Non-canonical RNA substrates of Drosha lack many of the conserved features found in primary microRNA stem-loops

Gu, Karen; Mok, Lawrence; Wakefield, Matthew J.; Chong, Mark M. W.

doi:10.1038/s41598-024-57330-5

Download PDF

Article
Open access
Published: 20 March 2024

Non-canonical RNA substrates of Drosha lack many of the conserved features found in primary microRNA stem-loops

Karen Gu^1,2,
Lawrence Mok¹,
Matthew J. Wakefield^3,4 &
…
Mark M. W. Chong^1,2

Scientific Reports volume 14, Article number: 6713 (2024) Cite this article

233 Accesses
Metrics details

Subjects

Abstract

The RNase III enzyme Drosha has a central role in microRNA (miRNA) biogenesis, where it is required to release the stem-loop intermediate from primary (pri)-miRNA transcripts. However, it can also cleave stem-loops embedded within messenger (m)RNAs. This destabilizes the mRNA causing target gene repression and appears to occur primarily in stem cells. While pri-miRNA stem-loops have been extensively studied, such non-canonical substrates of Drosha have yet to be characterized in detail. In this study, we employed high-throughput sequencing to capture all polyA-tailed RNAs that are cleaved by Drosha in mouse embryonic stem cells (ESCs) and compared the features of non-canonical versus miRNA stem-loop substrates. mRNA substrates are less efficiently processed than miRNA stem-loops. Sequence and structural analyses revealed that these mRNA substrates are also less stable and more likely to fold into alternative structures than miRNA stem-loops. Moreover, they lack the sequence and structural motifs found in miRNA stem-loops that are required for precise cleavage. Notably, we discovered a non-canonical Drosha substrate that is cleaved in an inverse manner, which is a process that is normally inhibited by features in miRNA stem-loops. Our study thus provides valuable insights into the recognition of non-canonical targets by Drosha.

Secondary structure RNA elements control the cleavage activity of DICER

Article Open access 19 April 2022

Sequence determinant of small RNA production by DICER

Article 22 February 2023

AGO-bound mature miRNAs are oligouridylated by TUTs and subsequently degraded by DIS3L2

Article Open access 02 June 2020

Introduction

MicroRNAs (miRNAs) are small RNAs, ~ 22 nucleotides (nt) in length, that play a crucial role in regulating gene expression^1,2. Mature miRNAs are generated from stem-loop structures embedded within primary (pri)-miRNA transcripts^3,4. The biogenesis of most miRNAs requires two RNase III enzymes, Drosha and Dicer, that process precursors in a stepwise manner^3,4. Drosha interacts with a dimer of the RNA-binding protein Dgcr8 to form the microprocessor complex^5,6,7. In the nucleus, the microprocessor binds to and cleaves the stem-loop structure in the pri-miRNA, releasing a precursor (pre)-miRNA stem-loop intermediate from the flanking single-stranded RNA (ssRNA) segments^3,4. The pre-miRNA is exported to the cytoplasm where it is further processed by Dicer. Dicer cleaves two helical turns (~ 21–22 bp) away from the Drosha cleavage sites, generating a miRNA duplex^3,4. This miRNA duplex is then bound by an Argonaute protein and one (passenger) strand quickly dissociates while the other (guide) strand remains with the Argonaute to form the core of the miRNA-induced silence complex (miRISC)⁸. The mature miRNA guides the complex to target mRNAs that have complementarity to its seed region (2–7 nt from the 5’ end of the miRNA) to induce gene repression^2,3.

Accurate processing of miRNA precursors is crucial because even a single nucleotide shift in the seed sequence can result in miRISC binding to a different set of mRNA targets. Because cleavage of the pre-miRNA by Dicer is at least partially dependent on the position of Drosha cleavage⁹, precise cleavage by Drosha is particularly critical. Pri-miRNAs have been shown to contain multiple sequence and structure motifs that ensure accurate Drosha cleavage.

Structurally, an optimal miRNA stem-loop has an extensively base-paired stem of ~ 35 ± 1 bp¹⁰ and a terminal loop that is larger than 10 nt¹¹. Two strands of the stem are connected by a ssRNA apical loop and flanked by unstructured ssRNA segments at the base. The microprocessor binds to the dsRNA stem, with Dgcr8 positioned towards the apical loop and Drosha towards the base with the catalytic site located one helical turn (~ 11 bp) away from the basal junction^12,13. Multiple motifs have been identified in miRNA stem-loops that are crucial for ensuring accurate Drosha cleavage. These motifs include the basal UG motif, the apical UGU/GUG (UGU) motif, the CNNC motif, the mGHG motif, and the midBMW motif^10,14,15,16. These appear to be redundant as the presence of a single motif is sufficient to enable efficient and accurate cleavage by Drosha¹⁰. This explains why most natural pri-miRNAs have only a subset of motifs. In addition to the motifs present on the pri-miRNA stem-loop and its adjacent regions, recent studies have shown that several suboptimal stem-loops require an optimal stem-loop on the same transcript for efficient processing^17,18.

Beyond pri-miRNAs, Drosha can also cleave mRNAs that harbor stem-loops resembling those in the pri-miRNAs¹⁹. However, the stem-loop products from this process rarely enter the miRNA biogenesis pathway. Instead, this cleavage primarily destabilizes the mRNA, leading to the repression of target gene expression. Direct Drosha cleavage-mediated gene repression has been demonstrated for the Dgcr8^20,21,22, Myl9, Todr1²³, Ngn2²⁴, and Nfib²⁵ mRNAs. Drosha cleavage of the Dgcr8 mRNA has been observed in many cell types. Because Dgcr8 is part of the microprocessor, this cleavage is thought to function as an autoregulatory mechanism. Drosha cleavage of the other mRNAs has largely been observed in pluripotent stem cells and cleavage plays a crucial role in maintaining the pluripotency of these cells¹⁹.

High-throughput sequencing suggests that there may be dozens of mRNAs that are cleaved by Drosha^26,27. While the features of pri-miRNA stem-loops have been extensively studied, the Drosha-targeted mRNAs have not been as well characterized. In this study, we employed high-throughput sequencing to capture Drosha-cleaved polyA RNA in mouse embryonic stem cells (ESCs) and characterized the features of these non-canonical stem-loop substrates of this enzyme. We show that while Drosha cleaved pri-miRNA stem-loops and mRNA stem-loops may appear similar at a superficial level, there are fundamental differences between these two groups of substrates.

Results

Capturing Drosha-cleaved polyA RNAs by Degradome-seq

Drosha is known to cleave mRNA in mouse ESCs²⁶. To characterize the non-canonical RNAs that are directly cleaved by Drosha in mouse ESCs, we employed Degradome sequencing (Degradome-seq) to capture polyA-tailed RNAs with a 5′ monophosphate (5′P), a hallmark of Drosha-mediated cleavage (Fig. 1a). The Degradome-seq libraries were constructed essentially as described by Karginov et al.²⁶ (Supplementary Fig. S1). PolyA RNA was first extracted, and an RNA linker was ligated to RNAs with a 5’P. The RNAs were then reverse transcribed with a random hexamer attached to a second (DNA) reverse linker sequence. PCR with primers against the 5′ and reverse linkers amplified fragments that have been endonucleolytically cleaved. To determine which cleavage sites are dependent on Drosha, we compared the sites between Drosha-deficient and control ESCs. These cells harbor a LoxP-flanked Drosha^fl/fl allele and CreERT2 knocked into the Rosa26 locus²⁸. Addition of 4-hydroxytamoxifen ablated Drosha expression and consequently the expression of canonical miRNAs, like miR-16 and miR-191, while the expression of the non-canonical miRNAs that are independent of Drosha, like miR-320 and miR-484²⁹, were unaffected (Supplementary Fig. S2a and b). Two independent libraries were generated from both Drosha-deficient and control cells, each resulting in ~ 20 million reads mapping uniquely to the mouse reference genome GRCm38 (mm10; Supplementary Table S1).

We first assessed the quality of our Degradome-seq libraries. Even though the libraries were polyA enriched, the reads that mapped to the protein-coding genes do not skew towards the polyA tail, indicating that the libraries were not position-biased (Supplementary Fig. S2c). The depth of the reads was highly correlated between the two replicates (Supplementary Fig. S2d). When comparing control and Drosha-deficient libraries, the coverage of miRNA genes was decreased (Supplementary Fig. S2c–e), indicating that our libraries captured cleavage targets that are dependent on Drosha.

