|Home | About | Journals | Submit | Contact Us | Français|
It is apparent that microRNAs (miRNAs) are important components in the regulation of genetic networks in many biological contexts. Based on computational analysis, typical miRNAs are inferred to have tens to hundreds of conserved targets. Many miRNA-target interactions have been validated by various means, including heterologous tests in cultured cells and gain-of-function approaches that can yield striking phenotypes in whole animals. However, these strategies do not report on the endogenous importance of such miRNA activities. Likewise, studies of miRNA pathway mutants can suggest an endogenous role for miRNAs in a given setting, but do not identify roles for specific miRNAs. Therefore, these approaches must be complemented with the analysis of miRNA mutant alleles. In this review, we describe some of the lessons learned from studying miRNA gene deletions in worms, flies and mice, and discuss their implications for the control of endogenous regulatory networks.
microRNAs (miRNAs) are a class of ~21–24 nucleotide RNAs that associate with Argonaute (Ago) proteins and guide them to target transcripts for regulation.1,2 By and large, miRNA-mediated regulation results in target repression, by reducing mRNA stability and/or by inhibiting translation. Early genetic studies showed that biologically-important targets in worms3–5 and flies6–8 had imperfect complementarity to miRNAs,9,10 and that 7 nt of Watson-Crick pairing to the 5′ ends of miRNAs (usually positions 2–8, also referred to as the miRNA “seed”) frequently suffices for target repression.11–13 Computational surveys revealed that a substantial fraction of animal transcripts contain one or more predicted target sites in their 3′ UTRs that are well-conserved, and thus presumed to be functionally constrained.14–17 With the possibility of atypical targeting via seed-mismatched sites with 3′ compensatory pairing,13,18 and the existence of poorly-conserved functional sites,19 the breadth of the miRNA target network increases. There are even detectable “anti-targets”, or genes that actively avoid the acquisition of miRNA binding sites.20,21 These collected observations suggest that the activity of most animal genes is, at least to some degree, directly influenced by one or more miRNAs.
A few years ago, the terms ‘switch target’, ‘tuning target’ and ‘neutral target’ were coined to describe possible outcomes of miRNA-mediated repression.21 A switch target was defined as one whose protein output is reduced to functionally inconsequential levels by the miRNA. A tuning target was defined as one for which miRNA-mediated regulation acts to to ensure functional, but not undesirably high levels of the gene product. A neutral target was defined as one whose regulation by a miRNA is entirely incidental, such that any change in target levels brought about by loss of the miRNA has no biological consequence.
In some cases of miRNA control of developmental regulators, the elimination of the target is essential for a given biological process. However, it is instructive to note that the “switch” category, as defined above, does not necessarily imply that a miRNA-target relationship is of substantial importance.2 Since the level of transcription of many targets is already low in cognate miRNA-expressing cells,20,22 target outputs might essentially be inconsequential even without the involvement of a miRNA. Such “de-noising” roles in clearing away low levels of unwanted transcripts might be quite broad for miRNAs. For such targets we might expect their loss of miRNA-mediated regulation to be of minimal consequence.
Conversely, one should not assume that mild amounts of target regulation are correspondingly of little consequence. In some cases, a “thresholding” function of miRNAs has been posited, which sets a minimum level of target activity that must be exceeded in order to get the ball rolling on some larger chain of events.2,23 For such cases of finely balanced target activity, the loss of miRNA-mediated regulation might be quantitatively subtle, but nonetheless of great phenotypic significance if a pathway is inappropriately tripped.
These considerations make clear that it is not a trivial matter to infer substantial biological functions of miRNAs strictly from perusal of target predictions.24 Phenotype-driven gain-of-function screening is one expedient approach to identify specific activities of miRNAs. In cultured cells or in whole animals, many miRNAs produce spectacular phenotypes due to the ectopic repression of one or more targets.25–29 Such studies can lend insight into potent miRNA-target relationships, and strongly support the notion that miRNA deregulation can directly induce disease and cancer. Nevertheless, the activities observed in such experiments are not necessarily germane to the normal function of the miRNA. For instance, if a miRNA is misexpressed entirely outside of its normal domain of expression, it might encounter and repress completely inappropriate targets.
