Since their initial discovery,1–3 the field of microRNA (miRNA) has opened a new era in our understanding of small noncoding RNAs. By molecular cloning and bioinformatic approaches, miRNAs have been identified in viruses, plants and animals.4 The latest version of miRBase (release 10.0, August 2007) has annotated over 500 miRNA sequences in the human genome and this number is predicted to double as more miRNAs are awaiting experimental validation.5–7
miRNAs are predicted to negatively target up to one-third of human messenger RNAs (mRNAs).8 Because base-pairing with the mRNA 3′ untranslated region (3′ UTR) is generally imperfect, a single miRNA may target over 200 transcripts simultaneously.9 Therefore, regulation of miRNAs itself (e.g., at the epigenetic level) may well be a potent, albeit indirect, way to control simultaneously numerous genes. Here we review the recent work on miRNAs, with emphasis on their alterations and roles in human cancer.
miRNA biogenesis and silencing mechanism
miRNAs are endogenously produced RNA molecules ∼18–25 nucleotides (nt) long. With the exception of those within the Alu repeats transcribed by polymerase III (Pol III),10 most miRNA genes are derived from primary miRNA transcripts (pri-miRNAs) produced by Pol II and containing a 5′ cap and a poly(A) tail.11, 12 The pri-miRNA is cleaved within the nucleus by a multiprotein complex called Microprocessor, which is composed of the RNAse III enzyme Drosha and the double-stranded RNA-binding domain (dsRBD) protein DGCR8/Pasha,13–17 into a ∼70-nt hairpin precursor known as pre-miRNA.
Next, the pre-miRNA is exported into the cytoplasm by Exportin-5 via a Ran-GTP-dependent mechanism.18–20 The pre-miRNA is further cleaved into the mature ∼22-nt miRNA:miRNA* duplex by an RNAse III enzyme, Dicer, in association with its partners, TRBP/Loquacious and PACT in human cells.21, 22 Subsequently, an RNA-induced silencing complex called RISC is assembled with Argonaute (Ago) 2.23, 24 The miRNA strand is selectively incorporated into RISC25, 26 and guides the complex specifically to its mRNA targets through base-pairing interaction.
miRNAs downregulate the expression of their target genes in 2 modes depending on the complementarity between them. miRNAs with perfect or near-perfect complementarity to the target sequence induce the cleavage and degradation of the transcript initiated by deadenylation and decapping of the mRNA.27 However, most miRNAs bind imperfectly to their target sequence and rather repress the translational process, with the underlying molecular mechanisms studied intensively using cell-free in vitro systems.28 Recently, Wakiyama et al. established a cell-free extract system derived from human embryonic kidney (HEK) 293 cells and demonstrated that efficient miRNA-guided translational repression requires an m7G-cap as well as a poly(A) tail,29 consistent with a previous report using a rabbit reticulocyte lysate system.30 In addition, inhibition of translational initiation by affected ribosome recruitment to the mRNA and targeting of the mRNA cap structure was proven in a study utilizing extracts from mouse Krebs-2 ascites cells.31 The dilemma how miRNA ribonucleoprotein complexes (miRNPs) that are bound to the 3′ UTR of a target mRNA interfere with translational initiation was resolved by Kiriakidou et al.32 They identified a motif (MC) within the Mid domain of Ago proteins, which bears significant similarity to the m7G cap-binding domain of eIF4E, an essential translation initiation factor. In their model, the Ago proteins compete with eIF4E for cap binding and thus repress the translational initiation. Inhibition of translational initiation by miR-2 was similarly observed in a cell-free system from Drosophila embryos. Interestingly, miR-2 induced the formation of structures (heavier than 80S) as “pseudo-polysomes,” which might resemble cytoplasmic processing bodies (P bodies).33
Function and deregulation of miRNAs in cancer
Cancer is a complex genetic disease caused by abnormalities in gene structure and expression. Previous studies have heavily focused on the protein-coding genes; however, accumulating evidence is revealing an important role of miRNAs in cancer. Although the function of only a handful of miRNAs has been characterized, miRNAs are clearly shown to control cell proliferation, differentiation and apoptosis.34 Therefore, disturbance of miRNA expression and function may contribute to the initiation and maintenance of tumors. Such miRNAs have been referred to as “oncomirs” and they can serve as both tumor suppressors and oncogenes.8 The multiple mechanisms altering miRNA expression and/or function in human cancer are summarized in Figure 1.
