Plant miRNAs are a class of 21∼22 nucleotides (nt) endogenous small RNAs that repress gene expression either through target RNA cleavage and/or translational inhibition (Achkar et al., 2016; Yu et al., 2017b). In plants, miRNAs are involved in various cellular and physiology processes and impact plant development, growth, and stress responses (D'Ario et al., 2017; Yu et al., 2019). In the past years, the roles of miRNAs in many biological processes have been gradually revealed. In the meantime, researchers have also established the molecular framework of miRNA biogenesis and modes of action in plants ( Figure 1 ).

Plant MIRNA (MIR) genes are transcribed by RNA POL II rendering 5’ capped (m7Gppp) and 3’ polyadenylated primary transcripts (pri-miRNAs) (Xie et al., 2005; Knop et al., 2017). The transcription of MIR genes is similar to the POL II-mediated eukaryotic genes, with their own promoter region, and their expression is regulated by a variety of transcriptional processes (Zhang et al., 2015). Pri-miRNA is a long stem-loop structure with an imperfectly matched in the middle of the stem that is considered undergo two sequential reactions to form miRNA/miRNA* duplexes in highly specialized nuclear dicing bodies (D-bodies) (Fang and Spector, 2007). In the first step, pri-mRNA is processed into a short precursor (pre-mRNA) by the microprocessors with central components of DCL1, HYPONASTIC LEAVES 1 (HYL1, also known as DRB1), and SERRATE (SE) (Rogers and Chen, 2013; Achkar et al., 2016). Pre-miRNA is further processed by DCL1 into a double-stranded miRNA duplex (Yu et al., 2017b; Li and Yu, 2021a). The nascent miRNA/miRNA* duplexes usually contain 2 nt overhang and two hydroxyl groups at both 3’ ends. The 2’ OH of the 3’ terminal nucleotide is immediately methylated by HUA ENHANCER 1 (HEN1), and the guide strand (miRNA) in the double-stranded RNA is loaded into AGO1 or AGO1 paralogs to form the miRNA-induced silencing complex (miRISC), while the passenger strand (miRNA*) is degraded (Yu et al., 2005; Singh et al., 2018). The assembly process of miRISC mainly occurs in the nucleus, which is attributed to the nucleo-cytosolic shuttling protein AGO1 (Bologna et al., 2018). Furthermore, nuclear plant miRNAs-AGO1 are exported to the cytoplasm through the CRM1 (EXPO1)/NES-dependent manner via TREX-2 and nucleoporin protein (NUP1) facilitated pathway (Bologna et al., 2018; Zhang et al., 2020). It has been shown that the conformation of the AGO1 is modified to accommodate the process of RISC assembly, which also needs to be facilitated by many cofactors such as HEAT-SHOCK PROTEIN 90 (HSP90) and CYCLOPHILIN 40 (CYP40) (Iki et al., 2010; Iwasaki et al., 2010; Iki et al., 2012; Nakanishi, 2016). In addition, ENHANCED MIRNA ACTIVITY1 (EMA1) and TRANSPORTIN1 (TRN1) are members of the AGO1-interacting importin-β family of proteins, which negatively and positively regulate miRNA loading into AGO1, respectively (Wang et al., 2011; Cui et al., 2016).

