Essential role of the amino-terminal region of Drosha for the Microprocessor function

Summary Drosha is a core component of the Microprocessor complex that cleaves primary-microRNAs (pri-miRNAs) to generate precursor-miRNA and regulates the expression of ∼80 ribosomal protein (RP) genes. Despite the fact that mutations in the amino-terminal region of Drosha (Drosha-NTR) are associated with a vascular disorder, hereditary hemorrhagic telangiectasia, the precise function of Drosha-NTR remains unclear. By deleting exon 5 from the Drosha gene and generating a Drosha mutant lacking the NTR (ΔN), we demonstrate that ΔN is unable to process pri-miRNAs, which leads to a global miRNA depletion, except for the miR-183/96/182 cluster. We find that Argonaute 2 facilitates the processing of the pri-miR-183/96/182 in ΔN cells. Unlike full-length Drosha, ΔN is not degraded under serum starvation, resulting in unregulated RP biogenesis and protein synthesis in ΔN cells, allowing them to evade growth arrest. This study reveals the essential role of Drosha-NTR in miRNA production and nutrient-dependent translational control.


INTRODUCTION
Drosha and DiGeorge syndrome critical region 8 (Dgcr8) are components of the Microprocessor, a complex responsible for the biogenesis of miRNAs. 1,2The RNase III enzyme Drosha processes long primary-miRNAs (pri-miRNAs) to generate precursor-miRNAs (pre-miRNAs) with hairpin structures in the nucleus, 1,2 which then undergo secondary processing by Dicer in the cytoplasm to generate small RNA duplexes of $22-nucleotides (nt). 3The RNA duplexes are then loaded onto an Argonaute (Ago) protein to form an RNA-induced silencing complex (RISC), which unwinds RNA duplexes, removes the passenger strands, binds 3 0 UTR of mRNAs through the sequence element that is partially complementary to the miRNA sequence, and mediates destabilization and translational repression of target mRNAs. 3The Microprocessormediated processing is regulated by physiological stimuli, for example, upon activation of the TGF-b-family of growth factor signaling. 4 All four Ago proteins (Ago1-4) incorporate miRNAs in RISC, but Ago2 is distinct because it is capable of miRNA-directed target RNA cleavage through its intrinsic RNA slicing activity. 5It has also been found that Ago2, rather than Dicer, processes pre-miR-451, [6][7][8] suggesting that the slicing activity of Ago2 has broader functions beyond miRNA-mediated mRNA silencing.
The Microprocessor activity is regulated by various proteins that associate with the Microprocessor, such as the Smads. 4,22Smads 1, 5, and 8, the signal transducers of the bone morphogenetic proteins (BMPs) signaling pathway, interact with Drosha upon BMP4 stimulation and promote the processing of specific miRNAs, such as miR-21 and miR-199a. 23IP-immunoblot analyses showed that Smad1/5/8 proteins, phosphorylated by the BMP receptor kinase upon BMP4 stimulation (p-Smad), associated with FL-Drosha but not with DN-Drosha (Figure 2E, IP), although p-Smads were detected in the input samples of DN cells stimulated with BMP4 (Figure 2E, Input).In FL cells, the levels of pre-miR-21 and pre-miR-199a increased 9-fold and 7-fold, respectively, upon BMP4 stimulation (Figure 2F, black bars).However, neither pre-miR-21 nor pre-miR-199a showed any increase in DN cells after BMP4 treatment (Figure 2F, red bars).These results indicate the Drosha-NTR is essential for the BMP4-dependent increase in Microprocessor activity through association with Smad proteins.

