (NC), along with each of double luciferase reporter constructs entailed by 3′-UTRs of their indicated targets or mutants deciphered above. The 3′-UTRs-modulated luciferase reporter activity regulated by either miR-3187-3p or miR-1247-5p was measured and then subjected to calculation of relative results shown graphically as fold changes (Mean ± S.D.) with significant decreases (*, p < 0.05; **, p < 0.01), which were determined from three independent experiments each performed in triplicates (n = 3 × 3).

F to I. Both mRNA and protein expression levels of indicated target genes regulated by either miR-3187-3p or miR-1247-5p mimics or inhibitors were determined by real-time qPCR (left panels) and Western blotting with indicated primary antibodies (right panels). The data of real-time qPCR are also shown graphically as fold changes (Mean ± S.D.) with significant increases ($, p < 0.05; $$, p < 0.01) and significant decreases (*, p < 0.05; **, p < 0.01), which were determined from three independent experiments each performed in triplicates (n = 3 × 3). The intensity of all indicated immunoblots was quantified and normalized by that of β-actin, before being shown as fold changes (on the bottom of indicated blots). The data are representative of three independent experiments (n = 3) with significant increases ($, p < 0.05; $$, p < 0.01) and significant decreases (*, p < 0.05; **, p < 0.01), relative to their corresponding control values.

Next, we further evaluate putative biological effects of miR-3187-3p’s and miR-1247-5p’s mimics or inhibitors on their downstream EMT-relevant genes by both real-time qPCR and Western blotting, after they had been transfected into two distinct experimental cell lines (HepG2 and Hep3B), respectively. The results showed that the expression of SNAI1 and another mesenchymal marker CDH2 was significantly inhibited by miR-3187-3p mimics, but as accompanied by activated expression of the epithelial marker CDH1 at both its mRNA and protein levels (Figure 3(F), f1 to f4). By sharp contrast, transfection of miR-3187-3p inhibitor (150 nM of its nucleotides) caused a notable increase in the expression of SNAI1 alongside with certain promotion in the CDH2 expression but was also accompanied by significantly inhibited expression of CDH1, the extents of which were inversely correlated with those of elevated SNAI1 (Figure 3(G), g1 to g4). Furthermore, the miR-1247-5p mimics demonstrated a notable inhibitory effect on the expression of MMP15 and MMP17 (Figure 3(H), h1 to h3), but another increased effect of MMP15 and MMP17 was also caused by miR-1247-5p inhibitor (Figure 3(I), i1 to i3). In addition, the MMP25 expression levels were almost unaffected by miR-1247-5p mimic or its inhibitor (Figure 3(H,I)). Of note, all the aforementioned data obtained from such transfected HepG2 cell lines were also allowed to be experimentally repeatable in similarly treated Hep3B cells (as shown in Figure S7).

3.4. Differential regulation of targeted EMT genes by distinctive status of miR-3187-3p or miR-1247-5p.

To verify the upstream and downstream effects of SNAI1, we synthesized its target-specific siRNA to knockdown SNAI1, along with its downstream CDH1 and CDH2, as shown by both real-time qPCR and Western blotting (Figure 4(A)). Further examination revealed more significant downregulation of SNAI1 by a miR-3187-3p mimic occurring after co-transfection with siSNAI1 in HepG2 cells, also as accompanied by directly affected expression patterns of its downstream CDH1 and CDH2 (Figure 4(B)). Conversely, enhanced expression of SNAI1 by the miR-3187-3p inhibitor was prevented by siSNAI1 (Figure 4(C)), which also affected the according expression levels of CDH1 and CDH2. Altogether, these further corroborated the regulatory relationship between miR-3187-3p and SNAI1, and their downstream CDH1 and CDH2.

Figure 4.

Rescue experiments to verify an upstream or downstream relationship between miRnas and their targets.

A. HepG2 cells that had been transfected with SNAI1-specific siRNA (i.e. siSNAI1) or negative control (i.e. NC) were subjected to real-time qPCR and Western blotting assays to determine both mRNA and protein expression levels of SNAI1 and its downstream CDH1 and CDH2. The qPCR data are shown graphically as fold changes (Mean ± S.D.) with significant increases ($$, p < 0.01) and significant decreases (**, p < 0.01), which were determined from three independent experiments each performed in triplicates (n = 3 × 3). The intensity of all indicated immunoblots was quantified and normalized by that of β-actin, before being shown as fold changes (on the bottom of indicated blots). The data are representative of three independent experiments (n = 3) with significant increases ($, p < 0.05; $$, p < 0.01) and significant decreases (*, p < 0.05; **, p < 0.01), relative to their corresponding control values.

