the proteomic datasets of both SE and HYL1. Whereas Brr2a has two homologues, Brr2b (AT2G42270) and Brr2c (AT5G61140), in Arabidopsis, only Brr2a showed ubiquitous expression (Extended Data Fig. 2a) and thus became our focus.

Next, we validated the association of Brr2a with microprocessor components. In co-immunoprecipitation (co-IP) assay, we could readily detect HYL1, but not SE or the control protein HSP70, in Brr2a co-immunoprecipitants (Fig. 1b). Importantly, the HYL1-Brr2a interaction is RNA independent as the addition of RNase did not decrease the amount of HYL1 in the immunoprecipitants of Brr2a (Fig. 1b). Notably, bimolecular fluorescence complementation (BiFC) assays showed that Brr2a could complement HYL1 when each was fused with an N-or C-terminal truncated portion of yellow fluorescent protein (YFP) (Fig. 1c), but this type of fluorescence complementation was not observed in the combinations of Brr2a with SE or DCL1 (Fig. 1c). Moreover, Brr2a showed co-localization with HYL1 when expressed in Nicotiana benthamiana (Extended Data Fig. 2b). These results indicate that Brr2a is a bona fide partner of HYL1 and might transiently interact with SE and DCL1.

The null mutation of Brr2a (brr2a-1) is embryo lethal. A hypomorphic allele of Brr2a mutant, cäö, hereafter referred as brr2a-2, harbours a point mutation within the N-helicase cassette (Extended Data Fig. 2c) and displays developmental abnormalities, including serrated leaves and early flowering 23 . Since brr2a-2 is a weak allele, we generated transgenic plants using an artificial miRNA system 24 to knock down Brr2a transcripts (Extended Data Fig. 2d). The transgenic plants produced abundant amiR-Brr2a (Extended Data Fig. 2e), resulting in significant reduction of the Brr2a transcript in comparison with the wild-type control (Fig. 1d). Sequence alignment suggested that amiR-Brr2a might have certain sequence complementarity with two additional transcripts, AT2G18110 and AT1G27750, raising concerns of potential off-target effects (Extended Data Fig. 2f). However, these concerns were cleared by 5′ rapid amplification of cDNA ends-polymerase chain reaction (RACE-PCR) and quantitative PCR with reverse transcription (RT-qPCR) assays, which detected neither the AGO-cleaved products from the two transcripts nor their expression changes in the transgenic plants vs Col-0 (Extended Data Fig. 2g,h). Thus, we named these amiR-Brr2a transgenic lines as brr2a-3. Notably, brr2a-3 phenocopied the morphological defects of brr2a-2, but with more severe phenotypes, including narrow and curly leaves, delayed plant growth, dwarfism and lesser rosette leaves before flowering (Fig. 1e,f). These developmental defects are reminiscent of hyl1-2 and se-2 (Fig. 1e,f), suggesting that Brr2a might have shared functions with the microprocessor components in regulating plant development.

Next, we performed RNA-seq analysis to examine the impact of Brr2a on the transcriptome. RNA-seq results further showed the significant reduction of Brr2a (Extended Data Fig. 3a). Moreover, we found 7,069 differentially expressed genes (DEGs) in brr2a-3 vs Col-0 (Fig. 1g and Supplementary Table 3). Among these DEGs, 3,155 and 3,914 genes were decreased and increased, respectively. Meanwhile, we observed that 1,170 and 2,627 DEGs were respectively downregulated and upregulated in hyl1-2 (ref. 25) (Fig. 1h,i). The number of DEGs in hyl1-2 was much smaller than that of brr2a-3, suggestive of a broader impact of loss-of-function mutation in brr2a-3 on the splicing process. Furthermore, there was a significant overlap of DEGs, with 958 upregulated and 413 downregulated genes shared between hyl1-2 and brr2a-3 (Fig. 1h,i). Conversely, when comparing the DEGs with opposite expression patterns, no significant overlap was found (Fig. 1h,i). Moreover, a similar pattern of DEGs was also observed between brr2a-3 and se-2 (ref. 26) (Extended Data Fig. 3b). Thus, RNA-seq analysis indicated that Brr2a and HYL1/SE might impact some genetic pathways in common in vivo.

Subsequent Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis implicated the brr2a-3-upregulated genes in the spliceosome Some species of pri-miRNA contain introns, and splicing out of introns from the pri-miRNAs is an essential step ahead of their processing by the microprocessor. Thus, defective splicing in the mutants of spliceosome components can impair pri-miRNA processing 6,7 . Some spliceosome members can enhance the transcription of MIRNA loci and promote the co-transcriptional assembly of the microprocessor, exemplified by Prp40 which typically functions as a U1 small nuclear ribonucleoprotein (snRNP) auxiliary protein 8 . Furthermore, spliceosome components, including Serrate-Associated Protein 1 (SEAP1, homologous to human SART3 and yeast Prp24) 9 , AAR2 (a U5 snRNP assembly factor) 10 and JANUS (homologue of the U2 snRNP assembly factor) 11 , can facilitate the formation of D-bodies where the microprocessor is often present. However, the mechanisms of spliceosome components appear to be diverse and their direct impact on pri-miRNA processing remains underexplored.

Brr2, a component of the spliceosome subunit U5, is well known for its canonical helicase function in unwinding the U4/U6 duplex to activate the spliceosome 12 . In contrast to other members of the Ski2-like family, Brr2 is characterized by its notably large molecular weight and two repeat RH cores (referred to as the N-helicase and C-helicase cassettes, respectively) 13 . In metazoans, upon binding to substrates, the Brr2 N-helicase hydrolyses ATP to facilitate its helicase function, thereby reshaping RNA substrates, whereas the C-helicase cassette lacks such a function but is able to stimulate the activity of the N-helicase cassette 14 . Furthermore, its conserved motif II (DEIH) and motif III (SAT) within the N-helicase cassette have been reported to be associated with ATPase and RH activities, respectively 15,16 . It has been also reported that superfamily 2 RHs can not only translocate along RNA to reshape its configuration, but also perform local strand unwinding in a non-processive manner 1 . Nevertheless, whether and how Brr2a has a novel function in any molecular processes other than splicing events is unknown.

Here we report that Arabidopsis Brr2a, the homologue of human Brr2, is a bona fide partner of the microprocessor, promoting levels of miRNAs derived from both intron-containing and intron-lacking pri-miRNAs. Brr2a was able to directly interact with pri-miRNAs in vivo and in vitro. The binding to pri-miRNA significantly enhanced the ATPase activity of Brr2a. Importantly, Brr2a exhibited the capability to unwind pri-miRNA substrates and the function relies on its DEIH and SAT motifs. Consequently, transgenic plants with these variants failed to restore miRNA levels in brr2a-2. Finally, dimethyl sulfate mutational profiling with sequencing (DMS-MaPseq) of the RSS of pri-miRNAs elucidated that Brr2a plays a pivotal role in ensuring the proper folding of most tested pri-miRNAs that is conducive to subsequent processing by the microprocessor. Thus, this study revealed that Brr2a, beyond its conventional involvement in splicing, plays a novel role in miRNA production.

