MicroRNA Regulatory Networks in Cardiovascular Development

Abstract

The heart, more than any other organ, requires precise function on a second-to-second basis throughout the lifespan of the organism. Even subtle perturbations in cardiac structure or function have catastrophic consequences, resulting in lethal forms of congenital and adult heart disease. Such intolerance of the heart to variability necessitates especially robust regulatory mechanisms to govern cardiac gene expression. Recent studies have revealed central roles for microRNAs (miRNAs) as governors of gene expression during cardiovascular development and disease. The integration of miRNAs into the genetic circuitry of the heart provides a rich and robust array of regulatory interactions to control cardiac gene expression. miRNA regulatory networks also offer opportunities for therapeutically modulating cardiac function through the manipulation of pathogenic and protective miRNAs. We discuss the roles of miRNAs as regulators of cardiac form and function, unresolved questions in the field, and issues for the future.

Introduction

The heart is the first organ to form and function in the embryo, and all subsequent events in the life of the organism depend on its uninterrupted contractility. Heart development is an extraordinarily complex process involving the integration of multiple cell lineages into the three-dimensional organ and its seamless connection to the vascular system. Abnormalities in heart formation result in congenital heart disease, the most common human birth defect, and abnormalities in function of the adult heart result in a spectrum of fatal disorders, including arrhythmias, cardiomyopathies, heart failure and sudden death (Bruneau, 2008; Olson and Schneider, 2003).

Heart formation and function are precisely controlled by networks of transcription factors that link upstream signaling systems with the protein-coding genes required for cardiac myogenesis, morphogenesis and contractility (Olson, 2006). It has recently become apparent that cardiac signaling and transcriptional pathways are intimately intertwined with a collection of microRNAs (miRNAs) that act as “rheostats” and “switches” to modulate multiple facets of cardiac development, function and disease. The myriad roles of miRNAs in the cardiovascular system exemplify the power of miRNAs to modulate complex phenotypes at the cellular and organ levels and highlight many fascinating, unanswered questions about this new frontier in biology. Here we review recent studies that illustrate the mechanisms of miRNA action during cardiovascular development and the important issues for the future.

miRNA biogenesis and function

miRNAs are a class of 21-25 nucleotide non-coding RNAs that are evolutionarily conserved in metazoans (Bartel, 2004). There are estimated to be as many as 1000 miRNAs encoded by the human genome, about half of which have been cloned and confirmed. Based on their locations in the genome, these regulatory miRNAs can be classified as intergenic, intronic, and exonic miRNAs. Intergenic miRNAs are derived from their own transcriptional units in the intergenic regions of the genome (Bartel, 2004). Intronic and exonic miRNAs are located within the introns and exons of host genes (protein-coding or non-protein coding genes), respectively, and are usually co-transcribed and co-expressed with their host genes. A subset of miRNAs are also encoded within introns, but are transcribed in the opposite orientation of the host genes and have their own cis-regulatory elements. Approximately half of all miRNAs are encoded by polycistronic transcription units that generate multiple miRNAs (Bartel, 2004).

Most animal miRNAs share common biogenesis and effector pathways (Bartel, 2004). miRNAs are transcribed by RNA polymerase II as precursor molecules called pri-miRNAs, which can encode single or multiple miRNAs (Cai et al., 2004; Lee et al., 2004). Pri-miRNAs fold into hairpin structures containing imperfectly base-paired stems and are processed by the endonuclease Drosha and its cofactor DGCR8 into 60-100 nt hairpins known as pre-miRNAs (Fig. 1) (Denli et al., 2004; Gregory et al., 2004; Lee et al., 2003). The pre-miRNAs are exported from the nucleus to the cytoplasm by exportin 5 (Yi et al., 2003), where they are cleaved by the endonuclease Dicer to yield imperfect miRNA-miRNA* duplexes (Chendrimada et al., 2005). The miRNA strand is selected to become a mature miRNA, while most often the miRNA* strand is degraded. Occasionally, both strands give rise to functional miRNAs. The mature miRNA is incorporated into the RNA induced silencing complex (RISC), which recognizes specific targets and induces post transcriptional gene silencing (Khvorova et al., 2003; Schwarz et al., 2003). Several mechanisms have been proposed for this mode of regulation: miRNAs can inhibit translational initiation, mark target mRNAs for degradation by deadenylation, or sequester targets into cytoplasmic P bodies (Filipowicz et al., 2008). In rare cases, animal miRNAs also lead to mRNA cleavage, similar to plant miRNAs. While miRNAs act most commonly to repress their mRNA targets, in rare cases they have also been reported to stimulate mRNA translation (Vasudevan et al., 2007); in one case by over 100-fold (Cordes et al., 2009). Recently, miRNAs have been detected in circulating plasma microvesicles (also called exosomes), indicating that miRNAs can be secreted and may mediate intercellular communication (Gibbings et al., 2009; Hunter et al., 2008; Valadi et al., 2007).

Figure 1. MicroRNA biogenesis.

miRNAs are transcribed as long precursors (pri-miRNAs), which are processed by Drosha and DGCR8 into hairpins called pre-miRs, which are processed by Dicer and TRBP to form mature miRNAs as a heteroduplex with miRNA*. miRNAs regulate numerous processes as shown.

Multiplicity of miRNA targets

miRNA target identification remains a major challenge. Most metazoan miRNAs bind to the 3′ UTRs of target mRNAs, although there are also cases in which miRNAs repress mRNA targets through binding sites in 5′ UTRs or coding regions (Kloosterman et al., 2004; Lytle et al., 2007). Unlike plant miRNAs that base pair to mRNAs with nearly perfect complementarity and act as siRNAs to promote mRNA decay, metazoan miRNAs most commonly form imperfect base pairs with target sequences. Therein lies the difficulty in target identification. Nucleotides 2 to 8 of the miRNA, termed the “seed” sequence, are essential for target recognition and binding (Bartel, 2009). While pairing beyond the seed sequence has only modest consequences for target recognition (Bartel, 2009), in some cases, pairing to the 3′ portion of the miRNA can supplement the 5′ seed match and compensate for a single mismatch in the seed region (Vella et al., 2004) (Yekta et al., 2004). Other variables, such as positions of miRNA binding sites, site accessibility, RNA secondary structure, and proximity of sites for other miRNAs and RNA binding proteins also influence miRNA:mRNA interactions (Bartel, 2009).

MiRNAs have numerous high and low affinity targets, averaging 300 conserved targets per miRNA family (Bartel, 2009). Individual 3′ UTRs also contain binding sites for multiple miRNAs, allowing for redundancy and cooperation between miRNAs. In addition, the proteins encoded by mRNA targets of miRNAs can themselves modulate the expression of additional miRNAs, increasing the complexity of miRNA-dependent gene regulatory circuits. It is likely therefore that miRNAs participate in nearly every physiological process.

Because miRNAs generally exert their effects through relatively subtle modulation of numerous targets, rather than through dramatic regulation of single targets, they are often considered to function as “fine-tuners” that eliminate the expression of low level transcripts that are inappropriate for a particular cell type, thereby providing robustness to cellular phenotypes (Bartel, 2009). A classic example is the muscle-specific miRNA miR-1 which, when mis-expressed in non-muscle cells, shifts the mRNA expression profile toward that of muscle (Lim et al., 2005). However, there are also settings in which miRNAs function as “on-off” switches (Reinhart et al., 2000; van Rooij et al., 2007). In perhaps the clearest example of such regulation described to date, the cardiac-specific miRNA, miR-208, regulates a cohort of transcriptional repressors that synergize to silence the expression of slow muscle fiber genes (van Rooij et al., 2007, 2009). In the absence of miR-208, up-regulation of each repressor by roughly 2-fold results in cooperative and virtually complete inhibition of down-stream target genes of these repressors.

