Post-transcriptional control of miRNA biogenesis

1MicroRNAs (miRNAs) are small noncoding RNAs that negatively regulate the expression of a large proportion of cellular mRNAs. They have unique, diverse expression patterns and affect many cellular processes and developmental pathways. Most miRNA genes are transcribed by RNA polymerase II (Pol II), with the long primary transcript, termed pri-miRNA, harboring a hairpin structure, which comprises the miRNA sequence. Whereas many of these genes are transcribed as intronic clusters within protein-coding pre-mRNAs, others can be transcribed as independent gene units, or be encoded within long noncoding RNAs (lncRNAs).

2The biogenesis of miRNAs is carried out by two RNase III enzymes, DROSHA and DICER, which catalyze two subsequent processing events, in the nucleus and in the cytoplasm, respectively. The nuclear event is catalyzed by the Microprocessor complex, which comprises the RNase III type enzyme DROSHA, the double-stranded RNA-binding protein (RBP) DGCR8 (DiGeorge syndrome critical region 8 gene) and associated proteins. This nuclear processing event results in the production of ∼70 nucleotide (nt) stem–loop precursor miRNAs, termed pre-miRNAs, which are subsequently exported to the cytoplasm using the export receptor, Exportin-5. Once in the cytoplasm, pre-miRNAs undergo a final processing event, by another RNase type III enzyme, DICER, to give rise to miRNA duplexes. These are then incorporated into the RISC (RNA-induced silencing complex) together with an Argonaute (AGO) protein, where one strand is selected to become the mature miRNA. In addition, there are also noncanonical miRNA biogenesis pathways that lead to the production of functional miRNAs. These include mirtrons that are generated via pre-mRNA splicing and miRNAs generated from small nucleolar RNA (snoRNAs) precursors (for review, see ). Regulation of gene expression by miRNAs is also prevalent in plants; however, several aspects of their biogenesis and function differ (for comprehensive reviews, see;; ). Several excellent recent reviews have focused on the function of animal miRNAs. Here, we will focus on post-transcriptional mechanisms that regulate miRNA production in animals, with a particular focus on the role of RBPs in the post-transcriptional regulation of their biogenesis.

3The nuclear phase of miRNA processing occurs cotranscriptionally acting on both independently transcribed and intron-encoded miRNA. This cotranscriptional processing can be facilitated by HP1BP3, a histone H1-like chromatin protein, which interacts with both the Microprocessor and endogenous pri-miRNAs to promote cotranscriptional miRNA biogenesis in human cells. MiRNA precursors form RNA hairpins that need to be recognized by the Microprocessor. To distinguish primary miRNAs (pri-miRNAs) from other hairpin-containing transcripts, additional identifiers are required. These determinants comprise a ∼35 bp stem harboring a mismatched GHG motif and also include three primary-sequence elements, a basal UG motif, an apical UGUG motif and a CNNC motif, which binds the SR protein, SRSF3, and is found downstream from approximately 60% of all pri-miRNA hairpins (Fig. 1A;; ). Another proposed identifier of what constitutes a bona fide pri-miRNA stem–loop is the presence of an N6-methyladenosine (m6A) mark in the vicinity of the pri-miRNA stem–loop. This mark is bound by a reader, the hnRNP protein, A2/B1, which interacts with DGCR8 and stimulates miRNA processing.

4The precise mechanism by which the Microprocessor recognizes pri-miRNAs and catalyzes their processing is beginning to be fully understood. In brief, the Microprocessor is a heterotrimeric complex, comprising one DROSHA and two DGCR8 molecules. The DGCR8 dimer interacts with the stem and apical elements of the pri-miRNAs through its double-stranded RNA-binding domains (dsRNA) and RNA-binding heme domain, respectively, leading to accurate and efficient processing. In contrast, DROSHA serves as a ruler to measure an 11 base pair (bp) distance from the basal single-stranded RNA-double stranded RNA (ssRNA-dsRNA) junction and cleaves the stem–loop of primary miRNAs. Crucially, the orientation of this complex on the substrate is maintained by DROSHA and DGCR8 recognizing the basal UG and apical UGUG motifs, respectively. It has been recently shown that SRSF3, a member of the SR protein family of splicing factors promotes miRNA processing by recruiting DROSHA to the basal junction in a CNNC-dependent manner. The activity of the Microprocessor can be enhanced by the binding of heme, a ferric ion-containing porphyrin, which promotes the interaction between the DGCR8 dimer and the apical UGUG motif, promoting Microprocessor activity.

