|Home | About | Journals | Submit | Contact Us | Français|
Discovered less than a decade ago, micro-RNAs (miRNAs) have emerged as important regulators of gene expression in mammals. They consist of short nucleic acids, on average ~22 nucleotides in length. The miRNAs exert their effect by binding directly to target messenger RNAs (mRNAs) and inhibiting mRNA stability and translation. Each miRNA can bind to multiple targets and many miRNAs can bind to the same target mRNA, allowing for a complex pattern of regulation of gene expression. Once bound to their targets, miRNAs can suppress translation of the mRNA by either sequestration or degradation of the message. Thus, miRNAs function as powerful and sensitive posttranscriptional regulators of gene expression. This review will summarize what is known about miRNA biogenesis, expression, regulation, function, mode of action, and role in disease processes with an emphasis on miRNAs in mammals. We discuss some of the methodology employed in miRNA research and the potential of miRNAs as therapeutic targets. The role of miRNAs in signal transduction and cellular stress is reviewed. Lastly, we identify new exciting avenues of research on the role of miRNAs in toxicogenomics and the possibility of epigenetic effects on gene expression.
Micro-RNAs (miRNAs) were first discovered in the 1990s through genetic screens in Caenorhabditis elegans (Lee et al., 1993) and were subsequently shown to play a role in regulation of gene expression in virtually all eukaryotes. Through experimental approaches and the application of bioinformatics thousands of miRNAs have been identified cumulatively across all species examined. miRNAs constitute a class of evolutionarily conserved, endogenously expressed, noncoding RNAs which average ~22 nucleotides (nt) in length and which bind with imperfect complementarity to the 3′ untranslated region (3′-UTR) and negatively regulate target messenger RNAs (mRNAs) (Ambros, 2004; Bartel, 2004). They are widely expressed in virtually all tissues and all stages of development. Greater than one third of all human genes have been predicted to be miRNA targets (Lewis et al., 2005) and therefore miRNAs are an abundant and important class of regulatory molecules. First described as having a role in developmental processes in worms and flies, miRNAs have since been associated with development and differentiation of tissues in mammals (Anderson et al., 2006), learning and memory (Fiore and Schratt, 2007), maintenance of germline stem cells (Park et al., 2007), cell cycle progression (Carleton et al., 2007), apoptosis, as well as such diverse disease processes as tumorigenesis (He and Hannon, 2004; O'Donnell et al., 2005), cardiac hypertrophy (van and Olson, 2007a), Alzheimer's disease (Lukiw, 2007), and diabetes (Poy et al., 2004). In addition, some viruses have been shown to encode miRNAs (Pfeffer et al., 2004) which can regulate host gene expression (Stern-Ginossar et al., 2007).
Multiple endogenous RNA-based silencing mechanisms have been described, such as miRNA, siRNA (small interfering RNA), tasiRNA (trans-acting siRNA), natsiRNA (natural antisense siRNA), and piRNA (piwi-interacting RNA, found only in germline cells). The core mechanism of the silencing process is triggered by the presence of an RNA:RNA duplex (for review, see Chapman and Carrington, 2007). The binding of the RNA molecules is usually with less than perfect complementarity. The distinction in the various mechanisms results from the differences in the specific trigger, the proteins involved in processing the RNA and the class of small RNA. For the purposes of this review we will focus on miRNA.
Structurally, miRNAs average 22 nucleotides in length. The 5′ region of the miRNA, nucleotides 2–7, is defined as the seed region. The seed region binds the target mRNA by near-perfect Watson–Crick pairing (Lewis et al., 2005). The degree of complementarity determines the type of regulation of the mRNA, as detailed below. The remainder of the miRNA binds the target with a variable amount of complementarity, wobbles, bulges and non-Watson-Crick base pairing. Target sites often occur in clusters or multiples and may even overlap. It is likely that several miRNAs may bind one mRNA, working cooperatively to silence gene expression (Saetrom et al., 2007). On the other hand, we propose the possibility that miRNAs may bind a target mRNA in a competitive manner to modulate the degree of silencing.
The vast majority of miRNAs are transcribed as longer primary miRNAs (pri-miRNAs) by RNA polymerase II, contain a 5′-methyl7G cap structure and are polyadenylated, much like mRNAs (Cai et al., 2004). It is possible that some additional miRNAs may be transcribed by RNA polymerase III (Zhou et al., 2007). The pri-miRNAs bear hairpin structures, which are processed in a compartmentalized, stepwise fashion by the RNase III enzymes Drosha and Dicer (Fig. 1). Processing begins in the nucleus, where Drosha and a double-stranded RNA-binding protein, known as DGCR8 in mammals, cleave near the base of the double-stranded structure to release the 60–70 nt long precursor mi-RNA hairpin (Lee et al., 2003). This double-stranded RNA processing creates molecules with a 5′ phosphate group and a 2-nt overhang on the 3′ end (Lee et al., 2003). Export to the cytoplasm is mediated by exportin-5 and Ran/GTP (Yi et al., 2003). Once in the cytoplasm, Dicer cleaves on the loop side of the hairpin to generate an ~22nt long miRNA:miRNA* duplex (Du and Zamore, 2005). The mature miRNA can arise from either arm of the duplex. Although Dicer alone is sufficient for the cleavage reaction, the protein is found associated with an RNA-binding cofactor, such as transactivator RNA-binding protein (Chendrimada et al., 2005). The Dicer cofactors facilitate effector complex formation, removal and degradation of the complementary strand (miRNA*) of the duplex, stabilization of the miRNA and may aid in identification of target mRNAs (Liu et al., 2003). The final effector complex is known as miRISC, which stands for miRNA-containing RNA-induced silencing complex (Kim and Nam, 2006).
