MicroRNAs: A new class of gene regulators

Abstract

Elucidation of the molecular basis of disease depends upon continued progress in defining the mechanisms by which genomic information is encoded and expressed. Transcription factor‐mediated regulation of mRNA is clearly a major source of regulatory control and has been well studied. The more recent discovery of small RNAs as key regulators of gene function has introduced a new level and mechanism of regulation. Mammalian genomes contain hundreds of microRNAs (miRNAs) that each can potentially downregulate many target genes. This suggests a new source for broad control over gene regulation and has inspired extensive interest in defining miRNAs and their functions. Here, the identification of miRNAs, their biogenesis, and some examples of miRNA effects on biology and disease are reviewed and discussed. Emphasis is placed on the possible role for miRNA in nervous system development, function, and disease.

• Gene regulation
• microRNAs
• molecular mechanism of disease

Introduction

Regulation of gene expression has been extensively studied across organisms. In particular, transcriptional regulation has been very well studied for its role in converting genomic information to molecular and biological effects. Transcriptional regulation mechanisms and paradigms originally defined in viruses and bacteria, such as the lambda switch, set the stage for more detailed understanding of complex transcriptional regulation in animal species 1. Regulation of mRNA expression is critical for normal development and function of organisms and plays a major role in disease etiology. Moreover, changes in spatial and temporal regulation of genes is a likely source of the differences among species 2.

It is also apparent that we are still limited in our understanding of how gene regulation is encoded within the genome. In fact, despite the completion of the genome project, we are not able to accurately predict which sequences are critical for regulation and transcription factor binding. Moreover, while we have become quite adept at defining the transcriptional profile within a cell, it should be noted that protein profiles are often distinct from mRNA profiles 3. There likely exist many other mechanisms of both pre‐ and posttranscriptional gene regulation that are critical to development, function, and disease.

The recent discovery of small RNAs that function to alter gene expression has defined a new, unexplored layer of genome regulation 4. An early consequence of this discovery has been the rapid appreciation and application of RNA interference (RNAi) as a tool in basic research and a potential therapeutic agent against disease. These efforts to harness the power of RNAi and small interfering RNAs (siRNAs) have been closely followed by efforts to understand the role of endogenous RNAs in the global regulation of genes. A large class of small RNAs, called microRNAs (miRNAs), has received the most attention. This is due to the fact that there are hundreds of miRNAs, and each of them can potentially regulate hundreds of genes. In this review we will provide an overview of miRNA mechanisms and biological function, with an emphasis on their role in the neural systems.

How were miRNAs discovered?

In 1993, Ambros and colleagues cloned the gene responsible for the lin‐4 phenotype in the roundworm Caenorhabditis elegans and discovered that it was a small noncoding RNA 5. They speculated that lin‐4 regulated the lin‐14 target gene by binding to a region of its 3′ untranslated region (UTR), causing posttranscriptional downregulation 5, 6. While the mechanistic implications of this discovery appear to be obvious in hindsight, it was not until the cloning of another miRNA, let‐7, and demonstration of its conservation throughout the animal kingdom, that the full implications of this work became clear 7, 8. Moreover, other work began to demonstrate that RNAs can have potent and specific effects on gene regulation via a mechanism now called RNAi. Fire, Mello, and colleagues first described RNAi in C. elegans whereby the introduction of a double‐stranded copy of a gene could result in the degradation of the endogenous gene product 9, similar to unexplained effects of transgenic ‘cosuppression’ of petal color reported in petunias 10–12. Work in mammals demonstrated that smaller RNAs could be used to achieve the same gene knockdown effects while avoiding nonspecific responses seen with introduction of larger RNAs 13. These small interfering RNAs (siRNAs) share mechanisms of gene regulation with miRNAs as discussed below.

Key messages

• MicroRNAs (miRNAs) are novel regulators of gene expression.

• Hundreds of miRNAs have the potential to regulate thousands of target genes.

