Plant DCL Proteins

Plant DCL Proteins Abstract Dicer, a double-stranded RNA (dsRNA)-specific endoribonuclease, plays an essential role in triggering both transcriptional and post-transcriptional gene silencing in eukaryotes by cleaving dsRNAs or single-stranded RNAs bearing stem-loop structures such as microRNA precursor transcripts into 21- to 24-nt small RNAs. Unlike animals, plants have evolved to utilize at least four Dicer-like (DCL) proteins. Extensive genetic studies have revealed that each DCL protein participates in a specific gene silencing pathway, with some redundancy. However, a mechanistic understanding of how the specific action of each DCL protein is regulated in its respective pathway is still in its infancy due to the limited number of biochemical studies on plant DCL proteins. In this review, we summarize and discuss the biochemical properties of plant DCL proteins revealed by studies using highly purified recombinant proteins, crude extracts, and immunoprecipitates. With help from co-factor proteins and an ATP ase/DExH-box RNA-helicase domain, the microRNA-producing enzyme DCL1 recognizes bulges and terminal loop structures in its substrate transcripts to ensure accurate and efficient processing. DCL4 prefers long dsRNA substrates and requires the dsRNA-binding protein DRB4 for its activity. The short-dsRNA preference of DCL3 is well suited for short-RNA transcription and subsequent dsRNA formation by coupling between a plant-specific DNA-dependent RNA-polymerase IV and RNA-dependent RNA-polymerase 2 in the transcriptional gene silencing pathway. Inorganic phosphate also seems to play a role in differential regulation of DCL3 and DCL4 activities. Further development of biochemical approaches will be necessary for better understanding of how plant DCL proteins are fine-tuned in each small RNA biogenesis pathway under various physiological conditions. Introduction RNA silencing, also known as RNA interference (RNAi), is one of the fundamental molecular mechanisms conserved in most eukaryotes to regulate gene expression both transcriptionally and post-transcriptionally. In both situations, what triggers the RNA silencing pathway is a small RNA molecule, 21 to 24 nt in length, called small interfering RNA (siRNA) or microRNA (miRNA) depending on its origin and the downstream pathways involved. The class 3 endoribonuclease (RNase) III enzymes known as Dicer are responsible for producing siRNA from longer double-stranded RNAs (dsRNAs) and miRNA from single-stranded RNAs with internal stem-loop structures   by a dsRNA-specific endoribonuclease. Therefore, the activity and regulation of Dicer-family proteins in a cell are vital to many biological processes requiring flexible adjustments at the level of gene expression, such as development, organogenesis, the circadian rhythm, biotic and abiotic stress responses, and defense against viruses an d transposons. Biochemical characterization of Dicers in animals The Dicer family is a unique class of RNase III enzymes due to the presence of an ATPase/DExD/H-box helicase domain at the N-terminus, a Piwi/Argonaute/Zwille (PAZ) domain in the middle and dual RNase III domains followed by one or two dsRNA-binding domains in the C-terminal half (exception: Giardia intestinalis) (Figure 1) (Bernstein et al. 2001). In general, the helicase domain serves as a protein-protein interaction surface recruiting co-factor regulatory proteins (Lee et al. 2006; Ma et al. 2008; Ye et al. 2007). It also utilizes ATP hydrolysis to achieve processive cleavage of the long dsRNA substrate (Cenik et al. 2011; Welker et al. 2010). The PAZ domain contains a conserved pocket for recognizing the terminus of the dsRNA substrate, and the distance between PAZ and the RNase III catalytic center determines the product sizes (MacRae et al. 2007; MacRae et al. 2006). Each of the two RNase III domains cuts one of the dsRNA strands, leaving a characteristic 2-nt overhang at 3-end of the product (Elbashir et al. 2001; Takeshita et