Cancer is a complex genetic disease in which oncogene amplication and/or tumor suppressor gene mutation leads to step-wise deregulation of cell proliferation and apoptosis. Evidence is emerging that particular microRNAs (miRNAs) may play a role in human cancer pathogenesis. MiRNAs are endogenous ∼22 nucleotide (nt) noncoding RNAs (ncRNAs), which can play important regulatory roles in animals and plants by pairing to the messenger RNAs (mRNAs) of target genes and specifying mRNA cleavage or repression of protein synthesis.1, 2, 3, 4 Recent evidence indicates that miRNAs exhibit important regulatory roles in development, cell proliferation, cell survival, and apoptosis, and thus play a central role in gene regulation in health and disease.5, 6, 7 The miRNAs that were first discovered, lin-48, 9 and let-7,10, 11 were identified genetically in Caenorhabditis elegans (C. elegans); it became clear that miRNAs represent novel means of regulating developmental timing in C. elegans. Subsequently, hundreds of nonprotein-coding miRNAs were identified across species, showing that they are highly conserved.6, 12, 13, 14
In practice, when a small RNA is discovered by cDNA cloning, typically, it can be classified as a miRNA if it meets 2 of the following 3 criteria: (i) its expression should be confirmed by Northern blotting, RT-PCR or RNase protection assay, and so on (i.e. expression criterion); (ii) the sequence should be present in 1 arm of the hairpin precursor, which lacks large internal loops or bulges (i.e. structure criterion); (iii) the sequences should be phylogenetically conserved (i.e. conservation criterion).15 If a small RNA is found by methods other than cDNA cloning, its expression must be demonstrated along with the precursor structure and conservation.15 To date, the public miRNA database, MiRBase (http://microrna.sanger.ac.uk; release 8.1), has collected 462 human and 340 mouse miRNA sequences. Both cloning and bioinformatic approaches have shown that the human genome contains a much larger number of miRNAs than previously appreciated.16 It has been suggested that the total number might be above 800 or even 1,000.5 MiRNA genes can be located in the introns and/or exons of protein-coding genes or in the intergenic regions between protein-coding genes. They can form polycistronic clusters or exist individually.1, 13, 17
Biochemical approaches have provided a basic understanding of the molecular details of miRNA biogenesis.15 As shown in Figure 1, a primary transcript RNA (pri-miRNA) transcribed from a miRNA gene by RNA polymerase II is first processed into a stem-loop structure of about 70–100 nt (precursor miRNA; pre-miRNA) by a double-strand (ds)-RNA-specific ribonuclease, Drosha, with the help of its binding partner DGCR8.18 These pre-miRNAs are transported into the cytoplasm via an Exportin-5-RanGTP dependent mechanism.19, 20, 21 In the cytoplasm, they are digested by a second, dsRNA-specific ribonuclease called Dicer with the help of TRBP and AGO2.22, 23, 24 The released 17–25 nt mature miRNA is bound by a complex called miRNA-associated RNA-induced silencing complex (miRISC). Two mechanisms for miRNA action are currently known: mRNA cleavage and translational repression of mRNA without RNA cleavage. In animals, the complex-bound, single-stranded miRNAs binds specific target mRNAs through the sequence that is significantly, though not completely complementary to the target mRNAs.25 By a mechanism that is not fully characterized, the bound mRNA remains untranslated, resulting in reduced expression of the corresponding genes. A single miRNA could regulate multiple target genes, while a single gene could be targeted by multiple miRNAs, suggesting that the miRNAome and mRNAome interaction is a complicated network. For more details, see reviews.1, 6, 15, 17, 26, 27, 28
Several excellent review papers focusing on the link between miRNAs and cancer, e.g., Refs.7, 26, 27 and 29, 30, 31, 32 have been published in recent years. Particularly, Esquela-Kerscher and Slack's review26 is a very elegant and comprehensive one on this topic, in which the connection of miRNAs and cancer has been discussed in depth. In the present review, we are trying to describe a more general and up-to-date picture regarding the association of miRNAs with cancer, as well as miRNA regulation, signaling pathways and target prediction and validation. Some recent advancements in this field that have not been included in the previous reviews are also discussed here.
