MicroRNA: master controllers of intracellular signaling pathways

Abstract

Signaling pathways are essential intracellular networks that coordinate molecular outcomes to external stimuli. Tight regulation of these pathways is essential to ensure an appropriate response. MicroRNA (miRNA) is a class of small, non-coding RNA that regulates gene expression at a post-transcriptional level by binding to the complementary sequence on target mRNA, thus limiting protein translation. Intracellular pathways are controlled by protein regulators, such as suppressor of cytokine signaling and A20. Until recently, expression of these classical protein regulators was thought to be controlled solely by transcriptional induction and proteasomal degradation; however, there is a growing body of evidence describing their regulation by miRNA. This new information has transformed our understanding of cell signaling by adding a previously unknown layer of regulatory control. This review outlines the miRNA regulation of these classical protein regulators and describes their broad effects at both cellular and disease levels. We review the regulation of three important signaling pathways, including the JAK/STAT, NF-κB, and TGF-β pathways, and summarize an extensive catalog of their regulating miRNAs. This information highlights the importance of the miRNA regulon and reveals a previously unknown regulatory landscape that must be included in the identification and development of novel therapeutic targets for clinical disorders.

Introduction

MicroRNA (miRNA) is a class of small, non-coding RNA of 19–25 nucleotides (nt) in length that regulate gene expression at a post-transcriptional level by binding to the 3′ UTR of the target transcript [1]. They are transcribed in the form of a primary transcript (pri-miRNA), either under the control of their own promoter regions or processed from a coding gene [2]. pri-miRNA is subsequently processed in the nucleus by RNase III-type endonuclease, Drosha, in association with an accessory double-stranded RNA (dsRNA)-binding protein, DiGeorge Critical Region 8 (DGCR8), into a stem-loop dsRNA pre-miRNA (or precursor miRNA), of 60–70 nt in length, with a 2 nt overhang at the 3′ end [3, 4]. This stem-loop dsRNA structure is transported from the nucleus to the cytoplasm by Exportin-5 [5, 6], where it is further processed by another RNase III-type endonuclease, Dicer, and, in some cases, also with the help of accessory dsRNA-binding proteins, PACT and TRBP, into a mature dsRNA duplex of 19–21 nt base pairs each side and a 2 nt overhang at each 3′ end [7, 8]. This duplex is then loaded onto an effector complex called RNA-induced silencing complex (RISC). One strand from the duplex acts as a guide to direct RISC in binding the target mRNA complementary sequence that mediates gene silencing [9] (Fig. 1). The discovery of microRNA in 1993 [10] revealed previously unknown layer of post-transcriptional control that revolutionized our concept of gene regulation. “Classical” protein regulators, such as suppressor of cytokine signaling (SOCS) and A20, are well known to quickly control signaling pathways through direct post-translational modification, such as phosphorylation or ubiquitination, of their target protein. miRNAs do not regulate the activity of existing proteins, but rather limit the synthesis of new proteins, providing an extra layer of control that is now accepted as being an essential component of pathway regulation in processes such as cell development or differentiation [11, 12]. However, as well as regulating key players within signaling pathways, miRNAs are increasingly being documented to regulate the “classical” regulators, thus providing additional control which we review in this manuscript. Here, we describe this novel mechanism of molecular regulation of three major signaling pathways: Janus kinase/signal transducer and activator of transcription (JAK/STAT), nuclear factor kappa-light-chain-enhancer of activated B cells (NF-κB) and transforming growth factor beta (TGFβ). We also outline how each miRNA plays a role in different cellular and clinical contexts. In fact, this review highlights that regulators of the cellular signaling pathways are important targets of regulation by miRNAs, and are significant targets for future research.

Fig. 1.

Biogenesis of microRNA. In the nucleus, miRNA is either transcribed from its own promoter in the intergenic region or processed from the intronic region of a coding gene as a primary transcript (pri-miRNA). It is processed by a RNase III-type endonuclease family protein Drosha, with an accessory dsRNA-binding protein DGCR8, into a precursor molecule (pre-miRNA) with stem-loop structure of around 60–70 nt in length and a 2 nt overhang at the 3′ end. It is then exported to the cytoplasm by a transport protein Exportin-5. In the cytoplasm, pre-miRNA is further processed by another RNase III-type endonuclease family protein Dicer, in some cases also with the help of accessory dsRNA-binding proteins PACT and TRBP, into a dsRNA duplex of 19–21 nt base pair region and a 2 nt overhang at each 3′ end. One of the two strands (guide strand) from this miRNA duplex is loaded onto RISC complex to effect its gene silencing function

JAK-STAT signaling pathway

The JAK-STAT signaling pathway is mainly adopted by cytokine receptors to effect their anti-viral, inflammatory, and cell-proliferative activity [13]. Upon binding of extracellular ligands, such as interferon (IFN)-α/β and interleukin (IL)-6, to their respective dimerized transmembrane receptors, the pre-associated JAK tyrosine kinases are brought into juxtaposition and activated by trans-autophosphorylation. Activated JAK kinases then mediate tyrosine phosphorylation on the conserved residue of the receptors, which causes the receptor recruitment of SH2-domain containing STAT proteins, which are in turn phosphorylated by JAKs, dissociate from the receptors, dimerize and translocate into the nucleus, where they act as transcriptional activators, driving expression of effector genes [14]. This pathway is central to the well being of our complex immune system, with dysregulation leading to serious lymphoproliferative and autoimmune diseases [13], and is therefore under tight regulation at multiple levels (reviewed in [15]).