al. 2007; Zhang et al. 2004). The C-terminal dsRNA-binding domains (dsRBDs) serve as a protein-protein interaction interface and nuclear localization signals, in addition to having dsRNA-binding function (Doyle et al. 2013; Hiraguri et al. 2005; Wostenberg et al. 2012). The specific functionality of each domain differs depending on the Dicer protein. Since the first demonstration of in vitro small RNA-producing activity of Dicer in the fruit fly Drosophila melanogaster (Bernstein et al. 2001), its biochemical properties and regulatory machinery have been extensively studied in humans, D. melanogaster and Caenorhabditis elegans. In humans, there is only one Dicer-family protein (hDicer), which cleaves short-hairpin pre-miRNAs produced by Drosha and dsRNA substrates into 20- to 22-nt small RNAs in an ATP-independent manner (Myers et al. 2003; Provost et al. 2002; Zhang et al. 2002). The cleavage activity requires a divalent metal cation such as Mg2+, Co2+ or Mn2+, and recognizes mainly the 5-end of the substrate to dictate the product length (Park et al. 2011). This 5-counting rule is reliant on the conserved 3-pocket motif within the PAZ domain and the 5-pocket motif, which is less conserved in Dicers of other eukaryotes. The binding of Dicer to a dsRNA substrate and its cleavage are uncoupled, because Dicer can bind to dsRNA with out Mg2+ or under low temperature (Provost et al. 2002; Zhang et al. 2002). The helicase domain of hDicer has an autoinhibitory function (Ma et al. 2008). In line with this, the activity of recombinant full-length hDicer protein can be improved under limited proteolytic conditions (Zhang et al. 2002). hDicer is responsible for both siRNA and miRNA production, and co-factor dsRNA-binding proteins TRBP and PACT dictate hDicer function in the two distinct small RNA production pathways (Chendrimada et al. 2005; Haase et al. 2005; Kok et al. 2007; Lee et al. 2013; Lee et al. 2006). In particular, the hDicer complex containing PACT disfavors siRNA precursor dsRNA and shows different cleavage patterns on the same pre-miRNA substrate than the hDicer-TRBP complex (Lee et al. 2013). The interaction with TRBP occurs through the hDicer helicase domain, and stimulates the hDicers catalytic activity. (Ma et al. 2008). Similarly, it has been reported that the C. elegans Dcr-1 interacts with a dsRNA-binding protein RDE-4 which enhances the Dicer activity toward long dsRNA substrates in siRNA production, while RDE-4 is apparently dispensable in miRNA production pathway (Parker et al. 2006; Parker et al. 2008; Tabara et al. 2002). D. melanogaster has two Dicer proteins, Dcr-1 and Dcr-2, which produce miRNA and siRNA, respectively (Lee et al. 2004; Miyoshi et al. 2010). Dcr-1 alone can process dsRNA into siRNA in vitro, but its interaction with the dsRNA-binding protein Loquacious isoform PB (Loqs-PB) confers pre-miRNA substrate specificity to the Dcr-1-Loqs complex by suppressing cleavage of long perfect dsRNAs and enhancing pre-miRNA processing activity (Saito et al. 2005; Zhou et al. 2009). Dcr-2 interacts with Loqs isoform PD and another dsRNA-binding protein, R2D2, in the siRNA production pathway (Liu et al. 2003; Liu et al. 2006; Miyoshi et al. 2010; Zhou et al. 2009). Dcr-2 alone is also capable of cleaving a pre-miRNA precursor in an ATP-independent manner, but R2D2 significantly suppresses Dcr-2 activity toward pre-miRNA, whereas Loqs-PD enhances the cleavage activity of Dcr-2 toward long perfect dsRNA precursors by boosting its affinity to the substrate (Cenik et al. 2011; Miyoshi et al. 2010). The pr ocessive processing of long dsRNA substrates by Dcr-2 depends on ATP hydrolysis by its ATPase/helicase domain, implying that one of the functions of the helicase domain is to allow Dcr-2 to produce multiple siRNAs from a single long dsRNA molecule before it dissociates from the substrate (Cenik et al. 2011). Such differential regulation of Dicer activity through specific interaction with co-factor dsRNA-binding proteins in distinct