To identify new factors with the microprocessor, we mined the interactome of HYL1 (ref. 17) and SE 18 . Apart from DCL1 (ref. 19), CHR2 (ref. 5), HEN2 (ref. 18), RH6/8/12 (ref. 20) and RH11/37 (ref. 21), some other RHs with varying peptide reads were also recovered in the MS/ MS analysis (Extended Data Fig. 1 and Supplementary Table 1). To explore whether the new candidates might have any potential correlation with the miRNA biogenesis pathway, we conducted an in silico transcriptome-wide association analysis with DCL1 (ref. 22). Among the RHs, RH11 and RH37, which act in small RNA (sRNA) loading and/or stabilization in extracellular vesicles 21 , showed only moderate association of its expression dynamics with that of DCL1 (Extended Data Fig. 1, and Supplementary Tables 1 and 2). In contrast, CHR2, Brr2a, Brr2c, CHR5 and UPF1 displayed the most significant co-expression patterns with DCL1 (Fig. 1a, Extended Data Fig. 1 and Supplementary Table 2). Among them, Brr2a (AT1G20960) was the only protein identified in Article https://doi.org/10.1038/s41477-024-01788-8 pathway, probably due to feedback regulation of defective splicing in the mutant (Extended Data Fig. 3c). Other enriched pathways included ribosome biogenesis, glucosinolate biosynthesis, DNA replication, RNA degradation, and alanine, aspartate and glutamate metabolism pathways (Extended Data Fig. 3c). On the other hand, the plant hormone signal transduction and circadian rhythm pathways exhibited downregulation in brr2a-3 vs Col-0 (Extended Data Fig. 3c). Notably, most of these pathways were also observed in DEGs shared by brr2a-3 and hyl1-2 (Fig. 1j), further suggestive of their possible co-regulatory impact on RNA processing. We next assessed sRNA profiles in brr2a-3 with Col-0 as a control (Extended Data Fig. 4a). The sRNA-seq analysis showed that the portions of miRNA and trans-acting small interfering RNA (tasiRNA) were clearly reduced, whereas the sRNAs derived from ribosomal RNA (rRNA) increased in brr2a-3 (Extended Data Fig. 4b). Further analysis showed that among 291 detectable miRNAs, 136 showed decreased expression, whereas only 24 were upregulated in brr2a-3 (Fig. 1k and Supplementary Table 4). Notably, most of the downregulated miR-NAs were HYL1 dependent 27 (Fig. 1l). Consistently, the expression of miRNA targets was broadly increased in brr2a-3 vs Col-0 (Extended Data Fig. 4c). Taken together, these findings indicate that Brr2a has physical and genetic associations with the microprocessor, with further suggestion that the protein might function in the miRNA pathway.

Brr2a boosts miRNA yield via a splicing-independent route Some pri-miRNAs contain introns, requiring coordinated processing with splicing and dicing events. It has been reported that alternative splicing (AS) of pri-miRNAs, as seen in se-2, can accumulate aberrant isoforms, thereby causing fluctuations in production of their cognate miRNAs 28 . We next detected whether Brr2a regulated AS of pri-miRNAs to modulate miRNA production. We found that pri-miR162a, which harbours two introns, exhibited increased isoforms of either skipped-exon or retention-intron. However, the functional isoform was reduced in brr2a mutants, reminiscent of se-2, compared with Col-0 and hyl1-2 (Fig. 2a). Consequently, miR162a levels were significantly reduced in brr2a mutants vs Col-0 (Fig. 2b). Similar splicing defects were also observed with pri-miR156d and pri-miR171b among others. These splicing defects probably reduced the amount of processable forms of pri-miRNAs and at least partially contributed to decreased accumulation of miRNAs in the mutants vs Col-0 (Fig. 2a,b). These results suggest that Brr2a could indeed exert a canonical role in splicing to promote miRNA production.

Notably, although pri-miR172b exhibited proper splicing in brr2a mutants, as observed in Col-0 and hyl1-2, it still produced less mature miR172b (Fig. 2a,b). Thus, we hypothesized that Brr2a might act in a splicing-independent mechanism to facilitate miRNA production. To test this, we revisited the sRNA-seq data, categorizing miRNAs into two groups on the basis of their origin-either intron-containing or intron-lacking pri-miRNAs. Surprisingly, we observed a notable reduction in mature miRNAs that were derived not only from intron-containing pri-miRNAs (33/87) (Fig. 2b) but also from those lacking introns (103/204) (Fig. 2c). Moreover, more cases were readily validated by RT-qPCR (Fig. 2d), suggestive of a potential novel function of Brr2a beyond its canonical role in splicing.

Splicing defects on the transcripts of microprocessor components could lead to a decrease in miRNA production 29 . To test this possibility, we assessed the transcript integrity of microprocessor and RNA-induced silencing complex components in brr2a mutants. First, both RNA multivariate analysis of transcript splicing (rMATS) and Integrative Genomics Viewer (IGV) files indicated that the splicing patterns of SE, DCL1 and HYL1 among other transcripts showed no obvious defect in brr2a-3 vs Col-0 (Extended Data Fig. 5a,b and Supplementary Table 5). Next, RT-PCR analysis further confirmed that the splicing machinery in brr2a mutants could still ensure the completion of DCL1 and AGO1 transcripts, probably due to the hypomorphic nature of brr2a alleles and/or the functional redundancy of Brr2b and 2c (Fig. 2e). Neither the transcript nor protein levels of AGO1, DCL1 and SE were reduced in brr2a-3 compared with those of Col-0 (Extended Data Fig. 5c,d). Intriguingly, we observed a significant upregulation of both RNA and protein levels of HYL1 in the brr2a mutants (Extended Data Fig. 5c,d). These accumulations probably resulted from feedback regulation, which parallels the previously noted upregulation of the spliceosome pathway in brr2a-3 (Extended Data Fig. 3c). Despite this, the formation of dicing bodies was seemingly unaffected in brr2a-3 vs Col-0 (Extended Data Fig. 5e). In addition, the expressions of other key components in the miRNA pathway were not reduced (Extended Data Fig. 5f). Hence, these data further suggest a splicing-independent function of Brr2a in miRNA production.

We then investigated whether Brr2a was involved in the transcriptional machinery at MIRNA loci. Our initial focus was on MIR159a and MIR159b, which generate pri-miR159a and pri-miR159b, respectively, with both lacking introns and producing less mature miRNAs in brr2a-3 (Supplementary Table 4). We crossed the pMIR159a/b::Flag-4xMYC (FM) -β-glucuronidase (GUS) transgenic lines with brr2a-3 separately and analysed the GUS reporter transcription in the brr2a-3 homozygote background from F 3 segregation lines. Through GUS staining assays and western blot analysis, we found comparable and slightly increased accumulation levels of GUS for MIR159a and MIR159b in brr2a-3 vs Col-0, respectively (Fig. 2f and Extended Data Fig. 5g). This prompted us to further investigate the transcription level of MIR159b in brr2a mutants via a Pol II chromatin immunoprecipitation (ChIP) assay. Again, the overall Pol II occupancy on MIR159b slightly increased in brr2a mutants vs Col-0 (Fig. 2g). For other tested loci, Brr2a did not alter their transcription, regardless of whether pri-miRNAs have introns or not (Fig. 2g). Taken together, we conclude that Brr2a does not influence miRNA production via promoting MIR transcription.

