MicroRNA in Human Malignancies

MicroRNA in Human Malignancies offers a deep overview of the role and translational significance of miRNAs in the development of cancer and other malignancies. The book establishes the foundations of the field by covering essential mechanisms and the translational potential of miRNAs in the field of oncology. Specific topics covered include invasion and metastasis, miRNAs and metabolism, and opportunities of miRNAs in therapeutics. Chapters on diseases include content on disease-related pathophysiology, as well as diagnostic, prognostic and predictive value. This book is an essential reference for students entering the field, as well as researchers and investigators.

• Provides fundamental and translational chapters that facilitate the acquisition of knowledge needed to design and perform innovative miRNA-related research studies
• Synthesizes current research, with a critical review on the field
• Offers in-depth research by leading experts in the field

• Language: English
• Publisher: Academic Press
• Release date: Feb 18, 2022
• ISBN: 9780128232743

Book preview

Chapter 1: Overview on miRNA classification, biogenesis, and functions

Jacob Anderson O’Briena; Vu Hong Loan Nguyena; Chun Penga,b    a Department of Biology, York University, Toronto, Canada

b Centre for Research in Biomolecular Interactions, York University, Toronto, Canada

Abstract

MicroRNA (miRNA) are small noncoding and single stranded RNAs involved in the regulation of gene expression. The core characteristic defining a miRNA molecule is the hairpin structure of RNA with central mismatches and key motifs involved in its processing. This primary miRNA, either contained within a host gene’s primary transcript or transcribed via its own promoter, is cleaved by a microprocessor complex to produce a pre-miRNA hairpin. Following transport to the cytoplasm, the terminal loop is removed by the endonuclease DICER resulting in the dsRNA miRNA duplex. One of the strands in the miRNA duplex is loaded into the AGO family of proteins to form the minimally competent miRNA-induced silencing complex (miRISC). This complex, with additional effector proteins, facilitates target RNA regulation typically by inhibiting translation initiation and promoting mRNA decay. Target specificity is mostly a result of complementarity between the 5′ seed region of miRISC-associated miRNA (nt 2-8) and target mRNA sequences, termed the miRNA response element (MRE). Although MRE have been located throughout the mRNA molecule, it is usually the 3′ untranslated region that elicits a regulatory potential. Beyond targeting cytoplasmic mRNA, miRISC can regulate target RNA within the nucleus, generally promoting its degradation. Nuclear miRISC can also promote or inhibit transcription through multiple mechanisms involving DNA and co-transcriptional RNA interaction, affecting chromatin state and transcriptional activity. The degree to which any specific target may be regulated is dependent on many factors but the affinity between miRNA and MRE, the total number of MRE and miRNA within the cell, and the subcellular localization are all crucial. In this chapter, we provide an overview of miRNA biogenesis and nomenclature, and discuss how miRNAs regulate gene expression.

Keywords

miRNA biogenesis; gene regulation; miRNA function; miRNA dynamics; miRNA classification; miRNA nomenclature

Introduction

Since the discovery of the first microRNA (miRNA), lin-4, in 1993 by the Ambros and Ruvkun groups in Caenorhabditis elegans (Lee, Feinbaum, & Ambros, 1993; Wightman, Ha, & Ruvkun, 1993), there have been more than 120,000 publications on the topic of miRNAs. Moreover, more than 1000 human miRNA genes have been validated while many hundreds more potential miRNA genes require further study (de Rie et al., 2017; Kozomara & Griffiths-Jones, 2011). MicroRNA are small noncoding RNA involved in the regulation of gene expression facilitated by the miRNA-induced silencing complex (miRISC). This complex, guided mostly by sequence complementarity between nucleotides 2–8 of miRNA and target RNA, allows for the highly specific regulation of gene expression. It is now well established that miRNAs play critical roles in normal developmental and physiological processes (Hayder et al., 2018; Rajman & Schratt, 2017) and their aberrant expression is associated with many human diseases, including cancer (Paul et al., 2018; Tufekci et al., 2014). In addition to cell-intrinsic gene expression regulation, miRNAs can also be secreted into extracellular fluids and play a role in mediating cell–cell communication (O'Brien et al., 2018). Importantly, extracellular miRNAs have been under intensive investigation for their potential to be used as predictive or prognostic biomarkers for human diseases (Hayes, Peruzzi, & Lawler, 2014; Huang, 2017; Wang, Chen, & Sen, 2016). It is no surprise then that miRNAs have become such a crucial element in our understanding of the cell and why miRNA as a field is so intensely studied.

In this chapter, we will start with an introduction of the complex network of gene regulation, followed by a short overview of miRNA biogenesis and nomenclature and a more in-depth discussion of miRNA functions in regulating gene expression and the underlying mechanisms. Each section will cover multiple independent but related topics that on their own provide a static view, a snapshot, of miRNA biology. Nonetheless, the aim of this chapter is not only to introduce fundamental miRNA theory but also to generate a deeper understanding of the potential reality of miRNA within mammalian cells. As such, in each section, the key concepts will be described assuming a framework of a dynamical nonlinear system to place the abstracted snapshot back into the living cells.

MicroRNA and the complex biological system of gene expression

A small noncoding RNA is typically defined as being between 18 and 31 nucleotides (nt) in length (Dozmorov et al., 2013), although this is not a strict definition but a commonly observed categorization; miRNAs are on average 22 nt in length. An RNA molecule that cannot be translated into a protein is termed noncoding RNA (ncRNA) as opposed to an RNA that contains a protein-coding sequence (CDS), such as messenger RNA (mRNA). The areas of gene regulation that miRNAs are involved in range from transcriptional/posttranscriptional to translational, meaning that miRNA can regulate the amount of mRNA in a cell as well as how much mRNA is translated into protein.

Both these processes can then be further differentiated theoretically and mechanistically. First, both transcription and translation are at least dynamical steady-state equilibrium processes. A dynamical system is a system that changes over time (Janson, 2012). If the amount of mRNA existing within any given cell is a dynamical steady-state process, this implies that the amount of mRNA both changes (regulated creation and destruction) and is maintained over time. There are many ways cells regulate the abundance of mRNA, generally termed mRNA metabolism, and they can be grouped into transcription, maturation, and decay/turnover.

Messenger RNA are transcribed by RNA polymerase II (POLII) stochastically (Cao & Grima, 2020; Elowitz et al., 2002), analogous to randomly. In other words, the average rate of transcription of any given gene may remain the same at steady state (e.g., 3 transcripts/min) but the time interval between each transcription event tends to vary within some probability distribution. Maturation of the primary transcript involves co/posttranscriptional splicing, 3′ poly-adenylation, m⁷G capping, and nuclear export to the cytoplasm (Ben-Yishay & Shav-Tal, 2019). Within the cell, mRNAs do not exist alone as naked RNAs but instead associate with a variety of proteins involved in the regulation of their stability (i.e., change in mRNA abundance over time) and translation potential (how likely the mRNA is to be translated into protein); protein-bound mRNAs are termed messenger ribonucleoproteins (mRNPs) (Borbolis & Syntichaki, 2015). It should be generally assumed that when discussing mRNAs, mRNPs are implied.

In the cytoplasm, various processes regulate mRNA turnover. Stability can be modulated by subcellular localization of the mRNA to areas that promote or suppress mRNA decay or as a result of sequence errors detected during translation that induce mRNA degradation (Borbolis & Syntichaki, 2015). Although the cause of degradation can vary, this usually involves poly-deadenylation, m7G cap removal (decapping), and/or endonucleolytic cleavage followed by 5′– 3′ or 3′– 5′ exonucleolytic decay.