5Besides its established role in miRNA biogenesis, noncanonical functions for the Microprocessor have also been suggested (for reviews, see; ). The first hint of more extended roles for the Microprocessor came from the observation that DROSHA cleaves pri-miRNA-like hairpins harbored within the 5′-UTR of the mRNA encoding the DGCR8 protein itself, providing a feedback loop to control DGCR8 levels. Furthermore, the phenotypic differences observed during early T-cell development in the mouse, following Dgcr8/Drosha and Dicer inactivation, were correlated to transcriptomic changes that were unique to Drosha but not Dicer highlighting the existence of DROSHA-dependent, DICER-independent processing of RNAs. Thus, these noncanonical activities of the Microprocessor could affect cellular RNAs, beyond the described autoregulatory feedback that controls levels of DGCR8 pre-mRNA. Identification of endogenous targets for DGCR8, revealed that the Microprocessor complex binds and regulates a large variety of cellular RNAs, other than miRNAs, including mRNAs, noncoding RNAs and transcripts derived from several human active retrotransposons (LINE-1, Alu). Accumulating evidence suggests that these noncanonical activities of the Microprocessor do indeed have physiological relevance in the turnover of cellular RNAs. For instance, DROSHA has been show to negatively regulate the expression of the transcription factor Neurogenin by cleaving evolutionarily conserved hairpins present in the Neurogenin mRNA that are similar to pri-miRNAs. Furthermore, miRNA-independent functions of DGCR8 were also shown to be essential for neocortical development in the mouse. This was attributed to the action of the Microprocessor directly regulating the cortical transcription factor, Tbr1, which also contains evolutionarily conserved hairpins that resemble miRNA precursors. Finally, DGCR8 can also associate with other nucleases, suggesting the existence of alternative DGCR8 complexes that may regulate the fate of a subset of cellular RNAs, as shown by the DGCR8-mediated cleavage of small nucleolar RNAs (snoRNAs), which functions independently of DROSHA.

6In the canonical pathway, pre-miRNAs are exported to the cytoplasm and assembled into a complex containing DICER and the Hsp90 chaperone. Subsequently, the pre-miRNA is cleaved by DICER, in tandem with TRBP (HIV-1 TAR RNA RBP) and PACT (protein activator of PKR) (Fig. 1A; ). In this reaction, DICER serves as a molecular ruler that measures the distance from the pre-miRNA basal end to the cleavage site adjacent to the TL (also known as apical loop). This cleavage liberates the pre-miRNA TL element and creates an RNA duplex that interacts with the Argonaute 2 protein (AGO2). Of note, due to variable structural features of pre-miRNAs, the DICER-depended cleavage is often imprecise, generating two or more miRNA duplex variants that will give rise to distinct mature miRNAs. The miRNA duplex is incorporated into an AGO2 protein to form the RNA-induced silencing complex (RISC), in an ATP-dependent manner with the assistance of HSC70/HSP90 chaperones. Subsequently, AGO2 unwinds the RNA duplex and evicts the passenger strand forming the mature RISC complex. The activated RISC then recognizes a specific mRNA sequence by complementary base-pairing resulting in translation inhibition and/or RNA degradation (for reviews, see; ).

7Due to the important role of miRNAs in the control of gene expression and organism development, the production of mature miRNAs is tightly regulated at multiple levels, including transcriptional and post-transcriptional steps. A variety of post-transcriptional mechanisms, which affect DROSHA and DICER processing, as well as miRNA modification and turnover have been previously described. Dysregulation of miRNA production can result in global defects in gene expression and lead to human disease. As an example, impaired miRNA processing promotes cellular transformation and tumorigenesis, and a global miRNA depletion is frequently observed in human cancers.

8Increasing evidence suggests that general RBPs, including splicing factors and other diverse RNA processing factors, act as post-transcriptional regulators of miRNA processing (for review, see ). In such a scenario, the binding of an RBP to the TL or a stem of a miRNA progenitor can positively or negatively affect the Microprocessor-mediated processing of pri-miRNA in the nucleus, and/or the DICER-mediated processing of a pre-miRNA in the cytoplasm (Fig. 1B,C). Below, we will discuss the role of trans-acting factors that bind to precursor miRNAs and influence their nuclear and cytoplasmic processing by the Microprocessor and DICER, respectively. We will also focus on the contribution of sequence variation, exemplified by single nucleotide polymorphisms (SNPs) present in the human genome that can have a role in the biogenesis of miRNAs. Finally, we will present attempts to target miRNA regulatory events using biological and synthetic compounds that could eventually lead toward the development of therapies that correct unbalanced miRNA production in disease.

