SKI2 mediates degradation of RISC 5'-cleavage fragments and prevents secondary siRNA production from miRNA targets in Arabidopsis

Abstract

Small regulatory RNAs are fundamental in eukaryotic and prokaryotic gene regulation. In plants, an important element of post-transcriptional control is effected by 20-24 nt microRNAs (miRNAs) and short interfering RNAs (siRNAs) bound to the ARGONAUTE1 (AGO1) protein in an RNA induced silencing complex (RISC). AGO1 may cleave target mRNAs with small RNA complementarity, but the fate of the resulting cleavage fragments remains incompletely understood. Here, we show that SKI2, SKI3 and SKI8, subunits of a cytoplasmic cofactor of the RNA exosome, are required for degradation of RISC 5', but not 3'-cleavage fragments in Arabidopsis. In the absence of SKI2 activity, many miRNA targets produce siRNAs via the RNA-dependent RNA polymerase 6 (RDR6) pathway. These siRNAs are low-abundant, and map close to the cleavage site. In most cases, siRNAs were produced 5' to the cleavage site, but several examples of 3'-spreading were also identified. These observations suggest that siRNAs do not simply derive from RDR6 action on stable 5'-cleavage fragments and hence that SKI2 has a direct role in limiting secondary siRNA production in addition to its function in mediating degradation of 5'-cleavage fragments.

Figure 1.

Isolation of the ski2–4 mutant. (A) Left panel, northern blot analysis of 20 μg of total inflorescence RNA from GFP171.1/rdr6 and ski2–4/rdr6. The same blot was hybridized consecutively to radiolabeled 5′ and 3′ probes as indicated in the schematic of the GFP171.1 transgene. FL, full length. rRNA, ethidium bromide stained ribosomal RNA in gel prior to blotting. Right panel, northern blot of total inflorescence RNA probed with end-labeled oligonucleotides complementary to miR171 and U6. (B) RNA and protein analysis from inflorescences performed as described in (A) above. (C) Fluorescence microscopy images of GFP accumulation in cross sections of stems of the indicated genotypes.

Figure 2.

Characterization of 5′-cleavage fragments in ski2–4. (A) 10 μg of total RNA was incubated in Terminator buffer with or without enzyme, and the RNA was analyzed by northern blot hybridization to a radiolabeled 5′ GFP probe (Figure 1A). (B) Fractionation of poly(A)+ RNA using oligo(dT) coupled beads. 10 μg of total RNA was used as input. The unbound supernatant fraction (SN) was ethanol precipitated and analyzed in parallel with bound and total fractions by northern blot as in (A). (C–F) Northern blot analyses of 20 μg of inflorescence RNA. Radiolabeled probes specific to either 5′ or 3′ parts of LOM2, AGO1, MYB33 and CSD2 mRNAs with respect to miRNA cleavage sites were hybridized to six northern membranes prepared from the same batch of inflorescence RNA. 5′ and 3′ probes for each transcript were hybridized to different membranes. AGO1 and MYB33 5′-probes were hybridized consecutively to the same northern membrane, while independent membranes were prepared for hybridizations to LOM2 5′, LOM2 3′, CSD2 5′ and CSD2 3'-probes. AGO1 and MYB33 3′-probes were hybridized consecutively to the same membrane.

Figure 3.