The first indication of miRNAs as tumor suppressors came from a report by Calin et al. when they found that patients diagnosed with B-cell chronic lymphocytic leukemia (B-CLL) have frequent deletions or downregulation of mir-15a and mir-16-1 located at chromosome 13q14.3.35 A follow-up study demonstrated that miR-15a and miR-16-1 negatively regulate the antiapoptotic protein BCL2 at a posttranslational level,36 and their function in leukaemogenesis and lymphomagenesis was supported by further studies.37–39 Some other miRNAs have also been shown to function as tumor suppressor genes, including the let-7 family which negatively regulates Ras40; the mir-17-92 cluster that breaks the positive-feedback loop between MYC and E2F1 in cellular proliferation41; as well as mir-143 and mir-145, which exhibit decreased abundance in colon cancer and various cancer cell lines.8
Other miRNAs can function as oncogenes. Indeed, the above mentioned mir-17-92 cluster was found by He et al. to be upregulated in 65% of B-cell lymphoma and its overexpression expedited the development of malignant lymphomas in a transplantation mouse model.42 Additionally, upregulation of mir-21 has been reported in glioblastomas43 and breast cancer,44 where it exerts an antiapoptotic function. mir-155 is remarkably overexpressed and linked to tumorigenesis in pediatric Burkitt, Hodgkin, primary mediastinal and diffuse large-B-cell lymphomas45–48 and breast cancer,44 likely in cooperation with MYC. Lastly, mir-372 and mir-373 have been implicated as oncogenes in testicular germ cell tumors.49
The causes of miRNA deregulation in human cancers are emerging by studies using genomic, molecular and functional approaches. First, it has been reported that more than half of the known human miRNAs are located at fragile sites (FRA), as well as in minimal regions of loss of heterozygosity (LOH), minimal regions of amplification (minimal amplicons) and common breakpoint regions, which are particular genomic regions prone to alteration in cancer cells.50 Furthermore, using a high-resolution (∼1 Mbp) array-based comparative genomic hybridization (aCGH), we analyzed 283 known miRNAs and found that a large proportion of miRNA gene-containing genomic loci exhibit DNA copy number alterations in ovarian cancer (37.1%), breast cancer (72.8%) and melanoma (85.9%),51 suggesting that miRNA deregulation stemmed at the genomic level may be frequent.
Second, abnormalities of the protein machinery involved in miRNA biogenesis may affect the global miRNA expression and/or processing. In a study of 67 non-small cell lung cancer (NSCLC) cases, lower Dicer1 expression levels were significantly associated with poor tumor differentiation and shortened postoperative survival.52 Recently, Lu et al. performed a large-scale profiling of 217 mammalian miRNAs from 334 samples using a novel bead-based flow cytometric method and found a global decrease of mature miRNA expression in human cancers.38 Kumar et al. expressed short hairpin RNAs (shRNAs) targeting the 3 critical components of the miRNA processing machinery, Drosha, Dgcr8 and Dicer1, in mouse lung adenocarcinoma LKR13 cells and observed a substantial decrease in the steady-state miRNA levels and a more pronounced transformed phenotype.53 More interestingly, miRNA processing-impaired cells formed tumors with accelerated kinetics in a mouse xenograft model and the conditional deletion of Dicer1 enhanced tumor development in a K-Ras-induced mouse model of lung cancer, indicating that abrogation of global miRNA processing promotes tumurigenesis.
Importantly, the epigenetic regulation of miRNA expression has been recently discovered. Since this particular mode may provide a more complex and sophisticated regulation of miRNAs, particularly in cancer, we will focus on it for the rest of this review.
The epigenetics of cancer
Epigenetics is defined as mitotically and/or meiotically heritable changes in gene expression that are not accompanied by changes in DNA sequence. Epigenetic regulation generally falls into 2 categories: DNA methylation and histone modifications.