By regulating the expression of multiple proteins that function at different steps in complex biological pathways, miRNAs are able to exert powerful effects on cellular processes such as growth, differentiation, metabolism and apoptosis (Ambros, 2004; Kloosterman and Plasterk, 2006). Thus, although the effects of an indivdual miRNA on a specific mRNA target may be relatively modest, the combined effects of a miRNA on multiple targets functioning within a common pathway can be synergistic. In this respect, miRNAs act in a manner distinct from that of classical drugs, which are designed to impose maximal inhibition on single targets. Instead, by modestly inhibiting multiple targets with shared functional consequences, miRNAs are able to circumvent possible redundant mechanisms that might bypass a single inhibited target.

Validation of miRNA targets and identification of specific targets responsible for miRNA-dependent phenotypes represent significant challenges in the field. Current experimental tools for target validation rely on reporter assays in which 3′ UTRs containing the putative miRNA binding site are linked to artificial reporter genes and repression by miRNAs is analyzed by reporter activities in transient overexpression assays. mRNA and protein levels of putative targets are also analyzed in vitro. However, because in vitro assays involve transient over-expression and knockdown of miRNAs, targets validated in vitro may not represent bona fide targets that respond to physiological levels of miRNAs in vivo.

In addition to regulating downstream mRNA targets, miRNAs often operate through feedback loops to ensure precise control of both their own expression and that of their targets (Fig. 2). In feed-forward loops, miRNAs can enhance gene expression through repressing negative regulators of a pathway, such as in the case of miR-1 and histone deacetylase 4 (HDAC4) in muscle cells. Repression of HDAC4 by miR-1 de-represses the activity of the MEF2 transcription factor which, in turn, activates expression of miR-1 and other target genes (Fig. 2A) (Chen et al., 2006). miRNAs can also establish negative feedback loops by repressing activators of miRNA/mRNA expression. The reciprocal regulatory interaction between miR-133 and SRF in muscle cells is an example of such a regulatory loop (Chen et al., 2006; Liu et al., 2008). In this case, SRF regulates the expression of miR-133, which represses expression of SRF, providing feedback control to the pathway (Fig. 2B).

Figure 2. miRNA feedback loops.

In feed-forward regulation (A), miRNAs repress a repressor, leading to activation of transcription factors that activate miRNAs expression. The regulatory loop formed between miR-1, HDAC4 and MEF2 exemplifies this form of regulation. In negative feedback loop (B), miRNAs repress transcription factors that are required for miRNA expression, leading to decreased expression of miRNAs. The regulatory loop formed between miR-133 and SRF exemplifies this form of regulation.

Especially intriguing is the finding that the actions and expression of certain miRNAs are modulated under conditions of cell stress (Leung and Sharp, 2007). This facet of miRNA biology is of particular importance in diseases of the heart in which miRNAs have been shown to exert protective and pathogenic functions (Divakaran and Mann, 2008; Latronico and Condorelli, 2009; van Rooij and Olson, 2007). The precise mechanisms whereby stress signals potentiate miRNA action have not been fully defined, but there is evidence for stress-dependent modulation of miRNA processing and association of miRNAs and mRNA targets with P bodies, the sites of mRNA repression (Bhattacharyya et al., 2006).

Requirements of cardiac miRNAs revealed by Dicer deletion

The consequences of inhibition of miRNA expression have been examined by deletion of Dicer, the RNase essential for miRNA biosynthesis. Cardiac deletion of Dicer, using Cre recombinase under control of cardiac regulatory DNA sequences, results in lethality at different developmental stages depending on the temporal expression pattern of the Cre transgene. Deletion using Nkx2.5 regulatory elements, which direct expression in cardiac progenitor cells by embryonic day (E) 8.5, causes lethality at E12.5, with hearts displaying pericardial edema and hypoplastic ventricular myocardium (Zhao et al., 2007). Post-natal deletion of Dicer in cardiomyocytes using Cre recombinase under control of the α-myosin heavy chain (MHC) promoter causes dysregulation of cardiac contractile proteins and sarcomere disarray, resulting in dilated cardiomyopathy, heart failure, and lethality (Chen et al., 2008). Similarly, deletion of Dicer in the adult heart using a tamoxifen-inducible Cre recombinase causes heart failure and death (da Costa Martins et al., 2008). While these studies demonstrate the importance of miRNAs in heart development and function, it is unclear whether the severe phenotypes resulting from Dicer deletion reflect requisite roles of specific miRNAs or, more likely, the collective functions of numerous miRNAs in multiple essential processes. Pinpointing specific miRNAs that mediate the actions of Dicer is complicated by the paradoxical finding that numerous miRNAs are up-regulated in hearts of adult mice following cardiac Dicer deletion, which likely reflects secondary changes in miRNA expression in non-cardiomyocytes within the heart (da Costa Martins et al., 2008).

The muscle-specific miR-1/133 gene clusters

miR-1 was the first miRNA to be implicated in heart development (Zhao et al., 2005) and is the most extensively studied miRNA in muscle. Nearly identical in sequence in organisms ranging from fruit flies to humans, miR-1 is expressed specifically in striated muscle cells in all organisms. In vertebrates, miR-1 and miR-133 are generated from a common bicistronic transcript (Fig. 3A), whereas these miRNAs are transcribed separately in invertebrates, suggesting they became genetically linked during the evolutionary emergence of vertebrates (Chen et al., 2006). In contrast to the muscle-specificity of miR-1, miR-133 is expressed in non-muscle cell types of invertebrates, indicating that it became incorporated during evolution into the miR-1 locus and thereby acquired muscle-specificity. The miR-1/133 locus also duplicated twice to yield three related miRNA clusters (miR-1-1/133a-2; miR-1-2/133a-2 and miR-206/133b) (Fig. 3A). The former two clusters are expressed in cardiac and skeletal muscle, whereas the latter cluster is skeletal muscle-specific (Chen et al., 2006; Lagos-Quintana et al., 2002; Rao et al., 2006; Sempere et al., 2004; Zhao et al., 2005).

Figure 3. Genomic organization and transcriptional regulation of the miR-1/miR-133 cluster.

(A) Bicistronic miR-1/miR-133 clusters and the muscle tissues in which they are expressed are shown. (B) Transcriptional regulation of miR-1/miR-133a genes in heart and skeletal muscle. The upstream and intragenic enhancers of each locus, and the transcription factors that act on these cis-regulatory elements are shown. The expression patterns of lacZ transgenes controlled by the intragenic enhancers of each locus are shown in mouse embryos at E11.5. The bottom panel shows expression patterns of these enhancers in transverse sections of the hearts. a, atrium; lv, left ventricle; rv, right venticle. Embryo images are from Liu et al 2007.