• The impact of miRNA control is widespread, including the regulation of nervous system development and function.

What are the mechanisms of miRNA biogenesis and action?

miRNAs are ∼22‐nucleotide noncoding transcripts that are processed from much longer transcripts termed primary miRNAs, or pri‐miRNAs 14. These pri‐miRNAs are transcribed by RNA polymerase II, sometimes as long polycistronic complexes, and contain a 7‐methylguanosine cap along with polyadenylation sequences that are found on most mRNAs 14–16. Initially thought to be primarily located in intergenic clusters within the genome 17, 18, it was determined recently that over 50% of mammalian miRNAs are located in introns of coding genes, suggesting possible coregulation at the level of transcription 19. More recently it has been shown that some human miRNAs are interspersed with Alu repeats and are transcribed by RNA polymerase III 20. As more miRNAs are identified, it is possible that a number of subclasses will exist, each with different genomic organization and possibly different mechanisms of biogenesis. This is exemplified by the recent detection of intronic miRNAs called mirtrons that bypass the Drosha cleavage step, yet still undergo nuclear export and Dicer‐mediated processing 21, 22.

pri‐miRNAs contain a local stem loop structure that encodes the functional miRNA sequences in the stem. This stem loop structure is cleaved by the nuclear RNAse III type enzyme Drosha, as part of a ∼650‐kDa complex called the Microprocessor that also contains the RNA‐binding protein Pasha/DGCR8 23–27. These processed ∼70‐nucleotide fragments of RNA, called precursor miRNAs or pre‐miRNAs, are transported from the nucleus to the cytoplasm via Exportin 5, a nuclear export factor that specifically binds to the pre‐miRNA in a Ran‐GTP dependent manner 28–30. Within the cytoplasm, a second cleavage event occurs via the RNAse III enzyme Dicer, which forms the mature ∼22‐nucleotide double‐stranded miRNA. Dicer also processes siRNA from dsRNA as part of RNAi pathways 31–35. Like Drosha, Dicer associates with dsRNA‐binding domain containing a protein partner known as Loquacious/TRBP (transactivator RNA binding protein) 36–38. The resulting RNA duplex is often referred to as the miRNA/miRNA*, where the miRNA or ‘guide’ strand is preferentially loaded into the RNA‐induced silencing complex (RISC), whereas the miRNA* or ‘passenger’ strand is likely degraded. The details of RISC loading have been best demonstrated for siRNAs, where the choice of guide strand depends on local thermodynamic stability; the strand whose 5′ end is less stably paired is preferentially loaded into RISC 39, 40. The main constituent of the RISC are members of the Argonaute (AGO) family, which in humans include eight members, four which are ubiquitously expressed and are associated with siRNAs and miRNAs, and four which seem restricted to expression in the germ line and whose function is poorly understood. Argonautes are ∼100‐kDa highly basic proteins consisting of a Piwi‐Argonaute‐Zwille (PAZ) domain that mediates nucleic acid binding, a middle domain (MID) determined through X‐ray crystallography to be critical for the association between the RNA and AGO, and a PIWI domain that is likely to contain ribonuclease activity 41, 42.

While RNAi/siRNA mechanisms certainly overlap with those of miRNA, there are likely distinctions that reflect differences in their effects. siRNAs designed for RNAi usually have 100% homology with target genes, which results in mRNA degradation through cleavage by AGO2 43. In contrast, miRNAs generally cause their negative regulation of gene expression by translational repression of the target, which may be due to the imperfect base‐pairing to their target sequences 44. However, evidence from the fruit fly Drosophila melanogaster suggests that different isoforms of Dicer are used for either siRNA or miRNA processing, and that different AGO subunits mediate the action of either miRNA or siRNA 45, 46. These results highlight a molecular means by which the fate of miRNA targets would be decided that may involve base‐pairing or might be dictated by the composition of the associated RISC, or both. Readers are referred to recent reviews that outline the current models for miRNA function for a more detailed account of known mechanisms 47, 48. It is clear that individual miRNAs can cause mRNA degradation, translational repression, or both, and analysis of target message and protein levels is essential for assessing the function of a miRNA.