The actions of cytoplasmic miRNAs occur mainly through the repression of their target mRNAs in two ways: transcripts cleavage and translational repression. miRISC-catalyzed mRNA cleavage is accomplished by the PIWI domain of AGO1 or other paralogs proteins. After cleavage, the 3’ fragments are subsequently degraded by 5’-3’ exonuclease EXORIBONUCLEASE 4 (XRN4) (Souret et al., 2004). The 5’ fragments are uridylated by HEN1 SUPPRESSOR 1 (HESO1) and UTP: RNA URIDYLYLTRANSFERASE 1 (URT1) at the 3’ end both in vivo and in vitro (Ren et al., 2014; Wang et al., 2015). In addition, RISC-INTERACTING CLEARING 3’-5’ EXORIBONUCLEASE 1 (RICE1) are involved in degradation of uridylated 5’ fragments (Zhang et al., 2017b). Moreover, the cytoplasmic exosome consists of subunits such as SUPERKILLER 2 (SKI2), SKI3 and SKI8, which are responsible for degrading the 5’ fragment generated by RISC (Branscheid et al., 2015). Translational repression also plays a vital role in miRNA-mediated post-transcriptional regulation. In fact, as translation repression was observed less frequently than transcript cleavage, the mechanism of translation repression still needs to be elucidated. Nonetheless, multiple factors are identified to participate in miRNA-mediated translational repression (Brodersen et al., 2008; Yang et al., 2012; Li et al., 2013; Reis et al., 2015). Notably, ALTERED MERISTEM PROGRAM1 (AMP1), an integral membrane protein associated with AGO1 on the endoplasmic reticulum (ER), which is responsible for miRNA-directed translational repression at the protein level (Li et al., 2013).

To maintain miRNA homeostasis, the abundance of miRNAs needs to be precisely regulated. The miRNA/miRNA* is stabilized by HEN1 through 3’-terminal 2’-O-methylation (Chen et al., 2002). HESO1 and URT1 can uridylate the 3’ unmethylated miRNA, leading to miRNA degradation (Ren et al., 2014; Wang et al., 2015). Free miRNAs can be degraded by SMALL RNA DEGRADING NUCLEASE 1 (SDN1) (Ramachandran and Chen, 2008). For miRNAs bound to AGO1, removal of methyl group at the 3’ end is also initially mediated by SDN1 (Ramachandran and Chen, 2008). Subsequently, a few of proteins are responsible for the degradation of miRNAs at different stages (Ramachandran and Chen, 2008; Ren et al., 2012a; Zhao et al., 2012b; Tu et al., 2015).

This review summarizes our current understanding of the MIR gene transcription, pri-miRNA cotranscriptional processing, pri-miRNA modification, miRNA AGO1 loading and nuclear export, turnover and degradation of miRNAs. We aim to provide an updated view of miRNA biogenesis pathways in plants.

As the substrates of the microprocessor, pri-miRNAs themselves also affect the processing efficiency in various ways (Li and Yu, 2021a). Unlike in metazoans, plant pri-miRNAs have highly variable sizes and secondary structures. Pri-miRNA secondary structure is one of the important factors which affecting the cleavage site selection and processing efficiency of DCL1 ( Figure 2B ). DCL1 functions as molecular ruler that measures and cleaves pri-miRNAs into 21 nt long mature miRNA duplexes (Macrae et al., 2006). DCL1 processes pri-miRNAs in a base-to-loop manner in two steps (Kurihara and Watanabe, 2004). The first cut is located 15 to 17 nt away from the base of the stem or a bulge or unstructured region within the loop-distal stem. The resulting precursor-miRNA (pre-miRNA) is further cleaved by DCL1 to produce a 21 nt miRNA/miRNA* duplex (Song et al., 2010; Liu et al., 2012). However, some pri-miRNAs with longer hairpins are processed in a base-to-loop manner. The first cut is ∼21 nt proximal to the miRNA/miRNA*, which is likely determined by the presence of a 15 bp stem below the cut site and additional cuts to generate miRNA/miRNA* duplex (Bologna et al., 2013; Zhang et al., 2015). In contrast, some pri-miRNAs such as pri-miR159a and pri-miR319a possess long upper stem and are cleaved by DCL1, resulting in loop-to-base processing mechanism (Bologna et al., 2009). Interestingly, a few pri-miRNAs contain multibranched terminal loops that can be processed bidirectionally, but the base-to-loop processing seem to be more effective, as in the case of pri-miR166C (Zhu et al., 2013). In addition, several young miRNAs (miR822 and miR839) from long inverted repeat (IR) locus are processed by DCL4 instead of DCL1 (Rajagopalan et al., 2006).

