Arabidopsis micro-RNA biogenesis through Dicer-like 1 protein functions

Plant Materials. Plant growth conditions were as described (14) and differed only in photoperiod length, which was 16 h/8 h in this study. DCL1-9/DCL1-9 homozygous plants were distinguished from DCL1-9/dcl1-9 heterozygous plants by PCR amplification of the octopine synthase terminator sequence of the T-DNA (portion of the tumor-inducing plasmid that is transferred to plant cells) by using genomic DNA as a template and the primers 5′-CTCCGTTCAATTTACTGATTGTAC-3′ and 5′-TTGAATGGTGCCCGTAACTTTCG-3′.

Northern Blot Analysis. Total RNA was extracted from the entire aerial portions of plants inoculated with mock or crucifer tobamovirus Cg (TMV-Cg) (14) by using TRIzol reagent (Invitrogen). Low molecular weight RNA plus (LMW RNA+) was isolated by anion-exchange chromatography (RNA/DNA Midi kit, Qiagen, Valencia, CA) according to the manufacturer's instructions with a minor exception: a specific elution buffer [50 mM Mops/1.0 M NaCl, pH 7.0/15% (vol/vol) ethanol] was used in the elution step. For detection of pri-miR163, pre-miR163, and miR163, LMW RNA+ was resolved with electrophoresis on a denaturing 5% polyacrylamide gel (8 M urea) and 15% polyacrylamide gel (7 M urea), respectively, in 0.5× TBE buffer (89 mM Tris/89 mM boric acid/2 mM EDTA, pH 8.3) and electroblotted onto Hybond-N⁺ membranes (Amersham Pharmacia). Radiolabeled DNA probes (probes 1 and 2) were constructed by random priming reactions in the presence of [α-³²P]dCTP by using the Megaprime DNA-labeling system (Amersham Pharmacia). Radiolabeled DNA oligonucleotide probes were constructed by end-labeling with [γ-³²P]ATP by using T4 polynucleotide kinase. Two probes, 1 and 2, contained the following sequences, with nucleotide numbers starting at the beginning of the pri-miR163: nucleotides 149-479 (probe 1) and nucleotides 1-146 (probe 2). Probes 3 and 4 were end-labeled DNA oligonucleotides: 5′-ATCGAAGTTCCAAGTCCTCTTCAA-3′ (probe 3) and 5′-CGATTTTATCCAAAGGGTTTTGGT-3′ (probe 4). Probe 1′ contained the region of DCL1-9 pre-miR163 corresponding to probe 1. The probes for detection of upper-left (UL), miR163^*, and lower-left (LL) small RNAs were also end-labeled DNA oligonucleotides that were complementary to their respective sequences.

Mapping of pri-miR163 Sequence. For mapping of the pri-miR163 sequence, RNA ligase-mediated rapid amplification of cDNA ends was performed by using the GeneRacer kit (Invitrogen) according to the manufacturer's instructions. Total RNA was extracted from whole Arabidopsis plants. Poly(A)⁺ RNA was purified by using the Oligotex-dT30 Super mRNA-purification kit (Takara) and ligated to the GeneRacer RNA oligo adapter (5′-CGACUGGAGCACGAGGACACUGACAUGGACUGAAGGAGUAGAAA-3′). The GeneRacer Oligo(dT) primer [5′-GCTGTCAACGATACGCTACGTAACGGCATGACAGTG(T)₁₈-3′] was then used to prime first-strand cDNA synthesis in the reverse-transcription reaction. The 5′ ends of pri-miR163 cDNA were amplified with the GeneRacer 5′ primer (5′-CGACTGGAGCACGAGGACACTGA-3′) and gene-specific primer (5′-ACACGGGGGATAATATCGAA-3′). The 3′ ends were amplified with the GeneRacer 3′ primer (5′-GCTGTCAACGATACGCTACGTAACG-3′), GeneRacer 3′ nested primer (5′-CGCTACGTAACGGCATGACAGTG-3′), and gene-specific primer (5′-ACCCGGTGGATAAAATCGAG-3′). PCR products were gel-purified and cloned into pCR2.1 TOPO vector (Invitrogen) for sequencing.

Determination of Cleavage-Site Sequence. Aliquots of 5 μg of low molecular weight RNA plus were self-ligated with T4 RNA ligase; the reaction volume was 100 μl, and the reaction was performed overnight at 14°C. After the ligation reaction, RNAs were phenol/chloroform-extracted and precipitated with ethanol, and pre-miR163s or remnants were amplified by RT-PCR with the gene-specific primers 5′-GAAGTAACTGCGCAGTGCTTA-3′ and 5′-ACCGTGCTCTTCCTAAGAAG-3′, which were used in both the reverse-transcription reaction and PCR. These two primers have sequences corresponding to nucleotides 375-394 (antisense) and 399-419 (sense) of pri-miR163, respectively. PCR products were gel-purified and cloned into pCR2.1 TOPO vector for sequencing.