As is the case for protein-encoding genes, then, loss-of-function analysis is needed to divine the endogenous function of miRNA genes. Antisense inhibitors of miRNAs are currently in broad usage for their technical convenience. However, the extent to which they might induce off-target effects is currently uncertain, and the phenotypes attributed to the inhibition of several miRNAs30 were not observed following the analysis of corresponding deletion mutants.31–33 Therefore, the unambiguous determination of miRNA requirements requires genuine loss-of-function alleles. A substantial number of miRNA deletions have now been isolated in worms, flies and mice (summarized in Table 1), and we review some of the lessons learned from their analysis.
The first miRNA genes were serendipitously identified from forward genetics in Caenorhabditis elegans. The nematode has been a favorable system to analyze the control of developmental timing by so-called “heterochronic” genes. One essential gene is lin-14, a master regulator for cell lineages executed during the first larval stage (L1). Lin-14 is normally expressed during L1 but must be inactivated during the transition to L2.34 Accordingly, lin-14 loss-of-function mutants skip L1 and precociously initiate later developmental programs, while lin-14 gain-of-function mutants repeat L1 and fail to progress to later larval stages.35 The gain-of-function mutations in lin-14 all have deletions or re-arrangements in the 3′ UTR,3 indicating that this region mediates post-transcriptional repression.
Loss-of-function mutations in another heterochronic gene, lin-4, yield phenotypes virtually identical to lin-14 gain-of-function.36 Identification of the lin-4 transcription unit revealed two small RNA products, which we now recognize as the pre-miRNA precursor hairpin and the mature 21 nt miRNA.10 lin-4 exhibits antisense complementarity to seven motifs in the lin-14 3′ UTR, many of which are missing in lin-14 gain-of-function alleles.4,10 Subsequent analysis demonstrated direct recognition of these motifs by lin-4.37 The power of genetic analysis is surely well-illustrated in this tale, in which the main biological function of a miRNA (lin-4), and one of its most important direct targets (lin-14) were elucidated without even knowing what a miRNA was, and lacking experimental knowledge of features of miRNA:target pairing. This founding paradigm served as a precedent in the study of a second direct heterochronic target of lin-4, namely lin-28.5
The second known miRNA was also initially characterized from a heterochronic mutant, this time one that affected the L4 to adult transition. Hypodermal blast cells normally exit the cell cycle at the end of L4 and fuse with neighbouring hypodermal seal cells to generate cuticular alae. In let-7 mutants, these cells maintain larval patterns of cell division and the animal enters an ectopic larval stage. Conversely, ectopic let-7 results in premature cell cycle exit of hypodermal blast cells at the L3–L4 molt.18 The let-7 phenotype can be mostly explained by misregulation of Lin-41. Increased Lin-41 phenocopies the most striking features of let-7 loss-of-function, and reduction in lin-41 strongly suppresses let-7.18,38 However, the regulation of other heterochronic targets is likely important for let-7 function. For example, let-7-mediated regulation of hbl-1 and daf-12 is also involved in the L4 to adult transition.39–41 let-7 was also reported to repress ras/let-60 during vulval formation.42
miR-48, miR-84 and miR-241 all share the same seed as let-7, and are thus predicted to have overlapping target potential. However, their expression is under distinct temporal control, reaching maximal expression at the earlier L3 stage.43 Individual mutations of these let-7 “sister” genes yielded few obvious defects. However, double and triple mutant combinations of mir-48, mir-84 and mir-241 exhibited strong heterochronic phenotypes including an apparent repetition of the L2 program of proliferation of the V lineage seam cells. Reiteration of the L2 program was suppressed by reduction of hbl-1, a hunchback like gene with multiple let-7-family binding sites in its 3′ UTR. miR-48, miR-84 and miR-241 are also necessary for repression of a GFP reporter with the hbl-1 3′ UTR.43 miR-84 was also reported to synergize with let-7 to promote seam cell differentiation via repression of nuclear hormone receptor nhr-25.44
These findings demonstrate that members of miRNA families can have overlapping activities, so that several genes may need to be deleted to uncover a substantial phenotype. However, one must also bear in mind that mutation of let-7 alone reveals an essential function; therefore, depending on their particular expression in time and space, similar miRNAs might still have very different genetic usage.