Another pathway for the biogenesis of miRNA in the nucleus, called mirtrons, has been described first in invertebrates (Okamura et al., 2004; Ruby et al., 2007) and more recently in mammals (Berezikov et al., 2007). These miRNAs are derived from short introns that can form hairpins, but lack the base pairing necessary for recognition and cleavage by Drosha. The mirtrons are spliced and debranched in the nucleus and then join the miRNA pathway, are exported to the cytoplasm, where they are cleaved by Dicer and exert their effects on their target mRNAs.
In mammals, miRNAs are found throughout the genome in all chromosomes with the exception of the Y chromosome (Ro et al., 2007). The miRNA genes are not randomly dispersed, but are instead found in clusters (Lau et al., 2001). These clusters may share a common regulatory region resulting in transcription of polycistronic RNA (Lee et al., 2002). The gene clusters are likely the result of gene duplication resulting in closely related miRNA genes, but may also contain unrelated miRNAs. The cluster of miRNA genes may be functionally related by targeting the same gene or members of a gene family. Another possibility is that multiple members of the same pathway are targeted by related miRNAs. miRNA genes are found in intergenic regions as well as in introns or exons of protein coding mRNAs and in introns of noncoding RNA (Rodriguez et al., 2004). The majority of miRNAs identified thus far have been shown to have orthologs in other species, suggesting the role of miRNA in gene regulation is evolutionarily conserved. In addition, when sequence variation occurs between homologs, it is often accompanied by a compensatory change in the other arm of the hairpin to maintain pairing, which provides phylogenetic support for the importance of these secondary structures (Lau et al., 2001).
Expression of miRNAs seems to be tightly regulated in a temporal (developmental) or tissue-specific manner. Because half of the known miRNA genes reside within the transcription units of protein coding genes and noncoding RNAs and can derive from introns or exons or both, it is likely that the regulation of expression of miRNAs is complex (Rodriguez et al., 2004). Some may be regulated at the posttranscriptional level by splicing factors and transport of the RNA. Therefore, studies on the promoter regions of miRNA genes have been performed mainly on the intergenic miRNA clusters. The promoters of miRNA genes share many characteristics with typical targets of RNA polymerase II such as TATA boxes (Houbaviy et al., 2005) and contain conserved motifs that may serve as transcription factor–binding sites (Kim and Nam, 2006). These putative cis-acting promoter elements may bind known transcription factors that regulate mRNA transcription or possibly new classes of transcription factors. A recent comprehensive comparison of the promoter regions of miRNAs in four species, Homo sapiens, C. elegans, Arabidopsis thaliana, and Oryza sativa, confirmed that the core promoter region of miRNA genes are located within 500 bp of the gene, showed that many miRNA genes are TATA-less, but also provided evidence that a small fraction of miRNA genes may be transcribed instead by RNA polymerase III (Zhou et al., 2007) and thus be regulated independently of canonical mRNA transcription units. The paper also describes conserved CT repeat microsatellites in the promoters of all four species and CpG islands close to the miRNA genes, similar to protein coding genes. Two recent reports identified a common human miRNA consensus sequence, CCCc/aCt/cCC that may be an important miRNA regulatory motif (Inouchi et al., 2007; Zhou et al., 2007). As more genomic sequence data becomes available for a wider variety of species, the relative conservation of promoter elements is being elucidated. One recent comparative genomics study shows that miRNA promoter regions are twice as conserved as mRNA promoters (Mahony et al., 2007) which underscores the important role that miRNAs play in the regulation of gene expression in animals.
