|Home | About | Journals | Submit | Contact Us | Français|
MicroRNAs (miRNAs) are critical post-transcriptional regulators that may collectively control a majority of animal genes. With thousands of miRNAs identified, a pressing challenge is now to understand their specific biological activities. Many predicted miRNA:target interactions only subtly alter gene activity. It has consequently not been trivial to deduce how miRNAs are relevant to phenotype, and by extension, relevant to disease. We note that the major signal transduction cascades that control animal development are highly dose-sensitive and frequently altered in human disorders. On this basis, we hypothesize that developmental cell signaling pathways represent prime candidates for mediating some of the major phenotypic consequences of miRNA deregulation, especially under gain-of-function conditions. This perspective reviews the evidence for miRNA targeting of the major signaling pathways, and discusses its implications for how aberrant miRNA activity might underlie human disease and cancer.
The past decade has seen the known catalogs of small regulatory RNAs grow tremendously. A family of 21-24 nucleotide RNAs derived from hairpin precursor transcripts, termed microRNAs (miRNAs), provided a pioneer glimpse into this small RNA universe. Although miRNAs were only broadly recognized in 2001,1-3 prior developmental genetic studies in Caenorhabditis elegans4-7 and Drosophila melanogaster8-11 had already unearthed many of the key mechanistic features of miRNA production and function. With the cloning of miRNAs came the realization of Watson-Crick base pairing between the 5' ends of miRNAs and several previously defined ~7 nucleotide (nt) 3' UTR repression elements,12 defining a key feature of miRNA target recognition. Canonical pairing between nucleotides 2-8 of the miRNA (i.e., the “seed”) and the 3'UTR is often sufficient to mediate target regulation.12-16 Using evolutionary conservation as a robust filter for functionally constrained target sites, a substantial fraction of animal genes are predicted as direct miRNA targets, with hundreds of targets often assigned to individual miRNAs.17-20 Additional non-canonical, seed-mismatched sites that are relevant to target regulation may remain to be elucidated.5,13 Moreover, regulation mediated by non-conserved sites might underlie species-specific or individual-specific characters.21-23
Despite great advances concerning the identity, biogenesis and biochemical properties of miRNAs, we currently understand comparatively little about their biological functions and key targets. It seems doubtful that very many of the 10,000s of highly conserved miRNA: target sites will prove to be of substantial impact to phenotype. At the same time, the seemingly all-encompassing breadth of the miRNA-mediated regulatory network has fueled great speculation about the potential relevance of miRNAs to human disease. Functional data are indeed quickly accumulating that implicate miRNAs in human diseases affecting lymphocyte development and immune function, muscle development and cardiac function, neurogenesis and neurophysiology, and a host of cancers, just to name a few.24
Because the large numbers of predicted miRNA targets do not conveniently lend themselves to hypothesis-based biological inquiries, rational approaches are needed to effectively prioritize functional investigations. The premise of this perspective is that activity-based studies of how miRNAs affect the major animal signaling pathways is likely to be an expedient strategy for uncovering phenotypically relevant connections between miRNAs and disease. We will review some of the major evidence for this notion, and how principles derived from these studies may inform future research.
A century of developmental genetic studies has led to the appreciation that cell-cell signaling pathways are typically highly dose-sensitive processes. For example, the Drosophila Notch pathway alone contains three haploinsufficient members (Notch, Delta and Hairless); in other words, genes for which loss of a single allele confers a completely penetrant morphologically mutant phenotype. Animal development relies upon a core set of cell signaling pathways,25,26 namely the Notch (N), Hedgehog (Hh), Wnt/Wingless (Wg), TGFβ, receptor tyrosine kinase (RTK), nuclear receptor, Jak/STAT, and Hippo pathways. All of these pathways are rich with components that exhibit genetic interactions in heterozygous state with other respective pathway components. This property has been the basis of highly successful screens for novel signaling genes by virtue of dominant suppression or enhancement of a sensitized pathway phenotype. The sensitivity of cell signaling often involves amplifying mechanisms that convert initially modest activity differences into an all-or-nothing output.
The precise level of transcriptional output of these cell signaling pathways is essential for normal cell specification and tissue growth in invertebrates and vertebrates. As a consequence, their dysfunction invariably induces developmental abnormality and/or adult disease. As each of the major signaling pathways is constantly reused during development, the same core pathway components are normally configured to generate very different transcriptional outputs in specific temporal or spatial settings. In disease situations, aberrant signaling may lead to phenotypic outputs that might normally be relevant in other tissue or temporal contexts. In sum, the fundamental control of cell division, growth and apoptosis by core cell signaling pathways makes many of their components potent oncogenes and tumor suppressors.