pathways is commonly found in most of the systems studied, including plants. DCL proteins in plants Plant genomes contain at least four distinct classes of DCL family proteins (DCL1-4). Like their animal counterparts, each class of DCL has evolved to participate in its primary pathway (Figure 2), but the three siRNA-producing DCLs (DCL2-4) function redundantly as well, because defects in one class of DCL can be compensated for by other classes in some cases (Gasciolli et al. 2005; Mukherjee et al. 2013; Xie et al. 2004). Because DCL1 is the only Dicer protein that produces most 21-nt miRNAs (Kurihara and Watanabe 2004; Reinhart et al. 2002), knockout mutants of DCL1 are embryonic lethal (Schauer et al. 2002). DCL4 is the major producer of 21-nt antiviral siRNA and endogenous siRNAs such as trans-acting siRNA and phased siRNAs (phasiRNA) (Bouche et al. 2006; Gasciolli et al. 2005; Mukherjee et al. 2013; Qu et al. 2008; Xie et al. 2005; Yoshikawa et al. 2005). DCL2 can compensate for the loss of DCL4 (Bouche et al. 2006; Gasciolli et al. 2005; Parent et al. 2015), although its major function remains unclear. DCL3 mainly produces 24-nt repeat-associated siRNAs derived from transposons and DNA repetitive elements, and participates in transcriptional gene silencing (TGS) through RNA-dependent DNA methylation, suppressing proliferation of these elements (Henderson et al. 2006; Pontes et al. 2006; Xie et al. 2004). In addition to the four classes of DCLs, monocots have another distinct class of Dicer, DCL5 (also known as DCL3b) (Margis et al. 2006). DCL5 is specifically expressed in developing panicles and is responsible for 24-nt reproductive phasiRNAs, although the biological significance of a reproductive-organ-specific 24-nt phasiRNA pathway mediated by this specific Dicer remains to be elucidated (Borges and Martienssen 2015; Fei et al. 2013; Kapoor et al. 2008; Song et al. 2012). This pathway might be analogous to the Dicer-independent PIWI-interacting RNA (piRNA) pathway in vertebrates, which suppresses transposons and other genes specifically in germlines (H irakata and Siomi 2016). Both forward and reverse genetics and physiological studies have successfully dissected the major RNA silencing pathways and allowed identification of the function of DCL genes in each pathway in plants. However, investigations on the molecular and enzymatic characteristics underlying the functional diversification and specificity of the DCL proteins are still in their infancy. Detection of DCL activity in crude extracts of various plants Biochemical characterization of plant Dicer activity was first demonstrated in wheat germ extract (monocot) and cauliflower extract (dicot), which contain multiple DCL activities producing ~21 nt and ~24 nt small RNAs with 2-nt 3-overhangs in the double-stranded form (Tang et al. 2003). These activities are weaker in the absence of ATP, consistent with characteristics of Dicer family proteins from Drosophila and C. elegans. Long dsRNA competitors effectively suppress both activities in wheat germ extract. The 24-nt small RNA producing activity was inhibited by 25-nt synthetic siRNA duplexes, whereas 21-nt small RNA production was unaffected by 21-nt synthetic siRNA duplex competitors, suggesting that two different enzymes with active sites that have distinct size-dependent binding properties are in the wheat germ extract (Tang et al. 2003). A recent study on wheat germ extract characterized these activities in further detail, revealing (1) that the 21-nt activity could be found in a much larger (~950 kDa) complex than the 24-nt activity, which had maximum activity in an approximately 450 kDa complex; and (2) the biochemical properties associated with the activities, such as divalent cation and NTP requirements, optimum NaCl concentration, temperature, and pH, and substrate length dependence (Shivaprasad et al. 2015). The identities of the DCL enzymes responsible for these activities in the wheat germ extract remain to be identified. A