Recent progress in miRNA biogenesis and decay

ABSTRACT

MicroRNAs are a class of small regulatory RNAs that mediate regulation of protein synthesis by recognizing sequence elements in mRNAs. MicroRNAs are processed through a series of steps starting from transcription and primary processing in the nucleus to precursor processing and mature function in the cytoplasm. It is also in the cytoplasm where levels of mature microRNAs can be modulated through decay mechanisms. Here, we review the recent progress in the lifetime of a microRNA at all steps required for maintaining their homoeostasis. The increasing knowledge about microRNA regulation upholds great promise as therapeutic targets.

KEYWORDS:

• miRNA biogenesis
• miRNA decay
• miRNA stability
• miRNA targeting
• miRNA therapeutics

Introduction

MicroRNA (miRNA) biogenesis is a complex process that involves multiple steps from transcription in the nucleus to mature miRNA in the cytoplasm [1]. This process is intensively studied due to the exciting regulatory roles of miRNAs, with applications in both basic research and therapeutics. The six primary processes that define miRNA levels include transcription of the primary miRNA (pri-miRNA) transcript; processing of the primary miRNA transcript into precursor miRNAs (pre-miRNA); export of the precursor miRNA to the cytoplasm; processing of the precursor miRNA into the mature miRNA; recognition, inhibition or cleavage of target transcripts; and mechanisms to regulate stability and decay of mature miRNAs (Figure 1).

Figure 1. Summary of the canonical miRNA biogenesis pathway.

Biogenesis of miRNAs consists of the stepwise processing of pri-miRNAs in the nucleus and the cytoplasm. Briefly, pri-miRNAs fold into distinct hairpin-like structures that are recognized and cleaved by the Microprocessor, a heterodimer complex formed by DROSHA and DGCR8. The Microprocessor cleavage site (blue arrows) is defined by multiple structural and sequence motifs. The resulting product is exported into the cytoplasm via XPO5, for a second cleavage event by DICER (red arrows). The resulting duplex is loaded into one of the AGO paralogs for strand selection and assembly of mature RISC complexes with TNRC6. Recognition chambers inside AGO facilitate the recognition of target RNAs, and the subset of nucleotides involved in these facilitated pairings is known as the seed (cerulean circles) and the supplemental pairing regions (tangerine circles). Upon stable miRNA:target recognition, TNRC6 will mediate the recruitment of deadenylation complexes, leading to translational repression and decay of target mRNAs.

RNA polymerase II first transcribes a pri-miRNA transcript from the DNA [2]. The pre-miRNA is contained in the primary sequence and forms a hairpin structure that is recognized and cleaved by the Microprocessor complex consisting of the RNase III enzyme DROSHA and its cofactor DGCR8 [3], resulting in the processed hairpin-shaped pre-miRNA.

The Exportin 5 (XPO5) protein then exports the pre-miRNA to the cytoplasm [4] where the pre-miRNA is further processed by the RNase III enzyme DICER and its cofactor TRBP to produce a double-stranded RNA duplex consisting of the mature miRNA and its complementary strand, known as the passenger strand [5]. The duplex is unwound, and the mature miRNA is loaded into the RNA-induced silencing complex (RISC). The RISC/miRNA complex then binds to target mRNAs and represses their translation or promotes their degradation [6].

The process of miRNA biogenesis is tightly regulated at each step, and dysregulation can result in a variety of diseases, including cancer, neurodegenerative disorders, and cardiovascular disease. Here we focus on mammalian miRNAs, and while most principles apply to other organisms, important differences exist and are reviewed elsewhere in the scientific literature.

Transcription

miRNAs are processed as long primary transcripts that can expand several kilobases long and present the opportunity for multiple promoters regulating a single miRNA [1]. In addition, polycistronic transcripts are responsible for the biogenesis of several miRNAs contained in the same primary transcript, combining the impact of transcription regulation and differential processing efficiency to obtain a dynamic range of miRNA regulation and expression [1]. In addition to complex promoter regulation, enhancers have also been proposed to be involved in both pri-miRNA transcription and processing, where biogenesis of master miRNAs essential for cell identity is regulated at both the transcription and pri-miRNA processing steps [7]. The identification of miRNA promoters has proved much harder than anticipated due to the long length and complexity of pri-miRNA transcripts. Early studies using RNA sequencing were able to map a subset of full-length pri-miRNA transcripts, revealing the position of promoters, especially following knockdown of DROSHA to stabilize the pri-miRNA transcripts and thus make the transcription start site and promoter location easier to detect in RNA sequencing experiments [1,2].

