MiRNA-140 acts as a liver tumor suppressor by controlling NF-κB activity via directly targeting Dnmt1 expression

Abstract

MicroRNAs (miRNAs) are small RNAs that regulate the expression of specific target genes. While deregulated miRNA expression levels have been detected in many tumors, whether miRNA functional impairment is also involved in carcinogenesis remains unknown. We investigated whether deregulation of miRNA machinery components and subsequent functional impairment of miRNAs are involved in hepatocarcinogenesis. Among miRNA-containing ribonucleoprotein complex components, reduced expression of DDX20 was frequently observed in human hepatocellular carcinomas, in which enhanced NF-κB activity is believed to be closely linked to carcinogenesis. Because DDX20 normally suppresses NF-κB activity by preferentially regulating the function of the NF-κB-suppressing miRNA-140, we hypothesized that impairment of miRNA-140 function may be involved in hepatocarcinogenesis. Dnmt1 was identified as a direct target of miRNA-140, and increased Dnmt1 expression in DDX20-deficient cells hypermethylated the promoters of metallothionein genes, resulting in decreased metallothionein expression leading to enhanced NF-κB activity. MiRNA-140-knockout mice were prone to hepatocarcinogenesis and had a phenotype similar to that of DDX20 deficiency, suggesting that miRNA-140 plays a central role in DDX20 deficiency-related pathogenesis.

Conclusion

These results indicate that miRNA-140 acts as a liver tumor suppressor, and that impairment of miRNA-140 function due to a deficiency of DDX20, a miRNA machinery component, could lead to hepatocarcinogenesis.

INTRODUCTION

Hepatocellular carcinoma (HCC) is the third most common cause of cancer-related mortality worldwide (1). Although multiple major risk factors have been identified, such as infection with hepatitis viruses B or C, the molecular mechanisms underlying HCC development remain poorly understood, hindering the development of novel therapeutic approaches. Therefore, a better understanding of the molecular pathways involved in hepatocarcinogenesis is critical for the development of new therapeutic options.

Nuclear factor-κB (NF-κB) is one of the best-characterized intracellular signaling pathways. Its activation is a common feature of human HCC (2–4). It acts as an inhibitor of apoptosis and as a tumor promoter (4, 5), and is associated with the acquisition of a transformed phenotype during hepatocarcinogenesis (6). In fact, studies using patient samples suggest that NF-κB activation in the liver leads to the development of HCC (7). Although there are conflicting reports (8), activation of the NF-κB pathway in the liver is crucial for the initiation and promotion of HCC (4).

MicroRNAs (miRNAs) are small RNA molecules that regulate the expression of target genes and are involved in various biological functions (9–12). Although specific miRNAs can function as either suppressors or oncogenes in tumor development, a general reduction in miRNA expression is commonly observed in human cancers (13–22). In this context, it can be hypothesized that deregulation of the machinery components involved in miRNA function may be related to the functional impairment of miRNAs and the pathogenesis of carcinogenesis.

Here we show that the expression of DDX20, a miRNA-containing ribonucleoprotein (miRNP) component, is frequently decreased in human HCC. Because DDX20 is required for both the preferential loading of miRNA-140 into the RNA-induced silencing complex and its function (23), we hypothesized that DDX20 deficiency would lead to hepatocarcinogenesis via impaired miRNA-140 function. MiRNA-140 knockout mice were indeed more prone to hepatocarcinogenesis, and we identified a possible molecular pathway from DDX20 deficiency to liver cancer.

EXPERIMENTAL PROCEDURES

Mice and liver tumor induction

MiRNA-140^−/− mice have been described previously (24). Recombinant murine TNF-α (25 μg/kg; Wako, Osaka, Japan) was injected into the tail vein, and the mice were sacrificed 1 h later. To induce liver tumors, 15-day-old mice received an intraperitoneal injection of diethylnitrosamine (DEN) (25 mg/kg body weight), and were sacrificed 32 weeks later. All animal experiments were carried out in compliance with the regulations of the Animal Use Committee of the University of Tokyo and the Institute for Adult Disease, Asahi Life Foundation.

