Isolation of a mutant with defects in miRNA biogenesis
We performed an ethylmethane sulfonate mutagenesis screen for Arabidopsis mutants in miRNA biogenesis, utilizing the vein-centered leaf beaching phenotype caused by the phloem-specific expression of an artificial miRNA (amiR-SUL) targeting SULFUR (SUL) as a visible marker for miRNA activity28. A series of mutants with reduced leaf bleaching was isolated and found to be in genes with known roles in miRNA biogenesis/activity, such as hyl1-11, dcl1-30, hen1-11 and ago1-25, suggesting that the genetic screen was effective (Supplementary Fig. 1a and b). The ago1-25 allele isolated in our study was thus named because it harbored the same mutation as the one previously reported 29. In addition, we isolated a new suppressor mutant with reduced leaf bleaching (Fig. 1a and b); the mutation was designated rbv-1 as it was later shown to be in a previously uncharacterized gene that we named REDUCTION IN BLEACHED VEIN AREA (RBV). This mutant exhibited pleiotropic developmental phenotypes, such as reduced root length, smaller plant size, narrow leaves, short stature, increased branching and reduced fertility (Fig. 1c and Supplementary Fig. 2). Northern blot analyses showed that both amiR-SUL and endogenous miRNAs (miR156, miR159, miR164, miR165, miR167, miR319 and miR390) were moderately reduced in abundance in 14-day-old amiR-SUL rbv-1 seedlings as compared to amiR-SUL seedings (Fig. 1d). We also preformed small RNA sequencing with 14-day-old amiR-SUL and amiRSUL rbv-1 seedlings. Clustering analysis showed that the three biological replicates for each genotype were highly reproducible (Supplementary Fig. 3a). Reads corresponding to miRNAs were normalized against total mapped reads and miRNAs with RPM (reads per million mapped reads)>10 in either genotype (average of three replicates) were included in our analyses (Supplementary Data 1). An overall reduction in miRNA abundance in the mutant relative to the amiR-SUL line was observed (Fig. 1e and Supplementary Fig. 3b), while there were some miRNAs such as miR845a and miR845b showing increased accumulation in the amiRSUL rbv-1 mutant. We also performed RT-qPCR to examine the expression of some miRNAs target genes. The transcript levels of SPL3 and SPL10 (targets of miR156), PHB, REV and PHV (targets of miR165/6), MYB33 and MYB65 (targets of miR159), ARF8 (a target of miR167), and CUC2 (a target of miR164) were de-repressed in the amiR-SUL rbv-1 mutant (Fig. 1f).
RBV encodes an evolutionarily conserved WD40 domain protein
The amiR-SUL rbv-1 mutant was backcrossed with the parental amiR-SUL line. In a total of 614 F2 plants, 141 (23%) exhibited the mutant phenotypes, which is consistent with the phenotype being caused by a single, nuclear, and recessive mutation (c2 =1.257; P =0.244; (Supplementary Table 1). In order to identify the causal mutation in amiR-SUL rbv-1, pooled DNA from mutant plants in the F2 of the amiR-SUL rbv-1 x amiR-SUL cross was used for whole-genome re-sequencing. The results revealed that the amiR-SUL rbv-1 phenotype was linked to a single nucleotide change (G-to-A) in the first exon of AT5G64730, causing the change of the encoded amino acid from glycine to glutamic acid (see Methods; Fig. 2a).
In order to confirm that RBV is indeed AT5G64730, a construct of RBV (AT5G64730)-eYFP driven by its native promoter was generated and introduced into amiR-SUL rbv-1. Phenotypes of the transgenic plants showed that pRBV:RBV-eYFP fully rescued the morphological defects of amiR-SUL rbv-1 (Fig. 2c). Furthermore, the transgene restored amiR-SUL accumulation as well as leaf bleaching in amiR-SUL rbv-1 (Fig. 2d). While the expression of the amiR-SUL target gene, SUL, was increased in the mutant at the RNA and protein levels, the transgene also restored SUL RNA (Fig. 2e) and SUL protein (Fig. 2f) in amiR-SUL rbv-1 to wild-type levels. Moreover, the transgene rescued the defects in miR159 and miR319 accumulation in the amiR-SUL rbv-1 mutant (Fig. 2d). Therefore, the miRNA biogenesis and morphological defects of amiR-SUL rbv-1 were attributable to the mutation in RBV (AT5G64730).