Increased production of RPs in DN cells
The Microprocessor potentiates the transcription of RPGs by binding to the 5 0 TOP sequence shared by all $80 RPGs. 18A ChIP assay indicated that both FL-and DN-Drosha associate with the Rps2, Rps10, and Rpl28 gene loci at similar levels (Figure 3A), indicating that the Drosha-NTR is dispensable for interaction with RPGs.We showed that, upon serum starvation, Drosha translocates from the nucleus to the cytoplasm and is degraded through the association of an E3 ubiqutin ligase Nedd4 at Drosha-NTR, resulting in the repression of RPG transcription. 18Consistently, the nuclear localization of FL-Drosha decreased from 36.3% to 8.7%, however, DN-Drosha remained in the nucleus upon serum starvation (Figure 3B).Furthermore, 16 h after serum starvation, the amount of FL-Drosha decreased to 24% (Figure 3C).The levels of both RPs (Figure 3C) and RP mRNAs (Figure S4A) in FL cells also decreased after starvation as previously reported. 18However, the amount of DN-Drosha remained the same after starvation (Figure 3C).Furthermore, the levels of RPs (Figure 3C) and RP mRNAs (Figure S4A) remained high in DN cells after starvation.The amount of Gata1 protein, which is sensitive to changes in ribosome abundance, 18 was reduced by 80% in FL cells after serum starvation (Figure 3C).In DN cells, however, Gata1 protein decreased only by 3% (Figure 3C).This is consistent with the absence of the reduction of RPs under starvation (Figure 3C).When FL-Drosha was exogenously expressed in DN cells, a reduction of FL-Drosha, RPs (Figure 3D), and RP mRNAs (Figure S4B) was observed following serum starvation.This indicates that exogenous FL-Drosha is capable of rescuing the control of RP biogenesis in DN cells.In contrast, when DN-Drosha was introduced in DN cells, the levels of RPs (Figure 3D), RP mRNAs (Figure S4B) and DN-Drosha remained unchanged under starvation.These results underscore the essential role of Drosha-NTR in controlling RP biosynthesis in response to changes in nutrient availability.An in vitro puromycin incorporation assay 24 showed that the amount of puromycin-incorporated nascent proteins was reduced to 38% in FL cells after serum starvation for 16 h (Figure 3E), indicating a reduced translation.In contrast, there was no decrease in translation upon starvation in DN cells (Figure 3E), which aligns with the continued RP biogenesis in DN cells during serum starvation (Figure 3C).When the proliferation of FL and DN cells was compared under normal growth conditions (10% serum) and starvation conditions (1% serum), the doubling time (Td) of FL cells increased from 20 h (10% serum) to 56 h (1% serum) (Figure 3F, black lines), indicating slower proliferation under serum starvation conditions.In contrast, DN cells proliferated at a consistent rate in both 10% serum (Td = 21 h) and 1% serum (Td = 21 h) conditions (Figure 3F, red lines).These results demonstrate that while the Drosha-NTR is dispensable for the association with RPGs, it is essential for starvation-induced degradation of Drosha.Consequently, DN cells cannot modulate ribosome abundance and cell proliferation rate to adapt to the changes in the environment.

