MicroRNAs (miRNAs) are a class of small RNAs that are approximately 22 nucleotides in length. Hundreds of miRNA genes are encoded in the animal genome, and each miRNA potentially regulates tens to hundreds of protein-coding transcripts post-transcriptionally. Experimental and bioinformatic approaches have shown widespread regulatory roles for miRNAs in metazoa including roles in cellular homeostasis and human diseases. Since the discoveries of let-7 and lin-4 miRNAs as regulators of developmental timing in Caenorhabditis elegans, functions of miRNAs in the context of animal development have been studied in many model organisms. Although miRNAs are essential to achieve complex developmental processes, the vast majority of animal miRNA functions have yet to be determined. The identification of miRNA-target interactions and the interpretation of their biological significance are often difficult due to the divergent functions of miRNAs in intricate gene regulatory networks. This review summarizes our current knowledge on miRNA functions in vertebrate development by focusing on the progress made in the vertebrate model organism zebrafish (Danio rerio). Studies of miRNA functions in this small teleost highlight several common principles underlying the functions of animal miRNAs.
Development of the multicellular organism requires complex and multi-layered regulations of gene expression. Since the discoveries of lin-4 and let-7 in Caenorhabditis elegans (Lee et al. 1993; Reinhart et al. 2000), microRNAs (miRNAs) have been recognized as a novel layer of the gene regulatory network in animals and plants. miRNAs are approximately 22 nucleotides (nt) in length and are generated from numerous distinct genetic loci that do not overlap with protein-coding sequences. miRNAs post-transcriptionally regulate protein expression mainly via the untranslated region (UTR) of protein-coding mRNAs (Bartel 2004). Previous studies showed widespread impacts of miRNAs on most biological processes, including physiological processes that regulate cellular homeostasis and pathological processes that contribute to human diseases (Bushati & Cohen 2007; Croce 2009; Small & Olson 2011). In silico analyses have estimated that 30–70% of protein-coding genes that are encoded in the human genome are under the influence of miRNAs (Lewis et al. 2005; Rajewsky 2006; Friedman et al. 2009). These observations indicate that miRNAs govern the functionality of “non-coding regions” in the genome, which have been expanded during the evolution of eukaryotes. However, the majority of miRNA functions are unknown. This review summarizes our current knowledge on miRNA functions in vertebrate development by focusing on the progress that has been made in the vertebrate model organism zebrafish (Danio rerio). Related studies in other animal models and the basic mechanisms of the miRNA pathway are discussed, whereas miRNA functions in plants are described elsewhere (Meins et al. 2005; Wollmann & Weigel 2010).
Principles of miRNA-mediated silencing
The animal miRNA pathway is shown as a schematic in Figure 1. In most cases, miRNAs are transcribed by RNA polymerase II as a long transcript called primary miRNA (pri-miRNA), comprising a hairpin structure that corresponds to the mature miRNA and its complementary strand (miRNA*; Lee et al. 2004). The length of the pri-miRNAs is variable, ranging from hundreds to thousands of nucleotides. Most pri-miRNAs are encoded in the intergenetic regions with their own promoters, whereas some are embedded in the intron of host genes, antisense strand of other genes, or viral genomes (Pfeffer & Voinnet 2006; Griffiths-Jones et al. 2008). These divergent origins of pri-miRNAs make it possible to express mature miRNAs in a temporally and spatially regulated manner. The pri-miRNA is first processed by the RNaseIII enzyme Drosha and its co-factor DGCR8 in the nucleus to generate a shorter hairpin RNA molecule called precursor miRNA (pre-miRNA; Han et al. 2004). The pre-miRNA is exported into the cytoplasm via Exportin-5 (Lund et al. 2004) and further processed by another RNaseIII enzyme, Dicer, and its co-factor TRBP (Rossi 2005). This reaction removes the loop of the pre-miRNA hairpin to generate an RNA duplex that is approximately 22 nt in length. The resultant miRNA/miRNA* duplex is loaded into a protein called Argonaute (Ago), forming the miRNA-induced silencing complex (miRISC), which maintains one strand of the miRNA duplex (Kawamata & Tomari 2010). A few miRNAs are generated via other processing pathways in addition to this canonical biogenesis pathway. The mirton pathway bypasses Drosha-mediated cleavage by generating pre-miRNA hairpins as spliced introns (Berezikov et al. 2007; Okamura et al. 2007). miR-451 bypasses Dicer-mediated processing by coupling miRNA processing and loading in Ago2 (Cheloufi et al. 2010; Cifuentes et al. 2010).