Identification of the SKI2 gene. (A) SHOREmap allele-frequency plot of a pool of genomic DNA sequences obtained from 99 ski2–4 mutants selected in the F2 population of a ski2–4/rdr6 (C24) x rdr6–3 (Ler) mapping population. Positive and negative values indicate enrichment of Ler and C24 alleles, respectively. (B) Schematic diagram of the SKI2 gene, indicating functional domains (Pfam nomenclature) and the positions of the ski2–4 point mutation and the ski2–2 and ski2–5 T-DNA insertions. DExH and Helicase C, domains typical of RNA helicases; rRNA processing, domains necessary for ribosomal RNA processing in the SKI2 homologue MTR4; DCHCT, C-terminal domain occurring in DOB1/SKI2/helY-like DExH-box helicases. (C) Northern blot analysis of total inflorescence RNA hybridized with LOM2 and AGO1 probes as in Figure 2C, D. For LOM2, this set of samples exhibits unusually weak overaccumulation of the 5′-cleavage fragment in ski2 mutant alleles compared to parental lines (the ski2/parental line ratio of 5′-cleavage fragment intensity normalized to full length transcript is 1.1 for the samples of the three ski2 alleles shown here) (D–F) Northern blot analysis of total inflorescence RNA hybridized with AGO1, MYB33, and CSD2 5′ and 3′ probes as in Figure 2D–F. The same two membranes were used for the four AGO1 and MYB33 hybridizations (pairs of AGO1 5′/MYB33 5′ and AGO1 3′/MYB33 3′ hybridized to the same membranes), while two different membranes were used for CSD2 5′ and CSD2 3′ hybridizations.

Figure 4.

Transitive GFP siRNAs in ski2–4. (A) Top panel, inflorescence RNA analyzed by high molecular weight northern blot with a 5′ GFP probe as in Figure 1A. Bottom panel, same RNA analyzed by low molecular weight northern blot with consecutive hybridizations to 5′-GFP, 3′-UTR (as in Figure 1A), miR159 and miR171 probes in that order. miR159 and miR171 hybridizations demonstrate the integrity of the membrane after stripping of the 5′-GFP signal. The 3′-GFP probe had a comparable number of cpm/ml as the one used for the hybridization shown in Figure 1A. (B) Overview of siRNAs mapping to the GFP171.1 transcript in SKI2/RDR6, SKI2/rdr6, ski2–4/rdr6 and ski2–4/RDR6. Ordinate, reads per million.

Figure 5.

SKI2 limits transitivity on miRNA targets. (A) Venn diagram showing categories of loci with numbers of siRNA reads in ski2–4/RDR6 significantly different from SKI2/RDR6 (exact negative binomial test, P < 0.05). The enrichment of miRNA targets in loci with higher numbers of siRNA in ski2–4/RDR6 is highly significant (Fisher-test: P < 2.2 × 10⁻¹⁶); this is not the case for the group of loci with lower numbers of siRNA in ski2–4/RDR6 (Fisher-test: P = 0.09). (B) Boxplot of normalized mapped siRNA read counts in SKI2/RDR6, SKI2/rdr6, ski2–4/rdr6 and ski2–4/RDR6 for loci enriched in siRNAs in ski2–4/RDR6 compared to SKI2/RDR6. Bars indicate medians, boxes indicate data between the first and third quartiles (Q1 and Q3, respectively) defining the interquartile range (IQR). Whiskers correspond to data between Q1 – 1.5 x IQR and Q3 + 1.5 IQR. Other values correspond to outliers and are represented as dots. (C) Examples of siRNA accumulation mapped to miRNA targets. Cleavage sites are indicated by dashed lines. PHO2 mRNA contains five closely spaced miR399 binding sites, all other dashed lines indicate the presence of a single miRNA binding site. The ordinate gives number of small RNA reads per 10 million, the abscissae depict TAIR9 coordinates corresponding to each gene. Small RNA libraries prepared from two replicates of each genotype were used for all analyses presented in this figure. (D) Strength of base pairing between miRNA and 5′ and 3′ cleavage fragments (cf). ΔGs were calculated for hybridization between free RNA strands with UNAFold (64) at 21°C, 150 mM NaCl, 5 mM MgCl₂. PHO2 contains five closely spaced miR399 sites of which one is shown. Four of the five sites are similar to the one shown, the last site has stronger pairing to the 5′-cleavage fragment (See Supplementary Table S2). For LOM1, only pairing to the most abundant isomiR (miR171a) is shown. The other members of the miR171 family (miR171b/c, miR170) show stronger relative base pairing to the 3′-cleavage fragment (See Supplementary Table S2). For other targets, differences between isomiRs are negligible.