We next examined the reads that map to annotated canonical miRNAs loci. As expected, these reads exhibited a homogenous 5′ stack precisely at the Drosha cleavage site on the 3′ arm of the stem-loop (Fig. 1b, left panel). This detection of the 3′ cleavage site is expected because this protocol captures polyA-tailed RNA fragments (Fig. 1a). These stacked reads are referred to as “pile-up reads”, with the homogenous 5′ end referred to as the “pile-up site”. The depth of the pile-up site at the example Mir290a was reduced in Drosha-deficient ESCs, confirming that cleavage is Drosha-mediated (Fig. 1b, left panel).

An analogous stacking pattern was observed at other mRNA targets of Drosha. For example, Drosha-dependent pile-up sites were observed at the two annotated stem-loops in Dgcr8, one in the 5’ untranslated region (UTR) and one in the coding domain sequence (CDS; Fig. 1b, right panel). These results demonstrate the effectiveness of our Degradome-seq libraries for capturing Drosha cleavage sites in both pri-miRNAs and mRNA substrates.

Taking advantage of the clear stacking pattern observed in known substrates of Drosha, we developed a bioinformatics pipeline to systematically identify all Drosha-dependent cleavage events (Fig. 1c). We first counted the depth at the start of reads for each locus. Drosha cleavage sites mapping to miRNAs (as listed in MirGeneDB⁵⁷) were used to train the classification algorithm, allowing it to learn the typical stacking pattern associated with Drosha cleavage. The successfully trained algorithm was then applied to the remaining loci to assess whether they exhibit a similar stacking pattern. This method identified 3632 pile-up sites in control cells, with 913 of these significantly decreased in Drosha-deficient cells (including pri-miRNAs; Fig. 1d; Supplementary Table S2). The pipeline successfully captured the cleavage of canonical pri-miRNAs, such as the Mir290 ~ 295 family, while excluding Drosha-independent pri-miRNAs, such as Mir5099 and Mir677³⁰. Furthermore, Drosha-dependency was validated for selected target by qRT-PCR (Supplementary Fig. S3). These findings indicate that our pipeline successfully identifies Drosha-dependent cleavage of polyA RNAs in ESCs.

Most of the Drosha-dependent cleavage sites are miRNA-independent

A Drosha-dependent pile-up site was identified in the Plekhm1 mRNA (Fig. 1d). The cleavage of Plekhm1 has previously been shown to be a miR-106b-5p-guided and is mediated by the slicer activity of Argonaute2²⁶. This also leaves a 5’P on the cleaved RNA (Supplementary Fig. S4). Drosha is required for miR-106b biogenesis, and thus loss of Drosha affects Plekhm1 cleavage indirectly.

To determine if other identified pile-up sites are also miRNA-dependent, we performed Degradome-seq on Dicer-deficient ESCs, in which the biogenesis of miRNAs will also be disrupted. Loss of miRNA expression was confirmed following Dicer1 gene inactivation (Supplementary Fig. S5a and b). We generated two independent replicate Degradome-seq libraries each from Dicer-deficient and control ESCs (Supplementary Table S1). The depth of the reads was highly correlated between the two replicates (Supplementary Fig. S5c). Following processing, we identified 4265 pile-up sites in the control cells, of which 115 were lost in Dicer-deficient cells (Supplementary Fig. S5d; Supplementary Table S3). Of these, only 31 sites were affected by both Drosha and Dicer deficiency, and presumably via loss of miRNAs (Supplementary Fig. S5e). This suggests that most Drosha-dependent RNA cleavage sites in ESCs are independent of Dicer or miRNA.

Inferring the secondary structure of RNAs that are cleaved by Drosha

Drosha usually cleaves stem-loops at two opposing strands to leave an RNase III enzyme signature 3′ 2 nt overhang³¹. To understand the nature of Drosha substrates, we first needed to identify the paired cleavage sites, i.e., the precise position of the cleavage on both the 5’ and 3’ arms of a putative stem-loop structure. Degradome-seq usually reveals the cleavage site on the 3’ arm, but not on the 5’ arm. To identify the 5’ cleavage site, we employed small (s)RNA-seq libraries, which capture miRNAs and other by-products of Drosha cleavage, such as miRNA-offset miRNAs (moRNAs) and other sRNAs. It has previously been shown that moRNAs are derived from the sequence immediately adjacent to the mature miRNA, with one end corresponding to the Drosha cleavage site and the other is thought to be processed by non-specific exonucleases³² (Fig. 2a). We reasoned that Drosha cleavage of non-miRNA stem-loops would also result in similar by-products that are captured by sRNA-seq.

We retrieved 38 sRNA-seq libraries of ESCs from the NCBI GEO database (Supplementary Table S4). Low-quality reads were first removed from each library, then the libraries were collapsed into unique reads and aligned to the mouse reference genome. Collapsing reads increased the representation of moRNAs and other low-frequency reads in the library to allow for easier identification. We confirmed that moRNAs do indeed map to miRNA loci immediately adjacent to the mature miRNA sequences, with the 3’ end of the moRNA-5p and the 5’ end of the moRNA-3p corresponding to the Drosha cleavage sites on the 5’ and 3’ arms of the pri-miRNA stem-loop (Fig. 2b). The distal termini of moRNAs were of varying lengths. This is different from the site-specific endonucleolytic cleavage signature and may indicate exonuclease-mediated degradation. Similar sRNAs were found adjacent to the Drosha-meditate cleavage site in the stem-loop at the Dgcr8 mRNA and other non-miRNA Drosha cleavage targets (Fig. 2b and Supplementary Fig. S6). Here were refer to these as “miRNA-like” and “moRNA-like” sRNAs. Such by-product sRNAs appear to be common to Drosha substrates and not unique to miRNA precursors.

An algorithm was developed to utilize these moRNA-like sRNA-seq reads to determine the position of the cleavage site on the opposite arm of Degradome-seq pile-up sites in Drosha targets (Fig. 2c and Supplementary Fig. S7a). For this, we collected the sRNA-seq reads that started or ended between 25 and 100 nt from the Degradome-seq site, which corresponds to the size of known pre-miRNA stem-loops (Fig. 2c). Drosha can also produce a single cleavage on the 5p arm³³. In this situation, the stem-loop and the sRNAs generated from the stem-loop would be downstream of the Degradome-seq pile-up site (Supplementary Fig. S7a). Therefore, the sRNA reads that are 25–100 nt downstream of the Degradome-seq site were also collected (Supplementary Fig. S7b). Of the Drosha-dependent sites identified by our Degradome-seq of ESCs, 801 exhibited this sRNA read pattern. If the sRNA-seq reads stacked at one of the termini and the depth of that terminus was greater than one read-per-million (RPM), it was regarded as a potential site of Drosha cleavage. The secondary structure of the sequence between the sRNA terminus and the corresponding Degradome-seq pile-up site, flanked by a 15 nt segment, was then predicted (Fig. 2c and Supplementary Fig. S7b). The flanking sequences were included to capture the entire putative stem-loop structure. A sequence that is predicted to fold into a stem-loop, with termini that formed an RNase III signature 3′ 2 nt overhang, was considered a cleavage target of Drosha (Fig. 2c and Supplementary Fig. S6b). This correctly identified the 5’ and 3’ Drosha-mediated cleavage sites in 68 miRNA stem-loops out of the 68 miRNAs that are highly expressed in ESCs, demonstrating the effectiveness of the algorithm. Using the same method, we were then able to reassemble 42 non-miRNA stem-loops that were cleaved by Drosha in ESCs (Supplementary Table S5).

The Drosha cleavage targets identified in this study largely overlapped with those captured in a previous analysis of mouse ESCs²⁶ (Supplementary Fig. S8). This further supports the authenticity of Drosha cleavage targets identified. That being said, our study identified many more Drosha cleavage targets, which can be attributed to considerably deeper sequence depth. However, we found minimal overlap in Drosha cleavage targets reported for HEK293T and Hela cells²⁷. This discrepancy is likely to be due to differences in the transcriptomes between mouse ESCs and immortal human cell lines. Moreover, many of Drosha targets with demonstrated biological function have been shown to be cell-type specific^23,24,25.