better understanding of the biochemical characteristics of individual plant Dicer proteins has come from the model plant Arabidopsis thaliana, which has four DCL proteins: DCL1, DCL2, DCL3 and DCL4 (summarized in Table 1). The first in vitro DCL activity in A. thaliana was demonstrated using a suspension cell lysate, a crude extract of inflorescence tissue, and an immunoaffinity-purified protein complex (Qi et al. 2005). Similar to the previous study using wheat germ extract or cauliflower, extracts from both Arabidopsis cultured cells and inflorescence tissue contained DCL dsRNA-cleaving activity producing 21- and 24-nt small RNAs from 400-bp dsRNA (Qi et al. 2005). The 21-nt producing activity and 24-nt producing activity were found in >660 kDa and ~400 kDa fractions, respectively, suggesting that these Dicers reside in protein complexes composed of multiple co-factors (Qi et al. 2005). In agreement with previous genetic studies showing CARPEL FACTORY/DCL1 is responsible for 21-nt miRNA production in vivo (Kurihara and Watanabe 2004; Reinhart et al. 2002), the 21-nt small RNA producing activity was DCL1 immunoaffinity-purified from inflorescence-derived crude extract by an anti-DCL1 antibody (Qi et al. 2005). The 24-nt activity was associated with anti-DCL3 antibody immunoprecipitate, and the activity was abolished when purified from a dcl3-1 mutant, showing that DCL3 is responsible for the 24-nt activity in Arabidopsis inflorescence extract. The immunoaffinity-purified DCL1 activity required ATP, whereas the activity of the DCL3 immunoprecipitate was ATP-independent (Qi et al. 2005). Interestingly, the dcl1-7 mutation did not abolish the 21-nt small RNA producing activity in the extract or immunoprecipitates, implying that the substitution (P415S) in its N-terminal helicase domain did not alter the enzymes catalytic activity itself (Qi et al. 2005); this study also found that the activity of DCL4 responsible for formation of 21-nt siRNA was present in the inflorescence extract. The presence of DCL4 activity in an Arabidopsis crude extract was demonstrated in later studies using 2-week-old seedlings as the starting material (Fukudome et al. 2011; Nagano et al. 2014), and will be discussed later in this review. In-depth biochemical characterization of DCL1, a microRNA-producing enzyme in plants  DCL1 activity requires DRB1/HYL1 and SERRATE for accurate processing of the miRNA precursor Both in wheat germ and Arabidopsis extracts, DCL activities are associated with size fractions larger than DCL monomeric form, implying that these DCLs form functional protein complexes composed of multiple co-factors in vivo. As summarized in an earlier section, such interactions between a Dicer and a co-factor protein are commonly found in mammals, nematodes and insects. One of the most characterized classes of co-factor proteins is a dsRNA-binding protein (dsRBP) harboring multiple dsRNA-binding domains or motifs. The A. thaliana genome encodes five dsRNA-binding (DRB) family proteins: DRB1/HYL1, DRB2 DRB3, DRB4, and DRB5. Multiple genetic and biochemical studies have demonstrated two specific interactions between DCLs and DRBs in A. thaliana: DCL1-DRB1/HYL1 and DCL4-DRB4 (Han et al. 2004; Hiraguri et al. 2005; Kurihara et al. 2006; Nakazawa et al. 2007). Arabidopsis DCL1, DRB1/HYL1, and another co-factor, SERRATE (SE), constitute an essential microRNA production pathway in vivo (Han et al. 2004; Lobbes et al. 2006). Unlike animals, which utilize two distinct RNase III enzymes, Drosha and Dicer, for the first and second cleavage of microRNA precursors, plants do not employ Drosha. Therefore, the DCL1-complex is responsible for the processing of both primary and precursor miRNA substrates. The detailed molecular machinery of the dual miRNA processing mediated by DCL1 and the co-factor proteins have been extensively studied biochemically using highly purified recombinant proteins produced in heterologous systems (summarized in Figure 3). One of the systems utilizes