Unlike DNA modifications that can dynamically modulate gene expression, some factors in mRNA modifications are critical to the biogenesis of miRNAs (Marquardt et al., 2022). It is known that N6-methyladenosine (m⁶A) modification occurs in plant pri-miRNAs, which can be introduced by adenosine methylesterase (MTA), a homolog of METTL3 (Bhat et al., 2020). MTA can interact with known miRNA biogenesis regulators, such as POL II and Tough (TGH), to promote miRNA biogenesis (Ren et al., 2012b; Bhat et al., 2020). The abundances of m6A and miRNAs are apparently reduced in the mta mutant (Bhat et al., 2020). The low level of m⁶A may affect the association between the stem-loop structure of pri-miRNAs and HYL1 (Bhat et al., 2020). Additionally, m⁶A can modulate splicing process through YTH Domain-Containing 1 (YTHDC1) and other several pre-mRNA splicing factors (Xiao et al., 2016). These studies indicate that RNA modifications can occur cotranscriptionally affecting downstream processing events (Marquardt et al., 2022).

Extensive studies were published to describe how the nascent pri-miRNAs are processed into mature miRNAs. In plants, DCL1 is the riboendonuclease to process the long pri-miRNAs into mature miRNAs (Macrae et al., 2006). The miRNA dicing complex is composed of DCL1, HYL1 and SE. In the past decades, many factors that directly or indirectly modulate microprocessor activity have been identified ( Figure 2C ) (Li and Yu, 2021a; Zhang et al., 2022). These factors include CDC5, PRL1, Phytophthora Suppressor of RNA Silencing 1 (PSR1), MODIFIER OF SNC1, 2 (MOS2), MAC3, MAC5, MAC7, CAP-BINDING PROTEIN 20 (CBP20) and CBP80, homolog of the DEAD-box pre-mRNA splicing factor Prp28 (SMA1), the pre-mRNA processing factor 6 homolog STABILIZED1 (STA1), HIGH OSMOTIC STRESS GENE EXPRESSION 5 (HOS5), ARGININE/SERINERICH SPLICING FACTOR 40 (RS40) and RS41, the U1 snRNP Subunit LETHAL UNLESS CBC 7 RL (LUC7rl), the THO2 in the THO/TREX complex, GLYCINERICHRBP 7 (GPR7), SE-Associated Protein 1 (SEAP1) etc. also interact with the microprocessor and positively regulate pri-miRNA processing (Laubinger et al., 2008; Ben Chaabane et al., 2013; Wu et al., 2013; Zhang et al., 2013; Koster et al., 2014; Zhang et al., 2014; Chen et al., 2015; Francisco-Mangilet et al., 2015; Qiao et al., 2015; Jia et al., 2017; Knop et al., 2017; Li et al., 2018a; Li et al., 2018b; Li SJ et al., 2020; Li et al., 2021b).

The 2’-O-methylation modification of mature miRNAs at their 3’ ends is catalyzed by the methyltransferase HEN1, which represents a key step in miRNA stabilization in plants ( Figure 1 ). Null hen1-2 mutant exhibits a series of developmental defects such as dwarfism, late flowering and short siliques, impaired photomorphogenic and skotomorphogenic (Chen et al., 2002; Tsai et al., 2014), and features of reduced miRNA abundance and miRNA size heterogeneity (Li et al., 2005). The crystal structure shows that the double-stranded RNA (dsRNA) of the miRNA duplex produced by DCL1 as a substrate can be preferentially bound by HEN1 (Huang et al., 2009). HEN1 is localized both in the nucleus and cytoplasm (Fang and Spector, 2007), and the place where methylation occurs remains unknown. Moreover, in vitro assays demonstrate HEN1 may directly interact with DCL1 and HYL1 (Baranauskė et al., 2015).