Deregulation of miRNA expression in cancer
Rapidly accumulating evidence has revealed that miRNAs are associated with cancer because of deregulation.7, 26, 29, 30, 33 Genome-wide studies demonstrated that miRNA genes are frequently located at cancer-associated genomic regions or in fragile sites, as well as in the minimal regions of loss of heterozygosity, in the minimal regions of amplifications or in the common breakpoint regions. This suggests that miRNA might be a new class of genes involved in human tumorigenesis. Downregulation of miRNAs is observed frequently in cancer samples. For example, miR-15 and miR-16 at 13q14 were deleted and/or downregulated in about 68% of B-cell chronic lymphocytic leukemia (CLL) cases.34, 35 MiR-143 and miR-145 consistently display reduced steady-state levels of the mature miRNAs in colon cancer36 and in lung cancer37 compared to the levels in normal tissues. Three research groups37, 38, 39 subsequently reported the reduced expression of the miRNA let-7 in human lung cancers relative to normal tissues, which indicates a poor prognosis. Lu et al.40 found that the absolute expression levels of many miRNAs were reduced significantly in tumors. Thus, the frequently observed downregulation of miRNAs in cancer tissues relative to normal tissues suggests that many miRNAs may function as tumor suppressor genes. Because the abrogation of differentiation is a hallmark of cancers, the low-level expression of such miRNAs may be related to the state of poor differentiation of cancer cells. If so, the expression of these miRNAs would be upregulated when the cancer cells are induced to differentiation. Indeed, Sempere et al. showed that 19 miRNAs existing in brain tissue (including lin-4 and let-7 orthologs) were downregulated in both human and mouse embryonal carcinoma cells, and their expression is upregulated during neuronal differentiation induced by retinoic acids.41 Similarly, Lu et al. treated the myeloid leukemia cell line HL-60 with all-trans retinoic acid for 5 days, and miRNA profiling revealed the induction of 59 miRNAs being coincident with differentiation.40 Therefore, a general downregulation of miRNAs in tumors and an upregulation of these same miRNAs during differentiation, indicate that miRNA expression reflects the state of cellular differentiation. Thus, Lu et al.40 speculated that abnormalities in miRNA expression might similarly contribute to the generation or maintenance of “cancer stem cells”, which was proposed to be responsible for cancerous growth in both leukemias and solid tumours.42 In addition, restoring the expression of such miRNAs in cancer cells probably could promote differentiation and induce malignant cells into a benign or normal state.
On the other hand, upregulation of miRNAs is also observed frequently. For example, over-expression of precursor miR-155/BIC RNA was reported in children with Burkitt lymphoma,43 human B-cell lymphomas44 and Hodgkin lymphoma.45 Further, Yanaihara et al. found that high expression of miR-155 has a significantly worse prognostic impact on patients with lung adenocarcinoma as an independent risk factor, and therefore could serve as a marker for survival. Although miR-155 is overexpressed in several types of human cancers, its biological function remains uncertain. However, a previous study has shown that BIC (host gene of miR-155) is implicated as a collaborator with c-myc in an avian lymphoma model system.37 A polycistronic miRNA cluster, mir-17-92 (including 7 miRNAs: miR-17-5p, miR-17-3p, miR-18a, miR-19a, miR-20a, miR-19b-1 and mir-92-1), residing in the C13orf25 gene locus on chromosome 13, was upregulated in 65% of B-cell lymphoma samples.47, 48 The increased expression of these miRNAs correlates with the presence of the 13q31 amplicon.47 The high-level amplification of 13q31 has been observed in hematological and other solid neoplasms,48, 49 including diffuse large B-cell lymphoma,50 primary cutaneous B-cell lymphoma,51 follicular lymphoma,52 mantel cell lymphoma,53 nasal-type natural killer/T-cell lymphoma,54 glioma,49 nonsmall cell lung cancer,49 bladder cancer,49 squamous-cell carcinoma of the head and neck,49 peripheral nerve sheath tumor,55 malignant fibrous histiocytoma,56 alveolar rhabdomyosarcoma57 and liposarcoma.58 The oncogenic function of the mir-17-92 cluster is further validated by a number of pieces of evidence as follows. MiR-19a and miR-92-1 are overexpressed in cells from patients with B-cell CLL.35 MiR-17-5p is overexpressed in breast, colon, lung, pancreas and prostate tumors, while miR-20a is overexpressed in colon, pancreas and prostate tumors.59 MiR-17-3p is overexpressed in lung adenocarcinoma and is related to the patient's survival.37 MiRNAs from the mir-17-92 cluster