Perhaps one of the most intriguing features of the miRNA pathway is the recognition of target RNAs. In the cases of the RNAi pathway in animals and the miRNA pathway in plants, small RNA and its target mRNA form nearly perfect base-pairing that leads to endonucleolytic cleavage of the target RNA by catalytically active Ago (Hutvagner & Simard 2008). In contrast, animal miRNAs recognize their regulatory target via a less stringent base-pairing with several mismatches. This “loose” binding allows a given miRNA to regulate hundreds of different target mRNAs, which makes it difficult to identify bona-fide miRNA targets using a simple computational search. The sequence determinants for miRNA target recognition have been identified by combining experimental and bioinformatic approaches (Lewis et al. 2003; Doench & Sharp 2004; Brennecke et al. 2005; Lim et al. 2005). These analyses have revealed that a majority of miRNA target genes contain one or more hexanucleotide sequences that are complementary between the second to seventh positions from the 5′ end of a given miRNA (Figure 2). This miRNA region is called the seed sequence, and its complementary target site is called the seed-matched site. Seed-matched sites are further divided into subclasses in which the 8-mer site is the most effective site and the 6-mer site is the least effective site for miRNA-mediated silencing (Lewis et al. 2005; Figure 2A–C). Atypical sites such as an offset 6-mer and an imperfect seed-matched site with a 3′ supplementary site were also observed (Figure 2D). The efficacy of each target site is influenced by additional parameters such as the local sequence context and the relative position within a transcript (Grimson et al. 2007). In general, seed-matched sites that are located in the 3′ UTR and surrounded by AU-rich sequences are the preferential target sites for miRNAs because these sites are free from the translation apparatus and RNA secondary structures. In addition, the presence of multiple target sites in close proximity synergistically enhances the regulatory competence of miRNAs (Grimson et al. 2007).
Upon recognition of its target mRNA, miRISC induces post-transcriptional silencing. If the miRNA and its target mRNA form nearly perfect base-pairing, the target RNA strand may be subjected to endonucleolytic cleavage via an RNAi-like activity (Hutvagner & Simard 2008). However, this reaction requires extensive base-pairing between the miRNA and its target RNA with perfect base-pairing at the center of the miRNA. In addition, only Ago2 possesses endonuclease activity among the four vertebrate Ago proteins (Ago1-4; Meister et al. 2004). Due to these limitations, only a few animal miRNA targets that are silenced via endonucleolytic cleavage were reported (Yekta et al. 2004; Davis et al. 2005). Instead, miRNAs induce the silencing of seed-matched targets via translational inhibition (Fabian et al. 2010). The miRNA targets are also subjected to mRNA degradation that is independent of Ago2-mediated endonucleolytic cleavage (Lim et al. 2005). The miRNA-mediated mRNA degradation is probably initiated by shortening the poly(A) tail of mRNA (deadenylation), which is a widespread consequence of miRNA binding (Giraldez et al. 2006; Wu et al. 2006). Although the relative contributions of translational inhibition and mRNA degradation to the overall silencing effect of miRNAs require further validation, recent genome-wide analyses strongly suggested a model showing that most of the miRNA target mRNAs are eventually degraded to the levels that correlate with the reduction of their protein products (Baek et al. 2008; Selbach et al. 2008; Guo et al. 2010). Therefore, reductions in both protein and mRNA amounts are hallmarks of miRNA-mediated repression. The mechanisms underlying these silencing processes are under intense debate, and more details have been discussed elsewhere (Fabian et al. 2010; Huntzinger & Izaurralde 2011).