While great strides have been made in elucidating the biogenesis and mechanism of action, there is much more to learn. For example, in addition to AGO proteins, a number of other proteins have been found to be associated with the RISC: such as fragile X mental retardation protein (FMRP), vasa intronic gene (VIG), tudor staphylococcal nuclease, R2D2, the Gemin family of helicases, Armitage, and the helicase RCK/p54 49–55. Much like the complex transcriptosome that controls transcription of mRNA 56, the RISC will likely take many years of biochemical and biophysical study before being fully described and understood. A major consideration is also the cellular context within which these processes are occurring. It is likely that cell‐type‐specific subunits of RISC exist that may lead to differences in miRNA function. Subcellularly, miRNAs and their target mRNAs are found in P‐bodies, small foci that are enriched in factors involved in mRNA decay and translational repression 57, 58. Interestingly, mRNAs targeted to a P‐body are not absolutely destined for degradation and can be shuttled in and out to initiate or alleviate translational repression 59. For example, the cationic amino acid transporter 1 (CAT‐1), a target of miR‐122, exits the P‐body via 3′‐UTR‐mediated binding to HuR, an AU‐rich element binding protein 60. Although the number and size of P‐bodies are correlated to global miRNA biogenesis, it is also clear that visible P‐bodies are not essential for miRNA function 55.

How many miRNAs are present in the genome?

The demonstration of widespread conservation of let‐7 and its target lin‐41 suggested a common role for small RNA function in regulating gene expression throughout the metazoan kingdom 7, 8. The RNAse III enzyme Dicer is essential for the processing of siRNAs as well as the small temporal RNAs found in C. elegans, and Dicer‐processed products have characteristic features. These features formed the basis of screening efforts to clone and sequence small RNAs 17, 18, 61. Briefly, small RNAs were separated from total RNA and DNA linkers were added to capture the specific products of Dicer cleavage. Reverse transcriptase polymerase chain reaction (RT‐PCR) was performed, the products were ligated together to form a concatamer, and finally cloned and sequenced. A variant approach was to generate cDNA from total RNA and size‐select for small cDNAs, followed by cloning and sequencing. These studies indicate that a large number of miRNAs are present in worms, flies, and mammals.

These studies also indicate that miRNAs have tissue‐specific expression patterns. Distinct miRNAs are found in different mouse tissues, also providing evidence that individual miRNAs are expressed at greatly different levels within a cell. In fact, some miRNAs appear to be highly enriched in specific tissues or cell types, such as miR‐1 (heart), miR‐122 (liver), and miR‐124a (brain) 62. Strategies continue to be developed to allow for isolation of new miRNAs 63. For example, a modification of the SAGE (serial analysis of gene expression) method was developed that increased the number of RNA tags per clone and allowed for significantly more coverage of the microRNAome, creating the potential to discover additional novel miRNAs 64. Finally, microarray analysis and other approaches are being applied to better describe, in high‐throughput fashion, the tissue distribution of miRNAs 65, 66.

It should be noted that identification of these sequences is not sufficient to warrant classification as a new miRNA since it could belong to a family of other small RNAs, such as endogenous siRNAs. A series of guidelines have been produced suggesting that a new miRNA candidate meet both expression and biogenesis criteria 67. Expression can be determined by Northern blotting or by finding the transcript in a size‐selected cDNA library, whereas biogenesis criteria require prediction of a fold‐back precursor structure, phylogenetic conservation, or reduced expression in a Dicer‐deficient context.