are overexpressed in lung cancers and enhance cell proliferation.60 Moreover, introduction of the expression construct of the mir-17-92 cluster, but not the putative open reading frame of C13orf25, could enhance lung cancer cell growth.60 Most importantly, to test the hypothesis directly that increased expression of this cluster contributes to cancer formation, He et al.47 overexpressed the mir-17-19b-1 cluster (the vertebrate-specific portion of the mir-17-92 cluster) using a mouse model of human B-cell lymphoma. Hematopoietic stem cells (HSCs) derived from fetal livers of c-myc transgenic mice generate B-cell lymphomas by 4–6 months of age when transplanted into lethally irradiated recipients.61 Irradiated animals that received HSCs overexpressing both Myc and the mir-17-19b-1 cluster developed malignant lymphomas faster (∼51 days) than those animals that received HSCs expressing Myc alone, or in combination with either 96 unrelated, single miRNAs or any individual miRNA of the mir-17-19b-1 cluster (3–6 months).47 More recently, Volinia et al.59 identified a large number of overexpressed miRNAs, including miR-17-5p, miR-20a, miR-21, miR-92, miR-106a and miR-155, in solid tumor samples. Therefore, overexpression of miRNAs appears to play an important role in tumorigenesis, with the miRNAs functioning directly or indirectly as oncomiRNAs. The aberrant expression of miRNAs in tumor cells appears to be strongly associated with cancer development.
A current challenge is to reveal the mechanism of regulation of miRNA gene expression itself, i.e. why and how the miRNAs are deregulated. It is very important for us to understand further, the mechanism of the development of miRNA-associated cancer. In some cases, the cause of miRNA deregulation is clear. For example, in the case of the mir-17-92 cluster, it is now very clear that the increase expression of these miRNAs correlates with the presence of the 13q31 amplicon.47, 48 Similarly, the downregulation of miR-15 and miR-16 at 13q14 observed in about 68% of B-cell CLL cases is largely due to a deletion at 13q14.34, 35 However, in most of cases, there is no clue as to why and how the miRNAs are deregulated. Given that miRNAs, like protein-coding genes, are transcribed by RNA polymerase II,62 and that many miRNAs are located within the introns of protein coding genes,63 the regulation of miRNA gene expression is probably similar to the regulation of protein-coding gene expression. For example, CpG-island methylation and/or histone deacetylation might be a possible mechanism for downregulation of miRNAs,64 although it might not be a common mechanism in the regulation of miRNAs.37, 65 In addition, some miRNAs probably are regulated along with their host genes, e.g., the mir-17-92 cluster is upregulated along with the host gene C13orf25, because of locus amplification.47, 48 Furthermore, the expression of miRNAs could be regulated through regulating microprocessors of the miRNAs, for example, by regulating the expression of Dicer.27, 66
Computational prediction of targets of miRNAs
Another challenge is to accurately identify targets that are regulated by miRNAs.26 In animals, miRNAs confer inhibitory regulation of the target-gene translation level through binding in the 3′ UTRs. This was first evidenced by lin-4 targeting of the lin-14/lin-28 genes,8, 9 and let-7 targeting of the lin-41 gene11 in C. elegans. Recently, 4 bioinformatics approaches have been developed for predicting miRNA targets in humans.3, 67, 68, 69, 70 Most bioinformatic algorithms use an “miRNA seed” that encompasses the first 2–8 bases of the mature miRNA sequence to search for complementarity to sequences in the 3′ UTRs of all expressed genes.71 These studies have revealed that a single miRNA might bind to hundreds of gene targets which can be diverse in their function, and that 1 target gene might be regulated by several different miRNAs.26 Briefly, Lewis et al.3, 67 developed a sophisticated algorithm, called TargetScan, and its improved version, TargetScanS. TargetScan3 searches the UTRs for segments of perfect Watson-Crick complementarity to base 2–8 of the miRNA (numbered from the 5′ end), calculates a folding free energy G to each miRNA-target site interaction using RNAeval, and then assigns a Z score to each UTR. TargetScanS67 is an improved algorithm that requires a shorter seed (6 nt), which is preceded by an adenosine and is located in a short “island” of conservation in at least 4 mammalian genomes (i.e., human, mouse, rat and dog). It simply relies on perfect Watson-Crick seed matches that are conserved in the UTR regions of whole genome alignments, and does not rely on free-energy calculation.67 John et al.70 developed an algorithm named miRanda, which optimizes