Investigations of the transcriptional regulation of individual miRNAs have shown that canonical mRNA transcription factors are indeed utilized in the regulation of miRNA expression. One study demonstrates the roles of nuclear factor I/A (NFI-A) and CCAAT/enhancer binding protein alpha (C/EBPα) in regulation of miR-223 in human granulocyte differentiation (Fazi et al., 2005). NFI-A binds to the promoter of miR-223 and inhibits transcription. Interestingly, the NFI-A mRNA is a target of miR-223. Retinoic acid treatment of cells releases NFI-A, allows the binding of C/EBPα, which causes transcriptional activation of miR-223 and subsequent inhibition of NFI-A translation. This molecular circuit thus allows for a sustained alteration in gene expression in a feed-forward loop leading to terminal differentiation in granulopoiesis. In a study performed on the mouse miR-223, a different promoter was defined for the miR-223 gene, which is regulated by two conserved PU.1 sites as well as C/EBPα (Fukao et al., 2007). Such conflicting reports demonstrate the need for a better understanding of what defines miRNA promoters. Other examples of known transcription factors involved in miRNA expression are p53 and c-Myc, two important regulators of the cell cycle. p53 has been shown to bind to and regulate expression of the miR-34 cluster promoter (Corney et al., 2007; He et al., 2007b, c). The miR-34 miRNAs repress translation of genes involved in cell proliferation and anchorage independent growth, thus demonstrating that miRNAs are effectors of p53 signaling in the cell (Corney et al., 2007). c-Myc has been shown to bind to E-box sequences in the miR-17 promoter (O'Donnell et al., 2005) and to regulate expression of several miRNAs including the miR-17 cluster in B cells (Zeller et al., 2006). Aberrant expression of the miR-17 miRNAs can hasten oncogenesis in a mouse model of lymphoma (He et al., 2005). A study that further details the regulation of the miR-17 cluster has revealed an interesting transcriptional network (Woods et al., 2007). The promoter of miR-17 contains a nonconsensus TATA box and two E2F transcription factor–binding sites. The authors used chromatin immunoprecipitation analysis to show that E2F3, a proliferation signal transducer, binds to the miR-17 promoter. E2F1 is a target of miR-17 repression (O'Donnell et al., 2005) and a proapoptotic protein. Ironically, c-Myc also induces transcription of E2F1 (Coller et al., 2007). Thus, expression of miR-17 is part of a complex network of gene regulation and facilitates the fine-tuning of the opposing proliferative and apoptotic signals. In another report, Peroxisome proliferator-activated receptor (PPAR)alpha was shown to regulate expression of Let-7c which targets the c-Myc mRNA (Shah et al., 2007). Inhibition of c-Myc in turn reduces the expression of the miR-17 cluster of miRNAs. The expression of miRNAs and interaction with their target sites form an elaborate network of feedback and feed-forward loops in gene regulation (Shalgi et al., 2007; Tsang et al., 2007).
The nature of miRNAs and their mode of binding their targets have made the task of target identification a difficult one. Only the short seed sequence in nucleotides 2–7 at the 5′ end of an miRNA binds the complementary target site on an mRNA. Bulges or G-U wobble matches can further confound efforts to determine targets. This small region of variable complementarity suggests that each miRNA could potentially bind to multiple targets, greatly extending the impact of these small RNAs on gene expression (Lewis et al., 2005). Existing algorithms for statistical sequence matching depended on longer stretches of sequence similarity (Karlin and Altschul, 1990) and so new computational methods have been developed to study miRNA–mRNA interactions. The first successful large-scale target prediction tool was used by Stark et al. (2003) to identify regions of high complementarity on mRNA targets in Drosophila. The method made use of the comprehensive database of defined Drosophila mRNA 3′-UTRs, identified potential binding sites and then looked for conserved sites among three species of flies. The thermodynamic stability of the predicted miRNA:mRNA association was examined using mFOLD (Zuker et al., 1999). This method identified known targets and revealed additional targets, some of which were validated experimentally. The 3′-UTRs of human transcripts are less defined than those of flies and potentially complicated by alternative splicing of genes. The miRanda algorithm is similar to its predecessor, but incorporates a statistical weighting of matches in the seed region of the miRNA, which reduced the rate of false-positive matches (John et al., 2004). Other algorithms for the prediction of miRNA targets include TargetScan, PicTar, and DIANA-microT and use different filters or statistical models for target identification. Other prediction methods use motif scanning of 3′-UTR sequences to look for potential targets and then look for complementary miRNAs or even whole-genome approaches. These computational methods for miRNA research are rapidly evolving. As an alternative to computational target prediction, a functional genetic screen was used in mutant zebrafish with defective Dicer. This forward genetics approach elucidated a number of targets that are normally regulated by miRNAs (Giraldez et al., 2005).
A number of freely accessible miRNA databases are available on the web such as: miRBase (http://microrna.sanger.ac.uk/) which can be used to search for miRNAs and their targets in multiple databases, miRNAviewer (http://cbio.mskcc.org/mirnaviewer/) allows the user to search for miRNAs and targets in human, Drosophila, or zebrafish databases, Argonaute (http://www.ma.uni-heidelberg.de/apps/zmf/argonaute/) is a database that allows the user to search by miRNA name and lists the tissues in which miRNAs have been reported to be expressed, and miRanda (http://www.microrna.org/miranda_new.html) is an open source algorithm for finding genomic miRNA targets. The most comprehensive database is MicroRNAdb (http://bioinfo.au.tsinghua.edu.cn/micrornadb/), which, as of this writing, lists 732 entries of micro-RNA sequences, and 439 entries of detailed annotations. TarBase (http://www.diana.pcbi.upenn.edu/tarbase.html) is an especially nice database that provides comprehensive information on experimentally supported miRNA targets in at least eight organisms. mirGen, which is linked to TarBase (http://www.diana.pcbi.upenn.edu/miRGen.html) provides information on whole-genome collections of miRNAs, on predicted miRNA clusters and provides specific functional information on the targets of miRNAs within each cluster. And finally, miRGator (http://genome.ewha.ac.kr/miRGator/) is comprised of a miRNA database with a navigator tool to integrate expression profiling with target gene prediction to aid in functional interpretation.