baculovirus-mediated recombinant protein production in Sf21 insect cells, followed by two-step affinity purification (Dong et al. 2008). The highly purified recombinant DCL1 protein alone could process a 94-bp dsRNA substrate with a 2-nt 3-overhang into 21-nt small RNA in an ATP/Mg2+ dependent manner. The optimum NaCl concentration for the activity was 25-50 mM, and a NaCl concentration higher than 100 mM severely impaired the activity (Dong et al. 2008). While the recombinant DCL1 protein alone could produce 21-nt small RNA from both primary and precursor miRNA (pri-/pre-miR167b) substrates in vitro, DRB1/HYL1 and SE recombinant proteins co-incubated in the same reaction mixtures significantly increased both yield and accuracy of the processing (Dong et al. 2008). Without these co-factors, more than 80% of 21-nt small RNA products from the DCL1-alone reaction were due to incorrect processing from the end of the primary miRNA substrate, whereas the processing mediated by the DCL1-DRB1/HYL1-SE complex produced accurate 21-nt products with a sequence identical to miR167b/miR167b*, amounting for up to 81% of the products (Dong et al. 2008). This demonstrated that accurate processing of miRNA precursors by DCL1 requires the co-factors DRB1/HYL1 and SE. Cons istent with a previous study, the interaction between DCL1-DRB1/HYL1 through the second dsRNA-binding motif of DCL1 is important for the precise processing of pri-miRNA in A. thaliana (Dong et al. 2008; Kurihara et al. 2006). Also, using highly purified recombinant proteins and surface plasmon resonance analysis, it has been suggested that DCL1 changes its structural conformation when it binds RNA and exposes more binding sites for SE (Iwata et al. 2013). Binding to substrate dsRNA or miRNA precursors might be an important regulatory step for DCL1 dicing activity, as its dsRNA-binding domains exhibit the strongest binding to dsRNA among the four Arabidopsis DCLs (Hiraguri et al. 2005). ATPase/DExH-box RNA-helicase domain of DCL1 suppresses its dicing activity, confers ATP dependence, and influences processing accuracy In addition to its RNase III and dsRNA-binding domains, the helicase domain of DCL1 plays a significant role in regulating its dicing activity. Two independent forward genetic studies have identified two dcl1 mutant alleles, dcl1-13 (E395K) and dcl1-20 (R363K), as hyl1 suppressors, and the amino acid substitutions of both alleles occur within the ATPase/DExH-box RNA-helicase domain. These dcl1 mutations partially rescue the accumulation of some miRNAs in a hyl1-2 mutant (Liu et al. 2012; Tagami et al. 2009), and dcl1-13 was at least partially able to restore the phenotypic defects of hyl1-2 such as a reduced number of rosette leaves and a   leaf shape (Tagami et al. 2009). Highly purified recombinant DCL1-20 protein exhibited enhanced catalytic activity (Kcat/Km) toward pri-miRNA156a compared to wild-type DCL1 (Liu et al. 2012). Similarly, the helicase domain-deleted DCL1 recombinant protein (DCL1à ¢Ã‹â€ Ã¢â‚¬  Helicase) showed higher processing activity in vitro and was no l onger dependent on ATP for its activity toward pri-miRNA156a (Liu et al. 2012), suggesting that the helicase domain of DCL1 might have an autoinhibitory function like that of human Dicer (Ma et al. 2008; Provost et al. 2002). The in vivo miRNA processing imprecision in hyl1-2, however, was not restored by a dcl1-20 mutation, implying that the partial recovery of the hyl1-2 mutant, including miRNA accumulation, was due to the enhanced catalytic activity resulting from the substitution in the helicase domain (Liu et al. 2012). Interestingly, the effect and magnitude of DRB1/HYL1 and DCL1 helicase domain seem to vary among miRNA precursors. For example, the in vivo processing accuracy of miR156a is much more severely affected by hyl1-2 mutation than miR166b is (Liu et al. 2012). pri-miR156a is processed from the loop-proximal site to the loop-distal base in vitro (Liu et al. 2012), which is considered unusual for plant