Tumor‐suppressive microRNA silenced by tumor‐specific DNA hypermethylation in cancer cells

Abstract

MicroRNA (miRNA) genes, located in intergenic or intragenic non‐coding regions of the genome, are transcribed and processed to small non‐protein‐coding RNA of approximately 22 nucleotides negatively regulating gene expression. Some miRNA have already been reported for their genetic alterations, aberrant expression and oncogenic or tumor‐suppressive functions. After 2008, there has been a striking increase in the number of publications reporting tumor‐suppressive miRNA (TS‐miRNA) silenced epigenetically in various types of cancers, suggesting important clinical applications for miRNA‐based molecular diagnosis and therapy for cancers. Here, we introduce a correlation of the gene silencing of TS‐miRNA through CpG island hypermethylation with the genomic distances between intergenic and intragenic miRNA genes or protein‐coding host genes and CpG islands located around these genes. Furthermore, we also discuss the potential of miRNA replacement therapy for cancers using double‐stranded RNA mimicking TS‐miRNA. (Cancer Sci 2012; 103: 837–845)

MicroRNA (miRNA) are small, non‐coding and single‐strand RNA of 19–22 nucleotides with a primary role in post‐transcriptional silencing generally through imperfect pairing with the 3′‐UTR of protein‐coding transcripts.1 Approximately 98% of the human genome is known to be non‐coding DNA harboring a large number of intergenic and intronic miRNA genes.1, 2, 3 Intergenic miRNA genes are located in the non‐coding regions between genes, while intragenic miRNA genes, or intronic miRNA genes, are harbored within introns of their protein‐coding host genes. In normal cells, some of these endogenous RNA play crucial roles in many processes, such as proliferation, development, differentiation and apoptosis.4, 5, 6, 7 In cancer cells, several studies demonstrate the deregulation of miRNA expression and the genetic aberration of a few miRNA genes within amplified or deleted regions,8, 9, 10 showing that miRNA can contribute to the multistep processes of carcinogenesis as oncogenes or tumor suppressor genes (TSG).6, 11, 12, 13 Recently, several miRNA genes have also been demonstrated to have copy number variations (CNV), although whether CNV affect miRNA genes in human cancers remains unclear.14

DNA hypermethylation of CpG sites within CpG islands is known as an epigenetic aberration leading to the inactivation of tumor‐suppressive miRNA (TS‐miRNA) in cancer cells,15 in the same manner as that of many classical TSG.16 In fact, the expression of several miRNA is generally downregulated in malignant tissues compared with corresponding non‐malignant tissues. Recent studies, including our own, clearly demonstrate DNA methylation‐mediated downregulation of TS‐miRNA gene expression in various types of cancers.17, 18, 19, 20, 21, 22, 23, 24 Although the genomic distances between the 5′‐end of intergenic miRNA genes or host genes harboring intronic miRNA and their proximal CpG islands vary, these distances might provide more important information for the understanding of silencing of TS‐miRNA genes through DNA hypermethylation. However, few studies have focused on these genomic distances.

The many achievements in the field of TS‐miRNA discovery and in vitro/in vivo delivery technology may offer the possibility of new therapeutic approaches for cancer. Because one miRNA can target many messenger RNA (mRNA) of protein‐coding genes, the in vivo applications of miRNA for cancer therapies are considered better than those of short interfering RNA (siRNA). In addition, among miRNA‐based in vivo delivery approaches, including the use of DNA plasmids or viral vectors, miRNA replacement therapy using double‐stranded RNA (dsRNA) mimicking TS‐miRNA is one of the most promising, offering hope for new cancer therapies.25

We recently evaluated the genomic distribution of 1523 miRNA genes and their CpG islands, and then determined the genomic distance between these genes and CpG islands located within 10 kb upstream of miRNA gene using the miRBase database (Release 18: November 2011) and the UCSC Genome Browser on Human February 2009 Assembly (hg19). In this review, based on information from these databases, we provide insights into the relationship of the gene silencing of TS‐miRNA through aberrant DNA hypermethylation with the genomic distances between TS‐miRNA genes or protein‐coding host genes harboring intragenic miRNA and related CpG islands. We also discuss the potential of these TS‐miRNA as therapeutic agents for cancer.