We next hypothesized that RH Brr2a might bind to pri-miRNAs and modulate their RSS, thereby affecting miRNA production. Since Brr2a's canonical unwinding activity relies on the U4/U6 duplex structure 14,30,31 , we first examined whether pri-miRNA could serve as a substrate for Brr2a. To test this, we generated transgenic plants carrying pBrr2a::FM-gBrr2a in the se-2 and hyl1-2 backgrounds, where pri-miRNAs were substantially accumulated due to defective processing. We then performed ribonucleoprotein immunoprecipitation (RIP) assays using anti-Flag antibodies, with se-2 as a control. The RIP-qPCR results revealed a significant enrichment of pri-miRNAs in Brr2a's immunoprecipitates compared with the control IP, regardless of whether they contain introns (for example, pri-miR156d, pri-miR162a, and pri-miR171b and 172b) or lack introns (for example, pri-miR159b, pri-miR162b and so on) (Fig. 3a). Conversely, IGN5, a non-coding RNA transcribed by Pol V 32 , was only detectable in the input but not in the immunoprecipitates (Fig. 3a). Notably, the association of Brr2a with pri-miRNAs was also detectable in the hyl1-2 background, but the amount was significantly reduced in hyl1-2 vs se-2 (Fig. 3a). Taken together, we conclude Fig. 3 | RH Brr2a binds and remodels pri-miRNAs. a, RIP assays show that Brr2a could bind intron-containing and intron-lacking pri-miRNAs in vivo. IPs were conducted with anti-Flag M2 beads, and the resulting Brr2a-bound RNAs were then used for RT-qPCR analysis. The se-2 and Pol V transcript IGN5 served as negative control plant and RNA, respectively. Unpaired two-way ANOVA with Dunnett's multiple comparisons test; NS P ≥ 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. na, no signal detected. b-g, EMSA and binding curves exhibit the distinct binding affinity of C-Brr2a (b and c), N-Brr2a (d) and its variants, N-EQ (e) and N-AA (f), to pri-miR159b in vitro. g, K d , appK d and R 2 values were calculated with s.d. of 3 independent replicates fitting a Hill slope model. h,i, TLC assays (h) and statistical analysis (i) display the ATPase activity of truncated Brr2a and its variants in the presence or absence of pri-miR159b RNAs in different time courses. j, Schematic illustrating the procedures for generating three kinds of dsRNA duplexes and performing unwinding assay. See Methods for details. r.t., room temperature. k-m, Results of RNA native PAGE gels reveal the accumulation of released ssRNA from pri-miR159b duplex (k), pri-miR166f duplex (l) and dsRNA duplex with a 3′ overhang (m) in unwinding assays. n, Statistical analysis of unwinding assays. Both paired and unpaired RNAs were quantified to generate the curves fitting the Michaelis-Menten model, with s.d. from 3 independent experiments. b-i, N/C-Brr2a represents N/Cterminal truncated Brr2a; N-EQ or N-AA represents the variants of N-Brr2a carrying mutations E640Q or S676A and T678A. j-n, Brr2a-EQ and Brr2a-AA represent the variants of full-length Brr2a carrying mutations E640Q and S676A T678A, respectively. Data are mean ± s.d.

Article https://doi.org/10.1038/s41477-024-01788-8

that Brr2a directly binds to both intron-containing and intron-lacking pri-miRNAs in vivo and the association is facilitated by HYL1. This observation is further supported by a recent RNA pull-down analysis using intron-lacking pri-miR398 as the bait that recovered Brr2a as one of pri-miRNA binding proteins 33 . We next studied whether Brr2a bound to pri-miRNAs in vitro. Since purification of the full-length Brr2a was undoable from E. coli, we generated N-(487-1,288 aa, N-Brr2a) or C-(1,289-2,171 aa, C-Brr2a) helicase cassettes of Brr2a variants, each encompassing a full-length helicase cassette (Extended Data Figs. 2c and 6a). Electrophoretic mobility shift assays (EMSAs) showed that C-Brr2a exhibited a weak binding affinity to folded pri-miR159b, with a K d of ~221.2 ± 2.2 nM (Fig. 3b,c and Extended Data Fig. 6b). In sharp contrast, N-Brr2a displayed a robust binding affinity to folded pri-miR159b, with a K d of ~13.3 ± 1.7 nM (Fig. 3d,g and Extended Data Fig. 6b).

To further study whether the association of N-Brr2a with pri-miRNAs depended on its ATPase or helicase activity, we generated N-Brr2a variants with compromised ATPase activity (N-EQ, E640Q) or defective helicase function (N-AA, N-S676A T678A), targeting two motifs that are highly conserved in animals, plants and fungi (Extended Data Figs. 2c and 6a). EMSA assays uncovered distinct effects of the two mutations on RNA binding affinity. The N-EQ seemingly increased RNA retention within the RNP, with a 1.5-fold enhancement in binding affinity with an appK d (apparent K d ) of ~8.2 ± 2.0 nM (Fig. 3e,g and Extended Data Fig. 6b). Conversely, the N-AA mutation hindered the formation of the RNP complex, resulting in a 3.5-fold reduction in RNA binding affinity compared with that of N-Brr2a with a K d of ~47.8 ± 1.7 nM (Fig. 3f,g and Extended Data Fig. 6b). However, when switching to ssRNA, all proteins exhibited negligible RNA binding affinity (Extended Data Fig. 6c-e). Taken together, we conclude that Brr2a, mainly relying on its N-terminal helicase domain, could bind to pri-miRNAs.

We then investigated whether Brr2a exhibited ATPase activity. In vitro ATPase reconstitution followed by thin layer chromatography (TLC) assays showed that both N-Brr2a and C-Brr2a exhibited weak capability to hydrolyse ATP in the absence of RNA. Specifically, N-Brr2a hydrolysed ~0.093% 32 [P]ATP supplied per min, showing slightly stronger activity than C-Brr2a, which hydrolysed ~0.07% ATP per min (Fig. 3h,i, Extended Data Fig. 7a and Supplementary Table 6). Notably, the ATP hydrolysis activity of N-Brr2a, but not C-Brr2a, was significantly stimulated upon the supply of pri-miR159b, resulting in ~3-fold increase, hydrolysing ~0.27% ATP per min (Fig. 3h,i, Extended Data Fig. 7a and Supplementary Table 6). Notably, both ATPase activity and the stimulation were eliminated by EQ and AA mutations, as the two variants barely hydrolysed the ATP (Fig. 3h,i, Extended Data Fig. 7a and Supplementary Table 6). These observations are reminiscent of previous findings that motif II (DEIH) accounts for ATPase activity 15,16 and motif III (SAT motif) is required for both helicase activity and ATPase activity [34][35][36] . Thus, these results suggest that pri-miR159b could stimulate the ATPase activity of Brr2a.

We hypothesized that Brr2a might unwind pri-miRNAs in vitro. To test this, we first detected whether truncated Brr2a and its variants could unwind a dsRNA duplex of pri-miR159b. This dsRNA duplex was generated by annealing the 3′ flanking single-stranded region of pri-miR159b with a short 32 P-labelled RNA fragment, resulting in a complicated structure with a 3′ stem loop, central ssRNA and stem region, and a 5′ stem loop, reminiscent of core domain and stem loops of U4 in the U4/U6 duplex 37 (Fig. 3j). Unwinding assays found that the N-terminal cassette, but not the C-terminal cassette, of Brr2a could release certain portions of the ssRNA from the dsRNA duplex (Extended Data Fig. 7b). Furthermore, the helicase activity of N-Brr2a depended on its DEIH and SAT motifs (Extended Data Fig. 7b). We next investigated the helicase activity of full-length Brr2a and its variants with compromised activities of ATPase or helicase immunoprecipitated from transgenic plants (Extended Data Fig. 7c). For this assay, we generated two other types of dsRNA duplexes: a nicked pri-miR166f duplex that contained a stem-loop RNA structure 5 and a full-paired dsRNA duplex with a ~110 nt 3′ overhang 38 (Fig. 3j). Interestingly, wild-type Brr2a readily displaced a majority of the short RNA fragments from the pri-miR159b duplex, whereas two variants (Brr2a-EQ and Brr2a-AA) did not, as the amount of released free short RNA from these variants was comparable to that from the control IP (Fig. 3k and Extended Data Fig. 7d). Furthermore, Brr2a could still unwind the nicked pri-miR166f and dsRNA duplex and exhibited comparable efficiencies for these substrates (Fig. 3l,m and Extended Data Fig. 7d), although it was less efficient than with the pri-miR159b duplex (Fig. 3n). Together, these results indicate that plant Brr2a could promiscuously remodel substrates with different structures.