Prediction of RNA Secondary Structure. The potential secondary structures of pre-miR163 were predicted by using mfold 3.1 (http://molbio.info.nih.gov/molbio-nih/mfold.html).

Characterization of pri-miR163. Most of the genes coding miRNA sequences are located in intergenetic regions of the genome. In the Arabidopsis genome, the miR163 gene is located on chromosome 1 between the At1g66720 and At1g66730 genes (Fig. 1A). We discovered that both pri-miR163 and pre-miR163 could be detected easily (unpublished data; Fig. 2). To examine the detailed process of miRNA biogenesis, we first determined the full sequence of pri-miR163 by 5′ and 3′ rapid amplification of cDNA ends methods using oligo(dT) primers. Interestingly, we obtained two types (types 1 and 2) of pri-miR163 sequences [GenBank accession nos. AY615373 (type 1) and AY615374 (type 2)] (Fig. 1B). Both have a polyadenylation signal, but type 2 has an intron bearing exact splice sites. Because the polyadenylation signal of type 1 was located in the intron of type 2, it was thought that type 2 pri-miR163 emerged by splicing through exclusion of the polyadenylation signal for type 1 pri-miR163. Another possible polyadenylation signal for type 2 pri-miR163 was found downstream from the intron. In addition, we could find a TATA box-like sequence upstream from the transcribed region. Taken together, it is obvious that pri-miR163 is transcribed by RNA polymerase II.

miR163 Biogenesis Requires at Least Three Cleavage Steps. We performed Northern blot analysis by using a probe complementary to the putative stem-loop region (Fig. 2A, probe 1) and then detected some species of RNAs in uninfected (mock) and TMV-Cg-infected plants. TMV-Cg infection could increase accumulation of most miRNAs (unpublished data). Previous work showed that expression of viral suppressors of RNA silencing increased miRNA accumulation (15-17). We preferentially used a TMV-Cg-infection condition for this Northern blot analysis, because it dramatically increased the accumulation of band d, described below, as well as miR163. Judging from the molecular lengths, it was assumed that bands a, c, and d correspond to pri-miR163, pre-miR163, and remnant after release of mature miR163, respectively. Band b probably corresponds to the precursor of pre-miR163. To confirm our assignments, we used three other probes (probes 2-4) that are complementary to the 5′ end sequence of pri-miR163, the 24-nt miR163 sequence, and the stem root of pri-miR163 (Fig. 2A). Probe 2 detected band a and new reproducible band e, probe 3 detected bands a-c, and probe 4 detected bands a and b (Fig. 2B, probes 2-4) as expected. Fig. 2C shows detection of mature miR163 using probe 3.

Based on the results shown in Fig. 2B and C, we suggest a model for miR163 biogenesis (Fig. 2D). The pri-miR163 (a) is processed into long pre-miR163 (b) at the root of the stem-loop structure, releasing 5′ remnant (e), which corresponds to 5′ end of pri-miR163 (first step). At the next step, long pre-miR163 is cleaved into short pre-miR163 (c) at the edge of the miR163 sequence (second step). Last, short pre-miR163 is cleaved at the opposite edge of the miR163 sequence, producing mature miR163 and remnant (d) (third step). Thus, miR163 biogenesis requires at least three cleavage steps, probably by RNase III-like enzymes.

Determination of Positions and Forms of RNase III-Cleavage Sites. To determine the forms of three cleavage sites and confirm whether they are catalyzed by RNase III proteins, we used a method described previously (18). RNAs were self-ligated, followed by RT-PCR using gene-specific primers (Fig. 3A). Three bands detected between 200- and 400-nt molecular markers were cloned and sequenced (Fig. 3B). The result is shown in Fig. 3C. All three identified cleavage sites had a 3′ overhang of 2 or 3 nt, which is a feature of RNase III cleavage (Fig. 3C). It is notable that the length of two intervals between separate cleavage sites was 21 nt long if a few bulges were not counted. These intervals indicate that any DCL proteins that are involved in miRNA biogenesis probably measure ≈21 nt in length from one end to the next cleavage site, and then the site was processed. This hypothesis is parallel with a previously proposed cleavage model of animal Dicer (19). Our result also suggested that the enzyme measures 21 nt on one strand of the duplex RNA, because the miR163 strand is actually 24 nt in length, whereas the other strand is 21 nt (Fig. 3C, gray box).