Distribution of Micro RNA Genes and Their Related CpG Islands in the Human Genome

We examined the genomic distribution of the 1523 genes registered as human miRNA genes in the miRBase database (Release 18: November 2011), (Fig. 1A). Interestingly, 20.0% (304/1523) of these miRNA genes are located on chromosomes 14, 19 and X. Notably, some of these genes are concentrated at 14q32.31, 19q13.42 and Xq27.3, and lie within limited regions of 44, 122 and 33 kb, respectively. In 19q13.42 and Xq27.3, the chromosome 19 miRNA cluster (C19MC) and chromosome X miRNA cluster were revealed as primate‐specific.26 Some C19MC miRNA were described as expressed at a very low level in most human tissues.27 A correlation between their expression patterns and the methylation status of a distal CpG‐rich region located approximately 17.6 kb upstream of the C19MC region has also been demonstrated in gastric cancer cells.28

Figure 1.

Genomic feature of micro RNA (miRNA) genes. (A) Genomic distribution of 1523 human miRNA genes registered in the miRBase database (Release 18: November 2011). (B) Ratio of intergenic and intragenic miRNA genes in 1523 human miRNA genes. (C) Genomic distribution of 878 intergenic (left) and 645 intragenic (right) miRNA genes. These data were obtained from the miRBase database (http://www.mirbase.org/index.shtml) and UCSC Genome Browser on Human February 2009 Assembly (hg19) (http://genome.ucsc.edu/cgi-bin/hgGateway). In our database analyses, miRNA genes, which were located within introns of protein‐coding host genes and considered to be transcribed in the same direction as those of their host genes, were analyzed as intragenic miRNA genes.

Tumor‐specific downregulation of subsets of miRNA has been generally observed in various types of human cancer,11 suggesting that some of these miRNA act as TSG. Because the downregulation of some TS‐miRNA has been shown to be tightly linked to CpG island hypermethylation, the aberrant DNA hypermethylation of CpG islands located around TS‐miRNA genes, similar to various protein‐encoding TSG, has been recognized as one of the main epigenetic alterations in cancer cells.17, 18, 19, 20, 21, 22, 23, 24 The most recent study indicated 11.6% (122/1048) of miRNA to be epigenetically regulated in 23 cancer types,29 and 19.5% (26/133) of the 133 miRNA genes transcribing these 122 miRNA to have a CpG island within 5 kb upstream. In addition, 14.2% (19/133) of these miRNA genes were also demonstrated to reside within CpG islands. Few studies have examined the relationship of the transcriptional regulation of intergenic, intragenic miRNA genes or host genes harboring intronic miRNA genes through proximal CpG island hypermethylation with genomic distances between these individual genes and their related CpG islands. We discuss these relationships in the following sections.

Intergenic Tumor‐Suppressive Micro RNA Genes Silenced by CpG Island Hypermethylation in Cancer Cells

A recent study showed that RNA polymerase II (Pol II) promoters driving miRNA expression contained most of the features of the protein‐coding gene promoters and that intergenic and some intragenic miRNA were transcribed by RNA Pol II at a distance that could be as large as 40 kb from the miRNA genes.30 In addition, a computational approach demonstrated that 81.9% (59/72) of predicted promoters of intergenic miRNA genes (37 miRNA clusters among 46 miRNA clusters) contained or overlapped with at least one CpG island.31

In our database analyses, intergenic and intragenic miRNA genes made up 57.6% (878/1523) and 42.4% (645/1523), respectively, of human miRNA genes (Fig. 1B,C). Among intergenic miRNA genes examined in our database analyses, 15.3% (134/878) were located within 500 bp downstream of CpG islands (Fig. 2A). These 134 intergenic miRNA genes included 40 genes whose name contained a number lower than 700. Several research groups, including ours, have been investigating some of intergenic miRNA since the early days of miRNA study, and 47.5% (19/40) of these genes are well‐known TS‐miRNA (Table 1). Moreover, gene silencing of these known TS‐miRNA has already been reported to be related with DNA hypermethylation in several types of cancers.

Figure 2.