miRNAs (Addo-Quaye et al. 2009; Mateos et al. 2010). Accurate processing of pri-miRNA166b by native DCL1 is largely dependent on the presence of ATP, and processing by DCL1à ¢Ã‹â€ Ã¢â‚¬  Helicase is less accurate than that of native DCL1 (Liu et al. 2012). In contrast to miR156a, the processi ng precision of which is markedly affected by hyl1-2, that of miR166b was much more impaired by dcl1-20 mutation than hyl1-2 (Liu et al. 2012). Also, the effect of the other helicase mutant allele, dcl1-13, on miRNA production was shown to depend on the presence or absence of DRB1/HYL1 in vivo (Tagami et al. 2009). These observations indicate that efficient processing of different miRNA precursors by DCL1 have different reliance upon DRB1/HYL1 and DCL1helicase domain that potentially depends on structural determinants of the miRNA precursors. Structural determinants for efficient and accurate processing of miRNA precursors by DCL1 Primary transcripts of miRNA (pri-miRNA) have a characteristic secondary structure: a loop-distal stem (lower stem), a miRNA duplex, a loop-proximal stem (upper stem) and a terminal loop (Figure 3). Typical miRNA maturation from these precursors requires at least two cleavages occurring at the lower and upper stems. In animals, the single-stranded base region of the loop-distal stem is recognized by the dsRNA-binding protein DGCR8, which guides the processing center of Drosha to the correct position, which is 11 nt from the base of the stem (Han et al. 2006). However, this distance-from-base rule is not sufficient for plants because the length of the loop-distal stem of plant pri-miRNAs is highly variable (Song et al. 2010). Several structural features of pri-miRNAs that influence the activity, binding position and directionality of the processing by DCL1 have been elucidated genetically and biochemically (Figure 3a, b). One structural determinant lies within the loop-distal stem of pri-miRNA. For the first cleavage at the loop-distal stem, bulges and unpaired regions play a major role in the efficiency of miRNA processing. Mutant pri-miRNAs with closed bulges were processed at the correct position, but resulted in the accumulation of unprocessed pre-miRNAs in vivo, indicating that the rate of subsequent processing at the loop-proximal stem was impaired (Song et al. 2010). In pri-miR171a, which has a long loop-distal stem, the first cleavage position was determined by the distance from a relatively unstructured region instead of the base of the stem; the conserved distance from an unstructured region of the lower stem important for miRNA processing was found to be approximately 15 nt (Figure 3a) (Mateos et al. 2010; Song et al. 2010; Werner et al. 2010). The 15-nt rule was essentially reproduced in an in vitro miRNA processing system using highly purified DCL1-DRB1/HYL1-SE recombinant proteins and an artificial pri-miRNA substrate bearing another unstructured region in the elongated lower stem. In addition to the canonical processing, another type of processing occurred at 15 nt from the artificially introduced unstructured region, validating the functionality of the 15-nt rule (Song et al. 2010). The importance of bulges and unpaired regions in the lower stem for processing by DCL1 might explain why some miRNAs with a near-perfect matched stem seem to be DCL4-dependent, rather than DCL1-dependent (Rajagopalan et al. 2006; Song et al. 2010). On the loop-proximal and terminal loop side, a branched terminal loop (BTL) or a large terminal loop was found to be an essential structural factor that may alter directionality of processing by DCL1 and the resultant miRNA-accumulation (Figure 3b). BTL induces abortive processing of pri-miR166c both in vivo and in vitro (Zhu et al. 2013), meaning the first cleavage of the pri-miRNA occurs in the loop-proximal stem as opposed to the normal productive processing beginning in the loop-distal stem. The molecular basis of this bidirectional processing by DCL1 was further investigated using an in vitro