Regulation of SOCS by microRNAs

The best studied regulators of the JAK-STAT pathway are SOCS proteins, which constitute a family of 8 members, including SOCS1–7 and cytokine-inducible Src homology 2 protein (CIS) [16]. SOCS proteins bind phosphorylated tyrosines of JAKs and/or the receptor via their SH2-domains, thus blocking STAT recruitment [14]. Additionally, the SOCS box domain recruits elongin B and C-containing ubiquitin E3 ligase complexes and effectively mediates receptor degradation through the proteasome [14, 15]. Basal expression of SOCS is low, but can be up-regulated by cytokine stimulation, providing an essential and effective negative feedback loop for the activated pathway [16]. Recent publications have documented that miRNA regulation of SOCS expression is also key to optimal performance of the JAK/STAT pathway. The role of miR-155 in regulating SOCS1 protein expression has been implicated across a spectrum of cell types. It was first described by Rudensky’s group in the context of Foxp3 expression of regulatory T (Treg) cell homeostasis. The authors noticed that an up-regulation of SOCS1 protein level was detected in miR-155-deficient Treg cells, a phenotype which could be reverted by the re-introduction of miR-155 [17]. Not limited to T cell biology, the regulatory role of miR-155 on SOCS1 has also been implicated in NK cell development and functions. In this study, NK cells from miR-155 knockout mice had elevated SOCS1 expression, and suffered from both impaired NK cell generation and response to viral infection [18]. The functional consequences of SOCS1 regulation by miR-155 are clearly illustrated by early work from Cao’s group, showing that even though miR-155 did not alter IFN expression in virally infected macrophages, its suppression of SOCS1 levels increased STAT1 phosphorylation and downstream IFN stimulated gene (ISG) induction [19]. Interestingly, the miR-155-SOCS1 relationship has also been actively implicated in the field of cancer biology. An inverse correlation of miR-155 and SOCS1 expression was observed in breast cancer patients and cell lines, in which miR-155 conferred enhanced oncogenic properties [20]. In hepatocellular carcinoma (HCC), miR-155 regulation of SOCS1 increases STAT3 signaling, in turn stimulating matrix metalloproteinase (MMP)9 expression and increasing tumor invasion [21]. Other miRNAs that have been shown to regulate SOCS1 expression include miR-30d in prostate cancer [22], miR-122 in Huh-7 hepatocyte cells [23], and miR-150 in lupus nephritis pathogenesis model [24]. The miR-19a/b family was up-regulated in multiple myeloma (MM) and acted as an oncogenic regulator via suppression of SOCS1, an important inhibitor of IL-6-mediated growth in MM pathogenesis [25]. The miR-19a-SOCS1 relation has also been implicated in gastric cancer [26]. To our surprise, when we analyzed the predicted targets of miR-19a within the JAK-STAT pathway by bioinformatics, we discovered that the miR-19a target sequence is conserved between SOCS1 and SOCS3, and found that SOCS3 is a putative target of miR-19a and modulates the activity of JAK-STAT signaling in response to IFN-α and IL-6 [27]. Together these results demonstrate a realization of how evolution has “chosen” a single miRNA species to regulate multiple cellular targets that converge onto the same signaling pathway to exert an amplified combinatory effect on the pathway [12]. This broad effect of miR-19a provides the cell with a convenient switch to control the expression of a set of genes with powerful effect. While miR-19a regulates multiple targets of the JAK-STAT pathway, its regulation also extends to the NF-κB signaling pathway [28]. Similarly, miR-155 inhibits SOCS1 and SOCS3 expression, which enhances IFN production during persistent viral infection, demonstrating its ability to also control several SOCSs that regulate more than one pathway [29].

Targeting of SOCS by multiple miRNA also seems to be a common strategy adopted by miRNA to regulate the JAK-STAT pathway. miR-203 was demonstrated by two independent groups to regulate SOCS3 expression in different cellular contexts. Ru and colleagues reported that miR-203 was up-regulated in breast cancer and its knock-down correlated with enhanced level of SOCS3 expression and improved chemosensitivity toward cisplatin [30]. Moffatt and Lamont demonstrated that gingival epithelial cells infected with Porphyromonas gingivalis had increased cellular miR-203 expression, resulting in SOCS3 down-regulation and enhanced STAT3 activation [31]. On top of this, the latter showed that SOCS6 is also a putative target of miR-203 [31]. In another example, miR-221 was implicated in regulating SOCS3 expression level, which conferred anti-tumorigenic effects in prostate cancer patients [32]. TargetScan predicts that SOCS1 and SOCS7 are additional targets for this miRNA [33], highlighting that our current knowledge and understanding of SOCS-targeting microRNAs is in its infancy and that future investigations may reveal an even more complex and intricate network of intracellular pathway regulation.