Ago2-dependent processing of the miR-183 cluster
Among a small number of miRNAs expressed in DN cells at a level equivalent or slightly higher than FL cells were miR-183, -96, and -182 (Figure 2B, left; Figures S5A and S5B).These miRNAs belong to the miR-183 cluster and are transcribed as a single long polycistronic transcript (pri-miR-183) with three hairpin structures corresponding to mature miR-183, -96, and -182. 25qRT-PCR confirmed the small RNA-seq data that the levels of these miRNAs were slightly higher in DN cells compared to FL cells (Figure 4A).Consistently, we observed reduced mRNA levels of targets of miR-183/96/182 in DN cells compared to FL cells (Figure S5C). 25 The levels of pri-miR-183 transcripts were similar in FL and DN cells (Figure 4A), indicating no difference in the miR-183 cluster gene transcription between FL and DN cells.When Drosha was depleted by siRNA (siDrosha) in FL cells, the levels of miR-183 cluster miRNAs were slightly elevated (Figure 4B) while the levels of control miRNAs (miR-21, -24, -105, and -330) were diminished (Figure 4B; Figure S6).The small RNA-seq data revealed a large fraction of miR-183 and miR-96 sequence variants (isomiRs) (Table 1), which contain 1 or 2 additional nucleotides at the 5 0 -end with respect to the reference sequence (Table 1), in DN cells, suggesting that the cleavage of these pri-miRNAs in DN cells occurred 1-or 2-nt upstream of the conventional Drosha cleavage site.Unlike miR-183 and miR-96, we did not find 5 0 -end sequence variants of miR-21, miR-10a, miR-34a, and miR-105 in FL or DN cells (Table 1).
Together with the result of IVP assay (Figure 2C), these results support the hypothesis that an enzyme other than Drosha processes hairpins in pri-miR-183 when Drosha is depleted or unable to bind pri-miRNA like DN-Drosha.
Previously, high-throughput sequencing following cross-linking immunoprecipitation (HITS-CLIP) analysis detected an association of Ago2 with miR-183/96/182. 26Ago2 is uniquely capable of directly cleaving RNA targets 5,27,28 and localizes in both the cytoplasm and the nucleus. 29e found that $40% of Ago2 is localized in the nucleus in both FL and DN cells (Figure S7).Therefore, we hypothesized that nuclear Ago2 might be processing miR-183/96/182 hairpins under the circumstance that Drosha amount or activity is compromised.As expected, the RIP assay showed an association of FL-Drosha with both pri-miR-183 and pri-miR-21 (control) in FL cells (Figure 4C, black, siCtrl).This interaction was diminished upon Drosha depletion (Figure 4C, black, siDrosha), however, Ago2 depletion did not affect the interaction in FL cells (Figure 4C, black, siAgo2).Unlike in FL cells, no association of DN-Drosha with pri-miR-183 or pri-miR-21 was detected in DN cells (Figure 4C, red, siCtrl), despite the comparable amounts of pri-miR-183 or pri-miR-21 between DN and FL cells (Figure S8A) and 7-fold higher amount of DN-Drosha than FL-Drosha was immunoprecipitated in the RIP samples (Figure S9).The RIP assay confirmed the association of Ago2 with pri-miR-183 in DN cells (Figure 4D, red, siCtrl).When Ago2 was depleted by siRNA (siAgo2), the RIP signal in DN cells (siCtrl) diminished (Figure 4D, red, siAgo2), confirming the specific detection of the Ago2-pri-miR-183 interaction by RIP.Ago2 did not associate with pri-miR-21 in either DN-Drosha or FL Although similar amounts of Ago2 protein were immunoprecipitated in DN and FL cells (Figure S8B), the Ago2-pri-miR-183 interaction was not detected in FL cells presumably because FL-Drosha prevented the association of Ago2 with pri-miR-183 (Figure 4D, black, siCtrl).Only when FL-Drosha was depleted by siDrosha, the Ago2-pri-miR183 interaction became detectable in FL cells (Figure 4D, black, siDrosha).When exogenous FL-Drosha was introduced at 3.3-fold higher amount than the amount of DN-Drosha into DN cells (Figure 4E, top right), the Ago2-pri-miR-183 interaction was diminished (Figure 4E, top left) and, instead, the association of FL-Drosha with pri-miR-183 was detected (Figure 4E, top middle).It is also noted that the levels of Dgcr8 mRNA was reduced by the introduction of FL-Drosha in DN cells (Figure 4E, top right).These results indicate that FL-Drosha and Ago2 compete for pri-miR-183 binding.There was no significant change in the levels of miR-183, -96, or -182 by exogenous FL-Drosha in DN cells (Figure 4E, bottom), suggesting that both FL-Drosha and Ago2 are capable of processing pri-miR-183 with similar efficiency.We also detected the association of Drosha with pri-miR-21 (Figure 4E, top middle) and the amount of miR-21 was elevated 2.2-fold (Figure 4E, bottom) as expected when FL-Drosha was exogenously expressed in DN cells.These results indicate that exogenous FL-Drosha is capable of associating with pri-miR-21 and promoting the biogenesis of miR-21 in DN cells, providing evidence that DN cells do not lack any essential components of the Microprocessor or miRNA biogenesis pathway.Unlike Ago2, Dgcr8 interacted with both pri-miR-183 and pri-miR-21 in FL-Drosha and DN cells at similar levels, which were not affected by the depletion of Ago2 or Drosha (FL or DN) (Figures S10A and S10B), indicating the association of Dgcr8 with pri-miRNAs is independent of Drosha or Ago2.Depletion of Ago2 by siAgo2 did not alter the amount of miR-183 cluster miRNAs in FL cells (Figure 4F, black), but significantly diminished it in DN cells (Figure 4F, red).When wild type (WT) Ago2 was introduced to DN cells, in which endogenous Ago2 had been reduced by siRNA to $40%, the levels of miR-183, -96, and -182 increased $2to 3-fold compared to cells transfected with a control vector (vec) (Figure 4G).In contrast, when the endonuclease-inactive Ago2 mutant (D597A) 27 was introduced to DN cells in an amount similar to Ago2 (WT), the levels of these miRNAs remained similar to those in control cells (Figure 4G).The amount of miR-21 (control) was affected by neither Ago2 (WT) nor Ago2 (D597) (Figure 4G).These results demonstrate that the processing of the miR-183 cluster, but not other miRNAs, is dependent on the catalytic activity of Ago2.The IVP assay confirmed that the depletion of Drosha or Ago2 had no effect on the processing of pri-miR-183 in FL cells (Figure 4H; Figure S11).However, in DN cells, the levels of pri-miR-183 processing decreased by 50% and 82% when Ago2 and Drosha+Ago2 were depleted, respectively (Figure 4H; Figure S11).In contrast, pri-let7b (control) processing was reduced by 65% and 64% when Drosha and Drosha+Ago2 were depleted in FL cells, respectively (Figure 4H; Figure S11).When Ago2 was immunopurified from DN cells and added to the IVP reaction, pri-miR-183 processing was $2-fold more efficient compared to that by FL-Drosha (Figure 4I).In contrast, the cleavage of pri-let7b (control) by either Ago2 or DN-Drosha remained lower than that of FL-Drosha (Figure 4I).These results demonstrate Ago2-dependent processing of pri-miR-183 in DN cells.IP-immunoblot analysis detected the Drosha (FL and DN)-Dgcr8 association (Figure 4J, green boxes) and the Dgcr8-Ago2 association (Figure 4J, yellow boxes), but not the Drosha-Ago2 association (Figure 4J, blue boxes).These results suggest that the association of Ago2 with pri-miR-183 is facilitated by the physical interaction between Ago2 and Dgcr8.Furthermore, RIP assay showed that the Ago2-pri-miR-183 interaction was diminished (Figure 4K, siDgcr8) when Dgcr8 was knocked down in DN cells (Figure S10C).These data uncover a noncanonical processing of pri-miR-183 by Ago2 and Dgcr8 when Drosha is absent or missing the NTR and unable to associate with pri-miRNAs.