To further investigate whether helicase activity is necessary for Brr2a to promote miRNA production, we generated transgenic complementation plants in the brr2a-2 background expressing either ATPase activity-compromised variant Brr2a-EQ (brr2a-2 pBrr2a::FM-Brr2a-EQ, E640Q) or RH function compromised variant Brr2a-AA (brr2a-2 pBrr2a::FM-Brr2a-AA, S676A T678A) driven by the Brr2a native promoter. While pBrr2a::FM-Brr2a transgenic plant could fully rescue the morphologic defect of brr2a-2 (Fig. 4a,b), the transgenic plants carrying the two variants still exhibited developmental defects as seen in brr2a-2 (Fig. 4a,b). Consistently, these plants showed significant reduction in the levels of miRNAs derived from both intron-containing and intron-lacking pri-miRNAs in the transgenic plants (Fig. 4c). The mutations in ATPase and helicase activities did not impact the Brr2a interaction with HYL1 (Fig. 4d). However, the RIP assays showed that the EQ variant was enriched with even more pri-miRNAs compared with Brr2a. Clearly, the Brr2a-EQ-bound pri-miRNAs did not proceed for processing in the hypomorphic alleles (Fig. 4e and Extended Data Fig. 7e). By contrast, Brr2a-AA variant formed less stable or impaired pri-miRNA-protein complexes (Fig. 4e and Extended Data Fig. 7e). These results are consistent with their differential RNA-binding affinities in EMSA (Fig. 3b-g). Taken together, these results indicate that the proper helicase activity of Brr2a is indeed essential for miRNA production in plants.

We next assessed how Brr2a impacted RSS of pri-miRNAs in vivo by conducting DMS-MaPseq. Briefly, in DMS-MaPseq, DMS treatment of living Arabidopsis plants induces modifications specifically on unpaired adenosine (A) and cytosine (C) residues. These modified bases introduce mutations during the reverse transcription process that can be decoded via high-throughput sequencing analysis, enabling us to profile mutational rates to capture RSS of pri-miRNAs in vivo 39 (Extended Data Fig. 8a). In DMS-MaPseq analysis, we found a significant increase in mismatch ratios at A and C positions compared with G and U in all samples, confirming the success of the DMS treatment (Extended Data Fig. 8b). We next calculated Gini indexes of pri-miRNAs to reflect the variation of RSS in the transcripts. Typically, a higher Gini index indicates more structured and paired regions in RNAs. Interestingly, the Gini index was significantly increased in brr2a-3 compared with Col-0, indicating that pri-miRNAs were indeed more structured in the mutant vs Col-0 (Fig. 5a).

Because of the heterogenicity in shapes and lengths, plant pri-miRNAs can be processed from the base to loop (BTL), from loop to base (LTB) or even bidirectionally 4,40 . In this scenario, we divided the pri-miRNAs with reduced miRNA production in brr2a-3 into two categories of BTL and LTB, and then performed ensembled DMS-MaPseq analysis separately. For the BTL-processed pri-miRNAs, DMS displayed higher but varying activities at the sites of -17 to -13 nt and -9 to -5 nt in the 5′ end lower stem regions of pri-miRNAs in Col-0, indicative of internal loops or bulges in the regions (Fig. 5b). Importantly, DMS activity profiling along the lower stem regions showed altered patterns in the brr2a-3 vs Col-0 (Fig. 5b). For instance, we observed decreased DMS activities at -17 to -15 nt in the 5′ lower arms in brr2a-3, but increased and varying DMS activities at -14 nt to -13 nt, indicating a shifting of internal reference loops from -17 to -15 nt to -14 to -13 nt in the 5′ lower arms of pri-miRNAs in brr2a-3 vs Col-0 (Fig. 5b, green dashed box). In parallel, we detected decreased DMS activities at 15 nt and 12 nt in 3′ arms lower stems in brr2a-3, and increased DMS activities at 14 nt and 13 nt, which suggested a marginal shift of the reference loops from 15-14 nt to 13-12 nt at the 3′ arms of lower stems at these pri-miRNAs in brr2a-3 vs Col-0 (Fig. 5c, green dashed box). Furthermore, increased sizes of the internal loops at the -9 to -6 nt (Fig. 5b, red dashed box) positions at the 5′ arm and 4-6 nt (Fig. 5b, pink dashed box) and 13-14 nt (Fig. 5b, blue dashed box) positions in the miRNA/* duplexes, as well as coordinated increase of 6-3 nt (Fig. 5c, red dashed box) positions at the 3′ arm and -5 to -7 nt (Fig. 5c, pink dashed box) and -13 to -14 nt (Fig. 5c, blue dashed box) positions in the miRNA/* duplexes, were seemingly observable in the pri-miRNAs in the brr2a-3 vs Col-0. All these changes could be expected to have adverse impacts on pri-miRNA processing.

Computational analysis of the LTB-processed pri-miRNAs became more complicated than the BTL-patterned ones as many of the LTB pri-miRNAs were sequentially processed before reaching the miRNA/* duplexes 41 . However, we could observe variations in DMS activity in the upper stems and terminal loops between brr2a-3 and Col-0 (Extended Data Fig. 8c). Such difference might be meaningful as the internal loops in this region significantly impacted the production of mature miRNAs. In addition, the high and contrasted variations in DMS activities in the duplex regions of both brr2a-3 and Col-0 were also detectable.

Due to heterozygosity of pri-miRNAs in plants, ensembled DMS-MaPseq did not allow us to simply conduct statistical significance tests of DMS reactivity in each position because individual pri-miRNAs have their internal loops/bulges that are of viable sizes and in different sites. Alternatively, Brr2a might not act on the same locations of all pri-miRNAs. To accurately decipher the precise impact of Brr2a on RSS remodelling for all pri-miRNAs, we focused on a handful pri-miRNAs that were of sufficient reads, and thoroughly assessed the structural changes of the pri-miRNAs in brr2a-3 vs Col-0. Pri-miR158a follows a BTL-typed

a Col-0 Brr2a brr2a-2; pBrr2a::FM-brr2a-2 c miR159 1.5 1.0 0.5 0 * NS NS NS Intron-lacking Mixture Introncontaining Relative expression of miRNAs * ** NS * * NS * ** NS ** *** **** * ** NS NS 2.0

Col-0 brr2a-2; pBrr2a::FM-Brr2a brr2a-2; pBrr2a::FM-Brr2a-EQ brr2a-2; pBrr2a::FM-Brr2a-AA *** e brr2a-2; pBrr2a::FM-Brr2a brr2a-2; pBrr2a::FM-Brr2a-EQ brr2a-2;pBrr2a::FM-Brr2a-AA processing. In this scenario, the microprocessor sets the first cleavage site at 15-17 nt away from its ssRNA-dsRNA internal reference site at the basal region (Fig. 5d). Whereas DCL1 prefers to cut pri-miRNAs at the edge of small internal loops or mismatches in wild-type plants, an enlarged internal loop at the first cleavage site would suppress miR158 production in the brr2a-3 vs Col-0. Similar patterns were also observed with pri-miR170 and pri-miR172b. For these pri-miRNAs, the initial reference ssRNA-dsRNA junction regions at the lower stems became obscure in brr2a-3 vs Col-0 (Extended Data Fig. 9a). Furthermore, the numbers and sizes of internal loops/bulges were increased in both the lower stems and the miRNA/* duplex regions of pri-miRNAs in the brr2a-3 vs Col-0. Two pri-miRNAs, pri-miR172a and pri-miR844a, also followed this change. Moreover, these two pri-miRNAs also exhibited increased DMS activities in the upper stems, indicative of unpaired structures in brr2a-3 vs Col-0 (Fig. 5e and Extended Data Fig. 9b). This change would increase the likelihood of pri-miRNA processing from LTB, consequently impairing productive processing. Pri-miR159a is processed from an LTB direction, and four sequential cleavages are entailed to successfully release mature miR159a/* duplex. This pri-miRNA produced a remarkable high reads number of mature miR159 but significantly reduced reads of intermediate products (~60,000 vs 300) in Col-0, whereas the opposite pattern (35,000 vs 2,000) was observed in brr2a-3 (Fig. 5f). This contrast was pronounced and clearly resulted from the structural difference of pri-miR159a in Col-0 vs brr2a-3. The microprocessor favours the cutting sites at or close to the internal loops. Here, pri-miR159a had several hotspots of DMS reactivity, indicative of internal loops or mismatches, along the upper stems in both Col-0 and brr2a-3 (Fig. 5g). However, these internal loops or mismatches were intriguingly located at or adjacent to the cutting sites in Col-0, potentially facilitating processivity of DCL1 activity from the terminal loop to the lower base. By contrast, the cleavage sites were in relatively more folded regions of the upper stem in brr2a-3 which would be considered as a suboptimal structure for DCL1 function, hindering its processivity. Furthermore, DMS activity hotspots were also detected in the middle of the first set of sRNA/* duplexes (with 1,267 and 404 reads) in brr2a-3 that did not promote DCL1 to proceed with the processing. In addition, the lower stem of pri-miR159a harboured a bigger basal loop in Col-0, whereas the nucleotides in the counterpart regions were much more paired in brr2a-3 (Fig. 5g). This conformation is likely to further inhibit accurate and efficient processing of miRNA/* from pre-miRNAs in the mutant vs Col-0, as both animal Dicer and plant DCL1 recognize the loop-bulge structures in addition to the 5′ and 3′ ends of pre-miRNAs for accurate processing 4,42 .