Long non-coding (lnc)RNAs are a class of RNA that are longer than 200 nt. They usually possess a 5′ m⁷G cap and a 3′ polyA tail, but they do not encode functional proteins³⁴. Many miRNA stem-loops are mapped to lncRNA genes, and the level of mature miRNAs derived from these transcripts is relatively high (Fig. 2d and e). Thus, these lncRNA are actually pri-miRNAs. One-third of the miRNA stem-loops were located in mRNA introns (Fig. 2d), which is consistent with previous studies³⁵. Only three annotated miRNA stem-loops were located within an exon of a mRNA and these miRNAs tended to be expressed at low levels (Fig. 2e). Given the low level of mature miRNAs, these exon-located stem-loops are unlikely to be functional miRNA precursors. In contrast, most of the Drosha-targeted non-miRNA stem-loops were located in the exon of mRNAs (Fig. 2d).

Drosha-cleaved non-miRNA stem-loops are less thermodynamically stable than miRNA precursor stem-loops

Drosha cleavage of exonic stem-loops is expected to destabilize the mRNA^22,23,25. Stem-loops are one of the most common RNA structures³⁶, and yet Drosha cleaves only some stem-loops but not others. We thus sought to understand the nature of the stem-loops that are recognized and cleaved by Drosha. For this, we systematically characterized and compared the features of miRNA and mRNA stem-loop targets of Drosha. If more than one alternative Drosha cleavage sites were identified for a stem-loop, the site with the highest number of sRNA-seq reads was selected for analysis.

We found that non-miRNA stem-loops had a significantly higher minimum free energy (MFE) compared to miRNA stem-loops, indicating that they were less thermodynamically stable (Fig. 3a). The stability can be affected by several properties, including RNA length, base pairing availability, and base pairing composition. We found that non-miRNA stem-loops are slightly shorter but had a significantly higher variance in length compared to the miRNA stem-loops (Fig. 3b). In addition, miRNA stem-loops had a higher base pairing frequency and a longer base pairing stacking (Fig. 3c-e). These results suggest that base pairing plays a significant role in the stability of miRNA stem-loops.

The composition of base pairing appeared to have less of an impact. Both miRNA and non-miRNA stem-loops exhibited similarly high G-C base pairing and low A-U and G-U base pairing in the lower-stem (Fig. 3f, left panel). However, only non-miRNA stem-loops exhibited a preference for high G-C base pairing in the upper-stem whereas there was a similar frequency of A-U and G-C base pairing in the upper-stem of miRNA stem-loops (Fig. 3f, right panel). Mature miRNAs originate from the upper-stem. This similar A-U and G-C usage in the upper-stem of miRNA stem-loops likely relates to the requirement for sequence diversity in miRNAs. Therefore, the stability of miRNA stem-loops is primarily achieved through extensive base pairing, while non-miRNA stem-loop relies mainly on the G-C base pairing.

Non-miRNA stem-loops display more alternative structure

An analysis of positional base pairing entropy shows that miRNA stem-loops have low entropy in the stem region, with a rise in entropy in the ssRNA region, including the unstructured flanking region and terminal loop region (Fig. 4a). These suggest that the stem of miRNA stem-loops is unlikely to form alternative base pairing, which likely ensures precise cleavage by Drosha and therefore the sequence of the mature miRNA that is eventually produced. In contrast, the entropy for non-miRNA stem-loops was consistently high, implying the presence of numerous alternative base pairing possibilities (Fig. 4a). This is further supported by the high ensemble diversity of non-miRNA stem-loops, which reflects the diversity of secondary structures that a non-miRNA stem-loop can adopt (Fig. 4b).

Structural differences between miRNA and non-miRNA stem-loops

We next compared the fine structure of the stem-loops. Twenty-five nt of the sequence flanking the Drosha-cleaved stem-loop structure was included for these analyses. In agreement with the established understanding of the ideal length of miRNA stem-loops¹⁰, the miRNA stem-loops identified were found to be extensively base-paired between positions -13 and 22, creating a dsRNA stem of ~ 35 ± 1 bp in length (Fig. 4c). In contrast, non-miRNA stem-loops were extensively base-paired only between positions -13 and 13, resulting in a stem that is on average only ~ 26 bp in length (Fig. 4d). In addition, miRNA stem-loops have a conserved terminal loop region starting from position 25, while such a region was absent from the non-miRNA stem-loops. This is likely due to the variable stem length of non-miRNA substrates. Instead, a “weak” terminal loop can be observed starting from position 17 (Fig. 4d), suggesting that some non-miRNA stem-loop may have large terminal loops.

Non-miRNA stem-loops also displayed more asymmetrical internal loops compared to miRNA stem-loops. The miRNA stem-loops displayed small symmetrical internal loops in the lower-stem and the middle of the upper-stem (Fig. 4c and e). The internal loops at these positions have previously been shown to affect Drosha processing^10,15,33. Large asymmetrical internal loops were mainly observed in the region 21 bp and above. These large asymmetrical internal loops serve as terminal loop in some miRNAs, such as in Mirlet7b (Supplementary Fig. S9). In contrast, numerous mismatches were observed in non-miRNA stem-loops, most of which formed asymmetrical internal-loops (Fig. 4d and e).

Despite the numerous structural differences between miRNA and non-miRNA stem-loop, their basal junctions were strikingly similar. A sharp decrease in pairing was observed at position -13 in both types of stem-loops (Fig. 4c and d), suggesting that a clear ssRNA-dsRNA junction is a crucial feature for Drosha cleavage. Similar to miRNA stem-loops, Drosha appears to bind to this ssRNA-dsRNA junction in non-miRNA substrates to cleave ~ 13 bp away from the junction.

Non-miRNA stem-loop lacks canonical sequence motifs

Several conserved sequence motifs have been identified in miRNA stem-loop precursors and these have been shown to ensure efficient and accurate binding and processing by Drosha. These motifs include the basal UG motif, the apical UGU motif, the CNNC motif and the mGHG motif^10,14. All sequence motifs were found in the Drosha processed miRNA stem-loops identified in ESCs (Fig. 5a). The apical UGU motif was found to be the most common motif in the miRNAs expressed in ESCs. In contrast, none of these miRNA motifs were detected in non-miRNA stem-loops (Fig. 5b).

Sequence motifs are important for the proper orientation of Drosha when it binds to the stem-loops. Without these motifs, Drosha can bind to the apical junction and cleave the stem-loop inversely³⁷. While this phenomenon has been demonstrated using pri-miRNA variants in vitro with processing assays, whether it occurs in vivo has been unclear. Our analysis revealed a stem-loop in Lrrc59 that is inversely cleaved by Drosha in ESCs. This stem-loop is characterized by an unusually large terminal loop of 32 nt. The Drosha cleavage site is ~ 13 bp away from the apical junction and has no clear basal junction (Fig. 5c), suggesting that Drosha binds to the apical junction and cleaves the stem-loop in an inverse manner. Together, these findings indicate that Drosha recognizes and processes miRNA and non-miRNA stem-loops differently.

Drosha cleavage does not necessarily repress gene expression

While Drosha cleavage of transcripts can cause the significant destabilization and downregulation of transcript levels, we found this is not the case for many of the identified Drosha cleavage targets in our study. Differential gene expression analysis revealed that 1088 genes were significantly upregulated, and 375 genes were significantly downregulated (|logFC|> 1.5 and FDR < 0.05) in Drosha-depleted ESCs (Supplementary Fig. S10 and Supplementary Table S6). Except for Dgcr8, the cleavage of non-miRNA stem-loops results in only moderate to little reduction in transcript levels. This observation suggests that the cleavage of non-miRNA stem-loops might be either inefficient due to the absence of features that facilitate Drosha processing, or Drosha interaction serves a function other than destabilization for these transcripts.

Discussion

Drosha cleaves many non-miRNA stem-loops in mouse ESCs. Our analysis of these non-canonical stem-loop substrates revealed fundamental differences between these and miRNA stem-loops. Specifically, we found that non-miRNA stem-loops are less thermodynamically stable and more likely to fold into alternative structures. Moreover, they lack the sequence and structural motifs normally found in miRNA stem-loops that ensure Drosha cleavage at single nucleotide precision. Consequently, Drosha can cleave these non-canonical substrates at positions that are typically inhibited in miRNA stem-loops.

