The Significance of the DUF283 Domain for the Activity of Human Ribonuclease Dicer

Dicers are multidomain proteins, usually comprising an amino-terminal putative helicase domain, a DUF283 domain (domain of unknown function), a PAZ domain, two RNase III domains (RNase IIIa and RNase IIIb) and a dsRNA-binding domain. Dicer homologs play an important role in the biogenesis of small regulatory RNAs by cleaving single-stranded precursors adopting stem-loop structures (pre-miRNAs) and double-strand RNAs into short RNA duplexes containing functional microRNAs or small interfering RNAs, respectively. Growing evidence shows that apart from the canonical role, Dicer proteins can serve a number of other functions. For example, results of our previous studies showed that human Dicer (hDicer), presumably through its DUF283 domain, can facilitate hybridization between two complementary RNAs, thus, acting as a nucleic acid annealer. Here, to test this assumption, we prepared a hDicer deletion variant lacking the amino acid residues 625-752 corresponding to the DUF283 domain. The respective 128-amino acid fragment of hDicer was earlier demonstrated to accelerate base-pairing between two complementary RNAs in vitro. We show that the ΔDUF(625-752) hDicer variant loses the potential to facilitate RNA-RNA base pairing, which strongly proves our hypothesis about the importance of the DUF283 domain for the RNA-RNA annealing activity of hDicer. Interestingly, the in vitro biochemical characterization of the obtained deletion variant reveals that it displays different RNA cleavage properties depending on the pre-miRNA substrate.


Introduction
Dicer belongs to the ribonuclease III (RNase III) family of double stranded RNA (dsRNA)-specific endoribonucleases [1]. Dicer proteins generate short RNA duplexes containing functional microRNAs (miRNAs) or small interfering RNAs (siRNAs) [2], which are loaded into a multi-protein complex referred to as the RNA-induced silencing complex (RISC) [3]. During RISC activation, one strand of the RNA duplex (called the "passenger" strand) is released and degraded and the other strand (called the "guide" strand) remains in the complex and acts as a sequence-specific probe guiding RISC to complementary transcripts [4]. Depending on the degree of complementarity between the small RNA (miRNA or siRNA) and the targeted transcript, RISC binding results in either translational repression or mRNA cleavage and degradation [5]. The vast majority of cellular processes are controlled by miRNAs or siRNAs, e.g., developmental timing, growth control, differentiation [6], apoptosis, chromatin rearrangements [7] and even viral defense [8]. Consequently, a disruption of Dicer activity can initiate pathological processes like carcinogenesis, neurodegenerative, rheumatic or immune system disorders [1,[9][10][11].
Human Dicer (hDicer) contains 1992 amino acid residues (220-kDa) and is one of the most structurally complex members of the RNase III family. It comprises an amino (N)-terminal helicase domain, a domain of unknown function (DUF283), Platform, Piwi-Argonaute-Zwille (PAZ) domain, a Connector helix, two RNase III domains (RNase IIIa

Production of the hDicer DUF283 Deletion Variant
To investigate whether the DUF283 domain is indeed indispensable for hDicer RNA-RNA base pairing activity, we obtained the hDicer variant lacking the amino acid residues 625-752 corresponding to the DUF283 domain (a variant named ∆DUF(625-752)) ( Figure 1a). This 128-amino acid fragment of hDicer was earlier demonstrated to accelerate base-pairing between two complementary RNA or DNA molecules in vitro [17]. The expression plasmid encoding the hDicer ∆DUF(625-752) variant was produced by a PCR approach using as a template the plasmid encoding the wild-type full-length hDicer fused with the 3xFlagtag (called "hDcr") that was earlier obtained in our laboratory [25]) (Figure 1a,b) (for details please see Materials and Methods). The expression plasmid was used for the transfection of 293T NoDice cells (human cells that do not produce hDicer) [26]. A total of 72 h after transfection, the protein was isolated and purified from cell extracts by immunoprecipitation with anti-Flag antibody conjugated to agarose beads (Figure 1c). ) and hDcr preparations. The Cterminally 3xFlag-tagged proteins were expressed in 293T NoDice cells, then they were purified by immunoprecipitation and analyzed by SDS-PAGE followed by Coomassie Blue Staining.