Regulation of PIAS by microRNAs

While JAK kinase- and receptor-mediated signaling are directly regulated by SOCS, the downstream STAT proteins, are regulated by protein inhibitor of activated STAT (PIAS), which effectively fine-tunes pathway activity. The PIAS family in mammals is composed of 4 members: PIAS1, PIAS3, PIASx, and PIASy, recognized to target STAT1, STAT3, STAT4, and STAT1, respectively [15]. Each member of the PIAS family contains a RING-finger-like zinc-binding domain (RLD), which confers small ubiquitin like modifier (SUMO) E3-ligase activity, thus mediating SUMOylation and consequential deactivation of STATs [34]. PIAS proteins also regulate STATs independently of SUMOylation. Other mechanisms include direct blockage of STAT-DNA binding and recruitment of co-repressors, such as histone deacetylase (HDAC) (reviewed in [34]). Although PIAS-STAT interaction is believed to be cytokine-dependent [34], the recent discovery of their regulation by miRNAs could be crucial in understanding the maintenance of the pathway integrity and cellular homeostasis.

Since PIAS3 negatively regulates STAT3, a key player in the IL-6-mediated ISG induction [15], dysregulation of PIAS3 by miRNA could have devastating outcomes. Regulation of PIAS3 by miRNA was first proposed by Brock and colleagues, who showed that miR-18a targeted the 3′ UTR of PIAS3 mRNA, which suppressed its protein expression, resulting in IL-6-induced STAT3 activation in hepatocytes and the acute-phase response [35]. Since the dysregulation of JAK-STAT signaling, via altered expression of SOCS by miRNA, is evident in many cancer models, it is no surprise to see that the same is true for PIAS-mediated regulation. Indeed, modulation of PIAS3 by miR-18a has been implicated in gastric adenocarcinoma, in which a clinical correlation has been established between miR-18a, PIAS, JAK-STAT pathway activity and downstream anti-apoptotic and cell-proliferative genes [36]. Using proteomics, PIAS3 was also identified as a cellular target of miR-21, which was highly expressed in MM [25], resulting in similar pathological outcomes to IL-6-induced JAK-STAT pathway activation [37]. PIAS3 is also a target for miR-125b in glioblastoma stem cells [38], further highlighting the multifaceted nature of its regulation and importance as a gatekeeper of oncogenesis. Interestingly, microRNA regulation of PIAS3 even controls T cell development. In fact, inhibition of miR-301a expression in myelin auto-antigen exposed CD4⁺ T helper cells altered their cytokine expression profile and hampered their differentiation into Th17 cells, which was thought to be controlled by PIAS3 regulation of IL-6-STAT3 signaling [39].

Although there are four mammal PIAS proteins, evidence for their expression being regulated by miRNA has only been reported for PIAS3. This exclusivity may be partly explained by the length of the 3′ UTR of their mRNA, since only human PIAS3 mRNA carries a 3′ UTR that spans nearly 1000 nt long, while mRNA from the other PIAs family members are just a few hundreds. This speculation is supported by the study that some housekeeping genes, which have strong preference to minimize miRNA regulation, tend to evolve with a shorter 3′ UTR, thus avoiding miRNA binding, which consequently minimizes the risk of their accidental and undesirable shutdown [40]. However, the physiological relevance of the PIAS’s 3′ UTR length in regulating the JAK-STAT pathway warrants further investigation. In summary, it is clear that the integrity of a functional JAK-STAT pathway is essential for cellular homeostasis. Dysregulation of this signaling by miRNA may attribute to many cancers and autoimmune diseases. However, it is important to remember that there are other signaling pathways, such as the NF-κB and TGFβ cascades, which also determine the outcome of effective cellular reactions and are now known to be under the regulation of miRNAs.

NF-κB signaling pathway

NF-κB mediates diverse biological processes at the cellular level, including growth, development, and inflammatory responses [41]. The canonical NF-κB pathway mainly utilizes the RelA (also known as p65):p50 heterodimer as a transcription factor to activate downstream target genes. In unstimulated cells, the RelA:p50 heterodimer is bound to inhibitor of NF-κB (IκB), making it inaccessible to the nucleus and thus blocking gene transcription [42]. To remove this suppressive constraint, IκB protein must be phosphorylated by IκB kinase (IKK) complex, which constitutes two catalytic subunits, IKKα and IKKβ, and one regulatory subunit, NF-κB essential modulator (NEMO) (also known as IKKγ). This phosphorylation initiates K48-polyubiquitination and subsequent proteasomal degradation of IκB protein [41]. In the non-canonical (or alternative) NF-κB pathway, which utilizes the RelB:p52 heterodimer, p100, the p52 predecessor, acts like IκB to suppress translocation and transcription activation of RelB when bound under unstimulated conditions [42]. Upon activation, NF-κB-inducing kinase (NIK), with the help of IKKα, induces phosphorylation of p100, which is then subjected to ubiquitination and processing into p52, which along with RelB, serves as a heterodimeric transcription factor (reviewed in [43]) (Fig. 2). Dysregulation of the NF-κB pathway accounts for many autoimmune, chronic inflammatory and cancerous diseases [41], therefore, as with the JAK/STAT pathway, multiple levels of regulation must be adopted to avoid disease (reviewed in [41]).