9The first described example of an RBP regulating miRNA biogenesis at the post-transcriptional level involved the role of the pluripotency promoting proteins LIN28A and LIN28B in the regulation of the let-7 family of miRNAs in pluripotent embryonic stem cells (ESCs). These proteins are enriched in undifferentiated cells and their expression is gradually switched off during differentiation. LIN28 proteins harbor two RNA-binding domains, a cold shock domain (CSD) and two zinc knuckle domains that mediate recognition of the TL of let-7 in a sequence-specific manner. Binding of LIN28 proteins to let-7 precursors blocks their processing by different mechanisms at either the DROSHA and/or DICER level. Structural studies of LIN28 proteins in complex with sequences from several let-7 precursors revealed a bipartite recognition signal within the TL of let-7 precursors. The LIN28 CSD domain, which has a limited sequence specificity, binds to a let-7 closed loop to induce a conformational change of this precursor that facilitates binding of the CCHC zinc knuckles to a GGAG motif. LIN28B binds to the TL of let-7 precursors and affects their processing by blocking the activity of the Microprocessor in the nucleus. In contrast, LIN28A functions in the cytoplasm, where it recruits a TUTase (either TUT4 or TUT7) that adds a short oligo (U) stretch to the 3′-end of precursor miRNAs and blocks their processing by DICER. The E3 ligase TRIM25 acts as an auxiliary factor for LIN28A by binding to the let-7a precursor and stimulating TUT-mediated uridylation. Subsequently, recruitment of the 3′–5′ exoribonuclease DIS3L2 causes the degradation of the uridylated pre-let-7. This inhibitory effect of let-7 production is important to block miRNA-mediated differentiation in stem cells. LIN28A also inhibits the biogenesis of the neuro-specific miRNA-9 during neuronal differentiation of mouse cells, albeit using an uridylation-independent mechanism.

10Use of an unbiased in vivo cross-linking and immunoprecipitation protocol (CLIP) searching for RNA targets of the hnRNP protein, hnRNP A1, identified the miRNA precursor, pre-mir-18a. This miRNA is expressed as part of the miR-17-92 cluster encoded as an intronic polycistron, that includes six individual miRNAs (miR-17, 18a, 19a, 20a, 19b-1, and 92a-1) and is frequently amplified and/or overexpressed in human cancers, being also termed oncomiR-1. Of interest, hnRNP A1 has been functionally characterized as a general RBP, with a role in many aspects of RNA processing, including alternative splicing regulation, IRES (internal ribosome entry site)-mediated translation and even telomere maintenance (;;; for review, see ). HnRNP A1 has two RNA recognition motif (RRM) domains, each harboring conserved RNP-1 and RNP-2 submotifs that represent the RNA-binding region and a C-terminal glycine-rich domain. Mechanistically, we showed that hnRNP A1 binds to the TL of pri-mir-18a and induces a relaxation at the stem–loop structure near the DROSHA cleavage site, resulting in increased Microprocessor-mediated processing (Fig. 2A,B; ). The processing of the other miRNAs in this cluster is not affected, indicating that hnRNP A1 acts locally to influence the processing of its target pri-miRNA. Importantly, we also observed phylogenetic sequence conservation of TL sequences of precursor miRNAs, suggesting that these sequences act as a landing pad for regulatory factors that could have a positive or negative role in miRNA production, either at the level of DROSHA and/or DICER. More recently, using an integrative structural biology approach combined with biochemical and functional assays, we were able to demonstrate that hnRNP A1 forms a 1:1 complex with pri-mir-18a, in which the tandem RRM domains of hnRNP A1 recognize two UAG motifs in the pri-mir-18a TL and the proximal stem region (Fig. 2B). This structural approach also confirmed that binding of hnRNP A1 to the TL induces an allosteric destabilization of base-pairing in the pri-mir-18a stem that promotes its processing. Of interest, binding of hnRNP A1 to the conserved TL of a precursor miRNA does not always result in enhanced miRNA processing. We showed that hnRNP A1 binding to the TL of pri-let-7 has an inhibitory role in let-7 production in differentiated cells. This is due to a different mechanism from that shown for pri-mir-18a, and involves antagonistic roles for hnRNP A1 and another hnRNP protein, the KH-type splicing regulatory protein, KSRP, which was shown to promote let-7 biogenesis in differentiated cells (Fig. 2C;; ). Interestingly, KSRP not only regulates the processing of let-7, but also binds to the TL of a subset of pri- and pre-miRNAs that includes miR-20, miR-26b, miR-106a, miR-21, miR-16, and promotes both DROSHA- and DICER-mediated steps (Table 1;, ). Thus, these findings with LIN28 and hnRNP A1 suggested a previously uncharacterized role for general RBPs as auxiliary factors that influence the processing of specific miRNAs and prompted the search for novel regulators.