Reversible DNA methylation regulates the normal development of mammals. It occurs almost exclusively within 5′-CpG-3′ dinucleotides by adding a methyl group to the 5 position of the cytosine by DNA methyltransferases (DNMTs). There are 3 major forms of DNMTs: DNMT1 that maintains the methylation patterns throughout each cell division and DNMT3A and 3B that catalyze de novo methylation during early development or in germ cells.54 Vertebrate genome is generally depleted in the dinucleotide CpG; however, some regions contain a high frequency of CpGs called CpG islands.55 29,000 CpG islands are predicted in the human genome by computational analysis,56, 57 with many positioned at the 5′ end of genes.
Aberrant DNA methylation occurs in most cancers.58 The most well-known and intensively studied epigenetic abnormality in cancer is the hypermethylation of CpG islands in the promoter region and resulted silencing of tumor suppressor genes. For example, the hypermethylation of p16ink4A in lung cancer,59VHL in renal carcinoma60 and BRCA1 in breast cancer61 are all been reported to contribute to tumorigenesis. Another form of abnormality is loss of imprinting (LOI) in cancers. LOI for IGF2 gene leads to biallelic expression and overproduction of this potent growth factor and has been related to increased risk of colon cancer.62
Histones comprise the protein backbone of chromatin. They can undergo numerous modifications, including lysine acetylation, lysine and arginine methylation, serine and theronine phosphorylation, glutamic acid ADP-ribosylation, lysine ubiquitination and sumoylation.62 Such covalent modifications can directly alter the physical properties of the chromatin fiber, leading to changes in higher-order structures or to recruit or stabilize the localization of specific binding partners to chromatin.63 Furthermore, there exist noncovalent mechanisms such as nucleosome remodeling and the incorporation of specialized histone variants (e.g., H3.3, H2A.Z) to provide the cell with additional tools for introducing variation into the chromatin template.63
Even a subtle alteration of the chromatin structure can affect gene expression. It is therefore conceivable that aberrant histone modification constitutes an important mechanism in cancer development. For example, histone deacetylases (HDACs) remove the acetyl groups from histone, allowing compacted chromatin to reform and repress transcription of the gene. Dysregulation of those enzymes has been related to cancer initiation and progression. It has been reported that overexpression of HDAC1 represses the tumor suppressors p53 and VHL but induces the hypoxia-responsive genes HIF-1α and VEGF and increases angiogenesis.64
The 2 modes of epigenetic regulation, methylation and histone modification, can actually cooperate tightly in their function. For example, select type I and II HDACs are associated both with complexes involving each of the DNMTs and with a family of methyl CpG-binding domain proteins (MBDs) that interpret and mediate the transcriptional repressive activities of DNA methylation.62 And effective transcriptional reactivation of certain genes cannot be achieved unless both DNA methylation and HDAC inhibition are eliminated.
The cancer epigenome
With the sequencing of the human genome completed, investigators now seek a comprehensive view of the epigenetic changes that determine how genetic information is made manifest across an incredibly varied background of developmental stages, tissue types and disease states.65 There comes the term, epigenome, which refers to the complete description of epigenetic changes across the genome. It includes but is not limited to the large-scale bisulfite sequencing of the methylation status of CpG and non-CpG islands called mythylome, landscape of activating histone modifications and genomewide study of the targets of Polycomb-group (PcG) complexes.65
In addition to the local epigenetic regulation of discrete genes as mentioned earlier, global alterations have been observed in the cancer genome. For example, an overall decrease in the 5′mC content of cancer genomes has been reported.58 In a recent study of epigenetic silencing in colorectal cancer, Frigola et al. found common repression of the entire 4-Mb band of chromosome 2q.14.2, associated with global methylation of histone H3-lysine9.66 DNA hypermethylation within the repressed genomic neighborhood was localized to 3 separate enriched CpG island “suburbs,” with the largest hypermethylated suburb spanning 1 Mb. Both DNA-methylated and neighboring unmethylated genes can be coordinately suppressed by global changes in histone modification. Therefore, advanced study of epigenetic alterations at a global level should contribute to our appreciation of the mechanisms involved.