Gene silencing by the effector complex, miRISC, can proceed by two main mechanisms, mRNA cleavage (Zamore et al., 2000) or translational repression (Olsen and Ambros, 1999), which is more common in mammals. It is thought that the degree of complementarity between miRNA and mRNA determines whether the mRNA will be repressed or cleaved (Hutvagner and Zamore, 2002); near-perfect matching leads to degradation of the message. A key component of the miRISC complex is the Argonaute (Ago) protein. The Ago proteins are a family of proteins (Ago1–4 in humans) that are widely expressed in somatic tissues. Of the family members, only Ago2 possesses endonucleolytic activity (Liu et al., 2004; Meister et al., 2004) and can therefore cleave mRNA. Near-perfect pairing between miRNA and target mRNA provides a good substrate for recognition and cleavage by Ago2. If the miRNA binds its target with internal mismatched bulges there is no cleavage, but translation of the message is repressed. Several factors that may affect the efficiency of repression are the number and configuration of mismatches, flanking sequences and the number of target sites on the mRNA (Standart and Jackson, 2007). Translationally suppressed mRNAs can be found associated with polysomes (Nottrott et al., 2006) and actively translating ribosomes (Maroney et al., 2006) or sequestered from the translation apparatus in the cytoplasm. In either case, it appears that the m7G cap structure on the mRNA is involved, as cap-independent translation is not repressed (Pillai et al., 2005). Recently, Kiriakidou et al. (2007) identified a motif within Ago proteins similar to that the m7G cap binding domain of eIF4E, suggesting that Ago can compete for binding the message and exclude eIF4E. miRISC complexes along with suppressed mRNAs accumulate in the cytoplasm in foci called processing bodies (P-bodies) (Liu et al., 2005b; Pillai et al., 2005). P-bodies have been well-characterized in eukaryotic cells and contain decapping proteins and the exoribonuclease, XrnI (Fig. 1) and can therefore degrade the Ago2 cleaved mRNAs. Direct interaction of the P-body protein GW182 with Ago has been shown to be essential for the translational silencing activity (Liu et al., 2005a). Recent evidence suggests that miRNA suppression may be reversible and that mRNAs may leave the P-bodies and enter polyribosomes (Bhattacharyya et al., 2006b). Similar to P-bodies, stress granules (SGs) are sites of accumulated nontranslating mRNAs associated with ribosomes that form in response to osmotic or oxidative stress or heat shock (Anderson and Kedersha, 2006). Leung et al., demonstrated that Ago proteins translocate from their distribution in the cytoplasm to SGs in response to stress and that this movement is dependent on miRNAs (Leung et al., 2006), suggesting that during stress miRNAs bind translating mRNAs, associate with Ago proteins and re-direct the complex into SGs. When the cell recovers from stress, the mRNAs can either re-enter polyribosomes to resume translation or transfer to P-bodies for degradation. An example of such miRNA mediated repression and derepression is the regulation of the cationic amino acid transporter 1 (CAT-1) mRNA and miR-122 (Bhattacharyya et al., 2006a). Interactions with the HuR protein, which binds AU-rich elements (AREs) in the 3′-UTR, and perhaps other RNA-binding proteins, cause the release of CAT-1 mRNA from P-bodies and association with polysomes.
Thus, the storage of RNAs and proteins in P-bodies is a dynamic process that depends on accumulation and turnover of repressed mRNAs through degradation or release. Less is currently known about the mechanism facilitating the transport of repressed mRNAs back to actively translating ribosomes. However it has been reported that the cellular apolipoprotein B mRNA-editing enzyme catalytic polypeptide-like 3G protein (A3G) plays a significant role in counteracting the miRNA inhibition of protein synthesis through interactions with Ago proteins, poly (A)-binding protein and polysomes (Huang et al., 2007). A3G and other related protein family members were shown to overcome the inhibition by a variety of miRNAs, facilitate the exit of the complexes from the P-bodies and target them to polysomes where translation is resumed.
It is interesting to note that a new functional role has recently been proposed for miRNAs. miRNAs in conjunction with AREs in the tumor necrosis factor (TNF)-α mRNA were shown to function as positive regulators of translation during cell cycle arrest (Vasudevan et al., 2007). These same factors function as repressors of translation of the same mRNA during cell proliferation and thus miRNAs may have dual roles as suppressors and activators of translation during the cell cycle.
Although first thought to be a part of a developmental process, miRNAs have been shown to play a role in virtually every cellular process; and like other key regulators, aberrant expression of miRNAs is associated with cellular dysfunction and disease. Most attention has focused on the contribution of miRNAs in carcinogenesis, but altered miRNA expression levels contribute to insulin resistance in 3T3 adipocytes and diabetes in rats (He et al., 2007a), cardiac hypertrophy (van and Olson, 2007a), Alzheimer's disease (Lukiw, 2007), and may even play a role in mental retardation (Qurashi et al., 2007) and Tourette's Syndrome (Chang and Mendell, 2007).