The miR-183 hairpin-dependent processing of the miR-96 hairpin
According to the miRbase database and the Mfold RNA folding software, the predicted length of the stem structure in the miR-96 hairpin is 27-bp. 25This length is shorter than the optimal stem length of hairpins that are cleaved by Drosha ($35 bp). 30,31Thus, we hypothesized that the processing of suboptimal miR-96 hairpin might be assisted by other hairpins in the same cluster.To test the hypothesis in the endogenous context, we generated two HEK293T mutants (183KO1 and 183KO2), in which biallelic deletion of the miR-183 hairpin was introduced by CRISPR-Cas9 gene editing (Figure S12), and the amounts of miR-183, -96 and -183 were examined by qRT-PCR.miR-183 was not detected in 183KO1 or 183KO2 cells, validating the deletion of the miR-183 hairpin in 183KO clones (Figure 5A).Although the level of miR-182 in 183KO clones was similar to that in WT cells, the level of miR-96 in 183KO clones decreased to 15% of that in WT cells, even though the amount of pri-miR-183 in 183KO and WT cells was comparable (Figure 5A).These results suggest that the processing of the miR-96 hairpin is dependent on the presence of the miR-183 hairpin, which is located 135-nt apart.The association of Drosha with the miR-96 hairpin was undetectable in 183KO cells by RIP assay (Figure 5B, left).This is consistent with the depletion of miR-96 in 183KO cells (Figure 5A).Unlike the miR-96 hairpin, the miR-182 hairpin associated with Drosha in both WT and 183KO cells (Figure 5B, right), indicating that the processing of the miR-182 hairpin, which is located > 4200-nt apart from the miR-183 and miR-96 hairpins, is independent of the miR-183 hairpin.When Drosha was depleted by siDrosha, Ago2 was recruited to the miR-96 hairpin in WT cells, but not in 183KO cells (Figure 5C).In WT cells, the depletion of Drosha did not affect miR-96 levels since its processing was maintained by Ago2 (Figure 5D).However, in 183KO cells, miR-96 levels remained as low as those observed in WT cells (Figure 5D), suggesting that Ago2 requires the presence of the miR-183 hairpin to bind to and process the The average number reads (reads per million; RPM) of the miR-183 cluster (miR-183-5p, -182-5p, -96-5p) and control miRNAs (miR-21-5p, -10a-5p, -34a-5p, and -105-5p) in FL or DN-Drosha cells are shown.Reference sequences of miR-183-5p, -182-5p, -96-5p are based on the miRBase.Extra or missing nucleotides at the 5 0 -end in respect of the reference sequence are shown in bold.The ratio (DN-Drosha/FL) of the isomiR-183 that is 1-nt shorter than the reference sequence is close to the ratio of total number of miR-183-5p reads.The ratio of isomiR-183-5p (10.9 and 7.4) and isomiR-95-5p (5.3) are higher than the ratio of total number of miR-183-5p (1.4) and miR-96-5p (1.0), respectively.No isomiRs of control miRNAs was found.miR-96 hairpin similar to Drosha.These results demonstrate that the processing of the miR-96 hairpin is assisted by the miR-183 hairpin (Figure 5E), a mechanism similar to that observed in the miR-144/451 cluster. 32