RNase Activity of the hDicer DUF283 Deletion Variant
First, we tested the RNase activity of the ΔDUF(625-752) variant. The variant was assayed for the RNase activity using two pre-miRNA substrates: pre-mir-21 and pre-mir-16-1 and a 30-base pair (bp) RNA duplex having 2-nt 3′-overhanging ends (called "dsRNA"). The cleavage assays involved 18 nM of the protein and ~5 nM of either 5′-32 Plabeled pre-mir-21, pre-mir-16-1 or dsRNA in which one of the two strands was 5′-32 Plabeled. Thus, the reactions were performed under the low-turnover conditions, i.e., more than threefold molar excess of hDcr over a substrate was used. Two control reactions without the protein were also prepared: one containing only the substrate in the reaction buffer (C-) and the other containing the substrate in the reaction buffer with the addition of the Mg 2+ -chelating agent, EDTA (25 mM), (C+). As the Dicer cleavage activity is dependent on Mg 2+ , EDTA would abrogate this activity of Dicer. Thus, another control reactions included the substrate, protein and 25 mM EDTA (+EDTA). Yet another control contained the substrate and 18 nM hDcr. All reactions were carried out at 37 °C and then they were halted by adding 1 volume of 7 M UREA loading buffer and heating for 5 min at 95 °C. Reaction mixtures were separated on a 15% polyacrylamide gel with 7 M urea and 1xTBE (PAGE) and visualized by phosphorimaging ( Figure 2). The cleavage assays were conducted in triplicate. ) and hDcr preparations. The C-terminally 3xFlag-tagged proteins were expressed in 293T NoDice cells, then they were purified by immunoprecipitation and analyzed by SDS-PAGE followed by Coomassie Blue Staining.