Taken together, mRNA metabolism is at a steady state when there is a balanced rate of transcription, maturation, and turnover. If any one of these equilibriums are shifted, such as by increasing transcription, the entire cellular system equilibrates to its new steady state, which may or may not be an increase in mRNA abundance. Moreover, the complexity of the system allows even a small change in gene expression to result in significant and generally nonlinear effects on the overall cellular system (Janson, 2012). There are thousands of genes expressed at any given time, called the transcriptome (Mele et al., 2015), all of which in principle exert some degree of direct or indirect regulation on each other. Consequently, when we talk about gene regulation in general, it is important to keep in mind the interconnectedness of the cellular system.

Lastly, how are we defining a system? In our first example, the system was the entire body, but we could easily define our system as a tissue, single cell, subcellular compartment, signaling pathway, etc. What’s important is that elements (nodes) of the system, for example a single gene, are described relative to all other elements in the system. With the application of rules that define the consequence of all pairwise interactions of nodes, the system can be modeled given some initial state. The models of biological behaviors that scientists have generated are described as complex systems, in some cases through networks (Yan et al., 2018) and others through analytical solutions (Janson, 2012). All the interesting biological behaviors that we have come to understand well describe a complex system, one that is made of many heterogeneous elements, filled with nonlinear outputs, stochasticity, dynamics, massive interconnectedness, and feedbacks (Ma’ayan, 2017). A complex system is one where the same change in one parameter can lead to an entirely different set of outcomes with a simple change of the initial state.

MicroRNA biogenesis, nomenclature, and classification

Origins of the canonical primary miRNA transcript

The production of miRNA is a multistep, hierarchical process that involves the processing of RNA transcripts with a characteristic stemloop secondary structure to that of the single-stranded mature miRNA loaded into the Argonaute (AGO) family of proteins at which point they can elicit their gene regulatory potential. An important factor in understanding the first step of miRNA biogenesis, the synthesis of the primary miRNA (pri-miRNA) by POLII, is in the distribution of miRNA genes throughout the genome (Ha & Kim, 2014). Intragenic miRNA genes are located within the body of another protein-coding or noncoding gene called the host gene, and they make up approximately half of all identified miRNA genes (de Rie et al., 2017; Kim & Kim, 2007). Intragenic miRNA can also be further categorized depending on the position they occupy within the host gene, mainly exonic or intronic, with the majority of miRNA genes being intronic (Chang et al., 2015; de Rie et al., 2017). Intergenic miRNAs, as the name suggests, exist independent of a host gene and are transcriptionally regulated by their own promoters (de Rie et al., 2017).

What exactly then is the pri-miRNA? For miRNAs whose expression is coupled to their host gene, the pri-miRNA would be the pre-mRNA of the host gene. Conversely, intergenic miRNAs are transcribed into their own 5′ m⁷G capped and 3′ polyadenylated pri-miRNA transcript like any other POLII transcripts (Cai, Hagedorn, & Cullen, 2004; Lee et al., 2004). Unfortunately, this tidy categorization did not hold over time as new discoveries were made. For example, multiple intragenic miRNAs are regulated by their own promoters and can also be expressed independent of their host gene (de Rie et al., 2017; Kim & Kim, 2007; Marsico et al., 2013). Moreover, intragenic miRNA on the antisense strand of the host gene can be produced through antisense transcription (Georgakilas et al., 2014). Intergenic pri-miRNA transcripts can also be coupled to a proximal host gene through either divergent transcription or transcription that carries on past the end of the gene (readthrough transcription) (Georgakilas et al., 2014).

From canonical primary miRNA to mature miRNA

Regardless of the source of pri-miRNA, a common characteristic shared by canonical pri-miRNA is a stemloop (hairpin) structure with a central mismatch embedded within the transcript. The single strands of the stemloop are what will ultimately make up the mature miRNA. The stemloop is cleaved from the pri-miRNA near the basal junction (the base/end of the stemloop structure) into the ~  70 nt in length precursor miRNA (pre-miRNA) (Denli et al., 2004). The cleavage process typically results in a 2 nt 3′ overhang, which plays a role in strand selection and pre-miRNA cleavage in subsequent biogenesis processes (Han et al., 2004; Tian et al., 2014). The cleavage event is accomplished through the ~  365 kDa microprocessor complex consisting of an RNA binding protein DiGeorge Syndrome Critical Region 8 (DGCR8) dimer (Senturia et al., 2010) and the ribonuclease III enzyme DROSHA (Denli et al., 2004). In addition, there are various motifs, structures, and RNA modifications associated with the pri-miRNA stemloop that promote pri-miRNA processing into pre-miRNA. This is mainly accomplished by promoting DROSHA or DGCR8 binding and/or correct positioning to facilitate efficient cleavage of the stemloop (Auyeung et al., 2013; Chaulk et al., 2011; Conrad et al., 2014; Fang & Bartel, 2015; Roden et al., 2017) (Fig. 1.1).

Fig. 1.1

Fig. 1.1 Canonical and noncanonical microRNA biogenesis. Like mRNA, most miRNA genes are transcribed by RNA polymerase II in the nucleus to produce long transcripts termed pri-miRNA. In the canonical pathway, the microprocessor complex, which comprises Drosha and DGCR8, recognizes target pri-miRNA and cleaves both strands of pri-miRNA near the base of the stemloop structure, generating a ~  70 nt hairpin RNA precursor termed pre-miRNA. Upon the processing event, pre-miRNAs are transported to the cytoplasm by the Exportin 5/Ran-GTP complex, where the final processing event occurs. In the cytoplasm, RNase III endonuclease Dicer recognizes and cleaves pre-miRNA near the apical loop, giving rise to a small miRNA duplex. Subsequently, either strand (-5p or -3p) of the miRNA duplex is preferentially loaded into an AGO protein to form an miRNA-induced silencing complex (miRISC). In addition to the canonical pathway, functional miRNAs can be produced from noncanonical pathways. Some of the examples of the noncanonical pathway include mirtrons, m7G-pre-miRNA, snoRNA, and shRNA. While the production of miRNA from mirtron, m7G-pre-miRNA, and snoRNA is independent of the DROSHA/DGCR8 complex, shRNA has been shown to be processed independently of DICER.

The pre-miRNA is then exported to the cytoplasm via the Exportin-5 (XPO5)/RanGTP complex, where it is processed by the RNase III endonuclease Dicer (Denli et al., 2004; Okada et al., 2009) in cooperation with double-stranded RNA binding protein (dsRBP) trans-activation-responsive RNA-binding protein (TRBP) (Chendrimada et al., 2005). The 2 nt 3′ overhang of the pre-miRNA initially binds to DICER; this interaction is stabilized with the help of TRBP bound to both the pre-miRNA and DICER (Tian et al., 2014; Wilson et al., 2015). The apical loop of the pre-miRNA is then cleaved, leaving the mature miRNA duplex bound to DICER (Lau et al., 2012; Zhang et al., 2004) (Fig. 1.1).

The final step in canonical miRNA biogenesis is the loading of one of the miRNA duplex strands into the Argonaute family of proteins (AGO1-4) to form the miRNA-induced silencing complex (miRISC). This is accomplished through the TRBP-stabilized interaction of AGO and DICER to form the RISC loading complex (Chendrimada et al., 2005; Fareh et al., 2016; Wang et al., 2009). The exact mechanism of loading is not fully understood, but the prevailing theory suggests that there is a direct handoff of the miRNA from DICER to AGO (Nakanishi, 2016). The strand that is loaded into AGO is termed the guide strand and the unloaded is called the passenger strand. Depending on the cellular environment, there can be a strand bias toward either the 5p or 3p strands, ranging from near 50:50 proportions to predominantly one or the other (Medley, Panzade, & Zinovyeva, 2020; Meijer, Smith, & Bushell, 2014). Two identified mechanisms that mediate the strand bias are a 5′ adenine or uracil at position 1 of the mature miRNA or the thermodynamic stability at the 5′ end of the miRNA (Frank, Sonenberg, & Nagar, 2010; Khvorova, Reynolds, & Jayasena, 2003). It is thought that the strand with the weakest pairing at the 5′ end of the miRNA duplex is preferentially loaded into AGO; however, many factors modulating 5p/3p abundance have been identified (Medley et al., 2020). The process of unwinding of the guide strand from the passenger strand is still unclear, but AGO may be directly involved (Kwak & Tomari, 2012; Park & Shin, 2015). The transition of AGO from a high-energy to a low-energy conformational state may provide the required force to unwind the strands (Gu et al., 2012; Sheu-Gruttadauria & MacRae, 2017). Regardless, the end result is AGO bound to a mature miRNA through stable interactions with both the 5′ and 3′ nucleotides of the miRNA (Frank et al., 2010; Ma, Ye, & Patel, 2004; Wang et al., 2008) (Fig. 1.1).