Pri-miRNA stem-loops are typically thermodynamically stable and unlikely to fold into alternative structures. Many non-coding RNAs adopt specific secondary structures to carry out their functions, such as the clover-shaped transfer-RNAs. These RNAs have evolved over time, leading to exceptional thermodynamic stability and low structural diversity, which facilitate their processing and functionality^38,39. Similarly, the stem-loop structure of miRNA plays a crucial role in its biogenesis³. Thus, like other functional non-coding RNAs, miRNA precursors have a stable structure that enables their efficient and accurate processing by Drosha and Dicer to produce functional miRNAs. However, a stable stem-loop in mRNA would hinder other processes, such as the progression of ribosomes along an mRNA thereby impairing protein expression⁴⁰. Therefore, a stable stem-loop would be undesirable in mRNAs, which could explain why non-miRNA stem-loops are generally less stable.

The structure and sequence motifs of miRNA stem-loops are important for accurate binding and processing by Drosha. Drosha needs to bind to the ssRNA-dsRNA junction at the base and cleave ~ 11 bp away to produce miRNA intermediates^12,13. However, due to the symmetrical structure of a stem-loop (i.e. ssRNA-dsRNA-ssRNA), Drosha can also bind to the apical junction, which could result in inverse cleavage of the stem, closer to the loop than to the base. To prevent this, several miRNA stem-loop sequences and structure motifs have been found in miRNA stem-loops to encourage binding of Drosha to basal junction and therefore inhibit inverse cleavage. This includes the UGU motif at the apical junction that interacts with Dgcr8, positioning Drosha towards the lower-stem⁴¹, while the CNNC motif interacts with Srsf3 to recruit Drosha to the lower-stem^14,42. The basal UG and mGHG motifs also serve as Drosha-interacting motifs, enabling precise binding to the stem-loop^10,13. Recently, a MidMW10 motif, located 10–12 nt away from the Drosha cleavage site in the upper-stem has also been shown to be essential for preventing inverse cleavage¹⁵. None of these structure and sequence motifs were found in non-miRNA stem-loops that are cleaved in ESCs. Consequently, inverse cleavage of non-miRNA substrates by Drosha, such as the stem-loop in Lrrc59 that we described, is possible.

A difference between miRNA and non-miRNA stem-loops was anticipated because Drosha cleavage of mRNA stem-loops generally does not produce meaningful quantities of mature miRNAs. Consequently, many features required for precise miRNA production are unnecessary for non-miRNA stem-loop cleavage. However, the absence of Drosha recognition and processing features renders Drosha cleavage less efficient in repressing the expression of transcripts harboring non-miRNA stem-loops. This raises the question of why Drosha cleaves them at all. One possibility is that the cleavage serves a purpose other than destabilization, such as promoting alternative intron splicing. This has been demonstrated by Havens et al.⁴³ and Lee et al⁴⁴, where cleavage of a stem-loop promotes alternative splicing but has little impact on the gene expression level.

Even though the cleavage of most non-miRNA stem-loops is not efficient enough to repress gene expression in the ESCs, the possibility remains that trans- or cis-acting elements may facilitate Drosha-mediated cleavage of non-canonical stem-loop substrates to achieve spatiotemporally regulation of specific targets. It has been shown that post-transcriptional modification of target RNA can affect Drosha processing. N⁶-methyladenosine (m⁶A) on the RNA upstream of the stem-loop has been shown to recruit Drosha to the vicinity of pri-miRNA stem-loops and enhance Drosha processing efficiency⁴⁵. Such modification may also be present upstream of non-miRNA stem-loops to ensure Drosha recruitment and to enhance cleavage efficiency. Additionally, the stability of stem-loops may be affected by ADAR enzyme-dependent A-I editing, thereby altering Drosha's processing efficiency on the stem-loop⁴⁶.

Accessory proteins of the microprocessor may also affect non-canonical stem-loop processing. Although the minimal microprocessor complex of Drosha and Dgcr8 alone is sufficient to process pri-miRNA stem-loops, numerous accessory proteins that interact with the complex have been identified. These include the DEAD-box helicases (Ddx5) and tumor suppressor p53 that are required for Drosha-mediated processing of a subset of miRNAs⁴⁷. Additionally, hnRNP TAR DNA-binding protein 43 (Tdp43) can interact with Drosha to increase its stability and promote processing⁴⁸. Such accessory proteins may facilitate the processing of non-canonical miRNA stem-loops.

Furthermore, post-translational modifications can also regulate the microprocessor. Acetylation of lysine residues within the N-terminal of Drosha has been found to repress ubiquitin-mediated proteasome decay⁴⁹. Deacetylation of Dgcr8 by histone deacetylase 1 (Hdac1) has been found to increase the affinity of Dgcr8 for a subset of pri-miRNA⁵⁰. At least 23 phosphorylated amino acids have been found on Dgcr8. Phosphorylation appears to increase the stability of Dgcr8 without affecting its ability to interact with Drosha⁵¹. Similar post-translational modifications can regulate both protein–protein and protein-RNA interactions, thereby affecting Drosha’s processing efficiency on non-canonical stem-loop targets.

Our analysis is likely underestimating the number of possible Drosha targets. We focused on stem-loops where the two cleavage sites are at most 100 nt apart on the same RNA. Structural studies of Drosha^12,13 and Drosha targets (Fig. 4c and d) suggest that the ssRNA-dsRNA junction is a crucial feature for successful Drosha cleavage. Such a ssRNA-dsRNA junction can also be present in longer stem-loops or between a pair of sense:antisense transcripts. It is possible that these might also be recognized and cleaved by Drosha but further studies are required to explore these possibilities.

Our study has provided a comprehensive analysis of stem-loops cleaved by Drosha in ESCs, revealing that the non-canonical stem-loop substrates of Drosha differ significantly from pri-miRNA stem-loops. Drosha cleavage-mediated gene repression has been shown to be critical for safeguarding the pluripotency of the stem cells^23,24,25. Determining how this Drosha cleavage of non-canonical targets is achieved will therefore be critical for understanding the mechanisms regulating stem cell pluripotency, and the knowledge from this study provides a foundation for such future studies into the regulation of Drosha function.

Materials and methods

Generation of conditional Dicer ESCs

Conditional LoxP-flanked Dicer^fl/fl ESCs were derived from blastocysts obtained from Dicer1^fl/fl mice⁵². A time-mated female mouse was euthanized by CO₂ asphyxiation 3 days after the vaginal plug was confirmed. The uterine horns were harvested, and blastocysts flushed out with PBS. Individual blastocysts were then deposited into wells containing Mitomycin C-inactivated mouse embryonic fibroblasts as feeder cells and cultured in KnockOut DMEM (Gibco) supplemented with 20% Knockout Serum Replacement (Gibco), non-essential amino acids (Gibco), 0.1 mM β2-mercaptoethanol and 10³ U/mL LIF (Peprotech). The wells were monitored daily for outgrowth from the blastocysts. Upon outgrowth, the well was trypsinized into a single cell suspension and transferred into a larger well for picking of individual colonies. This animal work was approved by the St Vincent’s Hospital Animal Ethics Committee. The experiments were performed in accordance with relevant guidelines and regulations under the Australian code for the care and use of animals for scientific purposes.

Mouse ESCs culturing

Conditional LoxP-flanked Drosha^fl/fl Gt(ROSA)26Sor^CreERT2 ESCs have been described previously ²⁸. All ESCs were cultured in KnockOut DMEM, supplemented with 10% heat-inactivated fetal bovine serum (GE Healthcare Life Sciences), 5% KnockOut Serum Replacement (Gibco), 1% sodium pyruvate (Gibco), 1% non-essential amino acids, 1% penicillin–streptomycin-glutamine (Gibco), 0.1 mM 2-mercaptoethanol and 10³ U/mL LIF on a Mitomycin C-inactivated mouse embryonic fibroblasts.

Deletion of the floxed Drosha allele was achieved by adding 100 nM 4-hydroxytamoxifen (Sigma-Aldrich) to the Drosha^fl/fl Gt(ROSA)26Sor^CreERT2 ESCs for 72 h. The medium was then replaced (without 4-hydroxytamoxifen) for a further 48 h to allow for depletion of Drosha-dependent miRNAs before analysis.

Deletion of the floxed Dicer allele was achieved by transducing the Dicer^fl/fl ESCs with a Cre-expressing retrovirus⁵³. The virus also contained a GFP reporter that allowed for the sorting of transduced cells. GFP⁺ ESCs were sorted 3 days after transduction, then cultured for a further 2 days before analysis.