While the analysis of mutants was pivotal for the initial identification of miRNAs, the characterization of additional mutants has only led to a handful of new miRNAs compared to the above‐described sequencing approaches. In addition to lin‐4 and let‐7, lsy‐6 has been found in worms, and bantam, mir‐14, and mir‐278 were identified in flies 68–71. There are a number of reasons why so few mutations have been found in miRNAs. First, the small size of the miRNA provides a smaller ‘target’ for random mutations. Second, research has focused on understanding sequence changes in protein‐coding regions that would result in clear functional changes. It is likely that new mutations will be found, and polymorphisms that were previously characterized and considered nonfunctional (e.g. in the 3′UTR) may prove to be due to changes in miRNA binding sites. Finally, functional redundancy among miRNAs may limit phenotypes caused by mutations in a single miRNA 72.

The availability of sequenced genomes along with the large number of newly identified miRNAs provided enough common structural features to begin finding miRNAs using strictly bioinformatic approaches. Using a computational procedure termed MiRscan, which considers phylogenetically conserved stem loop precursors to detect candidate miRNAs in the genome, the existence of 200 to 255 human miRNAs was predicted 73, which would account for nearly 1% of all predicted genes in humans. This was expected to be an underestimate 74, and more recent predictions suggest an even greater number of miRNAs in humans 75, 76. Large‐scale sequencing and bioinformatics identification also predict upwards of 800 miRNAs in humans 77, 78. As with the cloning‐based identification strategy, these new miRNAs need to fulfill both the expression and biogenesis criteria. In addition, using solely bioinformatics leads to difficulty in defining the exact 5′ and 3′ ends of the processed miRNA. Advances such as the determination of the molecular basis of recognition of pri‐miRNAs by the Drosha and DGCR8 complex will aid in making better predictions of exact miRNA sequences from genomic sequence 79.

These bioinformatics studies also found numerous miRNAs that are only present in primates, indicating the existence of more recently generated miRNAs. The presence of species‐specific miRNAs suggests that miRNAs are evolving at a considerable rate. Interestingly the conserved miRNAs tend to be expressed at higher levels within a cell and have fewer predicted targets, while the nonconserved miRNAs tend to be expressed at lower levels and have more targets 4. This suggests an evolutionary interplay between the miRNA and its targets and dramatic effects of miRNAs on molecular evolution.

Recently, the use of ‘deep sequencing’, using either massively parallel signature sequencing or high‐throughput pyrosequencing, has been used to identify an even greater number of miRNAs 66, 80, 81. These studies characterized 100,000s of sequences and found novel miRNA transcripts. Although these approaches have the potential to uncover even the most rare transcripts, they have certain limits due to base composition of the miRNAs or posttranscriptional modifications that may impede isolation. The need to archive and organize this data is also being met with the generation and refinement of a miRNA database, miRBase 82. As of this writing there are 474 accepted human miRNAs, 377 found in mouse, and 337 found in zebrafish using a standardized definition of a miRNA. However, it appears that more miRNAs will be found using the strategies described above.

What do miRNAs do?

Identifying miRNA targets

The discovery of novel miRNAs lends itself to high‐throughput analysis. However, as with targets for transcription factors, defining miRNA target genes is challenging. While early studies on lin‐4 and let‐7 relied on known pathway components for identification of targets 5–7, more objective techniques are needed to identify all potential miRNA targets. For example, bantam was the first miRNA where bioinformatics was used to predict a target (hid) based on sequence complementarity alone 83. Using sequence analysis as the primary way to screen for potential targets, a series of papers demonstrated that this could produce lists of potential target genes 84–88. An important finding from these studies is the importance of the 5′ end of the miRNA, especially nucleotides 2–7, in directing pairing with the target. This region is referred to as the ‘seed sequence’ 82.