Adult worms have bilaterally paired taste receptor neurons that exhibit left-right asymmetry in their expression of guanylyl cyclase chemoreceptors. The existence of live fluorescent markers for left and right ASE neurons (ASEL and ASER) made this an excellent system for genetic screening to dissect left-right neural specification. One such mutant, lsy-6, exhibited no ASEL and 2 ASER neurons, indicating the conversion of ASEL into ASER.45 This locus was found to encode a miRNA that is specifically expressed in ASEL, namely the cell whose fate it controls. In fact, ectopic expression of lsy-6 in ASER was sufficient to redirect it into ASEL fate.
ASE cell fate is also controlled by the antagonistic expression and activity of the Die-1 and Cog-1 transcription factors. Die-1 is normally expressed in ASEL, whereas Cog-1 is normally expressed in ASER; die-1 mutants exhibit two ASER neurons, whereas cog-1 mutants differentiate two ASEL neurons.45,46 The similarity between cog-1 loss-of-function and lsy-6 gain-of-function (and vice versa) suggested a direct link. Indeed, target sequences in the cog-1 3′ UTR mediate direct repression by lsy-6, and their mutation permits cog-1 to be ectopically active in ASEL.
Curiously, subsequent studies revealed that miRNAs of the miR-273 family are reciprocally active in ASER and function there to repress die-1 directly.46 Therefore, bilaterally asymmetric ASEL and ASER miRNAs are integral components of a double negative feedback loop that ensures proper specification of these neural fates.
The above examples illustrate that C. elegans miRNAs can reveal strong phenotypes when mutated. This suggested that wide-ranging roles for miRNAs might be revealed via systematic deletion of all nematode miRNAs. A collaborative effort generated and analyzed deletions covering 89 miRNA genes.47 Homozygous mutant animals were subjected to a range of assays to identify phenotypes affecting morphology, growth, development, and a variety of behaviors such as locomotion, pharyngeal pumping, defecation, egg-laying, dauer formation and presence of chemosensory neurons.47 Strikingly, out of this entire collection, only four deletions produced detectable phenotypes, and two of these were attributable to loss of the protein-encoding gene that hosted an intronic miRNA. Bona fide miRNA mutant phenotypes were observed only for the mir-240/mir-786 deletion, which resulted in a prolonged defecation cycle, and for a deletion removing the mir-35-41 cluster, which exhibited temperature sensitive lethality. This study concluded that the vast majority of C. elegans miRNAs are individually dispensable for major aspects of development, viability or simple behaviors.47
miR-1 is highly conserved from C. elegans to humans, both in primary sequence and in muscle-specific expression. Although initial analysis of the worm miR-1 mutant showed it to be viable, fertile and to exhibit normal muscle cell number and morphology,47 more detailed studies of this mutant recently revealed striking defects in muscle function.48
Levamisole activates a subset of nicotinic acetylcholine receptors (nAChRs), causing muscle contraction and paralysis in wild-type worms but not in mir-1 mutants. Subsequently, two components of the levamisole-sensitive nAChRs were identified as direct miR-1 targets. Their increased levels in the mir-1 mutant correlated with its decreased sensitivity to levamisole. Interestingly, synaptic transmission at neuromuscular junctions (NMJ) was also impaired in mir-1 mutants. This seems to be explained by the overabundance of another miR-1 target, the transcription factor encoded by mef2, which causes synaptic accumulation of a Rab-3 transgene and a pre-synaptic defect.48 Thus, nematode miR-1 regulates both pre-synaptic and post-synaptic aspects of neuromuscular junctions.
This study highlights how seemingly healthy and fertile mutants might still exhibit compelling phenotypes when subjected to appropriate scrutiny. Indeed, this study is a general reminder of the fact that mutants in protein-encoding genes frequently have no overt phenotype in typical laboratory settings. Yet their active preservation in genomes is undoubtedly a reflection of their endogenous biological usage in the wild. We expect that other nematode miRNA mutants will yield compelling phenotypes once examined in just the right way.