Quantitative (q)RT-PCR

Total RNA was extracted from the cells using TRIsure (Bioline) following the manufacturer’s instructions. For measuring mRNA expression, 1 μg total RNA was reverse transcribed with 50 ng random hexamers using M-MuLV reverse transcriptase (NEB). 1/20th of the resulting cDNA was used and then analyzed by qRT-PCR using GoTaq qPCR master mix (Promega). The following primer pairs were used:

5′-GACGACGACAGCACCTGTT-3′ and 5′-GATAAATGCTGTGGCGG-ATT-3′ for Drosha;

5′-TCTGCAGGCTTTTACACACG-3′ and 5′-CAGCCAATGATGCAAA-GATG-3′ for Dicer;

5′-CACAGCTTCTTTGCAGCTCCTT-3′ and 5′-CGTCATCCATGGCGAAC-TG-3′ for β-actin.

Taqman miRNA assays (Thermo Fisher) were used to quantify the expression of mature miRNAs and U6 snRNA (control). For each target, 10 ng total RNA was reverse transcribed with its specific reverse transcription primer using 25 U Multiscribe reverse transcriptase. qRT-PCR then performed on 1/15th of the cDNA with Taqman universal PCR master mix and 1X Taqman miRNA assay mix (miR-16 assay ID: 000391; miR-191 assay ID: 002299; miR-320 assay ID: 002277; miR-484 assay ID: 001821; U6 snRNA assay ID: 001973).

Degradome sequencing library construction

Degradome-seq libraries were constructed based on a previously described protocol²⁶. In brief, polyA-tailed RNA was isolated from 75 μg of total RNA using the Dynabeads mRNA direct purification kit (Thermo Fisher) following the manufacturer’s instructions. An RNA linker (5′-CACGACGCUCUUCCGAUCU-3′) was ligated to the polyA RNA with T4 RNA ligase (Thermo Scientific) to capture RNAs with a 5’P, a hallmark of RNase III cleavage. The ligation products were then purified with the Dynabeads mRNA direct purification kit and reverse transcription was performed using Superscript III reverse transcriptase (Invitrogen) and random hexamer primer attached to reverse adaptor sequence (5′-AGACGTGTGCTCTTCCGATCNNNNNN-3′). The cDNA was cleared of RNA by treating it with RNase H (NEB). Half the cDNA was subjected to 2nd strand synthesis using Phusion high-fidelity DNA polymerase (NEB) with primers to the 5′ RNA linker and reverse adaptor (5′-ACACTCTTTCCCTACACGACGCTCTTCCGATCT-3′ and 5′-GTGACTGGAGTTCAGACGTGTGCTCTTCCGATC-3′) using the cycling conditions: 98 °C for 1 min, (98 °C for 30 s, 58 °C for 30 s, 72 °C for 1 min) × 7 cycles, and 72 °C for 5 min. The resulting cDNA was resolved on a Low Melting Point agarose gel (Scientifix) and cDNA corresponding to 200–400 bp were gel purified. This cDNA library was then PCR barcoded with PCR Primer 1.0 (5′-AATGATACGGCGACCACCGAGATCTACACTCTTTCCCTACACGACG-3′) and a sample-specific barcode primer (5′-CAAGCAGAAGACGGCATACGAGATNNNNNNGTGACTGGAGTTCAGACGTG-3’) using MyTaq Red Mix (Bioline) for 15 cycles. The resulting cDNA libraries were sequenced on the NextSeq 500 platform (Illumina) for 75 cycles in single-end high-output mode at the Australian Genome Research Facility.

Gene-specific degradome qRT-PCR

PolyA-tailed RNA was isolated from 75 μg of total RNA using the Dynabeads mRNA direct purification kit following the manufacturer's instructions. An RNA linker (5′-CACGACGCUCUUCCGAUCU-3′) was ligated to the polyA RNA with T4 RNA ligase to capture RNAs with a 5′P. The ligation products were then purified with the Dynabeads mRNA direct purification kit and reverse transcription was performed using Superscript III reverse transcriptase and random hexamer primers.

One-twentieth of the resulting cDNA was pre-amplified for 20 cycles using 5 U Taq DNA polymerase (NEB), with 1 μM forward primer to the 5′ RNA linker (5′-ACTCTTTCCCTACACGACGC-3′) and 1 μM gene-specific outer reverse primers: 5′-TTCATGGGGCAGCACTTGGA-3′ for Dgcr8; 5′-TGGCCGAATCTGCTACTTCAC-3′ for Rcan3; and 5′-TCTGTCCGTCACCTTGCCTT-3′ for Chpf2. One-hundredth of resulting pre-amplified cDNA was then analyzed by qPT-PCR using GoTaq qPCR master mix with 0.5 μM forward primer to the 5′ RNA linker (5′-ACACTCTTTCCCTACACGACGCTCTTCCGATCT-3′) and 0.5 μM gene-specific inner reverse primers: 5′-AGGGACTCTCATATGTCTCCA-3′ for Dgcr8; 5′-CTCAGAGTGCACAGTCCAGC-3′ for Rcan3; and 5′-TGTTGGCCTGTTCCTGTTCA-3′ for Chpf2.

Sequence processing and alignment of Degradome-seq libraries

Processing and alignment of Degradome-seq libraries were performed on the Galaxy Australia platform⁵⁴. The raw reads were processed using the Trimmomatic program⁵⁵ (version 0.38.0) to remove Illumina platform-specific adaptor sequences and low-quality bases (below 20 across 4 nt). The processed reads were then aligned to the mouse reference genome GRCm38 (mm10) using RNA STAR with default parameters⁵⁶ (version 2.7.8a).

Degradome-seq analysis pipeline

The depth of Degradome-seq reads at each locus was determined by counting the number of reads starting at that locus. The raw counts were normalized as the reads per million (RPM). The depth of read at the 5’ end was denoted by $C5$. For each genomic locus i, we generated a counting vector ${C}_{i}$ by

$${C}_{i}=C{5}_{i-10}, C{5}_{i-9}, C{5}_{i-8}, \dots , C{5}_{i}, C{5}_{i+1}, \dots , C{5}_{i+10}.$$

We retrieved a list of miRNAs from MirGeneDB⁵⁷. The genomic loci immediately upstream of the 5' end of the 5' arm miRNA or downstream of the 3' end of the 3' arm miRNA were considered as known Drosha cleavage sites. The counting vector of these known Drosha cleavage sites was utilized as the positive training dataset. A set of 1000 loci was randomly selected from the rest of the genome as a negative training dataset. These positive and negative training datasets were used to train a generalized linear model using the cv.glmnet function in the R package glmnet⁵⁸, with the following parameters: family = “binomial”, type.measure = “class”, nfolds = 10. The training error was accessed by sensitivity, specificity, and the area under the curve (AUC). The AUC was calculated using the R package pROC⁵⁹ (version 1.18.0). The sensitivity and specificity were calculated using the formulas below:

$$Sensitivity= \frac{TP}{TP+FN} ,$$

$$Specificity= \frac{TN}{TN+FP} ,$$

where $TP$ is true positive, $FN$ is false negative, $TN$ is true negative, and $FP$ is false positive. We considered the model trained successfully if it had sensitivity > 0.9, specificity > 0.9, and AUC > 0.9. We then applied the model to the rest of the genomic loci to identify those with a similar stacking pattern (i.e., pile-up site) to the known pri-miRNAs. The step between randomly selecting 1000 negative samples and applying the model to the rest of the genomic loci was repeated until the datasets were classified 1000 times. A locus was considered a true pile-up site if it was classified as a pile-up site in more than 900 iterations in both control ESC replicates.

To determine whether a pile-up site was dependent on Drosha or Dicer, we performed differential pile-up analysis between Drosha deficient and control ESCs or Dicer deficient and control ESCs using the edgeR package⁶⁰ (version 3.26.1). Raw read counts starting at each genomic locus were normalized by the trimmed mean of M values (TMM), and the dispersion was estimated by calling the function estimateDisp(). The testing of differential pile-ups was performed by calling functions glmQLFit() and glmQLFTest(). A genomic locus with a log₂ fold change (log₂FC) < (− 1.5) and false discovery rate (FDR) < 0.05 were considered significant.