Because precise output levels of these signaling pathways are so crucial for appropriate cell behavior, a dizzying array of positive and negative regulatory mechanisms have evolved to regulate cell signaling. These can occur at the transcriptional, post-transcriptional, post-translational levels, and involve control of RNA and protein stability, activity or localization. It follows that there are many mechanisms by which miRNAs might influence cell signaling. For example, since many signaling pathways are under varying degrees of default repression, miRNA targeting of transduction-limiting repressors could increase pathway activity. Conversely, miRNAs might repress signal-regulated events by targeting positively-acting pathway components. Since miRNAs usually have many targets, it might also be necessary to consider that miRNAs could have complex influences on multiple pathway components, and perhaps even on components of different signaling pathways. Evidence for all of these scenarios has indeed emerged in various model organism studies.
Prior to the formal discovery of miRNAs, there were already substantial data for important roles for miRNAs in influencing cell-cell signaling. Fundamental evidence came in the form of post-transcriptional regulation mediated by ~7 nt 3'UTR elements that were pervasive amongst two distinct classes of Drosophila Notch target genes,8-11 encoding Bearded proteins and Enhancer of split [E(spl)] basic-helix-helix (bHLH) repressors.27-31 These elements, termed the Brd box, GY box and K box, were originally noticed on the basis of hypermorphic alleles of Bearded and E(spl)m8,8,32 both of which lacked 3' UTR sequences containing these control elements.
The study of these “box” elements showed that they mediated both translational repression and transcript instability, and the latter was associated with target deadenylation.9,10 These box motifs proved to define the major strategy for animal miRNA target recognition, via Watson-Crick complementarity to the 5' ends of miRNAs.12 Mutational analysis showed that disruption of these motifs induced Bearded and E(spl)m8 hyperactivity during Notch-mediated sensory bristle and eye development,9,10 and misexpression studies of the cognate miRNAs similarly revealed specific disruption of Notch-mediated developmental programs.15
In worms, miR-61 is a direct transcriptional target of lin-12/ Notch, and proposed to function as a Notch effector that inhibits EGFR/MAPK signaling via the vav-1 target gene.33 More recently, the nematode let-7 miRNA was found to be upregulated downstream of activated lin-12/Notch.34 The observation of genetic interactions between lin-12/Notch and let-7 were consistent with the utilization of this regulatory link during the worm developmental timing program. Therefore, the transcriptional hierarchy downstream of activated Notch in worms appears to include multiple miRNAs, and this may conceivably mediate crosstalk between different developmental pathways.
The first Drosophila miRNA subjected to functional study was miR-14. It was isolated as a putative cell death-related gene; in particular, both loss- and gain-of-function studies revealed it to be a potent suppressor of pro-apoptotic insults.35 While the basis of this activity is incompletely understood, it was recently shown that endogenous miR-14 regulates nuclear receptor signaling by the steroid hormone Ecdysone, where it limits the expression of the Ecdysone Receptor.36 In this setting, miR-14 participates in a mutual antagonistic feedback loop, wherein miR-14 inhibits Ecdysone Receptor positive autoregulation, and Ecdysone Receptor reciprocally inhibits miR-14 expression. Interestingly, ecdysone has complex effects on miRNA expression, activating some (such as let-7, miR-100 and miR-125) while repressing others (such as miR-34) in an Ecdysone Receptordependent fashion.37-39
In worms, members of the let-7 family (including let-7 and miR-84) were shown to act together to suppress the nhr-23 and nhr-25 nuclear receptors. Genetic reduction of these miRNAs results in an additional molt and a failure of normal terminal differentiation.40 Such phenotypes can be interpreted as reflecting a normal role of miRNAs in limiting cell proliferation.
Two additional Drosophila miRNAs, bantam and miR-278,41-43 have both anti-apoptotic and growth promoting activity, dual functions that suggest oncogene-like function. Bantam provided a particularly interesting precedent as a miRNA whose endogenous function is critical to growth regulation, since bantam loss-of-function mutants exhibit undergrown tissues.44 The ability of bantam to suppress apoptosis is explained in large part by its strong repression of the pro-apoptotic gene head involution defective,41 although target(s) that explain its growth-promoting function are still wanting. Interestingly, bantam is a transcriptional target of Hippo signaling;45,46 thus, deployment of a non-coding RNA is a critical arm of this cell signaling pathway.
miR-278 similarly inhibits apoptosis and promotes growth, although in this case its loss-of-function phenotype is manifest in energy homeostasis.42,43 Direct repression of expanded, a growth regulator in the Hippo pathway, can explain the phenotypic effects of miR-278 deregulation with respect to tissue proliferation. This provides a notable precedent in which the inappropriate silencing of a miRNA target can lead to a profound cancer-relevant defect in a setting unrelated to the normal function of the miRNA.