Fig. 2.

Regulation of key regulators of cellular pathways by miRNAs. Left JAK-STAT pathway. Binding of cytokine to its cognate receptor pair activates and phosphorylates receptor-associated JAK kinase, which in turn phosphorylate receptor residues, providing docking sites for the downstream transcription factors, STATs, which are phosphorylated by JAKs. STATs then become activated and dissociate from the receptor to dimerize and translocate to the nucleus where they promote transcription of target genes from promoter carrying interferon-stimulated responsive element (ISRE) or IFN-γ activated sequence (GAS). SOCS proteins negatively regulate JAK kinase by blocking the binding with STATs and promoting the degradation of cytokine receptors; and PIAS proteins negatively regulate STAT transcription factors by blocking its binding to DNA and recruiting co-repressors to inhibit transcription. Both SOCS and PIAS are under tight regulation by miRNAs. Centre NF-κB pathway. Both canonical and non-canonical NF-κB pathways are activated through similar mechanisms. Under unstimulated conditions, the activity of NF-κB is suppressed by an inhibitory signal (canonical: inhibitor of κB (IκB); non-canonical: ankyrin repeats on p100). To remove this inhibitory constrain, the inhibitory signal is labeled by a K48-polyubiquitin chain and targeted for  proteasomal processing. Prior to K48-polyubiquitination by E3 ligase, the target protein must be phosphorylated by an NF-κB-activating kinase, also known as mitogen-activating protein kinase (MAPK) [for example, canonical: IκB kinase (IKK); non-canonical: NF-κB-inducing kinase (NIK)]. In order to become activated and mediate the downstream phosphorylation event, MAPK needs to be phosphorylated by an upstream kinase, also known as MAPK kinase (MAPKK) [for example TGFβ activated kinase-1 (TAK1)]. This activating phosphorylation can be removed by protein phosphatase PP2A/C. Like MAPK, activation of MAPKK requires the phosphorylation of another upstream kinase, also known as MAPKK kinase (MAP3K). MAP3K can be activated by K63-polyubiquitination in response to external stimuli. In terms of tumor necrosis factor (TNF) stimulation, the MAP3K protein, receptor-interacting protein (RIP)1 kinase, can be deactivated by deubiquitinase A20, as well as other A20-related proteins, including its binding partner Ring Finger Protein 11 (RNF11) and CYLD, by removing its K63-polyubiquitin chain and recruiting E3 ligase to tag a K48-polyubiquitin chain to promote its degradation. All these regulators, A20, RNF11, CYLD, and PP2A/C, can be regulated by miRNAs. Right TGFβ pathway. Binding of TGFβ to its cognate receptor phosphorylates and activates receptor Smad (R-Smad) proteins, such as Smad2 and 3. R-Smad proteins bind other coactivators and translocate into the nucleus to drive transcription of target genes with the Smad-binding element (SBE) in their promoter. Activation of R-Smad proteins can be inhibited by inhibitory Smad (I-Smad) such as Smad7 through direct blockage of receptor and recruitment of other deactivating enzymes such as Smad ubiquitin regulatory factor (Smurf) E3 ligase. TGFβ is normally secreted in a closed immature form into the extracellular matrix. Glycoprotein A repetitions predominant protein (GARP) expressed by T regulatory (Treg) cells can facilitate the maturation of TGFβ, thus acting as a positive regulator of the pathway. I-Smad, Smurf and GARP can be regulated by miRNAs. miRNAs that can regulate multiple cellular targets from the same signaling pathway are underlined

Regulation of PP2A/C by microRNAs

Although the canonical and non-canonical pathways mobilize different cellular modulators, they are regulated using a similar mode of action: phosphorylation, ubiquitination, and proteasomal processing of the inhibitory binding partner of NF-κB. Due to the importance of NF-κB regulation, it is not surprising to see that some of these steps are also controlled by miRNA. In both pathways, IKKα/β is the major protein kinase, engaging in the initial phosphorylation and subsequent processing or degradation of the inhibitory binding partner [42]. In order to become activated, IKKα/β complex requires trans-autophosphorylation and phosphorylation from another upstream kinase, such as TGF-β activated kinase-1 (TAK1) or NIK. These phosphorylation sites are subjected to dephosphorylation by a group of protein phosphatases called PP2A/C [44]. Regulation of PP2A/C by miRNA is evident in cancer models. Two papers recently reported that miR-520h targets PP2A/C and promotes NF-κB-driven tumorigenic gene expression in breast cancer and ovarian cancer cell lines, as well as in lung cancer patient samples [45, 46]. The significance of regulating PP2A/C is evident in the broad spectrum of miRNAs that control its expression, including miR-1, miR-19, miR-31, and miR-133. While these miRNAs have not been shown to impact the NF-κB pathway in disease models, they will most likely affect responses to the pathway and thus identify an area of research that warrants investigation [47–49].