The first report of aberrant miRNA expression related to cancer was shown in patients with B cell chronic lymphocytic leukemia (Calin et al., 2002). Since then there have been reports of miRNAs involved in tumor development based on their substantially altered expression patterns and a detailed portrait of the function and dysfunction of miRNAs in tumorigenesis is emerging (reviewed in He et al., 2007c; Kent and Mendell, 2006). Expression profiling of mRNAs in a variety of normal and transformed cell lines (e.g., MCF-7, MDA-231) and patient samples, have generally proven to be more tissue-specific rather than tumor specific (Ramaswamy et al., 2001). Reports of miRNA profiling, in contrast, have shown globally reduced miRNA expression among tumors as compared with normal tissues and revealed more tumor specific miRNA profiles (Iorio et al., 2005; Lu et al., 2005). Preliminary data from our laboratory comparing miRNA profiles from MCF10A normal breast epithelial cells and MCF10A1a tumor cells have revealed distinctly different miRNA profiles (Dombkowski, Miller, Cuckovic, Zheng and Novak, unpublished data). It is hoped that a detailed portrait of the miRNA expression profiles in specific tumor types will result in the development of biomarkers for refined cancer diagnosis and prognostic tools for management of the disease (Waldman and Terzic, 2007). The specific targets of the miRNAs are for the most part unknown, however, targets of a few miRNAs that are overexpressed in tumors have been identified. For example, miRNA-21 has been shown to repress the expression of a tumor suppressor gene, tropomyosin 1 in MCF-7 cells (Zhu et al., 2007). A complex network of regulation is emerging that shows how miRNAs interact with classic oncogenic and tumor suppressor genes such as c-Myc and p53 as mentioned above. In particular, the expression level of miR-34, which is regulated by p53, was found to be reduced in several human cancers (He et al., 2007d). In fact, miRNAs themselves may act directly as tumor suppressors or oncogenes (Zhang et al., 2007). Further evidence from a study of rats treated with a hepatocarcinogen indicates that changes in expression of miRNAs occur early on and not subsequent to tumorigenesis (Pogribny et al., 2007).
Standard methods for studies of mRNAs have been applied to the study of miRNAs. These include northern blot analysis for identification of individual miRNAs by hybridization of specific probes, and miRNA microarray analysis and quantitative real-time PCR for profiling of miRNA expression. Global transcriptional profiling allows the identification of miRNA genes expressed differentially or uniquely in various cell types or during pathogenesis and the elucidation of molecular signatures. These standard methods have been modified in order to be applied to the study of miRNAs due to their small size and short regions of homology. A recent review summarizing many methods used to examine miRNA function, including protocols is available (Clancy et al., 2007). Microarray analysis of miRNAs is challenging and can result in many false-positive and false-negative results. Our laboratory has successfully utilized Agilent arrays for miRNA profiling (Dombkowski, Miller, Cuckovic, Zheng and Novak, unpublished data). Isolation of miRNAs from total RNA is inefficient and can lead to a misrepresentative population of miRNAs. We use Agilent's direct label method on the miRNAs without prior purification from total RNA which results in high quality, reproducible results.
Other applications include genomic, bioinformatic and systems biology approaches (Gusev et al., 2007). Genomic miRNA bioinformatic analysis, termed miRNomics, uses in silico approaches to gain insight into the biological and potential therapeutic impact of miRNAs (Ghosh et al., 2007). Comparative genomics and transcriptomics help to explore evolutionarily conserved and divergent transcriptional patterns and functional landscapes within tissues, and identify fundamental and species-specific mechanisms controlling biological processes. Epigenetic analysis reveals methylation patterns and miRNA profiles unique to specific tissues. New applications of high-throughput technology are rapidly evolving. For example, a new sensitive PCR and sequencing method for measuring miRNA expression is called mRAP amplification profiling and is based on reverse transcribed complimentary DNAs primed from ligated adapters and uses very little starting material (Takada and Mano, 2007). This method may be useful in clinical applications.
An important development is that of chemically modified, synthetic nucleotides called locked nucleic acid (LNA) oligonucleotides which have higher melting temperatures and improved specificity (Grunweller and Hartmann, 2007) that aids in mismatch discrimination. LNA antagomirs have recently been used in mice to selectively inhibit miR-122 translation silencing (Elmen et al., 2007). LNA oligonucleotides may be especially useful for antisense therapeutic use due to their stability and low toxicity.
In order to look at the processing of miRNAs in cells, one group used in situ fluorescence microscopy (Ohrt et al., 2006). Synthetic, double-stranded, fluorescently labeled RNAs were injected into cells and observed with confocal microscopy. Such methods may be also useful for studying transport, stability and localization of miRNAs or antagomirs in cells.
The functional effects of miRNAs have been shown by comparing the overall level of the target mRNA with the protein product. The target mRNA is measured by northern blotting, RNAse protection assay or quantitative PCR, whereas the protein product is measured by western blot. The presence and stability of the mRNA in the absence of protein product is indicative of selected translational suppression. Other functional assays for miRNAs include the application of reporter constructs containing a fluorescent protein or luciferase gene with miRNA targets in the 3′-UTR (Clancy et al., 2007). These can be used in vivo by transfection into cells that express the miRNAs, but can also be used in in vitro translation assays. Alternatively, miRNAs and reporter plasmids may be cotransfected into any suitable cell line. Individual miRNAs can be tested for their ability to silence translation of the reporter construct. Controls used are miRNAs of nonhomologous sequence or reporters with scrambled target sequence.