From this result, it was suspected that the other three cleavage products [which we named upper left (UL), lower left (LL), and miR163^*, respectively; Fig. 3C] might be miRNAs. Therefore, we performed Northern blot analysis for detection of these products and miR163 by using probes complementary to them. Small RNA UL and miR163 were detected easily, but small RNA LL and miR163^* were not detected easily (Fig. 3D). We noticed that small RNA UL was identical to small RNA 91, which has been designated already (3). Thus, the small RNA UL (or 91) could be a candidate for miRNA. However, we could not discover complementary target mRNAs to small RNA UL by using the conventional blast method, probably because of lower complementarity.

Disruption of miR163 Biogenesis in dcl1 Mutants. miR163 biogenesis in dcl1-7 [sin1-1; Landsberg erecta (Ler) background] and dcl1-9 (carpel factory) mutants were examined by Northern blot analysis. The dcl1-7 mutant carries an amino acid substitution (P415S) in the RNA helicase domain of the DCL1 protein (10, 20), whereas the dcl1-9 mutant has a T-DNA insertion in the coding region for a second dsRNA-binding domain of DCL1 protein (10, 21). Both mutants fail to accumulate miRNAs (11, 16, 22). First, we performed Northern blot analysis by using probe 1 (shown in Fig. 2A). In dcl1-7 mutants, the accumulation of long pre-miR163 was at a similar level to that seen in wild-type Ler plants (Fig. 4A, band b), whereas the accumulation of short pre-miR163 and miR163 were at lower levels than those seen in wild-type plants (Fig. 4A, band c). When the accumulation levels of band b were normalized to 1.0 and the relative digital density of band c was evaluated, the c/b ratio in the case of dcl1-7 was lower than that seen in the case of wild-type Ler (Fig. 4A, under the panels).

The background in DCL1-9 plants was not wild-type Ler (at least with respect to the miR163 gene). The pri-miR163 gene in the DCL1-9 plants had an unexpected 12-nt deletion and 60-nt insertion in the loop-forming region of its transcript (Fig. 6, which is published as supporting information on the PNAS web site), although the structure of the stem-forming region of pre-miR163s was completely the same. Therefore, the molecular length of long pre-miR163, short pre-miR163, and the remnant should be 48 nt longer than those of Col-0 or Ler plants. In dcl1-9/dcl1-9 homozygous mutants, the accumulation of short pre-miR163 was at a lower level than that seen in DCL1-9/DCL1-9 homozygous or DCL1-9/dcl1-9 heterozygous plants (Fig. 4B, band g), whereas the accumulation of long pre-miR163 was at a similar level to that seen in DCL1-9/DCL1-9 or DCL1-9/dcl1-9 plants (Fig. 4B, band f). The g/f ratio was also lower than that of DCL1-9/DCL1-9 and DCL1-9/dcl1-9 plants (Fig. 4B, under the panels). These results indicate that DCL1 catalyzes at least the second cleavage step from long pre-miR163 to short pre-miR163.

Next, we performed Northern blot analysis by using probe 2 (shown in Fig. 2A). In dcl1-7 mutants, the 5′ remnant released after the first cleavage step was not detected (Fig. 4C, band e), and the accumulation of pri-miR163 increased (Fig. 4C, band a). This result indicates that DCL1 also catalyzes the first cleavage step from pri-miR163 to long pre-miR163. In all DCL1-9/DCL1-9, DCL1-9/dcl1-9, and dcl1-9/dcl1-9 backgrounds, the bands corresponding to the 5′ remnant were not detected (data not shown).

Previous work showed that in both dcl2 and dcl3 mutant plants, the accumulation of miRNAs was not affected compared with wild-type plants (13). We examined whether the defects in dcl2 and dcl3 mutants could affect miR163 biogenesis. The dcl2-2 (SALK_123586) and dcl3-1 (SALK_005512) mutants contain T-DNA insertions in the DCL2 (At3g03300) and DCL3 (At3g43920) genes, respectively. Accumulation of all miR163-related RNAs, including miR163 itself, was not affected in dcl2-2 and dcl3-1 mutants (Fig. 4D).

Aberrant Positioning of Cleavage Sites in the dcl1-9 Mutant. We examined the forms of three cleavage sites of miR163 and its precursors in dcl mutant backgrounds in the same way as described above. As shown in Fig. 5A,in dcl1-7 and dcl1-9/dcl1-9 mutants, RT-PCR products had a relatively slow mobility compared with that of the wild type (Ler or DCL1-9/DCL1-9). This result was consistent with the results shown in Fig. 4A and B, in which the accumulation of lower molecular RNA species (short pre-miR163 and remnant) was lower than that of the wild type (Fig. 4A and B).