Genomic distances between intergenic and intragenic micro RNA (miRNA) genes or protein‐coding host genes and their related CpG islands. Each map indicates the relationship between miRNA genes or host genes and CpG islands located around these genes on the genome. All intragenic miRNA genes examined in our database analyses were considered to be transcribed in the same direction as those of their protein‐coding host genes. TSS, transcription start site. Each pie graph shows results of our database analyses for genomic distances between CpG islands and the 5′‐end of intergenic (A) or intragenic miRNA genes (B and D) or protein‐coding host genes (C). These data were obtained from the miRBase database (Release 18: November 2011) (http://www.mirbase.org/index.shtml) and UCSC Genome Browser on Human February 2009 Assembly (hg19) (http://genome.ucsc.edu/cgi-bin/hgGateway).

Table 1.

Intergenic and intragenic tumor‐suppressive micro RNA (TS‐miRNA) genes (miRNA number in a name < 700): These intergenic TS‐miRNA genes and protein‐coding host genes harboring intragenic TS‐miRNA genes are located within 500 bp downstream of CpG islands

miRNA genes (ID < 700)	Loci	Distances from 5′ end of miRNA genes to 3′ end of their CpG islands (bp)	Targets	References indicating tumor suppressive functions of miRNA	Host gene	Distances from TSS of host genes to 3′ end of their CpG islands (bp)	Distances from 5′ end of miRNA genes to 3′ end of CpG islands around TSS of host genes (bp)
Intergenic miRNA genes
hsa‐mir‐9‐3	15q26.1	0	NF‐kappaB1, Androgen receptor (AR)	Guo, 2009; Ostling, 2011	None	—	—
hsa‐mir‐34b	11q23.1	0	CDK6, c‐MYC, E2F3, MET, CCNE2, CDK4, CAV1, MYB, SFRS2, CREB	Corney, 2007t; Kozaki et al.19; Toyota et al.20	None	—	—
hsa‐mir‐34c	11q23.1	272	CDK6, c‐MYC, E2F3, MET, CCNE2, CDK4, CAV1, MYB, SFRS2, Androgen receptor (AR)	Corney, 2007; Toyota et al.20; Ostling, 2011	None	—	—
hsa‐mir‐92b	1q22	0	PRMT5	Pal, 2007	None	—	—
hsa‐mir‐124‐1	8p23.1	0	CDK6, C/EBPα, SMYD3, VIM, IQGAP1, IGFBP7	Silber, 2008; Furuta, 2010	None	—	—
hsa‐mir‐124‐3	20q13.33	0	CDK6, C/EBPα, SMYD3, VIM, IQGAP1, IGFBP7	Silber, 2008; Furuta, 2010	None	—	—
hsa‐mir‐127	14q32.2	0	BCL6	Saito et al.17	None	—	—