Noncanonical miRNA biogenesis and isomiRs

An intriguing aspect of miRNA biogenesis is its usage of RNA substrates that can arise from various disparate cellular processes rather than entirely miRNA-specific RNA substrates. The multiple sources of pri-miRNA are the first example of substrate dependence rather than miRNA-specific RNA sources; sets of differently sourced RNA transcripts can all act as substrates for nuclear DROSHA/DGCR8 and cytoplasmic DICER. By definition, all canonical miRNAs require both the microprocessor complex and DICER to complete their maturation. Noncanonical miRNAs are currently classified by their independence of either the microprocessor complex or DICER to complete their maturation (Fig. 1.1). Mirtrons, pre-miRNA produced from small introns as a result of splicing, resemble DICER substrates that are not processed by DROSHA and are thus classified as DROSHA-independent (Babiarz et al., 2008; Rorbach, Unold, & Konopka, 2018; Ruby, Jan, & Bartel, 2007). Mirtron pre-miRNAs are then exported to the cytoplasm and processed like canonical pre-miRNA. In addition, 7-methylguanosine (m⁷G)-capped pre-miRNA (Xie et al., 2013) and small nucleolar RNA (snoRNA) (Ender et al., 2008) can act as DICER substrates without processing by DROSHA. m⁷G-capped pre-miRNA are shuttled to the cytoplasm by exportin 1 (XPO1) and cleaved by DICER. A strong 3p strand bias has been observed in this case, likely due to the cap interfering with the 5′ nucleotide AGO-binding of the 5p strand (Xie et al., 2013). The DICER-independent endogenous short hairpin RNA (shRNA) are another identified source of noncanonical miRNA (Yang et al., 2010). These transcripts are cleaved into pre-miRNA by the microprocessor and transported to the cytoplasm by XPO5 but are too small (dsRNA length of ~  18 bp) to act as DICER substrates (Yang et al., 2010). Instead, the entire miRNA duplex is loaded into AGO2 and undergoes AGO2-dependent slicing of the 3p strand followed by 3′-5′ trimming of the remaining 3p fragment to complete AGO2 loading (Cheloufi et al., 2010).

The positioning of the microprocessor complex on pri-miRNA and DICER on pre-miRNA is critical to the final mature miRNA sequence (Neilsen, Goodall, & Bracken, 2012). Differential positioning of these nucleases can lead to pri/pre-miRNAs that undergo alternative cleavage, producing 5′ and 3′ sequence variants called isomiRs (Vermeulen et al., 2005; Wu et al., 2009). Variations to the 5′ seed region have been shown to affect miRNA targeting (Manzano et al., 2015), whereas 3′ variations typically affect miRNA abundance and activity (Yu et al., 2017a). In some cases, single miRNA genes can lead to dozens of expressed 5′/3′ isomiRs (Parafioriti et al., 2020). Nontemplated nucleotide addition (NTA) of adenine or uracil at 3′ ends has been shown to directly affect miRNA stability (Neilsen et al., 2012). Terminal nucleotidyltransferase 2 (TENT2, also known as GLD2) has been shown to catalyze 3′ adenylation and promote miRNA stability (D'Ambrogio et al., 2012; Katoh et al., 2009), although destabilization has also been observed (Boele et al., 2014). Uridylation of 3′ miRNA termina by multiple terminal uridylyltransferase (TUT) family members has been associated with reduced activity and stability (Gutierrez-Vazquez et al., 2017; Jones et al., 2009; Kim et al., 2015). Internal variations in miRNA sequence are termed polymorphic isomiRs. These miRNAs undergo RNA editing, typically adenosine to inosine via adenosine deaminase RNA specific (ADAR) (Nishikura, 2016) or less frequently cytidine to uracil by apolipoprotein B mRNA editing enzyme catalytic (APOBEC) family members (Salter, Bennett, & Smith, 2016). Depending on the position of sequence variation, RNA-edited miRNA can show differences in RNA targets (Kawahara et al., 2007) or miRNA abundance (Kawahara et al., 2008; Vesely et al., 2014).

Nomenclature and classification

Given the relatively complex process of miRNA biogenesis, various naming conventions have been developed to aid in inferring the source of the miRNA, stage of biogenesis, and which arm of the pri-miRNA the mature miRNA is embedded within (Budak et al., 2016). Moreover, various other miRNA features, such as sequence homology, have been used in naming groups of miRNA genes to infer a relation between them (Seal et al., 2020). We will discuss in the next section why sequence homology is a defining feature of the mechanism of miRNA function and why this would infer a relation between miRNAs. It should be noted that naming conventions are still evolving as our understanding of miRNA biology is incomplete.

As discussed previously, miRNA biogenesis involves the processing of primary miRNAs to precursor and finally to mature miRNA. The primary miRNA is transcribed either with a host gene or from its own gene. Regardless if the miRNA is intra- or intergenic, each miRNA is given a gene symbol. In humans, miRNA is shortened to MIR and a number is assigned to it, such as MIR25, while in mice only the M is capitalized, such as Mir25. The number is assigned by chronological order of discovery. There are exceptions, like with the let family of miRNA, but this is an example of a historical naming convention.

The pri/pre-miRNA adopts a different naming structure. Using MIR25 as an example, it would be written hsa-mir-25 for humans or mmu-mir-25 for mice. The species should always be added to the miRNA name within the abstract, the first time it is used in the text, or as needed to avoid confusion if the species cannot be inferred. This will be ever more important in the future as more computer-based natural language processing is applied to miRNA studies. Unfortunately, this naming convention is not well followed.

When referring to the mature, single-stranded miRNA, we capitalize the R, hsa-miR-25, and identify which arm of the pre-miRNA the mature miRNA was embedded within, hsa-miR-25-5p or hsa-miR-25-3p (Fig. 1.2). A previous naming practice employed the use of the asterisk (*) to denote a passenger strand that was predominantly degraded in place of the strand identifier (i.e., 5p or 3p) but this has been largely deprecated. A common example was with the hypoxamir (an miRNA that is induced by hypoxia) hsa-miR-210 and its degraded passenger strand hsa-miR-210* (Huang, Le, & Giaccia, 2010). But this raises the question, which exact miRNA are we referring to when we say miR-210? Without the RNA sequence or a priori knowledge, it would remain ambiguous. As such, we should never refer to a mature miRNA without first identifying the arm it was embedded within, like with the hypoxamir miR-210-3p. It is also common practice to drop the strand identifier when referring to the 5p but for the same reasons above, this should be avoided. Evolving on past nomenclature, when referring to a common and predominant strand bias, the addition of the asterisk can be used in conjunction with the strand identifier (Budak et al., 2016). In the case of miR-210, we could refer to the 5p strand as miR-210-5p* as it is predominantly the passenger strand. Likewise, miRNA* would refer to the mature miRNAs that are predominantly passenger strands.

Fig. 1.2

Fig. 1.2 Structure and naming of miRNA. After transcription of the primary (pri-) miRNA, the pri-miRNA is cleaved by nuclear DROSHA ( red arrows ) to produce the precursor (pre-) miRNA. Cleavage by cytoplasmic DICER ( blue arrows ) produces the miRNA duplex. Either the 5p or 3p mature strand is then loaded into the AGO family of proteins, blue and red strands, respectively. MIR25 pri/pre miRNA are labeled as mir-25 whereas mature miRNAs include the strand identifier (5p or 3p) and a capital R, miR-25-5p.