RBV encodes a previously uncharacterized protein with seven WD40 repeats (Fig. 2a). In plants, WD40 repeat proteins are numerous, interact with diverse proteins, and act in a variety of biological processes, such as plant development and immunity30, 31. To investigate whether RBV is an evolutionarily conserved protein, a phylogenetic tree including homologs of RBV in various plants was generated (Supplementary Data 2); Supplementary Fig. 4a). According to the phylogenetic analysis, RBV orthologs can be found in plants ranging from single-cell green algae to core eudicots and grasses, and there is no close paralog of RBV in the latest common ancestor of land plants. In most eudicots that have undergone gamma whole genome duplication (WGD) and Brassicaceae that has also undergone beta and alpha WGDs, RBV remained as a single copy. Only in species that have undergone recent, specific WGDs, such as apple, soybean, and maize, there are multiple copies of RBV (Fig. 2b and Supplementary Fig. 4). This result indicated that RBV was functionally conserved and extra copies might be deleterious.
We also obtained two lines with T-DNA insertions in or near RBV (Salk_126634 and Salk_075672) (Supplementary Fig. 5a and 5c). Interestingly, the phenotypes of the T-DNA mutants were the same as Col (Supplementary Fig. 5b). Next, we examined RBV transcripts in the two T-DNA insertion lines. RT-PCR showed that transcripts corresponding to the full-length coding region of RBV were present in the two mutants (Supplementary Fig. 5d). In addition, real-time RT-PCR showed that the levels of RBV transcripts were only mildly reduced in the two mutants (Supplementary Fig. 5e), consistent with the lack of morphological phenotypes.
In order to remove the amiR-SUL transgene background, we crossed the amiR-SUL rbv-1 mutant with wild-type (Col) plants. In the F2 population, we identified homozygous rbv-1 plants without the amiR-SUL transgene through genotyping; these plants showed the same pleiotropic phenotypes as amiR-SUL rbv-1 (Fig. 3a). The morphological phenotypes of rbv-1 were completely rescued by the pRBV:RBV-eYFP transgene (Fig. 3a). We further confirmed that the rbv-1 mutant without the amiR-SUL transgene had defects in miRNA accumulation. RNA gel blots showed that the levels of several endogenous miRNAs, such as miR156, miR159, miR164 and miR167, were mildly reduced in rbv-1 as compared to wild-type Col plants. Other miRNAs, such as miR319 and miR398, were unaffected (Supplementary Fig. 6a). To rule out the possibility that RBV indirectly regulated miRNA accumulation through affecting the expression of the miRNA biogenesis machinery, we examined the expression of the known genes involved in miRNA biogenesis, and no significant changes were observed in rbv-1 (Supplementary Fig. 6b and 6c).
RBV promotes the transcription of MIR genes
We next investigated how RBV promotes miRNA biogenesis. MIR gene transcription to produce pri-miRNAs is the first step in miRNA biogenesis. RT-qPCR showed that the levels of pri-miRNAs were reduced by ~30-50% in the rbv-1 mutant (Fig. 3b). For the pri-miRNAs, 298 pri-miRNAs annotated in Araport11 were examined by RNA-seq. In total, only 41 pri-miRNAs gave reads in any sample (Supplementary Fig. 7). In the RNA-seq data, we only detected three (pri-miR159b, pri-miR166a and pri-miR319b) of the seven pri-miRNAs that we examined by qRT-PCR in Fig. 3b. All three pri-miRNAs showed reduced abundance in the rbv-1 mutant, which is consistent with results from qRT-PCR. The reduction in pri-miRNA abundance could be attributed to impaired MIR genes transcription, reduced stability of pri-miRNAs or enhanced pri-miRNA processing. To determine whether transcription was affected in the rbv-1 mutant, we crossed rbv-1 with a GUS reporter line (pMIR167a:GUS) under the control of the MIR167a promoter and obtained pMIR167a:GUS rbv-1 (with both the transgene and the mutation being homozygous). GUS activity was visibly lower in pMIR167a:GUS rbv-1 than in pMIR167a:GUS as revealed by GUS staining (Fig. 3c). RT-qPCR analysis confirmed that the rbv-1 mutant had lower GUS transcript levels (Fig. 3d). Thus, reduced transcription of MIR genes could be one of the problems in miRNA biogenesis in rbv-1.
To further confirm a positive role of RBV in MIR gene transcription, the occupancy of Pol II at MIR loci was determined by chromatin immunoprecipitation (ChIP) with an antibody against the Pol II C-terminal repeats. ChIP without an antibody served as the negative control. MIR166a, MIR167a and MIR171a promoter regions, as well as C1 (a region between the genes AT2G17460 and AT2G17470 known to not engage Pol II7, were examined by RT-qPCR after ChIP. The MIR166a and MIR167a promoter regions were enriched in the immunoprecipitates in both rbv-1 and Col relative to C1 (Fig. 3e). Pol II occupancy at MIR166a and MIR167a promoter regions was reduced in the rbv-1 mutant relative to Col, while the signal on MIR171a was too low to be calculated (Fig. 3e). Thus, RBV facilitates the recruitment of Pol II to MIR loci.