The first Drosophila miRNA mutant reported was mir-14. This locus was originally studied as a cell death regulator, since null mutants of mir-14 enhanced a pro-apoptotic phenotype, whilst extra copies of mir-14 suppressed apoptosis. Around 80% of flies lacking miR-14 died during pupal development, and the adult survivors had decreased lifespan and lower tolerance to environmental stresses. These mutants also exhibited increased levels of triacylglycerides and diacylglycerides.49
Subsequent studies revealed the ecdysone receptor as one key direct target of miR-14.50 The steroid hormone ecdysone signals via the Ecdysone receptor (EcR) to initiate metamorphosis. Activation of EcR increases EcR transcription, and it is proposed that this autoregulatory loop is normally repressed by miR-14. Accordingly, the viability of mir-14 mutants was partially rescued by removing one allele of EcR. Interestingly, the level of miR-14 is decreased by ecdysone signaling, suggesting a mutually inhibitory effect between miR-14 and EcR. These data led to the proposal that miR-14 dampens the autoregulatory loop of EcR activity, perhaps to avoid accidental triggering of metamorphosis. Only with strong induction of ecdysone signalling is enough EcR maintained to overcome miR-14. This mode of action illustrates how a miRNA can lend robustness to an established developmental program.
The bantam miRNA was found in a gain-of-function screen for loci that could induce growth phenotypes.29,51 Activation of bantam causes profound tissue overgrowth, whereas deletions of bantam exhibit severe growth deficiencies and early larval death.51 Detailed study of bantam activity showed that it actually has genetically separable growth-promoting and death-inhibiting activities. The latter is due to its ability to directly suppress the pro-apoptotic gene hid;29 however, its growth target(s) remains to be elucidated. The bantam locus was subsequently found to be a crucial transcriptional target of the growth-controlling Hippo pathway, which coordinately regulates proliferation and apoptosis.52,53
miR-278 was isolated as another locus whose overactivation yielded tissue overgrowth; it also has certain anti-apoptotic activities.54,55 Unlike bantam, the null mutant of this locus does not have an obvious growth defect, and there is no miR-278-related locus that might supply redundant function. Close examination revealed that mir-278 mutant flies were lean and exhibit decreased triglyceride levels and increased levels of insulin-like peptides.54 miR-278 targets expanded, a negative regulator of growth. Its downregulation by ectopic miR-278 can explain the growth inducing properties of this miRNA. Overaccumulation of Expanded in mir-278 mutants also correlates with the lean phenotype. These findings yield the lesson that while miRNA misexpression phenotypes can give hints about key targets, such miRNA: target relationships may normally be relevant in settings unrelated to that used for miRNA activity screening.
Drosophila miR-1 is strongly and specifically expressed in the developing mesoderm and somatic musculature during embryogenesis and early larval development.31,56,57 Loss of miR-1 is lethal at early larval stages. While modest defects in muscle specification were described in this mutant,56 it appears that the major defects are manifest during rapid muscle growth that initiates in early larvae.31
Newly hatched wild-type larvae arrest their development when deprived of amino acids and reared on sucrose as a sole energy source.58 Upon the addition of dietary amino acids, these larvae re-initiate development and grow rapidly, strongly increasing both mitosis index and endocycling of fat body, gut and salivary gland cells.58 Interestingly, the patterning and function of the musculature of miR-1 mutant 1st instar larvae held under such developmental stasis remains largely normal. The 2nd instar paralysis phenotype of mir-1 mutant larvae is thus strongly dependent on a nutritional input, indicating that the main requirement for miR-1 is in the post-mitotic growth of larval muscle.31 The early requirement for miR-1 in Drosophila muscle growth has precluded analysis of its function in mature muscles, but in light of the recent observations of C. elegans miR-1,48 it is tempting to speculate about its potential roles in muscle physiology.
Drosophila miR-7 is an intronic miRNA that resides in the gene bancal. In the developing eye, miR-7 is specifically expressed in photoreceptors. Overexpression of miR-7 results in extra photoreceptors, but flies lacking mir-7 do not show any gross morphological phenotypes. However, photoreceptors mutant for mir-7 have elevated levels of Yan, a transcriptional repressor that inhibits photoreceptor specification. miR-7 and Yan form a bistable feedback loop: Yan directly represses mir-7, while miR-7 represses Yan.59 The endogenous requirement for miR-7 was best illustrated in a sensitized eye that expresses an activated form of Yan; the deleterious effects of Yanact were far more potent in eyes mutant for mir-7. Therefore, this loop reinforces the photoreceptor fate decision by helping to ensure that Yan is active in cells that should not be differentiating, and inactive in cells that adopt the neural fate.