B,C. Two distinct rescue experiments were conducted by co-transfection of siSNAI1 with a miR-3187-3p mimic (B) or its inhibitor (C), before the expression abundances of SNAI1 and its downstream CDH1 and CDH2 were determined by immunoblotting with distinct antibodies. The intensity of all those indicated immunoblots was quantified and normalized by that of β-actin, before being shown as fold changes (on the bottom of indicated blots), which are representative of three independent experiments (n = 3) with significant increases ($, p < 0.05; $$, p < 0.01) and significant decreases (*, p < 0.05; **, p < 0.01), relative to their control values.

D. The mRNA and protein expression levels of MMP15 or MMP17 were determined by real-time qPCR Western blotting in HepG2 cells that had been interfered by siMMP15 or siMMP17, respectively. The qPCR data are shown graphically as fold changes (Mean ± S.D.) with significant decreases (**, p < 0.01), which were determined from three independent experiments each performed in triplicates (n = 3 × 3). The intensity of indicated immunoblots was quantified and normalized by that of β-actin, before being shown as fold changes (on the bottom of indicated blots), which are representative of three independent experiments (n = 3) with significant increases ($, p < 0.05; $$, p < 0.01) and significant decreases (*, p < 0.05; **, p < 0.01), relative to their control values.

E to H. Distinct rescue experiments were carried out by co-transfection of siMMP15 or siMMP17 with a miR-1247-5p mimic (E,F) or its inhibitor (G,H), before relevant protein abundances were determined by immunoblotting with distinct antibodies. The intensity of indicated immunoblots was quantified and normalized by that of β-actin, before being shown as fold changes (on the bottom of indicated blots), which are representative of three independent experiments (n = 3) with significant increases ($, p < 0.05; $$, p < 0.01) and significant decreases (*, p < 0.05; **, p < 0.01), relative to their corresponding control values.

Next, the downstream regulation of miR-1247-5p was also verified in a manner similar to the aforementioned approach. That is, the basal expression of MMP15 or MMP17 at their mRNA and protein levels was effectively silenced by siMMP15 or siMMP17, respectively, transfected in HepG2 cells (Figure 4(D)). Furthermore, markedly reduced expression of MMP15 or MMP17 was observed following co-transfection of siMMP15 or siMMP17 with miR-1247-5p mimics (Figure 4(E,F)), also with a synergistic effect exerted by miR-1247-5p mimics with siMMP15 or siMMP17. Conversely, upon co-transfection of miR-1247-5p inhibitor with either siMMP15 or siMMP17, the inhibitor-leading increases in the expression of MMP15 or MMP17 were partially mitigated by siMMP15 or siMMP17, respectively, hence forming a resistance effect on either MMP15 or MMP17 (Figure 4(G,H)). Of crucial importance, it is plausible that all the aforementioned data obtained from HepG2 cells (Figure 4) were experimentally repeatable in Hep3B cells (as shown in Figure S8).

In order to further confirm the above-described effects of miR-3187-3p or miR-1247-5p on their cognate targets, the indicated miRNAs were constructed into the pLKO.1-TRC cloning vector (for their transient expression) or the lentiviral expression system (subjected to their stable expression, along with the negative controls, i.e. NC) in HepG2 cells (Figure 5(A)). As expected, the expression levels of each of indicated miRNAs (i.e. miR-3187-3p or miR-1247-5p) were substantially incremented to considerably higher extents in all those examined cases, when compared to their negative controls (i.e. NC) (Figure 5(B,C)). Further experimental examinations of such highly expressing miRNAs’ effects on their respective targets revealed that SNAI1 and CDH2 were significantly downregulated by miR-3187-3p (no matter whether it had been transiently or stably expressed), as accompanied by enhanced expression of CDH2 (Figure 5(D,F)). Similarly, MMP15 and MMP17 were also indeed downregulated by miR-1247-5p, but as accompanied by unaffected expression of MMP25 (Figure 5(E,G)). In addition, all the above-mentioned data were also repeatable in Hep3B cells (as deciphered in Figure S9).