One can also monitor the effect of miRNAs on the subcellular location of the target message by sucrose density gradient centrifugation and polysomal analysis. Binding of miRNAs can change the association of the message from the high density polysomes to a lower density fraction. This method has long been used to study translational regulation of mRNAs (Rosen et al., 1982). A new high-throughput application of polysomal profile analysis utilizes spotted DNA microarrays to examine the translational regulation of total cellular mRNAs (Melamed and Arava, 2007) and will therefore be useful for studying changes in translation of miRNA targets.
A new bioinformatic tool for prediction of functional effects of miRNAs under different conditions called miRNAXpress has been developed for plants. It uses computed matrices of expression patterns, miRNAs, and miRNA targets to predict biological outcomes under specific conditions (Zilberstein et al., 2006). Such sophisticated algorithms will be of great importance in the development of miRNA therapeutics in mammals.
The obvious importance of miRNAs for regulation of gene expression and their role in pathogenesis suggests that they would make promising new therapeutic targets (Pappas et al., 2008). Although pharmaceutical drugs aim to develop highly specific targets, due to the nature of the action and interaction of miRNAs, any perturbation of a specific miRNA will likely affect multiple targets within a gene network and substantially alter a disease phenotype and therefore could prove more efficacious than traditional drug therapies. On the other hand, the challenges associated with specificity and multiple targets could lead to unintended side effects, so the therapeutic potential of miRNAs must be approached with caution. Many miRNAs are a part of closely related families of miRNAs that are expressed in many different tissues, so targeted delivery systems are necessitated to avoid unwanted consequences.
Chemically modified, cholesterol conjugated, single-stranded oligonucleotides that are complementary to specific miRNAs are called antagomirs and effectively inactivate miRNAs through base pairing. Antagomirs have been shown to be effective at disrupting the action of specific miRNAs in a wide variety of tissues in mice following intravenous administration (Krutzfeldt et al., 2005). The silencing was specific and resulted in increases in the mRNA targets of specific miRNAs. A novel alternative approach is that of gene-specific miRNA mimics and miRNA masking, which utilize RNA oligonucleotides that are complementary to gene-specific 3′-UTR sequences (Xiao et al., 2007). The antisense oligonucleotides bind to the 3′-UTR of the mRNA, effectively competing with the miRNA, but do not repress translation or initiate degradation of the message. Although not yet tested in vivo, the new method shows promise by increasing specificity and limiting the chance of unwanted consequences. Another possible therapeutic application of miRNAs is the introduction of novel, artificial miRNAs to regulate specific genes (van and Olson, 2007b).
Endocrine regulation of energy homeostasis is dependent on the effects of insulin and glucagon. Control of energy metabolism by insulin is mediated by a signal transduction pathway beginning with the activation of the insulin receptor which initiates a phosphorylation cascade causing the activation of phosphatidylinositol 3-kinase (PI3K), Akt/protein kinase B, and the mammalian target of rapamycin which has been shown to regulate CYP2E1 and gammaglutamylcysteine ligase expression, a rate-limiting enzyme for production of glutathione in response to oxidative stress (for a current review, see Kim and Novak, 2007). Activation of this signaling pathway results in alterations of the transcription and translation of genes involved in fat, glucose and glycogen metabolism. In all of the major tissues involved in energy homeostasis, pancreas, liver, muscle and adipose tissue, specific miRNAs have been identified that regulate mammalian metabolic processes. miRNAs have been shown to modulate various aspects of the insulin pathway, beginning with the discovery that insulin secretion from the pancreas is regulated by miR-375 (Poy et al., 2004). miR-375 binds to and inhibits translation of the mRNA of the myotrophin gene which controls glucose-stimulated insulin exocytosis in the pancreas, effectively regulating blood glucose levels. Another miRNA expressed in cells of the pancreas is miR-124a. miR-124a has been demonstrated to bind to a specific site in the 3′-UTR of the transcription factor Fox2a and inhibit its translation. Overexpression of mir-124a shows a concomitant decrease in the expression of Fox2a protein and Fox2a target genes, such as the preproinsulin gene (Baroukh et al., 2007). The Let-7c miRNA inhibits expression of PPARgamma, a key regulator of function in adipocytes (Esau et al., 2004). Growth factors and serum have been shown to induce several miRNAs in cultured fibroblasts and this response is dependent on signaling via the PI3K pathway (Gu and Iyer, 2006). Other components of the insulin signaling pathway are regulated by miRNAs as well. Let-7a-3 miRNA reduces expression of insulin-like growth factor II (Lu et al., 2007) and the expression of insulin receptor substrate-1 can be modulated by miR-145 (Shi et al., 2007). In Drosophila, mutants lacking the miR-278 miRNA are defective in fat storage and are insulin resistant, indicating that this miRNA contributes to the regulation of energy metabolism in adipose tissue through modulation of insulin responsiveness (Teleman et al., 2006). Recently, miR-29, a miRNA which is upregulated in diabetic (Goto-Kakizaki) rats, was shown to repress glucose uptake in response to insulin when overexpressed in 3T3-L1 adipocytes (He et al., 2007a). In addition, the authors provide evidence that the impaired insulin signaling effected by miR-29 occurs upstream of Akt in the phosphorylation cascade. Taken together, these data strongly suggest that miRNAs are key regulators of endocrine control of energy metabolism. Hence, miRNAs are good potential targets for therapeutic intervention in metabolic disorders.