RNase Activity of the hDicer DUF283 Deletion Variant
First, we tested the RNase activity of the ∆DUF(625-752) variant. The variant was assayed for the RNase activity using two pre-miRNA substrates: pre-mir-21 and pre-mir-16-1 and a 30-base pair (bp) RNA duplex having 2-nt 3 -overhanging ends (called "dsRNA"). The cleavage assays involved 18 nM of the protein and~5 nM of either 5 -32 P-labeled pre-mir-21, pre-mir-16-1 or dsRNA in which one of the two strands was 5 -32 P-labeled. Thus, the reactions were performed under the low-turnover conditions, i.e., more than threefold molar excess of hDcr over a substrate was used. Two control reactions without the protein were also prepared: one containing only the substrate in the reaction buffer (C-) and the other containing the substrate in the reaction buffer with the addition of the Mg 2+ -chelating agent, EDTA (25 mM), (C+). As the Dicer cleavage activity is dependent on Mg 2+ , EDTA would abrogate this activity of Dicer. Thus, another control reactions included the substrate, protein and 25 mM EDTA (+EDTA). Yet another control contained the substrate and 18 nM hDcr. All reactions were carried out at 37 • C and then they were halted by adding 1 volume of 7 M UREA loading buffer and heating for 5 min at 95 • C. Reaction mixtures were separated on a 15% polyacrylamide gel with 7 M urea and 1xTBE (PAGE) and visualized by phosphorimaging ( Figure 2). The cleavage assays were conducted in triplicate.   We found that in the assay with pre-mir-21, ∆DUF(625-752) generated only traces of 21-nt miRNA products (faint bands observed at the position of miR-21 generated by hDcr in the control experiment). Under the same reaction conditions, within 2 h incubation, hDcr processed almost all of the substrate (Figure 2a). In the case of the reactions carried out with the pre-mir-16-1 substrate, both ∆DUF(625-752) and hDcr produced efficiently miR-16-1. After 2 h incubation time, all pre-mir-16-1 was cut by both enzymes (Figure 2b). For the cleavage assay involving 30-bp dsRNA and ∆DUF(625-752), we noticed faint bands migrating as fast as 21-nt RNA that was generated efficiently by hDcr in the control experiment ( Figure 2c). Collected data revealed that the deletion of amino acids 625-752 affects hDicer's ability to cleave pre-mir-21 and 30-bp dsRNA, but not pre-mir-16-1, which suggests that this fragment of hDicer might be involved in substrate recognition and discrimination.
The aforementioned experiments were carried out under the in vitro conditions, consequently we asked the question about the ability of the ∆DUF(625-752) variant to process pre-mir-21 and pre-mir-16-1 under in cellulo conditions. By applying the RT-qPCR approach, we assessed the relative level of miR-21-5p and miR-16-1-5p produced in 293T NoDice cells expressing the ∆DUF(625-752) variant. As a negative control, we used 293T NoDice cells treated only with a transfection reagent and as a positive control, we used 293T NoDice cells transfected with the plasmid expressing the wild-type full-length hDicer, hDcr. All cells were harvested 72 h after transfection. We found that, under similar protein expression levels (Figure 3a), both miR-21-5p and miR-16-1-5p were less abundant in 293T NoDice cells expressing ∆DUF(625-752), than in 293T NoDice cells expressing hDcr (Figure 3b,c). Precisely, a reduction in miRNA level was by about 40% (for miR-21-5p) and by about 30% (for miR-16-1-5p) compared to the respective positive control reactions. Thus, we found that miR-21-5p and miR-16-1-5p were produced with different efficiency, depending on the used system: in vitro (Figure 2a hDicer, apart from the canonical substrates, can bind and cut miRNA-size RNA molecules [24,25,[28][29][30]; therefore, we assume that Dicer proteins might be involved in the miRNA turnover [25]. Accordingly, we performed cleavage assays with a 21-nt RNA substrate, called "RNA21". We found that ∆DUF(625-752) generated 15-nt, 6-nt and 2-nt RNA products (Figure 4), as it was earlier observed for hDcr and the hDicer variants lacking the PAZ domain [25]. With incubation time, we observed disappearance of 15-nt and 6-nt RNA products and accumulation of 2-nt RNAs. Based on these results, we conclude that 2-nt RNAs were cut off from 15-nt and/or 6-nt RNAs. Consequently, 15-nt and 6-nt RNAs can be defined as primary cleavage products, while 2-nt RNAs, as secondary cleavage products. We also noticed that, under the adequate reaction conditions, processing of RNA21 was much more efficient in the case of ∆DUF(625-752), compared to hDcr.  hDicer, apart from the canonical substrates, can bind and cut miRNA-size RNA molecules [24,25,[28][29][30]; therefore, we assume that Dicer proteins might be involved in the miRNA turnover [25]. Accordingly, we performed cleavage assays with a 21-nt RNA substrate, called "RNA21". We found that ΔDUF(625-752) generated 15-nt, 6-nt and 2-nt RNA products (Figure 4), as it was earlier observed for hDcr and the hDicer variants lacking the PAZ domain [25]. With incubation time, we observed disappearance of 15-nt and 6-nt RNA products and accumulation of 2-nt RNAs. Based on these results, we conclude that 2-nt RNAs were cut off from 15-nt and/or 6-nt RNAs. Consequently, 15-nt and 6-nt RNAs can be defined as primary cleavage products, while 2-nt RNAs, as secondary cleavage products. We also noticed that, under the adequate reaction conditions, processing of RNA21 was much more efficient in the case of ΔDUF(625-752), compared to hDcr.   hDicer, apart from the canonical substrates, can bind and cut miRNA-size RNA molecules [24,25,[28][29][30]; therefore, we assume that Dicer proteins might be involved in the miRNA turnover [25]. Accordingly, we performed cleavage assays with a 21-nt RNA substrate, called "RNA21". We found that ΔDUF(625-752) generated 15-nt, 6-nt and 2-nt RNA products (Figure 4), as it was earlier observed for hDcr and the hDicer variants lacking the PAZ domain [25]. With incubation time, we observed disappearance of 15-nt and 6-nt RNA products and accumulation of 2-nt RNAs. Based on these results, we conclude that 2-nt RNAs were cut off from 15-nt and/or 6-nt RNAs. Consequently, 15-nt and 6-nt RNAs can be defined as primary cleavage products, while 2-nt RNAs, as secondary cleavage products. We also noticed that, under the adequate reaction conditions, processing of RNA21 was much more efficient in the case of ΔDUF(625-752), compared to hDcr.  To investigate how removing the DUF283 domain from hDicer influences its RNA-RNA base pairing activity, we carried out the annealing assay involving a pair of complementary RNAs: 21-nt RNA (RNA21) and a 50-nt RNA (called "RNA50") adopting a hairpin structure and containing the fully complementary RNA21 target site, schematically described in Figure 5a. This pair of complementary RNAs was used in our earlier annealing assays [24]. Before the reaction, RNA50 was incubated at 95 • C for 5 min and then slowly cooled to room temperature to ensure proper folding. Then, RNA50 was mixed in annealing buffer with 5 -32 P-labeled RNA21 at a molar ratio of approximately 1:1 and incubated for 30 min with increasing amounts of ∆DUF(625-752) (1.88, 3.75, 7.5, 15, 30 nM) at 37 • C (Figure 5b). The same set of reactions was carried out for hDcr. Spontaneous annealing was determined by excluding the enzyme in the assay mixture (control reaction (C-)). Based on the results obtained from three independent experiments, for each reaction we calculated the percentage ratios between the fraction containing the RNA21-RNA50 duplex and the free RNA21 fraction. The average percentage content of the RNA21-RNA50 duplex was plotted against the protein concentration (Figure 5b, right panel). The collected results demonstrated that, in contrast to hDcr, ∆DUF(625-752) did not support base pairing between RNA21 and RNA50 under the applied reaction conditions. Neither we observed the RNA-RNA base pairing activity for the ∆DUF(630-709) variant ( Figure S2). Altogether, collected data strongly support our hypothesis that the DUF283 domain is indispensable for hDicer RNA-RNA annealing activity.
of the reaction buffer with 25 mM EDTA. The asterisk indicates the 32 P 5′-end label. The reproducible results were obtained using at least two batches of recombinant ΔDUF(625-752) and hDcr.