Figure 5.

Similar but nuanced effects of transiently or stably expressing miR-3187-3p and miR-1247-5p on their targets.

A. Schematic representations of miR-3187-3p or miR-1247-5p that were subjected to transient or stable expression by indicated pLKO.1-TRC vector and Lentiviral expression systems, respectively.

B,C. HepG2 cells had been transiently transfected or stably infected with each of the above-indicated miRNAs-expressing systems, before being subjected to real-time qPCR analysis. Their relative expression levels are shown graphically as fold changes (Mean ± S.D.) with significant increases ($$, p < 0.01), which were determined from three independent experiments each performed in triplicates (n = 3 × 3).

D,E. The transiently expressing miR-3187-3p’s or miR-1247-5p’s effects on their targets were examined by real-time qPCR and Western blotting with indicated antibodies. The qPCR data are shown graphically as fold changes (Mean ± S.D.) with significant increases ($$, p < 0.01) and significant decreases (**, p < 0.01), which were determined from three independent experiments each performed in triplicates (n = 3 × 3). The intensity of all the indicated immunoblots was quantified and normalized by that of β-actin, before being shown as fold changes (on the bottom of indicated blots). The data are representative of three independent experiments (n = 3) with significant increases ($$, p < 0.01) and significant decreases (*, p < 0.05; **, p < 0.01), relative to their control values.

F,G. The stably expressing miR-3187-3p’s or miR-1247-5p’s effects on their targets were examined by real-time qPCR and Western blotting with indicated antibodies. The qPCR data are shown graphically as fold changes (Mean ± S.D.) with significant increases ($$, p < 0.01) and significant decreases (*, p < 0.05; **, p < 0.01), which were determined from three independent experiments each performed in triplicates (n = 3 × 3). The intensity of all indicated immunoblots was quantified and normalized by that of β-actin, before being shown as fold changes (on the bottom of indicated blots), which are representative of three independent experiments (n = 3) with significant increases ($$, p < 0.01) and significant decreases (*, p < 0.05; **, p < 0.01), relative to their cg control values.

3.5. Cellular mechanisms by which miR-3187-3p and miR-1247-5p inhibit the invasive and migratory ability of HCC

Herein, to ascertain the migratory and invasive capabilities of miR-3187-3p and miR-1247-5p, a series of scratch and transwell experiments were also conducted, after miR-3187-3p and miR-1247-5p had been allowed for transient or stable expression in relevant plasmid-transfected or lentivirus-infected HepG2 cells, respectively. As anticipated, the results demonstrated that both miR-3187-3p and miR-1247-5p are indeed capable of inhibiting the migration and invasion of hepatoma cells to considerably lower extents compared to control values (Figure 6(A), a1 to a6). Further investigation of clone proliferation ability of stably-expressing miR-3187-3p or miR-1247-5p lentivirus-infected cell lines demonstrated that Lenti-miR-3187-3p cells exhibited relatively slower growth of hepatoma cells that had been seeded across a range of cell densities (500–1500) (Figure 6(C)), whereas Lenti-miR-1247-5p cells were only manifested with minimal changes.

Figure 6.

Inhibition of clone proliferation, invasion and migration of hepatoma cells by miR-3187-3p and miR-1247-5p.

A. Scratch and transwell tests revealed that the migration and invasion ability of hepatoma cells were inhibited by miR-3187-3p or miR-1247-5p, which had been transiently or stably expressed in HepG2 cells. The results were also quantified and shown graphically as fold changes (Mean ± S.D.) with significant decreases (*, p < 0.05; **, p < 0.01), which were determined from three independent experiments each performed in triplicates (n = 3 × 3).

B. The apoptosis of stably expressing miR-3187-3p or miR-1247-5p cell lines was determined by flow cytometry.

C. The colony formation assay of stably expressing miR-3187-3p or miR-1247-5p cell lines was carried out. The results were quantified and shown graphically as fold changes (Mean ± S.D.) with significant decreases (*, p < 0.05; **, p < 0.01), which were determined from three independent experiments each performed in triplicates (n = 3 × 3).