miRNAs are essential for vertebrate embryogenesis
Because Dicer is required for mature miRNA biogenesis, loss of Dicer function should inhibit production of all Dicer-dependent miRNAs. Using this approach, the global roles of vertebrate miRNAs in embryonic development have been addressed. In zebrafish, zygotic dicer mutant embryos were indistinguishable from wild type, were capable of producing mature miRNAs and were developed normally into larvae during the first week (Wienholds et al. 2003). The normal development of zygotic dicer mutant embryos is mediated by the presence of the maternal dicer transcript and Dicer protein that were deposited into the fertilized egg. Eventually, pre-miRNAs began to accumulate in dicer mutant larvae. The mutant larvae showed retarded growth and died 2 weeks after fertilization. These results suggest that miRNA functions are essential for post-embryonic development. To study miRNA functions in early embryogenesis, zebrafish dicer mutant embryos that were devoid of both maternal and zygotic dicer contributions were generated (maternal and zygotic dicer mutant, MZdicer; Giraldez et al. 2005). As expected, MZdicer embryos did not process pre-miRNAs and therefore lacked all canonical miRNAs. MZdicer embryos showed severe morphological defects from the mid-gastrulation stage onward, and died before hatching. Interestingly, MZdicer embryos developed normal anterior-posterior and dorsal-ventral axes and formed major cell lineages including muscles, neurons and hematopoietic systems. When transplanted into wild type hosts, primordial germ cells (PGCs) with the homozygous dicer mutation gave rise to functional sperms and oocytes that generated offspring over multiple generations. These observations in zebrafish suggest that miRNAs play major roles during embryonic development by regulating morphogenesis and coordinating differentiation after the establishment of fundamental cell lineages.
In contrast to these observations in zebrafish, zygotic dicer knockout mice demonstrated embryonic lethality before gastrulation (Bernstein et al. 2003). The dicer knockout embryo failed to express stem cell markers, suggesting that a loss of Dicer caused defects in stem cell development and/or maintenance. The conditional loss of Dicer activity in PGCs caused poor proliferation and retarded differentiation (Hayashi et al. 2008), whereas the loss of Dicer in oogenesis resulted in infertility (Murchison et al. 2007). In contrast, the dicer null embryonic stem (ES) cells, which were generated by conditional knockout, expressed stem cell markers and showed a morphology typical of ES cells (Kanellopoulou et al. 2005). However, the dicer knockout ES cells proliferated slowly and had defects in differentiation in vitro and in chimeric embryos. These observations suggest that miRNAs are required for the establishment of stem cells and subsequent differentiation in mice. Differential phenotypes of dicer-deficient germline cells in zebrafish and mice are attributed to the differential functions of Dicer-dependent small RNAs other than miRNAs. Indeed, mouse oocytes that are deficient for DGCR8, which is required for pri-miRNA processing, were fertilized normally and the resultant zygote developed until the pre-gastrulation stage (Suh et al. 2010). The requirement of miRNAs in mammalian germline cells therefore awaits further analysis.
Classification of miRNA functions in animal development
Given the essential role of the miRNA pathway in embryogenesis, developmental functions of individual miRNAs have been analyzed in many studies. Although the loss of some specific miRNAs causes striking phenotypes, the loss of individual miRNAs tend to result in less drastic phenotypes. Indeed, comprehensive analysis of miRNA loss-of-function phenotypes in C. elegans revealed that the majority of miRNA mutants did not show detectable phenotypes until they were combined with other miRNA mutants or with sensitized genetic backgrounds (Brenner et al. 2010; Alvarez-Saavedra & Horvitz 2011). Consistent with the relatively mild single miRNA loss-of-function phenotypes, a genome-wide analysis revealed that the repressive effect of a given miRNA on its target genes rarely exceeded a fourfold decrease at the protein level (Selbach et al. 2008). These observations suggest that although miRNAs have a capacity to regulate numerous target genes, their regulatory functions are quantitatively and qualitatively distinct from previously known regulatory mechanisms. Based on the evidence that has been presented in zebrafish, miRNA functions in animal development can be categorized into several groups as described below.
miRNAs mediate developmental switches
The earliest known role of the animal miRNAs is the regulation of developmental timing by let-7 and lin-4 miRNAs in C. elegans (Lee et al. 1993; Reinhart et al. 2000). A similar role for miRNAs was discovered in zebrafish by analyzing MZdicer embryos. In zebrafish, as well as in most animals, early development of the embryo largely relies on maternal factors, which are proteins and mRNAs that are stored in the egg. Maternal mRNAs are degraded during the period called maternal to zygotic transition (MZT), which denotes the time when the zygotic genome is activated for de novo transcription (Schier 2007). This massive degradation of old mRNAs is crucial to initiate developmental programs that are under the control of the embryonic genome. However, the molecular mechanism underlying this process had not been identified.