After miRNA maturation, nuclear HYL1 may be required for selecting the miRNA strand from the miRNA duplex and interact with other cofactors to load the guide strand into AGO1 ( Figure 1 ) (Eamens et al., 2009; Manavella et al., 2012). AGO1 has conserved NLS and NES sequences conferring the ability to shuttle between the cytoplasm and nucleus (Bologna et al., 2018). This challenges the notion that the loading of mature miRNAs into AGO1 in plants occurs primarily in the cytoplasm as in animals. A recent report showed that RBV can promote the loading of miRNAs into AGO1 (Liang et al., 2022). Some of miRNAs, such as miR165, miR166 and miR160, are found to be poorly loaded into AGO1 after being processed in a co-transcriptional manner, implying that miRNAs processing during transcription take a nuclear export route that avoids AGO1 loading, resulting in the cytoplasmic pool of “free” miRNAs that can move from cell to cell (Gonzalo et al., 2022). This could also explain why HST plays a co-transcriptional role in miRNA biogenesis and cell-autonomously promotes cell-to-cell movement (Brioudes et al., 2021; Cambiagno et al., 2021). For miRNAs loaded illegitimately into AGO1, or non-functional AGO1-miRNA pairs and unloaded AGO1, the F-box protein FBW2 assembles an SCF complex to promote AGO1 degradation, and regulation of FBW2 activity also requires the CURLY LEAF (CLF), a subunit of the polycomb repressor complex 2 (PRC2) (Dalmadi et al., 2019; Ré et al., 2020; Hacquard et al., 2022). The export process requires AGO1 to form a complex with chaperone HSP90 during its association with the methylated mature miRNA/miRNA* duplex (Iki et al., 2010). The activity of HSP90’s chaperone may alter the conformation of AGO1 when it binds or separates from it (Rogers and Chen, 2013). However, this process is promoted by Cyclophilin 40/Squint (CYP40/SQN) containing the TPR domain, and is suppressed by Protein Phosphatase 5 (PP5) (Iki et al., 2010; Iwasaki et al., 2010; Iki et al., 2012).

The nuclear envelope (NE) separates nuclear and cytoplasmic to control the macromolecular exchange, and it has a highly organized double membrane that is morphologically continuous with the endoplasmic reticulum (ER) of eukaryotic cells (Nigg, 1989). The nuclear pore complex (NPC), one of the structural components of NE, is composed of multiple copies of nucleoporin proteins (NUPs) (Strambio-De-Castillia et al., 2010). In Arabidopsis, The TREX-2 complex includes SAC3A, SAC3B, SAC3C, DSS1-(I), DSS1-(V), CEN1 and CEN2, which are localized in the NPC basket and are essential for transcription and mRNA export (Lu et al., 2010). It has been reported that the nuclear basket protein and TREX-2 complexes are linked by THP1-NUP1 interactions (Lu et al., 2010). NUP1 and THP1 have also been shown to facilitate nuclear export of miRNAs and AGO1 (Zhang et al., 2020). This links miRNA biosynthesis, assembly of miRISC and nuclear export processes through a cascade of dynamic regulation ( Figure 1 ).

The abundance of miRNAs in cells is balanced by the rate of biogenesis and degradation. In addition to miRNA biogenesis, another factor affecting miRNA levels is miRNA degradation which also includes miRNA 3’ uridylation and 3’ truncation ( Figure 1 ) (Yu et al., 2017b). As mentioned above, HEN1 mediated miRNA methylation can protect the small RNA duplex from degradation. For those unmethylated small RNAs, the addition of a U-rich tail at 3’ end is governed by the activity of the nucleotidyl transferase HESO1 and URT1, which is considered as a signal for the degradation (Ren et al., 2012a; Zhao et al., 2012b; Tu et al., 2015). HESO1 is in both the nucleus and cytoplasm, and URT1 is predominantly distributed in the cytoplasm. The null heso1 mutant partially rescues the developmental defects of the hen1 mutant, coupled with an increase in miRNA accumulation and a reduction in 3’ uridylation (Zhao et al., 2012b). The point mutation of URT1 (urt1-3) partially rescues the fertility defect of hen1 heso1-2 mutant, and the triple mutant of hen1-2 heso1-2 urt1-3 does not result in the complete absence of miRNA uridylation, presumably in the presence of other uridyltransferases (Wang et al., 2015). In addition, nucleotidyl transferase proteins 4 (NTP4) was recently reported as a close homolog of HESO1 and URT1 to direct the asymmetric uridylation of miRNA/miRNA* (guide/messenger) duplexes to regulate miRNA accumulation (Kong et al., 2021).