In some cases, multiple proximal miRNA genes are transcribed within a single RNA molecule. These miRNAs are referred to as clusters (Desvignes et al., 2015). A well-studied example is the mir-19/72 cluster (MIR17HG, HG meaning host gene), consisting of members mir-17, mir-18a, mir-19a, mir-20a, mir-19b-1, mir-92a-1, also known as oncomiR-1 (Mogilyansky & Rigoutsos, 2013), with oncomiR being the miRNA equivalent of an oncogene. Cluster members also tend to share similar expression profiles but depending on the cellular context, can be differentially expressed (Fang et al., 2017). In some cases, cluster members may also cooperate in the regulation of similar pathways within the cell, as with the mir-17/92 cluster (Mogilyansky & Rigoutsos, 2013) (Fig. 1.3A).

Fig. 1.3

Fig. 1.3 MicroRNA clusters and families. (A) MicroRNA clusters are single RNA molecules containing multiple precursor miRNA, such as with the mir-19/72 cluster. (B) MicroRNA families are classified by shared sequence identities; similar colors indicate 100% sequence identity. MicroRNA genes that share at least one identical mature sequence have a numerical suffix added to the miRNA gene symbol ( MIR218-1 / MIR218-2 ). In this example, the 5p strands of miR-218-1/2 share 100% sequence identity. In addition, because both miR-218-1/2-5p strands are identical, the family suffix can be dropped when referring to the mature miRNA, that is, miR-218-5p refers to both miR-218-1-5p and miR-218-2-5p. However, the mature 3p strands are not identical and must then be referred to as miR-218-1-3p and miR-218-2-3p.

Depending on sequence homology, miRNAs can also be grouped into families (Desvignes et al., 2015). MicroRNA genes that share a 100% sequence identity between mature miRNA sequences have an added incrementing number suffix while maintaining the same gene name. For example, hsa-mir-218-1 (chr4: MIR218-1) and hsa-mir-218-2 (chr5: MIR218-2) contain a 5p strand with identical sequences (Fig. 1.3B). Because the mature miRNAs share a 100% sequence identity, we can refer to the pool of miR-218-1-5p and miR-218-2-5p within the cell as just miR-218-5p. Likewise, exogenous sources of miRNA mimics for the 5p strand would also be referred to as miR-218-5p mimics. The 3p strands of MIR218-1/-2 are not identical and thus, they are referred to as miR-218-1-3p and miR-218-2-3p. Lastly, miRNAs that share a sequence homology between positions 2–7 but do not share a 100% sequence identity are given an added incrementing letter suffix instead to denote their membership in an miRNA family. For example, hsa-miR-19a-5p (5′ AGUUUUGCAUAGUUGCACUACA) and hsa-miR-19b-1-5p (5′ AGUUUUGCAGGUUUGCAUCCAGC) share a sequence homology only at positions 1–9 and 13–17 (Fig. 1.3A). It should be noted that miRNA family classification when dealing with evolutionary relation is more complex (Zou et al., 2014) but the nomenclature is the same.

Mechanisms of miRNA-mediated gene regulation

Posttranscriptional regulation of mRNA stability and translation

Once the mature miRNA is loaded into miRISC, this forms a minimally competent complex in which the miRNA acts as a guide to facilitate mRNA target recognition while AGO acts as a mediator of the regulatory potential of the complex (Huntzinger & Izaurralde, 2011; Ipsaro & Joshua-Tor, 2015; Kawamata & Tomari, 2010). Target specificity is a result of sequence complementarity between the miRNA and mRNA target (Jo et al., 2015). The site of interaction on target mRNA is called the miRNA response element (MRE) (Nair et al., 2020). Generally, nucleotides 2–8 of the miRNA are sufficient in conveying robust target specificity (Bartel, 2009; Ellwanger et al., 2011; Xu et al., 2014). Given its importance, it was termed the 5′ seed region of the miRNA and was sufficient for the in silico prediction of many MRE (Quillet et al., 2020). Additional pairing throughout the remaining 3′ region of the miRNA has been shown to increase specificity and stability (Sheu-Gruttadauria et al., 2019a). This would be expected given our current understanding of the miRNA:MRE interaction dynamic. When a miRNA is bound to AGO, the seed region of the miRNA is oriented such that nucleotides 2–4/5 can rapidly bind complementary RNA (Chandradoss et al., 2015; Salomon et al., 2015; Schirle, Sheu-Gruttadauria, & MacRae, 2014). This interaction is metastable and miRISC will quickly dissociate if the remaining seed region nucleotides are not complementary and cannot stabilize the interaction (Chandradoss et al., 2015). This promotes efficient scanning of potential target sites and more stable, long-lived interactions with true MREs (Chandradoss et al., 2015).

miRISC associations with target MREs are indiscriminate, and have the potential to interact with an MRE anywhere within the RNA molecule (Plotnikova, Baranova, & Skoblov, 2019; Xu et al., 2014). This assumes that the MRE is available to interact with miRISC. However, this does not mean that any given MRE will have equivalent regulatory potential. MRE that are complementary to a 5′ seed region of 7 nucleotides and contain an adenine opposite to position 1 (termed t1A) of the miRNA (8-mer) are generally the highest affinity MRE (Agarwal et al., 2015) where 7/6-mer MRE, lacking the position 1 adenine or with mismatches to the seed region, have a lower affinity and as a result, the miRISC:MRE interaction is less stable (Agarwal et al., 2015; Friedman et al., 2009; Nielsen et al., 2007). RNA secondary structures and RNA-binding proteins can inhibit or promote miRISC interactions to potential MREs (Kedde et al., 2010; Min et al., 2017; Nam et al., 2014). Typically, MREs within the 3′ untranslated region (UTR) of target mRNA lead to translation initiation inhibition and mRNA decay, thereby resulting in a decrease in mRNA and protein levels of the target genes (Brkic et al., 2018; Huntzinger & Izaurralde, 2011; Ipsaro & Joshua-Tor, 2015). Functional MREs within the 5′ UTR and CDS have also been reported but they are uncommon (Forman, Legesse-Miller, & Coller, 2008; Zhang et al., 2018a). The current model suggests that translating ribosomes actively lead to the efficient dissociation of miRISC, preventing miRISC activity (Gu et al., 2009). However, high affinity 3’ miRNA:MRE interactions within the CDS have been shown to induce ribosome stalling and mRNA degradation hypothesized via a nonstop decay mechanism (Zhang et al., 2018b).

Of the four human AGO family members (AGO1-4), only AGO2 is known to have endonuclease activity. The degree of complementarity between miRNA and MRE determines if target mRNA can be cleaved. A fully complementary miRNA:MRE interaction induces AGO2 endonuclease activity (Fig. 1.4A) while any central mismatches (positions ~  9–12) in this interaction prevent mRNA cleavage (Jo et al., 2015). The overwhelming majority of identified human MREs contain central mismatches with their cognate miRNA and as such, mRNA cleavage in humans is generally rare (Hansen et al., 2011; Jonas & Izaurralde, 2015). For clarification, cognate miRNAs are the set of miRNAs that are capable of interacting with a given MRE, whereas cognate MRE are the set of MRE capable of interacting with a given miRNA.