The RNA-RNA Base Pairing Potential of the hDicer DUF283 Domain Deletion Variant
To investigate how removing the DUF283 domain from hDicer influences its RNA-RNA base pairing activity, we carried out the annealing assay involving a pair of complementary RNAs: 21-nt RNA (RNA21) and a 50-nt RNA (called "RNA50") adopting a hairpin structure and containing the fully complementary RNA21 target site, schematically described in Figure 5a. This pair of complementary RNAs was used in our earlier annealing assays [24]. Before the reaction, RNA50 was incubated at 95 °C for 5 min and then slowly cooled to room temperature to ensure proper folding. Then, RNA50 was mixed in annealing buffer with 5′-32 P-labeled RNA21 at a molar ratio of approximately 1:1 and incubated for 30 min with increasing amounts of ΔDUF(625-752) (1.88, 3.75, 7.5, 15, 30 nM) at 37 °C (Figure 5b). The same set of reactions was carried out for hDcr. Spontaneous annealing was determined by excluding the enzyme in the assay mixture (control reaction (C-)). Based on the results obtained from three independent experiments, for each reaction we calculated the percentage ratios between the fraction containing the RNA21-RNA50 duplex and the free RNA21 fraction. The average percentage content of the RNA21-RNA50 duplex was plotted against the protein concentration (Figure 5b, right panel). The collected results demonstrated that, in contrast to hDcr, ΔDUF(625-752) did not support base pairing between RNA21 and RNA50 under the applied reaction conditions. Neither we observed the RNA-RNA base pairing activity for the ΔDUF(630-709) variant ( Figure  S2). Altogether, collected data strongly support our hypothesis that the DUF283 domain is indispensable for hDicer RNA-RNA annealing activity.