Regulation of CYLD by microRNAs

Ubiquitination of target proteins is arguably one of the most important and influential molecular events within a cell and is thereby controlled by a series of processes. Polyubiquitination does not only enable IκB degradation or p100 processing, but is also implicated throughout the NF-κB pathway (reviewed in [50]). In the canonical pathway, NEMO undergoes K63-polyubiquitination in response to TNF stimulation, which facilitates the recruitment of upstream activating factors and, in turn, the activation of IKK complex [51]. To regulate IKK activation, the tumor suppressor deubiquitinase, CYLD (cylindromatosis), removes polyubiquitin chains from NEMO [42]. The regulation of CYLD is now known to involve miRNAs, which have been shown to play an important role in cancer pathogenesis. For example, miR-181b-1 was found to be up-regulated in an oncogenic Src kinase transformed model and manipulation of cellular miR-181b-1 levels altered CYLD expression, NF-κB activity and mammary epithelial cell line transformation [52]. The targeting of CYLD by miR-181b-1 was later implicated in pancreatic cancer, in which increasing miR-181b-1 levels confer cell line chemoresistance to gemcitabine, via the down-regulation of CYLD and up-regulation of NF-κB activity [53]. In gastric cancer patients, miR-362 was up-regulated in tumor tissue samples, which inversely correlated with CYLD expression, suggesting that miR-362 regulation of CYLD promoted NF-κB activity and subsequently enhanced cell proliferation and apoptotic resistance [54]. In addition, CYLD mRNA has been shown to be directly targeted by miR-182 and miR-486, which promoted tumor aggressiveness of gliomas, again through NF-κB dysregulation [55, 56]. The broad inhibitory remit of both these miRNAs is clearly demonstrated in their spectrum of targets, with miR-486 also regulating Cezanne (A20 family deubiquitinase) and A20-interacting partners, TNF-α-induced protein 3 (TNFAIP3) interacting protein (TNIP)1/2/3; and miR-182 regulating TNIP1, optineurin ubiquitin-binding protein (OPTN), and the deubiquitinase ubiquitin-specific protease 15 (USP15) [55, 56].

Regulation of A20 by microRNAs

Another deubiquitination enzyme that has received much attention is A20 (also known as TNFAIP3). A20 contains an N-terminal ovarian tumor (OTU) deubiquitination domain and a C-terminal zinc finger (ZnF) E3 ligase domain, which are thought to have dual functions in K63-linked deubiquitination and K48-linked polyubiquitination of substrates, such as receptor-interacting protein (RIP)1 kinase [57]. RIP1 is an upstream activating kinase of TAK1, and its K63-linked polyubiquitination is indispensible for IKK activation in the TNF-induced NF-κB pathway [41]. Indeed, A20 mediates deubiquitination of K63-linked polyubiquitin chain on RIP1, but K48-linked polyubiquitination is actually mediated by A20-binding partner, ITCH (also known as itchy E3 ubiquitin protein ligase), which targets RIP1 for proteasomal degradation, thus terminating the transduced signal [42] (Fig. 2). With its ability to negatively regulate NF-κB activity, A20 is regarded as a tumor suppressor and its inactivation is frequently observed in various cancer models [57]. miRNA has also been documented to manipulate A20 expression levels with obvious consequences for NF-κB activity in tumor cells.

Gantier and colleagues reported that global depletion of miRNA expression, through conditional Dicer knock-out, impaired pro-inflammatory cytokine induction [28]. Initially using TargetScan, they predicted that negative regulators of NF-κB, including A20 and other related proteins, such as its binding partners (Ring Finger Protein 11 (RNF11) and ITCH), A20 regulator (TNIP1) and other deubiquitinases (CYLD and Cezanne), were targets of an oncogenic miRNA cluster, miR-17-92 [28, 33]. Among the four miRNA families expressed in this cluster, miR-19 was demonstrated to have a significant impact on NF-κB activity [28]. In this study, A20 and RNF11, as well as two other regulators of NF-κB (KDM2A and ZBTB16), were shown to be validated targets of miR-19, whereas the suppressive effect on other predicted targets, including CYLD, was not observed [28]. However, in T cell acute lymphoblastic leukemia (T-ALL) patient samples and cell lines, up-regulation of miR-19 inhibited CYLD expression, leading to sustained NF-κB activity [58], clearly demonstrating that miRNAs regulate multiple targets from the same pathway in a cell- and disease-type dependent fashion and highlighting the vast chasm of knowledge still to be explored. In addition, apart from miR-19, miR-146a was also able to regulate the expression of RNF11, which facilitated Hendra virus replication in a NF-κB-dependent manner [59]. miR-18a, from the miR-17-92 cluster, also reduced A20 expression in a model of rheumatoid arthritis (RA) and enhanced NF-κB-dependent expression of the matrix degrading enzyme, MMP1, and inflammatory cytokines, such as IL-6, in synovial fibroblasts [60]. Similar establishment of NF-κB dysregulation by miRNA-targeting of A20 was observed during Japanese encephalitis virus infection. This virus induced cellular miR-29b expression, which regulated A20 expression in a microglial cell line, thus enhancing NF-κB activity [61]. In stark contrast, another miRNA from the same miR-29 family, miR-29c, was found to be down-regulated in a Hepatitis B Virus (HBV)-related HCC cell line and patient samples. This loss of miR-29 expression up-regulated A20 levels, resulting in restricted cell proliferation and enhanced apoptosis [62]. While miRNA is thought to predominately suppress target gene expression, in a sarcoma model, miR-29 bound the A20 3′ UTR and prevented physical association of an RNA-binding protein HuR, thus protecting A20 mRNA from de-stabilization and degradation. In the same study, the authors showed that miR-125 could also regulate A20 expression [63], an observation mirrored in a macrophage polarization and diffuse large B-cell lymphoma model [64, 65].