The zebrafish miR-430 family, which is first expressed during MZT (2.75 h postfertilization, hpf), is the most abundant miRNA family during early embryogenesis (Chen et al. 2005). Surprisingly, injection of the mature miR-430 duplex into MZdicer embryos rescued most of the phenotypes that were observed during the first 24 hours, including gastrulation defect, brain malformations, body curvature and immobility (Giraldez et al. 2005). Therefore, miR-430 is an essential miRNA during zebrafish development with striking impacts on morphogenesis. To identify mRNAs that are regulated by miR-430, mRNA profiling was performed with wild type, MZdicer and MZdicer + miR-430 embryos (Giraldez et al. 2006). This analysis successfully identified more than 200 mRNAs that were downregulated by miR-430 and contained one or more seed-matched target sites for miR-430. A detailed analysis of these transcripts revealed that more than 40% of the miR-430 targets were maternal transcripts that were degraded at MZT. Conversely, the 3′UTR of maternal transcripts displayed a fourfold increase in miR-430 seed-matched sites. These analyses illustrated that zebrafish miR-430, which is expressed during MZT, acts as a developmental switch by clearing maternal transcripts to facilitate the transition to zygotic programs.
miRNAs mediate spatial expression patterns
The miRNAs not only regulate developmental timing but also regulate gene expression spatially during embryogenesis. Earlier studies in flies and mammals revealed that miRNAs and its target mRNAs tend to be expressed in a mutually exclusive manner (Farh et al. 2005; Stark et al. 2005). Conversely, mRNAs that are co-expressed with a given miRNA are selectively devoid of the seed-matched site. Ubiquitously expressed genes such as ribosomal proteins have evolved shorter 3′UTRs than tissue-specific transcription factors, and thereby generally avoid miRNA targeting (Stark et al. 2005). The discovery of this miRNA-target mRNA expression pattern argued that miRNAs confer accuracy of gene expression patterns by acting as a fail-safe mechanism of transcriptional regulation (Figure 3B, left). Therefore, the relatively mild phenotypes of many miRNA mutants might not be so surprising because embryogenesis is well controlled by the main regulatory mechanisms even in the absence of individual miRNAs unless genetic or environmental perturbations occur.
An alternative explanation for the observed inverse expression patterns of a miRNA and its target gene is that target genes that are co-expressed with a cognate miRNA in the same domain are barely detectable due to miRNA-mediated silencing. This explanation predicted an instructive model, in which miRNAs actively modulate the transcriptional input to spatially regulate gene expression domains (Figure 3B, right). In this model, the loss of any tissue-specific miRNA would expand the expression domain of its target genes. To test the global contributions of the fail-safe and instructive regulations of miRNA target genes in tissue-specific expression patterns, a genome-wide miRNA target analysis was performed in zebrafish embryonic skeletal muscle (Mishima et al. 2009). The combination of MZdicer embryos and morpholino antisense oligos (MOs) targeting individual miRNAs (Kloosterman et al. 2007) revealed miR-1/206 and miR-133 as the most active miRNAs in the zebrafish embryonic skeletal muscle. The two miRNAs directly regulate several hundred mRNAs. The expression analysis of the muscle miRNA targets revealed two interesting observations. First, consistent with previous studies, the majority of the identified targets were expressed at lower levels in muscle compared with surrounding tissues. Conversely, the genes that were expressed at lower levels in muscle compared with those in other tissues were enriched for miR-1 and miR-133 target sites. Second, the lower expression bias in muscle was considerably weakened but not diminished following miR-1 inhibition. A detailed analysis of individual miR-1 target genes revealed that there were three classes of target genes. The miR-1 targets such as pfn2l and atp6v1ba were primarily detected in non-muscle cells in wild type embryos but were readily detectable in muscle following miR-1 inhibition. These examples represent the role of miRNAs in shaping spatial gene expression. On the other hand, some genes displayed functional miR-1 target sites in their 3′UTRs but remained excluded from muscle, even in the absence of miR-1. These genes fulfilled the criteria of fail-safe targets. Other miR-1 targets were also bona-fide targets because their expression levels were quantitatively altered by miR-1. However, this change did not correlate with the qualitative alternation of its spatial expression pattern. These miR-1 targets might represent an intermediate state between the fail-safe and shaping targets. The analysis of neuronal miR-124 and its targets in zebrafish reported similar findings (Shkumatava et al. 2009).