11A flexible and long TL of approximately 10 nt was proposed to be required for efficient DROSHA processing; however, specific sequences at the TL region of some pri- and pre-miRNAs were shown to only have a minor effect on miRNA production. In agreement, most precursor miRNAs display a poor phylogenetic conservation in the TL region, when compared with the high level of conservation observed in mature miRNA sequences. However, a phylogenetic analysis of human pri-miRNAs sequences revealed that approximately ∼14% (74 out of 533) of all miRNAs displayed high conservation of the TL sequence, indicating that these sequences could act as a landing platform for the binding of auxiliary factors, such as hnRNP A1, that influence the post-transcriptional regulation of miRNA production. This was further validated by the use of 2′O-methyl oligonucleotides complementary to conserved TLs, which we termed LooptomiRs (for loop-targeting oligonucleotide anti miRNAs) that block the in vitro processing of precursor miRNAs. We attributed this to a block exerted by looptomiRs on conserved sequences within the TL that are recognized by auxiliary factors required for the efficient processing of these targeted miRNAs in vitro. Conversely, it is also likely that looptomiRs could block the access of factors that negatively regulate processing of target precursor miRNAs.

12In several cancers, the tumor suppressive role of let-7 is abrogated by the increased expression of its negative regulator, LIN28. Building on this concept, short, loop-targeting oligoribonucleotides were used to block binding of the negative regulator LIN28 to the precursor of let-7. These looptomiRs selectively antagonized the docking of LIN28, but still allowed the processing of pre-let-7a-2 by DICER, leading to suppression of growth in cancer cells. An RNA aptamer that specifically targets the pri-mir-17-92 cluster was identified through an in vitro selection process. This aptamer binds to the TL of pri-mir-18a and inhibits the biogenesis not only of miR-18a, but also of all other five miRNAs within this cluster. In contrast, looptomiR-targeting miR-18a only affects the processing of this miRNA, reflecting mechanistic differences on how these two reagents influence miRNA biogenesis. These evidences strongly suggest that TL recognition by RBPs could constitute a general mechanism to regulate miRNA biogenesis that operates via different mechanisms, such as altering the RNA structure of the precursor itself, recruiting additional RNA enzymes and/or affecting the recruitment and/or activity of core processing complexes associated with the Microprocessor and/or DICER. Indeed, a growing number of canonical and newly characterized RBPs have been shown to bind to TLs and regulate miRNA biogenesis (Table 1; Fig. 1B,C;; ).

13An increasing number of SNPs and rare mutations within precursor and/or mature miRNA sequences linked to human disease have been reported. Despite correlations between the presence of polymorphisms in pri-and pre-miRNAs and the corresponding levels of mature miRNAs, the mechanism by which sequence variation and RNA structure control miRNA biogenesis remains mostly enigmatic. A rare genetic variation in the TL of pri-miR-30c-1 (G27 to A) that was found in breast and gastric cancer patients results in increased levels of mature miR-30c. This genetic variant directly affects the Microprocessor-mediated processing of pri-mir-30c-1 by inducing a secondary RNA structure rearrangement that opens up the pri-miRNA stem and facilitates binding of the trans-acting factor SRSF3, a factor which was described to promote Microprocessor activity on a subset of miRNAs (Fig. 3; ). This finding raises the interesting hypothesis that primary sequence determinants in conjunction with RNA structure can act as regulators of miRNA biogenesis. Although this constitutes a largely unexplored area, the emerging picture is that human genetic variation could indeed not only affect miRNA function by targeting either miRNA-binding sites in the 3′-UTRs of target genes and/or miRNA seed sequences, but it could also have an essential role in the modulation of miRNA biogenesis. Pri-miRNA secondary structure can also influence miRNA biogenesis, as observed under limiting levels of DROSHA, when miRNAs without mismatches are processed more efficiently than mismatched miRNAs. Another systematic study identified additional structural elements and sequence distribution for optimal DROSHA processing. This study also predicts that a small but significant fraction of human SNPs could alter pri-miRNA processing, which ultimately could influence the levels of mature miRNA and their biological function. These findings highlight the interplay between genetics, RNA structure, and post-transcriptional regulation of miRNA biogenesis.

14In addition to LIN28 proteins and hnRNP A1, several other RBPs recognize the TL of miRNA precursors and influence, either positively or negatively, their processing (Table 1; Fig. 1B,C). In the following section, we will illustrate the role of a number of RBPs in the regulation of miRNA production.

15The hnRNP protein, TDP-43 (TAR DNA-binding protein-43), also promotes the processing of a subset of precursor miRNAs acting at both the level of DROSHA and DICER processing and this activity is required for neuronal outgrowth. It was shown that TDP-43 is indeed a component of the Microprocessor complex and its Microprocessor-related role affects the biogenesis of a subset of at least six miRNAs, including miR-558-3p, miR-574-3p, and both strands of miR-132 and miR-143. In contrast, cytoplasmic TDP-43 was shown to bind to the TL of pre-miRNAs and interact with the DICER complex.