D. Flow cytometry analysis of stably expressing miR-3187-3p or miR-1247-5p cell cycles were conducted (left three panels), as graphically shown by the percentage of distinct cell populations at different phases (right panel).

Further insights into cellular apoptosis of stably expressing miR-3187-3p or miR-1247-5p hepatoma cells by flow cytometry, the results demonstrated that the former Lenti-miR-3187-3p slightly enhanced early apoptosis but reduced late apoptosis (Figure 6(B), middle panel), whilst the latter Lenti-miR-1247-5p can consistently promote early and late apoptosis of HepG2 cells (Figure 6(B), lower panel), when compare to the controls. Such nuanced effects of miR-3187-3p or miR-1247-5p on hepatoma cell cyles were also further analysed by flow cytometry. The results unravelled that an increased proportion of cells at the S phase was observed following stable expression of miR-3187-3p, as accompanied by another obviously decreased proportion of cells at the G2/M phase, but with no changes in the G0/G1 phase proportion (Figure 6(D)). Rather, the stable expression of miR-1247-5p only caused a slight increase in the S phase proportion of hepatoma cells, but also as accompanied by largely unchanged proportions of the cells in the G2/M and G0/G1 phases (Figure 6(D)), when compared with their controls.

3.6. Molecular mechanisms by which distinct miRnas contribute to the opposing effects of Nrf1 and Nrf2 on EMT

Herein, to ascertain whether Nrf1 exerts a direct influence on miR-3187-3p, we constructed a series of luciferase reporter genes driven by 3× ARE-adjoining sequences in the miR-3187-3p promoter region (as deciphered in Figure 7(A)). Each of indicated ARE-driven reporters was co-transfected into HepG2 cells, together with an expression construct for Nrf1 or empty vector. As expected, the results revealed that Nrf1 exerted notable trans-acting effects on miR-3187-3p by its directly transcriptional regulation occurring primarily at its #1 and #4 ARE sites, while additional relative weaker transacting effects were executed by the other two ARE sites of #2 and #5 (Figure 7(B)).

Figure 7.

Distinct contributions of miR-3187-3p and miR-1247-5p to the opposing effects of Nrf1 and Nrf2 on the EMT.

A. A schematic representation of ARE-driven luciferase reporters that were constructed from each of those indicated ARE-adjoining sequences within the miR-3187-3p and miR1247–5 promoter regions.

B. The effects of ectopic Nrf1α factor on the transcriptional activity of miR-3187-3p-derived ARE-luc reporter genes were determined. The results were graphically shown as fold changes (Mean ± S.D.) with significant increases ($, p < 0.05; $$, p < 0.01), which were determined from three independent experiments each performed in triplicates (n = 3 × 3).

C. Distinct effects of ectopic Nrf1α on the expression of miR-3187-3p and its downstream proteins were determined by real-time qPCR and Western blotting analysis of HepG2 cells being allowed for overexpression of Nrf1α. The qPCR data are shown graphically as fold changes (Mean ± S.D.) with significant increases ($$, p < 0.01), which were determined from three independent experiments each performed in triplicates (n = 3 × 3). The intensity of indicated immunoblots was quantified and normalized by that of β-actin, before being shown as fold changes (on the bottom of indicated blots), which are representative of three independent experiments (n = 3) with significant increases ($, p < 0.05; $$, p < 0.01) and significant decreases (*, p < 0.05), relative to their controls.

D. The impact of a miR-3187-3p mimic on the examined proteins in Nrf1α^–/– cells was determined by Western blotting of both WT and Nrf1α-deficient cell lines that had been transfected with this miRNA mimic or its negative control (NC). The intensity of immunoblots was quantified and normalized by that of β-actin, before being shown as fold changes (on the bottom of indicated blots), which are representative of three independent experiments (n = 3) with significant increases ($, p < 0.05) and significant decreases (*, p < 0.05), relative to their controls.

E. Distinct effects of ectopic Nrf2 on the transcriptional activity of miR-1247-5p-derived ARE-luc reporter genes were determined. The resulting data were graphically shown as fold changes (Mean ± S.D.) with a significant increase ($$, p < 0.01) and another significant decrease (*, p < 0.05; **, p < 0.01), each of which were determined from three independent expSeriments performed in triplicates (n = 3 × 3).