Discussion
Initially, it was suggested that the DUF283 domain is critical for pre-miRNA processing, because the in vitro cleavage activity is lost in human and Drosophila melanogaster Dicer variants lacking both DUF283 and the helicase domain [31,32]. However, subsequent in vitro studies demonstrated that the deletion of only DUF283 decreases the cleavage of dsRNAs by hDicer, without affecting the cleavage of pre-miRNA substrates [27]. The mentioned DUF283 deletion variant, produced by Ma E. and colleagues, lacks the region spanning 630-709 amino acid residues (the ∆DUF(630-709) variant), ( Figure S1) [27]. The DUF283 deletion variant produced in our laboratory lacks the region between amino acids 625 and 752, including a linker joining the DUF283 domain to Platform (the ∆DUF(625-752) variant), (Figure 1a,b). Although removing this linker might affect the spatial organization of hDicer domains, we found that the ∆DUF(625-752) variant is successfully expressed in NoDice cells (Figure 1c). After purification, this protein retained its integrity and activity up to 12 months when stored at −20 • C. The three-dimensional structure of Dicer resembles the letter L [18,33]. Within this structure, the head, core and base can be distinguished. The head constitutes the PAZ and the Platform domains, the RNase III domains are in the core, whereas the helicase domain forms the base (Figure 1b). We can deduce that removing the linker, that joins the base to the head, precludes a proper positioning of the base (i.e., the helicase domain) in relation to the head (i.e., PAZ and Platform). Consequently, we assume that the orientation of the helicase domain differs between the ∆DUF(625-752) variant and the wild-type hDicer or the ∆DUF(630-709) variant. However, the helicase domain was shown to be dispensable for processing of both pre-miRNA and dsRNA substrates by hDicer [27]. Thus, mispositioning of the helicase domain should not abolish the cleavage activity of the ∆DUF(625-752) variant. Indeed, we found that, under the in vitro conditions, the ∆DUF(625-752) variant processed pre-mir-16-1 as efficiently as the wild-type enzyme (Figure 2b). In contrast to the case of pre-mir-16-1, the deletion of amino acids 625-752 compromised hDicer ability to cleave pre-mir-21 and 30-bp dsRNA in vitro (Figure 2a,c). What are the differences between pre-mir-16-1 and pre-mir-21 substrates? We can notice that pre-mir-16-1 contains large internal loops and bulges and has relaxed terminal loop region, whereas pre-mir-21 adopts more compact structure, with the small terminal loop (Figure 2). Therefore, taking into account the content of paired nucleotides, the structure of pre-mir-21 is more similar to the structure of the dsRNA substrate, rather than to the structure of pre-mir-16-1. Single-stranded regions of pre-mir-16-1 provide the greater flexibility of this substrate, compared to the other two substrates. Accordingly, we assume that pre-mir-16-1, due its flexibility, may better fit into the cleavage center of the ∆DUF(625-752) variant. In contrast to ∆DUF(625-752), the ∆DUF(630-709) variant processed all tested canonical substrates at least as efficiently as the wild-type enzyme ( Figure S1). These data further support our hypothesis that the spatial orientation of the domains differs between ∆DUF(625-752) and ∆DUF(630-709) variants. Additionally, the collected data suggest that not only the helicase domain [15], but the DUF283 domain may as well play an important role in substrate recognition and discrimination. Importantly, as mentioned above, the removal of both the helicase and the DUF283 domains was shown to abolish in vitro processing of pre-miRNA substrates by the respectively truncated hDicer variant [31]. Collectively, based on the literature data and results of the cleavage assays generated for ∆DUF(625-752) and ∆DUF(630-709), we hypothesize that factors influencing the spatial orientation of hDicer domains may impact the enzyme's substrate specificity and its cleavage properties.
Nevertheless, in contrast to the in vitro RNase cleavage assays (Figure 2), we found that in NoDice cells expressing ∆DUF(625-752), the levels of miR-21-5p and miR-16-1-5p were similarly decreased (by about 40% and 30%, respectively) compared to the control cell lines producing hDcr (Figure 3). These results indicate the possible involvement of cellular factors, interacting with pre-miRNAs, in the studied process.
We also conclude that the RNase III cleavage center of the ∆DUF(625-752) variant is accessible for miRNA/siRNA-size RNAs (Figure 4). The collected data showed that ∆DUF(625-752) generates the same cleavage pattern for 21-nt RNA as the wild-type enzyme and the PAZ deletion variant, which was reported in our previous studies [25]. Accordingly, taking into account the results of this and earlier studies [25], we deduce that the cleavage of miRNA-size RNAs by hDicer neither involves DUF283 nor PAZ domains. As we proposed earlier, we believe that Dicer might be involved in the miRNA turnover by promoting degradation of passenger strands that are discarded from the RISC complex [25].
Finally, we found that the deletion of the DUF283 domain (either the fragment spanning amino acids 625-752 or the fragment spanning amino acids 630-709) abolishes hDicer RNA-RNA annealing activity in vitro ( Figure 5) and ( Figure S2). Altogether the presented data provide further insight into the role of the DUF283 domain for the activity of human ribonuclease Dicer. The next challenge will be to translate all these results into the in vivo activity of Dicer-type proteins.