Fig. 1.4

Fig. 1.4 Cytoplasmic and nuclear functions of miRNA. The repressive effect of miRNA at the posttranscriptional level is facilitated through the promotion of either mRNA cleavage or mRNA decay and repression of translation initiation. The sequence of the loaded miRNA gives the miRISC complex target specificity via complementary miRNA:MRE interaction. While perfect complementarity of miRNA and MRE of the target mRNA results in mRNA cleavage (A), a majority of miRNA:MRE interactions have at least a central or 3′ mismatch in mammalian cells. Most miRISC binding in mammalian cells leads to translation initiation inhibition and mRNA decay through the recruitment of other factors (B). Mechanically, translation initiation inhibition is a consequence of eukaryotic translation initiation factor 4A1 (EIF4A1)/EIF4A2 dissociation. In addition, the binding of protein TNRC6A to AGO of the miRISC complex provides a binding platform for poly(A) binding proteins, including PAN2/3 and CCR4/NOT deadenylase complexes, which initiates and completes poly-A tail deadenylation of the target mRNA, respectively. The target mRNA is further decapped by decapping protein 1/2 (DCP1/2) complexes, and then degraded by 5′–3′ exoribonuclease 1 (XRN1). In addition, miRISC can be transported into the nucleus via the Importin 8/RAN-GTP complex and functions in regulating transcription at the target loci (C).

Most miRNA:mRNA interactions lead to target repression (Salem et al., 2018; Valinezhad Orang, Safaralizadeh, & Kazemzadeh-Bavili, 2014; Wang et al., 2019). In order for miRISC to affect mRNA stability without directly cleaving the target mRNA, other factors are recruited to miRISC to inhibit translation initiation and/or to promote mRNA decay (Fig. 1.4B). Inhibition of translation initiation has been shown to involve the dissociation of eukaryotic translation initiation factor 4A1 (EIF4A1)/EIF4A2, preventing formation of the EIF4F translation initiation complex (Behm-Ansmant et al., 2006; Fukao et al., 2014; Meijer et al., 2013). The current model suggests that mRNA decay is accomplished by localizing the poly(A) tail of mRNA to miRISC, followed by the recruitment of multiple deadenylases that can ultimately promote degradation of the mRNA (Christie et al., 2013; Jonas & Izaurralde, 2015). The crucial adapter protein that facilitates poly(A) localization to miRISC is the trinucleotide repeat containing adaptor 6A (TNRC6A), also called GW182 (Liu et al., 2005a). The N-terminal domain of TNRC6A interacts with AGO while the C-terminal domain interacts directly with poly(A)-binding protein C (PABPC) (Huntzinger et al., 2013). The C-terminal also recruits poly(A)-nuclease deadenylation complex subunit 2 (PAN2)/PAN3 and the carbon catabolite repressor protein 4 (CCR4):NOT complex (Braun et al., 2011; Chen et al., 2009). PAN2/3 initiates the deadenylation of target mRNA and CCR4:NOT completes the reaction (Jonas & Izaurralde, 2015). Deadenylation promotes the decapping of target mRNA by the mRNA-decapping enzyme subunit 1 (DCP1)/DCP2 complex (Behm-Ansmant et al., 2006). Decapped mRNA can then be degraded by 5′–3′ exoribonuclease 1 (XRN1) (Fig. 1.4B) (Braun et al., 2012).

Models of miRNA-mediated gene regulation within the nucleus

The role of miRNAs within the nucleus is an emerging area of study. MicroRNAs have been detected in the nucleus and several potential functions of miRNAs in the nucleus have been suggested. Perhaps the most striking is their potential to interact with DNA and regulate transcription. However, the nucleus also contains large amounts of RNA ranging from pre-mRNA to ribosomal RNA, raising the possibility that miRNAs may interact with these RNA molecules to regulate their levels.

Depending on cell type, the ratio of nuclear to cytoplasmic miRNA varies, but a significant population of nuclear miRNA has been observed in many cell types ranging from embryonic to neural cells in mice and humans (Catalanotto, Cogoni, & Zardo, 2016; Fan et al., 2020; Sarshad et al., 2018; Turunen et al., 2019; Zhang, Shen, & Cui, 2019). Changes in cell state have been shown to affect miRNA distributions within the cell with senescence increasing nuclear miRNA abundance (Benhamed et al., 2012) and serum starvation depleting it (Kucherenko & Shcherbata, 2018). These large-scale changes in miRNA localization tend to involve interactions between miRISC and cellular proteins whose localization is affected by the change in cell state. For example, unliganded estrogen receptor β (ERβ) led to the nuclear shuttling of ERβ:miRISC complexes (Tarallo et al., 2017) and the inactivation of canonical Wnt signaling increased the nuclear localization of miRNA (Di Mauro et al., 2019). In the case of hypoxia, the differential expression of miRNA was observed between the cytoplasm and nucleus, notably leading to significant increases of hsa-miR-210-3p within the nucleus during hypoxia (Turunen et al., 2019).

Multiple studies have attempted to identify various miRNA sequence motifs regulating steady-state nuclear or cytoplasmic localization, but these have been met with conflicting results. The current model involves a constant two-way shuttling of free miRISC:TNRC6A complexes (miRISC not bound to target RNA) between the nucleus (importin-8; IPO8) and cytoplasm (XPO1) (Nishi et al., 2013; Nishi et al., 2015; Pitchiaya et al., 2017). Instead of a sequence motif mediating localization (Hwang, Wentzel, & Mendell, 2007), the abundance of target RNA within the nucleus is likely to be the most significant miRNA-specific factor modulating miRISC retention within the nucleus, that is, an increase in target RNA within the nucleus is sufficient to mediate cognate miRNA nuclear retention. In the nucleus, free-miRISC diffuses more quickly than bound-miRISC, and as such, miRISC binding to more nuclear RNA targets shifts the steady-state distribution toward the nucleus (Pitchiaya et al., 2017). The current model may ultimately explain the previous conflict of trying to identify a nuclear localization signal (NLS) within miRNA. An enrichment of MRE within the nucleus can promote the increased retention of cognate miRNA. This results in an increase of miRNA within the nucleus containing the sequence complementary to the enriched MRE. Consequently, when applying statistical tools to identify overrepresented nuclear miRNA sequence motifs, such a sequence may be found. But the error then may lie in the assumption that the sequence is an NLS. This may explain why predicted miRNA NLS motifs may not promote nuclear localization in different cell types or why miRNA with similar seed regions tend to localize to the same subcellular compartments (Jeffries, Fried, & Perkins, 2011; Khudayberdiev et al., 2013).

In theory, mature miRNA alone should be insufficient to mediate any of its gene regulatory potential as AGO recruits effector proteins such as TNRC6A and CCR4:NOT. Within the nucleus, miRISC has been found in at least two major forms, a low molecular weight complex of just miRNA and AGO and a high molecular weight complex of miRNA, AGO, TNRC6A, and multiple poly(A) deadenylases (Sarshad et al., 2018). The degree to which either the low or high molecular weight complex is found within the nucleus may depend on cell types in which more stem-like cells contain higher proportions of high molecular weight miRISC. As a result, cell types with the high molecular weight miRISC have been shown to facilitate mRNA decay within the nucleus (Sarshad et al., 2018). Moreover, MRE within the CDS of target mRNA have been shown to elicit mRNA decay (Sarshad et al., 2018), in contrast to cytoplasmic targets, which are predominantly located within the 3′ UTR. Nuclear miRNAs have also been reported to affect alternative splicing (Allo et al., 2009) and target rRNA (Wang et al., 2015).

Nuclear miRISC has also been shown to regulate transcriptional activity both positively and negatively (Benhamed et al., 2012; Dharap et al., 2013; Miao et al., 2016; Nishi et al., 2013; Pitchiaya et al., 2017; Truesdell et al., 2012) (Fig. 1.4C). Genome-wide studies have found that miRISC tends to associate with actively transcribing loci (Xun et al., 2020) and can alter histone modifications to promote or inhibit euchromatin formation (Benhamed et al., 2012; Roberts, 2014). However, the way in which miRISC is known to interact with chromatin is still incomplete (Stavast & Erkeland, 2019). The most plausible model of interaction involves miRISC identifying MREs in nascent RNA cotranscriptionally, enriching chromatin modifiers to sites of active transcription by direct interaction with AGO and adapter proteins (Matsui et al., 2013; Pisignano et al., 2017; Schwartz et al., 2008). Nevertheless, there is growing evidence that miRISC may also interact directly with DNA through Watson-Crick base pairing (Younger, Pertsemlidis, & Corey, 2009) and triplex forming Hoogsteen base pairing (Paugh et al., 2016). The deletion of DNA-associated MRE, determined by sequence complementarity, can reduce the miRNA regulatory effect (Miao et al., 2016). Beyond enriching chromatin remodelers at miRNA target sites on the DNA, miRISC was also shown to sterically hinder the binding of transcription factors, leading to decreased transcription (Miao et al., 2016). Taken together, miRISC may function similarly to transcriptional activators and suppressors, guided to target loci by miRNA:DNA and miRNA:RNA sequence complementarity.