Some miRNAs can also be truncated before being uridylated in hen1 mutant, implying divergent degradation fates (Zhai et al., 2013). Degradation of miRNAs from both directions involves two separated mechanisms. The involvement of 3’ to 5’ exonucleases in the degradation of mature small RNAs is the SDN family, which belongs to the DEDD superfamily ( Figure 1 ) (Ramachandran and Chen, 2008; Wang et al., 2018). In vitro assays show SDN1 is unable to degrade U-tailed miRNAs, by contrast, several exonucleases in other eukaryotes prefer uridylated RNAs as substrates in mammals and yeast (Ramachandran and Chen, 2008). Additionally, it has been demonstrated that SDN1 only works on short ssRNAs, not small RNA duplexes, pre-miRNAs, or larger RNAs in vitro (Ramachandran and Chen, 2008). Analysis of the small RNA-seq data of hen1 and hen1 sdn1 sdn2 plants shows the 3’ truncation of some miRNAs is reduced when there is a lack of SDN1 and SDN2, demonstrating the SDNs are responsible for the 3’ truncation of miRNAs (Yu et al., 2017a). It is possible that SDN1 and HESO1/URT1 modulate the miRNAs that are coupled to AGO1 because truncated and uridylated miRNAs are associated with AGO1 in vivo (Zhao et al., 2012a).

The 10 AGO proteins are divided into three clades containing the conserved PAZ, MID and PIWI domains (Carbonell, 2017). Of these, AGO1 not only acts as an effector but may also serve as a refuge to protect miRNAs from degradation, since the abundance of many miRNAs is reduced in ago1 mutants (Vazquez et al., 2004). As a paralog of AGO1, AGO10 plays an opposite role to AGO1 in miRNA stabilization. AGO10 is involved in regulating stem cell and leaf polarity, miR165/166 is at high abundance in ago10 mutants, while overexpression of AGO10 reduces the accumulation of miR165/166 (Yu et al., 2017a; Wang JL et al., 2019). AGO10 binds miR165/6 with higher affinity than AGO1 and promotes its degradation, and rather than protecting miR165/6 (Zhu et al., 2011; Zhou et al., 2015). Therefore, different effects on the stability of the miRNAs are determined by which AGO proteins are loaded.

In order to adapt to environmental variability for survival, sessile organisms have adopted a range of responsive strategies. MiRNAs are essential players in the regulation of proper growth and development by cleaving of target mRNA and translational repression in various manners. Players of the miRNA biogenesis pathway and their activities are central to the generation of miRNA, and they can be mediated by regulators at different tissues and developmental stages. In this review, we focused on the dynamic control of the key steps in miRNA biogenesis pathway and also exhibit a hierarchical model of how these processes are regulated. We also emphasized the importance of how the miRNA processing is linked with MIR gene transcription and miRNA export, especially on the subcellular regulation of plant miRNAs. Nevertheless, studies on how and why plant miRNA biogenesis steps have to be tightly linked are still largely unknown. In addition, how the major protein complexes in each step dynamically interact or communicate is still unknow. For example, there is no studies on whether POL II dynamically associate with D-bodies and nuclear pore complex, or whether the miRNA loci have direct interaction with D-bodies. Moreover, how does the nuclear miRNA loading cooperate with miRISC export still remain to be further explored. Furthermore, Arabidopsis or other terrestrial or semiterrestrial plants with multi-organizational systems pose research barriers compared to animal cells, so a breakthrough in deep sequencing technology at the single-cell level in plants is still needed. However, in plants, utilizing canonical forward or reverse genetic merits can serve us to screen novel cofactors that regulate miRNA production pathways.

ND drafted the manuscript and the figures. BZ conceived the idea and revised the manuscript. All authors contributed to the article and approved the submitted version.