DISCUSSION
In this work, we demonstrated that DN-Drosha fails to associate with pri-miRNAs, and thus Drosha-NTR is essential for the Microprocessor activity.A previous in vitro study on an N-terminus truncated Drosha (DN-390; aa 390-1374), also lacking the P-rich region and the R/S-rich region as the DN-Drosha used in this study, showed that processing of pri-let7a1 by DN-390 was comparable to that of Drosha (WT), 9 leading to the conclusion that the Drosha-NTR (aa 1-389) was dispensable for the catalytic activity.Based on the immunoblot presented, though, it is possible that a higher amount of DN-390 than the WT protein might have been used in the IVP assay, decreasing the difference in activity between the two Drosha forms. 9The structure of the NTR-truncated Drosha (aa 353-1372), a partial Dgcr8 (aa 223-751), and a pri-miRNA-16-2 by cryo-electron microscopy demonstrate that the basal tip of the central domain (CED; aa 353-960) of Drosha wraps around the dsRNA-single stranded RNA (ssRNA) junction of the pri-miRNA. 30,31Our result that DN-Drosha, which lacks 43 aa of the N-terminus of the CED, is defective in the association with pri-miRNAs, underscores the significance of the CED for the interaction with pri-miRNAs.Pri-miRNA hairpin structures contain several key structural features that facilitate processing by the Microprocessor. 33,34Our results show that miR-96, which has a suboptimal short stem structure, requires the miR-183 hairpin to facilitate cleavage by Drosha or Ago2.This is similar to the miR-144/451cluster, where the presence of the miR-144 hairpin with an optimal structure aids in the recruitment of the Microprocessor and the cleavage of a suboptimal miR-451 hairpin; the mechanism is called ''cluster assistance.'' 32Given that the miR-144 hairpin and miR-451 hairpin are only 100-nt apart similar to the miR-183 and miR-96 hairpins that is 135-nt apart, the proximity of the two hairpins in the cluster appears to be critical, however, the exact mechanism of ''cluster assisted'' processing remains to be elucidated.miR-182, which has an optimal hairpin structure but is located > 4200-nt apart from the miR-183 and miR-96 hairpins, is bound and cleaved by Drosha regardless of the presence of the miR-183 hairpin and is unable to facilitate the cleavage of the miR-96 hairpin.We found that the miR-3136 hairpin, which has a long stem of 40-nt like the miR-183 and miR-182 hairpins (Figure S13A), does not associate with Ago2 in DN cells (Figures S13B and S13C), indicating that structural characteristics of the miR-183/96/182 cluster hairpins, other than the stem length, play an important role in the recruitment of Ago2.
The nuclear localization of Drosha is mediated by a predicted nuclear localization signal (NLS) in the R/S-rich region. 19An alternatively spliced form of Drosha, missing exon 6 which encodes NLS, localizes both in the nucleus and the cytoplasm. 35,36Furthermore, the phosphorylation of the serine (Ser)-300 or Ser-302 residue by glycogen synthase kinase 3b (GSK3b) is required for the nuclear retention of Drosha. 37,38e previously showed that p38 MAPK-dependent phosphorylation of Drosha at Ser-355 contributes to the nuclear-to-cytoplasmic shuttling of Drosha upon nutrient starvation. 39Because DN-Drosha is missing the NLS, Ser-300, Ser-302, and Ser-355, we predicted that DN-Drosha would localize exclusively in the cytoplasm; however, the results indicate that $70% of DN-Drosha is localized in the nucleus, indicating the presence of additional NLS or nuclear retention signals in aa 396-1374 of Drosha.Despite DN-Drosha being expressed from the endogenous Drosha loci and the mRNA amount being similar between DN-Drosha and FL, we noted the higher protein quantity of DN-Drosha than FL-Drosha.We previously reported that Drosha is degraded upon ubiquitination by Nedd4, 39 which requires the PPGY motif at aa 169-172 located in the NTR of Drosha associating with the WW domains of Nedd4. 39Additionally, it has been reported that the stability of Drosha protein can be modulated by the acetylation of lysine (Lys)-48 in Drosha-NTR by p300, CBP, and GCN5, which competes with ubiquitination of the same Lys residues. 40Because DN-Drosha lacks both the PPGY motif and Lys-48, we speculate that DN-Drosha resists ubiquitin-proteasome-dependent degradation, resulting in a higher protein amount than FL-Drosha.
HHT is an autosomal dominant vascular disorder caused by the loss-of-expression or loss-of-function mutations in the mediators of the BMP signaling pathway, such as Acvrl1, Endoglin, and Smad4. 41Nonsynonymous mutations in the Drosha-NTR, such as P32L, P100L, K226E, and R279L, have been identified in individuals with HHT without other mutations associated with HHT. 16HHT patients with Drosha mutations exhibit a range of vascular defects stemming from abnormal vascular endothelial cell structure and functions, including epistaxis, telangiectasias, and arteriovenous malformations (AVMs). 16Similar to the DN-Drosha, Drosha mutants P100L and R279L fail to associate with Smad proteins and is unable to mediate BMP4-Smad1/5/8 dependent induction of miR-21 and miR-199a. 16Because DN-Drosha is unable to control RP biogenesis, protein synthesis, and cell proliferation upon nutrient deprivation, we speculate that the vascular phenotypes in Drosha mutation carrier patients result from the inability of vascular endothelial cells to generate miRNAs upon BMP stimulation and the inability to adapt to changes in the extracellular environment.It is also plausible that the reduction of Smad4, a common target of miR-182 and miR-183, 42,43 contribute to the vascular remodeling in HHT patients with Drosha mutations.
Although the majority of the miR-183 cluster miRNAs were identical to the reference sequence, DN cells also contained miR-183 and miR-96 isomiRs that were 1-or 2-nt longer at the 5 0 -end.This suggests that the Ago2-dependent cleavage site is less specific than the Drosha cleavage site and the addition of extra nucleotides at the 5 0 -end of isomiR-183 might alter the seed sequence, which result in the alternative recognition and/or silencing of target mRNAs in DN cells.The miR-183 cluster is abundantly expressed in the retina and plays an essential role in its development and homeostasis. 456][47][48] It has been reported that Ago2 is present in both retinal neurons and glia, 26 and depletion of Ago2 results in the reduction of miR-182 and miR-183. 49Furthermore, an HITS-CLIP assay using whole retina lysates find that the miR-183 cluster is the most abundant group of miRNAs bound to Ago2, 26 which is consistent with our finding of the Ago2 role in pri-miR-183 processing.In the retina, depletion of Ago2 results in a reduction of miR-183 and -182 and retina degeneration, 49 despite the presence of Drosha, suggesting that Ago2-mediated pri-miR-183 processing might be a predominant mechanism in the retina, similar to Ago2-dependent pre-miR-451 processing in erythrocytes. 6Increased expression of the miR-183 cluster is associated with various human disorders, including cancer, autoimmune diseases, and neuronal diseases. 25Furthermore, an increase in Ago2 protein amount along with post-translational modifications of Ago2, such as phosphorylation of tyrosine and serine residues and acetylation of lysine residues, is associated with poor prognosis and survival of cancer patients. 50,51Our study suggests that an increase in the protein stability, catalytic activity, or nuclear localization of Ago2 may underlie the increased expression of miR-183/96/182 in cancer. 25