While miRNA promoter identification is experimentally challenging, recent computational tools and databases have added significantly to our knowledge of miRNA promoters. The RSmirT compared 14 studies on miRNA TSS identification from 2007 to 2017 to provide a set of miRNA TSSs supported by several studies and highlighting the challenges in computational miRNA promoter identification [8], providing a good overview of the different methods applied during those 10 years. More recently, the DIANA-miRGen v4 uses CAGE and ChIP-seq data to collect information on cell type-specific miRNA promoters and provide extensive prediction of dynamic miRNA promoters across tissues and cells [9], giving a good starting point for further characterization and experimental validation. Similarly, a deep learning model called D-mirT allows for computational prediction of condition-specific miRNA TSSs based on epigenetic features and sequencing data [10]. The development of these tools to predict dynamics and tissue- and condition-specific use of miRNA promoters is a very important foundation for further understanding the functional impact of differential miRNA promoter usage in biology.

This is well in line with studies coupling miRNA processing to transcription for efficient maturation, showing that transcripts expressed from a miRNA promoter are retained at chromatin and are more efficiently processed than pri-miRNAs coming from an mRNA promoter [11].

Primary transcript processing

The pri-miRNA transcript is processed to an approximately 80 nts long pre-miRNA by the Microprocessor complex, consisting of the core proteins DROSHA and DGCR8 along with a number of co-factors that have been well characterized and described [3,12–15].

Several regulatory steps ensure proper processing of the pre-miRNA, including binding of co-factors to the Microprocessor complex. Both sequence context, structure of the pri-miRNA and co-factors contribute to the efficiency, precision and dynamics of pri-miRNA processing. In particular, sequence elements upstream (UG) and downstream (CNNC) of the precursor (Figure 1) are affecting processing efficiency, by affecting binding by co-factors as SRSF3 [1,16]. The precision of the processing by the Microprocessor is to a large extent regulated by the GHG motif (Figure 1), embedded in the lower stem region of the pri-miRNA, supported by structural studies to assist in RNA recognition by the double-stranded RNA-binding domain of DROSHA and ensuring nucleotide-precision cleavage of the pri-miRNA into the pre-miRNA [17–19]. Computational studies further suggest that the GHG motif is a better predictor of DROSHA cleavage site present as a structural rather than sequence motif [20].

Several methods have been developed to quantify the efficiency of pri-miRNA processing, and high-throughput RNA sequencing analysis shows differential regulation of pri-miRNA processing between healthy and disease clinical samples in the liver [21]. An in vitro assay combined with sequencing provides a comprehensive dataset on DROSHA dependency of pri-miRNAs as well as cis-elements with impact on processing and alternative processing [22,23]. A study on the chromosome 19 miRNA cluster, containing 46 miRNAs, shows that alternative promoter usage and enhancer regulation of the pri-miRNA can as well affect tissue-specific processing of pri-miRNAs [24].

The cleavage of suboptimal pri-miRNAs can be enhanced by a phenomenon called cluster assistance, where neighbouring miRNAs can compete or assist in biogenesis of miRNAs within the same cluster [25]. Examples of this include the Enhancer of Rudimentary Homolog (ERH), a protein identified to act as a co-factor of the Microprocessor complex for processing suboptimal hairpins [26], that can bind to helper hairpins and facilitate processing of otherwise defective hairpins, as, e.g., miR-451 where miR-141 binding helps processing by Microprocessor [27,28]. In another study, scaffold attachment factor B2 (SAFB2) was identified as a Microprocessor co-factor mediating a miR-16-1-dependent miR-15 cleavage [29]. Interestingly, XPO5 previously thought to purely mediate the nuclear export, has been reported to enhance Microprocessor of certain pri-miRNAs and clusters [30]. These studies demonstrate the requirement for specific co-factors for efficient processing of subsets of miRNAs with a suboptimal structure in the primary transcript.