The dynamics of miRNA-mediated gene regulation

Perhaps one of the most exciting shifts in biology over recent decades has been the increased resolution and granularity with which we are able to observe the cell (de Rie et al., 2017; Yan et al., 2018; Zheng et al., 2017). This experimental and technical innovation has also correlated with a significant increase in the amount of information we are able to collect, and as such, the inferences we can draw can be substantially more accurate, encompassing, and complex. MicroRNA research has of course benefited from these advances and we are now capable of observing changes in entire populations of RNA molecules, including their spatiotemporal interactions and localizations. Spatio- refers to the physical locations within two-dimensional (2D) and/or three-dimensional (3D) space and temporal referring to time and changes over time.

The modeling of nonlinear dynamics in complex systems, such as with miRNA biology, is by its nature computationally complex (Walpole, Papin, & Peirce, 2013). In cases where miRNAs are expressed at unphysiologically high levels, a simple systems approach can be sufficient. That is, if we have a target mRNA with a predicted MRE and if the prediction is accurate, overexpression of the cognate miRNA will typically downregulate the target. But this can be a somewhat artificial set-up (Jin et al., 2015), analogous to naively overexpressing a constitutively active kinase, and misses the complexities of the system. Excluding the spatiotemporal regulation of miRNA and cognate mRNA, a key factor in the overall output of the system is the concept of MRE load and affinity (Denzler et al., 2016). Any given miRNA will typically have many interactable MRE distributed among numerous RNA molecules, that is, one miRNA can target many mRNAs (Baek et al., 2008). Reciprocally, a single mRNA molecule can be targeted by multiple miRNAs (Chou et al., 2018), in some cases synergistically (Denzler et al., 2016; Elkayam et al., 2017; Lemus-Diaz et al., 2017; Li et al., 2014; Pons-Espinal et al., 2017; Wu et al., 2010; Yao et al., 2014). Moreover, MREs for a single miRNA can vary by sequence, such as ACACGAA vs. ACACGAU, leading to cognate MRE with different affinities (Agarwal et al., 2015; Denzler et al., 2016). As affinities are highly important in determining more stable, long-lived interactions between miRISC and target mRNA, affinities can directly modulate the occupancy of any given MRE (Bosson, Zamudio, & Sharp, 2014). In addition, mRNA 3’ UTR isoforms (Nam et al., 2014) and RNA binding proteins can all alter the available pool of interactable MRE (Bottini et al., 2017; Kedde et al., 2010).

At steady state, assuming miRNA and MRE are concomitantly and homogeneously spatiotemporally regulated, cognate miRNA would be distributed such that MRE with the highest affinity would have the highest relative miRNA occupancy, that is, the cognate MRE with the highest affinity toward a cognate miRNA are, at any given time, most likely to be bound to that miRNA (Bosson et al., 2014; Denzler et al., 2016). Depending on the different affinities between cognate MRE, and the cognate miRNA is not in excess, this can drastically affect MRE occupancy such that low-affinity MRE can be predominantly unbound even though low-affinity cognate MRE are typically much more abundant relative to high-affinity MRE (Bosson et al., 2014). As such, higher-affinity MRE typically show the greatest sensitivity to miRISC regulation (Baek et al., 2008; Denzler et al., 2016; Molotski & Soen, 2012). Combined with relatively short interaction times (low affinity ≈ 1 s and high affinity ≫ 5 min (Chandradoss et al., 2015)), miRISC can efficiently sample many targets and respond quickly to stochastic fluctuations of transcription or dynamic changes in gene expression to change their MRE distribution profiles (Wee et al., 2012).

The distribution of miRNA among its MRE pool is at the heart of the competitive endogenous RNA (ceRNA) model of gene regulation. The primary aspect of the theory proposes that a sufficient increase of cognate MRE load within the cell will decrease the occupancy of any given MRE enough to result in a derepression of the target genes, assuming the miRNA was acting to downregulate its target genes (Bosson et al., 2014; Kartha & Subramanian, 2014) (Fig. 1.5). Typically, this mechanism is attributed to the upregulation of an mRNA, long noncoding RNA, circular RNA, or pseudogene that is the source of the MRE (Barrett & Salzman, 2016; Liu et al., 2019; Militello et al., 2017; Tay, Rinn, & Pandolfi, 2014; Yu et al., 2017b). RNA molecules with repeated high-affinity MRE are termed miRNA sponges and can be potent ceRNA (Militello et al., 2017). This theory, however, has been viewed with controversy due to the potentially unphysiological levels of ceRNA required for the derepression of miRNA targets. Mathematical models have suggested that at least 50% of total cognate MRE from a single ceRNA would be needed in order to sequester enough miRNAs from other targets to promote their derepression (Denzler et al., 2016). Nevertheless, to suggest whether an RNA molecule is acting as a ceRNA depends on the ratio of miRNA to MRE and the overall abundance of miRNA within the cell (Bosson et al., 2014; Liu et al., 2019). MicroRNA that are in great excess relative to their MRE pool (a high miRNA:MRE ratio) and have high copy numbers per cell (in the tens of thousands) tend to be robust to ceRNA derepression. However, as the miRNA copy number and miRNA:MRE ratio decrease, the miRNA can be more susceptible to ceRNA derepression. Although single genes acting as potent ceRNAs may be relatively infrequent, most reported miRNA:MRE ratios tend to be much less than one (Bosson et al., 2014), suggesting a potential wide-spread steady-state ceRNA crosstalk involving complex gene regulatory networks.

Fig. 1.5

Fig. 1.5 Competitive endogenous RNA (ceRNA) model. The ceRNA model predicts that the significant overexpression of RNAs containing high-affinity MRE results in the sequestering of miRNA from other endogenous targets leading to target derepression. When miRNA abundance is in excess to their MRE load (a high miRNA:MRE ratio; top left), the upregulation of ceRNA (shown as steel-blue RNA molecules) would not produce any appreciable derepression of endogenous targets (bottom left) as endogenous and ceRNA MRE are both maximally occupied. As miRNA:MRE ratios and miRNA abundance decrease, miRNA targets become more susceptible to ceRNA derepression. Before ceRNA upregulation, miRNA targets are still efficiently regulated (top right); however, an increase in ceRNA high-affinity MREs can sequester a limited pool of miRNA away from other targets, leading to target derepression (bottom right).