16The splicing factor SRSF1 (also known as SF2/ASF), which is the prototype of the SR family of splicing regulators, can also promote the maturation of a subset of miRNAs, including miR-7, miR-29b, miR-221, and miR-222. In the case of pri-mir-7, this was shown to be independent of splicing and involved SRSF1 binding to a consensus motif in the stem–loop of pri-mir-7 leading to an increase in DROSHA activity, yet the exact mechanism by which this occurs remains enigmatic.

17Two members of the TET family of proteins, FUS (also known as TLS, translocated in liposarcoma) and Ewing's sarcoma (EWS), have been shown to affect miRNA biogenesis. FUS/TLS, which is associated with familial forms of amyotrophic lateral sclerosis (ALS), promotes the biogenesis of a subset of miRNAs in neuronal cells. It is recruited to chromatin, associates with DROSHA and facilitates the recruitment of the Microprocessor complex to substrate pri-miRNAs, promoting the biogenesis of neuronal miR-9, miR-125b, and miR-132. Interestingly, mutations in both TDP-43 and FUS have been linked with the etiology of ALS, suggesting a possible link between mutations in these miRNA regulators and altered miRNA biogenesis in ALS.

18The EWS protein has a dual and opposing role in miRNA production. On the one hand, EWS has been shown to down-regulate DROSHA at the transcriptional level, but it also has a positive role in miRNA production by binding to flanking sequences in the stem–loop region of pri-miRNAs and promoting the cotranscriptional recruitment of the Microprocessor to chromatin.

19RBFOX3 binds to a subset of pri-miRNAs and directly regulates the Microprocessor-mediated processing of selected pri-miRNAs in neuronally differentiated P19 cells and mouse brain, with stimulatory or blocking effects, depending on the miRNA. Binding of RBFOX3 to the TL or to the stem of individual pri-miRNAs results in recruitment or exclusion of the Microprocessor differentially affecting the processing of the respective pri-miRNAs. Surprisingly, this role in miRNA biogenesis was independent of the cognate binding site for RBFOX3, the UGCAUG motif. In contrast, a more recent study also found a role for RBFOX proteins, in the regulation of miRNA biogenesis, but in this case involved sequence-specific binding of the conserved RBFOX2 RRM to pri-mir-20b and pri-mir-107 containing the cognate motif in their TLs. This binding alters the conformation of these precursors leading to inhibition of DROSHA processing.

20Several miRNAs are expressed in a tissue- or cell type-specific manner, thereby contributing to tissue and cell identity and function. An example of an RBP determining the tissue-specific expression of a target miRNA is illustrated in the case of the brain-enriched expression of miR-7, which is processed from the ubiquitous hnRNP K pre-mRNA transcript. This brain specificity is achieved by the inhibition of pri-mir-7 processing in nonneural cells by the combined action of two RBPs, Musashi RNA-binding protein 2 (MSI2) and the Hu antigen R (HuR), which bind to the TL of pri-mir-7. Mechanistically, HuR binds to the TL of pri-mir-7 and recruits MSI2. Both proteins act then synergistically to stabilize the pri-mir-7 stem–loop structure and inhibit Microprocessor cleavage. This is in agreement with a study showing that HuR depletion results in a significant increase of miR-7 without a noticeable change in the pri-mir-7 levels. Notably, mature miR-7 is sequestered by a circular RNA, ciRS-7, which is primarily expressed in the cerebellum, indicating a sophisticated mechanism of miR-7 regulation. Other miRNAs, such as miR-505, miR-92a-1, or miR-224 are also sensitive to MSI2 and HuR depletion but the precise mechanism of action awaits further characterization. The biogenesis of miR-675 is inhibited by HuR in intestinal tissue by blocking processing of lncRNA H19. Moreover, maturation of miR-199a is blocked by HuR in hepatocellular carcinoma (HCC) in hypoxic conditions that promote glycolytic metabolism and cancer proliferation. Finally, HuR inhibits processing of miR-133b from linc-MD1 noncoding RNA contributing to early stages of the muscle differentiation program. It is yet to be established whether HuR controls these processes alone or in a complex with MSI2.