Plasmids

FLAG-tagged human DDX20-expressing plasmids were as described previously (23). The pGL3-based reporter plasmid containing Dnmt1 3′ UTR sequences was provided by Dr. G. Marucucci (25).

Detailed experimental procedures

The detailed experimental procedures of clinical samples, cells, plasmids, reporter assays, RT-PCR analysis, antibodies, western blotting, cell assays, immunohistochemistry, microarray analysis, methylation analysis, and electrophoretic mobility-shift assay are described in the Supplementary Experimental Procedures.

Statistical analysis

Statistically significant differences between groups were determined using Wilcoxon rank-sum test. Wilcoxon signed-rank test was used for statistical comparisons of protein expression levels between HCC and surrounding non-cancerous tissues.

RESULTS

DDX20 expression is frequently decreased in HCC

The expression levels of proteins reported to be miRNP components (Dicer, Ago2, TRBP2, DDX20 [also known as Gemin3], Gemin4) (26) were initially determined by immunohistochemistry in HCC and non-cancerous background liver tissues from 10 patients. DDX20 expression was lower in HCC tissue compared with the surrounding non-cancerous tissue in 8 of 10 cases, while expression of the other genes was unchanged (Table 1 and Supplementary Fig. S1). Therefore, and because DDX20 was recently identified as a possible liver tumor suppressor in mice (27), we determined its role as a human HCC suppressor.

Table 1.

Number of cases with differential expression levels of miRNP components in HCC (n = 10). The expression levels of each miRNP component were determined by immunohistochemistry.

The numbers indicate the number of cases which had the differential expression levels (decreased, increased, or no change) in HCC tissues compared with those in the surrounding liver tissues.

Gene ID	Gene symbol	Decreased	Increased	No change
23405	Dicer1	2	1	7
27161	EIF2C2 (AGO2)	1	1	8
6895	TARBP2(TRBP2)	2	0	8
11218	DDX20 (GEMIN3)	8	0	2
50628	GEMIN4	1	0	9

DDX20 protein expression was lower in several HCC cell lines, such as Huh7 and Hep3B (Fig. 1a), compared with normal hepatocytes. DDX20 protein levels were also lower in human HCC needle biopsy specimens than in surrounding non-cancerous liver tissue (Fig. 1b). Immunohistochemical analysis confirmed that DDX20 expression was frequently lower in HCC than in surrounding non-cancerous liver tissue (Fig. 1c, d). Specifically, 47 of 70 cases examined showed reduced DDX20 protein expression in HCC versus background non-cancerous liver tissue (Fig. 1d and Supplementary Table S1). These results indicate that the expression of DDX20, a miRNP component, is frequently reduced in human HCC, and suggest that this reduced DDX20 expression might be involved in the pathogenesis of a subset of HCC cases.

Figure 1. Reduced DDX20 expression levels in hepatocellular carcinoma.

a, DDX20 protein expression in hepatocellular carcinoma (HCC) cell lines. Numbers between the panels indicate DDX20 protein levels normalized to β-actin levels. Lysates of 293T cells transiently transfected with a FLAG-tagged DDX20-expressing plasmid yielded two DDX20 bands corresponding to the endogenous DDX20 protein and the transfected FLAG-tagged DDX20 protein (*) as a positive control (p.c.; far right lane). Data are the results of three independent determinations. b, DDX20 protein expression in four HCC needle biopsy specimens and in the surrounding non-cancerous background liver tissue (back). * positive control. c, Immunohistochemical analysis of DDX20 protein expression in HCC and surrounding tissues (background liver). Two representative cases are shown. Scale bar, 500 μm. The lower panels display magnified images of the boxed areas in the upper panels. d, Grid summarizing DDX20 immunohistochemical staining data from 70 cases. In 47 cases (shaded in pink), DDX20 protein levels were lower in the HCC tissues than in the surrounding tissues (p < 0.05; Wilcoxon signed-rank test).