Cardiac- and skeletal muscle-specific expression of the miR-1/133a clusters are regulated by upstream and intragenic enhancers controlled by combinations of the MADS-box transcription factors SRF and MEF2 and the basic helix-loop-helix (bHLH) protein MyoD (Fig. 3B) (Liu et al., 2007; Zhao et al., 2005). The same transcription factors activate protein coding genes involved in muscle function (e.g. sarcomere genes), demonstrating the intimate integration of miRNA and mRNA regulatory networks. The existence of multiple independent enhancers of the miR-1/133 genes allows for fine-tuning of temporal-spatial control of gene expression, and provides a means of reinforcing their cardiac and skeletal muscle specific expression during development.

The regulatory connection between SRF/MEF2 and miR-1 was conserved during evolution. In flies, for example, SRF also regulates the cardiac expression of miR-1, and MEF2 cooperates with the bHLH transcription factor Twist to control expression in the skeletal musculature (Kwon et al., 2005; Sokol and Ambros, 2005). Thus, the key transcription factors for myogenesis were locked in with muscle-specific miRNAs to establish an integrated transcriptional-miRNA regulatory network that controls fundamental aspects of myogenesis.

The remarkable evolutionary conservation of miR-1 and miR-133 sequences raises interesting questions about the mechanism of action of these miRNAs and the basis for the apparent selection pressure to maintain their sequences across vast evolutionary distances. If the primary determinant of mRNA target recognition is the seed sequence and variability is tolerated at other positions, why are these miRNAs so highly conserved across their entire lengths? Most likely, conserved nucleotides outside the seed play important roles in aspects of miRNA biogenesis or functions independent of mRNA recognition (e.g. interactions with accessory factors or other regulatory components) or perhaps these nucleotides mediate completely different functions (e.g. transcriptional control). It is also intriguing that these miRNAs are so conserved both in their sequence and their muscle-specific expression while their mRNA targets are not evolutionarily conserved. The divergence of mRNA targets suggests that the regulatory circuits controlled by these miRNAs are relatively plastic and acquired or lost during evolution, whereas there must be strong evolutionary pressure to maintain the precise sequences of these miRNAs. Alternatively, over the course of evolution, mRNAs that are crucial for essential cellular functions are likely under selective pressure to avoid being repressed by miRNAs by either changing the cell types where targets are expressed so that targets and miRNAs are not present in the same tissue, or by mutations in the 3′ UTRs so that recognition by seed sequences is dampened (Bartel, 2009). Therefore, in some cases, nonconserved targets may represent important species-specific repression, which may be critical for the functions of miRNAs.

Control of cardiac cell fates by miR-1 and miR-133

Studies in embryonic stem (ES) cells have revealed functions for miR-1 and miR-133 in specification of mesodermal cell fates. miR-1 and miR-133 are not expressed in undifferentiated ES cells and are up-regulated in early precardiac mesoderm (Ivey et al., 2008). These two miRNAs function in concert to promote mesoderm differentiation in ES cells, while suppressing differentiation into the endodermal or neuroectodermal lineages (Ivey et al., 2008). Remarkably, reintroduction of miR-1 into SRF-null ES cells restores mesodermal gene expression, suggesting that the early functions of SRF in mesodermal lineages are mediated by the regulation of miR-1/133 expression (Ivey et al., 2008).

In contrast to their parallel roles in repressing nonmuscle lineages, miR-1 and miR-133 have antagonistic effects on further differentiation of muscle lineages: miR-1 promotes differentiation of ES cells toward a cardiac fate, whereas miR-133 inhibits differentiation into cardiac muscle (Ivey et al., 2008). miR-1 has been proposed to exert its effects by targeting the Notch ligand Delta-like (Dll-1) (Ivey et al., 2008) (Fig. 4). The levels of miR-1 and miR-133 also increase during spontaneous myocardial differentiation in ES cells in 2-dimensional culture (Takaya et al., 2009). However, in this setting, forced expression of miR-1 and 133 was reported to suppress expression of cardiac genes (Takaya et al., 2009). The discrepancy between these findings in terms of miR-1 function may be attributable to the differences in cell-cell interactions under different culture conditions. Nonetheless, these studies clearly demonstrate the importance of miR-1 and miR-133 in regulating cardiac cell fates during ES cell differentiation.

Figure 4. Functions of miR-1 and 133 in heart and skeletal muscle.

(A) Targets of miR-1 and processes they regulate are shown.

(B) Targets of miR-133 and processes they regulate are shown.

In Drosophila, miR-1 is also expressed in early mesoderm. Mutations in the miR-1 gene cause a spectrum of defects in muscle differentiation ranging from embryonic death to later demise in the larval stages after hatching (Kwon et al., 2005; Sokol and Ambros, 2005). In a subset of miR-1 mutant flies, muscle progenitors are arrested in a proliferative state and accumulate ectopically, implicating miR-1 in regulating the decision to proliferate versus differentiate (Kwon et al., 2005). As in mammalian mesodermal progenitors, regulation of the Notch ligand Delta by miR-1 may contribute to the actions of this miRNA on Drosophila cardiogenesis.

Control of cardiac growth and development by miR-1

miR-1 and miR-133a regulate fundamental aspects of heart development in vivo (Fig. 4). Over-expression of miR-1 in the embryonic heart under the control of β-MHC promoter inhibits cardiomyocyte proliferation and causes lethality at E13.5 due to deficiency of cardiomyocytes (Zhao et al., 2005). Inhibition of cardiac growth by miR-1 was proposed to be mediated by translational inhibition of Hand2, a bHLH transcription factor required for right ventricular growth (Srivastava et al., 1995; Srivastava et al., 1997; Zhao et al., 2005). Although miR-1 and Hand are conserved from flies to mammals, miR-1 does not target Hand in flies, suggesting that the miR-1:Hand regulatory link appeared later in evolution.

Targeted deletion of miR-1-2 in mice causes lethality with incomplete penetrance between late embryogenesis and birth, due to ventricular-septal defects (VSDs) (Zhao et al., 2007) (Table 1). Consistent with the proposed role of miR-1 as a negative regulator of myocyte proliferation via Hand2, miR-1-2 deficient mice display an increase in Hand2 expression and an increased number of proliferating cardiomyocytes (Zhao et al., 2007). Because miR-1-2 null mice express miR-1-1 at normal levels, understanding the complete function of miR-1 in cardiogenesis awaits the compound deletion of both miR-1-1 and miR-1-2 genes in mice. However, the fact that only a 50% reduction in the overall level of miR-1 expression in these mice results is such profound cardiac abnormalities underscores the importance of the precise dosage of this miRNA for normal cardiac development and function.

Table 1.