Epigenetic alterations of miRNAs in cancer
A close look at the epigenetic regulation on miRNAs was derived from an extensive expression profiling of miRNAs in T24 human bladder cancer cells and LD419 human normal fibroblast cells treated with some chromatin-modifying drugs.67 The study revealed that 17 of 313 human miRNAs examined were upregulated by simultaneous treatment with the DNA-demethylating agent, 5-Aza-2′-deoxycytidine (5-Aza-CdR) and the histone deacetylase inhibitor 4-phenylbutyric acid (PBA). Interestingly, these upregulated miRNAs were quite different between these 2 cell lines, indicating that DNA methylation status and chromatin structure around miRNA genes are different between cancer and normal cells, although tissue-specific expression may not be completely ruled out.
Among them, mir-127 was focused on because it was dramatically upregulated (49-fold) in T24 cells following the 5-Aza-CdR and PBA treatment, and the gene is embedded within a CpG island that is methylated in most tissues. Further examination revealed that mir-127 was silenced in various cancer cells but was expressed in normal fibroblasts. Although located within a cluster with several other miRNAs on chromosome 14q32.31, Northern blot indicated that miR-127 was induced from its own promoter, which was identified by 5′RACE. By bisulfite genomic sequencing, the promoter region of mir-127 was found heavily methylated in LD419 fibroblasts (97%), and the methylation level was decreased to 81% after 5-Aza-CdR and PBA treatment. On the other hand, the methylation level of the same region in T24 cancer cells was only 60%, and it was further decreased to 41% after treatment. In addition, chromatin immunoprecipitation (ChIP) assay revealed that acetylated histone H3 and methylated histone H3-lysine 4, both associated with open chromatin structure and active gene expression, were increased in the mir-127 promoter region in T24 cells following treatment. mir-127 is downregulated in various primary human tumors. A candidate target of miR-127, the proto-oncogene BCL6, was translationally suppressed after miR-127 upregulation by 5-Aza-CdR and PBA treatment, suggesting that DNA demethylation and histone deacetylase inhibition may activate expression of miRNAs acting as tumor suppressors.67
The effect of methylation on miRNA expression was further confirmed by several groups using the same genetic model, Dnmt1 and Dnmt3b knockout HCT116 colorectal cancer cells. Lujambio et al. performed microarray profiling of 320 human miRNAs and found that 18 miRNAs were upregulated by >3-fold in the Dnmt1/Dnmt3b double knockout HCT116 cells.68 One of the main targets was mir-124a, which underwent transcriptional inactivation by CpG island hypermethylation in human tumors from different cell types. Interestingly, they also functionally linked the epigenetic loss of mir-124a with the activation of cyclin D kinase 6, a bona fide oncogenic factor, and the phosphorylation of Rb, a tumor suppressor gene. Indeed, CpG island hypermethylation of tumor suppressor miRNAs has been proposed as an important mechanism in tumorigenesis.69
Brueckner et al. noticed that the human let-7a-3 gene on chromosome 22q13.31 is associated with a well-defined CpG island.70 Also using Dnmt1 and Dnmt3b knockout HCT116 cells, they proposed that the miRNA gene methylation is cooperatively maintained by both methylases. The let-7a-3 gene was heavily methylated in normal human tissues but hypomethylated in some lung adenocarcinomas. Hypomethylation facilitated epigenetic reactivation of the gene and the elevated expression of let-7a-3 in a human lung cancer cell line induced enhanced tumor phenotypes and oncogenic changes in transcription profiles. Their results thus identified let-7a-3 as an epigenetically regulated miRNA gene with oncogenic function and suggested that aberrant miRNA gene methylation might contribute to human cancer.
Han et al. found that the expression of about 10% miRNAs was regulated by DNA methylation in Dnmt1/Dnmt3b double knockout HCT116 cells.71 In addition, neither 5-Aza-CdR treatment nor deletion of Dnmt1 alone recapitulated miRNA expression profile seen in the double knockout cell line, suggesting that miRNA expression was tightly controlled by DNA methylation and partial methylation reduction was not sufficient for miRNA reexpression. They also found that Hoxa3 and Hoxd10 were putative targets of miR-10a, one of the differentially expressed miRNAs that is located in the Hox gene cluster.