NF-κB activity is enhanced by DDX20 deficiency

Because DDX20 knockout mice are embryonic-lethal (28), DDX20 has been suggested to have important biological roles. DDX20, a DEAD-box protein (29), was originally found to interact with survival motor neuron protein (30). Later, it was identified as a major component of miRNPs (26, 31), which may mediate miRNA function. As we reported previously, DDX20 is preferentially involved in miRNA-140-3p function (23), acting as a suppressor of NF-κB activity in the liver (32). DDX20-knockdown PLC/PRF/5 cells exhibit enhanced NF-κB activity (23) (Fig. 2a). While the proliferation rates of DDX20-knockdown cells and control cells were comparable (Fig. 2b), apoptotic cell death after stimulation with tumor necrosis factor (TNF)-related apoptosis-inducing ligand (TRAIL), which induces both cell apoptosis and NF-κB activation (33), was significantly reduced in DDX20-knockdown cells (Fig. 2c). Similar results were obtained using DDX20-knockdown HepG2 cells (Supplementary Fig. S2a, b, c, and d). Conversely, NF-κB activity was reduced, but cell proliferation remained unchanged, in Hep3B cells stably overexpressing DDX20 (Fig. 2d, e). Sensitivity to TRAIL-induced apoptosis was restored in these cells (Fig. 2f). Similar results were also obtained using Huh7 cells (Supplementary Fig. S2e, f, g, and h). These data confirm a previous report that DDX20 deficiency enhances NF-κB activity and the downstream events of this pathway.

Figure 2. Modulation of downstream events of the nuclear factor-κB pathway by DDX20.

a, Left panel: establishment of stable DDX20-knockdown (DDX20 KD) PLC/PRF/5 cells. *, positive control; p.c. Right panel: DDX20 deficiency enhances tumor necrosis factor (TNF)-α-induced nuclear factor (NF)-κB activity. NF-κB reporter plasmids were transiently transfected into control (Ctrl) or DDX20-knockdown (KD) PLC/PRF/5 cells. The cells were then treated with TNF-α (5 ng/mL) or vehicle for 6 h. *p < 0.05. Data represent the mean ± standard deviation (SD) of three independent determinations. b, Cell proliferation rates were comparable for control (Ctrl) and DDX20-knockdown (KD) PLC/PRF/5 cells. Data represent the mean ± SD of three determinations. c, DDX20 deficiency reduces TNF-related apoptosis-inducing ligand (TRAIL)-induced apoptotic cell death. Control (Ctrl) and DDX20-knockdown (KD) PLC/PRF/5 cells were incubated with 25 ng/mL TRAIL. Data are cell viability after TRAIL stimulation (gray bars) relative to the number of vehicle-treated cells (white bars). *p < 0.05. Data are the mean ± SD of triplicate determinations. d, Left panel: establishment of stable DDX20-overexpressing cells. Hep3B cells were infected with control or FLAG-tagged DDX20-overexpressing lentiviruses and selected on puromycin. Western blot analysis confirmed increased expression of DDX20 protein. Right panel: DDX20 overexpression suppresses TNF-α-induced NF-κB activity. NF-κB reporter plasmids were transiently transfected into Hep3B control (Ctrl) and DDX20-overexpressing (DDX20) cells treated with TNF-α for 6 h. Data are the mean ± SD of three independent determinations. *p < 0.05. e, Proliferation of control (Ctrl) and DDX20-overexpressing (DDX20) Hep3B cells was measured as described in (b). f, DDX20 overexpression reduces TRAIL-induced apoptotic cell death. Control (Ctrl) and DDX20-overexpressing (DDX20) Hep3B cells. *p < 0.05.