Cardiovascular Phenotypes of miRNA Mutant Mice

miRNA	Phenotype	Reference
miR-1-2	~50% viable; ~50% w/ VSDs in embryos, and 50% w/ arrhythmias and ecotpic cardiomyocyte proliferation in surviving adults, ~15% w/ cardiac dilation, ventricular dysfunction, and death	Zhao et al., 2007
miR-126	~60% viable; ~40% w/ lethal embryonic and perinatal hemorrhages	Wang et al., 2008 Kuhnert et al., 2008
miR-133a-1	no apparent phenotype	Liu et al., 2008
miR-133a-2	no apparent phenotype	Liu et al., 2008
miR-133a dKO	~56% w/ lethal VSDs at P1; ~24% viable, with disrupted sarcomeres, ectopic cardiomyocyte proliferation, and smooth muscle gene expression	Liu et al., 2008
miR-143	no apparent phenotype	Xin et al., 2009
miR-145	reduced blood pressure, resistance to vascular stenosis	Xin et al., 2009
miR-143/145 dKO	similar to miR-145 KO	Xin et al., 2009 Boettger et al., 2009 Elia et al., 2009
miR-208a	viable; block to stress-dependent cardiac remodeling, reduced slow/increased fast muscle gene expression; cardiac conduction defects	van Rooij et al., 2007 Callis et al., 2009
miR-208b	viable, no apparent phenotype	van Rooij et al., 2009
miR-499	viable, no apparent phenotype	van Rooij et al., 2009
miR-208b/499 dKO	viable; reduced slow/increased fast muscle gene expression	van Rooij et al., 2009
miR-17~92	perinatal lethality w/ VSDs	Ventura et al., 2008
miR-106a~363	viable, no apparent phenotype	Ventura et al., 2008
miR-106b~25	viable, no apparent phenotype	Ventura et al., 2008
miR-17~92/106b~25 dKO	embryonic lethality w/ VSDs, ASDs, thin myocardium	Ventura et al., 2008
miR-17~92/106a~363/ 106b~25 tKO	same as dKO	Ventura et al., 2008

Control of cardiac rhythm and remodeling by miR-1

miR-1 modulates numerous ion channels involved in cardiac conduction. In rats subjected to myocardial infarction (MI), over-expression of miR-1 increased the occurrence of arrhythmias (Yang et al., 2007). Delivery of an antisense oligonucleotide against miR-1 in the rat infarct model reversed the predisposition to arrhythmias. miR-1 was shown to regulate cardiac electrophysiology by regulating the cardiac gap junction protein connexin 43 and the potassium channel subunit Kir2.1 (Yang et al., 2007). Consistent with these findings, miR-1-2-null mice exhibit a spectrum of conduction abnormalities, including a shortened PR interval and a prolonged QRS complex, which cause sudden death (Zhao et al., 2007). Elevated levels of the transcription factor Irx5, another target of miR-1 and a regulator of cardiac conduction channels, may also contribute to cardiac arrhythmias in miR-1-2 null mice (Zhao et al., 2007).

miR-1 is downregulated in several models of cardiac hypertrophy and heart failure (Care et al., 2007) (Ikeda et al., 2009; Sayed et al., 2007) and attenuates cardiomyocyte hypertrophy when overexpressed in cultured cardiomyocytes and in adult heart (Ikeda et al., 2009). This function of miR-1 has been attributed to the inhibition of MEF2A (Ikeda et al., 2009), a transcriptional mediator of pathological cardiac remodeling (Potthoff and Olson, 2007). miR-1 also inhibits calmodulin expression and down-regulates the calmodulin-calcineurin-NFAT signaling pathway (Ikeda et al., 2009), which activates MEF2 and promotes cardiac growth (Crabtree and Olson, 2002).

Control of cardiac growth and development by miR-133

Unlike the sensitivity of the heart to haploinsufficiency of miR-1 expression, mice lacking either miR-133a-1 or miR-133a-2 genes do not display obvious cardiac abnormalities, whereas deletion of both genes results in late embryonic and neonatal lethality due to VSDs and chamber dilatation (Liu et al., 2008) (Table 1). The small subset of miR-133a-1/133a-2 double knockout (dKO) mice that survives to adulthood develops dilated cardiomyopathy, cardiac fibrosis, and heart failure (Liu et al., 2008). miR-133a null mice exhibit excessive proliferation and apoptosis of cardiomyocytes, sarcomere disarray and cardiac fibrosis (Liu et al., 2008). Consistent with the enhanced proliferation of cardiomyocytes in mutant mice, several genes involved in cell cycle control such as cyclins D1, D2 and B1 are upregulated in miR-133a-null hearts, and cyclin D2 was shown to be a direct target of miR-133a (Liu et al., 2008). Conversely, over-expression of miR-133a under the control of β-MHC promoter inhibits cardiomyocyte proliferation in the heart, which results in embryonic lethality due to thinning of ventricular walls and VSDs (Liu et al., 2008). These consequences of over-expression compliment the loss-of-function data to demonstrate that miR-133a inhibits cardiomyocyte proliferation in the developing heart.

miR-133a null mice also display ectopic activation of smooth muscle genes in the developing heart (Liu et al., 2008). During normal heart development, smooth muscle genes are transiently expressed at the heart tube stage (Li et al., 1996; McHugh, 1995; Ruzicka and Schwartz, 1988). The aberrant expression of smooth muscle genes in these mutant mice suggests therefore that miR-133a null hearts are less mature than wild-type hearts. miR-133a represses smooth muscle gene expression in the heart by directly targeting several smooth muscle mRNAs, as well as SRF, a regulator of smooth muscle gene expression (Liu et al., 2008). The genetic interaction between miR-133a and SRF creates a negative feedback loop in which the upregulation of miR-133a by SRF results in increased repression of SRF (Fig. 2B).

miR-133a is down-regulated in ventricular tissue of mice during hypertrophy and various heart diseases (van Rooij et al., 2006) (Care et al., 2007; Duisters et al., 2009). Accordingly, over-expression of miR-133a in cultured cardiomyocytes blunts hypertrophic responses to agonist treatment (Care et al., 2007). Delivery of an antagomiR that blocks activity of miR-133a was shown to induce cardiac hypertrophy in mice, which was mediated by derepression of the Ras homology gene family member A (RhoA) and cell division cycle 42 (cdc42) (Care et al., 2007). In contrast, miR-133a dKO mice develop dilated cardiomyopathy before succumbing to heart failure and sudden death, without evidence of cardiac hypertrophy (Liu et al., 2008). Several possibilities may explain the different consequences of miR-133a inhibition by antagomiRs versus genetic deletion. Because antagomiRs do not completely eliminate the target miRNAs, in contrast to genetic deletion, residual levels of miR-133a might be sufficient for certain functions. In addition, genetic deletion eliminates miRNA expression throughout the life of the organism, which might allow for compensatory pathways that are not operative during transient antagomiR-based knockdown.

miR-133a has been proposed to inhibit apoptosis by targeting of caspase 9 in cultured cardiomyocytes (Fig. 4) (Xu et al., 2007). Interestingly, however, caspase 9 protein levels were unaffected in hearts of miR-133a dKO mice despite increased apoptosis of cardiomyocytes (Liu et al., 2008). miR-133a regulates structural changes in the extracellular matrix of the myocardium by targeting a profibrotic factor, connective tissue growth factor (CTGF) (Duisters et al., 2009). miR-133a has also been implicated in metabolic control of cardiomyocytes by targeting KLF15, thus repressing expression of GLUT4, a glucose transporter important for myocardial energy supply (Horie et al., 2009). miR-133a also represses a potassium channel gene KCNH2, contributing to QT prolongation and the associated arrhythmias in diabetic rabbit hearts (Xiao et al., 2007), and has been implicated in atrial fibrillation (Shan et al., 2009). In addition, miR-133a regulates alternative splicing during myoblast differentiation by regulation of the alternative splicing factor nPTB (Boutz et al., 2007). It is worth noting that most of the above conclusions are based on over-expression of miR-133a, and only a handful of genes shown to be targeted by miR-133a in vitro are affected in miR-133a dKO mice. For example, the surviving adult miR-133a dKO mice do not exhibit defects in cardiac conduction or KCNH2 expression.