Although there were some experimental validations performed on predicted targets, these target prediction algorithms did not generate largely overlapping lists using the same miRNA. Because the seed sequences used were so small, minor differences in the algorithm could lead to such variation. In addition, these data sets use different methods for 3′UTR annotations, conservation, and alignment 89, 90. It is clear that each miRNA has the potential for having multiple targets, and individual target genes frequently have target sites for multiple miRNA in their 3′UTR. Another strategy to predict targets seeks to identify an inverse relationship between the expression of a miRNA and the potential target. Overexpression of miR‐124 and miR‐1 identified over 100 genes subsequently downregulated. In fact the cells took on the expression characteristics of the tissue in which the miRNA was expressed, showing the role of these tissue‐specific miRNAs in cell differentiation 91. These studies are limited to the identification of mRNA targets that are destabilized by miRNA action and will miss the targets that are strictly translationally repressed. Analysis of miRNA and gene expression data sets has suggested that some mRNAs expressed in the same tissue as a given miRNA have acquired mutations in their 3′UTR to avoid miRNA‐mediated repression 92, 93. These ‘antitargets’ are particularly interesting and suggest a significant role for miRNAs in influencing the evolution of gene expression.

Using information derived from target validation studies, a new generation of target prediction algorithms has generated a number of new target gene search engines (reviewed in 90). While these may share common theoretical elements, it is clear that there are multiple ways to predict miRNA targets, and the field has not coalesced around a single prediction program 94. A further refinement of a target prediction algorithm has indicated that sequences outside of the critical seed sequence contribute to miRNA targeting specificity 95. A limitation of this prediction strategy is the incomplete identification of 3′UTR sequences for all the genes. Moreover, the secondary structure of the target 3′UTR might have to be considered as part of the prediction program 96–98. However it is likely that an average of 200 targets per miRNA exist, which suggests that 30% or more of the mammalian genome is regulated by miRNAs 99.

How many predicted mRNA targets are biologically relevant? Genetics studies in flies and worms have suggested that a single dominant target may exist for certain miRNAs 100. It has been demonstrated in some cases that a phenotype caused by loss of a miRNA can be almost fully rescued by reducing the levels of a single target 101–103. In contrast, it was recently shown in mice that miR‐181a regulates T cell sensitivity by modulating a number of distinct phosphatases 104. The preponderance of predicted targets and the potential for combinatorial control have led to a thoughtful model in which miRNA sites act analogously to rheostats. This level of fine‐tuning allows for precise control of gene expression and is consistent with the relatively minimal regulation of reporter constructs containing miRNA binding sites 105. It is entirely possible that the number of targets will be different depending on which model system is used, which miRNA is being studied, and which tissue type is under investigation. As with transcription factors, confirmation of physiologically relevant targets is likely to be one of the greatest challenges in this field. To add an additional layer of complexity, miRNAs undergo RNA editing, and the substitution of a single nucleotide can alter the entire repertoire of predicted targets 106.

miRNA function in development, cell survival, and metabolism

The identification of large numbers of miRNAs via sequencing and bioinformatics has far surpassed our ability to discern their individual functions. However, studies have thus far indicated that miRNAs are involved in a multitude of cellular processes. The aberrations seen in the original miRNAs, lin‐4 and let‐7, indicated a role in animal development. Other studies have indicated roles for miRNAs in specific cell or tissue types, including the development and specification of myoblasts, adipocytes, and hematopoietic and neuronal stem cells 107–110. For example, Drosophila with mir‐1 mutations have been shown to have defects in cardiogenesis and muscle development 111, 112. Mice with a targeted mutation in the ortholog, mir‐1‐2, have profound defects in cardiac development, highlighting the conservation in function for this miRNA throughout metazoans 113. Moreover, mir‐208 plays a role in cardiac stress responses 114. Knockout experiments have also established a role for mir‐155 in immune response and lymphocyte function 115, 116. These papers likely represent the beginning of a series of targeted deletion papers that will help establish the diverse roles for miRNAs in development and adult organ function.