Since fly studies showed that miRNAs are components of the Hippo pathway, a universal mechanism for growth control,47 presumably human miRNAs that can regulate Hippo components might be relevant to oncogenesis. Indeed, human miR-372/miR-373 were implicated as testicular germ cell oncogenes that promote cellular transformation, at least partly through the repression of the LATS2 tumor suppressor.48 LATS2 is the human homolog of Warts, a kinase component of the Hippo pathway that was first identified in Drosophila.
An unusual screening strategy was recently employed in Xenopus to identify miRNAs that might affect TGFβ signaling mediated by the Nodal ligand subfamily. By purifying a collection of early embryonic low molecular weight RNAs and re-injecting them into Xenopus oocytes, the authors identified regulation of a Nodal reporter construct that was mediated, at least in part, by the miR-15/16 cluster.49 In this developmental context, the ventral enrichment of miR-15/16 appears to be important for the proper dorsal induction of the Spemann organizer. Curiously, the graded expression of these miRNAs may be a consequence of repression by dorsal Wnt signaling. In this fashion, it was proposed that miR-15/16 functions at the crossroads of these signaling pathways.49
Concurrent studies in zebrafish identified a distinct miRNA, miR-430, that appears to be important for balancing the early developmental activity of Nodal signaling.50 The activation of this pathway is already balanced by positively-acting ligands (i.e., Squint) and negatively-acting ligands (i.e., Lefty proteins). In the early fish embryo, miR-430 simultaneously represses both squint and lefty to balance the phenotypic output of this pathway. This study innovated the use of “target protectors” that specifically block miRNA binding to particular target sites.50 Curiously, the loss of miR-430 is in some respects intermediate to loss of regulation of either target, indicating that the function of miR-430 is in the balance of both ligands.
Adding to the list of miRNA-regulated signaling pathways, Hedgehog (Hh) signaling has been identified as a target of zebrafish miR-214. Chemical inhibition of miR-214 results in the loss of slow-twitch muscles via deregulation of the negative Hh component suppressor of fused.51 Expression profiling was also done in the zebrafish system to identify miRNAs that are transcriptional targets of Hh.52 This study revealed not only miRNAs whose expression is modulated by Hh signaling, but also many miRNAs that are apparently regulated by Notch signaling. It was proposed that a subset of miRNAs might act as points of crosstalk between these fundamental signaling pathways.
In Drosophila, the miR-12/miR-283/miR-304 genomic cluster was recovered as a locus whose misexpression in imaginal discs yielded a mirror image wing duplication similar to that produced by ectopic Hh.53 This study showed that multiple transcripts that encode members of a multicomponent cytoplasmic complex in the Hh pathway were directly repressed by miR-12 and/or miR-283, including costal-2, smoothened and fused. These were demonstrated as endogenous targets since imaginal disc cells lacking the miRNA cluster exhibited modestly increased levels of transgenic 3'UTR sensors for these target genes. Nevertheless, this was insufficient to lead to a phenotypic defect in Hh pathway output in miRNA mutant cells.53 Possibly the miRNAs are required only for fine-tuning signaling. On the other hand, since Costal-2 and Smoothened are antagonistically-acting components of the Hh pathway, it is conceivable that the miRNA deletion situation represents an intermediate, phenotypically suppressed condition. This might be analogous to the usage of zebrafish miR-430 to repress both agonist and antagonist of the Nodal/BMP pathway.50
The members of the miR-12/miR-283/miR-304 cluster collectively have hundreds of conserved targets. Nevertheless in the ectopic situation, this miRNA cluster drives a characteristic activated Hh phenotype. Since aberrant Hh is causal to several types of tumors, this suggests that deregulated miRNAs with many targets can still have specific consequences on cancer-relevant, developmental cell signaling pathways.