The abnormalities in cardiac structure in miR-1-2 and miR-133a mutant mice are similar to those described in mice lacking the miR-17~92 cluster (also known as Oncomir-1), which gives rise to a precursor trancript encoding six miRNAs that can be grouped into four classes based on their seed sequence (Ventura et al., 2008). Mice homozygous for a deletion of this miRNA cluster die perinatally and show VSDs (Ventura et al., 2008) (Table 1). miRNAs encoded by the miR-17~92 cluster are thought to promote cell survival by inhibiting expression of the proapoptotic protein Bim1 (Ventura et al., 2008). Deletion of two related miRNA clusters, miR-106b~25 or miR-106a~363, has no effect on heart development but combination of either of these null alleles with the miR-17~92 null allele results in embryonic lethality accompanied by severe VSDs, ASDs and thin-walled myocardium (Ventura et al., 2008). Together, these studies of miR-1-2, miR-133a and miR-17~92 demonstrate how relatively subtle changes in cardiac gene expression resulting from miRNA deletion can severely perturb cardiac development, causing VSDs, which are the most common form of congenital heart disease in humans (Hoffman and Kaplan, 2002).

Functions of miR-1 and miR-133 in zebrafish

Experimental and computational studies in zebrafish point out key functions of both miR-1 and miR-133 in shaping muscle gene expression and regulating sarcomeric actin structure. miR-1/206 and miR-133 are the two miRNAs with the strongest influence on gene expression in embryonic muscle (Mishima et al., 2009). More than half of the muscle mRNAs upregulated in zebrafish embryos mutant for maternal and zygotic Dicer function are predicted targets of miR-1 and miR-133a (Mishima et al., 2009). miR-1 and miR-133 targets are enriched for actin-binding and vesicle transport functions, and interfering with the function of these miRNAs disrupts sarcomeric organization of actin in fast skeletal muscle.

miR-133 is highly expressed in adult caudal fin and was shown to negatively regulate fin regeneration in zebrafish (Yin et al., 2008). FGF signaling, a critical pathway in regeneration, down-regulates miR-133 during regenerative outgrowth. Forced expression of miR-133 attenuates fin regeneration, whereas antagonism of miR-133 is sufficient to partially rescue fin regeneration that has been inhibited by FGF receptor blockade (Yin et al., 2008). The Mps1 kinase, a positive regulator of blastemal proliferation, is a target of miR-133 such that the absence of miR-133 would be predicted to potentiate signaling through this regenerative pathway. Whether miR-133a plays a role in heart regeneration in zebrafish or other organisms remains to be addressed.

A family of miRNAs encoded by myosin genes

Cardiac muscle contraction depends on the expression of two MHC proteins α-MHC (Myh6) and β-MHC (Myh7). In rodents, α-MHC, a fast ATPase, is the predominant myosin isoform in the adult heart, whereas β-MHC, a slow ATPase, is highly expressed in the developing heart and is down-regulated after birth (Morkin, 2000). Cardiac stress and hypothyroidism result in a switch in myosin content (Gupta, 2007; Krenz and Robbins, 2004), up-regulating β- and downregulating α-MHC, which has profound effects on cardiac contractility and function. β-MHC shares extensive homology with an ancient myosin Myh7b, the function of which is unknown (McGuigan et al., 2004). These three muscle-specific myosin genes (Myh6, Myh7, and Myh7b) encode a family of intronic miRNAs (miR-208a, miR-208b, and miR-499), called MyomiRs, which control pathological cardiac remodeling, muscle myosin content, myofiber identity, and muscle performance (Fig. 5) (Callis et al., 2009; van Rooij et al., 2009; van Rooij et al., 2007).

Figure 5. Functions of MyomiRs in cardiac and skeletal muscle.

MyomiRs are encoded by myosin heavy chain (MHC) genes. Targets of MyomiRs and processes they regulate are shown.

miR-208a, the sole heart-specific miRNA, is embedded in intron 29 of the α-MHC gene (Fig. 5) (van Rooij et al., 2007). Similarly, the β-MHC gene co-expresses a closely related miRNA, miR-208b (Callis et al., 2009; van Rooij et al., 2009). miR-208a and miR-208b have the same seed sequence and differ at only three nucleotides in the 3′ region. A third member of the MyomiR family, miR-499, is encoded by intron 19 of the mouse Myh7b gene (Rossi et al., 2009; van Rooij et al., 2008a; van Rooij et al., 2009). Although the sequence of miR-499 differs substantially from miR-208a and miR-208b, the seed region contains 6 over-lapping bases, implying that these miRNAs share overlapping functions and regulate similar sets of targets. The MyomiRs are conserved from fish to humans, but do not exist in invertebrates; their appearance in evolution correlates with the diversification of different myosin isoforms and the acquisition of stress- and thyroid hormone (T3) -responsiveness of muscle, two key gene programs that they regulate. Interestingly, the MyomiR sequences are the only conserved intronic sequences in these myosin genes, indicating that there was strong selection pressure to maintain these sequences through evolution.

Expression of the MyomiRs parallels the expression of their host genes during development and in response to stress and T3 signaling. Like their host genes, miR-208a is highly expressed in adult mouse heart; whereas miR-208b is abundant in embryonic heart but is expressed at a low level in adult heart (Callis et al., 2009; van Rooij et al., 2009; van Rooij et al., 2006). Myh7b and miR-499 are constitutively expressed at a high level in the heart (van Rooij et al., 2009). While miR-208a is only expressed in the heart, both miR-208b and miR-499 are also expressed in slow skeletal muscles. In response to cardiac stress such as pressure overload, β-MHC/miR-208b is upregulated in the adult heart (Callis et al., 2009; van Rooij et al., 2009). T3 signaling induces α-MHC/miR-208a, whereas hypothyroidism results in a dramatic increase in β-MHC/miR-208b expression (Callis et al., 2009; van Rooij et al., 2009). Inhibition of T3 signaling also upregulates the expression of Myh7b/499 (van Rooij et al., 2009).

Control of myosin expression and muscle performance by MyomiRs

MyomiRs play central roles in the regulation of the α-MHC to β-MHC switch and in other aspects of the stress response and T3 sensitivity of the heart (Fig. 5). Mice lacking miR-208a are viable but are resistant to fibrosis and cardiomyocyte hypertrophy in response to stress (Callis et al., 2009; van Rooij et al., 2007). This function of miR-208a is mediated by repression of T3 receptor coregulator 1 (THRAP1), an activator or repressor of the T3 receptor, and myostatin, a repressor of hypertrophic growth of both skeletal and cardiac muscles (Callis et al., 2009; Cook et al., 2002; Ito and Roeder, 2001; Lee, 2004; van Rooij et al., 2007). Most notably, miR-208 null mice fail to up-regulate β-MHC in response to stress and inhibition of T3 signaling in the adult heart. In addition, fast skeletal muscle genes are inappropriately expressed in miR-208a null hearts (van Rooij et al., 2007) (Table 1).

miR-208a controls not only the expression of β-MHC/miR-208b in response to stress and hypothyroidism, but also the expression of Myh7b/miR-499 in adult heart. Both Myh7b and miR-499 are completely absent in adult hearts of the miR-208a null mice (van Rooij et al., 2009). Over-expression of miR-499 in the heart of miR-208a null mice is sufficient to reactivate the expression of β-MHC/miR-208b and to repress ectopic expression of fast muscle genes, suggesting that miR-499 is a downstream mediator of miR-208a actions (van Rooij et al., 2009). Over-expression of miR-499 in miR-208 null hearts also reactivates the expression of Myh7b, which suggests the existence of a positive autoregulatory loop whereby miR-208a regulates Myh7b and its intronic miRNA, miR-499, which regulates β-MHC (van Rooij et al., 2009) (Fig. 4).