hsa‐mir‐129‐2	11p11.2	0	SOX4	Dyrskjot, 2009; Huang, 2009	None	—	—
hsa‐mir‐137	1p21.3	0	CDK6, MITF, Cdc42	Kozaki et al.19; Bemis, 2008; Liu et al.65	None	—	—
hsa‐mir‐148a	7p15.2	406	DNMT‐1, TGIF2	Braconi et al.60	None	—	—
hsa‐mir‐193a	17q11.2	0	E2F6, c‐kit	Kozaki et al.19; Gao, 2011	None	—	—
hsa‐mir‐196b	7p15.2	0	c‐myc	Bhatia, 2010	None	—	—
hsa‐mir‐203	14q32.33	0	ABCE1, ABL	Kozaki et al.19; Bueno, 2008; Furuta, 2010	None	—	—
hsa‐mir‐210	11p15.5	0	FGFRL1	Camps, 2008; Tsuchiya, 2011	None	—	—
hsa‐mir‐212	17p13.3	0	MeCP2, PED	Incoronato, 2010; Wada, 2010	None	—	—
hsa‐mir‐375	2q35	0	PDK1, 14‐3‐3zeta, RASD1, YAP	Tsukamoto, 2010; Liu, 2010; de Souza Rocha Simonini, 2010	None	—	—
hsa‐mir‐409	14q32.31	0	RDX	Zheng, 2011	None	—	—
hsa‐mir‐424	Xq26.3	0	PLAG1	Pallasch, 2009	None	—	—
hsa‐mir‐663	20p11.1	0	JunB, JunD, TGFβ1, p21/Waf1/Cip1	Pan, 2010; Tili, 2010; Tili, 2010	None	—	—
Intragenic miRNA genes
hsa‐mir‐1‐1	20q13.33	3	HDAC4, FoxP1, MET	Datta et al.44	C20orf166	0	3726
hsa‐mir‐15b	3q25.33	3498	BCL2	Cimmino, 2005; Xia, 2008	SMC4	0	3498
hsa‐mir‐16‐2	3q25.33	3655	BCL2	Cimmino, 2005; Xia, 2008	SMC4	0	3655
hsa‐mir‐23b	9q22.32	554	GLS, uPA, c‐Met	Gao, 2009; Salvi, 2009	C9orf3	0	358 318
hsa‐mir‐26a‐1	3p22.2	24 730	EZH2, cyclins D2, cyclin E2	Lu et al.48; Sander, 2008; Kota et al.47	CTDSPL	0	106 856
hsa‐mir‐26a‐2	12q14.1	20 365	EZH2, cyclins D2, cyclin E2	Sander, 2008; Kota et al.47	CTDSP2	0	20 365
hsa‐mir‐26b	2q35	1813	SLC7A11	Ma, 2010 #133; Liu et al.65	CTDSP1	0	1813
hsa‐mir‐27b	9q22.32	317	CYP1B1	Tsuchiya, 2006	C9orf3	0	358 555
hsa‐mir‐30c‐1	1p34.1	26 046	BCL2‐like 11 (BIM)	Garofalo, 2011	NFYC	0	64 981
hsa‐mir‐33a	22q13.2	8601	Pim‐1	Thomas, 2011	SREBF2	0	66 964
hsa‐mir‐95	4p16.1	65 175	SNX1	Huang, 2011	ABLIM2	0	152 620
hsa‐mir‐101‐2	9p24.1	45 517	EZH2, COX‐2, Mcl‐1	Varambally, 2008; Strillacci, 2009; Su, 2009	RCL1	0	56 744
hsa‐mir‐107	10q23.31	51 653	CDK6, PLAG1, HIF‐1β, PKCɛ	Lee, 2009; Pallasch, 2009; Yamakuchi, 2010; Datta, 2011	PANK1	281	51 653
hsa‐mir‐126	9q34.3	0	IRS‐1, SLC7A5, SOX2, VEGFA, PIK3R2	Tavazoie et al.34; Zhang et al.35	EGFL7	0	4049
hsa‐mir‐128‐1	2q21.3	76 097	E2F3a, Bmi‐1, NTRK3	Zhang, 2009; Godlewski, 2008; Guidi, 2010	R3HDM1	0	133 252