RBV is localized in the nucleoplasm and required for the proper localization of HYL1 in D-bodies
We studied the expression of RBV in various tissues and the subcellular localization of the protein. RT-PCR analyses of RNAs from seedlings, roots, cauline leaves, rosette leaves, stems and inflorescences showed ubiquitous RBV expression (Supplementary Fig.8). RBV:RBV-eYFP transgenic plants exhibited YFP fluorescence in the nucleoplasm but not the nucleolus. SE is a component of the microprocessor in pri-miRNA processing. An mRuby3-tagged SE protein driven by the SE promoter9 co-localized with RBV-eYFP in the nucleoplasm (Fig. 4a).
HYL1 and DCL1 form nuclear foci known as Dicing bodies (D-bodies), which are sites of pri-miRNA processing21, 32. To determine the status of D-bodies in the rbv-1 mutant, we crossed a HYL1-YFP transgene21 into the mutant and obtained plants homozygous for both the transgene and the rbv-1 mutation. D-body numbers were determined in 420 and 537 root nuclei of wild-type and rbv-1 plants, respectively. The number of HYL1-YFP D-bodies was significantly decreased in rbv-1 (Fig. 4b and 4c), suggesting that mutation of RBV leads to a defect in D-body formation or the localization of HYL1 to D-bodies.
Mutation of RBV leads to a defect of miRNA loading into AGO1
miRNAs are loaded into AGO1 to form miRISCs that execute the silencing of target RNAs23, 25, 33. To determine whether RBV affects the formation of RISCs, we performed AGO1 IP followed by sRNA-seq with both input and IP samples from wild-type and rbv-1 seedlings. Three replicates were performed and showed high reproducibility (Supplementary Fig. 9). From input samples, many miRNAs showed small but statistically significant reduction in abundance in the rbv-1 mutant (Supplementary Data 3). A few miRNAs, such as miR845a, miR845b and miR843, were elevated in abundance in the mutant (Fig. 5a; Supplementary Data 3). The loading status of each miRNA was expressed as the ratio of miRNA abundance between AGO1 IP and input and differences between wild type and rbv-1 were evaluated by student’s t-test. A global reduction in the AGO1 loading of miRNAs was observed in the rbv-1 mutant (Fig. 5b; Supplementary Data 4). Notably, miR845a and miR845b, which exhibited elevated levels, were less associated with AGO1 in the mutant, indicating that RBV is crucial for miRISC formation. RNA gel blots were also performed to validate the sRNA-seq results. miR159 and miR166 levels were lower in input and further reduced in AGO1 IP in the mutant (Fig. 5c). miR845a levels were strongly increased in the rbv-1 input sample as compared with Col input but greatly reduced in the rbv-1 AGO1 IP sample as compared with Col AGO1 IP (Fig. 5c; Supplementary Data 4). Thus, the rbv-1 mutant exhibits a global miRNA loading defect.
To determine how RBV might promote the loading of miRNAs into AGO1, we first examined whether RBV interacts with AGO1. Co-IP was performed with pRBV:RBV-eYFP plants using anti-GFP and anti-AGO1 antibodies, but no interaction between RBV and AGO1 was detected. We next sought to determine whether the bi-molecular feature of miRISCs is affected in the mutant. Protein extracts from pRBV:RBV-eYFP rbv-1 and rbv-1 plants were subjected to gel filtration followed by western blotting to detect AGO1 and northern blotting to detect miR159. Among the 44 fractions, AGO1 was distributed in both high molecular weight (HMW) (fractions 9-11) and low molecular weight (LMW) (fractions 16-17) complexes in pRBV:RBV-eYFP rbv-1, consistent with previous findings27. Notably, the LMW complexes, which corresponded to monomeric AGO1 in size, also showed peak levels of miR159, suggesting that they represent miRISCs with one AGO1 protein and one miRNA. However, such miRISCs were greatly reduced in rbv-1, with AGO1 being in HMW complexes instead (Fig. 5d). The distribution of miR159 also shifted towards HMW complexes in rbv-1 (Fig. 5d). Western blot analyses with anti-GFP antibody showed that RBV-eYFP was present in fractions 8-16 with estimated molecular weights much higher than that of an RBV-eYFP monomer, suggesting that RBV itself also resides in protein complexes (Supplementary Fig. 10). Therefore, RBV promotes the formation of miRISCs that contain only AGO1 and miRNAs. The HMW AGO1 complexes may represent intermediates in RISC formation.