Defects in cardiac conduction have also been reported in miR-208a-null mice. miR-208a is likely to control cardiac conduction by regulating expression levels of the gap junction protein connexin 40 and the homeodomain-only protein (Hop) and GATA4 transcription factors (Callis et al., 2009) (Fig. 5). Consistent with these findings, mice over-expressing miR-208a develop cardiac hypertrophy and arrhythymias (Callis et al., 2009).

In addition to their functions in the heart, MyomiRs also control the myofiber gene program in skeletal muscle (Fig. 5). Gain-of-function and loss-of-function analyses demonstrated that miR-208b and miR-499 redundantly specify muscle fiber identity by activating slow and repressing fast myofiber genes (van Rooij et al., 2009). The actions of MyomiRs in skeletal muscle are mediated by a collection of transcriptional repressors of slow muscle genes, including Sox6, Purβ, and Sp3, which are all repressors of β-MHC expression, and HP-1β, a repressor of MEF2 (van Rooij et al., 2009).

Taken together, these studies demonstrate that MyomiRs, which are embedded in myosin genes, in turn regulate myosin gene expression and muscle identity in response to physiological and pathological signaling. The incorporation of MyomiRs into the introns of myosin genes that they regulate provides an efficient means of ensuring co-regulation of the miRNA and the gene program under its control, rather than creating separate sets of cis-regulatory elements to control expression of the miRNA and the myosin gene. Intriguingly, the regulatory circuitry among MyomiRs and their host genes is only seen in the adult but not in the fetal heart, as β-MHC/miR-208b and Myh7b/miR-499 are expressed normally in neonatal hearts of miR-208a-null mice (van Rooij et al., 2009).

Control of cardiac patterning by miR-138

Zebrafish have a two-chambered heart containing a single atrium and ventricle separated by the atrioventricular canal (AVC) (Stainier, 2001). miR-138 is specifically expressed in the ventricular chamber of the zebrafish heart. Knockdown of miR-138 by morpholino and antagomiR in zebrafish embryos led to expansion of AVC gene expression into the ventricular chamber and failure of ventricular cardiomyocytes to fully mature, indicating that miR-138 is required for cardiac maturation and patterning in zebrafish (Morton et al., 2008). It is worth noting that the requirement of miR-138 during cardiac development is limited only to a distinct temporal window of early cardiac looping, and miR-138 was dispensable thereafter (Morton et al., 2008). miR-138 restricts AVC gene expression in the ventricle by negatively regulating the cell adhesion molecule versican and repressing the retinoic acid signaling pathway (Morton et al., 2008). It will be interesting to study whether other miRNAs have similar functions like miR-138 to “clear” unwanted mRNA expression from a specific region of the heart, thus establishing cardiac patterning. miR-138 is conserved from zebrafish to humans, so it will also be interesting to determine whether it plays similar roles in patterning of the mammalian four-chambered heart.

Control of vascular development and stress-responsiveness by miR-143 and miR-145

In addition to regulation of cardiomyocyte development and function, recent reports have revealed roles of miRNAs in vascular development. Smooth muscle cells (SMCs) are highly plastic and possess the ability to modulate their phenotypes between differentiated, proliferative or migratory states in response to vascular injury or growth factor signaling. This plasticity contributes to a variety of vascular disorders including atherosclerosis, restenosis following angioplasty and hypertension (Owens et al., 2004). The cardiovascular specific miRNAs, miR-143 and miR-145, play key roles in modulating vascular SMC (VSMC) phenotypes.

Regulation of miR-143/145 expression during cardiovascular development

miR-143 and miR-145 have distinct sequences, but are transcribed as a bicistronic unit. The miR-143/145 cluster is expressed in cardiac progenitors in the cardiac crescent as early as E7.5 and throughout the developing heart until E16.5, as well as in the dorsal aorta and other developing vessels (Boettger et al., 2009; Cordes et al., 2009; Xin et al., 2009). In adult mice, expression is restricted to vascular and visceral smooth muscle cells. Independent studies from two groups have revealed a single SRF binding site (CArG box) in the enhancer region that is required for expression in smooth muscle cells (Cordes et al., 2009; Xin et al., 2009). However, conflicting results were obtained in terms of whether SRF also controls cardiac expression. Xin et al. reported that mutations in the single CArG box eliminated both cardiac and smooth muscle expression in embryos, thus making SRF and its co-activators myocardin/MRTF-A solely responsible for both cardiac and SMC expression of miR-143/145. However, Cordes et al. reported that cardiac expression depends on an evolutionarily conserved binding site for Nkx2.5. Although the basis for these differing results is unclear, in SRF-null embryonic heart at E9.5, expression of miR-143 is significantly downregulated, supporting the role of SRF in controlling cardiac expression of the miR-143/145 gene (Niu et al., 2008).

Control of smooth muscle fate by miR-143/145

miR-143 and miR-145 are downregualted in injured and atherosclerotic vessels and in cultured dedifferentiated VSMCs (Cheng et al., 2009; Cordes et al., 2009; Elia et al., 2009). VSMC differentiation genes are upregulated by over-expression of miR-145 and downregulated by inhibition of miR-145 in cultured VSMCs (Cheng et al., 2009). Restoration of miR-145 expression in balloon-injured arteries inhibits neointimal growth (Cheng et al., 2009). These studies demonstrated that miR-145 promotes VSMC differentiation and represses proliferation in response to vascular injury, thus making miR-145 a novel modulator of VSMC plasticity.

Work by Cordes et al. has provided intriguing insights into the roles of miR-143 and miR-145 in controlling VSMC cell fate. Introduction of miR-145 but not miR-143 into neural crest stem cells was sufficient to drive these multipotent cells into functionally mature VSMCs (Cordes et al., 2009). miR-145 can also robustly potentiate myocardin-dependent conversion of fibroblasts into VSMCs. The SRF coactivator myocardin is a master regulator of smooth muscle cell fate, and it is sufficient to reprogram fibroblast cells into VSMCs (Pipes et al., 2006; Wang et al., 2003). Although neither miR-143 nor miR-145 alone was sufficient to reprogram fibroblasts, overexpression of miR-145 in the presence of myocardin robustly enhances the conversion of fibroblasts into VSMCs, and inhibition of miR-145 blocks myocardin-dependent conversion (Cordes et al., 2009). Thus, miR-145 is apparently required for myocardin activity in reprogramming fibroblasts. Especially striking is the finding that miR-145 directly targets myocardin by enhancing rather than decreasing its expression (Cordes et al., 2009). This stimulatory effect is mediated by a miR-145 target sequence in the 3′ UTR of the myocardin mRNA. The mechanistic basis for the dramatic stimulatory effect of miR-145 on myocardin translation remains to be defined. These findings are in contrast to the generally accepted inhibitory roles of miRs on protein expression. In addition to targeting myocardin, miR-145 represses expression of kruppel-like factor 4 (Klf4) and calmodulin kinase II-delta (CamkIId), both of which positively regulate SMC proliferation (Cordes et al., 2009). miR-143, on the other hand, directly represses translation of Elk1, a myocardin competitor and an activator of VSMC proliferation (Cordes et al., 2009; Wang et al., 2004). Based on these results, a model has been proposed in which miR-145 and miR-143 modulate co-activator and co-repressor activity of SRF to promote differentiation and repress proliferation of VSMCs (Fig. 6). In addition to controlling VSMC differentiation in vitro, miR-145 was shown to silence the self-renewal program of human ES cells by directly repressing the core pluripotency factors Oct4, Sox2, and Klf4, thus facilitating ES cell differentiation (Xu et al., 2009).