F. Distinct effects of ectopic Nrf2 on the expression of miR-1247-5p and its downstream proteins were determined by real-time qPCR and Western blotting analysis of HepG2 cells allowed for overexpression of Nrf2. The qPCR data are shown graphically as fold changes (Mean ± S.D.) with a significant increase ($$, p < 0.01) and another significant decrease (**, p < 0.01), which were determined from three independent experiments each performed in triplicates (n = 3 × 3). The intensity of indicated immunoblots was quantified and normalized by that of β-actin, before being shown as fold changes (on the bottom of indicated blots), which are representative of three independent experiments (n = 3) with significant increases ($, p < 0.05; $$, p < 0.01), relative to their controls.

G. The impact of another miR-1247-5p inhibitor on the examined proteins in Nrf2^–/– cells was determined by Western blotting of both WT and Nrf2-deficient cell lines that had been transfected with this miRNA inhibitor or its negative control (NC). The intensity of immunoblots was quantified and normalized by that of β-actin, before being shown as fold changes (on the bottom of indicated blots), which are representative of three independent experiments (n = 3) with significant decreases (*, p < 0.05; **, p < 0.01), relative to their controls.

H. A model is proposed to give a concise explanation of the finding that the EMT of HCCs is negatively regulated by intact Nrf1α-activating miR-3187-3p, but also positively promoted by Nrf2-inhibiting miR-1247-5p. For detailed descriptions, please see supplemental Figure S11.

Next, we further examined the effects of Nrf1-forced expression on miR-3187-3p and its downstream targets. The results showed a significant increase in the miR-3187-3p expression caused by over-expression of Nrf1α in HepG2 cells (Figure 7(C), c2 vs c1). Of particular interest, it was observed that overexpression of Nrf1α led to the evident inhibition of SNAI1 and CDH2 (Figure 7(C), c3) but also was accompanied by substantial activation of both CDH1 and HO-1 (as a canonical downstream enzyme of Nrf1 to exert its antioxidant and detoxifying responses). Further examinations of miR-3187-3p’s effects on the EMT-relevant genes in Nrf1α^–/– cells revealed that this miR-3187-3p mimic was capable of attenuating the elevation of SNAI1 caused by the loss of Nrf1α, such that reduced SNAI1 also led to the restoration of CDH1 abundances, along with another reduction in the CDH2 expression, when compared to their relevant controls (Figure 7(D)).

Similar experiments were performed as described above, to verify a direct effect of Nrf2 on miR-1247-5p. A series of ARE-driven luciferase reporter genes were constructed from the miR-1247-5p promoter region (Figure 7(A), a3). The results demonstrated that Nrf2 exerted a pronounced inhibitory effect on miR-1247-5p at its two ARE sites of #2 and #4 within its promoter region (Figure 7(E)), but as accompanied by a modest transacting effect on its #6 ARE site. The latter weak effect appeared to be nearly inconsequential, due to a quite far distance of the 6th ARE site from its transcription start site. This is supported by further experiments showing that overexpression of Nrf2 led to an overall significant inhibitory effect on the miR-1247-5p expression in HepG2 cells (Figure 7(F), cf. f2 with f1). Such reduction of miR-1247-5p by ectopic Nrf2 caused obvious increases in the expression of MMP15 and MMP17, but not MMP25, aside from transactivation of the cognate downstream target GCLC by this CNC-bZIP factor (Figure 7(F), f4 to f6). Conversely, following the transfection of a miR-1247-5p inhibitor into Nrf2 ^–/– cells, it was shown that this miRNA inhibitor enabled to mostly rescue the declined expression of MMP15 and MMP17 caused by the loss of Nrf2 (Figure 7(G)), but almost no effects were observed on MMP25.

In addition, the above-mentioned experimental data were all repeatable to be obtained from Hep3B cells (as shown in Figure S10). Taken altogether, these results demonstrated that the EMT of HCCs is negatively regulated by Nrf1α-activating miR-3187-3p, but quite positively promoted by Nrf2-inhibiting miR-1247-5p (Figures 7(H) and S11).