Different types of cellular stress have been shown to regulate miRNA levels. Many miRNAs target stress-associated genes, such as superoxide dismutase (SOD). Oxidative stress induces SOD genes in a transcription-independent manner as a result of the derepression of translation by miR-398 (Sunkar et al., 2006).
Under hypoxic stress, translation is generally suppressed (Liu et al., 2006) and yet translation of a subset of genes is specifically upregulated (Koritzinsky et al., 2005). This selective control of translation is mediated by miRNAs (Rocha, 2007). In particular, hypoxia-responsive transcription factors such as nuclear factor-kappa B (Taganov et al., 2006) and p53 (He et al., 2007b) have been shown to induce miRNAs.
Likewise, cold stress is known to induce expression of the RNA-binding protein Rbm3, which is thought to act as an RNA chaperone to facilitate translation during hypothermia (Danno et al., 1997). Interestingly, it has been shown that Rbm3 induction during cold stress alters miRNA levels, suggesting that miRNAs and Rbm3 have opposing roles in hypothermia and in cellular homeostasis (Dresios et al., 2005).
Nutrient deprivation is another form of cellular stress. Folate deprivation has been shown to alter miRNA expression profiles (Marsit et al., 2006). Specifically, miR-222 is significantly overexpressed in response to folate deficiency in cells in culture as well as in peripheral blood from humans with low folate in their diets. Serum starvation of cells in culture results in specific upregulation of TNF-α mediated by the ARE sequences in the 3′-UTR, miRNAs and Ago proteins (Vasudevan and Steitz, 2007).
The implications of such studies are that long-term nutrient deficiency or chronic oxidative stress may have an impact on miRNAs and on global gene expression, perhaps leaving tissues more susceptible to pathogenic processes. It is unclear what impact chronic alterations in gene expression could have on human health, but it has been suggested that long-term, chronic changes in gene expression can hasten the progression of diseases like diabetes (Hudder et al., 2007). In light of the role of miRNAs in metabolism and signal transduction, chronic oxidative stress or other cellular stress can initiate cellular reprogramming through alterations in miRNA expression or action. Furthermore, changes in miRNA expression in feed-forward loops may lead to sustained changes in gene expression and cellular physiology.
The rapidly progressing field of miRNA research has delineated new avenues of research and provided new insights on control of developmental pathways and disruption of processes that lead to tumorigenesis. Can miRNAs provide a link to environmental influences on gene expression?
The application of microarray technology to toxicology has spawned the discipline of toxicogenomics (Pennie et al., 2001). Although much useful data from studies of gene expression profiling in response to toxins and toxicants has been generated, a clear signature of exposure is often lacking. Similar to what has been revealed in tumor profiling, perhaps miRNA profiling in response to toxic compounds will provide toxicant-specific profiles.
It has been long observed that enzymes involved in xenobiotic metabolism such as the cytochrome P450 (CYP) family members, CYP 2B and 2E are posttranscriptionally regulated (de et al., 1995). Earlier work in our laboratory identified the 3′-UTR as having a role in the stability and translational efficiency of the CYP2E1 mRNA (Kocarek et al., 2000). With the discovery of miRNAs, these translationally regulated mRNAs are being evaluated for regulation by miRNAs. CYP 1B1 was found to be regulated by miR-27b in MCF-7 cells in culture (Tsuchiya et al., 2006). Recently we have identified miRNAs that are responsible for the downregulation of CYP2E1 expression in primary rat hepatocytes and experimentally confirmed their function (Overton, Hudder and Novak, unpublished data). It has been suggested that other CYP family members may be regulated by miRNAs (Ingelman-Sundberg et al., 2007), although experimental validation is needed.
In a recent study, Moffat et al. (2007) looked at the effects of dioxin treatment on miRNA in mice, dioxin-resistant rats (Han/Wistar; Kuopio) and dioxin-sensitive rats (Long–Evans; Turku/AB). Although few changes were observed it is interesting to note that the dioxin-sensitive rats had more affected miRNAs, but the overall fold changes were small. p53 is a known molecular target of dioxin and its expression is downregulated in response to treatment (Tijet et al., 2006). Because p53 has recently been shown to be a transcriptional activator of miRNA expression, it stands to reason that dioxin treatment may result in reduction of expression of selected miRNAs. The authors did observe a slight downregulation of expression of a subset of miRNAs, including the p53 target mir-34, shortly after dioxin exposure (Moffat et al., 2007). Is this biologically relevant? It is possible that changes in expression level of miRNAs of less than twofold could have substantial biological effects due to the number of potential targets regulated by individual miRNAs (Calin and Croce, 2006). In one study, cells treated with gamma-irradiation showed no alteration in miRNA expression level, but exposure to sodium arsenite resulted in significant increases in miRNA levels (Marsit et al., 2006). In another recent report, the hepatotoxicants acetaminophen and carbon tetrachloride were shown to cause changes in expression of a number of miRNAs in rat liver (Fukushima et al., 2007). These changes occurred within 3h of exposure during the early phase of toxicity, prior to cellular necrosis. Two of the miRNAs involved in the toxicant response were miR-298 and miR-370, which are known to bind to and regulate expression of genes involved in the cellular response to oxidative stress (Fukushima et al., 2007). Metal sulfates have been shown to generate reactive oxygen species and trigger the expression of specific miRNAs (Lukiw and Pogue, 2007). Thus miRNAs may play an important role in the cellular response to toxicants.