Several studies have developed tools to study the molecular features determining pri-miRNA processing efficiency and accuracy. One of these studies provides a functional atlas of elements promoting pri-miRNA processing and suggests that low Shannon entropy favours DROSHA cleavage [23]. In complementary studies such as the MapToCleave additional new sequence- and structural features impacting pri-miRNA processing are identified, as, e.g., the GHG motif mentioned earlier in this review [20]. Other studies support that it is primarily structure, rather than sequence, that determines the affinity of the Microprocessor complex for binding to the pri-miRNA processing site [31], and that differences in structure can mediate dynamic processing efficiency of pri-miRNAs [32]. In summary, both structure and sequence of pri-miRNAs can affect biogenesis, highlighted by diverse studies taking different approaches to reveal various aspects of the complex regulation of pri-miRNA processing. Processing can also be affected by SNPs in the pri-miRNA sequence, either by affecting the sequence of Microprocessor co-factor binding sites or by mediating structural changes [33]. Supporting the structural importance of the pri-miRNA for processing, cryo-EM structures of DROSHA in complex with pri-miRNA has revealed the molecular details of how Microprocessor recognizes hairpins for processing and how efficiency and accuracy of pri-miRNA processing is achieved [18,19].

Cytoplasmic processing

The pre-miRNA is further processed in the cytoplasm by DICER into a 21–25 nucleotides (nts) double-stranded miRNA that is subsequently incorporated into the RISC complex. The recognition and cleavage of pre-miRNAs by DICER strongly relies on 2 nts overhangs left by DROSHA, with a distinctive 5′-terminal phosphate group and 3′-OH termini. Insight from the resolved structure of DICER and pre-miRNA substrates has revealed multiple large conformational changes that enable the complex to precisely bind bona fide substrates and reach a catalytically active state [34–37]. During these structural rearrangements, multiple domains are involved in locking pre-miRNAs in a specific position through sequence-independent [38] and sequence-specific features (‘GYM motif’) [39]. The combination of these features leads to the cleavage of the pre-miRNA around 21–25 nucleotides away from the 2 nts overhangs. DICER binding of pre-miRNA is aided by TRBP (transactivation response element RNA-binding protein) that ensures efficient and accurate processing [40–42]. Since then, it has been shown that TRBP facilitates pre-miRNA recognition in RNA crowded environments [43]. The loading of the resulting miRNA duplex from DICER into AGO involves a complex of chaperones (HSC70/HSP90) and co-chaperones that includes RPAP3 and UNC45 among others [44]. These chaperones not only stabilize empty AGO but also promote the transfer of miRNA duplexes [45]. Next, one of the strands from the duplex is selected (named guide strand) to form the miRNA-Induced Silencing Complex (miRISC), while the other strand is released from the complex (named passenger strand) and degraded [46]. The selection of either the 5’ proximal or 3’ strand from duplexes gives rise to so-called 5p and 3p miRNAs, respectively. Unwinding of miRNA duplexes and subsequent strand selection is an important process that can switch functions within a miRNA locus (reviewed in [46]). Some of the determinants of the strand selection include intrinsic sequence and structure features [47,48], but it can also be regulated via the interaction with cellular factors [49]. Overall, the misprocessing of pre-miRNA substrates can lead to mature miRNAs with altered target mRNA specificities (‘seed shift’) and/or changes in the miRNA strand loaded (‘arm switch’) [49,50], putting emphasis on proper pre-miRNA processing for correct miRNA function.

Inhibition and cleavage of the target transcript

The mature miRNA in the RISC complex binds its target mRNAs and regulates protein synthesis. This miRNA target interaction is based on base-pairings between the miRNA and the mRNA, governed by a region known as the ‘seed sequence’ (nts 2–8 of the mature miRNA) [6]. Quantitative biochemical analysis of hundreds of thousands of putative sites has revealed many miRNA-specific differences in target-site affinities for canonical and non-canonical sites, as well as AGO2-mediated slicing [6,51]. Despite that miRNA:target interactions can be detected all over the transcript body [52], most interactions happen in the 3’UTR [53,54]. On top of that, the topology has a strong impact on repression efficacies, and masking of individual miRNA sites by intramolecular RNA structures or RNA binding proteins can affect efficiency of miRNA regulation [55,56]. Recent studies have shown that upon target recognition, both AGO and miRNA complexes undergo critical structural rearrangements, some of which are determined by the architecture of specific miRNA target base-pairing interactions [57,58].

While AGO proteins provide target specificity to the RISC complexes via the loaded miRNAs, TNRC6 (trinucleotide repeat-containing gene 6 protein family) binds AGO’s PIWI domain and mediates the recruitment of decapping and deadenylation complexes. Single-molecule studies indicate that miRNAs bind target mRNAs immediately after their nuclear export, but their effects on translational repression and mRNA decay occur 30–60 minutes later [59].