Limitations of the study
This study has some potential limitations.The DN-Drosha cells were generated using HEK293T cells, which raises the possibility that the pri-miRNA processing activity of DN-Drosha, as well as the role of Ago2 in the processing of pri-miR-183, may be observations specific to this cell type.Further research is needed to elucidate the mechanism by which Drosha-NTR interacts with pri-miRNAs, particularly concerning the structure of Drosha-NTR.Lastly, further investigation is warranted to understand the mechanism by which Ago2 facilitates the processing of pri-miR-183, as well as the potential contribution of Ago2 in the deregulation of the miR-183 cluster of miRNAs in various human disorders.

Lead contact
Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Akiko Hata (akiko.hata@ucsf.edu).

Ago2 expression plasmids
Wild type Ago2 expression construct was reported previously. 52The coding sequence of Ago2-D597A mutant was PCR amplified from the lentiviral vector carrying Ago2-D597A 53 and subcloned into the pcDNA3.1(+)vector at HindIII and EcoRI restriction sites.
Quantitative reverse transcriptase-polymerase chain reaction (qRT-PCR) analysis Total RNAs were extracted by RNeasy Mini Kit (#74104, Qiagen) and subjected to generate cDNAs by RT reaction (#17088890, Bio-Rad).qPCR was performed using iQ SYBR Green Supermix (#1708882, Bio-Rad).All reactions were run in triplicates.The relative expression values were determined by normalization to GAPDH transcript levels and calculated using the DDCT method.Primers used for qRT-PCR are listed in supplemental information.

Quantitative miRNA analysis
Mature miRNAs were determined using TaqMan microRNA Assays (Applied Biosystems Inc.).Normalization was performed with the small nuclear RNA U6 (RNU6B; Applied Biosystems Inc.).All real-time reactions, including no-template controls and real-time minus controls, were run using the CFX Connect Real-Time PCR System (Bio-Rad) and performed in triplicate.Relative expression was calculated using the DDCT method.

Immunoprecipitation assay
Cells were lysed in SBB buffer (1% Triton X-100, 150mM NaCl, 50mM Tris-Cl at pH 7.5, 1mM EDTA) supplemented with protease inhibitors (Sigma, P8340) and phosphatase inhibitor (Sigma, P5726).Cell lysates were incubated at 4 C for 30 min and centrifuged at 12,000 g for 10 min at 4 C. Lysates were incubated with anti-Drosha, anti-Ago2, anti-Dgcr8, and anti-IgG (negative control) nutating overnight at 4 C followed by the addition of Dynabeadsä Protein A/G (Invitrogen, 10002D/10004D) and rocking for 4 hr at 4 C.The magnetic beads were precipitated and rinsed with SBB buffer for 5 min at 4 C for three times, followed by adding sample buffer (Invitrogen, NP0007) with reducing agent (Invitrogen, NP0009) and heated at 95 C for 3 min.

Nuclear and cytoplasmic fractionation
Cells were washed with PBS twice, scrape off and pelleted by centrifuging at 4,500 g for 5 min.Cells were then swelled by adding 5 volume of lysis buffer (10 mM HEPES, pH 7.9, with 1.5 mM MgCl 2 , 10 mM KCl, 1mM DTT and protease inhibitor, sigma, P8340) and homogenized.After centrifugation at 10,000 g for 15 min, the supernatant was collected as a cytoplasmic fraction.The crude nuclei pellet was resuspended in 2/3 volume extraction buffer (20 mM HEPES, pH 7.9, with 1.5 mM MgCl 2 , 0.42 M NaCl, 0.2 mM EDTA, 25% (v/v) Glycerol, 1mM DTT and protease inhibitor sigma P8340) and homogenized with a tissue homogenizer.After centrifuging at 20,000 g for 5 min, the supernatant was collected as a nuclear fraction.
Next generation RNA-seq and small RNA-seq and analysis Total RNAs were extracted from cells using TRIzol (Invitrogen).The quality of RNAs was evaluated by 2100 Bioanalyzer Instrument (Agilent Technologies) and the samples with RIN>8.0 were sent to Beijing Genome Institute (BGI) for RNA sequencing and small RNA sequencing.
The sequencing was performed with DNBSEQä technology platforms.Adapter removed clean data were generated by BGI.Quality control, index generation and mapping for RNA sequencing were done with Salmon software tool.Differential gene expression was analyzed with R package DESeq2_1.4.5.Quality control for small RNA sequencing data were done by fast quality filter and fastx trimmer.FASTQ sequences were aligned to the human reference genome (GRCh38) by miRDeep2.The reads were aligned against known miRNAs from miRBase (version 19.0).To process paired-end sequencing, reads were aligned separately covering the mature miRNA both on the forward and reverse read and the obtained number of matches were averaged.The match counts were normalized by a linear scaling method: Trimmed Mean of M-values (TMM) and tested for differential expression using the edgeR package with default settings.To calculate the fold change (FC) of miRNA levels in +/+ cells and Dex5/Dex5 cells, normalized miRNA counts in +/+ cells (n=4) or ex5/Dex5 cells (n=5) were averaged, and the FC of miRNAs was calculated.To analyze the change in proportion of trimmed or added reads at each ends of miRNAs listed in Table 1, we calculated the number of miRNAs reads with the size shorter or longer than the reference sequence as reads per million (RPM), followed by calculating the ratio of RPM in Dex5/Dex5 cells: +/+ cells.