Metallothionein expression is decreased by DDX20 deficiency

Next, to investigate the biological consequences of DDX20 deficiency, we examined the changes in transcript levels in DDX20-knockdown cells using microarrays (GEO accession number: GSE28088). The expression of genes driven by NF-κB that are related to carcinogenesis, such as FASLG, IRAK1, CARD9, and Galectin-1, were enhanced significantly in DDX20-knockdown cells, as expected (Table 2). To determine the mechanism underlying the enhanced NF-κB activation in DDX20-deficient cells, we searched for candidate genes and noticed that the expression levels of a group of metallothioneins (MTs), such as MT1E, MT1F, MT1G, MT1M, MT1X, and MT2A, were all significantly decreased when DDX20 was deficient (Table 3). The decreased expression of MTs in DDX20-knockdown HepG2 and PLC/PRF/5 cells was confirmed by quantitative reverse-transcriptase polymerase chain reaction (qRT-PCR; Fig. 3a and Supplementary Fig. S3). Expression of MT-3, which was not altered in the microarray analysis, was similarly unaltered in qRT-PCR analysis. Notably, it was already known that MTs are frequently silenced in human primary liver cancers (34–36). In addition, MT knockout mice have enahnced NF-κB activity, likely due to reactive oxygen species, and these mice are more prone to hepatocarcinogenesis (37). These results suggest that DDX20 deficiency enhances NF-κB activity by decreasing the expression of MTs, which could facilitate the development of liver cancer.

Table 2.

Increased expression of NF-κB-related genes in DDX20-knockdown HepG2 cells compared to wild-type cells. The genes were identified as NF-κB-related based on the Gene Ontology and the GeneCodis Databases.

RefSeq ID	Symbol	Description	Ratio	Representative gene function
NM_000639	FASLG	Fas ligand	3.5	NF-kappaB target, apoptosis
NM_052813	C9orf151	CARD9	2.5	NF-kappaB cascade, NF-kappaB target
NM_014959	CARD8	Tumor-up-regulated CARD-containing antagonist of CASP9 (TUCAN).	2.2	NF-kappaB target
NM_131917	FAF1	FAS-associated factor 1 (hFAF1)	1.9	Cytoplasmic sequestering of NF-kappaB, NF-kappaB target
NM_020644	TMEM9B	Transmembrane protein 9B precursor	1.9	Positive regulation of NF-kappaB transcription factor activity
NM_017544	NKRF	ITBA4 protein	1.9	Negative regulation of transcription
XM_375632	PPP5C	Protein phosphatase T	1.8	Positive regulation of NF-κB cascade
NM_020345	NKIRAS1	KappaB-Ras1	1.8	NF-kappaB cascade
NM_001569	IRAK1	IRAK-1	1.7	Positive regulation of NF-kappaB transcription factor activity
NM_177951	PPM1A	Protein phosphatase 1A	1.7	Positive regulation of NF-κB cascade
NM_018098	ECT2	Epithelial cell-transforming sequence 2 oncogene	1.6	Positive regulation of NF-κB cascade
NM_002305	LGALS1	Galectin-1(putative MAPK-activating protein MP12).	1.6	Positive regulation of NF-κB cascade
NM_015093	TAB2	TAK1-binding protein 2	1.6	Positive regulation of NF-κB cascade
NM_004180	TANK	TRAF-interacting protein	1.5	NF-kappaB cascade
NM_014976	PDCD11	Programmed cell death protein 11	1.5	rRNA processing
NM_015336	ZDHHC17	Putative NF-kappa-B-activating protein 205	1.5	Positive regulation of NF-κB cascade
NM_002503	NFKBIB	IKB-beta	1.5	Cytoplasmic sequestering of NF-kappaB
NM_138330	ZNF675	zinc finger protein 675	1.5	Negative regulation of NF-kappaB transcription factor activity

Table 3.

Decreased expression levels of metallothionein genes in DDX20 knockdown HepG2 cells as compared with wild-type cells. Bold numbers indicate below 0.5.