We performed cloning and sequencing of the RT-PCR products from dcl1-7, dcl1-9/dcl1-9, DCL1-9/DCL1-9, dcl2-2, and dcl3-1 mutants. Three cleavage sites were consistent (as shown in Fig. 3C) among dcl1-7, DCL1-9/DCL1-9, dcl2-2, and dcl3-1 mutants (data not shown), whereas the cleavage sites were different in dcl1-9/dcl1-9 mutants. The result of the dcl1-9/dcl1-9 mutant is shown in Fig. 5B. Cleavage sites were not strictly identical or mapped within 4-5 nt as depicted in Fig. 5B (expressed in the gray zones). The first cleavage site was mapped within zone f (Fig. 5B), and the second cleavage site was within zone g. Both sites shifted to the loop side compared with those of the DCL1-9/DCL1-9 plant. It is interesting that the length of the interval between these two cleavage sites is ≈21 nt long. It was thought that dcl1-9 protein could not process the regular cleavage position(s) because of its disrupted second dsRNA-binding domain. Therefore, dsRNA-binding domains of DCL1 protein have an important role in determination of the cleavage positions. We sequenced 30 clones, but the third cleavage site was not identified, probably because the position of the 21-nt sequence from the second cleavage site was not in the dsRNA-forming region.

Taken together with the results shown in Fig. 4, this result in dcl1-9 mutants suggests that DCL1 catalyzes at least the first and second cleavage steps.

miRNA Biogenesis in Plants. In this article we investigated detailed miR163 biogenesis, which is mediated through Dicer-like1 protein functions. Our data indicate that DCL1 not only catalyzes at least the first cleavage step from pri-miR163 to long premiR163 and the second cleavage step from long pre-miR163 to short pre-miR163, releasing another miRNA candidate, small RNA UL, but also is involved in positioning of the cleavage sites. We also suspected that the third cleavage step might be catalyzed by DCL1. Therefore, it is suggested that all cleavage steps are probably catalyzed by the DCL1 protein. Thus, miR163 biogenesis is a powerful tool to investigate plant miRNA-maturation processes.

miR163 biogenesis requires three cleavage steps by RNase III-like enzymes. Because the dsRNA-forming stem regions of most of the other predicted pre-miRNAs in Arabidopsis were much shorter than that of pre-miR163, and they do not have the two 21-nt-length sequences, it is supposed that most plant miRNA biogenesis generally requires only two cleavage steps, from pri-miRNA to pre-miRNA and from pre-miRNA to the remnant and miRNA.

Previous work has shown that pre-miR167 is found in the nucleus but not in the cytoplasm, whereas mature miR167 is found in the cytoplasm but not in the nucleus in the tomatoes (23). It has also been shown that Arabidopsis DCL1 protein contains two putative nuclear localization signals (NLSs) (10) and partial DCL1 (including NLSs); green fluorescent protein fusion proteins localize to nuclei in the onion cells (24). Taken together, our results suggest that all processes of plant miRNA biogenesis occur in the nuclei.

In animal miRNA biogenesis, the first cleavage step from pri-miRNA to pre-miRNA is catalyzed by an RNase III enzyme, Drosha, but not Dicer (5). However, whole-genome analysis suggested that the Arabidopsis genome does not encode a Drosha homologue. In Arabidopsis, our data suggest that the Dicer homologue DCL1 has a Drosha function to catalyze the first cleavage step as well as a Dicer function. In addition, plant pre-miRNAs are usually much longer than those of animals. These differences between plant and animal miRNA biogenesis are very interesting, although it is unknown why they occur.

It was shown that dsRNA-binding domains of DCL1 protein have an important role in determination of the first and second cleavage positions (Fig. 5B). However, it was observed that the 21-nt interval between the first and second cleavage positions was maintained in dcl1-9/dcl1-9 mutants (Fig. 5B). Therefore, we supposed that dcl1-9 proteins processed pri-miR163 at an irregular position but could exactly measure 21 nt from the first cleavage position to the next position.

This paper was submitted directly (Track II) to the PNAS office.

Abbreviations: miRNA, micro-RNA; pri-miRNA, miRNA primary transcript; pre-miRNA, miRNA precursor; DCL, Dicer-like; dsRNA, double-stranded RNA; T-DNA, portion of the tumor-inducing plasmid that is transferred to plant cells; TMV-Cg, tobamovirus Cg; Ler, Landsberg erecta.

Data deposition: The sequences reported in this paper have been deposited in the GenBank database [accession nos. AY615373 (pri-miR163 sequence type 1) and AY615374 (pri-miR163 sequence type 2)].

We thank the Salk Institute and Arabidopsis Biological Resource Center for providing the seeds for dcl2-2 (SA LK_123586), dcl3-1 (SALK_005512), dcl1-7 (sin1-1; stock number CS3089), and dcl1-9 (carpel factory; stock number CS3828). This work was supported in part by Molecular Mechanisms of Plant-Pathogenic Microbe Interaction Grant-in-Aid 08680733 and Spatiotemporal Network of RNA Information Flow Grant-in-Aid 14035215 for scientific research, priority area (A), awarded by the Ministry of Education, Culture, Sports, Science, and Technology of Japan (to Y.W.).