21RNA editing and RNA editing enzymes can also act to regulate miRNA biogenesis. Adenosine deaminases acting on RNA (ADARs) are responsible for the editing of adenosine residues to inosine in dsRNA. They also affect RNA interference (RNAi) and miRNA processing by deamination of specific adenosines to inosine. RNA editing of pri-mir-142, which is expressed in hematopoietic tissues, blocks its DROSHA-mediated processing. The resulting edited pri-miR-142 is subsequently degraded by Tudor-SN, a component of RISC and also a ribonuclease specific to inosine-containing dsRNAs. In contrast, RNA editing of pri-mir-151 blocks it processing by DICER in the cytoplasm. It was also demonstrated that ADAR proteins can influence miRNA biogenesis independently of their enzymatic activity, as evidenced by the role of ADAR2 in blocking the Drosha-mediated processing of miR-376a-2, independently of its catalytic RNA editing activity.

22The heterodimer NF45-NF90 is an RBP complex that regulates the post-transcriptional expression of a large number of cellular RNAs. It also has a negative role in the processing of pri-mir-7 in HCC. The expression of this heterodimer is elevated in primary HCC tissues compared with adjacent nontumor tissues. The NF90-NF45 heterodimer binds to pri-mir-7-1 and blocks its processing. The biological effect of this repression is the elevation of EGF receptor levels that results in the promotion of cell proliferation in HCC.

23The Y box-binding protein (YB-1), a member of the DNA/RNA-binding family of proteins with an evolutionarily conserved CSD, is a modulator of miRNA processing in glioblastoma multiforme (GBM). A CLIP approach revealed that YB-1 binds to the TL of pri-/pre-mir-29b-2 and regulates its processing by blocking the recruitment of the Microprocessor and DICER to its precursors. Down-regulation of miR-29b by YB-1, which is up-regulated in GBM, is crucial for cell proliferation.

24MBNL1 stimulates the production of miR-1 by binding to a UGC motif located within the TL of pre-miR-1 and competing for the binding of the negative regulator LIN28A in the cytoplasm. In myotonic dystrophy, which is an RNA gain-of-function disease caused by expansions of CUG or CCUG repeats, MBNL1 is sequestered by these expansions. This results in a decreased miR-1 processing in heart samples from patients with myotonic dystrophy, contributing to pathophysiology.

25The RNA-methyltransferase, BCDIN3D, regulates miRNA biogenesis by methylating the 5′-monophosphate end of precursor miRNAs, thus blocking the recognition of 5′-monophosphate by Dicer and inhibiting miRNA processing. In particular, it was shown that BCDIN3D phospho-dimethylates pre-mir-145 both in vitro and in vivo leading to a reduced processing by Dicer in vitro.

26Post-translational modifications of miRNA processing factors have been identified, including phosphorylation as well as ubiquitination and sumoylation that can affect DGCR8, DROSHA and/or DICER complex components (for review, see ). It has been shown that miRNA biogenesis can also be regulated in a cell density-dependent manner, and this is mediated by the tumor-suppressive Hippo pathway. At low cell density, when the Hippo signaling is suppressed, its component YAP relocalizes to the nucleus where it binds and sequesters a Microprocessor-associated component, the RNA helicase DDX17 (also known as p72), thus, down-regulating Microprocessor activity. In contrast, at high cell density the Hippo-induced cytoplasmic retention of YAP restores the association of DDX17/p72 with the Microprocessor stimulating its activity. Thus, the frequent inactivation of the Hippo pathway or expression of constitutively active YAP observed in many cancers results in a widespread miRNA suppression in cells and tumors, which explains the global down-regulation of miRNAs during cancer.

27In particular, the Microprocessor-mediated step of miRNA biogenesis can be regulated by multiple signaling pathways, such as the transforming growth factor beta (TGF-β) and bone morphogenetic protein (BMP) pathways. Mechanistically, TGF-β and BMP-specific SMAD signal transducers are recruited to pri-mir-21 in a complex with the RNA helicase p68 and facilitate its DROSHA-mediated processing. The induction of miR-21 promotes the contractile phenotype in human vascular smooth muscle cells (VSMCs); thus, regulation of miRNA biogenesis by ligand-specific SMAD proteins acts to control the VSMC phenotype.

28As noted above, a global down-regulation of miRNAs is commonly observed in human cancers. The tumor suppressor, p53, has been shown to act as an enhancer of miRNA biogenesis in response to DNA damage, promoting the post-transcriptional maturation of a subset of miRNAs with growth-suppressive function, which includes miR-16-1, miR-143, and miR-145. Mechanistically, p53 interacts with the Microprocessor complex by binding to the DEAD-box RNA helicase p68 (also known as DDX5) in HCT116 cells and human diploid fibroblasts and promotes the processing of pri-miRNAs. The tumor suppressor breast cancer 1 protein (BRCA1) has also been shown to promote the processing of a subset of pri-miRNAs, which include let-7a-1, miR-16-1, miR-145, and miR-34a. BRCA1 interacts with components of the Microprocessor complex, namely DROSHA and the RNA helicase DDX5, as well as with SMAD3, p53, and the DHX9 RNA helicase. This novel function of BRCA1 in miRNA biogenesis could be linked to its well-established roles in tumor suppression and maintenance of genomic stability.