Beyond miRNA and MRE abundance and differential MRE affinities, the subcellular localization of miRISC is integral to its regulatory potential, both in terms of control and efficiency. As was discussed earlier, the nuclear localization of miRISC is necessary to regulate nuclear RNA. This process is at least in part dependent on nuclear cognate RNA abundance where increased target abundance increases nuclear miRNA retention. Outside the nucleus, miRISC has been detected within the cytosol, microtubule networks, processing (P)-bodies, stress granules, lysosomes, rough endoplasmic reticulum (rER), polysomes, Golgi apparatus, endosomes, and mitochondria. miRISC can be transported along microtubule networks within the cytosol, which can lead to the nucleation of P-bodies enriched with miRNA, cognate mRNA, deadenylases, and ribonucleases, promoting the efficient degradation of target mRNA (Carbonaro, O'Brate, & Giannakakou, 2011; Eulalio et al., 2007; Liu et al., 2005b; Rajgor et al., 2014). In response to cell stress, miRNA and target mRNA can be exported from the cell or shuttled to stress granules that lack miRISC effector proteins and reduce miRNA posttranscriptional regulation (Detzer et al., 2011; Kucherenko & Shcherbata, 2018; Wang et al., 2010). Stress granules are transient and their components, including miRISC, can return to the cytosol and in some cases be degraded within lysosomes (Leung, Calabrese, & Sharp, 2006; Protter & Parker, 2016; Rajgor et al., 2014). Localization of miRISC:polysome complexes to the rER has been associated with increased silencing of targets, where MRE within the mRNA of polysomes were observed to reach up to ~  50% occupancy compared to 5%–20% for a typical MRE (Barman & Bhattacharyya, 2015; Bosson et al., 2014; Molotski & Soen, 2012; Pillai et al., 2005). miRISC within the Golgi and endosomes have been associated with miRISC activity and either miRNA export from the cell or recycling back into the cytosol (Bose et al., 2017). Endosomes are also important players in miRNA import from extracellular fluids, potentially regulating which imported miRNA are cycled back into the cytosol or deposited into lysosomes (Bose et al., 2017; Gibbings et al., 2009). Within the mitochondria, miRISC has been shown to promote translation (Barrey et al., 2011; Zhang et al., 2014).

MicroRNA abundance and turnover

The abundance of miRNA within the cell is dependent on at least the cell state and cell type (Trapnell, 2015), with state referring to the current set of programs (gene expression networks) running within the cell, and cell type referring to stable morphological and/or phenotypically distinct populations of cells. Put another way, any given cell type can run many different cell states and the set of potential cell states is limited by the cell type.

There are substantial differences in miRNA expression profiles within and between different cell types. A large-scale study by the Functional Annotation of the Mammalian Genome (FANTOM5) consortium found that for any given cell type, the top five expressed miRNA represent on average ~  50% of the total miRNA pool within the cell (de Rie et al., 2017). Of the total set of expressed human miRNA, half were cell type enriched, a quarter broadly expressed, and the remaining identified miRNA were expressed at low levels regardless of cell type.

Generally, miRNA abundance is strongly correlated with pri-miRNA expression (Kingston & Bartel, 2019). Moreover, the expression rate of an miRNA is considered high if an individual miRNA molecule can be generated at >  100 molecules per minute per cell (Kingston & Bartel, 2019). The overall lifespan of miRNA is also long lived; in some cases, half-lives of more than 100 h have been observed. Depending on the cell type, the median half-life can vary drastically, ranging from less than 10 h to more than 30 h. This is in contrast to median mRNA half-life, which is commonly observed between 2 and 10 h (Kingston & Bartel, 2019). These differences in expression rate and half-life result in typically more than 1000 molecules per miRNA per cell vs. ~  18 molecules per mRNA per cell (Denzler et al., 2016; Schwanhausser et al., 2011). Although in some cases, more than 100,000 molecules per miRNA per cell have been observed (Song et al., 2017).

Various mechanisms of specific miRNA turnover have been identified, whereas factors affecting global miRNA turnover rates still remain poorly understood (Gebert & MacRae, 2019). Global turnover rates may correlate with the speed at which cells need to remodel their gene expression. This is exemplified in mouse embryonic stem cells (mESC) in which median global miRNA turnover rates are less than 6 h versus >  20 h in other differentiated cell types (Kingston & Bartel, 2019). Long miRNA half-lives would in theory act to help maintain the cell type/state, rendering the system more robust to perturbations and changes to cell state (Kingston & Bartel, 2019; Siciliano et al., 2013).

Although global miRNA turnover rates may be more stable, individual miRNAs are susceptible to target RNA-directed miRNA degradation (TDMD), a consequence of miRNA interacting with MRE that have significant pairing in the 3′ region of the miRNA (Ameres et al., 2010; Guo et al., 2015; Sheu-Gruttadauria et al., 2019b). This pairing is hypothesized to dissociate the 3′ end of the miRNA from AGO, which allows 3’ NTA (tailing) and trimming followed by 3′–5′ exonuclease degradation (Ameres et al., 2010). However, what factors are necessary to induce TDMD are uncertain as it is difficult to predict (Haas et al., 2016). Regardless, the 5′ seed region may be an important factor in TDMD, even though it is not always necessary (Park, Shin, & Shin, 2017), as it stabilizes miRISC association with cognate MRE, which then allows more stable 3′ pairing to occur. It can also be the case that 5′ seed interactions with MRE stabilize cognate miRNA levels (Pitchiaya et al., 2017). Nevertheless, the miRNA sequence has been shown to be critical in determining miRNA stability (Guo et al., 2015). For example, miRNA:MRE interactions with 100% complementarity generally lead to unloading of the miRNA from AGO and promote the miRNA’s degradation (Ameres et al., 2010; De et al., 2013). This is the mechanism utilized to achieve exogenous antisense miRNA knockdown (Krutzfeldt et al., 2005).

Concluding remarks

The topic of miRNA is an immensely rewarding avenue of research, which is in part due to the strongly dynamical and nonlinear nature of the complex miRNA system. A nonlinear system is one where the sum of the inputs does not necessarily equal the sum of its outputs, in contrast to a linear response where a change in A equates to a proportional change in B. A dynamical system is one that evolves (changes) over time. And lastly, a systems approach takes into consideration a network of connections that describes how each node relates to all the other nodes in the network. In biology, the node could represent a molecule, a set of molecules, a single gene, a protein, a cellular function, or even an entire cell or tissue. The key idea is that when thinking about miRNA or any biological system, keeping in mind this complex interplay of connections is crucial. By thinking of miRNA biology through this framework, it will offer an understanding that cannot be inferred by a more linear approach alone.

MicroRNA, a subset of small noncoding RNA, arise from various RNA sources but the end result is the loading of mature single-stranded miRNA into the Argonaute family of proteins forming the miRNA-induced silencing complex. This complex is shuttled within the cell between the cytoplasm and nucleus, existing both diffusely within these compartments and localized to specific subcellular regions. The recognition of miRNA response elements by the 5′ seed region of the miRNA (nucleotides 2–8) and supplemental 3′ miRNA complementarity within target RNA and/or DNA are sufficient to elicit miRISC regulatory potential. Depending on the effector proteins complexed with miRISC and the subcellular localization, translation initiation inhibition, ribosome stalling, poly(A) deadenylation and mRNA decay, translation activation, and transcriptional suppression and promotion have all been observed. Although miRNAs are most often observed to negatively correlate with cognate mRNA levels, our current understanding of the mechanistic and regulatory potentials of miRNA has so greatly expanded over the last 20 years that miRISC has become insufficiently named to encompass the wide breadth of roles it plays within the cell.

The spatiotemporal steady-state levels of miRNA globally and relative to their mRNA targets are of crucial importance when trying to understand the dynamics of miRNA:mRNA interactions and their regulatory potential. In a system where the number of miRNA molecules is in a 1:1 proportion to target mRNA, what effect would that have on target mRNA stability? Is it sufficient to promote mRNA decay? What are the interaction dynamics like at that ratio? At any given time, would each target mRNA molecule be bound by at least one cognate miRNA? In other words, would there be a 100% occupancy of target mRNA by miRISC at any given time? Does miRNA interaction with MRE have any influence on the miRNA itself? What if cognate miRNA and mRNA are independently spatially regulated, what effect would that have on interaction dynamics and the models we generate to predict those effects? Although we have no definitive answer and no single model to fully explain miRNA biology, there has been truly impressive work in these areas that has greatly impacted and expanded our view of miRNA biology.