Similar to pri-miR159a, pri-miR162a and pri-miR400 are processed from LTB. The pri-miRNAs exhibited a smaller terminal loop that were less instructive to DCL1 (Extended Data Fig. 9c). Notably, there were larger loops in the lower stems but located ~10 nt away from the miRNA/* duplexes of pri-miR162a and pri-miR400 in brr2a-3, but how this change affects miRNA production is uninterpretable based on our current understanding of the microprocessor mechanism.

Pri-miRNAs with branched terminal loops or big terminal loops are subjected to bidirectional processing, leading to a dynamic of productive and abortive processes 5 . Pri-miR165a has previously been predicted to have a structure of branched terminal loops according to RNAfold that is solely grounded on a thermostability algorithm 4 . However, it has been a mystery that this pri-miRNA produces substantial amount of miR165a even with a predicted branched terminal loop. In fact, DMS activity was detectable but weak in the upper stem in Col-0, suggesting that pri-miR165a is in a dynamic transition between the linear folded hairpin and a structure with a terminal branched loop in vivo, underscoring its capacity to produce abundant miR165 as observed in pri-miR165b 4 (Fig. 5h). Surprisingly, pri-miR165a exhibited much stronger DMS activity indicative of a structure with standardized branched terminal loops in brr2a-3, which would trigger abortive processing and resultant lower accumulation of miR165 in the mutant.

Different from other miRNAs, miR157a and miR2112, whose pri-miRNAs are processed from LTB, were accumulated in brr2a-3 vs Col-0. Interestingly, both pri-miR157a and pri-miR2112 have more unpaired regions in the upper stem in brr2a-3 vs Col-0, indicating a greater propensity to produce higher levels of miR157a and miR2112, respectively (Extended Data Fig. 9d). Our DMS-MaPseq analysis also detected increased DMS activities in pri-miR164a, suggestive of more mismatches in the mutant vs Col-0 (Extended Data Fig. 9e). Interestingly, these positions with increased DMS activities were mainly located within pre-existing internal and terminal loops, which did not obviously alter the RSS of pri-miR164a and miR164 production (Fig. 2d and Supplementary Table 4). Taken together, these results suggest that Brr2a is typically involved in remodelling the secondary structure of pri-miRNAs for their proper processing, whereas in a very few cases, the protein might hinder the production of miRNAs (Fig. 6).

Here we report a non-canonical function of Brr2a and elucidate its new mechanism in miRNA biogenesis. While Brr2a could act as an essential spliceosome component to regulate the splicing of intron-containing pri-miRNAs to generate processable forms of pri-miRNAs for the microprocessor, this very protein could also be recruited by HYL1 and directly associate with pri-miRNAs, utilizing its helicase activity to modulate their RSS to present generally more optimally structured substrates for DCL1, resulting in increased accumulation of miRNAs.

Mechanistically, Brr2a represents a distinctive paradigm wherein helicase activity is employed to restructure pri-miRNAs, thereby facilitating their processing and increasing miRNA yield. This is different from our earlier-reported RH CHR2 which typically remodels pri-miRNAs to inhibit their processing, leading to reduced miRNA production 5 . Several other RHs, such as RH6/8/12 (ref. 20), MAC7 (ref. 17) and Prp28 (SMA1) 29 , have been reported to contribute to miRNA Fig. 5 | Brr2a remodels RSS of pri-miRNAs for efficient processing in vivo. a, The Gini index shows an increase in pri-miRNA structural complexity in brr2a-3 vs Col-0. The Gini index was calculated using sliding windows with a size of 50 nt and a step length of 25 nt. A larger numerical value indicates greater structural complexity. Unpaired two-sided t-test; **P < 0.01. b,c, Meta profiles show the structural switch at the 5′ arm (b) and the 3′ arm (c) of BTL-processed pri-miRNAs in brr2a-3 vs Col-0. Green dashed boxes, the shifted loops. Red and pink/blue dashed boxes, increased sizes of internal loops at the 5′ arm and miRNA/* duplexes, respectively. In the RNA plot and the x axis of the statistical analysis graph, lower stem and miRNA/* duplex regions are in red and pink/ blue, respectively. Position 0 and numbers, the base and the distances to the processing site in the miRNA/* region. d, RSS of pri-miR158a, BTL-typed pri-miRNAs, exhibits structural switch at the first cleavage site in brr2a-3 vs Col-0. e, RSS of pri-miR172a, BTL-typed pri-miRNAs, exhibits structural switch at the upper stem and miRNA/* duplex regions in brr2a-3 vs Col-0. f,g, RSS of pri-miR159a, LTB-typed pri-miRNAs, exhibits structural switch from the upper stems to the lower stem in brr2a-3 vs Col-0. f, the IGV profiles of miR159a in the sRNA-seq, with y axis shown in log scale. g, numbers and grey vertical lines are the reads of the major cleavage products of pri-miR159a in Col-0 and brr2a-3 (g). h, RSS of pri-miR165a, bidirectionally processed pri-miRNAs, exhibits structural switch at the branched terminal loops in brr2a-3 vs Col-0. The BTL and LTB processing of pri-miR165a produces functional and abortive products, respectively. Two dynamic structures of pri-miR165a in Col-0 are labelled with #1 and #2. Lines, medians; boxes, quartiles; whiskers, minimum to maximum (a-c). Molecular rulers are indicated in grey (d and e). For d-h, pink and blue circles/ short lines respectively represent miRNA and miRNA*; grey arrowheads label the first cutting sites; differences are highlighted by green dashed boxes; DMS activity from high to low is colour coded in a gradient from red to green and grey, as specified in the relevant figure legends. production, and their mechanisms are interpreted to impact the formation of D-bodies, transcriptional regulation of pri-miRNAs and splicing of the microprocessor components' transcripts. However, whether these proteins have helicase activities and/or whether their functions require the helicase activities are not clear. In contrast, we demonstrated that Brr2a could hydrolyse ATP and unwind dsRNA in vitro (Fig. 3h,j and Extended Data Fig. 7a). We also showed that Brr2a could reshape plant pri-miRNAs, altering specific structural features in vitro and in vivo (Fig. 5 and Extended Data Fig. 9). Importantly, the ATPase or helicase activity-compromised variants of Brr2a could not