In general, aberrant NF-κB activity resulting from the dysregulation of its regulators by miRNA drives the expression of numerous pro-inflammatory cytokines and chemoattractants at the site of injury, and confers aggressiveness and apoptotic tolerance to tumors at the cellular level. These events have been evident in many of the aforementioned examples and are believed to be at the center of many inflammatory diseases and cancers. Therefore, further elucidating the control of NF-κB regulators by miRNA will help us better understand the development and progression of these diseases and reveal much needed therapeutic targets.

TGFβ signaling pathway

From the beginning of life, the TGFβ pathway is indispensable in coordinating cell development and differentiation and is essential for sustaining a functional immune response [66]. The signal begins when mature TGFβ is freed by an endoprotease from a latent complex held within the extracellular matrix [67]. The binding of TGFβ to its cognate receptor complex, consisting of two type-I receptors and two type-II receptors, triggers serine/threonine kinase activity on the type-II receptor molecule, which subsequently activates and phosphorylates the type-I receptor at its cytoplasmic domain [68]. Smad2 and Smad3 (Receptor (R)-Smad proteins) are recruited to the phosphorylation site and are themselves phosphorylated by the activated type-I receptor complex. Phosphorylated Smad2/3 can then form a trimeric complex with other coactivators, such as Smad4 and TIFIγ, to regulate gene expression in the nucleus [67]. Alternatively, the receptor complex can activate a Smad-independent pathway through modulating Rho, GTPase, MAP kinase and PI3K signaling pathway activity, which regulates a different set of target genes [66] (Fig. 2). To achieve optimized signaling activity, the TGFβ signaling pathway output is tightly regulated at multiple stages.

Regulation of Smad7 by microRNAs

While R-Smad proteins convey activatory downstream signals, inhibitory Smad (I-Smad) proteins regulate this intracellular transduction. Smad7, for example, is expressed in response to TGFβ pathway activation and provides efficient negative feedback through several mechanisms [67]. It can physically bind to the type-I TGFβ receptor, acting as a direct competitor to R-Smad [67], or it can further recruit other regulatory proteins, including PP1 phosphatase and the Smad ubiquitin regulatory factor (Smurf) E3 ligase, which inactivates and promotes degradation of the receptor molecule, respectively [67, 68]. We now know that expression of Smad7 is regulated by multiple miRNAs. The up-regulation of miR-21 expression has been shown to suppress Smad7 expression in HCV-infected liver biopsies. Interestingly, these findings correlated with patient HCV load and fibrosis stage, indicating a role for miR-21 in accelerating fibrogenesis [69]. Others observed that miR-21-mediated reduction of Smad7 correlated with expression of TGFβ-induced fibrotic markers, such as alpha smooth muscle actin (α-SMA) and fibronectin (Fn), which promoted epithelial–mesenchymal transition (EMT) in lung fibrosis [70, 71], renal fibrosis [72, 73], and systemic sclerosis [74, 75]. EMT also plays an important role in the initiation of cancer metastasis. Complementary to this notion, miR-21 was up-regulated in the invasive ductal carcinoma region of breast cancer and knock-down of miR-21 restored Smad7 levels in a breast cancer cell line [76]. miR-21-mediated reduction of Smad7 expression has also been implicated in the generation of carcinoma-associated fibroblasts (CAFs), which confer tumorigenesis, proliferation and invasiveness characteristics of tumors [77]. However, elevated miR-21 levels are not always associated with enhanced proliferation and differentiation. In the case of myelodysplastic syndromes (characterized by ineffective hematopoiesis), suppression of Smad7 by miR-21 was found to decrease erythroid colony formation of CD34⁺ cells, while inhibiting miR-21 could rescue red blood cell count and stimulate erythropoiesis in transgenic mice [78]. As seen with miR-21, suppression of endogenous Smad7 is a common carcinogenic mechanism that promotes EMT. Other examples include the miR-216a/217 cluster in HCC [79], miR-20a in gall bladder carcinoma [80], miR-181a in ovarian cancer [81], and miR-106b-25 in breast cancer [82]. While the oncogenic role of these microRNAs has been described, the tumor suppressive role of a selection of them has also been reported, with reduced miR-25 in colon cancer [83] and attenuated miR-181c in metastatic neuroblastoma patients [84]. Interestingly, both reports reasoned that the tumor suppressive roles of miR-25 and miR-181c were accounted for by reduced Smad7 protein expression and TGFβ activity, which would otherwise stimulate tumor growth and migration [83, 84].