Oligonucleotides
DNA primers and RNA oligonucleotides were purchased from Genomed (Warsaw, Poland) and FutureSynthesis (Poznan, Poland), respectively. Sequences of all oligonucleotides used in this study are listed in Table 1.

32 P Labeling of Oligonucleotides
The 5 -end labeling of oligonucleotides was performed using 10 pmol of RNA, 1 µL γ-32 P-ATP (Hartman Analytic GmbH, Braunschweig, Germany) and 10 U T4 polynucleotide kinase (Thermo Fisher Scientific, Waltham, MA, USA) in a final volume of 10 µL. The composition and reaction conditions were in accordance with the manufacturer's instructions. Before preparing the reaction mixture, the RNA was denatured by heating to 95 • C and rapidly cooling on ice. The radiolabeled oligonucleotides were PAGE-purified in 8% denaturing polyacrylamide gels and resuspended in water to a final concentration of approximately 10,000 cpm/µL. The homogeneity of the material after gel purification was checked by electrophoresis in 15% PAA under denaturing conditions. Gels were exposed to a phosphorimager plate, which was subsequently scanned with FLA-5100 Fluorescent Image Analyzer (Fujifilm, Minato, Tokyo, Japan) to visualize the bands.

Preparation of dsRNA
To prepare a dsRNA substrate, non-labeled strand (RNA_sense) was hybridized, at a molar ratio of approximately 1:1, with 32 P-labeled complementary strand (RNA_ov) in buffer containing 50 mM NaCl, 2.5 mM MgCl 2 and 20 mM Tris-HCl, pH 7.5, by heating up to 95 • C and then slowly cooling down to room temperature. Next, the reaction mixtures were analyzed on 10% native polyacrylamide gels to check whether the double-stranded complexes were free of single-stranded species.

Immunoprecipitation
Cells were harvested 72 h after transfection by centrifugation at 1500 rpm for 3 min and suspended in lysis buffer (30 mM Hepes pH 7.4, 100 mM KCl, 5 mM MgCl 2 , 10% glycerol, 0.5 mM DTT and 0.2% Tergitol) containing 1 × protease inhibitor mix without EDTA (Roche, Warsaw, Poland) and broken by passing through a 0.9 mm × 40 mm needle. Lysates were centrifuged at 13,000 rpm for 5 min at 4 • C. The supernatant was incubated overnight on a rotator with ANTI-FLAG ® M2 Affinity Gel (Merck, Darmstadt, Germany) that was pre-washed with TBS buffer. After incubation, the beads were washed five times with TBS buffer. Isolated protein was eluted with 100 µg/mL 3XFlag Peptide (Sigma, Kawasaki, Japan). Purified protein was suspended in the buffer (20 mM Tris pH 7.5, 50 mM NaCl, 10% glycerol and 0.25% Triton X-100). Protein concentration was estimated by the Bradford assay (BioRad, Irvine, CA, USA), relative to a BSA standard curve and on SDS-PAGE with BSA as a standard. Proteins were concentrated using Amicon filters (Merck) in the respective buffer (as described above) enriched with 40% glycerol and stored at −20 • C.