The small GTPase Ras is a classical oncogene that is the central culprit in a wide variety of human cancers. Its normal function is as a component of a receptor tyrosine kinase cell signaling cascade. Unrestrained Ras drives aggressive tissue proliferation, and so it follows that Ras repressors might be candidate tumor suppressors. Based on a genetic regulatory relationship between let-7 and the worm Ras homolog let-60, let-7 was shown to regulate human Ras as well.54 In particular, regulation of Ras genes by human let-7 genes has continued to be a focus of studies of lung cancer55 and breast cancer.56
Studies of Drosophila miR-7 revealed another type of interaction between miRNAs and RTK signaling. During eye development, miR-7 promotes photoreceptor development by repressing the RTK transcriptional repressor Yan.57 Similar to the example of miR-14/ Ecdysone Receptor, miR-7/Yan participate in a mutually negative feedback loop. miR-7 directly represses Yan at a post-transcriptional level, while Yan directly represses miR-7 at the transcriptional level.57 Since miR-7 controls both EGFR and Notch activity,15,57 miR-7 may represent yet another point of crosstalk between signaling pathways.
The aforementioned examples provide ample precedent that the existence of miRNA-mediated control of the fundamental signaling pathways can lead to profound phenotypic consequences to miRNA dysfunction. We recently expanded on this premise to initiate systematic functional screens for the ability of individual miRNAs to influence the fundamental signaling pathways. Our strategy takes advantage of previously described transcriptional reporters for these pathways, which enable high-throughput, quantitative, cell-based screening. We generated a library of miRNA expression plasmids and used these to assess the ability of individual miRNAs to alter the nuclear output of various signaling pathways. Since miRNA control of Wingless (Wg)/Wnt signaling had not yet been described, we focused on this pathway for our proof-of-principle studies.58
We screened cells that harbored a luciferase reporter controlled by a multimer of binding sites for TCF, the main sequence-specific transcription factor in the Wg pathway. Since the cells used did not express Wg ligand natively, we screened for alteration of TCF-luc by individual miRNAs in the presence or absence of exogenous Wg. miR-315 emerged from this screening as a potent (~10-fold) activator of TCF-luc; curiously, this activity was evident only in cells that were not stimulated with Wg. We verified that this activity was independent of the Wg receptor or co-receptor; therefore, miR-315 is capable of autonomously activating Wg signaling in cells that are not normally engaged in Wg signaling.58
This regulatory relationship was corroborated in the animal by the demonstration that ectopic miR-315 was sufficient to respecify notum as wing, yielding 4-winged flies. A similar phenotype was previously attributed to misexpression of Wg itself.59 The ability of miR-315 to activate Wg signaling suggested that it repressed one or more negatively-acting components of this pathway. Molecular and genetic tests showed that direct repression of Axin and Notum by miR-315 was sufficient to explain its ability to activate Wg signaling.58
We draw attention to the fact that miR-315 has hundreds of conserved targets; thus in retrospect, a particular connection between miR-315 and Wg signaling could not have been anticipated from the outset. Nevertheless, regulation of Axin, Notum and Wg signaling dominate the functional consequence of expressing miR-315 in the wing disc. In particular, ectopic Notum or Axin are able to rescue significant aspects of mutant phenotypes induced by ectopic miR-315. Conversely, miR-315 gain-of-function renders animals extremely sensitive to heterozygosity for Axin or Notum, indicating that this genetic condition reduces these target genes to a phenotypically critical level.
The first few miRNA loci identified in worms were noticed on account of their essential genetic requirements in developmental cascades.4,5,60,61 What a surprise, then, when a systematic reverse genetic screen revealed that vanishingly few of the remaining ~100 miRNAs of C. elegans are required for viability, fertility or gross morphology or behavior.62 Given that miRNAs have been placed on a par with transcription factors as executive-level regulators of gene expression, the initial results from this arduous project were underwhelming. Certainly this unique mutant collection will be the basis for many future biological discoveries, and one may posit that genetic redundancy or highly-cell specific requirements masks the requirement of many of these miRNAs. On the other hand, any random selection of 100 protein-encoding gene mutants—and certainly 100 transcription factor mutants—would undoubtedly have revealed more obvious defects.
Phenotypes of a growing number of Drosophila and mammalian miRNA deletion mutants have recently been reported.24,63 However, the worm studies indicate that it is not a foregone conclusion that miRNA loss-of-function, even for highly conserved and highly expressed miRNAs with many highly conserved targets, will necessarily be of substantial consequence to an individual. Based on the “mutual exclusion” model for miRNA and target expression,18,64 miRNA-mediated regulation may often be secondary to that of transcriptional regulation, in which case miRNAs might be expected to more generally serve as backup regulators.