Regulation of Smurf by microRNAs

Reduced Smad7 expression in bronchial epithelial cells was also associated with overexpressed miR-15b in a chronic obstructive pulmonary disease model [85]. In their cell-based assays, the authors also found that, apart from Smad7, Smurf2 expression was affected by miR-15b [85]. Smurf proteins are E3 ubiquitin ligase proteins, recruited by Smad7 to the type-I TGFβ receptor complex. Their recruitment promotes proteasomal degradation of the receptor complex, thus restricting signal transduction [86]. The antagonizing miR-322 and miR-503 action on Smurf2 regulation was shown to inhibit the phosphorylation of Smad2 [87]. As miRNA-suppression of Smad7 was observed in many cancers, it is not surprising to see that in an aggressive breast cancer model, Smurf2 was down-regulated by miR-15, miR-16, and miR-128 [88]. miR-15b also targets Smurf1 during osteoblast differentiation, as a way to activate the expression of a master transcription factor, Runx2 [89]. Smurf1 is also a target of miR-17 and miR-497; both of these studies showed reduced miRNA and up-regulated Smurf1 expression in periodontitis and metastatic ovarian cancer, respectively [90, 91].

Regulation of GARP by microRNAs

TGFβ is secreted and stored in the extracellular matrix inside a large latency complex. The cytokine remains inactive and bound to the latency-associated peptide until positive regulators increase the efficiency of its dissociation from the large latency complex and it is processed into a mature form. Glycoprotein A repetitions predominant protein (GARP) is expressed by T regulatory (Treg) cells and tightly associated with the latency-associated peptide bound to immature TGFβ [67]. GARP is essential for TGFβ activation [92], as it frees immature TGFβ molecules from the latency complex [67]. miR-142-3p regulates GARP expression and thus controls Treg cell proliferation [93]. Specifically, the authors observed decreased expression of miR-142-3p in CD25⁺ CD4 T cells and manipulation of miR-142-3p levels resulted in altered proliferation of these cells [93], which is consistent with the concept of TGFβ-mediated Treg cell proliferation. No matter whether a miRNA is targeting the positive or negative regulators of the TGFβ signaling pathway, any upset in the homeostatic balance could lead to serious pathological consequences, like fibrosis or oncogenesis.

Regulation of microRNA levels

So far this review has discussed how miRNAs participate in signaling pathway regulation and provide an additional layer of supervision on pathway regulators, but, in order to fully understand how these “master regulators” control our signaling networks it is important to note how miRNAs themselves are regulated. Endogenous levels of miRNA are significantly linked to the final output of signaling pathways and are under the control of several factors. The majority of miRNAs are located in either intragenic or intergenic regions and are transcribed together with their host gene or from their own promoter [2]. The miR-106b-25 cluster is an example of intragenic miRNA, which sits itself within intron 13 of the miniature chromosome maintenance 7 (MCM7) gene [94]. This miRNA cluster encodes three miRNAs, namely miR-106b, miR-93, and miR-25, two of which, miR-106b and miR-25, we described above as regulators of Smad7. These miRNA were found to be frequently co-expressed and probably co-regulated with their host mRNA [95] and amplification of the MCM7 gene locus and its elevated expression with miR-106b-25 cluster have been associated with human malignancies [94, 96]. On the other hand, the miR-17-92 cluster (also known as oncomir-1) is an example of intergenic miRNA that is expressed and processed from the C13orf25 transcript [96]. This miRNA cluster encodes six mature miRNAs, namely miR-17, miR-18a, miR-19a, miR-20a, miR-19b-1, and miR-92-1, which have broad effects on multiple pathway regulators, including SOCS, PIAS, A20, Smad7, and Smurf1.

It is interesting to note that the expression of both MCM7/miR-106b-25 and miR-17-92 genes are induced by common transcription factors, E2F1 and MYC [96]. E2F1 and MYC are involved in a positive feedback loop, whereby both proteins are capable of regulating one another [97]. The entire regulatory network may be further complicated by the negative regulation of E2F1 expression by miR-106b and miR-20a [96]. These constitute an important regulatory mechanism that allows these miRNAs to be expressed at optimized levels. It is foreseeable that in the event of sub-optimal miRNA levels, the negative feedback constrain is lenient, so the positive feedback loop of the transcription factors encourages the expression of these miRNAs. Alternative, when miRNA expression becomes excessive, it places a heavy negative feedback constrain on the transcription factors, so that the continuous expression of miRNA can be eventually shut off. Therefore, it is evident that the sequence of molecular events responsible for miRNA expression can be more complex than their regulation of target genes and networks, and hence should be analyzed case by case. In fact, when considering the mechanisms that alter the steady-state level of any miRNA, we should also take into account the cell type, its half-life and transient intracellular turnover (reviewed in [98]).