Figure 6. Model for the regulation of smooth muscle phenotypes and actin dynamics by miR-143 and 145.

miR-143 and 145 are cotranscribed as a bicistronic unit. The targets of these miRNAs and the processes they regulate are shown.

Control of stress-dependent vascular remodeling by miR-143/145

One major question from in vitro cell culture studies is whether miR-145 and miR-143 control VSMC and ES cell differentiation in vivo. In particular, is miR-145 required for myocardin activity on VSMC differentiation or stem cell pluripotency in vivo as reported in vitro? If miR-145 indeed is required for myocardin function, one might expect genetic deletion of miR-145 in mice to cause phenotypes similar to those of myocardin-null mice, which die at E10.5 due to a failure in VSMC differentiation (Li et al., 2003). On the contrary, mice homozygous for miR-143/145 deletion are viable without obvious abnormalities during early embryogenesis (Boettger et al., 2009; Elia et al., 2009; Xin et al., 2009) (Table 1). Thus, neither miR-143 nor miR-145 is essential for VSMC differentiation in vivo, which is in contrast to their reported functions in vitro. Perhaps compensatory mechanisms in vivo override the requirement of miR-143 and/or miR-145 in VSMC differentiation.

In spite of the dispensable roles of miR-143 and miR-145 during embryogenesis the mice lacking these miRNAs do exhibit smooth muscle abnormalities, including thinning of the muscular wall of the aorta and other arteries and a shift from a differentiated toward a synthetic phenotype (Boettger et al., 2009; Elia et al., 2009; Xin et al., 2009). Correspondingly, the contractility of arteries is compromised, accompanied by reduced expression of a subset of differentiated VSMC markers and an increase in synthetic VSMC markers. The smooth muscle defects can be attributed, at least in part, to abnormalities in actin dynamics and cytoskeletal assembly (Xin et al., 2009). The number and prominence of actin stress fibers in VSMCs of mice lacking miR-145 and 143 are significantly reduced. Whether the thinner smooth muscle layers of the aorta is due to defects in differentiation of the neural crest derived SMC lineage during development remains to be determined. Mice lacking miR-145 also display reduced blood pressure (Boettger et al., 2009; Elia et al., 2009; Xin et al., 2009). Especially intriguing is the finding that neointima formation in response to carotid artery ligation is profoundly impeded in mice lacking miR-145, or both miR-143 and miR-145. The dKO mice also develop neointimal lesions in the femoral arteries at 18-months of age (Boettger et al., 2009). miR-143 and 145 both target numerous genes involved in the regulation of SRF activity and actin dynamics, thus creating a complex set of feedback loops to modulate cytoskeletal assembly and dynamics (Xin et al., 2009). An unbiased genomic/proteomic analysis allowed identification of other miR-143/145 targets involved in contractility of VSMCs, such as angiotensin-converting enzyme (ACE) (Boettger et al., 2009). Knockdown of miR-145 in zebrafish results in defects in gut smooth muscle and epithelium maturation, which have been ascribed to dysregulation of GATA6 (Zeng et al., 2009). Together, the in vivo studies of miR-143 and miR-145 have demonstrated that these two miRNAs function in multiple pathways to control VSMC phenotypes and actin remodeling (Fig. 6). Despite being expressed at high levels in the developing heart, no cardiac function of miR-143 or miR-145 has yet been found.

It is puzzling that transient over-expression of miR-145 in vitro or in the vessel wall in vivo blocks smooth muscle cell growth and promotes differentiation, while genetic deletion of miR-145 in mice also appears to block vascular growth in response to mechanical injury. How might these seemingly contradictory findings be reconciled? One explanation is that transient over-expression assays reveal a subset of potential miRNA functions resulting from acute modulation of down-stream targets in response to non-physiological levels of miRNA expression, whereas genetic deletion of miRNAs in vivo reveals the function of the miRNA in the setting of persistent deletion and associated compensatory mechanisms that do not exist in short-term over-expression studies. In addition, there is precedent in other systems for over-expression and knockdown of miRNAs having similar consequences, which likely reflects their roles in buffering fluctuations in gene expression such that tilting the balance of a particular target one way or the other perturbs that process (Choi et al., 2007).

Control of angiogenesis by miR-126

Endothelial cells (ECs) form the internal surfaces of the vasculature and are essential for vascular development and disease (Carmeliet, 2003). Mounting evidence indicates that miRNAs are important regulators of vascular development and angiogenesis. Mice with EC-specific deletion of Dicer display defects in postnatal angiogenesis (Suarez et al., 2008). Numerous miRNAs have also been shown to regulate tumor angiogenesis in vitro and in vivo, such as miR-378, miR-296 and the miR-17~92 cluster (Dews et al., 2006; Lee et al., 2007; Wurdinger et al., 2008).

The EC-specific miR-126 is encoded by an intron of the EGF-like domain 7 (Egfl7) gene, which encodes a growth factor involved in cell migration. EC-specific expression of Egfl7/miR-126 is controlled by Ets transcription factors (Wang et al., 2008). Deletion of miR-126 in mice causes leaky vessels, hemorrhaging, and partial embryonic lethality (Kuhnert et al., 2008; Wang et al., 2008). Similarly, knockdown of miR-126 in zebrafish results in loss of vascular integrity and hemorrhage during embryonic development (Fish et al., 2008). Surviving miR-126-null mice display defects in cardiac neovascularization following MI, resulting in cardiac rupture and death. miR-126-null ECs are also defective in angiogenesis in response to angiogenic factors (Wang et al., 2008). Recently, knockdown of miR-126 with an antagomiR has been shown to inhibit the ischemia-induced angiogenic response (van Solingen et al., 2009). The pro-angiogenic actions of miR-126 have been ascribed to the negative regulation of Spred1 (the Sprouty-related protein 1) and PIK3R2 (phosphatidylinositol 3-kinase p85beta), negative regulators of the MAP kinase and the PI3 kinase signaling pathways, respectively (Fish et al., 2008; Wang et al., 2008) (Fig. 7). miR-126 also plays a role in vascular inflammation by targeting vascular cell adhesion molecule 1 (VCAM-1) (Schmidt et al., 2007) (Fig. 7). In addition, miR-126 is downregulated in many cancer lines and inhibits tumorigenesis by targeting CT10 regulator of kinase (CRK), PIK3R2, insulin receptor substrate-1 (IRS-1), and other genes (Crawford et al., 2008; Guo et al., 2008; Tavazoie et al., 2008; Zhang et al., 2008) (Fig. 7). Interestingly, a recent report showed that miR-126 inhibits proliferation of non-small cell lung cancer (NSCLC) cells by positively regulating expression of its host gene Egfl7 (Sun et al., 2010).