Conversely, the collected observations strongly suggest that excess or ectopic miRNA function in both invertebrates and vertebrates quite frequently generates dramatic and deleterious phenotypes. We may attribute this to the fact that miRNA gain of function operates in a dominant fashion, and is not limited to tissues with endogenous miRNA expression. In addition, ectopic miRNAs can enforce strong suppression of target genes, whereas the consequences of miRNA loss might be compensated for by other regulatory mechanisms. With respect to model organism genetics, many scientists view gain-of-function phenotypes with measured skepticism, since they may not be reflective of a gene's endogenous function. However, with regard to how miRNAs impact human disease or cancer, one may argue that it does not matter if the endogenous regulatory requirement is subtle. If a miRNA is pathogenic in disease situations, then that is information that we want to know.48,65,66
As mentioned, because signaling pathways are amongst the most highly dose-sensitive biological processes known, we argue that the dominant repression of key signal components can be of especial phenotypic consequence. For example, while GY box and K box Drosophila miRNAs have very large predicted target sets,67 many of the morphological consequences of their misexpression can be specifically attributed to repression of Notch signaling.15 Similarly, miR-315 has one of the largest catalogs of conserved miRNA targets of all Drosophila genes. Nevertheless, its misexpression in an animal, at least in the context of wing-notum development, yields a simply interpreted phenotype that can be accounted for largely by 2 targets in the Wg pathway.58 The same is true for the Drosophila miR-12/miR-283/miR-304 cluster, which collectively have hundreds of conserved targets, but whose misexpression induces specific Hh pathway activation.53 These observations provide ample precedent for the notion that miRNA targets in developmental cell signaling pathways may be of particular relevance to disease settings involving miRNA deregulation.
Computationally-driven target analysis has generated substantial insight into the nature of the miRNA regulatory network. On the other hand, biologically-driven experiments have shown that target predictions are often a poor guide of miRNA targets whose modulation is of phenotypic consequence. In the case of Drosophila Wg signaling, a significant number of miRNAs are predicted to target the same negative regulators that mediate miR-315's activation of Wg signaling, yet these miRNAs do not activate pathway activity and fail to strongly repress the endogenous 3'UTRs they are predicted to target.58 Reciprocally, the Notum gene is a mediocre predicted target, even in the latest prediction schemes;68 yet its repression is essential for understanding the phenotypic consequence of ectopic miR-315.58 These data highlight that target predictions have room to improve as we gain additional understanding of how 3' UTR secondary structure and associated RNA binding proteins impact miRNA-mediated regulation.
Overall, these findings support the rationale of approaching miRNA activities from an unbiased position wherein the potential for signal pathway regulation is assessed first by activity-based screening, followed by consultation with target prediction algorithms and the appropriate genetic validation tests. Tools for systematic activity based screening of mammalian miRNAs have been generated, and used to great effect to examine how miRNAs influence proliferation and metastasis pathways.48,69,70 With an established battery of cell signaling assays readily available for high-throughput screening, it is possible to extend the approach taken in model organisms to better understand the activity of human miRNAs in signal transduction settings.
miRNAs are integrated into vast regulatory networks that impinge upon a broad spectrum of biological events. While we are decades from a complete understanding of the endogenous functions of miRNAs, we now have tools with which to rapidly assess the potential contributions of miRNAs to discrete biological events. Amongst these are a growing number of examples where miRNAs influence core, conserved cell-cell signaling pathways that are central to both development and disease. Taking advantage of the dose-sensitive nature of signaling pathways, miRNAs represent an elegant layer above transcriptional control for both fine-tuning and dramatically altering the activity and output of cell signaling. In addition, miRNAs may serve as points of crosstalk between signaling pathways, by integrating transcriptional inputs or by their functional regulatory output on different pathways. We believe that directed investigation, using activity based screening of cell signaling pathways, will yield many additional examples of functionally relevant regulatory relationships. Their elucidation will define endogenous functions and, perhaps more critically, provide insight into the roles of miRNAs in human disease.
J.W.H. was supported by the Mildred and Emil Holland Scholarship, the Margaret and Herman Sokol Fellowship, NIH MSTP grant GM07739, and Ruth L. Kirschstein pre-doctoral fellowship 1F30NS062501-01 (NINDS). E.C.L. was supported by grants from the Leukemia and Lymphoma Society, the Burroughs Wellcome Foundation, the V Foundation for Cancer Research, the Sidney Kimmel Cancer Foundation, and the National Institutes of Health (R01-GM083300 and U01-HG004261).