4. Discussion

Clearly, human liver cancer is one of the most prevalent malignant neoplasms worldwide with a high incidence of morbidity and mortality. Of pivotal significance, thus it is how to overcome the metastasis of hepatocellular carcinoma (HCC) because it represents a crucial step in advancing the treatment of this disease. For this end, it is of importance to elucidate the cellular and molecular mechanisms whereby the EMT exerts a critical role in the promotion of HCC initiation, progressive development and malignant metastasis. For instance, by suppression of the E-cadherin (CDH1) expression during EMT, it can indeed enable for transformation of a typical polygonal, cobblestone-like epithelial cell shape to acquire another spindle-type mesenchymal morphology. As a consequence, this is also allowed to acquire the additional ability to invade and metastasize remotely. In this process, a progressively increased expression pattern of those mesenchymal-associated markers (including N-calmodulin (CHD2), vimentin and fibronectin, as well as matrix metallopeptidease proteins (MMPs) [29]) is subsequently accompanied by another decreased expression pattern of the epithelial-associated proteins. Overall, these prior studies have also demonstrated that a canonical EMT has been successfully achieved through relevant transcription factors, including Snail, Twist, and ZEB families, which all enable to exert a predominant influence to tightly govern the expression abundances of both epithelial and mesenchymal markers. In the present study, it is found that the EMT of HCCs is negatively regulated by antioxidant transcription factor Nrf1α-activating miR-3187-3p signalling, but also positively promoted by its homological factor Nrf2-inhibiting miR-1247-5p network (as illustrated in Figure 7(H), and also see Figure S11).

Since the EMT is dictated by the expression profiling of specific marker genes regulated by relevant transcription factors, it is inferable that microRNAs also play a pivotal role in the transcriptional regulation of those indispensable genes for EMT [36,37]. A previous study demonstrated that the forced expression of miR194 in hepatic stromal cells results in an reduction of N-calmodulin, which, in turn, prevents the cell migration and invasion [12]. Furthermore, miR148a has also been shown to reduce the accumulation of SNAI1 by binding to Met (a proto-oncogene-type receptor tyrosine kinase) in hepatocyte growth, which enables to activate the downstream Akt phosphorylation at its Ser⁴⁷³ but also inhibit the phosphorylation of GSK-3β at Ser⁹, thereby leading to remission of EMT in HCCs [38]. Besides, it has also been proposed that Nrf2 plays a role in the EMT process of liver cancer cells [27,39], although antioxidant transcription factors Nrf1 and Nrf2 are known to play a vital role in maintaining redox homoeostasis and organ integrity (including morphostasis) across within organisms [25,40]. Nevertheless, the involvement of Nrf1- and Nrf2-mediated microRNAs in EMT regulation in liver cancer cells remains to be not yet well documented, as far as we know from the current literature. In this study, we investigated the differential roles of miRNAs regulated by Nrf1 and Nrf2 in putative EMT of HCC cells, along with such a novel view to elucidate the impact of these regulatory processes on the EMT of liver cancer cells. The cellular-molecular mechanisms accounting for the phenotypic differences between Nrf1α^–/– and Nrf2^–/– cell lines derived from their WT cells were further verified by bioinformatics analysis of small RNA-sequencing and transcriptome, in combination with a series of the classic ‘wet’ experiments.

Here is a major object of this study to aim at ascertaining how miR-3187-3p and miR-1247-5p are regulated by Nrf1 and/or Nrf2. Amongst those accurately screened downstream targets of miR-3187-3p and miR-1247-5p, SNAI1, MM15, MMP17, and MMP25 were determined as the key EMT molecules for further investigation. Subsequently, the functional effects of miR-3187-3p and miR-1247-5p on EMT were further identified by administering their mimics and inhibitors. The mechanisms underlying the action of miR-3187-3p and miR-1247-5p were explored in depth by using their transient or stable expression cell lines. As expected, the results demonstrated that miR-3187-3p could target and inhibit the expression level of SNAI1, as described elsewhere [9] and that, as a typical EMT-related transcription factor, SNAI1 could precision-regulate the expression of its downstream CDH1 and CDH2 to certain homeostatically balanced extents, thereby inhibiting EMT. By contrast, miR-1247-5p has been shown to directly target both MM15 and MP17, reducing the invasive capacity of liver cancer cells and hence inhibiting its EMT process. Such differential regulation of miR-3187-3p and miR-1247-5p by Nrf1 and/or Nrf2 and their distinct contributions to the opposing effects of these two CNC-bZIP factor on the EMT of liver cancer cells. Altogether, the evidence has been provided herein, revealing that miR-3187-3p is positively regulated by Nrf1, whilst miR-1247-5p is negatively regulated by Nrf2 (albeit as a dominant transactivator). Therefore, it is quite plausible that the occurrence of EMT in liver cancer cells is tightly confined by Nrf1-activating miR-3187-3p towards the SNAI1-CDH1/2 signalling axis, but promoted by Nrf2-inhibiting miR-1247-5p signalling to the MMP15/17 network.