sequence complementarity using position-specific rules and relies on strict requirements of interspecies conservation. Kiriakidou et al.69 developed an algorithm called DIANA-MicroT, which is trained to identify miRNA targets having a single binding site. This algorithm takes an approach that is different from those of the other algorithms described earlier71: (i) It focuses on single-binding-site targets and (ii) it seeks binding sites that have a typical central bulge and require 3′-binding, beyond the obligatory 5′-seed. Krek et al.68 developed an advanced algorithm called PicTar, which is trained to identify both binding sites targeted by a single miRNA, and those that are coregulated by several miRNAs in a coordinated manner. It employs pair-wise alignment for accurate filtering for binding sites that are conserved across 8 vertebrates, and takes into account clustering and coexpression of miRNAs, ontologic information, as well as free energy of RNA:RNA duplexes by using RNA hybrid.72
Recently, we collected all the putative miRNA-target pairs predicted by each algorithm to set up a complete list of all the predicted human miRNA-target pairs, because there is no evidence as yet to show that one is much better than the others. Briefly, we collected 21,293 unique miRNA-target pairs predicted by TargetScanS,67 46,155 unique miRNA-target pairs predicted by PicTar,68 54,578 unique miRNA-target pairs predicted by miRanda70 and 185 unique miRNA-target pairs predicted by DIANA-MicroT.69 After combination, we obtained 101,031 predicted unique miRNA-target pairs. Of the 101,031 pairs, 0.01%(12), 2.8% (2,817), 15.4% (15,510) and 81.8% (82,692) were predicted by 4, 3, 2 and a single algorithm, respectively (Sun M, Chen J. unpublished). Notably, most of the pairs (81.8%) were predicted by only a single algorithm. Although the false positive rate of each of the 4 algorithms was estimated to be 20–30%,71 it is unlikely that most of the 82,692 pairs are false-positive ones. Instead, it is possible that each algorithm may only predict a part of the whole miRNA-target pairs. For example, although the KIT gene was predicted only by miRanda as a target of miR-221, it has been biologically validated to be a real target.73, 74 Thus, the combination of these 4 predictions might provide a much more comprehensive list of putative miRNA-target pairs than do any single prediction.
Current methodologies of miRNA target validation
Because of the potential high false-positive rate of computational prediction of miRNA targets, validating target predictions is very important. MiRNA target prediction algorithms may be informatically validated by evaluation of an algorithm's success in correctly identifying known miRNA targets (i.e. targets that have already been validated biologically, and scoring them highly), or they may be validated by comparing the number of postulated binding sites that an algorithm finds for a real miRNA with that found for a control group of artificially generated “fictitious miRNAs”. Nonetheless, all of these strategies merely provide indirect validation, and they have several limitations.71 However, whereas the ultimate validation of predicted miRNA targets is biological validation, at present there is no a high-throughput method for biologically validating miRNA targets.71 The conventional biological validation methodologies are still extremely labor-intensive and do not allow high-throughput target validation. The commonly used validation methodologies include luciferase reporter assays,3, 39, 68, 69, 75 mutation studies,69, 75 gene-silencing techniques,39, 75, 76 rescue assays77 and classic genetic studies.8, 39, 77 Overall, only about 30 animal miRNA targets have been validated to date with use of these various techniques.71
A new mechanism of miRNA-mediated gene regulation in animals and a potential high-throughput method for identifying miRNA-target pairs
In a classic model of miRNA function in animals,1, 2 miRNAs that form imperfect duplexes with their targets inhibit protein expression without affecting mRNA levels (as illustrated in Fig. 1). However, recent findings indicate that miRNAs that share only partial complementarity with their targets can also induce mRNA degradation.74, 78, 79, 80, 81, 82, 83 For example, in C. elegans, regulation by the let-7 miRNA results in degradation of its lin-41 target mRNA. Furthermore, lin-14 and lin-28 mRNA levels significantly decrease in response to lin-4 miRNA expression.81 In addition, transfection of exogenous miRNA duplexes into HeLa cells can cause moderate downregulation of hundreds of mRNAs, many of which contain the recognition motif of the overexpressed miRNA in their 3′ UTR.80 Moreover, in vivo knockdown of a liver-specific miRNA (miR-122) has shown that hundreds of mRNAs, many of them seemingly to be direct targets of this miRNA, were moderately upregulated.78 Taken together, these studies suggest that mRNAs containing partial miRNA complementary sites can be targeted for degradation in vivo, and that miRNA-dependent regulation of mRNA stability may be more common than previously appreciated.82