29The initial findings that RBPs bind to TL regions of miRNAs and influence their processing, such as the examples described above (Table 1), prompted a global search for additional factors that control miRNA processing. Several strategies have been developed with the aim of identifying precursor miRNAs (both pri- and pre-miRNAs) whose biogenesis is affected by the binding of RBPs. These include the identification of additional miRNA precursors bound by cognate RBPs, and identification of novel RBPs for a particular precursor miRNA sequence/s. A biochemical method involving an RNA pull-down combined with SILAC mass spectrometry (RP-SMS) led to the identification of trans-acting factors that regulate the processing of miR-7, miR-9, and let-7. In the case of miR-9, which is specifically expressed in the brain, this approach led to the identification of LIN28A, which binds to pri-mir-9 in differentiating cells and induces the degradation of its precursor through a uridylation-independent mechanism.

30A recent proteomics-based pull-down approach focused on the identification of RBPs that recognize 72 different pre-miRNA hairpins used as baits in 11 different cell lines. This identified approximately 180 RBPs that interact specifically with this subset of precursor miRNAs, including known RBPs, splicing factors, as well as other mRNA processing factors. Interestingly, this approach revealed that both TLs, but also stem regions of miRNA precursors, could be specifically recognized by regulatory RBPs. In several cases, loss-of-function experiments validated the impact of these RBPs in the regulation of mRNA biogenesis, although their mechanism of action remains enigmatic. A recent study relied on a high-throughput computational screen to assess a role for 126 RBPs in miRNA biogenesis, using available eCLIP data sets from ENCODE. Those RBPs with enriched binding in the vicinity or within miR-encoding genomic loci represent candidate trans-acting factors for miRNA processing. This exercise resulted in the identification of 116 putative regulators that bind at 1871 annotated human precursor miRNA loci. These candidate RBPs have a potential role either positive or negative at different steps of the miRNA biogenesis cascade. Of interest, some of the interactions of individual RBPs with subsets of precursor miRNAs seem to be cell-type specific. Importantly, this difference was noted even when the corresponding pri-miRNA is expressed in both cell lines. The authors further showed that most RBPs bind fewer than 25 unique miRNA loci and in most cases, depletion of individual RBPs affects the corresponding mature miRNA levels. The precise mechanisms of post-transcriptional regulation that these proteins use to control miRNA biogenesis are currently unknown.

31A role for lncRNAs in the post-transcriptional regulation of miRNA processing has also been recently described (Table 2). Uc.283+A, a lncRNA transcribed from an ultraconserved region, inhibits the Drosha-mediated processing of pri-mir-195. This regulatory event requires complementarity between the lower stem region of the pri-mir-195 transcript and an ultraconserved sequence in Uc.283+A. The proposed mechanism involves lower-stem strand invasion by Uc.283+A, which impairs Microprocessor recognition and blocks pri-miRNA processing. RNA 4 (RNCR4), a retina-specific lncRNA, stimulates the timed processing of the pri-mir-183-96-182 cluster, which is repressed at an earlier developmental stage by the RNA helicase Ddx3x, during mouse retina development. Finally, the heterodimer of splicing factors, NONO/PSF, which are components of the paraspeckles, bind a large number of pri-miRNAs and promote the Microprocessor-mediated processing of these precursors in HeLa cells. The lncRNA, NEAT1, interacts with NONO/PSF and scaffolds RBPs and the Microprocessor to globally promote miRNA processing.

32Given the central role that miRNAs have in controlling the expression of target mRNAs, it is unsurprising that dysregulation of miRNA production leads to aberrant gene expression due to the misregulated expression of target mRNAs. This can affect cellular homeostasis, as well as many developmental pathways and have an impact on the development of human disease. One of the most studied examples is the altered expression of a variety of miRNAs in many different types of cancer, which arises as a consequence of mis-regulation of miRNA production, but also through the presence of mutations in miRNA processing components (for review, see ). Mutations in Microprocessor components, DROSHA and DGCR8 have been identified in Wilms tumors, a pediatric kidney tumor. Conversely, DICER mutations have also been linked to several human conditions, including early childhood tumors (;; for reviews, see; ).

33The regulation of let-7 miRNA precursors by the LIN28 proteins is a clear example of how alterations to post-transcriptional regulation of miRNA precursors can indeed lead to cancer. In the section below, we will discuss novel approaches used in an attempt to influence miRNA biogenesis by targeting regulatory RBPs that regulate the production of miRNAs.