Briefly, TNRC6 recruits the poly(A)-nuclease (PAN) and the CCR4–NOT complex to the poly(A) tail of the target mRNA for the shortening of the poly(A) tail, which is then followed by mRNA destabilization through decapping and 5′-to-3′ exonucleolytic degradation. While recent studies indicate that miRNAs specifically trigger a fast decay of target mRNAs with already short poly(A) tails [60], in early embryos the primary mechanism of miRNAs on target mRNAs is modulation of translational efficiency as poly(A) tail shortening is decoupled from mRNA stability on early embryos [61]. This piece of evidence is also supported by single-molecule studies that indicate that miRNAs can promote mRNA degradation without interfering with translation [62]. TNRC6 and the CCR4-NOT complex also contribute to the recruitment of DDX6 which interacts with eIF4E (competing with eIF4G) thus enhancing decay and translational repression of target mRNAs (reviewed in [63]). TNRC6 is also recruited to less extent by TRIM71 and UPF1, thus suggesting a competition between the miRNA and other RNA silencing pathways [64]. Detailed characterization of each TNRC6 paralog indicates that their functions are mostly redundant, despite some evidence indicating that TNRC6A may regulate some additional subsets of genes [65]. Interestingly, there is evidence that in stem cells AGO-miRNA complexes can silence target mRNAs in the nucleus through the recruitment of CCR4-NOT complex by TNRC6 [66]. In the nucleus, AGO-miRNA complexes are unrestricted, and they can repress by binding to coding sequences and introns [66]. Of note, an engineered mouse model expressing dominant negative peptide of TNRC6 was developed to study in vivo miRNA function without affecting miRNA biogenesis or abundance [67].

Stability and decay

The levels of mature miRNAs within cells are a function of synthesis, biogenesis, and degradation. Stability of the mature miRNA is an important step for regulating its function, as long half-lives provide a very cost-efficient regulation of targets, whereas short half-lives allow quick and focused responses to regulate gene expression. Several studies have addressed the half-lives of miRNAs and reported ranges spanning from minutes to days, e.g. cardiac-specific miR-208 has a half-life of ~2 weeks [68], while others, such as miR-16 can decay within minutes in response to cell cycle phases [69] or changes in light-darkness in retinal neurons [70], underlining the importance of tight control of miRNA degradation for proper target regulation.

miRNA decay follows several different pathways as outlined in Figure 2. One layer of regulation of the stability of AGO-bound miRNAs arises from the attack by different RNA nucleases. Exonucleases DIS3L2 and PARN are specialized in targeting miRNAs that are marked post-maturation with oligo(U) and oligo(A) tails, respectively [71,72]. These tailing marks on miRNAs are mediated by terminal nucleotidyltransferases (TENTs), a family of RNA polymerases specialized in the non-templated addition of nucleotides (reviewed in [73]). Non-templated uridylation of miRNAs is catalysed by TUT4 and TUT7 [74], despite the contribution of each of them to the addition of mono(U) or oligo(U) tails can be miRNA and cell-type dependent [75,76]. By contrast, broad miRNA monoadenylation depends on TENT2 [76,77], while TENT4B (PAPD5) adds oligo(A) tails to a specific subset of miRNAs (miR-21-5p, miR-181b-5p, miR-92b-3p, and let-7e-5) [72]. Structural and biochemical studies have shown that tailing of AGO-bound miRNAs is dependent on the release of the miRNA 3’ end from the PAZ domain [76,78]. This dislocation of the 3’ end is promoted through extensive base-pairings between the miRNA and target mRNAs [79]. Of note, other miRNA decay mechanisms include the endonuclease Tudor-SN that cleaves internal [C/U]A dinucleotides around five nucleotides away from the 3’ end of the mature miRNA [80].

Figure 2. Summary of the mature miRNA decay pathways.