Sequence processing and alignment of small RNA sequencing libraries

Small RNA sequencing (sRNA-seq) libraries used in this study are retrieved from the NCBI Gene Expression Omnibus (GEO) database (Supplementary Table S1) and the processing and alignment of reads were performed on the Galaxy Australia platform⁵⁴. The quality of the reads and the representation of the sequencing adapters were accessed by the FastQC program (version 0.72), and the reported adapter sequences were trimmed using cutadapt⁶¹ (version 1.16). The processed reads were collapsed into unique reads using an in-house. The unique reads were then aligned to the mouse reference genome GRCm38 (mm10) using RNA STAR with default parameters⁵⁶ (version 2.7.8a).

Identification of Drosha-cleaved stem-loops

Positions identified as Drosha-dependent cleavage sites in the Degradome-seq analysis were denoted as ${D}_{i}$. Reads starting or ending between ${D}_{i-100}$ and ${D}_{i-25}$ or between ${D}_{i+25}$ and ${D}_{i+100}$ were collected. If the sRNA-seq reads stack at one of the termini and the depth of that terminus is > 1 RPM (denoted as ${S}_{j}$), the RNA sequence between ${S}_{j-15}$ and ${D}_{i+15}$ or between ${D}_{i-15}$ and ${S}_{j+15}$ was used to predict the secondary structure using RNAfold program in the ViennaRNA package⁶² (version 2.4.17) with the following option: RNAfold -p -d2 –noLP –MEA. The 3′ overhang was calculated in the structure generated by RNAfold. The structure was considered an authentic Drosha-cleaved stem-loop if ${S}_{j}$ and ${D}_{i}$ form a 3′ 2 ± 2 nt overhang.

Overlapping analysis

To obtain a list of Drosha cleavage targets captured in Degradome-seq libraries from Karginov et al.²⁶, we reanalyzed the Degradome-seq dataset using the method developed in this study. The Drosha cleavage targets identified in HEK293T and HeLa cells²⁷ were converted to mouse genes using “Mouse/Human Orthology with Phenotype Annotations” obtained from Mouse Genome Informatics (MGI). The identification of overlapping Drosha targets was performed using the ggVennDiagram package⁶³.

Differential gene expression analysis

Stranded RNA sequencing (RNA-seq) libraries were prepared using the TruSeq Stranded mRNA Sample Preparation Kit from 3 μg total RNA. The libraries were sequenced on the Illumina NovaSeq 6000, generating 20 million 100 bp single-end reads at Australian Genome Research Facility. The raw reads were aligned to the mouse reference genome GRCm38 (mm10) using RNA STAR with default parameters. The differential gene expression analysis was performed using edgeR package⁶⁰. Genes with |log₂FC|> 1.5 and FDR < 0.05 were considered to be significantly differentially expressed.

Analysis of stem-loop thermodynamic properties

The length of the stem-loop was determined as the distance between ${D}_{i}$ and ${S}_{j}$ with a 15 nt flanking sequence. The base pairing frequency and maximum stacking of base pairing were normalized to the length of the stem-loop. The base pair composition was normalized to the total number of base pairs in the stem.

Analysis of stem-loop structural diversity

The ensemble diversity of the stem-loop was obtained from the RNAfold output using the option described above. The positional entropy of the stem-loops was calculated using script mountain.pl in the ViennaRNA package⁶² (version 2.4.17).

Information bits plot of structural features

The secondary structure between ${D}_{i}$ and ${S}_{j}$ with a 25 nt flanking sequence was predicted using the RNAfold program with the following option: RNAfold -p -d2 –noLP -C. The position of paired bases in secondary structure between ${D}_{i}$ and ${S}_{j}$ with a 15 nt flanking sequence was used as a constraint to inform the prediction. This ensured that paired bases remained paired when incorporated with longer flanking segments.

An in-house script was used to determine whether a position forms a base pair, an internal loop (symmetrical or asymmetrical), a terminal loop, or flanking segments. The resulting data was used to generate information bits plots with the Python package Logomaker⁶⁴ (version 0.8).

Sequence Logo

The sequence logo was generated using the Python package Logomaker⁶⁴ (version 0.8) and the sequence between ${D}_{i}$ and ${S}_{j}$ with a 25 nt flanking sequence.

Statistical analysis

Two-tailed t-tests were used to perform statistical analysis on the qPT-PCR data. If not indicated, two-tailed unequal variance t-tests (Welch’s t-test) were used to analyze the differences between miRNA and non-miRNA stem-loop features. Results were considered statistically significant at p < 0.05.

Data availability

The data underlying this article are available in NCBI GEO and can be accessed with accession number GSE228299. All methods are reported in accordance with ARRIVE guidelines (https://arriveguidelines.org).