In addition to methylation-mediated regulation, histone modification alone may also regulate the miRNA expression. Scott et al. reported rapid alteration of miRNA levels in response to the potent hydroxamic acid HDAC inhibitor LAQ824 in the breast cancer cell line SKBr3.72 Using miRNA microarray analysis, they measured significant changes in 40% of the >60 different miRNA species expressed in SKBr3 cells, with 22 miRNA species downregulated and 5 miRNAs upregulated.
The above evidence raises important questions on the mechanisms that are responsible for the epigenetic regulation of miRNAs. Interestingly, interleukin-6 (IL-6) is overexpressed and contributes to tumor cell growth in cholangiocarcinoma. Meng et al. reported that overexpression of IL-6 increased the DNMT1 and expression profiling revealed that 7 miRNAs were significantly downregulated by IL-6 overexpression (<0.4-fold) and upregulated (>2-fold) by 5-Aza-CdR.73 One of these, mir-370, is embedded in a CpG island and its upregulation by 5-Aza-CdR was exclusively observed in malignant cells. Oncogene mitogen-activated protein kinase kinase kinase 8 (MAP3K8), a potential target of miR-370, was decreased by 5-Aza-CdR in cholangiocarcinoma cells. Furthermore, overexpression of IL-6 reduced miR-370 expression and reinstated MAP3K8 expression in vitro as well as in tumor cell xenografts in vivo. Thus, the modulation of tumor development through miRNAs may be mediated via certain factors, such as IL-6, in an epigenetic manner.
It is noteworthy that the epigenetic regulation of miRNAs, albeit potent, may well be cell-type specific. For instance, although the miR-127 expression was significantly upregulated by 5-Aza-CdR and PBA treatment in several cell lines, including HCT116, HeLa, NCCIT embryonic carcimoma, Ramos lymphoma and CFPAC-1 pancreatic carcinoma,67 such effect was absent in MCF7 breast carcinoma and CALU-1 lung carcinoma cells. In addition, Diederichs and Haber have reported the absence of significant alterations in miRNA expression patterns following either DNA demethylation or HDAC inhibitor treatment in A549 lung cancer cells.74
Exploiting miRNA epigenetics for cancer
As discussed earlier, both miRNAs and epigenetics play an active role in cancer biology. The complexity of miRNA epigenetic regulation reveals novel and exciting dimensions of cancer biology. Indeed, miRNA epigenetic profiling could be used to develop diagnostic or prognostic biomarkers.
One exciting example of miRNA profiling as a diagnostic tool came from the study by Lu et al. Using a bead-based flow cytometric profiling method, they observed differential expression of nearly all miRNAs across a large panel of samples representing diverse human tissues and tumor types. Unsupervised hierarchical clustering of the samples using miRNA profiles revealed developmental origin of the tissues, i.e., epithelial or hematopoietic. Moreover, miRNAs partitioned tumors within a single lineage, in that bone marrow samples obtained from patients with acute lymphoblastic leukaemia (ALL) were grouped into 3 major branches consistent with their differential genetic lesion. Most strikingly, they successfully employed miRNA profiling to address a very challenging diagnostic problem, that of tumors of histologically uncertain cellular origin. It is estimated that 2–4% of all cancer diagnoses represent cancers of unknown origin or diagnostic uncertainty. By analyzing 17 poorly differentiated tumors with nondiagnostic histological appearance, they showed that the miRNA-based classifier was much more accurate at establishing the correct diagnosis of the samples than the mRNA classifier. Their work thus demonstrated the feasibility and utility of monitoring the expression of miRNAs in human cancer.38
Use of miRNA profiling as a prognostic tool has also been shown, e.g., in lung cancer. In unsupervised hierarchical analysis of 143 lung cancer cases, Takamizawa et al. classified them into 2 major groups according to let-7 expression. Significantly shorter survival of patients was related to reduce let-7 expression. Multivariate COX regression analysis showed this prognostic impact to be independent of disease stage.75 Their study was potentiated by Yanaihara et al. when they correlated high miR-155 and low let-7a-2 expression with poor survival of 104 lung cancer patients from US institutions by univariate analysis of miRNA profiling.76
Epigenomic profiles have been used as markers for cancer diagnosis and prognosis. The promoter CpG islands of a number of tumor suppressor genes have been shown to be hypermethylated in cancer.60, 77–79 One of the first global analyses to define methylation status in cancer was carried out by Costello et al. on 1,184 unselected CpG islands in each of 98 primary human tumors using restriction landmark genomic scanning (RLGS).80 They estimated that an average of 600 CpG islands of the 45,000 in the genome were aberrantly methylated in tumors, including early stage tumors. Shared patterns of CpG-island methylation were also identified within each tumor type, together with patterns and targets that displayed distinct tumor-type specificity. In some cases, e.g., primary prostate carcinomas, hypermethylation of a single gene locus, glutathione S-transferase pi (GSTP1), is found in >90% tumors but not in normal prostatic tissue or other normal tissues.81, 82 Hypermethylation of death-associated protein kinase (DAPK), p16ink4A and epithelial membrane protein 3 (EMP3) have been linked to aggressive tumor phenotypes in lung, colorectal and brain cancer patients.79 Finally, changes in global levels of individual histone modifications, e.g., acetylation and dimethylation of 5 residues in histones H3 and H4, are predictive of clinical outcome of prostate cancer.83 It will be thus important to explore methylation patterns of miRNAs and their potential for cancer classification in the near future.