system that reconstitutes the DCL1-processing machinery. For this purpose, DCL1, DRB1/HYL1 and SE harboring Agrobacterium tumefaciens were co-infiltrated to Nicotiana benthamiana leaves, and the transiently expressed DCL1-DRB1/HYL1-SE complex was immunoaffinity-purified two days after infiltration (Zhu et al. 2013). The reconstituted DCL1 complex cleaves the substrate pri-miRNA 16-17 nt fro m the unpaired region of the lower stem, supporting previous studies (Mateos et al. 2010; Song et al. 2010; Werner et al. 2010). By disrupting one of the two RNase III domains of DCL1 alternately and using 5- or 3-end labeled pri-miR166c substrates, the bidirectional nature of both productive and abortive processing was demonstrated (Zhu et al. 2013; Figure 3b). The helicase domain of DCL1 fine-tunes the position of both productive and abortive processing by DCL1 in an ATP-dependent manner (Zhu et al. 2013). DCL1à ¢Ã‹â€ Ã¢â‚¬  Helicase complex could not abortively process a substrate with BTL. Also, wild-type DCL1 required ATP for abortive processing, but not productive processing, indicating that the ATPase-driven helicase activity is necessary in abortive processing to unwind the structured BTL (Zhu et al. 2013; Figure 3b). In productive processing, the effect of helicase deletion and ATP depletion depend on the distance between the processing site and the bulge in the lower stem. Many potential byproducts of the abortive processing of pri-miRNA precursors with BTL can be found in publically available high-throughput small RNA sequencing data from both Arabidopsis and rice, implying that both substrate structure and the functionality of the ATPase/helicase domain of DCL1 are conserved mechanisms to regulate miRNA biogenesis in higher plan ts (Zhu et al. 2013). Dissecting distinct characteristics of DCL3 and DCL4 activities DCL4 activity requires the dsRNA-binding protein DRB4 In A. thaliana, DCL2, DCL3 and DCL4 are responsible for producing various siRNAs 21-24 nt in length. The dsRNA-cleaving activities of DCL3 and DCL4 can be detected in crude extracts prepared from 2-week-old seedlings (Fukudome et al. 2011). Extracts from wild-type seedlings cleave 500-bp dsRNA substrates into 21-nt and 24-nt small RNAs. In this system, the 21-nt and 24-nt small RNA-producing activities can be attributed to DCL4 and DCL3 respectively, because the dsRNA-cleaving activity of the corresponding size was abolished in each of the single mutants (Fukudome et al. 2011). Also, a mutation in the dsRNA-binding protein DRB4, which interacts with DCL4 (Hiraguri et al. 2005; Nakazawa et al. 2007), abolished DCL4 activity in seedling extracts. The DCL4 activity could be further purified by immunoprecipitation with anti-DCL4 or anti-DRB4 antibodies. The immunoaffinity-purified DCL4 requires Mg2+ and ATP for its activity, and is inhibited by >200 mM NaCl. This property is similar to t hat of recombinant DCL1 protein (Dong et al. 2008). The DCL4 complex immunoprecipitated from the drb4-1 mutant did not show dsRNA-cleaving activity, but the addition of recombinant DRB4 protein to the complex restored the 21-nt producing activity in vitro, showing that DRB4 functions as an essential co-factor for the dsRNA-cleaving activity of DCL4 (Figure 4b). In this system, mutant DRB4 proteins harboring substitutions in the conserved amino acid residues that form a hydrogen bond with the phosphodiester backbone of dsRNA at the dsRNA-binding site (H32A in the first dsRBD and K133A in the second dsRBD of DRB4) lost their ability to interact with dsRNA and DCL4, and did not restore DCL4 activity. The second substitution (K133A) alone impaired its interaction with the C-terminal half of DCL4 containing two RNase III domains and two dsRBDs in a GST pull-down assay using recombinant proteins, but was not sufficient to block restoration of DCL4 activity when added to DCL4 immunopurified from a drb4-1 mutant extract. There might be an add itional interaction surface between DCL4 and DRB4 involving dsRBD1 of DRB4 