Thus, as shown by some pilot studies,74, 82, 83 the predicted miRNA-target pairs could be roughly validated on a large scale by comparison of expression patterns between miRNAs and the predicted targets through conducting microarray analysis on both miRNAs and protein-coding genes. However, the aims of the pilot studies are not to set up a high-throughput method for biological validation of the predicted miRNA-target pairs, and the specificity, robustness and reproducibility of the approaches have not been systematically tested and evaluated. In fact, because only a few miRNA targets have been biologically validated to be affected by miRNAs at the mRNA level,78, 80, 81 it is unclear whether the majority of miRNA targets can be significantly degraded at the mRNA level by miRNAs in animals. In addition, although it was reported that the fold decrease of lin-14 mRNA is consistent with that in LIN-14 protein levels mediated by lin-4 miRNA,81 it was suggested that the changes in target mRNAs are usually weak.82 It is unclear to what extent the mRNA-level changes of the miRNA targets can reflect the changes at their protein level. Thus, a large-scale protein-level analysis technique, such as a protein array, probably should also be integrated in the potential high-throughput technique for miRNA-target-pair validation.
MiRNA signaling pathways in cancer development
As described earlier, thousands of target genes of human miRNAs have been predicted using bioinformatic tools. Interestingly, predominant miRNA targets are transcription factor or kinases.3 Researchers are currently attempting to dissect the target genes of miRNAs and their signaling pathways involved in cancer. Several miRNA signaling pathways have been carefully studied (Fig. 2). As an example, Johnson et al. showed that the 3′ UTR of the human RAS gene contains multiple let-7 complimentary binding sites, allowing let-7 to regulate RAS expression. Furthermore, the overexpression of let-7 in human cancer cell lines results in a decreased level of RAS compared with untreated cells. Conversely, knockdown of let-7 in human cancer cell lines, in which let-7 is normally expressed at a high level, leads to a remarkable increase in RAS protein expression, providing a mechanism for let-7 negatively regulating RAS. Consistent with the observations in vitro, expression of let-7 is lower in lung tumors than in normal lung tissue, while that of RAS protein is significantly higher in lung tumors.39 Introduction of let-7 isoforms (let-7a and let-7f) into a lung cancer cell line A549 resulted in a 78.6% reduction in the number of colonies,38 indicating that growth inhibition may be the biological significance of let-7, presumably through negatively regulating RAS signaling pathways. A few additional genes have also been predicted to be a potential target for let-7, such as LIM kinase 2,3 which belongs to a gene family having a role in the regulation of cell shape and motility and possibly in metastasis as well.
As aforementioned, the mir-17-92 cluster is overexpressed in multiple malignancies. To reveal the mechanisms of the mir-17-92 oncogenic effect directly, He et al.47 utilized a mouse B-cell lymphoma model and demonstrated that enforced expression of the mir-17-92 cluster gene acted with c-myc expression to accelerate tumor development in this model. Tumors derived from haematopoietic stem cells expressing a subset of the mir-17-92 cluster and c-myc possesses less apoptosis than c-myc-induced lymphomas, which underwent extensive apoptosis. Because of its oncogenic function, the authors nominate this miRNA cluster as the first candidate noncoding oncogene, “oncomiR-1.”47 Interestingly, O'Donnell et al.75 independently found that c-Myc binds directly to the miR-17-92 locus and activates expression of a cluster of 6 miRNAs. One of the transcriptional targets of c-Myc, E2F1 is negatively regulated by 2 miRNAs in this cluster, miR-17-5p and miR-20a. This study expands the known classes of transcripts within the MYC target gene network, and uncovers a mechanism through which MYC simultaneously activates E2F1 transcription and limits its translation by the miR-17-92 cluster of miRNAs, allowing a tight control of proliferative signal. Collectively, these findings suggest that the miR-17-92 cluster acts as an oncogene at one set of conditions, while as a tumor suppressor at another.