References

Ameres, S. L. & Zamore, P. D. Diversifying microRNA sequence and function. Nat. Rev. Mol. Cell Biol. 14, 475–488. https://doi.org/10.1038/nrm3611 (2013).
Article CAS PubMed Google Scholar
Jonas, S. & Izaurralde, E. Towards a molecular understanding of microRNA-mediated gene silencing. Nat. Rev. Genet. 16, 421–433. https://doi.org/10.1038/nrg3965 (2015).
Article CAS PubMed Google Scholar
Bartel, D. P. Metazoan MicroRNAs. Cell 173, 20–51. https://doi.org/10.1016/j.cell.2018.03.006 (2018).
Article CAS PubMed PubMed Central Google Scholar
Ha, M. & Kim, V. N. Regulation of microRNA biogenesis. Nat. Rev. Mol. Cell Biol. 15, 509–524. https://doi.org/10.1038/nrm3838 (2014).
Article CAS PubMed Google Scholar
Denli, A. M., Tops, B. B., Plasterk, R. H., Ketting, R. F. & Hannon, G. J. Processing of primary microRNAs by the Microprocessor complex. Nature 432, 231–235. https://doi.org/10.1038/nature03049 (2004).
Article ADS CAS PubMed Google Scholar
Gregory, R. I. et al. The Microprocessor complex mediates the genesis of microRNAs. Nature 432, 235–240. https://doi.org/10.1038/nature03120 (2004).
Article ADS CAS PubMed Google Scholar
Han, J. et al. The Drosha-DGCR8 complex in primary microRNA processing. Genes Dev. 18, 3016–3027. https://doi.org/10.1101/gad.1262504 (2004).
Article CAS PubMed PubMed Central Google Scholar
Medley, J. C., Panzade, G. & Zinovyeva, A. Y. microRNA strand selection: Unwinding the rules. Wiley Interdiscip. Rev. RNA 12, e1627. https://doi.org/10.1002/wrna.1627 (2021).
Article CAS PubMed Google Scholar
Park, J. E. et al. Dicer recognizes the 5’ end of RNA for efficient and accurate processing. Nature 475, 201–205. https://doi.org/10.1038/nature10198 (2011).
Article CAS PubMed PubMed Central Google Scholar
Fang, W. & Bartel, D. P. The menu of features that define primary MicroRNAs and enable de novo design of MicroRNA genes. Mol. Cell 60, 131–145. https://doi.org/10.1016/j.molcel.2015.08.015 (2015).
Article CAS PubMed PubMed Central Google Scholar
Zeng, Y. & Cullen, B. R. Efficient processing of primary microRNA hairpins by Drosha requires flanking nonstructured RNA sequences. J. Biol. Chem. 280, 27595–27603. https://doi.org/10.1074/jbc.M504714200 (2005).
Article CAS PubMed Google Scholar
Jin, W., Wang, J., Liu, C. P., Wang, H. W. & Xu, R. M. Structural Basis for pri-miRNA Recognition by Drosha. Mol. Cell 78, 423-433 e425. https://doi.org/10.1016/j.molcel.2020.02.024 (2020).
Article CAS PubMed Google Scholar
Partin, A. C. et al. Cryo-EM structures of human Drosha and DGCR8 in complex with primary MicroRNA. Mol. Cell 78, 411-422 e414. https://doi.org/10.1016/j.molcel.2020.02.016 (2020).
Article CAS PubMed PubMed Central Google Scholar
Auyeung, V. C., Ulitsky, I., McGeary, S. E. & Bartel, D. P. Beyond secondary structure: Primary-sequence determinants license pri-miRNA hairpins for processing. Cell 152, 844–858. https://doi.org/10.1016/j.cell.2013.01.031 (2013).
Article CAS PubMed PubMed Central Google Scholar
Li, S., Nguyen, T. D., Nguyen, T. L. & Nguyen, T. A. Mismatched and wobble base pairs govern primary microRNA processing by human Microprocessor. Nat. Commun. 11, 1926. https://doi.org/10.1038/s41467-020-15674-2 (2020).
Article ADS CAS PubMed PubMed Central Google Scholar
Li, S., Le, T. N., Nguyen, T. D., Trinh, T. A. & Nguyen, T. A. Bulges control pri-miRNA processing in a position and strand-dependent manner. RNA Biol. https://doi.org/10.1080/15476286.2020.1868139 (2020).
Article PubMed PubMed Central Google Scholar
Fang, W. & Bartel, D. P. MicroRNA clustering assists processing of suboptimal MicroRNA hairpins through the action of the ERH protein. Mol. Cell 78, 289-302 e286. https://doi.org/10.1016/j.molcel.2020.01.026 (2020).
Article CAS PubMed PubMed Central Google Scholar
Hutter, K. et al. SAFB2 enables the processing of suboptimal stem-loop structures in clustered primary miRNA transcripts. Mol. Cell 78, 876-889 e876. https://doi.org/10.1016/j.molcel.2020.05.011 (2020).
Article CAS PubMed Google Scholar
Gu, K., Mok, L. & Chong, M. M. W. Regulating gene expression in animals through RNA endonucleolytic cleavage. Heliyon 4, e00908. https://doi.org/10.1016/j.heliyon.2018.e00908 (2018).
Article PubMed PubMed Central Google Scholar
Han, J. et al. Posttranscriptional crossregulation between Drosha and DGCR8. Cell 136, 75–84. https://doi.org/10.1016/j.cell.2008.10.053 (2009).
Article CAS PubMed PubMed Central Google Scholar
Shenoy, A. & Blelloch, R. Genomic analysis suggests that mRNA destabilization by the microprocessor is specialized for the auto-regulation of Dgcr8. PLoS One 4, e6971. https://doi.org/10.1371/journal.pone.0006971 (2009).
Article ADS CAS PubMed PubMed Central Google Scholar
Triboulet, R., Chang, H. M., Lapierre, R. J. & Gregory, R. I. Post-transcriptional control of DGCR8 expression by the Microprocessor. RNA 15, 1005–1011. https://doi.org/10.1261/rna.1591709 (2009).
Article CAS PubMed PubMed Central Google Scholar
Johanson, T. M. et al. Drosha controls dendritic cell development by cleaving messenger RNAs encoding inhibitors of myelopoiesis. Nat. Immunol. 16, 1134–1141. https://doi.org/10.1038/ni.3293 (2015).
Article CAS PubMed Google Scholar
Knuckles, P. et al. Drosha regulates neurogenesis by controlling neurogenin 2 expression independent of microRNAs. Nat. Neurosci. 15, 962–969. https://doi.org/10.1038/nn.3139 (2012).
Article CAS PubMed Google Scholar
Rolando, C. et al. Multipotency of adult hippocampal NSCs in vivo is restricted by Drosha/NFIB. Cell Stem Cell 19, 653–662. https://doi.org/10.1016/j.stem.2016.07.003 (2016).
Article CAS PubMed Google Scholar
Karginov, F. V. et al. Diverse endonucleolytic cleavage sites in the mammalian transcriptome depend upon microRNAs, Drosha, and additional nucleases. Mol. Cell 38, 781–788. https://doi.org/10.1016/j.molcel.2010.06.001 (2010).
Article CAS PubMed PubMed Central Google Scholar
Kim, B., Jeong, K. & Kim, V. N. Genome-wide mapping of DROSHA cleavage sites on primary MicroRNAs and noncanonical substrates. Mol. Cell 66, 258-269 e255. https://doi.org/10.1016/j.molcel.2017.03.013 (2017).
Article CAS PubMed Google Scholar
Cheloufi, S., Dos Santos, C. O., Chong, M. M. & Hannon, G. J. A dicer-independent miRNA biogenesis pathway that requires Ago catalysis. Nature 465, 584–589. https://doi.org/10.1038/nature09092 (2010).
Article ADS CAS PubMed PubMed Central Google Scholar
Babiarz, J. E., Ruby, J. G., Wang, Y., Bartel, D. P. & Blelloch, R. Mouse ES cells express endogenous shRNAs, siRNAs, and other Microprocessor-independent, Dicer-dependent small RNAs. Genes Dev. 22, 2773–2785. https://doi.org/10.1101/gad.1705308 (2008).
Article CAS PubMed PubMed Central Google Scholar
Stavast, C. J. & Erkeland, S. J. The non-canonical aspects of MicroRNAs: Many roads to gene regulation. Cells https://doi.org/10.3390/cells8111465 (2019).
Article PubMed PubMed Central Google Scholar
Nicholson, A. W. Ribonuclease III mechanisms of double-stranded RNA cleavage. Wiley Interdiscip. Rev. RNA 5, 31–48. https://doi.org/10.1002/wrna.1195 (2014).
Article CAS PubMed Google Scholar
Shi, W., Hendrix, D., Levine, M. & Haley, B. A distinct class of small RNAs arises from pre-miRNA-proximal regions in a simple chordate. Nat. Struct. Mol. Biol. 16, 183–189. https://doi.org/10.1038/nsmb.1536 (2009).
Article CAS PubMed PubMed Central Google Scholar
Nguyen, T. L., Nguyen, T. D. & Nguyen, T. A. The conserved single-cleavage mechanism of animal DROSHA enzymes. Commun. Biol. 4, 1332. https://doi.org/10.1038/s42003-021-02860-1 (2021).
Article CAS PubMed PubMed Central Google Scholar
Statello, L., Guo, C.-J., Chen, L.-L. & Huarte, M. Gene regulation by long non-coding RNAs and its biological functions. Nat. Rev. Mol. Cell Biol. 22, 96–118. https://doi.org/10.1038/s41580-020-00315-9 (2021).
Article CAS PubMed Google Scholar
Rodriguez, A., Griffiths-Jones, S., Ashurst, J. L. & Bradley, A. Identification of mammalian microRNA host genes and transcription units. Genome Res. 14, 1902–1910. https://doi.org/10.1101/gr.2722704 (2004).
Article CAS PubMed PubMed Central Google Scholar
Ritchie, W., Legendre, M. & Gautheret, D. RNA stem-loops: To be or not to be cleaved by RNAse III. RNA 13, 457–462. https://doi.org/10.1261/rna.366507 (2007).
Article CAS PubMed PubMed Central Google Scholar
Nguyen, H. M., Nguyen, T. D., Nguyen, T. L. & Nguyen, T. A. Orientation of human microprocessor on primary MicroRNAs. Biochemistry 58, 189–198. https://doi.org/10.1021/acs.biochem.8b00944 (2019).
Article CAS PubMed Google Scholar
Clote, P., Ferre, F., Kranakis, E. & Krizanc, D. Structural RNA has lower folding energy than random RNA of the same dinucleotide frequency. RNA 11, 578–591. https://doi.org/10.1261/rna.7220505 (2005).
Article CAS PubMed PubMed Central Google Scholar
Moss, W. N. The ensemble diversity of non-coding RNA structure is lower than random sequence. Noncoding RNA Res. 3, 100–107. https://doi.org/10.1016/j.ncrna.2018.04.005 (2018).
Article CAS PubMed PubMed Central Google Scholar
Bao, C. et al. mRNA stem-loops can pause the ribosome by hindering A-site tRNA binding. eLife 9, e55799. https://doi.org/10.7554/eLife.55799 (2020).
Article PubMed PubMed Central Google Scholar
Dang, T. L. et al. Select amino acids in DGCR8 are essential for the UGU-pri-miRNA interaction and processing. Commun. Biol. 3, 344. https://doi.org/10.1038/s42003-020-1071-5 (2020).
Article CAS PubMed PubMed Central Google Scholar
Kim, K., Nguyen, T. D., Li, S. & Nguyen, T. A. SRSF3 recruits DROSHA to the basal junction of primary microRNAs. RNA 24, 892–898. https://doi.org/10.1261/rna.065862.118 (2018).
Article CAS PubMed PubMed Central Google Scholar
Havens, M. A., Reich, A. A. & Hastings, M. L. Drosha promotes splicing of a pre-microRNA-like alternative exon. PLoS Genet. 10, e1004312. https://doi.org/10.1371/journal.pgen.1004312 (2014).
Article CAS PubMed PubMed Central Google Scholar
Lee, D., Nam, J. W. & Shin, C. DROSHA targets its own transcript to modulate alternative splicing. RNA 23, 1035–1047. https://doi.org/10.1261/rna.059808.116 (2017).
Article CAS PubMed PubMed Central Google Scholar
Alarcon, C. R., Lee, H., Goodarzi, H., Halberg, N. & Tavazoie, S. F. N6-methyladenosine marks primary microRNAs for processing. Nature 519, 482–485. https://doi.org/10.1038/nature14281 (2015).
Article ADS CAS PubMed PubMed Central Google Scholar
Tomaselli, S. et al. ADAR enzyme and miRNA story: A nucleotide that can make the difference. Int. J. Mol. Sci. 14, 22796–22816. https://doi.org/10.3390/ijms141122796 (2013).
Article CAS PubMed PubMed Central Google Scholar
Suzuki, H. I. et al. Modulation of microRNA processing by p53. Nature 460, 529–533. https://doi.org/10.1038/nature08199 (2009).
Article ADS CAS PubMed Google Scholar
Kawahara, Y. & Mieda-Sato, A. TDP-43 promotes microRNA biogenesis as a component of the Drosha and Dicer complexes. Proc. Natl. Acad. Sci. U S A 109, 3347–3352. https://doi.org/10.1073/pnas.1112427109 (2012).
Article ADS PubMed PubMed Central Google Scholar
Tang, X. et al. Acetylation of drosha on the N-terminus inhibits its degradation by ubiquitination. PLoS One 8, e72503. https://doi.org/10.1371/journal.pone.0072503 (2013).
Article ADS CAS PubMed PubMed Central Google Scholar
Wada, T., Kikuchi, J. & Furukawa, Y. Histone deacetylase 1 enhances microRNA processing via deacetylation of DGCR8. EMBO Rep. 13, 142–149. https://doi.org/10.1038/embor.2011.247 (2012).
Article CAS PubMed PubMed Central Google Scholar
Herbert, K. M., Pimienta, G., DeGregorio, S. J., Alexandrov, A. & Steitz, J. A. Phosphorylation of DGCR8 increases its intracellular stability and induces a progrowth miRNA profile. Cell Rep. 5, 1070–1081. https://doi.org/10.1016/j.celrep.2013.10.017 (2013).
Article CAS PubMed PubMed Central Google Scholar
Harfe, B. D., McManus, M. T., Mansfield, J. H., Hornstein, E. & Tabin, C. J. The RNaseIII enzyme Dicer is required for morphogenesis but not patterning of the vertebrate limb. Proc. Natl. Acad. Sci. U S A 102, 10898–10903. https://doi.org/10.1073/pnas.0504834102 (2005).
Article ADS CAS PubMed PubMed Central Google Scholar
Zou, Y.-R. et al. Epigenetic silencing of CD4 in T cells committed to the cytotoxic lineage. Nat. Genet. 29, 332–336. https://doi.org/10.1038/ng750 (2001).
Article CAS PubMed Google Scholar
Afgan, E. et al. The Galaxy platform for accessible, reproducible and collaborative biomedical analyses: 2016 update. Nucleic Acids Res. 44, W3–W10. https://doi.org/10.1093/nar/gkw343 (2016).
Article CAS PubMed PubMed Central Google Scholar
Bolger, A. M., Lohse, M. & Usadel, B. Trimmomatic: A flexible trimmer for Illumina sequence data. Bioinformatics 30, 2114–2120. https://doi.org/10.1093/bioinformatics/btu170 (2014).
Article CAS PubMed PubMed Central Google Scholar
Dobin, A. et al. STAR: Ultrafast universal RNA-seq aligner. Bioinformatics 29, 15–21. https://doi.org/10.1093/bioinformatics/bts635 (2013).
Article CAS PubMed Google Scholar
Fromm, B. et al. MirGeneDB 2.0: the metazoan microRNA complement. Nucleic Acids Res. 48, D132-d141. https://doi.org/10.1093/nar/gkz885 (2020).
Article CAS PubMed Google Scholar
Friedman, J., Hastie, T. & Tibshirani, R. Regularization paths for generalized linear models via coordinate descent. J. Stat. Softw. 33(1), 1–22 (2010).
Article PubMed PubMed Central Google Scholar
Robin, X. et al. pROC: An open-source package for R and S+ to analyze and compare ROC curves. BMC Bioinform. 12, 77. https://doi.org/10.1186/1471-2105-12-77 (2011).
Article Google Scholar
Robinson, M. D., McCarthy, D. J. & Smyth, G. K. edgeR: A Bioconductor package for differential expression analysis of digital gene expression data. Bioinformatics 26, 139–140. https://doi.org/10.1093/bioinformatics/btp616 (2010).
Article CAS PubMed Google Scholar
Martin, M. Cutadapt removes adapter sequences from high-throughput sequencing reads. EMBnet 17, 3. https://doi.org/10.14806/ej.17.1.200 (2011).
Article Google Scholar
Lorenz, R. et al. ViennaRNA Package 2.0. Algorithms Mol. Biol. 6, 26. https://doi.org/10.1186/1748-7188-6-26 (2011).
Article PubMed PubMed Central Google Scholar
Gao, C. H., Yu, G. & Cai, P. ggVennDiagram: An intuitive, easy-to-use, and highly customizable R package to generate Venn diagram. Front. Genet. 12, 706907. https://doi.org/10.3389/fgene.2021.706907 (2021).
Article PubMed PubMed Central Google Scholar
Tareen, A. & Kinney, J. B. Logomaker: Beautiful sequence logos in Python. Bioinformatics 36, 2272–2274. https://doi.org/10.1093/bioinformatics/btz921 (2019).
Article CAS PubMed Central Google Scholar