More generally, miRNA activity has dramatic effects on cell survival and growth as well as metabolism. The bantam miRNA was found in Drosophila to be involved in mediating apoptosis through its interaction with the cell death gene hid69. mir‐14 was similarly found to be a cell death suppressor in flies, but its deletion also led to increased levels of triacylglycerol and diacylglycerol, indicating a role for miRNA regulation of metabolism 70. mir‐278 was found as a gain of function mutation involved in cellular growth in flies, but a targeted mutation in the gene also showed an opposite metabolic phenotype to mir‐14, with a reduction in total body triglycerides 71. Another important miRNA for metabolic function is miR‐375, which regulates insulin secretion by the beta cells of the pancreas 117. In contrast to the above studies using genetic mutations, this result was generated using antisense 2′‐O‐methyl oligoribonucleotides directed towards miR‐375. For in vivo miRNA manipulation, these oligoribonucleotides have been modified to obtain cell‐permeable characteristics and are called antagomirs 118. The results suggest that miRNAs are likely to regulate cellular and animal physiology across a broad range of species.

miRNA function in the development of the nervous system

Studies have indicated that miRNAs are critical for the normal specification and function of the nervous system. The lsy‐6 miRNA was found in a genetic screen in worms for mutants affecting left/right neuronal asymmetry and is only expressed in ∼10 neurons 68. Further work has shown that lsy‐6 and its counterpart miR‐273 are part of a regulatory cascade dictating which transcription factors are expressed in either the left or right neuron 119. Evidence of a more general role for miRNAs in the development of the nervous system comes from analysis of mutations affecting components of the miRNA processing pathway. Drosophila that are missing the maternal and zygotic contributions of AGO1 have broad defects in neural and glial development 120. AGO1 appears to solely mediate miRNA function in flies, thus suggesting a role for miRNA‐mediated regulation in neural development 46.

When miRNA processing was abolished in zebrafish using maternal‐zygotic dicer mutants, the general patterning of the organism seemed intact but there were severe defects in a few organ types, particularly the brain. The authors went on to demonstrate that the miR‐430 family was responsible for these defects since replacement of this family reversed a brain defect but not the ear and heart defects 121. Identical experiments are not possible in mice due the fact that the dicer loss‐of‐function mice die at embryonic day 7.5; however, conditional mutants have been generated and may be used to address the involvement of miRNAs in numerous tissues 122, 123.

More evidence suggesting that miRNAs are involved in regulating neuronal differentiation comes from the analysis of miR‐9, a highly conserved miRNA expressed in the nervous system. In Drosophila, mir‐9 mutant flies have ectopic neuronal precursors, whereas overexpression of miR‐9 leads to diminished neuronal precursors 103. The results suggest that miR‐9 plays an essential role in regulating the precise number of neuronal precursors during development. The expression of nervous system‐enriched miRNAs was analyzed during the course of mammalian embryonic stem cell‐derived neurogenesis using arrays, and miR‐9, along with miR‐124a, is highly regulated. Either gain‐ or loss‐of‐function manipulations of miR‐9, or gain‐of‐function manipulation of miR‐124a, led to alterations in neuronal or glial cell differentiation 109. However, an independent analysis of miR‐124a in the developing neural tube failed to show an effect on neuronal differentiation, potentially indicating specificity in miR‐124a‐mediated regulation of differentiation 124.

How do miRNAs interact with other master regulators, such as transcription factors, in control of neural development? Studies with known transcription factors, cyclic adenosine monophosphate response element binding protein (CREB), and Repressor Element‐1 silencing transcription factor (REST), have shown that miRNAs expressed primarily in brain‐related tissue are highly enriched for binding sites for these transcription factors 125. In nonneuronal cells, REST inhibits miR‐124a expression, whereas during differentiation REST leaves the mir‐124a locus, and this leads to the decrease of nonneuronal transcripts by miR‐124a targeting 126. miR‐132 was identified as a target of CREB, and overexpression causes neurite outgrowth, whereas loss of function led to attenuated neuronal outgrowth 127.