The shaping function of miRNAs in spatial expression patterns has been observed with other zebrafish miRNA-target pairs. The miR-10 family is encoded in the intron of hoxB4 and targets the neighboring Hox genes hoxB1 and hoxB3a. Loss of miR-10 function expanded the hoxB1 and hoxB3a expression domains into the Hox-4 expression domain, indicating that miR-10 shapes the boundaries of the Hox transcriptional network (Woltering & Durston 2008). miR-451 was required for normal erythrocyte maturation (Dore et al. 2008; Patrick et al. 2010). One of the in vivo targets for miR-451, gata2, is initially expressed during early somitegenesis in the anterior and posterior intermediate cell mass (ICM), which is the teleost equivalent of the mammalian yolk sac blood island (Detrich et al. 1995). After 19 hpf, the expression of gata2 was selectively decreased in the anterior ICM with concomitant expression of miR-451. MO-mediated knockdown of miR-451 caused the persistence of gata2 expression in the anterior ICM, demonstrating that miR-451 restricted gata2 expression to the posterior ICM (Pase et al. 2009). Other study demonstrated that miR-143 was expressed in heart ventricles and repressed the expression of retinaldehyde dehydrogenase type 2 (aldh1a2) and retinoid X receptor alpha b (rxrab) in the ventricle. Therefore, miR-143 regulates the gradient of retinoic acid signaling that is required for proper cardiogenesis (Miyasaka et al. 2011). These studies demonstrated that the miRNA-mediated instructive regulation of target gene expression is widespread in zebrafish embryos.
miRNAs dampen target gene expression levels
The genome-wide miRNA target gene analyses in zebrafish identified target genes that showed strong expression overlapping with the expression of a cognate miRNA (Giraldez et al. 2006; Mishima et al. 2009). For those targets, miRNAs appear to fine-tune expression levels by dampening their transcriptional inputs so that an optimal range is reached (Figure 3C). A good example for this class of target gene is the miR-430-mediated regulation of Nodal signaling components (Figure 3E). In zebrafish, the Nodal agonist, Squint, and its antagonist, Lefty, balance Nodal signaling during germ layer formation (Schier & Talbot 2005). Interestingly, both squint and lefty2 are targeted by miR-430. Loss of miR-430 regulation disrupted the overall balance of Nodal signaling and resulted in phenotypic consequences that indicated attenuated Nodal signaling (Choi et al. 2008).
The Sdf1a-miR-430 target pair is another intriguing example of miRNA-dependent dampening of signaling pathways (Figure 3F). Sdf1a, which is also known as Cxcl12a, is a chemokine ligand that attracts primordial gem cells (PGCs) via its receptor Cxcr4b (Doitsidou et al. 2002; Knaut et al. 2003). The precise expression of Sdf1a is essential to guide PGCs to the presumptive gonad region because Sdf1a is a potent regulator of multiple cell movements that occur in close proximity during development (David et al. 2002; Knaut et al. 2005; Hollway et al. 2007). Staton et al. (2011) revealed that miR-430 tightly regulated this signaling pathway by directly targeting sdf1a mRNA. In the absence of miR-430-mediated regulation, the expression level of sdf1a mRNA was increased and its expression domain was spatially expanded. Concomitantly, some PGCs stayed in the middle of the migration path, inappropriately migrated into neighboring tissues, or were recruited to another sdf1a expression domain that was outside of its original path. In this case, miR-430 shapes the migration path for PGCs by restricting the expression domain of sdf1a mRNA, and helps directed PGC movement by dampening the amount of Sdf1a within the range that generates an optimal signaling gradient. In addition to sdf1a, miR-430 also regulates Cxcr7b, which is a decoy receptor that sequesters Sdf1 (Boldajipour et al. 2008). The simultaneous regulation of Sdf1a and Cxcr7b by miR-430 conferred robustness for the PGC migration pathway against the effects of gene dosage alterations (Staton et al. 2011). Dampening of the signaling pathway is one of the common themes of miRNA-mediated regulation in animal development because other signaling pathways including Hedgehog, fibroblast growth factor (FGF), vascular endothelial growth factor (VEGF) and platelet-derived growth factor (PDGF) are under the control of miRNAs in zebrafish (Flynt et al. 2007; Eberhart et al. 2008; Fish et al. 2008; Leucht et al. 2008).