Such two key miRNAs, miR-3187-3p and miR-1247-5p identified in this study, could be characterized to function as tumour suppressors, which is rather not in contradiction with previous reports [41–43]. This notion is also further supported by prior demonstration that the miR-3187-3p expression is significantly diminished in many malignant tumours, while the elevated levels of miR-1247-5p are markedly linked to those favourable patient prognoses. Thereby, it is inferable that Nrf1, which also acts as a putative tumour suppressor, can regulate the EMT process of HCC cells by modulating miR-3187-3p in addition to controlling robust redox homoeostasis. By sharp contrast, Nrf2, as a crucial redox regulator within cells, has been demonstrated by numerous studies and clinical data analysis to exhibit aberrant expression of this CNC-bZIP factor, resulting in malignant proliferation of many tumours, with a poor prognosis [27]. Herein, Nrf2 has been shown to inhibit the expression of miR-1247-5p, which, in turn, can promote the EMT process of liver cancer. Taken together, both miR-3187-3p and miR-1247-5p enable to function as intermediate bridges of Nrf1 and/or Nrf2 between their regulatory networks involved in the activation of EMT signalling towards HCC progression [44] (Figure S11).

The SNAI1 expression is likely induced by Nrf2, acting as a powerful transactivator, but herein we found that Nrf2 enables to inhibit the expression of miR-1247-5p. This is fully consistent with our sequencing data obtained from the knockout of NRF2 that causes a substantial increase of miR-1247-5p (and even miR-3187-3p), as a consequence. Such Nrf2-transactivated inhibition was further confirmed by ARE-driven luciferase reporter gene experiments. The distinct roles of Nrf2 in regulating SNAI1 from miR-1247-5p are much likely interpreted by different functional heterodimers of active Nrf2 with each of small Maf or other bZIP proteins, which were divergently recruited to distinct target genes containing differential ARE-adjoining sequences in their promoter regions. Distinction in their DNA-binding activity is likely affected by chromosomal topological-associated domain adjacent to ARE sites and/or such a spatial repression effect of different functional heterodimer of Nrf2 binding to the ARE sites.

As such being the case, a limitation of this study should also be declared that all the relevant researches have been conducted primarily at the cellular and molecular levels. Therefore, it is imperative for us to develop animal translational models, aiming to elucidate the consequences of animal experimentation in light of the discrepancies between in vivo and in vitro experiments. This will be a key focus of our future study to explore distinct contributions of miR-3187-3p and miR-1247-5p to the opposing effects of Nrf1 and Nrf2 on the EMT during cancer development and malignant metastasis.

Funding Statement

This study was funded by the National Natural Science Foundation of China (NSFC, with three project grants [82073079],[81872336] and [82473147] awarded to Prof. Yiguo Zhang, and also partially by the Key project [BYKY2024002] of Bishan Hospital of Chongqing Medical University awarded to Medical Dr J.C. the Key project of Bishan Hospital of Chongqing Medical University

Disclosure statement

No potential conflict of interest was reported by the author(s).

Author contributions

J.C., J.F., Y-p.Z., and S.H. performed all the experiments, collected all the relevant data, and made the manuscript draft with figures and supplemental information. S.H. did bioinformatics analysis of datasets. Lastly, Y.Z. designed and supervised this study, analysed all the data, helped to prepare all figures with cartoons, wrote and revised the paper.

Availability of data and materials

All data needed to evaluate the conclusions in the paper are present in this publication along with the supplementary documents that can be found online. Additional other data related to this paper may also be requested from the corresponding author (with a lead contact at the Email: eaglezhang@fyust.edu.cn or yiguozhang@cqu.edu.cn).

Supplementary material

Supplemental data for this article can be accessed online at https://doi.org/10.1080/15476286.2025.2548628