Another miRNA involved in neural specification is miR-9a.32 Drosophila sensory bristle development is a classic paradigm of pattern formation. Sensory organ formation requires Senseless (Sens), a zinc finger transcription factor that is upregulated in sensory organ precursors (SOPs) and helps maintain high levels of expression of bHLH proneural genes.60 In the embryo, miR-9a is expressed in ectoderm and neurectoderm but is absent in neural precursors and SOPs,20 consistent with a role in suppressing neural cell fates.
miR-9a directly represses sens via its 3′UTR, and mir-9a deletion mutants are viable but have ectopic sensory organs. Conversely, ectopic miR-9a abolishes the formation of sensory organs.32 The role of miR-9a has been interpreted to prevent inappropriate induction of SOPs by dampening the ‘leaky’ expression of sens in cells not destined to become SOPs. This illustrates an example of a miRNA that exerts its biological function by thresholding.23
Flies deleted for mir-8 exhibit decreased survival and are uncoordinated.61 Amongst other genes, miR-8 directly regulates the transcription regulator Atrophin. The increased level of Atrophin causes excess brain apoptosis that appears to account for much of the behavioral deficits of the mir-8 mutant. Interestingly, the expression of an atrophin knockdown transgene specifically under control of miR-8 regulatory sequences caused reduced viability and produced wing phenotypes consistent with atrophin loss-of-function. This suggests that Atrophin is functionally required in miR-8-expressing cells, implying that the role of miR-8 is not to eliminate atrophin, but instead to “tune” it to an optimal level.
The classical homeotic genes are clustered and encode homeodomain proteins, whose precise deployment along the anterior-posterior axis of all animals is critical to the appropriate specification of body segments. Extensive genetic analyses suggested that the Bithorax cluster (BX-C) contains as many as 9 homeotic genes.62 However, genome sequencing and saturating loss-of-function analysis identified only 3 homeotic protein-coding genes, namely the Ultrabithorax, abdominal-A and Abdominal-B homeobox genes.
Recently, the classically mapped iab-4 locus, located between abdominal-A and Abdominal-B, was recognized to encode two miRNA genes that are transcribed from opposite DNA strands. These generate mir-iab-4 and mir-iab-8, which form similar hairpins as sense and antisense transcripts and are both processed into miRNAs that regulate homeobox genes of the BX-C.25,26,63,64 As one illustration of their regulatory activities, the misexpression of either miRNA transcript induces striking homeotic defects. On the other hand, a deletion mutant that removes the miRNA hairpin appears grossly normal and does not exhibit obvious homeotic segment transformations.64 However, the genetic requirement of the miRNAs is clear, since both sexes were sterile, due to male copulatory and female egg-laying defects.64
Clever genetic manipulations permitted the synthesis of strand-specific miRNA mutants, in which deletion allele was placed in trans to chromosome rearrangements that disrupt either mir-iab-4 or mir-iab-8 transcription. Loss of mir-iab-4 resulted in fertile animals, but animals specifically lacking mir-iab-8 recapitulated the sterile phenotype.64 It remains to be seen how the potential misregulation of homeotic genes, and/or other targets, yields these phenotypes. But these studies nicely highlight how the elegant genetics of model organisms can be exploited to study miRNA activities, even ones transcribed from either strand of the same DNA sequence.