Evidence for miRNA function in neuronal plasticity and behavior

Perhaps the most intriguing hypothesized function of neuronal miRNAs is an involvement in synaptic plasticity 128. There is significant interest in the localized translation of protein at synapses and how this may relate to synaptic plasticity 129. Components of the miRNA processing pathway, such as Dicer, have been found in postsynaptic densities in mouse brain 130. Components of RISC such as Armitage have been shown to be localized to dendritic spines in flies. Armitage is degraded during the acquisition of long‐term memory allowing for the synaptic translation of Ca2+ calmodulin‐dependent kinase II 131. miR‐134 is expressed at the developmental time when synapses are forming, and manipulation of miR‐134 function causes defects in dendritic spine volume 132. It will be particularly interesting to see how loss of miR‐134 function alters synaptic plasticity and related behavior in an intact mouse. These observations show that miRNA silencing is taking place at the synapse and may be a major determinant of localized translation found at the synapse. These cellular results are complemented by the suggestion that miR‐196 functions via regulation of the transcription factor Hoxb8 to control grooming behavior in rodents 133–135. More work will be needed to connect the effects of miRNAs on neurons with eventual control of animal behavior.

Are miRNAs associated with disease?

Considering the biological effects of miRNA manipulation and the large scale of potential gene regulation by miRNAs, a strong possibility exists that miRNA or target mutations have the potential to underlie human disease. The most extensive research in this field has been in cancer, originating with the suggestive finding that mir‐15a and mir‐16‐1 are located in a chromosomal region frequently deleted in patients with a common form of adult leukemia, B cell chronic lymphocytic leukemia 136. Further support for the idea that miRNAs act as tumor suppressors comes from studies on let‐7 and a target, the oncogene Ras 137. Activating mutations in Ras that result in increased expression causes cellular transformation. let‐7 family members are downregulated in human lung cancer, and it has been speculated that this leads to increased severity of cancer due to misregulation of Ras 138. Another example of let‐7 repression of an oncogene concerns a well known translocation near the High mobility group A2 (Hmga2) gene, which is normally targeted by let‐7 but becomes misregulated when its 3′UTR target sequence is deleted due to the translocation 139, 140. miRNAs may also act as oncogenes (e.g. miR‐21 is upregulated in glioblastomas and breast cancer); however, additional studies will be required to show a direct function in cancer progression 141–143. In addition to these functional studies, the expression profile of miRNAs has been examined in a variety of tumors, and it has been suggested that the miRNA expression profile may be a more effective diagnostic tool than traditional mRNA expression profiles 144. As an indication of the widespread potential for miRNAs in cancer, it has been reported that over 50% of miRNAs map to regions in the genome associated with chromosomal locations linked to cancer 145.

In addition to cancer, some neurological diseases have been associated with miRNA function. An example is fragile X syndrome, one of the most common forms of inherited mental retardation. It has been shown that a CGG repeat expansion in the 5′UTR of the fmr1 gene, which leads to methylation‐induced transcriptional inactivation, is the underlying cause of this syndrome in greater than 95% of the cases 146. fmr1 encodes FMRP (fragile X mental retardation protein), a selective RNA‐binding protein that associates with miRNAs, Dicer and Argonaute subunits 49, 51, 147. It has been proposed that FMRP mediates miRNA‐mediated translational repression, and defects in this process may lead to aspects of the phenotype of fragile X syndrome 148.

Recent analysis of brain region‐specific reduction of Dicer has suggested a role for miRNAs in mediating neurodegenerative disorders 149. Purkinje cell‐specific ablation of Dicer leads to progressive loss of miRNAs, which leads to cerebellar degeneration and development of ataxia. These results indicate that miRNAs are involved in the maintenance as well as the development of the nervous system, and efforts directed at demonstrating a role in human degenerative diseases are underway.