In most malignant tumors, the PI3K/Akt signaling pathway is activated and results in an aggressive proliferation. The tumor suppressor gene, phosphatase and tensin homolog deleted on chromosome ten (PTEN), inhibits PI3K function. Loss of PTEN expression is associated with PI3K/Akt signaling activation in many types of tumors.84 MiR-19a, b has been demonstrated to bind the 3′ UTR of PTEN mRNA in vitro.3 In lymphomas, overexpression of miR-19 is correlated with underexpression of PTEN protein,35 suggesting that overexpression of mir-19 in tumor cells could be an alternative mechanism by which the PI3K/Akt signaling pathway is activated.
Other studies have linked miRNAs to the regulation of apoptosis. Chan et al.85 revealed a strong over-expression of miR-21 in human glioblastoma tumor tissues and cell lines. Knockdown of miR-21 in cultured glioblastoma cells triggers activation of caspases and leads to increased apoptotic cell death, suggesting that aberrantly expressed miR-21 may contribute to the malignant phenotype by blocking expression of critical apoptosis-related genes. It appears that miR-21 has a broad function in tumor progression, because miR-21 is also upregulated in breast cancer86 and lung cancer.37 Cimmino et al.87 demonstrated that expression of miR-15a and miR-16-1 is inversely correlated to expression of BCL-2 in CLL and that both miRNAs negatively regulate BCL-2 protein at a posttranscriptional level; BCL-2 repression by these miRNAs induces apoptopsis in a leukemic cell lines. Cheng et al.88 employed a library of miRNA inhibitors to screen for miRNA involved in cell growth and apoptosis, and observed that miRNA-mediated regulation has a complexity of cellular outcomes and that miRNAs can be mediators of regulation of cell growth and apoptosis pathways.
KIT is an important tyrosine kinase receptor in cell differentiation and growth. Inhibiting the KIT signaling via miRNAs may contribute to uncontrolled cell growth in certain type of cells.73 He et al.74 reported that miR-221, -222 and -146 are upregulated in papillary thyroid carcinoma and are associated with dramatic decrease of expression of KIT at both transcript and protein levels. The authors further demonstrated that point mutations occurred frequently in the KIT loci (5 out of 10 samples) within the complimentary binding regions for these miRNAs, and therefore disrupted the miRNA–mRNA target interactions in these tumors.
Volinia et al.89 observed that miR-106a overexpression in colon carcinoma inversely correlates to RB1 protein expression, and that downregulation of miR-20a in breast cancer is associated to high expression of TGFBR2 protein, while overexpression of miR-20a in some lung cancer samples is associated with downregulation of TGFBR2 protein. These results suggest that miR-106a and miR-20a might negatively regulate RB1 and TGF β signaling, respectively, which was confirmed by a luciferase assay in vitro.
More recently, to perform genetic screens for novel functions of miRNAs, Voorhoeve et al.90 developed a library (miR-Lib) of vectors expressing the majority of cloned human miRNAs and made a corresponding microarray (miR-Array) containing all miR-Lib inserts. In a screen for miRNAs that cooperate with oncogenes in cellular transformation, the authors identified miR-372 and miR-373 as potential oncogenes in testicular germ cell tumors. Each of them permits proliferation and tumorigenesis of primary human cells that harbor both oncogenic RAS and active wild-type p53. These miRNAs neutralize p53-mediated CDK inhibition, possibly through direct inhibition of the expression of the large tumor suppressor homolog 2 (LATS2), a serine-threonine kinase whose deletion accelerates cellular proliferation and tumorigenic development. In all, the authors provided compelling evidence that these miRNAs are potential novel oncogenes participating in the development of human testicular germ cell tumors by numbing the p53 pathway, thus allowing tumorigenic growth in the presence of wild-type p53.