hsa‐mir‐133a‐2	20q13.33	10 216	FSCN1	Chiyomaru et al.55; Kano, 2010	C20orf166	0	14 332
hsa‐mir‐139	11q13.4	24 361	ROCK2	Wong, 2011	PDE2A	0	27 048
hsa‐mir‐149	2q37.3	0	Akt1, E2F1	Lin et al.27	GPC1	0	19 098
hsa‐mir‐152	17q21.32	0	DNMT1, E2F3, MET, Rictor	Huang et al.61; Das, 2010; Tsuruta et al.24	COPZ2	93	0
hsa‐mir‐153‐1	2q35	516	Bcl‐2, Mcl‐1	Xu, 2010	PTPRN	12	14 948
hsa‐mir‐153‐2	7q36.3	2006	Bcl‐2, Mcl‐1	Xu, 2010	PTPRN2	0	1 012 214
hsa‐mir‐185	22q11.21	11 562	Six1, RhoA, Cdc42, Androgen receptor (AR)	Imam, 2010; Liu et al.65; Ostling, 2011	C22orf25	0	11 562
hsa‐mir‐186	1p31.1	13 087	P2X7	Zhou, 2008	ZRANB2	189	13 087
hsa‐mir‐198	3q13.33	45 984	c‐MET	Tan, 2011	FSTL1	0	54 626
hsa‐mir‐218‐1	4p15.31	171 810	ECOP, IKK‐b, LASP1, PXN, Robo1, BIRC5, GJA1, Rictor	Martinez, 2008; Wu et al.56; Alajez et al.49; Uesugi et al.23	SLIT2	0	273 030
hsa‐mir‐218‐2	5q34	188 233	ECOP, IKK‐b, LASP1, PXN, Robo1, BIRC5, GJA1, Rictor	Martinez, 2008; Wu et al.56; Alajez et al.49; Uesugi et al.23	SLIT3	0	532 169
hsa‐mir‐326	11q13.4	15 994	Notch, MRP‐1/ABCC1	Kefas, 2009; Liang, 2010	ARRB1	0	15 994
hsa‐mir‐335	7q32.2	2841	SOX4	Tavazoie et al.34	MEST	0	2841
hsa‐mir‐338	17q25.3	2999	SMO	Huang, 2011	AATK	0	39 752
hsa‐mir‐340	5q35.3	55 894	MITF, c‐Met	Goswami, 2010; Wu, 2011	RNF130	0	55 894
hsa‐mir‐346	14q32.2	1172	Androgen receptor (AR)	Grady et al.43	GRID1	0	98 379
hsa‐mir‐449a	5q11.2	2333	HDAC‐1, E2F1, Androgen receptor (AR)	Noonan, 2009; Yang, 2009; Lize, 2009; Ostling, 2011	CDC20B	0	2333
hsa‐mir‐449b	5q11.2	2213	E2F1, Androgen receptor (AR)	Yang, 2009; Lize, 2009; Ostling, 2011	CDC20B	0	2213
hsa‐mir‐486	8p11.21	6311	OLFM4	Oh, 2011	ANK1	0	235 314
hsa‐mir‐488	1q25.2	134 811	Androgen receptor (AR)	Sikand, 2010	ASTN1	178	134 811
hsa‐mir‐489	7q21.3	90 754	PTPN11	Kikkawa, 2010	CALCR	43	90 754
hsa‐mir‐548d‐1	8q24.13	47 722	ERBB2	Chen, 2009; Heyn, 2011	ATAD2	0	47 722
hsa‐mir‐559	2p21	7558	ERBB2	Chen, 2009	EPCAM	343	7558
hsa‐mir‐562	2q37.1	178 223	EYA1	Drake, 2009	DIS3L2	0	210 527
hsa‐mir‐591	7q21.3	101 711	–	Shohet, 2011	SLC25A13	0	101 711
hsa‐mir‐593	7q32.1	22 182	PLK1	Ito, 2010	SND1	0	429 362
hsa‐mir‐634	17q24.2	47 907	Androgen receptor (AR)	Ostling, 2011	PRKCA	0	483 573