Insights derived from regulator-targeting miRNAs

Although the role played by each miRNA appears to be context-dependent in individual studies, some have collectively demonstrated their versatility in regulating multiple regulators of a signaling pathway. Examples include miR-15 (targeting Smad7, Smurf1, and Smurf2 in the TGFβ pathway) [85, 88, 89], miR-19 (targeting SOCS1 and SOCS3 in the JAK-STAT pathway; A20, CYLD, and RNF11 in the NF-κB pathway) [25–28, 58], miR-155 (targeting SOCS1 and SOCS3 in the JAK-STAT pathway) [29] and miR-203 (targeting SOCS3 and SOCS6 in the JAK-STAT pathway) [30, 31]. This versatility is indeed conferred by the ability of miRNA to regulate gene expression at post-transcriptional level by binding to and targeting the complementary sequence present on 3′ UTR of any genes, irrespective of their actual protein coding sequence, such that multiple genes that share similar gene function and acquire the same complementary sequence can be regulated simultaneously by a common set of miRNAs. The miRNA-target relationship is now thought to be under tight natural selection and is believed to have co-evolved with one another, as well as the whole regulatory network [99].

To date, the vast majority of publications have focused on the validation of a single gene targeted by miRNA that has implications in different biological models. While this has been limited by both our lack of knowledge and experimental capabilities, it has certainly led to an under-estimation of miRNA capacity in modulating the entire regulatory network as a “master regulon”. Fortunately, the recent advancement in next-generation sequencing and other molecular biological techniques, such as photoactivatable-ribonucleoside-enhanced crosslinking and immunoprecipitation (PAR-CLIP), have already improved our understanding and knowledge of the transcriptome-wide regulation of miRNA in many cellular contexts. Nevertheless, bioinformatic database analysis remains a cornerstone for the predictive analysis of miRNAs and their targets and will continue to be used to understand how miRNA can act beyond a single gene to regulate an entire network [100]. Careful data-mining procedures and the use of a systems biology approach will conserve these efforts in validating all of the predicted targets. Additionally, when regarding the timeframe and strength of miRNA regulation, our current knowledge is greatly limited by the use of existing non-physiological methodology that manipulates endogenous miRNA expression in cell-based systems. Furthermore, it is important to note, that in most cases, owing to direct gene amplification of the miRNA region or altered expression of the transcription factors responsible for regulating miRNA expression, miRNAs levels during disease pathogenesis are aberrantly expressed [96].

The plethora of current evidence outlined in this review identifies miRNAs as “master controllers” of intracellular signaling pathways in many disease models and in the era of new therapies against miRNA, this evidence highlights them as powerful targets for therapeutic development, with highly significant clinical applications. In many of the studies covered in this review, manipulation of endogenous miRNA levels by chemically synthesized analogs or inhibitors could revert the phenotype caused by the dysregulated miRNA, and therefore provide the proof-of-principle for potential drug development. While therapeutic development from “bench to bedside” is a long, arduous and expensive process, recent advances with the development of the first miRNA-targeting drug, miravirsen, (miR-122 targeting locked-nucleic acid (LNA)-modified inhibitor for treatment of hepatitis C virus infection, currently in phase 2 clinical trial), have brought the entire miRNA research community closer to therapeutic solutions than ever before [101]. Another miRNA-based drug, MRX34, is the first miRNA mimic for miR-34 and has entered phase 1 trial for treatment of HCC [102]. These outstanding advances demonstrate that miRNA are promising targets for therapeutic intervention and, with our advanced understanding of their regulation of different cellular pathways and disease pathogenesis, it is expected that more pre-clinically validated drugs will enter clinical trials and eventually be used in daily clinical practice.

Conclusion

This article has reviewed how miRNAs potently act as novel regulators of classical inhibitors of the JAK-STAT, NF-κB, and TGFβ signaling pathways. We now have a much deeper understanding of the way in which miRNAs regulate many pathological diseases and normal developmental processes. More importantly, we have identified a reiterating concept, whereby miRNAs bind 3′ UTRs of their target irrespective of the protein coding sequence, and regulate multiple targets, which usually work at different levels of the signaling cascade, within the same signaling pathway. This allows miRNA to provide another layer of signaling regulation, in order to achieve maximal effect and avoid detrimental responses to stimuli. With this concept in mind, it is essential for future research of miRNA-target identification to consider the regulation network (or regulon) of specific miRNA, in order to achieve total understanding of the mechanism of any cellular process. More advanced techniques that can reveal the transcriptome-wide regulation of miRNAs should be considered a standard and essential approach. This not only takes the concept of a single miRNA regulating different targets within the same pathway into account, but also provides a bigger picture of how miRNAs can collectively regulate several targets among different signaling pathways.