Symbol	Description	Ratio
MT1E	Metallothionein-1E	0.12
MT1F	Metallothionein-1F	0.36
MT1H	Metallothionein-1H	0.16
MT1G	Metallothionein-1G	0.06
MT1M	Metallothionein-1M	0.24
MT1X	Metallothionein-1X	0.27
MT2A	Metallothionein-2	0.28
MT3	Metallothionein-3	0.84
MTL5	Metallothionein-like 5 (Tesmin)	1.12

Figure 3. Targeting of Dnmt1 by microRNA-140-3p and reduced metallothionein expression.

a, The expression levels of metallothioneins (MTs) were determined by quantitative reverse-transcriptase polymerase chain reaction. The relative expression ratios of the MTs in control (white bars) and DDX20-knockdown (black bars) HepG2 cells were calculated by normalizing control cell values to 1.0. The data represent the mean ± SD of three independent determinations. *p < 0.05. b, Putative microRNA (miRNA)-140-3p target sites in the 3′ untranslated region (UTR) of human Dnmt1. Seed sequences are indicated in red. c, Dnmt1 expression was increased in DDX20-knockdown cells. Ctrl, control cells; KD, DDX20-knockdown cells. d, Left panel; schematic diagrams of wild-type (upper) and mutant (lower) luciferase reporter constructs (Luc-Dnmt1-3′ UTRs) carrying the Dnmt1 3′ UTR region harboring the putative miRNA-140-3p target site. The mutant seed sequence contained two nucleotide substitutions. Right panel; the Dnmt1 3′ UTR is targeted directly by miRNA-140-3p. Cells were co-transfected with Luc-Dnmt1-3′ UTR (wild-type or mutant) plus either an empty vector (white bars) or a plasmid expressing the miRNA-140 precursor (black bars). Data are the mean ± SD of three independent determinations. e, Overexpression of miRNA-140 reduces Dnmt1 expression in control cells. Values between the panels indicate Dnmt1 protein levels normalized to those of β-actin. KD, DDX20 knockdown cells. f, Representative histochemical images showing expression of DDX20, Dnmt1, and MT proteins in hepatocellular carcinoma (HCC; upper three panels) and surrounding tissue (lower panels). Compared with adjacent non-cancerous liver tissue, HCCs exhibited decreased DDX20 and MT expression, and increased Dnmt1 expression. Note that adjacent sections were stained for each protein. Scale bar, 50 μm.

MiRNA-140 directly targets Dnmt1

Because MT expression is regulated principally by CpG island methylation in their promoter regions (38, 39), we examined the quantitative methylation status of MT promoters in DDX20-knockdown cells. The CpG islands of the MT1E, MT1G, MT1M, MT1X, and MT2A promoters, and the CpG shores of the MT1F promoters, were significantly more highly methylated under DDX20-deficient conditions, as determined by the comprehensive Illumina Quantitative Methylation BeadChip method (Table 4, Supplementary Table S2, and GSE 37633). A crucial step in DNA methylation involves DNA methyltransferase (Dnmt), which catalyzes the methylation of CpG dinucleotides in genomic DNA (40). The methylation status of MT promoters is mediated specifically by Dnmt1 (41). Because Dnmt1 contains a predicted miRNA-140-3p target site in its 3′ untranslated region (UTR), with a perfect match to its seed sequences (Fig. 3b), and because the effects of miRNA-140-3p activity were impaired in DDX20-knockdown cells (23), it was hypothesized that while miRNA-140 normally targets and suppresses Dnmt1 protein expression, miRNA-140-3p dysfunction due to DDX20 deficiency results in enhanced Dnmt1 expression, leading to hypermethylation of MT promoters. Consistent with this, Dnmt1 expression was increased significantly in DDX20-knockdown cells (Fig. 3c). MiRNA-140 precursor overexpression suppressed the activity of the Dnmt1 3′ UTR reporter construct, the effect of which was lost when two mutations were introduced into its seed sequences (Fig. 3d). MiRNA-140 precursor overexpression suppressed Dnmt1 protein expression (Fig. 3e). These results indicate that miRNA-140 directly targets Dnmt1 and suppresses its expression in the normal state. Consistently, decreased DDX20, increased Dnmt1, and decreased MT expression were detected together in human clinical HCC samples, as determined by immunohistochemistry (Fig. 3f). By contrast, miRNA-140 precursor-overexpressing Huh7 cells showed increased expression of MTs and reduced NF-κB activity in vitro (Supplementary Fig. S4a, b). Moreover, the increase in the number of spheres formed from PLC/PRF/5 cells due to DDX20 knockdown was antagonized by treatment with an NF-κB inhibitor or a demethylating agent (Supplementary Fig. S5). Taken together, these results suggest that the upregulated Dnmt1 protein expression caused by functional impairment of miRNA-140-3p due to DDX20 deficiency results in decreased expression of MTs via enhanced methylation at the CpG sites in their promoters. This may lead to enhanced NF-κB activity and cellular transformation at least in vitro.