34Several strategies have been developed to affect the production of miRNAs, targeting their nuclear and cytoplasmic processing machineries and/or factors that regulate their biogenesis. We have described above the use of looptomiRs and Aptamers to target the recognition of TLs by RBPs and regulate miRNA processing. Several alternative approaches have been developed to identify small molecules that bind to miRNA precursors or to RBPs that regulate miRNA biogenesis. These include the identification of a peptoid ligand that interacts with the apical loop of pri-mir-21 and inhibits cleavage by DROSHA, of a benzimidazole that inhibits the biogenesis of miR-96 and of polyamine derivatives that block the DICER-mediated processing of pre-mir-372 processing. A different strategy was developed to inhibit miRNAs that are overexpressed in human cancers, such as miR-21, in which a cyclic β-hairpin peptidomimetic binds to RNA stem–loop structures of miRNA precursors, with potent affinity and specificity. This peptide was shown to recognize the DICER cleavage site and inhibit miR-21 processing. Another study focusing on miR-21, used small molecule screening and 3D structure modeling and identified AC1MMYR2 (2,4-diamino-1, 3-diazinane-5-carbonitrile) as a potent inhibitor of pre-mir-21 cleavage by DICER. This inhibitor reversed epithelial-to-mesenchymal transition and suppressed tumor growth. Oleic acid (a natural monounsaturated fatty acid produced by plants and animal cells) inhibits the RNA-binding activity of Musashi RNA-binding protein 2 (MSI2), a negative regulator of pri-mir-7 processing, by binding to its N-terminal RRM. Thus, the action of oleic acid could be used to disrupt the formation of a negative regulatory complex and lead to stimulation of miR-7 biogenesis. Such strategies could reduce the levels of miR-7 target genes such as the EGFR oncogene, to potentially alleviate its deleterious effects in high-grade glioblastomas, where miR-7 is post-transcriptionally down-regulated. Recent small-molecule screenings have identified compounds that inhibit Lin28 binding to RNA. One such compound, LI17, potently inhibited Lin28-mediated oligouridylation of pre-let-7 in vitro and in cells, causing a concomitant increase in mature let-7 levels. Likewise, another compound de-repressed let-7 and inhibited proliferation and stem-like properties in human cancer cells. These examples demonstrate that selective pharmacologic inhibition of RBPs involved in post-transcriptional regulation of miRNA biogenesis could provide a foundation for therapeutic intervention in diseases underpinned by deregulated miRNA levels.

35Several layers of tightly controlled regulation have evolved to maintain the levels of mature miRNAs in order to fine-tune gene expression during development and differentiation. Such multifaceted regulation ultimately prevents gross changes in gene expression that can contribute to numerous diseases. Among these mechanisms of control, post-transcriptional steps are predominant, and increasing evidence shows the central role of general RBPs in the control of miRNA production. The binding of RBPs to TL sequences within miRNA precursors (pri- and pre-miRNAs) has emerged as a general mechanism to regulate the activity of DROSHA and/or DICER. This can encompass different mechanisms, such as conformational changes and dynamic destabilization induced by the binding of these auxiliary factors. For example, the binding of hnRNP A1 or Rbfox proteins to pri-miRNAs leads to structural changes that affect Microprocessor binding and/or activity (Table 1;; ). Another common mechanism is antagonistic binding of the regulatory RBP to either a positive or negative regulator, as seen with the competitive binding of hnRNP A1 and KSRP to let-7 precursors in differentiated cells, or MBNL-1 antagonizing LIN28 binding to pri-mir-1. Thus, increased knowledge of binding sites and regulatory mechanisms could facilitate the manipulation of individual miRNA expression. Similarly, recent efforts have relied on the use of oligonucleotide approaches and/or chemical or biological compounds to target the LIN28/let-7 interaction. In principle, these approaches could be expanded to compete out the binding of positive or negative regulators to individual precursor miRNAs and modify the outcome of the biogenesis pathway to correct the unbalanced miRNA levels.

36Extensive genetic variation leading to altered miRNA biogenesis represents another largely unexplored mechanism of regulation. We and others have established that primary sequence determinants and RNA structure are important regulators of miRNA biogenesis. Of interest, polymorphisms and mutations within or proximal to miRNAs are frequently overlooked in disease and trait studies searching for functionally important variants, but could have an important role in determining levels of miRNA expression. In the near future, efforts will focus on the identification and functional characterization of additional RBPs or other regulators that affect miRNA biogenesis, as illustrated by recent genome-wide efforts. Our expanding knowledge about the mechanisms that regulate miRNA production will be essential to understand and treat human diseases that arise from deregulated gene expression.