There may be hundreds of miRNAs expressed at any given time within any given cells, but only a few miRNAs are highly expressed in any given cell types. These highly expressed miRNA tend to exist in great excess compared to their MRE load and function to broadly suppress the majority of their targets by maximally occupying cognate MRE and as a result, are robust to changes competitive endogenous RNA expression. However, when miRNAs are expressed at lower absolute abundances and are not in excess relative to their MRE load, these miRNA regulatory networks are more susceptible to changes in MRE load. The combination of low miRNA abundance and less than or equal miRNA:MRE proportions leads to quickly changing MRE occupancy distributions following the upregulation of high-affinity targets, which may result in the derepression of previously occupied MRE.

Taken together, miRNAs have evolved to dynamically and efficiently regulate a wide range of targets to maintain a steady-state equilibration of stochastic gene expression networks in an ever-changing environment. By considering global populations of gene expression agents, miRNA being of particular noteworthiness, the study of disease-perturbed and normal gene expression networks alike has led to models with ever greater fidelity, aiding researchers around the world to more completely and accurately understand the complex biological system that is our existence.

References

Agarwal V. Predicting effective microRNA target sites in mammalian mRNAs. eLife. 2015;4:e05005.

Allo M. Control of alternative splicing through siRNA-mediated transcriptional gene silencing. Nature Structural & Molecular Biology. 2009;16(7):717–U43.

Ameres S.L. Target RNA-directed trimming and tailing of small silencing RNAs. Science. 2010;328(5985):1534–1539.

Auyeung V.C. Beyond secondary structure: Primary-sequence determinants license Pri-miRNA hairpins for processing. Cell. 2013;152(4):844–858.

Babiarz J.E. Mouse ES cells express endogenous shRNAs, siRNAs, and other microprocessor-independent, dicer-dependent small RNAs. Genes & Development. 2008;22(20):2773–2785.

Baek D. The impact of microRNAs on protein output. Nature. 2008;455(7209):64–U38.

Barman B., Bhattacharyya S.N. mRNA targeting to endoplasmic reticulum precedes ago protein interaction and microRNA (miRNA)-mediated translation repression in mammalian cells. The Journal of Biological Chemistry. 2015;290(41):24650–24656.

Barrett S.P., Salzman J. Circular RNAs: Analysis, expression and potential functions. Development. 2016;143(11):1838–1847.

Barrey E. Pre-microRNA and mature microRNA in human mitochondria. PLoS ONE. 2011;6(5):e20220.

Bartel D.P. MicroRNAs: Target recognition and regulatory functions. Cell. 2009;136(2):215–233.

Behm-Ansmant I. mRNA degradation by miRNAs and GW182 requires both CCR4:NOT deadenylase and DCP1:DCP2 decapping complexes. Genes & Development. 2006;20(14):1885–1898.

Benhamed M. Senescence is an endogenous trigger for microRNA-directed transcriptional gene silencing in human cells. Nature Cell Biology. 2012;14(3):266–275.

Ben-Yishay R., Shav-Tal Y. The dynamic lifecycle of mRNA in the nucleus. Current Opinion in Cell Biology. 2019;58:69–75.

Boele J. PAPD5-mediated 3' adenylation and subsequent degradation of miR-21 is disrupted in proliferative disease. Proceedings of the National Academy of Sciences of the United States of America. 2014;111(31):11467–11472.

Borbolis F., Syntichaki P. Cytoplasmic mRNA turnover and ageing. Mechanisms of Ageing and Development. 2015;152:32–42.

Bose M. Spatiotemporal uncoupling of microRNA-mediated translational repression and target RNA degradation controls microRNP recycling in mammalian cells. Molecular and Cellular Biology. 2017;37(4):e00464–e00516.

Bosson A.D., Zamudio J.R., Sharp P.A. Endogenous miRNA and target concentrations determine susceptibility to potential ceRNA competition. Molecular Cell. 2014;56(3):347–359.

Bottini S. Post-transcriptional gene silencing mediated by microRNAs is controlled by nucleoplasmic Sfpq. Nature Communications. 2017;8(1):1189.

Braun J.E. GW182 proteins directly recruit cytoplasmic deadenylase complexes to miRNA targets. Molecular Cell. 2011;44(1):120–133.

Braun J.E. A direct interaction between DCP1 and XRN1 couples mRNA decapping to 5′ exonucleolytic degradation. Nature Structural & Molecular Biology. 2012;19(12):1324–1331.

Brkic J. MicroRNA-218-5p promotes endovascular trophoblast differentiation and spiral artery remodeling. Molecular Therapy. 2018;26(9):2189–2205.

Budak H. MicroRNA nomenclature and the need for a revised naming prescription. Briefings in Functional Genomics. 2016;15(1):65–71.

Cai X.Z., Hagedorn C.H., Cullen B.R. Human microRNAs are processed from capped, polyadenylated transcripts that can also function as mRNAs. RNA. 2004;10(12):1957–1966.

Cao Z.X., Grima R. Analytical distributions for detailed models of stochastic gene expression in eukaryotic cells. Proceedings of the National Academy of Sciences of the United States of America. 2020;117(9):4682–4692.

Carbonaro M., O'Brate A., Giannakakou P. Microtubule disruption targets HIF-1alpha mRNA to cytoplasmic P-bodies for translational repression. The Journal of Cell Biology. 2011;192(1):83–99.

Catalanotto C., Cogoni C., Zardo G. MicroRNA in control of gene expression: An overview of nuclear functions. International Journal of Molecular Sciences. 2016;17(10):1712.

Chandradoss S.D. A dynamic search process underlies microRNA targeting. Cell. 2015;162(1):96–107.

Chang T.C. Genome-wide annotation of microRNA primary transcript structures reveals novel regulatory mechanisms. Genome Research. 2015;25(9):1401–1409.

Chaulk S.G. Role of pri-miRNA tertiary structure in miR-17 similar to 92 miRNA biogenesis. RNA Biology. 2011;8(6):1105–1114.

Cheloufi S. A dicer-independent miRNA biogenesis pathway that requires Ago catalysis. Nature. 2010;465(7298):584–589.

Chen C.Y.A. Ago-TNRC6 triggers microRNA-mediated decay by promoting two deadenylation steps. Nature Structural & Molecular Biology. 2009;16(11):1160–U6.

Chendrimada T.P. TRBP recruits the dicer complex to Ago2 for microRNA processing and gene silencing. Nature. 2005;436(7051):740–744.

Chou C.H. miRTarBase update 2018: A resource for experimentally validated microRNA-target interactions. Nucleic Acids Research. 2018;46(D1):D296–D302.

Christie M. Structure of the PAN3 pseudokinase reveals the basis for interactions with the PAN2 deadenylase and the GW182 proteins. Molecular Cell. 2013;51(3):360–373.

Conrad T. Microprocessor activity controls differential miRNA biogenesis in vivo. Cell Reports. 2014;9(2):542–554.

D'Ambrogio A. Specific miRNA stabilization by Gld2-catalyzed monoadenylation. Cell Reports. 2012;2(6):1537–1545.

de Rie D. An integrated expression atlas of miRNAs and their promoters in human and mouse. Nature Biotechnology. 2017;35(9):872–878.

De N. Highly complementary target RNAs promote release of guide RNAs from human Argonaute2. Molecular Cell. 2013;50(3):344–355.

Denli A.M. Processing of primary microRNAs by the microprocessor complex. Nature. 2004;432(7014):231–235.

Denzler R. Impact of microRNA levels, target-site complementarity, and cooperativity on competing endogenous RNA-regulated gene expression. Molecular Cell. 2016;64(3):565–579.

Desvignes T. miRNA nomenclature: A view incorporating genetic origins, biosynthetic pathways, and sequence variants. Trends in Genetics. 2015;31(11):613–626.

Detzer A. Cell stress is related to re-localization of Argonaute 2 and to decreased RNA interference in human cells. Nucleic Acids Research. 2011;39(7):2727–2741.

Dharap A. MicroRNA miR-324-3p induces promoter-mediated expression of RelA gene.