Table 4.

Methylation levels in CpG islands of the metallothionein genes in DDX20-knockdown HepG2 cells compared with control cells, as determined by the quantitative Illumina Human Methylation BeadsChip. Bold numbers indicate increased methylation levels in DDX20 knockdown cells.

Symbol	CpG island methylation ratio	Target ID
MT1E	1.14	cg00178359
	1.29	cg06463589
	3.65	cg02512505
	1.02	cg15134649
MT1G	2.14	cg16452857
	1.03	cg27367960
	1.00	cg03566142
	0.99	cg07791866
MT1M	1.16	cg02132560
	0.98	cg02160530
	1.03	cg04994964
MT1X	1.24	cg05596720
	1.05	cg26802333
	1.06	cg09147880
	1.01	cg08872713
MT2A	2.06	cg07395075
	0.94	cg20430434

MiRNA-140 is a liver tumor suppressor

To further examine the biological consequences of functional impairment of miRNA-140 due to DDX20 deficiency, we determined the phenotypes of miRNA-140 knockout (miRNA-140^−/−) mice (Fig. 4a). Similar to the in vitro DDX20 knockdown results, Dnmt1 expression was increased and MT levels decreased in the liver tissue of these mice (Fig. 4b). NF-κB–DNA binding activity was enhanced in the livers of miRNA-140^−/− mice after the tail-vein injection of TNF-α, a crucial cytokine that induces NF-κB activity and hepatocarcinogenesis (Fig. 4c). As was found in MT knockout mice, phosphorylation of p65 at serine 276, which is critical for p65 activation, was significantly increased in the liver of miRNA-140^−/− mice after DEN exposure, which induces NF-κB activation and liver tumors (37) (Fig. 4d). Notably, the size and number of liver tumors that developed 8 months after DEN exposure were markedly elevated in miRNA-140^−/− mice compared with control mice (Fig. 4e, f). These results indicate that miRNA-140^−/− mice are indeed more prone to liver cancer development, and suggest that miRNA-140 acts as a liver tumor suppressor, probably by suppressing NF-κB activity, although we cannot at present completely exclude other molecular mechanisms. Nonetheless, these results also suggest that the impairment of miRNA-140 function due to DDX20 deficiency may lead to hepatocarcinogenesis in human, as we have observed in miRNA-140^−/− mice (Supplementary Fig. S6 and S7).

Figure 4. miRNA-140^−/− mice are prone to hepatocarcinogenesis.

a, Representative genotyping of mice with wild-type or mutant alleles. PCR genotyping was performed for microRNA (miRNA)-140 wild-type (419 bp; wild) and knockout (734 bp; mutant) alleles. (+/+), wildtype; (+/−), hetero; (−/−), knockout. b, Increased Dnmt1 expression and decreased metallothionein (MT)I/II expression in the liver tissues of miRNA-140^−/− mice compared with wildtype mice. Western blotting was performed using antibodies against the indicated proteins. Wildtype; (+/+), miRNA-140^−/−; (−/−). The image shown is representative of four independent experiments. c, NF-κB–DNA binding was assessed by gel-shift assay using equal amounts of liver nuclear extracts from untreated and TNFα-injected wildtype and miRNA-140^−/− mice. Wildtype; (+/+), miRNA-140^−/−; (−/−). Cold probe was added to TNFα-injected knockout mouse nuclear extract to test assay specificity. A result representative of four independent experiments is shown. d, Western blotting for phosphorylated p65 expression in the liver at 32 weeks after DEN treatment in miRNA-140^−/− mice compared to wildtype mice. A result representative of four independent experiments is shown. e, Representative histological images of mouse liver at 32 weeks after DEN treatment. Arrows indicate tumors. Lower panels: higher-magnification images of the highlighted areas in the upper panels. Scale bar, 500 μm. f, The number (left panel) and size (right panel) of tumors (five random sections per mouse treated with DEN) are shown as the means ± standard deviations (wildtype, n = 8; miRNA-140^−/− mice, n = 8). *, p < 0.05.