Decay of mature miRNAs can happen through multiple pathways. Up to date, three distinct mechanisms have been identified. (1) Tailing of miRNAs by terminal nucleotidyl transferases (TENTs) primes the decay via RNA nucleases. The example shows the decay via TUT4/7 mediated uridylation and DIS3L2. Similarly, TENT2 and TENT4B can mediate the adenylation that facilitates PARN mediated decay. (2) Other RNA nucleases, such as Tudor-SN, can also decay miRNAs independently of tailing. It is unclear whether certain target RNAs may facilitate the exposure of cleavable nucleotides. (3) ZSWIM8 can mediate the ubiquitination and proteolysis of AGO complexes. As a consequence, unprotected miRNAs can be degraded by cytoplasmic nucleases. In general, binding to target RNAs with extensive complementarity has been identified as a cue for most decay pathways.

Another layer of regulation comes via the decay of entire AGO:miRNA complexes. The proposed model here indicates that the structural changes in AGO induced by extensive base-pairing between the miRNA and certain target RNAs [78] are sensed by ZSWIM8 that acts as an adaptor for the cullin-RING ubiquitin ligase complex (CUL3) [81,82]. Ubiquitinated AGO:miRNA complexes are substrates for degradation by the proteasome, releasing naked miRNA for nuclease degradation [81,83]. Knock-out of ZSWIM8 orthologs in mice and fly displayed extensive organ and growth defects that lead to perinatal lethality [82–84], in both cases some of the phenotypes could be rescued by an additional deletion of individual ZSWIM8-regulated miRNAs [83,85]. Overall, this indicates the importance of timely miRNA decay for proper embryonic development.

Target RNAs can affect the stability of miRNAs, e.g. the long non-coding RNA Cyrano [86], the circRNA Cdr1as [87,88] and the mRNAs Nrep, Serpine1 and Bcl2l11 [89–91], showing an intricate relationship between mature miRNAs and their targets. Interestingly, genomic deletion of individual target sites for miRNA decay resulted in miRNA dysregulation [86,92] and increased susceptibility to environmental stressors [92].

Remarkably, both ‘tailing and trimming’ and ‘ubiquitination’ pathways are broadly conserved in insects, worms and plants [93], and it remains enigmatic how the different mechanisms of miRNA stability and decay are orchestrated [93,94]. Interestingly, a recent pan-cancer study found that trimmed miRNAs accumulate in cancer cells, and its restoration to its normal length had robust anti-tumoural effects [95].

microRNA biogenesis as a therapeutic target

Pri-miRNA processing is a complex and finely tuned process ensuring proper expression of mature miRNAs but also constituting a central step for therapeutic targeting, where manipulating miRNA biogenesis at an early step could provide more benefits than targeting overexpressed miRNAs at the mature level in disease.

miRNAs have been explored for several years as therapeutic targets at the mature level using antisense oligonucleotides. Other possibilities are antisense oligonucleotides targeting the stem of miRNA hairpin within the pri-miRNA [96], and recent work on developing small molecules that recognize and cleave the defined structure of a pre-miRNA stem has shown promise for experimental and clinical use [97], where e.g. miR-200 and miR17–92 have been successfully targeted and degraded by small molecules [98,99]. These studies show new and promising ways to manipulate miRNA expression at processing steps before mature miRNAs, with the potential to improve experimental and clinical application of miRNA manipulation, highlighting the application of recent progress in miRNA biogenesis regulation for therapeutic use.

Concluding remarks

The miRNA field has yielded a very detailed and quantitative picture in recent years. The major players in control of miRNA biogenesis and decay are now well established, although, it is to be expected that new factors that play important roles in the regulation of individual miRNAs in tissue- or cell-specific contexts will be identified. As the field continues to evolve, our understanding of the regulation of levels and function of mature miRNA may integrate into the broader cellular responses to environmental cues. In this line, phosphorylation cycles on AGO have been reported to impact its function [100–102], providing avenues for crosstalk of signalling pathways. Other responses may exploit the dynamic formation of phase-separated droplets already described for AGO and TNRC6 complexes [103,104] that might expand the roles and implications of miRNAs and miRNA biogenesis further.

Acknowledgments

Work in the author’s laboratories is funded by the Lundbeck Foundation (XB-DR, UAVØ), Novo Nordisk Foundation, Independent Research Fund Denmark, Danish Cancer Society, and the Carlsberg Foundation (UAVØ). We thank Itxaso Santiago and Tobias Gellrich for their feedback on the manuscript.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

No data availability statement.

Additional information

Funding

The work was supported by the Carlsbergfondet Danmarks Frie Forskningsfond Kræftens Bekæmpelse Lundbeck Foundation Novo Nordisk Fonden.