Taken together, the deregulation of miRNAs participates in activation of cell proliferation or inactivation of apoptotic signaling pathway in conjunction with other genetic changes leading to cancer pathogenesis.91 Further studies would reveal that miRNAs function as key regulators of many more cancer-related genes. Besides the fact that miRNAs inhibit the target genes directly, Gregory and Shiekhattar27 provided evidence that defect of microprocessor of miRNA is one alterative mechanism through which the expression of miRNA can be altered and miRNA perturbation is linked to cancer. Reduced expression of Dicer is associated with poor prognosis in lung cancer patients.66
MicroRNA expression profiles for cancer diagnosis
Emerging data indicate that miRNAs play an important role in regulating gene activation in cancer initiation and progression. Researchers are now using the miRNA-expression signatures to classify cancers and to define miRNA markers that might predict favorable prognosis.26 Various methods have been employed for this purpose. Northern blotting was used in early profiling studies, and is still widely used, because it can convincingly and simultaneously detect both the precursor and mature miRNAs. It provides data about the Dicer processing step, which may be regulated.92 However, Northern blotting is not a high-throughput and sensitive method. Thus, over the past years, real-time RT-PCR and microarray methods have been adapted to detect the expression of miRNAs. High-throughput real-time RT-PCR methods have been developed to screen the expression of precursor and/or mature miRNAs using either SYBR Green dye or TaqMan chemistry (probe) (e.g., Refs.93, 94, 95). To date, largely, there are 4 platforms of microRNA-arrays including dotted cDNA arrays,96 oligonucleotide arrays,97, 98, 99 bead-based flow cytometric miRNA expression profiling40 and “locked nucleic acid” (LAN)-modified oligonucleotide arrays100, 101 (for reviews, see Refs.92 and102). These technologies have been utilized in screening miRNA expression pattern in hematopoietic malignancies and solid tumors.17 For example, using Northern blotting, Chen et al.103 demonstrated that miR-181, miR-223 and miR-142s were differentially or preferentially expressed in hematopoietic tissues. Using a high-throughput Taqman real-time RT-PCR method, Jiang et al.94 screened 222 miRNA precursors in 32 commonly used cell lines of lung, breast, head and neck, colorectal, prostate, pancreatic and hematopoietic cancers. Using oligonucleotide microchips,99 Croce's group has revealed distinct signatures of miRNAs in human breast cancer,86 in human megakaryocytopoiesis104 and in B cell CLL.35, 105 They observed that a unique miRNA signature is associated with prognosis factors and disease progression in B cell CLL.105 Using a similar oligonucleotide microchip, Yanaihara et al.37 analyzed the miRNA expression in 104 pairs of primary lung cancers and corresponding noncancerous lung tissues, and found that the expression of 43 miRNAs was statistically differentially displayed in cancer tissues; their further analysis revealed that the high expression of miR-155 and the low expression of let-7a-2 are correlated with poor survival. In addition to studies on individual type of tumors, a large-scale microRNAome analysis was performed on 540 samples including lung, breast, stomach, prostate, colon and pancreatic tumors with oligonucleotide microchips; cancer-associated miRNAs were identified as miRNA signatures, including some well characterized ones such as miR-17-5p, miR-20a, miR-21, miR-92, miR-106a and miR-155 in solid tumors.59 Lu et al.40 used a new, bead-based flow cytometric miRNA expression profiling method to conduct a systematic expression analysis of 217 mammalian miRNAs on 334 samples, including multiple human cancers. They observed a general downregulation of miRNAs in tumors when compared with that in normal tissues. Furthermore, they successfully classified poorly differentiated tumors using miRNA expression profiles.
Taken together, miRNA expression profiles can be utilized to define the expression pattern of diverse cancers, to predict the prognosis, and to monitor response to therapy or toxicity. It is noteworthy that small RNAs can be measured easily from the formalin-fixed tissue specimens used routinely in hospital pathology laboratories. Thus, potential miRNA-based diagnostics could fit simply into the standard hospital workflow.98 To improve the specificity and sensitivity of miRNA microarrays, Neely et al.100 recently developed a new approach, LAN-modified oligonucleotide array, in which they designed 2 probes for each miRNA (each probe is half of the length of the mature miRNA) and labeled with different fluorophores. With such an approach, the authors were able to quantitate miRNA expression from as little as 50 ng total RNA with a high specificity.100 In addition, Kloosterman et al.101 designed LNA oligonucleotides and applied them for in situ hybridization to monitor the spatial and temporal expression of miRNAs in vivo. Such technique improvement and renovation will make it more promising to use miRNA expression profiles for cancer diagnosis, for defining a treatment strategy for patients, for prediction of the prognosis, and for monitoring treatment response.