DISCUSSION

Here, we report that miRNA-140^−/− mice have increased NF-κB activity and are more prone to HCC development. In addition, we show that DDX20, a miRNP component, is frequently decreased in human HCC tissues. Because DDX20 deficiency preferentially causes impaired miRNA-140 function (23), the functional impairment of miRNA-140 may result in phenotypes similar to those of miRNA-140^−/− mice and may lead to hepatocarcinogenesis. In support of the hypothesis that DDX20 dysfunction is involved in hepatocarcinogenesis, DDX20 is located at 1p21.1–p13.2, a frequently deleted chromosomal region in human HCC (27), and DDX20 was recently identified as a possible liver tumor suppressor in a functional screen in mice (27). While the possibility that intracellular signaling pathways other than miRNA-140 may also be involved in the biological consequences of DDX20 deficiency cannot be denied, we believe that functional impairment of miRNA-140 plays a major role in the phenotypes induced by DDX20 deficiency, based on the phenotypic similarities.

Changes in miRNA expression levels have been reported in various tumors (7, 12, 42). However, in this study, we found that reduced expression of a miRNA machinery component might lead to carcinogenesis, at least in part, through functional impairment of miRNAs. Recent studies have shown that components of the RNA interference machinery are associated with the outcome of ovarian cancer patients (43), and that single-nucleotide polymorphisms in miRNA machinery genes can be used as diagnostic risk markers (44, 45). Therefore, the impairment of miRNA function caused by deregulated miRNA machinery components may also be involved in carcinogenesis.

Our study identified Dnmt1 as a critical target of miRNA-140. The decreased MT expression due to the CpG promoter methylation induced by Dnmt1 resulted in enhanced NF-κB activity. This finding was consistent with the results obtained using MT gene knockout mice, in which enhanced NF-κB activation promoted hepatocarcinogenesis (37). The decrease in MT expression that results from the increased Dnmt1 expression caused by functional impairment of miRNA-140, together with the increased NF-κB activation and hepatocarcinogenesis in MT knockout mice (37), support the concept that the DDX20/miRNA-140/Dnmt1/MTs/NF-κB pathway may play a crucial role in hepatocarcinogenesis. However, we cannot fully exclude the possibility that other intracellular signaling pathways are also involved in the induction of hepatocarcinogenesis by miRNA-140 or DDX20 deficiency, because the precise role of NF-κB in hepatocarcinogenesis has not been clearly defined (8), although constitutive activation of NF-κB signaling has been frequently detected in human HCCs (46). The mechanisms by which DDX20 expression is initially decreases and the reason its locus is frequently deleted in HCC remain to be elucidated. However, because DDX20 expression is also regulated by methylation of its CpG promoter (47), once this pathway is deregulated, decreased DDX20 expression could be maintained by a positive feedback mechanism, even without deletion of its locus (27).

In conclusion, this study showed that miRNA-140 acts as a liver tumor suppressor. We showed that DDX20, a miRNP component, is frequently decreased in human HCC, which may induce hepatocarcinogenesis via impairment of miRNA-140 function. These results suggest the importance of investigations of not only aberrant miRNA expression levels (12, 14, 17, 48), but also deregulation of miRNP components (22), with subsequent impairment of miRNA function as molecular pathways and possible therapeutic targets for carcinogenesis and other diseases.