Figure 7. Model for miR-126 function in endothelial cells.

miR-126 is encoded by an intron of the Egfl7 gene. Targets of miR-126 and the processes they regulate are shown.

In miR-126 null mice, expression of the host gene Egfl7 is not changed at the mRNA or protein level (Wang et al., 2008). Interestingly, miR-126 null mice have similar vascular abnormalities to those of mice with a targeted deletion of the Egfl7 gene, raising the question whether the phenotype of this strain of Egfl7 null mice is attributable to loss of miR-126 function (Schmidt et al., 2007). Indeed, another recently generated strain of Egfl7 null mice, in which miR-126 expression is unaffected, is phenotypically normal (Kuhnert et al., 2008), suggesting that the vascular abnormalities of the former strain of Egfl7 null mice are attributable to the deletion of miR-126. This finding highlight the possibility other gene deletions in mice may have inadvertently deleted miRNAs embedded in introns of protein-coding genes resulting in phenotypes that actually reflect the absence of the embedded miRNA. Evaluation of intronic miRNAs should therefore be a general consideration in the design of gene targeting strategies.

Recently, miR-92a, another EC-specific miRNA, was shown to control angiogenesis and functional recovery of damaged tissues after MI (Bonauer et al., 2009). Numerous other miRNAs have been shown to play key roles in tumor angiogenesis (Wang and Olson, 2009), but their potential functions in blood vessel development during embryogenesis and cardiovascular development have not yet been explored.

miRNAs as modifiers of stress responses and heart disease

Even though we provide examples of miRNAs that play key roles in numerous facets of heart development and disease, many other miRNAs can be deleted in the heart with relatively minimal consequences on cardiac development. This observation suggests that subtle changes in gene expression (<2-fold) resulting from miRNA deletion can be tolerated in vivo (at least in a controlled setting) without perturbing tissue homeostasis. Similarly, there are numerous cases in Drosophila and C. elegans in which miRNA deletion evokes minimal phenotypes (Li and Carthew, 2005; Li et al., 2006; Miska et al., 2007), yet many miRNAs are evolutionarily conserved, indicative of important functions. In contrast, there are now numerous examples, particularly in the heart, in which the absence of specific miRNAs results in a pronounced response to stress – either exacerbating the consequences of the insult or protecting the heart from injury (van Rooij and Olson, 2007).

How might the absence of phenotypes of miRNA mutant mice, and the manifestation of phenotypes in response to stress be explained? One explanation for such tolerance is the phenomenon of network buffering, in which multiple, compensatory regulatory interactions buffer the effects of loosing one node in a complex web of regulatory interactions (Bartel, 2009), whereas under conditions of injury or stress, there may be heightened requirements for precise protein levels to maintain normal functions of a regulatory network. Stress is also likely to alter multiple nodes within a regulatory network such that compensatory interactions are diminished. Another explanation for the lack of developmental phenotypes in miRNA knockout mice is that most mRNAs are targeted by multiple unrelated miRNAs such that many miRNAs need to be disrupted before the depression of a specific mRNA has perceptible consequences.

Although miRNA gene deletions in mice have relatively minor consequences in the absence of stress and only in the case of miR-1 has a 50% reduction of a cardiac miRNA been shown to cause lethality (Zhao et al., 2007), haploinsufficiency in humans is the basis for numerous diseases, particularly in the heart, underscoring the potential impact of relatively subtle changes in protein dosage on heart development and function (Garg, 2006; Mori and Bruneau, 2004; Yamagishi, 2002). miRNAs may therefore exert stronger effects in humans than in animal models. In this regard, a hallmark of congenital heart disease in humans is variable penetrancy and expressivity, pointing to the existence of modifiers of cardiac phenotypes. Whether mutations in miRNAs or their target sites may cause heart disease in humans by either disrupting or creating miRNA:mRNA interactions remains to be determined. It is intriguing that a single nucleotide polymorphism in the 3′ UTR of the mRNA encoding myostatin, a member of the TGF-β family and negative regulator of muscle growth, has been shown to cause dramatic muscle hypertrophy in the Texel strain of sheep (Clop et al., 2006). This mutation creates a target sequence for miR-1, causing repression of myostatin expression and unrestrained muscle growth as seen in mice with myostatin loss-of-function mutations (Clop et al., 2006). Although human polymorphisms that influence miRNA target sites and contribute to disease have also been described, genetic and functional evidence for the causality of these polymorphisms in disease is lacking (Sethupathy and Collins, 2008).

The VSDs and ASDs in mice lacking miR-1, miR-133a, and miR-17~92 suggest the potential involvement of these and other miRNAs in congenital heart disease. It is intriguing in this regard that DiGeorge syndrome, which associated with cardiac abnormalities, is caused by a deletion of a region of chromosome 22 that encodes the DiGeorge syndrome critical region 8 (DGCR8), which encodes a component of the RISC complex (Gregory et al., 2004; Han et al., 2004). The fact that this syndrome results from haploinsufficiency of this locus raises the possibility that perturbation of miRNA expression could contribute to the gene dosage sensitivity of this disease by impacting numerous miRNA targets.

The seemingly minimal effects of miRNAs on unstressed adult tissues and their selective contributions to remodeling responses of diseased tissues make miRNAs attractive targets for therapeutic manipulation since strategies to inactivate disease-inducing miRNAs might have modest off-target effects on normal tissues. Therapeutic manipulation of miRNA levels in the heart represents a promising approach for the modification of disease phenotypes. In this regard, modulators of miRNAs, such as antagomiRs and miR mimics, have shown promising effects on a number of cardiovascular disorders in animal models (Bonauer et al., 2009; Care et al., 2007; Thum et al., 2008). Moreover, the sophisticated approaches for catheter-based delivery to the heart should facilitate the introduction of miRNA-based therapeutics directly to sites of cardiac injury, bypassing possible off-target effects on other tissues (van Rooij et al., 2008b). It is also likely that miRNAs will have diagnostic or prognostic applications for heart disease.

Signature patterns of miRNAs have been identified in a variety of cardiac disorders (e.g. hypertrophy, heart failure, ischemic cardiomyopathy, post-MI remodeling) (Ikeda et al., 2007; Sayed et al., 2007; Tatsuguchi et al., 2007; Thum et al., 2007; van Rooij et al., 2006; van Rooij et al., 2008c) and recent studies have reported miRNAs in the blood and the elevation of specific miRNAs following myocardial infarction (Laterza et al., 2009).

Concluding remarks

Despite current insights into the roles of miRNAs in heart development and disease, our understanding of how miRNAs function is far from complete and numerous conceptual and experimental questions remain. To date, only a handful of the hundreds of miRNAs expressed in the cardiovascular system has been functionally analyzed. Identification of additional miRNAs and analysis of the functions of their targets promises to provide new and unanticipated insights into mechanisms of cardiovascular development, function and dysfunction. The realization that miRNAs play central roles in modulation of congenital and acquired diseases of the heart and cardiovascular system provides a new perspective on these disorders and has revealed unanticipated cellular mechanisms of disease and potential new therapeutic targets. Finally, the functions of miRNAs in heart development and disease are undoubtedly representative of the roles of these regulatory RNAs in other tissues. Thus, we anticipate further lessons from the heart.