Although the simultaneous presence of a miRNA and its target transcript triggers silencing in theory, this is not necessarily true for particular target mRNAs. Nanos is an evolutionarily conserved RNA-binding protein (RBP) that is required for PGC migration and survival (Parisi & Lin 2000; Koprunner et al. 2001; Tsuda et al. 2003). In many animals, nanos mRNA is maternally supplied and asymmetrically inherited by a subset of cells that develop into PGCs. In zebrafish, the 3′UTR of nanos1 mRNA directs Nanos1 protein expression in PGCs via two post-transcriptional mechanisms: selective inheritance of mRNA by the PGCs, and translational silencing that is coupled with mRNA degradation in somatic cells (Koprunner et al. 2001). Notably, the nanos1 3′UTR contains a seed-matched site for miR-430. Reporter analyses revealed that the miR-430 target site in the nanos1 3′UTR was fully active to mediate silencing by miR-430 in somatic cells (Mishima et al. 2006). In addition, the nanos1 3′UTR resisted miR-430-mediated repression specifically in PGCs, thereby allowing protein expression only in PGCs (Figure 4). This activity was specific to the nanos1 3′UTR because most of the other miR-430 target genes were equally repressed both in somatic cells and PGCs. These results show that miR-430 is active in somatic cells and PGCs, and that PGC-specific mechanism(s) that bind to the nanos1 3′UTR inhibit miR-430-mediated repression. The same phenomenon has been reported with germline genes tdrd7 and hub (Mishima et al. 2006; Mickoleit et al. 2011). Two factors that are responsible for the anti-miRNA activity in PGCs have been reported previously. A germline-specific RBP, Dead end (DND), binds to its binding sites near the miR-430 seed-matched sites in the nanos1 and tdrd7 3′UTRs and protects those mRNAs from miR-430 by steric hindrance (Kedde et al. 2007). Another germline-specific RBP, Daz-like (Dazl), acts via multiple GUUC/A sequences in the tdrd7 3′UTR and counteracts the repressive effects of the miRNA, including mRNA deadenylation and degradation (Takeda et al. 2009). Analogously, the binding of the Elav-related protein ElrB1 to deadend mRNA prevents miR-18-mediated repression in Xenopus PGCs (Koebernick et al. 2010). The context-dependent regulation of miRNA target genes has been reported outside of the germline. In humans, the binding of HuR to CAT-1 mRNA confers differential regulation by miR-122 under stressed and normal conditions (Bhattacharyya et al. 2006). These observations indicate a novel class of miRNA targets, the conditional targets, which show differential susceptibility to miRNA-mediated repression in different cellular contexts.
From a technical point of view, the conditional regulation of miRNA target genes is applicable as a target protector, which was first developed by Giraldez and colleagues (Choi et al. 2007). In this case, MOs were used to block miR-430 binding to its target mRNAs, lefty and squint, by specifically masking the miR-430 target sites in those mRNAs. The target protector is a useful tool to inhibit a specific miRNA-target pair without affecting other miRNA-target interactions.