and the N-terminal half of DCL4, which contains an ATPase/DExH-box RNA-helicase domain and an RNA-binding domain (formerly known as domain of unknown function DUF283; Figure 1), as their specific interaction was reported in vitro (Qin et al. 2010). Short dsRNA preference of DCL3 activity orchestrates 24-nt siRNA biogenesis in TGS pathway Crude extracts from 2-week-old seedlings have also been used to characterize substrate specificity of DCL3 and DCL4. Consistent with the long dsRNA preference of Drosophila Dcr1 (Bernstein et al. 2001), DCL4 preferentially cleaves longer dsRNA substrates, and is less efficient in producing 21-nt siRNAs when the substrate is shorter than 50 nt (Nagano et al. 2014). On the other hand, DCL3 activity, producing 24-nt siRNAs, favors shorter substrates such as 30 nt and 37 nt dsRNA with a 1-nt or 2-nt 3-overhang (Nagano et al. 2014). It also favors substrate dsRNA with 5-adenosine or uridine. The 24-nt small RNA produced by DCL3 has a 2-nt 3-overhang, and the cleavage follows the 5-counting rule proposed for human Dicer (Park et al. 2011). DCL3 is not reliant on ATP hydrolysis for activity, as it can still process the short dsRNA substrate in the presence of a non-hydrolyzable ATP analog, adenosine 5-O-(3-thio)triphosphate (Nagano et al. 2014). Unlike DCL4, which targets long dsRNAs such a s RDR6-dependent TAS dsRNAs or exogenous viral dsRNAs in vivo (Bouche et al. 2006; Dunoyer et al. 2005; Qu et al. 2008; Yoshikawa et al. 2005), DCL3 may not need to perform a processive cleavage, which requires ATP hydrolysis, because the length of its targets allows only a single cut (Figure 4a). The DCL3 preference for short dsRNA substrate is consistent with the one precursor, one siRNA model for RNA polymerase IV (Pol IV)-dependent 24-nt siRNA biogenesis (Blevins et al. 2015; Zhai et al. 2015). In this model, a remarkably short (30- to 40-nt) transcript with 5-adenosine is produced by Pol IV and is simultaneously converted into double-stranded form by an RNA-dependent RNA polymerase, RDR2. The short dsRNA substrate is processed into 24-nt siRNA preferentially by DCL3 due to its length specificity, facilitating the subsequent RNA-directed DNA methylation process (Blevins et al. 2015; Zhai et al. 2015). The transcription of short RNAs by Pol IV, and the length and 5-adenosine substrate preference of DCL3 might be essential mechanisms to prevent other DCLs from processing specific dsRNA substrate needed for the TGS pathway. Such coupling of RDR-Dicer-RNAi is also known in fission yeast, where a Dicer physically interacts with an RNA-dependent RNA polymerase to form coupled ma chinery that drives siRNA-mediated TGS (Colmenares et al. 2007). In addition, DCL3 can participate in 24-nt siRNA production from longer transcripts with aid from another RNase III enzyme, RNase III-like 2 (RTL2). As a class II RNase III enzyme, RTL2 possesses one RNase III domain and two dsRBDs, and is involved in rRNA maturation [in vivo is implied]in A. thaliana (Comella et al. 2008). Recombinant RTL2 protein can cleave long dsRNA substrates into 25 bp or longer dsRNA in vitro (Kiyota et al. 2011). Recently, it has been shown that RTL2 processes a subset of Pol IV-dependent dsRNA into shorter intermediates, which are preferable for DCL3 activity in vivo (Elvira-Matelot et al. 2016). Although no direct interaction has been reported, RTL2 and DCL3 can be considered other examples of coordinated action of a dsRBD-containing protein and a Dicer in plants. DCL3 is also reported to physically interact with the dsRNA-binding protein DRB3 in the antiviral RNA-directed DNA methylation pathway (Raja et al. 2014). The function of DRB3 in DCL3 activity rem ains elusive. Inorganic phosphate, NaCl and KCl differentially regulate DCL3 and DCL4 activities In the same assay system using crude extracts, inorganic phosphate at a physiological concentration promotes DCL3 activity but suppresses DCL4 activity toward 50-nt dsRNA substrates (Nagano et al. 2014). The differential effe