A mutation affecting mir-279 was identified in a screen for mutants disrupting olfactory neuron circuitry.65 Drosophila neurons expressing CO2 sensing odorant receptors are normally confined to the antenna; however, loss of mir-279 results in the ectopic generation of CO2 sensing neurons in the maxillary palp. Computational analysis identified nerfin as a compelling target containing multiple conserved miR-279 sites, and Nerfin was elevated in the maxillary palp of mir-279 mutants. Additionally, the number of ectopic CO2 sensing neurons in the maxillary palp could be suppressed by removal of one copy of endogenous nerfin. However, ectopic Nerfin was itself insufficient to produce ectopic CO2 sensing neurons, suggesting that deregulation of other targets in mir-279 mutants contribute to its phenotype.65
As mentioned, the C. elegans let-7 and lin-4 miRNAs are important for regulating developmental timing. let-7 was the founding member of a miRNA family that is deeply conserved amongst Bilateria.66 Drosophila let-7 is encoded by a polycistron that also produces miR-100 and miR-125 (the orthologue of C. elegans lin-4), i.e., the let-7-Complex (let-7-C). Consistent with a role in regulating developmental timing, the let-7-C is expressed during metamorphosis and is induced by ecdysone, a hormone that controls developmental timing.67 Around 40% of flies deleted for the let-7-C were pupal lethal, and the adult survivors exhibited several behavioral phenotypes including locomotive defects and reduced fertility.68
The roles of the individual let-7-C miRNAs were examined by introducing rescuing transgenes into the deletion background. Interestingly, single mutants of let-7, mir-100 or mir-125 showed normal adult viability, male fertility and near-normal climbing activity, indicating functional overlap amongst these miRNAs. However, female fertility specifically required let-7 activity, while spontaneous locomotion was deficient in flies that lacked either let-7 or miR-125. Analysis of pupal and adult neuromusculature revealed a heterochronic phenotype; animals lacking let-7-C fail to complete larval to adult remodelling, with larval muscles and neurons that normally die soon after eclosion persisting to mature adult stages. let-7 proved to be primarily responsible for this heterochronic phenotype, demonstrating that let-7 has a conserved role in regulating developmental transitions.68
Many other miRNA genes reside in clusters, and it seems reasonable to suspect that they might be coexpressed for a good reason. However, there are few compelling examples of related functions amongst unrelated members of a miRNA cluster. Perhaps the best illustration of this is with the mir-309/mir-3/mir-286/mir-4/mir-5 mir-6-1,2,3 cluster (i.e., miR-309 cluster). This 8-gene cluster encodes 6 distinct miRNAs that are amongst the earliest zygotically transcribed miRNAs in Drosophila. Homozygous mutants exhibit 20% larval lethality, but the 80% that survived were viable and fertile as adults.
Microarray analysis revealed that mir-309 cluster mutants upregulated a large cohort of maternal genes just after the normal onset of zygotic transcription (~2 hr after egg laying). These upregulated genes were strongly enriched for occurrences of four different miRNA seeds represented by members of the miR-309 cluster.33 Importantly, these enrichments were present at 2–3 hr but not at 0–1 hr, consistent with the notion that the miR-309 cluster miRNAs act in concert to degrade maternal transcripts during the maternal-to-zygotic transition (MZT). Interestingly, a similar role was reported for the miR-430 family during zebrafish MZT.19 Since miR-430 is unrelated to any members of the miR-309 cluster, miRNAs seem to have been used in an evolutionarily convergent fashion to promote timely maternal-to-zygotic transition by preferential inhibition of maternally deposited transcripts.
Loss-of-function studies of individual miRNAs have been only recently been begun to be carried out on mice. The first murine miRNA to be deleted was mir-1-2, which is highly conserved both in primary sequence and in tissue of expression in multiple invertebrate and vertebrate species. Murine mir-1-1 and mir-1-2 reside on different chromosomes but produce identical miR-1 products; both are co-transcribed with paralogous copies of miR-133. Mice homozygous for a specific deletion of mir-1-2 exhibit a broad spectrum of heart defects, causing 50% lethality before weaning. The postnatal survivors exhibit a range of defects, including cardiomyocyte hyperplasia, defects in cardiac electrophysiology, and even sudden death.69 Therefore, miR-1 has been shown by loss-of-function studies to have substantial roles in muscle growth and/or physiology in worms, flies and mice. These findings underscore the possibility that segregating miRNA alleles might contribute to heart disease in humans.
The mammalian heart responds to stress by hypertrophic growth involving downregulation of alpha myosin heavy chain (αMHC) and upregulation of beta MHC (βMHC). αMHC hosts the cardiac miRNA mir-208. Specific deletion alleles of mir-208 (that do not affect αMHC function) were homozygous viable and fertile, but began to display heart defects after 20 weeks of age.70 These mutants had lowered expression of βMHC in adults, and under stress conditions that normally induce βMHC. Conversely, transgenic mice expressing mir-208 under the control of the αMHC promoter showed significantly increased levels of βMHC.
These regulatory effects involve direct repression of thyroid hormone receptor associated protein-1 (THRAP-1) by miR-208. THRAP-1 mediates the transcriptional response to thyroid hormone, including repression of βMHC. Thus, the elevated THRAP-1 in mir-208 mutant mice results in enhanced repression of βMHC transcription. This regulatory network provides an interesting precedent for how an embedded RNA gene (mir-208) can impart a regulatory function to a structural gene (αMHC), that allows it to influence the activity of its paralog (βMHC).