Proliferation assay
Cell growth was monitored by cell counting and 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) assay using MTT cell growth assay kit (#CT02, Millipore).FL or DN-Drosha cells (1x10 5 ) were seeded in 12-well plates and cultured in DMEM containing 10% or 1% FCS. 8, 16, 24, 32, or 40 hr after the media change, cells were harvested and counted by a hemocytometer.For MTT assay, MTT dye was added to each well, incubated at 37 C for 1 hr, followed by the addition of 0.1 mL isopropanol with 0.04 N HCl.The absorbance was measured at a wavelength of 570 nm.

Puromycin incorporation assay
Cells were cultured in the growth media (DMEM with 10% FCS) or starvation media (DMEM with 1% FCS) for 6 or 16 hr, followed by the treatment with 1mM puromycin at 37 C for 10 min.Total cell lysates were generated and subjected to SDS-PAGE and immunoblot with an antipuromycin antibody (Kerafast, EQ0001).

In vitro pri-miRNA processing (IVP) assay
A partial pri-let7b (431-nt) or pri-miR-183 (432-nt) sequence was amplified from human genomic DNAs and used as in vitro transcription template in the reaction using Riboprobe System-T7 Kit (P1440, Promega) with a fluorescence dye (ATTO 680)-conjugated UTP (Aminoallyl-UTP-ATTO-680, NU-821-680, Jena Bioscience).FL and DN cells ($3x10 6 cells) were harvested in 300ml sonication buffer (20 mM Tris-HCL pH=8.0, 100 mM KCL, 0.2 mM EDTA RNase-free) and sonicated with 20% intensity for 5 sec for 3 times.After the sonication, cell lysates were subjected to the centrifugation (12,000 rpm at 4 C for 15 mins).The fluorescein-labelled pri-let-7b or pri-miR-183 (0.1 mg) was mixed with the supernatant (total protein amount of 20-30 mg) supplemented with 6.4 mM MgCl 2 and 0.5 U/ml Recombinant RNase Inhibitor (Promega) in total volume of 30 ml.After the incubation at 37 C for 45 min, the reaction mixtures were separated on a 15% Urea-PAGE gel at 90 V for 150 min to separate the processing product [pre-let7b (82-nt) or pre-miR-183 (110-nt)] from the substrate (pri-let7b or pri-miR-183).The gel image was captured by Odyssey Dlx Imaging System (LI-COR) and the amounts of pri-and pre-let7b and miR-183 were quantitated.The relative processing activity was calculated by the amount of pre-miRNA divided by the sum of pri-miRNA and pre-miRNA after being normalized by the Drosha protein amount.The low range single stranded RNA ladder (#N0364N, NEB) was stained with SYBR Gold (#S11494, Thermo Fisher Scientific) according to the manufacturer's protocol and used as a molecular marker.

RNA immunoprecipitation (RIP) assay
HEK293T cells with FL and DN-Drosha Drosha were subjected to crosslinking with 1% formaldehyde for 15 min at room temperature.Nuclei were isolated and disrupted by sonication using Bioruptor (Diagenode).The sonicated lysates were cleared and subjected to immunoprecipitation with antibodies against RNA binding protein (RBP), such as Drosha, Ago2, or Dgcr8.After immunoprecipitation of RBP associated with RNAs, washing and elution, the pellets were subjected to 10 U DNase I treatment for 30 min at 37 C to remove any remaining DNA.Next, total RNAs were extracted using Trizol/phenol:chloroform (5:1), precipitated with ethanol, and dissolved in 20ul of nuclease free water.5ul of RNA was used for 20ul cDNA synthesis reaction.qRT-PCR reactions were performed using pre-miR-primers by real-time PCR machine (CFX96, BioRad). 100 U/ml RNase inhibitor (SUPERase$inä) was used throughout the experiment.Immunoblot analysis was performed to quantitate the amount of RBP in the precipitates.To calculate the fold enrichment of each RIP reaction from qPCR data, first normalize the Ct value of the target RNA to the Ct of GAPDH (control) mRNA.The data are displayed as 'fold enrichment' of RNAs in RBP IP relative to IgG IP after the normalization of the amount of RBP in the IP sample.