Western Blot Analysis
Obtained proteins were separated on 8% SDS-PAGE and electro-transferred onto a PVDF membrane (Thermo Fisher Scientific, Waltham, MA, USA). For hDicer and hDicer deletion variant, the blots were probed with a mouse monoclonal primary anti-Dicer antibody mapping at the C-terminus of hDicer (1:300, Santa Cruz Biotechnology, Dallas, TX, USA), for β-Actin, the blots were probed with a rabbit monoclonal primary anti-β-Actin antibody (1:1100, Cell Signaling Technology, Danvers, MS, USA) and subsequently with HRP-conjugated secondary antibody, anti-mouse or anti-rabbit (1:5000, Jackson ImmunoResearch Laboratories, Inc., Cambridgeshire, UK). The immunoreactions were detected using SuperSignal TM West Pico PLUS Chemiluminescent Substrate (Thermo Fisher Scientific, Waltham, MA, USA).

The RNA Cleavage Assay
The cleavage assay was performed in 10 µL reactions containing 50 mM NaCl, 2.5 mM MgCl 2 and 20 mM Tris-HCl, pH 7.5, 5 -32 P-labeled substrate (10,000 cpm, approximately 5 nM) and 18 nM of the protein (hDcr or ∆DUF(625-752), or ∆DUF(630-709)). In addition, a reaction mixture without the protein was prepared as a control. In controls including EDTA, the reaction buffer was supplemented with the chelating agent to the final concentration of 25 mM. Reactions were carried out at 37 • C for 10, 30, 60 and 120 min with the addition of the commercial RNase-inhibitor cocktail (NEB, Ipswich, USA). The reactions were stopped by the addition of 1 volume of 7 M urea loading buffer and heating for 5 min at 95 • C. Samples were separated on a 15% denaturing polyacrylamide gel in 1 × TBE running buffer.

Reverse Transcription and Quantitative PCR
The RNA was extracted using a standard TRIzol protocol (Invitrogen, Waltham, MA, USA). The RNA quantity and quality were measured using NanoDrop 2000 (Thermo Fisher Scientific, Waltham, MA, USA). For reverse transcription (RT), 8 µg of total RNA was used. RT was performed by using Mir-X™ miRNA First-Strand Synthesis Kit (Takara Bio, Canada, USA) in accordance with the manufacturer's instructions. RT-qPCR was performed using iTaq™ Universal SYBR ® Green Supermix (Bio-Rad, Irvine, CA, USA). For miRNA analysis, 8 ng of total cDNA was used. Amplification was performed using the Bio-Rad CFX96™ system (Bio-Rad) and the software determined C t thresholds. The relative expression level of miRNA was evaluated by the 2 −∆∆Ct method [34]. miRNA levels were normalized to U6 small nuclear RNA as a reference gene. Control reactions lacking template were performed to verify clean backgrounds in all samples. The results from three independent experiments were presented as mean ± S.D. One-way ANNOVA test followed by Dunnett's multiple comparisons test was used for statistical evaluation. Statistical analyses were performed using GraphPad Prism 6 Software (GraphPad Software, San Diego, CA, USA). The primer sequences used for qRT-PCR are listed in Table 1.

Gel Imaging and Analysis
The data were collected using a Fujifilm FLA-5100 Fluorescent Image Analyzer (Fujifilm, Minato, Tokyo, Japan). The amounts of 32 P-labeled substrates and products were determined from the intensity of the respective bands in the gels measured by