BIC is a non-coding RNA that was initially identified as a common retroviral insertion site in B cell lymphomas induced by avian leukosis virus,71 and appears to constitute the primary transcript for miR-155. Deletion of mir-155 results in viable and fertile mice, but they develop lung pathology with age, characteristic of increased remodelling of the airways; they were also immunodeficient.72 At the cellular level, B cells, T cells and dendritic cells were all abnormal in mice lacking mir-155.72,73 Looking more closely in B cells, a parallel study found that in animals lacking BIC/mir-155, the emergence of IgG1 class switched B cells producing high affinity antibodies was impaired in an immune response.74 This study also identified the transcription factor Pu.1 as a direct miR-155 target, whose overexpression in wild-type B cells also impairs the emergence of IgG1 switched cells.
Amongst other hematopoetic miRNAs, miR-150 is one that is expressed in mature resting B and T cells but not in their progenitor cells. Bioinformatic analysis suggested that the transcription factor c-Myb, which is required for multiple stages of B cell development, was a likely target of miR-150.75 Consistent with this, the expression of c-Myb is temporally complementary to the expression of miR-150. Mice lacking c-Myb in B cells display a severe block of B cell development at the pro to pre-B cell transition, and are deficient in the B1 subset of mature B cells.76 The precise level of c-Myb is so critical that even c-Myb heterozygotes show reduced numbers of B1 cells. In contrast, mice lacking miR-150 had elevated levels of c-Myb and increased B1 cells.75 Thus, miR-150 sets a precise level of c-Myb function in B cells, reminiscent of a tuning miRNA-target relationship.
Another miRNA that functions in the development of immune cells, although in the myeloid rather than lymphoid lineage, is miR-223. Expression of miR-223 is low in uncommitted myeloid precursor cells and increases steadily as granulocytic differentiation proceeds; miR-223 is subsequently repressed in the subset of cells that differentiate along the monocyte lineage.77 Mice deleted for mir-223 showed elevated levels of circulating neutrophils with hypermature morphology and granulocyte hyperplasia in the bone marrow. Interestingly, hematopoietic chimeras with equal numbers of wild-type and mir-223 mutant bone marrow cells revealed that mir-223 mutant neutrophils have a competitive advantage over wild-type cells. mir-223 mutant mice also have severe immune reactions to mild stimuli resulting in significant pathology.77
Computational predictions of revealed Mef2c as the strongest candidate for direct regulation with two potential binding sites in the 3′UTR. High levels of Mef2c enhance proliferation of granulocyte-monocyte progenitors,78 and Mef2c is indeed upregulated in mir-223 mutant granulocyte-monocyte progenitors. More compellingly, conditional loss-of-function of Mef2c in granulocyte-monocyte precursors in a mir-223 mutant background rescued the proliferation defect of mir-223 mutants, demonstrating it as a key functional target.
Overexpression of miRNAs is a feature of many cancers. A classic example is the mir-17/18a/19a/20a/19b-1/92-1 polycistronic miRNA transcript (the “miR-17-92 cluster”), which is overexpressed and likely causal to B cell lineage cancers.27,79,80 This cluster has two paralogous clusters; the mir-106a-363 and mir-106b-25 clusters. Mouse knockouts of the three clusters revealed that only mir-17-92 is required individually for mouse development. These mutants were smaller than their littermates and exhibited lung hypoplasia and a ventricular septal defect. Consistent with the fact that miR-17-92 can drive B cell cancers, zygotic or conditional loss of mir-17-92 exhibited a significant decrease in pre-B cells at fetal and adult stages. This is primarily due to an increase in apoptosis caused by accumulation of the pro-apoptotic protein Bim,81 a direct target of members of the miR-17-92 cluster and its paralogs.
Double and triple cluster mutants were also analyzed. Although no compounding effects of deletion of the mir-106a-363 cluster were observed in any combination, mice homozygous for deletions of both mir-17-92 and mir-106b-25 clusters showed greatly enhanced phenotypes, indicating that one or more members of the miR-106b-25 cluster can partially compensate for members of the miR-17-92 cluster.81 These studies nicely illustrate how careful genetic analyses can reveal both overlapping and distinct roles for nearly identical miRNA loci scattered about the genome.