AGO1 is known to associate with trans-acting small interfering RNAs (ta-siRNAs) in addition to miRNAs34, 35. Contrary to miRNAs, which are loaded into AGO1 in the nucleus, ta-siRNAs are loaded into AGO1 in the cytoplasm24. We investigated whether RBV affects the ta-siRNA-AGO1 association. We quantified 21-nt siRNAs that mapped to 100-bp windows that overlapped with TAS1A, TAS1B, TAS1C and TAS2 loci. The ta-siRNAs were not significantly altered in rbv-1 input as compared to Col input in each 100-bp window (Supplementary Fig. 11 a, c, e and g). We then analyzed their levels in AGO-IP vs. input, and found that the loading of ta-siRNAs into AGO1 was largely unaffected. Only ta-siRNAs in window 5 of TAS1A, window 3 of TAS2B and window 3 of TAS1C were affected (Supplementary Fig.11 b, d, f and h). The levels of miR173, the trigger of ta-siRNA biogenesis from TAS1 and TAS2 loci, were not significantly different between Col and the rbv-1 mutant (Supplementary Fig. 11 i). The fact that ta-siRNA loading was largely unaffected in the mutant is consistent with RBV being a nuclear protein.
RBV is required for the splicing of short introns in certain pre-mRNAs
In order to determine whether RBV affects the expression of protein-coding genes, we performed RNA-seq with 14-day-old seedlings of Col and rbv-1 in triplicates. The three biological replicates for each genotype were highly reproducible (Supplementary Fig. 12).
Differentially expressed genes (DEGs) were identified between Col and mutant samples with FPKM >1, fold-change >2 and FDR<0.05 as the cutoff. In total, we identified 632 up-regulated (hyper-DEGs) and 363 down-regulated (hypo-DEGs) genes, respectively (Supplementary Data 5 and 6, Supplementary Fig. 13a). Gene Ontology (GO) analyses showed that both the hyper-DEGs and hypo-DEGs were enriched in genes with roles in responses to stimuli (Supplementary Fig. 13b and 13c). We also examined the global transcript levels of miRNA targets from the RNA-seq data. Although several examined miRNA targets were found to be derepressed in the mutant by RT-qPCR, a global trend of increased levels of miRNA target transcripts was not observed (Supplementary Fig. 13d, Supplementary Data 7). It is possible that the RBV mutation leads to a defect in the transcriptional regulation of these genes as well.
It has been reported that MOS4-Associated Complex (MAC) components MAC3A, MAC3B, MAC7, and two WD40 repeat proteins (PRL1 and PRL2) affect both miRNA biogenesis and the splicing of protein-coding transcripts36, 37, 38. This prompted us to examine whether the rbv-1 mutant had splicing defects using the RNA-seq data (see Methods). In total, 474 Genes were found to have intron retention defects in the rbv-1 mutant compared to Col, with a total of 511 intron retention events (Supplementary Data 8). Two examples (At4g15790 and At1g03280) are presented in Fig. 6a. PI (percent of intron reads) was increased in the rbv-1 mutant (Fig. 6b). No differential expression of the genes with intron retention was observed between the rbv-1 mutant and wild type (Supplementary Fig. 14a). Next, we examined whether the genes with intron retention or the retained introns in the rbv-1 mutant had any common features. Intriguingly, the retained introns in the rbv-1 mutant seemed to be shorter as compared to the overall length distribution of introns (P= 1.984083e-22) (Fig. 6c). Besides, genes affected in rbv-1 tended to have more introns than all genes (P=2.823e-05) (Fig. 6d). Intron retention defects are also found in prl1 prl2 and mac3a mac3b mutants36. We compared the retained introns in rbv-1, prl1 prl2 and mac3a mac3b and found little overlap between the introns affected in rbv-1 with those affected in either prl1 prl2 or mac3a mac3b (Fig. 6d), suggesting that RBV affects different introns from the MAC components.
Many MIR genes have introns39, 40, 41. The RNA-seq did not detect intron retention events in pri-miRNAs, but the low abundance of pri-miRNAs could have precluded the detection of intron retention events. We performed RT-PCR to test the splicing of introns from three miRNA precursors (pri-miR163, pri-miR156 and pri-miR168) with intron-flanking primers. Genomic DNA was amplified with the same primers to indicate the size of the intron-containing fragments. No defects in pri-miRNA splicing were observed in the rbv-1 mutant (Supplementary Fig. 14b).