The genetics of human disease has demonstrated that many diseases are polygenic with small contributions arising from many genes. There are large‐scale efforts to define all of the single‐nucleotide polymorphisms (SNPs) present in human populations in order to increase mapping power as well as identify alleles that may directly contribute to disease. It is possible that some of these SNPs will modulate the binding or effects of miRNA regulation, since a single base mismatch in a miRNA target sequence can have a dramatic effect on targeting 150. In fact, recent analysis of public databases revealed little variation in miRNA sequence but did find hundreds of examples of variation in miRNA target sites 151. Some of these changes would lead to a loss in miRNA binding, whereas others would create a miRNA target site. A clear example of this comes from studies of muscularity in sheep. A quantitative trait‐mapping study identified a polymorphism in the 3′UTR of the myostatin (gdf8) gene in Texel sheep 152. This base pair change results in the generation of a miRNA binding site, subsequent downregulation of myostatin, and increased muscularity in Texel sheep. Further analysis was conducted on human 3′UTR sequences in databases to demonstrate that that many natural sequence polymorphisms result in the creation or disruption of a miRNA binding site, suggesting this as a general phenomenon that could be relevant to many disease states.

An example of how this type of polymorphism may relate to human disease was suggested by the study by Abelson et al. 153. They identified Slit and Trk‐like family member 1 (SLITRK1), a leucine‐rich transmembrane protein, as a candidate for involvement in Tourette's syndrome based on a de novo chromosomal inversion. This inversion causes a frameshift in SLITRK1, and they identified two independent probands with a mutation in the 3′UTR in a miR‐189 binding site. This mutation would be predicted to enhance the miRNA target binding, and they present evidence that this is the case in cell culture. Effects of regulatory small RNAs might be particularly significant in the case of complex psychiatric disease, where regulatory changes are likely to be important 154. A recent study demonstrated changes in miRNA levels in prefrontal cortex from schizophrenic patients 155, thus indicating that studies of variation in binding sites in other psychiatric diseases are also likely to yield interesting data.

Discussion

We have given a broad overview to introduce the reader to miRNAs and their function in animal development and disease. While we have not discussed the extensive literature demonstrating miRNA function outside of the animal kingdom; miRNAs are clearly an ancient mechanism for regulating gene expression 156. Throughout this review miRNA regulation is compared to regulation of gene expression by transcription factors; a single miRNA or transcription factor can potentially regulate hundreds of target genes. The extensive research on transcription factors also foretells challenges, such as the identification of relevant targets, which will need to be overcome to fully appreciate the effects of miRNAs on biology and disease.

The mechanistic distinctions between miRNAs and transcription factors are also notable. First, miRNAs are mainly acting on mRNA in the cytoplasm of the cell, versus transcription factors binding DNA in the nucleus. Second, the primary effect of miRNA is to lower gene expression, whereas transcription factors can either increase or decrease expression. While these distinctions cannot be ignored, the basic concepts of regulation might be shared. In fact, the finding of multiple miRNA target sequences within the same gene is also reminiscent of multiple transcription factor binding sites seen in well studied enhancer sequences 157. The common and unique elements seen in miRNA regulation will become apparent as they are catalogued and analyzed in more detail.

While it is tempting to speculate on the universal regulatory function of miRNAs, it is likely that they will serve many roles in many different contexts. Like other large classes of genes, they will likely have been co‐opted for function in many different pathways. Some miRNAs may function to canalize, or buffer, critical genetic pathways in order to increase robustness 93, 158. Other miRNAs seem to function as developmental switches 5, 7, 68, while some play key roles in the development of major organs 113. The next few years will undoubtedly yield a large set of studies looking at miRNA function in animals. It is likely that further study of these new molecular regulators will modify our views of molecular control over animal development, function, and disease.

Acknowledgements

We would like to thank Drs Amy MacQueen and Maysa Sarhan for valuable comments on the manuscript.