Targeting miRNA regulation with antisense oligonucleotides: a potential therapeutic intervention
To investigate the potential causative roles of deregulated miRNA processes in human disease, synthetic, chemically-modified antisense oligonucleotides targeting to miRNAs or protein-coding genes have been widely used. They have been proven to be a powerful research tool, and may eventually become a new therapeutic tool.106, 107, 108 On the one hand, if a disease phenotype is related to the overexpression of miRNAs, antisense oligonucleotides that are complementary to either the mature miRNAs or their precursors can be designed to specifically inhibit the activity of the miRNAs. There are 3 types of modifications for anti-miRNA oligonucleotides (AMO)106: (i) 2′-O-Methyl (OMe) AMOs, (ii) 2′-O-Methoxyethyl (MOE) AMOs and (iii) Locked nucleic acid (LNA) AMOs. Various AMOs have been implemented in experimental studies. For example, Cheng et al.88 applied a library of Me-AMOs to screen for miRNAs that are involved in cell growth and apoptosis. Esau et al.109 inhibited miR-122 in mice with a MOE-AMO to uncover the role of the liver-specific miR-122 in the adult liver and demonstrated that miR-122 is a key regulator of cholesterol and fatty-acid metabolism in the adult liver. Chan et al.85 successfully applied 2′-O-methyl- and DNA/LNA-mixed oligonucleotides to specifically knockdown miR-21 to investigate the potential contribution of this miRNA in the regulation of apoptosis-associated genes in glioblastoma cell lines. On the other hand, to enhance the function of miRNAs due to a deletion or a loss of function mutation, a therapeutic approach could entail exogenous delivery of corrective synthetic miRNAs in the form of (siRNA-like) double strand oligoRNAs.106 One example shows that enforced expression of let-7 in the lung adenocarcinoma cell line A549 inhibited lung cancer cell growth in vitro.38 This holds a promise that let-7 may be useful in treatment of lung cancer or enhancing the current treatments. Synthesized miRNA-14U was used to inhibit the overexpressed HER-2 protein in ovarian cancer cell line SKOV3.110
To develop a pharmacological approach for silencing miRNAs in vivo, Krutzfeldt et al.78 designed chemically modified, cholesterol-conjugated single-stranded RNA analogues (termed “antagomirs”), which are complementary to miRNAs. Their study shows that the termed antagomirs are a powerful tool to silence specific miRNAs in vivo.78 Therefore, we may expect that in the future, activation of miRNAs with “agomiR” or silencing of miRNAs with “antagomirs” could become a therapeutic strategy for cancer treatment. Nevertheless, there are still several difficulties to overcome before applying such strategies in clinic. For example, as miRNAs may have dual functions (oncogene or tumor suppressor) in different organs or conditions, is it possible to design “agomiR” or “antagomirs” that target specific tissues or disease sites?108 In addition, the regulation and stability of the injected “agomiR” or “antagomirs” in vivo, the proper dosages for the optimal effect, and the potential side effects, should be carefully examined and addressed. Moreover, because a single miRNA may regulate hundreds of different targets while a single target gene may be regulated by several different miRNAs, the regulation network may be very complex. Thus, more cautions should be taken in considering clinical applications.
The microRNA field is rapidly developing, as shown by the facts that the number of experimentally identified miRNAs and their targets is increasing, and that the functions and the signaling pathways of more and more miRNAs have been carefully studied. The recent findings suggest that miRNAs play an important role in regulation of gene activation by binding to the mRNAs of target genes, by switching the genes on or off, or by fine-tuning the genes, to control cell viability, cell cycle, proliferation or apoptosis. It was predicted that more than half of human genes might be regulated by miRNAs. Deregulation of miRNA expression is often associated with cancer. These studies may change the landscape of cancer genetics and uncover new mechanisms that contribute to cancer development. The future challenges are to learn how miRNAs are regulated in human genome, and to identify more biological targets of miRNAs and the relevant signaling pathways. Nonetheless, activation of miRNAs with “agomiR” or silencing of miRNAs with “antagomirs” could probably become a therapeutic strategy for cancer treatment in the future, although we are far from that point.
We thank Dr. Janet D. Rowley and Dr. Suzanne D. Conzen for their critical comments and valuable suggestions, and 3 anonymous referees for their constructive comments. We also appreciate the help of Ms. Mary Beth Neilly in manuscript revision. We apologize to our colleagues whose outstanding contributions to the growing miRNA field were not cited as primary references because of the space constraints.