The analysis of individual miRNAs and miRNA clusters by loss of function analyses has begun to shed light on the endogenous roles of miRNAs. Perhaps the most striking meta-observation of the studies to date is that while the reports on fly and mouse miRNA mutants collectively reveal a cornucopia of compelling phenotypes, the systematic deletion of all worm miRNAs revealed only a handful of loci with overt morphological or behavioral phenotypes. Why is this?
One might wonder if the published fly and mouse mutants represent “special” genes that warranted publication, and that the eventual systematic deletion of all miRNAs will reveal many more lacking overt phenotypes. Is there something about the identities of the published mutants that suggests they were more likely to have defects? miRNA conservation and predicted target numbers are correlated, so that well-conserved miRNAs seemingly have been more-entrenched in regulatory networks. However, many of the fly mutants with phenotypes are miRNAs that lack homologs in vertebrates (Table 1), so this does not explain much. Neither are the miRNA mutants preferentially ones that are single copy in the genome (Table 1). For example, miR-279, miR-9a and several genes in the miR-309 cluster each have multiple family members in the Drosophila genome. Reciprocally, mice lacking murine miR-1-2 and the miR-17-92 cluster have clear phenotypes, despite having identical copies located elsewhere in the genome. Therefore, there is no clear correlation between the evolutionary age of miRNA or its singularity in the genome, and whether its corresponding mutant yielded an obviously scorable defect (Table 1). This is true even for the few well-characterized worm mutants, since let-7 has an essential function despite six other let-7 family members in C. elegans, while lsy-6 has an essential neural function despite being a nematode-specific gene.
Another potential explanation regards the complexity of these organisms. C. elegans has several orders of magnitude fewer cells that Drosophila or mice and has a much shorter lifespan, yet it has comparable numbers of protein coding and miRNA genes. Perhaps C. elegans has greater developmental or physiological robustness as a result. At the same time, C. elegans is in many respects a highly derived organism that has deleted many conserved signaling genes that are otherwise essential components in flies and vertebrates.82 Perhaps worms have decreased their reliance on many miRNAs as well.
The sensitivity and specificity of phenotypic assays are also critical in determining miRNA function. Many of the murine miRNA deletions are viable and fertile. However, as many of the mouse miRNAs studied are specific to the immune system, highly sensitive immunological assays were available that proved to identify clear phenotypes. The adult sensory organ phenotype of the Dm-mir-9a mutant consists of just a couple extra bristles out of many hundreds on the fly exterior;32 nevertheless, this phenotype is easily distinguished by simple visual inspection of living animals. The lsy-6 miRNA is expressed by only a handful of cells in the entire nematode, and controls a single left-right neuron pair.45 In this case the phenotype is “obvious”, but only if you happen to be a laboratory that studies this particular left-right neuron pair. These considerations highlight the fact that the success of reverse genetic approaches will depend on the richness of available phenotypic assays, the ease of their execution, and the availability of corroborating information (such as expression pattern) that would inform their study. However, as many fly and mouse miRNA mutants seem to exhibit detectable phenotypes, this strongly supports the notion that individual miRNAs are candidates to influence human traits and disease.
miRNA mediated gene regulation in the biology of multicellular organisms is being investigated at an ever increasing rate. Since miRNAs exert their influence on mRNAs, they must necessarily act downstream of transcriptional regulatory mechanisms. Nevertheless, accumulating evidence suggests that the refinement of gene expression patterns and levels by miRNAs is of critical importance in maintaining the fidelity and precision of cellular and developmental programs. It appears that in many cases the consequence of loss of individual miRNAs is likely to have only subtle phenotypic effects, which are most likely to be readily observed in model organisms with higher levels of complexity and phenotype detection. Several mouse miRNA mutants are viable but exhibit with phenotypes reminiscent of cardiac or immune diseases. Future studies are sure to delve further into the contribution of miRNA loss-of-function as causative mutations for human diseases.
We thank Alex Flynt for critical reading and discussion. E.C.L. was supported by the V Foundation for Cancer Research, the Sidney Kimmel Cancer Foundation, and the National Institutes of Health (GM083300).