Figure 1 .Figure 2 .Figure 3 .
Figure 1.Generation of cell lines expressing Drosha truncated in the NTR (A) Schematic diagram of the domain structure of Drosha FL and DN-Drosha protein and the human Drosha gene structure (light blue).P: Pro-rich region, R/S: Arg/Ser-rich region, PAZ: Piwi Argonaut and Zwille domain, RIIID: RNase III domain, dsRBD: double-stranded RNA binding domain.(B) RNA-seq data of clones 4, 7, and 8 corresponding to the ex1-8 of the Drosha gene are shown.Each clone was sequenced twice.(C) Total cell lysates of clones 2, 4, 7, 8 or the original HEK293T cells (WT) were subjected to immunoblot analysis of Drosha and GAPDH (loading control).(D) Nuclear (Nuc) and cytoplasmic (Cyto) fraction of +/+ cells (FL) and Dex5/Dex5 cells (DN) cells were subjected to immunoblot analysis of Drosha, Lamin A/C (control for the Nuc fraction), and b-actin (control for the Cyto fraction) (top).Relative distribution (%) of FL and DN-Drosha in the nucleus vs. cytoplasm is shown (bottom).(E) Co-immunoprecipitation of Drosha (FL or DN-Drosha) and Dgcr8 in nuclear extracts of +/+ cells (+) and Dex5/Dex5 cells (Dex5).As control, non-specific IgG (control) was applied.The amount of Drosha after IP is shown.Input samples were subjected to immunoblot analyses of Drosha, Dgcr8, Ddx5, and Lamin A/C (loading control).(F) The level of the Drosha and Dgcr8 mRNA relative to GAPDH mRNA in +/+ cells (black), Dex5/+ cells (blue), and Dex5/Dex5 cells (red) were measured by qRT-PCR and plotted as mean G SEM. n = 3 independent experiments.See also and Figures S1-S3.

Figure 3 .
Figure3.Continued (E) FL or DN-Drosha cells were treated with or without serum starvation (1% serum) for 16 h, followed by puromycin treatment for 10 min and immunoblot analysis with anti-puromycin antibody and anti-b-actin antibody (loading control) (left).Each condition is shown in duplicate (left).The relative abundance of puromycinincorporated proteins normalized by b-actin is shown as mean G SEM (right).n = 3 per group.Unpaired two-tail t-test was used for the statistical analysis.(F) FL or DN-Drosha cells were cultured in growth media (10% serum) or starvation media (1% serum) and the cell number was counted at 8, 16, 24, 32, and 40 h after the media change.The result is plotted as mean G SEM. n = 5 independent samples.See also FigureS4.

Figure 5 .
Figure 5. miR-183 hairpin is required for assisting the processing of miR-96 hairpin (A) The amount of miR-183, -96, -182, control miRNAs (miR-21, and -103a) (normalized to U6 snRNA) and pri-miR-183/96/182 (normalized to GAPDH) in wild type HEK293T (WT) cells and two independent miR-183 knock out clones (183KO1 and 183KO2) was quantitated by qRT-PCR in triplicates.(B) Association of Drosha or Ago2 with the miR-96 or the miR-183 hairpin was assessed by RIP assay in WT and 183KO cells.The amount of the miR-96 hairpin or the miR-183 hairpin in the IP of anti-Drosha, anti-Ago2 antibody or nonspecific IgG was quantitated by qRT-PCR in triplicates.The relative enrichment of the miR-96 hairpin or the miR-183 hairpin (Drosha IP/IgG IP or Ago2 IP/IgG IP) is plotted as mean G SEM. n = 3. (C) Association of Drosha or Ago2 with the miR-96 hairpin or the miR-183 hairpin was assessed by RIP assay in WT and 183KO cells transfected with siCtrl or siDrosha in triplicates.The relative enrichment of the miR-96 hairpin or the miR-183 hairpin (Ago2 IP/IgG IP or Drosha IP/IgG IP) is plotted as mean G SEM. n = 3. (D) The amount of miR-183, -96, and -182 normalized to U6 snRNA in wild type HEK293T (WT) and 183KO cells transfected with siCtrl or siDrosha was quantitated by qRT-PCR in triplicates.The graph was plotted as mean G SEM. (E) Schematic description of the miR-183 hairpin-assisted recruitment of the Microprocessor to the miR-96 hairpin in WT cells.In the absence of the miR-183 hairpin (183KO cells), the Microprocessor (blue and pink circles) is not recruited to the miR-96 hairpin and therefore, the miR-96 hairpin cannot be processed.The miR-182 hairpin processing is independent of the miR-183 hairpin and is performed equally in WT and 183KO cells.Blue circle: Dgcr8.Pink circle: Drosha or Ago2.See also Figure S12.