miRNA miRNA* a b d f 0.15 0.10 0.05 0.02 0.01 0 -1 7 -1 5 -1 4 -1 3 -1 2 -1 1 -1 0 -9 -8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8 9 1 0 1 1 1 2 1 3 1 4 1 5 1 6 1 7 1 8 1 9 2 0 Col-0 brr2a-3 DMS activity 17 nt 5′ 3′ Base to loop 0.4 0.3 0.2 0.02 0.01 0 0.03 DMS activity Base to loop 5′ 3′ 15 nt c e Col-0 brr2a-3 DMS activity (mismatch ratio) 0.0030 0.0025 0.0010 0 pri-miR158a (BTL) 0.0002 miRNA miRNA* 52 reads 404 reads 185 reads 132 reads 59,295 reads 1,267 reads 35,094 reads DMS activity (mismatch ratio) 0.100 0.010 0.005 0 0.050 111 reads Col-0 brr2a-3 g Col-0 brr2a-3 miR159a miR159a* (0-1,462) RPM (0-1,980) RPM pri-miR159a (sequential LTB) Col-0 (#1) Col-0 (#2) 17 nt 16 nt Col-0 brr2a-3 DMS activity (mismatch ratio) 0.12 0.05 0.02 0 0.08 pri-miR172a (BTL) miRNA miRNA* 16 nt 16 nt D yn am ic tr an si tio n brr2a-3 miRNA miRNA* DMS activity (mismatch ratio) 0.010 0.002 0.001 0 0.003 h pri-miR165a (bidirection) 1 5 1 4 1 3 1 2 1 1 1 0 9 8 7 6 5 4 3 2 1 0 -1 -2 -3 -4 -5 -6 -7 -8 -9 -1 0 -1 1 -1 2 -1 3 -1 4 -1 5 -1 6 -1 7 -1 8 -1 9 -2 0 Col-0 brr2a-3 Gini index 1.00 0.75 0.25 0 1.25 0.50 ** b r r 2 a -3 C o l -0

rescue the miRNA defect in the complementation lines, despite retaining the ability to interact with the microprocessor (Fig. 4d). This underscores the indispensable role of Brr2a's ability to reshape pri-miRNA for efficient production of miRNAs. Thus, Brr2a exhibits a distinct mode of action from other documented RHs in promoting miRNA biogenesis. Importantly, Brr2a's impact on miRNA production is not only limited to intron-containing pri-miRNAs but also to intron-lacking substrates. Nevertheless, Brr2a also mechanistically differs from other reported spliceosome components that regulate miRNA accumulation through their cognate activities in splicing events 6,[9][10][11]43 .

The remodelling action of pri-miRNAs by Brr2 seemingly resembles the latter's classical unwinding mechanism of the U4/U6 duplex. The N-helicase cassette of Arabidopsis Brr2a, which directly interacts with pri-miRNAs, exhibits both ATPase and helicase activities to reshape RSS 30,34,44 , reminiscent of yeast and human counterparts. Meanwhile, the C-helicase cassette of Brr2a acts as an intramolecular modulator, enhancing the helicase activity of the full-length protein 14 . Notably, there are at least 147 different pri-miRNAs in plants, varying in size from ~100 to several hundred nucleotides, showing heterogeneous shapes and structures 40 . A common feature among all pri-miRNAs is that they carry a stem-loop structure and two flanking single-stranded basal regions. This configuration suggests that the 3′ extended single-stranded RNA tail might serve as a docking site for Brr2a, aligning with its canonical model for initial loading onto the ssRNA region of a U4/U6 RNA duplex in the splicing process. In this scenario, Brr2 could reshape the RSS by translocating along pri-miRNAs in a 3′ to 5′ direction in vivo 14 . For some pri-miRNAs, such as pri-miR396 and pri-miR170 (ref. 40), which have larger internal bulges and loops, respectively, Brr2a might load on these unpaired regions and then translocate along pri-miRNAs. Interestingly, we noticed that Brr2a, different from its human homologue, exhibits a significantly higher affinity to dsRNA vs ssRNA 45 . These results prompted us to hypothesize that Brr2a could directly interact with dsRNA regions of pri-miRNAs and then remodel their structure. However, this hypothesis awaits future testing, ideally via cryo-EM structural analysis of the Brr2a-pri-miRNA complex.

The expression of miRNAs necessitates frequent changes in response to environmental stress and stimuli 3 . Importantly, a growing body of research underscores the pivotal role of RSS in regulating messenger (m)RNA translation, stability, metabolism and other critical processes 46 . Pri-miRNAs exhibit sophisticated structures that not only instruct the microprocessor for processing but also contain additional layers of information to be decoded by other RNPs in various physiological contexts. Brr2-mediated remodelling of pri-miRNAs might represent such a novel avenue to reconfigure the pri-miRNA transcriptome for better adaptation to physiological changes. Notably, RSS can be influenced by temperature 46 , whereas Brr2, also known as Bad Response to Refrigeration, regulates plants' sensitivity to varying temperatures, as evidenced by mutants displaying a heightened sensitivity to both cold and heat treatments [47][48][49][50] . A retrospective thought of these findings is that Brr2 might serve as a rapid regulatory switch to finely tune the accumulation of miRNAs, in addition to its canonical role in splicing, in response to both physiological and environmental changes. Finally, the Brr2a protein is highly conserved in eukaryotes, and it would be tempting to learn whether its homologues have a similar function to the Arabidopsis counterpart.

plants, employing sRNA blot analysis and qRT-PCR assays, respectively. To generate brr2a-2 complementary lines, pBA-pBrr2a::FM-Brr2a, pBA-pBrr2a::FM-Brr2a-E640Q and pBA-pBrr2a::FM-Brr2a-S676A T678A were separately introduced into brr2a-2 backgrounds using the floral-dip transformation method 51 . T 2 transgenic lines containing FM-tagged Brr2a and its variants were identified through western blot analysis. In addition, brr2a-2 pBA-pBrr2a::FM-Brr2a was crossed into se-2 and hyl1-2 to obtain se-2 pBA-pBrr2a::FM-Brr2a and hyl1-2 pBA-pBrr2a::FM-Brr2a. F 3 plants were used for further analysis.

Most of the cloned coding sequences (CDSs) and genomic DNA sequences were initially introduced into the pENTR vector (Invitrogen, A10462) using the primers listed in Supplementary Table 7. Following confirmation through sequencing, they were subsequently cloned into the destination vectors through attL-attR (LR) recombination reactions.

The construction of pBA-pBrr2a::FM-gBrr2a, pBA-pBrr2a:: FM-Brr2a, pBA-pBrr2a::FM-Brr2a-EQ and pBA-pBrr2a::FM-Brr2a-AA was carried out as follows: a native promoter containing built-in Xba I and Asc I restriction enzyme cleavage sites, and a Brr2a genomic fragment, were applied using KOD polymerase with Col-0 genomic DNA as a template, along with the primers listed in Supplementary Table 7. The PCR product of the native promoter was cloned into Xba I/Asc I-digested pBA-FM-DC vector to generate pBA-pBrr2a::FM-DC through T4 ligation. The Brr2a genomic fragment was cloned into the pENTR vector, and then an LR reaction was performed with the destination vector pBA-pBrr2a::FM-DC to generate pBA-pBrr2a::FM-gBrr2a. For the mutated variants of Brr2a (EQ and AA), Brr2a cDNA from reverse transcription was cloned into the pENTR vector. Subsequently, the catalytically inactive forms were obtained through KOD amplification using the primers listed in Supplementary Table 7. The resulting products, after Dpn I digestion, were transferred into DH5a cells, resulting in pENTR-Brr2a-EQ and pENTR-Brr2a-AA, followed by sequence validation. Subsequently, pBA-pBrr2a::FM-Brr2a, pBA-pBrr2a::FM-Brr2a-EQ and pBA-pBrr2a::FM-Brr2a-AA were constructed through LR reactions.

pET28a-His-SUMO-N-Brr2a, pET28a-His-SUMO-N-EQ, pET28a-His-SUMO-N-AA and pET28a-His-SUMO-C-Brr2a were constructed as follows: the N-Brr2a (487 to 1,288 aa) and C-Brr2a (1,289 to 2,171 aa) fragments, each containing built-in BamH I and Sal I sites, were amplified using pENTR-Brr2a as the template and the primers listed in Supplementary Table 7. Similarly, the N-Brr2a-EQ and N-Brr2a-AA fragments were also amplified using pENTR-Brr2a-EQ and pENTR-Brr2a-AA as the template, respectively, and the primers listed in Supplementary Table 7. The PCR fragments of N-Brr2a, N-Brr2a-EQ, N-Brr2a-AA and C-Brr2a were cloned into BamH I/Sal I-digested pET28a-His-SUMO to generate final vectors, including pET28a-His-SUMO-N Brr2a, pET28a-His-SUMO-N-Brr2a-EQ, pET28a-His-SUMO-N-Brr2a-AA and pET28a-His-SUMO-C-Brr2a.