TSS, transcription start site. These data were obtained from the miRBase database (Release 18: November 2011) (http://www.mirbase.org/index.shtml), UCSC Genome Browser on Human February 2009 Assembly (hg19) (http://genome.ucsc.edu/cgi-bin/hgGateway) and PubMed (http://www.ncbi.nlm.nih.gov/pubmed). All intragenic miRNA genes, indicated in this table, are transcribed in the same direction as those of their host genes.

Pioneer studies of these intergenic TS‐miRNA demonstrated that miR‐127 was decreased by aberrant DNA methylation and histone modification in bladder cancer cells17 and that miR‐124 was inactivated by CpG island hypermethylation in several types of cancers.18 These studies suggest DNA hypermethylation to be an important molecular mechanism for downregulation of miRNA expression in cancers. Previously, we identified four intergenic TS‐miRNA silenced through DNA hypermethylation of CpG islands located within 500 bp upstream in oral squamous cell carcinoma (OSCC) and hepatocellular carcinoma (HCC), and also reported their targets (Table 1). We first identified miR‐137 and miR‐193a as an intergenic TS‐miRNA frequently silenced by tumor‐specific DNA hypermethylation in OSCC using expression‐based screening with a series of sequential analyses of expression profiles of 148 miRNA, DNA hypermethylation status of selected candidates, and their tumor‐suppressive activities in a panel of 18 OSCC cell lines and 11 primary tumors of OSCC with paired normal oral mucosa.19 Our study also revealed that miR‐137 and miR‐193 might induce cell cycle arrest at the G1‐S checkpoint and apoptosis, respectively, through direct binding to their target mRNA, CDK6 and E2F6, respectively, in OSCC cell lines. We performed methylation‐based screening, comparing methylation and expression status for 39 miRNA located at 43 loci containing CpG islands within 500 bp upstream of these miRNA genes in a panel of 19 HCC cell lines and 41 primary HCC tumors with corresponding non‐tumorous tissue, resulting in the discovery of hypermethylation‐mediated silencing of miR‐124 and miR‐203 genes as a relatively frequent molecular event in HCC.21 In this study, miR‐124 and miR‐203 were elucidated to exert cell growth‐inhibitory effects on HCC cell lines through the induction of cell cycle arrest at the G1‐S checkpoint and apoptosis, respectively, with the downregulation of protein expression of their targets, CDK6,VIM,SMYD3 and IQGAP1 or ABCE1, respectively. Based on these observations, we believe that the tumor‐specific DNA hypermethylation of CpG islands located immediately 5′‐upstream of intergenic miRNA genes is a useful landmark to explore novel TS‐miRNA silenced epigenetically in cancer cells, similar to classical TSG.

Intragenic Tumor‐Suppressive Micro RNA Genes and Their Host Genes Silenced by CpG Island Hypermethylation in Cancer Cells

Sizable regions of genomic DNA are recognized as introns harboring many intragenic miRNA genes, referred to as intronic miRNA genes. Approximately 37% of mammalian miRNA genes appear to be intronic miRNA genes.32 A recent computational approach demonstrated that 94.2% (49/52) of predicted promoter regions of intronic miRNA genes overlapped with promoters of their host genes.31 In addition, the majority of mammalian intronic miRNA genes are found to be frequently coexpressed with their protein‐coding host genes under the promoter‐driven regulation of the host gene, while nearly 26% of intronic miRNA genes are described to be transcribed from their own promoters.3, 30, 33 In our database analyses using the miRBase database (Release 18: November 2011), intragenic miRNA genes accounted for 42.4% (645/1523) of human miRNA genes (Fig. 1B). Notably, among these 645 intragenic miRNA genes, 13.0% (84/645) were located within 500 bp downstream of CpG islands (Fig. 2B). In contrast, the data on the genomic distance between the 5′‐end of protein‐coding host genes harboring intragenic miRNA genes and the 3 + +′‐end of CpG islands located at the 5′‐side of host genes show that 81.2% (477/581) of these host genes, containing 645 intragenic miRNA genes within their introns, were located within 500 bp of CpG islands (Fig. 2C), whereas only 6.4% (41/645) of intragenic miRNA genes were located within 500 bp downstream of these CpG islands (Fig. 2D). In addition, These 477 host genes harbored 534 intragenic miRNA genes, including 176 genes assigned a name with a number below 700. Among these 176 genes, no more than 23.9% (42/176) of intragenic miRNA genes have been identified as TS‐miRNA (Table 1).

Previous studies have shown tumor‐suppressive activities of miR‐126 and miR‐335 located within introns of the EGFL7 and MEST genes, respectively, in breast cancer,34 and report IRS‐1,35 SLC7A5,36 SOX2,37 VEGFA,38, 39 and PIK3R2 39 or SOX4 34 to be their targets, respectively. Coexpression of miR‐126 and EGFL7 is reported to be downregulated through histone modification and CpG island hypermethylation in the EGFL7 gene promoter in T24, HeLa and MCF7 cells and primary bladder and prostate tumors.40 Transcription of miR‐335 is also demonstrated to be coregulated with MEST by promoter hypermethylation in breast cancer cells.41 miR‐342 located within intron of the EVL gene is described as acting as an TS‐miRNA by targeting DNMT1 in colorectal cancer.42 CpG island hypermethylation upstream of EVL is indicated to suppress both EVL and miR‐342 expression.43 As regards to miR‐1‐1 located within the C20orf166 gene, miR‐1 is described as a TS‐miRNA targeting HDAC4,FOXP1 and MET, and is inactivated by DNA methylation of CpG island in HCC.44 However, these four host genes have never been examined for tumor‐suppressive activities.