Download references

Acknowledgements

This work was supported by grants and fellowships from the National Health and Medical Research Council, Australia (Grant numbers 1079586, 1122384, 1122395 and 1117154). Research performed at St Vincent’s Institute of Medical Research is made possible by the Victorian State Government Operational Infrastructure Support and the Independent Research Institutes Infrastructure Support Scheme of the National Health and Medical Research Council.

Author information

Authors and Affiliations

St Vincent’s Institute of Medical Research, Fitzroy, VIC, 3065, Australia
Karen Gu, Lawrence Mok & Mark M. W. Chong
Department of Medicine (St Vincent’s), University of Melbourne, Fitzroy, VIC, 3065, Australia
Karen Gu & Mark M. W. Chong
Walter and Eliza Hall Institute of Medical Research, Parkville, VIC, 3052, Australia
Matthew J. Wakefield
Department of Obstetrics and Gynaecology, University of Melbourne, Parkville, VIC, 3010, Australia
Matthew J. Wakefield

Authors

Karen Gu
View author publications
You can also search for this author in PubMed Google Scholar
Lawrence Mok
View author publications
You can also search for this author in PubMed Google Scholar
Matthew J. Wakefield
View author publications
You can also search for this author in PubMed Google Scholar
Mark M. W. Chong
View author publications
You can also search for this author in PubMed Google Scholar

Contributions

Conceptualization: K.G. and M.M.W.C.; Methodology: K.G., M.J.W. and M.M.W.C.; Software: K.G.; Validation: K.G. and L.M.; Formal analysis: K.G.; Investigation: K.G., L.M.; Data Curation: K.G.; Writing—Original Draft: K.G. and L.M.; Writing—Review & Editing: K.G. and M.M.W.C.; Visualization: K.G.; Supervision: M.J.W. and M.M.W.C.; Funding acquisition: M.M.W.C.

Corresponding author

Correspondence to Mark M. W. Chong.

Ethics declarations

Competing interests

The authors declare no competing interests.

Additional information

Publisher's note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Supplementary Figures.

Supplementary Tables.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions

About this article

Cite this article

Gu, K., Mok, L., Wakefield, M.J. et al. Non-canonical RNA substrates of Drosha lack many of the conserved features found in primary microRNA stem-loops. Sci Rep 14, 6713 (2024). https://doi.org/10.1038/s41598-024-57330-5

Download citation

Received: 12 November 2023
Accepted: 18 March 2024
Published: 20 March 2024
DOI: https://doi.org/10.1038/s41598-024-57330-5

Comments

By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.