miRNA and epigenetics as therapeutic targets
Since miRNAs could function as oncogenes or tumor suppressor genes, they provide a therapeutic target for cancer treatment. Modified antisense oligonucleotides (ASOs) complementary to miRNAs are used by many groups to inhibit miRNAs with oncogenic properties. For example, 2 types of modified ASOs have been used successfully to inhibit the liver-specific miR-122 in mice.84, 85 Krutzfeldt et al. intravenously (i.v.) administered a cholesterol-conjugated 2′-O-methyl modified ASO named “antagomirs” into mice and observed a marked reduction of corresponding miRNA levels in a variety of tissues. The silencing of endogenous miRNAs by this novel method was specific, efficient and long-lasting. Administration of 80 mg/kg antagomirs to normal mice resulted in increased levels of miR-122 target gene mRNAs in the liver, as well as successful reduction of plasma cholesterol and an apparent degradation of the miRNA as determined by northern blotting.84 Esau et al. achieved the same results by inhibiting miR-122 in normal mice using an unconjugated 2′-O-methoxyethyl phosphorothioate modified ASO, delivered intraperitoneally (i.p.) twice weekly in saline.85
To supplement and/or enhance the function of tumor suppressor miRNAs, enforced expression of a short hairpin RNA (shRNA) from a polymerase II or III promoter in a nonviral or viral vector, which can be further processed into mature miRNAs, has been tested.86 In addition, in vivo delivery of double-stranded miRNA mimics has been reported.86 However, concerns are recently raised about the gene therapy approach for shRNA expression. For example, Grimm et al. showed that overexpression of shRNA from AAV vectors in mice could saturate the miRNA pathway and cause severe liver toxicity.87 Therefore, precautions have to be applied when considering the practical application of those strategies in vivo.
An intriguing feature about epigenetic changes is that, unlike genetic changes, they are potentially reversible by certain drugs. For example, it has been known for years that gene silencing from DNA methylation can be reversed by demethylating drugs. Such drugs are not only available for cell culture based experiments, but have also been used in patients with a significant antitumor activity. For example, US Food and Drug Administration (FDA) has approved the use of 2 such agents, 5-azacytidine and 5-aza-2′-deoxycitidine, as selective treatment for a preleukaemic disease, myelodisplastic syndrome.88
With the demonstration that HDAC inhibitors have antitumor potential, they have begun to constitute another promising group of agents for the epigenetic therapy of cancer. One of the main therapeutic mechanisms lies in their transcriptional reactivation of “dormant” tumor-suppressor genes, such as p21WAF1.89 Indeed, FDA has recently approved the first drug of this type, suberoylanilide hydroxamic acid (SAHA), for the treatment of cutaneous T-cell lymphoma.88, 90 The development of new HDAC inhibitor drugs is now being actively pursued by pharmaceutical companies.91
Given the emerging role of epigenetic miRNA regulation mechanisms in cancer, it is conceivable that therapeutic targeting of these mechanisms may benefit cancer therapy greatly in the foreseeable future.