Co-immunoprecipitation was conducted in 3-week-old Arabidopsis plants. Total proteins were extracted from 0.2 g of finely ground powder using 1 ml of IP buffer (comprising 40 mM Tris-HCl pH 7.5, 150 mM NaCl, 5 mM MgCl 2 , 5 mM dithiothreitol (DTT), 0.5% Triton X-100, 2% glycerol, 1 mM phenylmethyl sulfonyl fluoride (PMSF), 50 μM MG-132 and 1 pellet per 10 ml of Complete EDTA-free protease inhibitor from Roche). Subsequently, protein extracts were subjected to immunoprecipitation using anti-Flag M2 magnetic beads (Sigma-Aldrich, M8823) at 4 °C for 2 h. For RNase treatment, 50 μl of RNase A (1 mg ml -1 ) (Thermo Fisher, EN0531) was introduced into 1 ml of the immunoprecipitation buffer during the incubation period. After the incubation, the beads were washed four times with IP buffer at 4 °C for 5 min each, followed by the application of SDS loading buffer at 95 °C for 10 min.

RIP and ChIP assays were conducted following established procedures 32 . The brr2a-2 pBrr2a::FM-Brr2a was crossed with se-2 and hyl1-2 to switch genetic backgrounds. The Brr2a variant transgenic plants were generated by complementing brr2a-2 with pBrr2a::FM-Brr2a-variants. Three-week-old seedlings were crosslinked using the buffer (20 mM HEPES pH 7.4, 1 mM PMSF, 1 mM EDTA, 0.4 M sucrose, 1% formaldehyde) and ground to fine powder. Plant nuclei were isolated from 2 g or 3 g plant powder and resuspended with nuclei sonication buffer (40 mM Tris-HCl pH 7.4, 150 mM NaCl, 5 mM MgCl 2 , 1% SDS, 1 mM DTT, 1% Triton X-100, 2% glycerol, 1 mM PMSF, 50 μM MG-132, and 1 pellet per 10 ml complete EDTA-free protease inhibitor and 10 U SUPERase-In RNase inhibitor (Thermo Fisher)) for sonication (15 cycles, 30 s sonication and 90 s pause). The supernatant was diluted 9 times with dilution buffer (40 mM Tris-HCl pH 7.4, 150 mM NaCl, 5 mM MgCl 2 , 1 mM DTT, 2% glycerol, 1 mM PMSF, 50 μM MG-132, 1 pellet per 10 ml complete EDTA-free protease inhibitor and 10 U ml -1 SUPERase-In RNase inhibitor). Either Flag antibodies or endogenous antibodies of NRPB2 (PhytoAB, PHY2429S) were used for RIP or ChIP, respectively. The immunoprecipitations were performed at 4 °C with agitation for 2 h (RIP) or overnight (ChIP). IP samples were washed with dilution buffer, high salt wash buffer (20 mM Tris-HCl pH 7.4, 500 mM NaCl, 5 mM MgCl 2 , 50 μM ZnCl 2 , 1 mM DTT, 1 mM PMSF, 0.5% Triton X-100, 50 μM MG-132, 1 pellet per 10 ml complete EDTA-free protease inhibitor and 10 U ml -1 SUPERase-In RNase Inhibitor), LiCl wash buffer (0.25 M LiCl, 1% NP-40, 1% sodium deoxycholate, 20 mM Tris-HCl pH 7.4, 1 mM EDTA) and Proteinase K buffer (20 mM Tris-HCl pH 7.4, 200 mM NaCl and 10 U SUPERase-In RNase inhibitor). RNA or DNA was eluted from de-crosslinking with Proteinase K solution (20 mM Tris-HCl pH 7.4, 200 mM NaCl, 1 mg ml -1 Proteinase K, 1% SDS and 10 U SUPERase-In RNase inhibitor, 10 mM EDTA) on a Thermomixer shaker at 65 °C for 2 h or 6 h, respectively. Following this, RNA underwent purification using RNA Clean & Concentrator kits (ZYMO, R1017) and was treated with DNase before RT-qPCR analysis. DNA underwent RNase treatment and subsequent purification using DNA Clean & Concentrator kits (ZYMO, D4004). For Pol II ChIP in brr2a mutants vs Col-0 and Brr2a RIP in hyl1-2 pBrr2a::FM-Brr2a vs se-2 pBrr2a::FM-Brr2a, quantification was performed by first normalizing the amount from IP to the input and then to that of the Col-0 and se-2 pBrr2a::FM-Brr2a which were arbitrarily set as 1 for both ChIP and the RIP assays. For the RIP of Brr2a and its variants, quantification was performed by normalizing the amount of Brr2a variant IP-derived RNA to that of the immunoprecipitated proteins and then to that of wild-type Brr2a where the ratio was arbitrarily set as 1.

Tobacco plants were used for BiFC assays 26 . Briefly, p35S::nYFP-Brr2a was co-infiltrated with p35S::cYFP-HYL1, p35S::cYFP-SE and p35S::cYFP-DCL1 in the leaves of 4-week-old tobacco plants. The fluorescence signals were detected 3 days later. YFP fluorescence signals were excited at 514 nm. The emissions for YFP and chlorophyll fluorescence are 525-550 nm and 661-700 nm, respectively. All images were acquired using a Leica stellaris 5 laser-scanning confocal microscope with LAS X Life Science microscope software.

Total RNA for RNA-seq and sRNA-seq was extracted from 3-week-old soil-grown plants of Col-0 and the brr2a mutants using TRI reagent (Sigma). Library preparation and analysis for RNA-seq and sRNA-seq using Illumina sequencing followed established protocols 26 . DESeq2 analysis was used for normalization in RNA-seq. Reads per million (RPM) of miRNAs were counted and normalized on the basis of the amount of sRNAs that were derived from rRNA.

Genomic DNA was removed by DNase before reverse transcription. Random hexamers and oligo(dT) 12-18 were used to detect mRNA and Article https://doi.org/10.1038/s41477-024-01788-8 pri-miRNAs. Primers and procedures of stem-loop qPCR detecting miRNA expression were adopted from a previous study 26 . Relative expression level was calculated with the 2 -ΔΔCt method.

The sRNA blot assays were conducted on 10-day-old seedlings using the primers listed in Supplementary Table 7, following an established method adopted from a previous study 39 . Total RNA was extracted using TRI reagent (Sigma-aldrich, T9424) and Turbo DNase treatment (2 U Turbo DNase per 50 μg of RNA) (Invitrogen, AM2239). Urea-polyacrylamide gels (15%) were used to separated RNA samples. Following electrophoresis, RNA was transferred onto a nylon membrane (GE Healthcare) using a semi-dry transfer apparatus. The sRNA and U6 probes were labelled using [γ-32 P] ATP with T4 PNK and 21-nucleotide DNA oligos that are complementary to the corresponding sequences. Signals were detected with a Typhoon FLA7000 imaging system (GE Healthcare). The stripped membrane was then rehybridized with a different sRNA and U6 probes following the same hybridization and washing procedures.

The blots were probed with specific antibodies as follows: actin (Sigma-Aldrich, A0480), NRPB2 (PhytoAB, PHY2429S), HSP70 (Agrisera, AS08371), MYC (Sigma, C3956), HYL1 (homemade), DCL1 (Agrisera, AS122102), AGO1 (Agrisera, AS09527) and SE (homemade). Secondary antibodies used in this study were anti-rabbit (GE Healthcare, NA934) and anti-mouse IgG (GE Healthcare, NA931).