In light of the mechanism of action of miRNAs in cells undergoing stress, the action and localization of miRNA-mRNA complexes may be of equal importance as their level of expression in toxicology. Most studies focus on the level of expression of miRNAs by microarray profiling and quantitative PCR methods rather than the binding of miRNAs to targets or their subcellular localization. New methodology will need to be developed to better evaluate the impact of miRNAs on translational regulation of specific mRNAs in response to environmental agents.
Epigenetics is defined as heritable changes in gene expression that do not involve alterations in DNA sequence. The primary types of epigenetic control are by histone modifications that alter chromatin accessibility and DNA methylation of CpG islands that silence gene expression (Reamon-Buettner and Borlak, 2007). Alterations in both histone deacetylation and methylation by chromatin-modifying drugs have been shown to result in alterations in miRNA expression (Chuang and Jones, 2007; Saito et al., 2006), suggesting that miRNAs are susceptible to epigenetic reprogramming. Several studies have identified alterations in methylation patterns in human cancers that result in changes in miRNA expression (Calin and Croce, 2007; Esteller, 2007; Weber et al., 2007). About half of miRNA genes have been shown to be associated with CpG islands, providing targets for the DNA methylation machinery and in fact, a high frequency of miRNA gene methylation has been found in both normal and cancer cells (Weber et al., 2007). Conversely, miRNAs can regulate factors involved in epigenetic control. Two miRNAs have been reported to target histone deacetylase 4, miR-140 during bone development (Tuddenham et al., 2006), and miR-1 during myoblast differentiation (Chen et al., 2006).
Such modifications can have immediate and long-term effects on the function of an organ or the health of an individual. Epigenetic modifications of germline cells will have transgenerational impact. piRNAs are small RNAs that function like miRNA, but are expressed exclusively in the germ cells and therefore could effect transgenerational epigenetic alterations in gene expression during development. Alterations in the pattern of expression of miRNAs in response to toxic chemical exposure in utero can result in sustained perturbations of function, developmental abnormalities or oncogenesis. It is interesting to note that congenital defects in cardiac and skeletal development associated with in utero exposure to xenobiotics such as valproic acid and dimethadione may be mediated through epigenetic modification of histones and possibly miRNAs (Weston et al., 2006).
The field of research spawned by the discovery of miRNAs is proceeding at a very rapid pace with over 800 publications on the topic in the past year alone. Writing a comprehensive review of the subject is a daunting task and it is impossible to give comprehensive coverage to all of the research that has contributed significantly to our current knowledge in the area. In this review we have highlighted some of the significant discoveries in defining miRNAs, their biogenesis, transcriptional regulation of expression and identification of miRNA targets, including a list of some of the available databases for up-to-date online information about miRNAs and their targets. The role of miRNAs in translational regulation and degradation of specific mRNA targets is discussed in addition to the formation and function of P-bodies and SGs. As there are already many reviews available on the role of miRNAs in cancer and other diseases, we offer only a brief summary of what is known about miRNAs and tumorigenesis. We have included an overview of methods used in analysis of miRNA expression and function and the application of miRNAs as both therapeutic tools and targets. The role of miRNAs in modulation of insulin signaling is intriguing as it provides a novel link between endocrine control of energy homeostasis and gene expression. As we learn more about the function and regulation of miRNAs, a new world of potential therapeutic targets is revealed. The evidence that the expression of miRNAs is affected by oxidative and other forms of cellular stress suggests an important role of miRNAs in toxicology. Indeed some very recent work has demonstrated that miRNA expression and function is altered by known toxicants. These observations lead us to consider new questions such as: does exposure to drugs, toxicants or carcinogens lead to sustained alterations in miRNA expression levels? Does sustained exposure lead to cumulative or progressive alterations in miRNA expression patterns? Do altered miRNA expression patterns correlate with altered phenotypes? This is an especially intriguing aspect of miRNA biology. If a toxicant alters miRNA expression or action, the net effect on the cell can be great due to the number of possible targets that individual miRNAs may regulate. The fact that miRNAs are involved in complex networks of regulation of gene expression in both feed-forward and feedback loops suggests that alterations in miRNA expression levels, cellular location or action can have far reaching effects on cellular physiology and even sustained alterations in cellular function. Long-term xenobiotic exposure, including in utero exposure, may result in the progressive reprogramming of the miRNA profile giving rise to an altered phenotype. Furthermore, it is possible that such chronic exposure can result in alterations in miRNA expression and function that lead to transgenerational effects through changes in germline DNA methylation patterns and posttranslational effects on histone proteins.
We wish to thank Jennifer Ortwine for her expert work in generating the diagram of miRNA processing and function and for her assistance in manuscript preparation.
FUNDING National Institutes of Health grants (ES 03656); and EHS Center Grant (P30 ES06639).