Conserved versus non-conserved targeting
Given the pervasive roles of miRNAs in zebrafish development, one question is whether these regulations are evolutionally conserved or not. In general, conserved target sites tend to be more effective than non-conserved sites, suggesting a strong correlation between conserved targeting and biological consequences (Farh et al. 2005; Lewis et al. 2005; Selbach et al. 2008). The genetically identified interaction between let-7 miRNA and lin-28 mRNA in C. elegans has been demonstrated in humans (Pasquinelli et al. 2000; Rybak et al. 2008). The miR-430 target sites in lefty and sdf1 are also well conserved in vertebrates (Choi et al. 2007; Rosa et al. 2009; Staton et al. 2011). In contrast to these potent targets, approximately 80% of the experimentally identified miR-430 target sites are not conserved even within closely related phylogeny (Giraldez et al. 2006). These less-conserved target sites should not be considered as less important sites because they might represent miRNA-mediated regulation of species-specific traits. For example, the targeting of the Nodal antagonist Lefty by miR-430/427/302 family is well conserved in Xenopus and humans. However, the targeting of Nodal agonists is variable between species (Rosa et al. 2009). The specific loss of squint regulation by miR-430 in zebrafish increased the mesoendoderm cell population (Choi et al. 2007), suggesting that the gain or loss of Nodal regulation by miR-430/427/302 family influenced the evolutionary changes in mesoderm formation.
miRNA-mediated regulation might not necessarily be conserved within a single gene because miRNA targeting might drift within a set of genes that belong to the same molecular pathway or a single functional unit. A previous study showed that experimentally identified miR-1 targets in zebrafish were enriched for genes that are functionally related to the regulation of actin dynamics (Mishima et al. 2009). Interestingly, predicted miR-1 targets in humans are also enriched for actin-related genes. However, many of the human miR-1 targets do not overlap with the zebrafish miR-1 targets. The rewiring of miRNA-target interactions within a set of related genes might help maintain the overall phenotypic consequences despite evolutionary changes in the target sites, and might allow subtle phenotypic change without disrupting the overall fitness of the system.
The switch function of miR-430 at MZT is conserved in vertebrates. miR-427, which is a miR-430 orthologue in Xenopus, was expressed at MZT and destabilized maternal cyclin B2 mRNA and probably other maternal mRNAs, such as cyclin A1 mRNA (Lund et al. 2009). Although whether this system is also used in mammals has not been addressed, the murine miR-290 cluster, which shares the same seed sequence with miR-430, was transcribed coincidently with de novo transcription during the two-cell stage (Svoboda & Flemr 2010). The switch function of miRNAs at MZT was also observed in fruit fly. Interestingly, however, in fly the switch function is mediated by a set of miRNAs in the miR-309 cluster, which have sequences that are distinct from those of miR-430 (Bushati et al. 2008). As observed in zebrafish, the loss of the miR-309 cluster in fruit fly embryos caused the stabilization of maternal transcripts after fertilization. Target sites for the miR-309 cluster were enriched in maternal transcripts but were depleted from zygotic transcripts. Therefore, the fly miR-309 cluster may represent a functional counterpart of the vertebrate miR-430 family, and the switch function of animal miRNAs at MZT might be a typical example of convergent evolution (Chen & Rajewsky 2007).
Conclusions and future perspectives
As described in this review, miRNAs play essential roles during the development of zebrafish and other animal species. miRNAs exert their widespread regulatory functions by regulating different target transcripts with distinct consequential effects that range from spatial and temporal regulation of gene expression to quantitative modulation of gene doses. Based on the multi-faceted functions of the miR-430-mediated regulation of distinct targets, it is compelling to envision that miRNAs are deeply embedded in intricate genetic programs. One important issue is the quantitative and qualitative characterization of the subtle effects of miRNA-mediated repression, which might be inconsequential under stable conditions in the laboratory. In fly sensory organ development, miR-7 acts against environmental fluctuation by forming interlocking feedback and feedforward loops (Li et al. 2009). Two recently published papers combined experimental and mathematical approaches to reveal the functions of miRNAs in generating thresholds of gene activity and cell fate decisions (Mukherji et al. 2011; Yoon et al. 2011). These studies indicate milestones for the next breakthrough in the miRNA field.
Due to their short sequences, miRNAs and their target sites are more frequently lost and acquired than protein-coding genes during evolution. Therefore, miRNAs may have played predominant but undiscovered roles during animal evolution. Further experimental analyses in multiple model animals and in animals with characteristic evolutionary traits will reveal the capacity of miRNAs in development and evolution.
The author is grateful for comments on the manuscript by Drs Kunio Inoue and Kazuhiro Fukumura. This work was supported by the Grants-in-Aid for Scientific Research from the Japan Ministry of Education, Culture, Sports, Science and Technology and funding from the Kishimoto Memorial Foundation.