Physiological or pathophysiological functions of host genes harboring intragenic TS‐miRNA mostly remain unclear, whereas miR‐26a and miR‐218 are unique intragenic TS‐miRNA located within introns of known TSG. miR‐26a‐1 located within an intron of CTDSPL/SCP3/HYA22/RBSP3 having a tumor‐suppressive function45 is reported to be silenced through CpG island hypermethylation.46 Cyclin D2 and E2 or EZH2 are described as miR‐26a targets in liver cancer47 and nasopharayngeal carcinoma (NPC),48 respectively. Regarding miR‐218, downregulation of its expression was demonstrated to be associated with DNA methylation of CpG islands at the 5′‐ends of SLIT2 and SLIT3 harboring miR‐218‐1 and miR‐218‐2, respectively, in NPC.49 The SLIT2 and SLIT3 genes are indicated to act as TSG through SLIT‐Robo signaling in breast cancer cell lines,50 and to be inactivated by their promoter hypermethylation in several types of cancers51, 52. BIRC5,49 ECOP,53 GJA1 49, IKK‐β,54 LASP1 55, PXN 56 and Robo1 57 are reported as targets of miR‐218. Recently, to explore TS‐miRNA having potential for miRNA replacement therapy, we performed function‐based screening using OSCC and endometrial cancer (EC) cell lines (Fig. 3A), and identified miR‐218 and miR‐152 as intragenic TS‐miRNA frequently silenced through tumor‐specific DNA hypermethylation in OSCC and EC, respectively.23, 24 COPZ2 is a protein‐coding host gene harboring the miR‐152 gene and has recently been shown to display no tumor‐suppressive activities, but, rather, to protect tumor cells from apoptosis induced by COPZ1‐knockdown.58 Moreover, our studies elucidated that Rictor was a novel direct target of these two TS‐miRNA. Rictor, together with mTOR, forms mTOR complex 2 (mTORC2), and the Rictor‐mTOR complex directly regulates the phosphorylation of Akt at Ser‐473, resulting in cell growth.59 In our study, miR‐218 was clearly demonstrated to act as a suppressor of the TOR‐Akt pathway, independently of the PI3K‐Akt pathway, in an OSCC cell line without a genetic alteration of EGFR,PIK3CA and PTEN (Fig. 3B).23 We also identified E2F3 and MET as direct targets of miR‐152 other than DNMT1, which had been reported previously.60, 61 A correlation between aberrant DNA methylation of CpG island of miR‐152 and a poor clinical outcome is reported in breast cancer62 and MLL‐rearranged acute lymphoblastic leukemia.63 These observations strongly support our notion that the tumor‐specific DNA hypermethylation of CpG islands located around protein‐coding host genes harboring intragenic miRNA genes, as well as intergenic miRNA genes, is a useful landmark to explore novel TS‐miRNA silenced epigenetically in cancer cells.

Figure 3.

Function‐based screening of tumor‐suppressive micro RNA (TS‐miRNA) for miRNA replacement therapy as a cancer treatment. (A) Strategy of our function‐based approach to the identification of epigenetically silenced TS‐miRNA in cancer cells. Previously, to identify novel TS‐miRNA having the great potential for miRNA replacement therapy, we performed function‐based screening combined with methylation and expression analyses in oral squamous cell carcinoma (OSCC) and endometrial cancer (EC) cell lines according to this strategy shown in this figure, resulting in identification of novel TS‐miRNA,miR‐218 and ‐152, respectively, directly targeting Rictor.23, 24 (B) A model summarizing the molecular mechanism of miR‐218,miR‐152 and their direct target Rictor in the TOR‐Akt signaling pathway. Our previous studies demonstrated that these TS‐miRNA acted as suppressors of the TOR‐Akt signaling pathway, independently of the PI3K‐Akt signaling pathway and that methylation‐mediated silencing of these TS‐miRNA might contribute to the pathogenesis of OSCC and EC through the activation of this signaling pathway. (C) Therapeutic effects of dsRNA mimicking miR‐152 (dsmiR‐152) or control non‐specific miRNA (dsNC) on tumor growth in vivo. Left panel: a schema indicating the protocol of in vivo analysis (upper) and bar graph showing effects of dsmiR‐152 on tumor growth in three SCID mice (lower). Right panel: photograph shows tumor‐bearing SCID mice (upper) and their subcutaneous tumors (lower) at the end of in vivo analysis. These findings strongly support the great potential of dsRNA mimicking miR‐152 to be applied to miRNA replacement therapy for cancers.