Staining assays were carried out using the F 3 generation of pMIR159a::FM-GUS and pMIR159b::FM-GUS in either Col-0 or brr2a-3 background. For each sample, tissues were collected from at least 10 individual plants. The plant tissues were incubated in GUS staining solution (120 mM Na 2 HPO 4 , 78 mM NaH 2 PO 4 , 2 mM potassium ferricyanide [K 3 Fe(CN) 6 ], 2 mM potassium ferrocyanide [K 4 Fe(CN) 6 ], 0.1% Triton X-100, 10 mM EDTA, 10% methanol and 2 mM X-Gluc (Thermo Fisher, R0851)) at 37 °C overnight 26 . After staining, the tissues were cleared with 70% ethanol to remove chlorophyll at 37 °C overnight with a horizontal shaker. GUS expression patterns were documented using a stereo microscope.

To purify recombinant truncated Brr2a and its variants from E. coli, BL21 (DE3) cells carrying plasmid were grown in Luria Broth (LB) medium at 37 °C until optical density (OD) 600 = 0.6. Recombinant protein expression was induced with 0.5 mM isopropyl-β-d-thiogalactopyranoside at 16 °C overnight. Then, bacterial cells were collected and resuspended in a lysis buffer that contained the following components: 40 mM Tris-HCl pH 8.0, 800 mM NaCl, 2% glycerol, 5 mM β-mercaptoethanol, 3 mM PMSF, 0.5% Triton X-100, 50 μM MG-132 and 1 pellet of complete EDTA-free protease inhibitor (Roche) per 50 ml of buffer. The resuspended cells were then disrupted using a high-pressure homogenizer (Microfluidics). Following cell disruption, the lysate was clarified through a combination of centrifugation (21,000 × g at 4 °C for 30 min) and filtration via a 0.4 μm filter. The cleared lysate was supplemented with imidazole to a final concentration of 20 mM. Finally, the prepared lysate was loaded onto a HisTrap HP column (GE Healthcare, 17524802).

The column was initially washed with 25 ml of a wash buffer composed of 40 mM Tris-HCl at pH 8.0, 800 mM NaCl, 2% glycerol and 20 mM imidazole. Subsequently, elution was carried out using a gradient elution buffer ranging from 20 to 300 mM imidazole. Peak fractions containing the recombinant truncated Brr2a or its variants were combined and concentrated using a 50 kDa molecular weight cut-off (MWCO) centricon (Millipore). The concentrated sample was then loaded onto a HiLoad 16/600 Superdex 200 pg column (GE Healthcare).

The gel filtration buffer for the Superdex column consisted of 20 mM Tris-HCl pH 8.0 and 500 mM NaCl. Peak fractions containing Brr2a or its variants were collected and subsequently dialysed with 1 l of dialysis buffer (10 mM Tris-HCl at pH 8.0, 150 mM NaCl, 30% glycerol and 1 mM β-mercaptoethanol) at 4 °C for 6 h. The final purified protein was quantified using SDS-PAGE and then divided into aliquots for storage at -80 °C.

EMSA was conducted as previously described 5 . In summary, recombinant Brr2a or its variants were mixed with labelled pri-miRNA or ssRNA in the EMSA buffer, consisting of 20 mM Tris-HCl (pH 8.0), 150 mM NaCl and 10% glycerol. The mixtures were incubated at room temperature for 30 min and then separated on a native 1% agarose gel. The images were quantified with ImageJ. Prism 9 (GraphPad) was used to calculate K d and appK d with a Hill slope model.

The procedure for the ATPase assay was adapted from a previously described method 5 . Fresh purified recombinant Brr2a or its variants at the respective concentrations were introduced into ATP hydrolysis reactions (10 μl) comprising 20 mM Tris-HCl pH 8.0, 150 mM NaCl, 5 mM MgCl 2 , 12% glycerol, 2 mM DTT and 750 ng of folded pri-miR159b. The above mixture was pre-incubated for 10 min before the addition of 1 mM cold ATP and 0.1 μl [γ-32 P] ATP (PerkinElmer, 6,000 Ci mmol -1 ). Reactions were incubated at 37 °C for the specified time intervals and then terminated by loading the samples onto a polyethyleneimine-cellulose (PEI-cellulose) TLC plate. The plate was developed in 0.5 M LiCl and 1 M formic acid. Liberated phosphate was analysed through TLC and phosphor imaging.

The unwinding assay was conducted with slight modifications to a previously published method 52 . Three dsRNA duplexes were made: (1) pri-miR159b duplex with a ~20-nt over-hung single-stranded RNA and a stem-loop on its 3′ end, (2) nicked pri-miR166f duplex with one nucleotide gap and (3) a model dsRNA duplex with a ~110 nt 3′ overhang. The duplexed dsRNAs were generated by co-folding long cold RNA and short hot RNA in vitro. The final purified RNAs were visualized through native PAGE gels to observe released short RNA and dsRNA that remained annealed. Freshly purified truncated Brr2a and its variants from E. coli, and freshly immunoprecipitated full-length Brr2a and its variant proteins from plants were used. Plant Brr2a and its variants were immunoprecipitated from transgenic plants of brr2a-2 pBA-pBrr2a::FM-Brr2a, brr2a-2 pBA-pBrr2a::FM-Brr2a-EQ and brr2a-2 pBA-pBrr2a::FM-Brr2a-AA, respectively, using M2 beads. Subsequently, the resulting beads were washed with the IP buffer and high salt buffer (comprising 40 mM Tris-HCl pH 7.5, 500 mM NaCl, 0.5% NP-40, 1 mM PMSF, 50 μM MG-132, 0.2 U μl -1 SUPERase-In RNase inhibitor and 1 pellet per 10 ml of complete EDTA-free protease inhibitor from Roche) for three times each. Each washing step included 5 min agitation at 4 °C for 5 min. Proteins were eluted using Flag peptide (Sigma-Aldrich, F4799). The eluate was detected by western blot before unwinding assays.

To prepare the pri-miRNA substrates, the indicated 5′-32 P-labelled short RNA fragment was hybridized with a truncated strand of pri-miR166f, pri-miR159b and LUC reverse. Annealed RNA duplexes were purified by native polyacrylamide gel electrophoresis (native PAGE). Subsequently, 200 nM recombinant proteins or freshly immunoprecipitated Brr2a or its variants were mixed with the RNA duplexes (5 fmol, signalling 500-2,000 c.p.m. per reaction) in a remodelling buffer containing 40 mM Tris-HCl pH 8.0, 75 mM NaCl, 1.5 mM DTT, 100 ng μl -1 acetylated BSA and 8% glycerol at room temperature for 10 min allowing RNP formation. Reactions were started by adding 5 mM MgCl 2 and 10 mM ATP per 1 mM GTP. The reaction mixtures were incubated at 25 °C for the specified time course and halted by adding 800 mM NaCl (to elute RNA from proteins) and 0.1 M of cold short RNA Extended Data Fig. 7 | Brr2a can exert its role as an RH using pri-miRNA as a substrate. a, Two independent replicates of TLC assays showed the distinct ATPase activity of truncated Brr2a and its variants in vitro. b, Two independent replicates of unwinding assays showed the helicase activity of truncated Brr2a and its variants in vitro. c, Western blot analysis validated the comparable IP product of Brr2a and its variants used for unwinding assay. The IPs were performed using M2 beads with brr2a-2 complementary lines, with brr2a-2 serving as a control. Function compromised variants were slightly more concentrated during elution. Eluted proteins were detected by an anti-MYC antibody, with HSP70 serving as a control. d, The Native-PAGE results from an additional independent replicate of unwinding assays showed the distinct RH activity of truncated Brr2a and its variants in vitro. e, Western blot analysis validated the successful immunoprecipitation of Brr2a and its variants from indicated brr2a-2 complementary lines in RIP assays. The IPs were performed using M2 beads with brr2a-2 complementary lines, with brr2a-2 serving as a control. Proteins were detected by an anti-MYC antibody, with HSP70 serving as a control. At least three independent experiments were performed, and representative images are shown (a-d). For a-e, N/C-Brr2a represented N/C terminal truncated Brr2a; N-EQ or N-AA represented the variants of N-Brr2a carrying mutations E640Q or S676A and T678A; Brr2a-EQ and Brr2a-AA represent the variants of full-length Brr2a carrying mutations E640Q and S676A T678A, respectively.