Future Perspectives on Tumor‐Suppressive Micro RNA Silenced by Tumor‐Specific DNA Hypermethylation in Cancer Research

Taking miRNA‐induced effects on normal cells into consideration, TS‐miRNA, the endogenous expression of which is sufficiently activated in normal cells and remarkably reduced in cancer cells, are assumed to have great potential for miRNA replacement therapy.25 The concept behind this therapy is a restoration of loss of function in cancer cells by exogenous expression of TS‐miRNA. A few TS‐miRNA, such as miR‐34a 64, 65 and let‐7,66 have already demonstrated therapeutic effects on tumor formation in vivo following the replacement of these miRNA using dsRNA mimicking their mature forms. Because dsRNA are unlike proteins, are substantially smaller than DNA plasmids or viral vectors, and have the ability to enter the cytoplasm of target cells and to be delivered systemically by technologies that have been used for siRNA,25 a dsRNA‐based approach may be better than other approaches for exogenous expression of miRNA in vivo. TS‐miRNA silenced through tumor‐specific DNA hypermethylation might be better suited as prime candidates for this therapy. Recently, we successfully showed for the first time that dsRNA mimicking miR‐152 administered with atelocollagen to SCID mice could suppress the in vivo growth of an EC cell line (Fig. 2C),24 leading us to consider the possibility of miRNA replacement therapy for cancer using dsRNA mimicking TS‐miRNA silenced by tumor‐specific DNA hypermethylation.

While the function‐based approach is a powerful tool for exploring the use of dsRNA with tumor‐suppressive effects, including TS‐miRNA and siRNA, as therapeutic agents for cancers, the tumor‐suppressive functions of miRNA eventually identified in our studies have been reexamined using two or three kinds of dsRNA purchased from independent companies to take into consideration the off‐target effects associated with dsRNA.23, 24 Although such effects have been known to complicate the interpretation of phenotypic effects in gene‐silencing experiments using siRNA,67 dsRNA mimicking miRNA might potentially cause these unwanted actions, similar to siRNA. These unpredictable target‐independent effects should be addressed during data interpretation in all dsRNA‐based studies related to functional genomics, drug target discovery and dsRNA‐therapeutics. In addition, these dsRNA have mainly been used at 1.0–50.0 nM in dsRNA‐based studies, and overexpression above their physiological concentrations might lead to toxic effects in normal cells with an accumulation of these exogenous dsRNA. In contrast, although atelocollagen24, 64 or lipid‐based delivery agents65, 66 have been used for in vivo dsRNA delivery, these systems are inadequate for clinical applications to fully restore downregulated miRNA in cancer cells. Therefore, the development of solutions attenuating the nonspecific off‐target effects associated with dsRNA mimicking TS‐miRNA and the advancement of dsRNA‐delivery technology may yield a new field of miRNA‐based cancer therapy.

Concluding Remarks

In this review, we focused on genomic distances between intergenic TS‐miRNA genes or protein‐coding host genes harboring intragenic miRNA genes and their related CpG islands, and illustrated how the tumor‐specific DNA hypermethylation of CpG islands located immediately 5′‐upstream of intergenic miRNA genes and host genes, as well as classical protein‐encoding TSG, might be a useful epigenetic marker for exploration of TS‐miRNA, cancer diagnosis and prognosis. Moreover, we also discussed the potential of miRNA replacement therapy for cancers using dsRNA mimicking TS‐miRNA silenced epigenetically in cancer cells. Further studies of molecular mechanisms of TS‐miRNA and significant advancement of dsRNA‐delivery technology are essential for the actualization of TS‐miRNA replacement therapy for several types of cancers.

Disclosure Statement

The authors declare no financial or commercial conflicts of interest.