In plants, decapping prevents RDR6-dependent production of small interfering RNAs from endogenous mRNAs

Plant material and growth conditions

All plants are in the Columbia accession with the exception of the dcp2^tdt-1 mutant (25), which has been back-crossed six times to Ler and which was kindly provided by L.E. Sieburth (8). Lines L1, Hc1 and 6b4, and mutants vcs-6 (SAIL_831_D08), dcp2-1 (SALK_000519) and rdr6^sgs2-1 have been previously described (6,8–9,26). The following mutants were identified in this study: dcp1-3 (SAIL_377_B10), vcs-8 (SAIL_1257_H12) and vcs-9 (SAIL_218_E01). For GUS analyses, plants were grown on Bouturage media (Duchefa) in standard long-day conditions (16 hours light, 8 hours dark at 20–22°C) and transferred to soil after two weeks and grown in controlled growth chambers in standard long-day conditions. dcp1-3, vcs-8 and vcs-9 were crossed to Hc1, which is in the Col ecotype, while dcp2^tdt-1 was crossed to Hc1/Ler, i.e. Hc1 that had been back-crossed 10 times to the Ler ecotype. For small RNAs profiling, plants were grown on Bouturage media (Duchefa) in long-day conditions (16 hours light, 8 hours dark) at 15°C and aerial portions were collected 12 dag (days after germination).

Molecular methods

RNA extraction, RNA gel blot analysis, GUS extraction and activity quantification were described before (20). See supplementary information online for more details.

Small RNAome profiling and analysis

For each genotype, the aerial portions of 20 12-day-old plants were pooled, and two biological replicates were assayed. RNA samples enriched for small fractions were obtained with mirVana™ miRNA Isolation Kit (#AM1560, Ambion®/Life Technologies Corporation). They were checked for their integrity on RNANano chip, using Agilent 2100 bioanalyzer (Agilent Technologies, Waldbroon, Germany). Small RNA-seq libraries were performed according to NEBNext® Multiplex Small RNA Library Prep Set for Illumina instructions with a different bar code for each sample (#E7300S, New England Biolabs, Inc.). Small RNA-seq libraries are checked for their quality on DNA1000 chip using Agilent 2100 bioanalyzer (Waldbroon, Germany) before Illumina sequencing (Illumina®, California, U.S.A.). The SmallRNA-seq samples have been sequenced in Single-End (SE) with a read length of 100 bases and a multiplexing rate of 10 SmallRNA-seq libraries on one lane of Hiseq2000 machine to obtained around 15 millions reads/sample. We extracted 18–26 nt reads from sequencing files, removed reads mapping the tRNA and rRNA and mapped the remaining reads to the A. thaliana genome (TAIR10, perfect matches) using Bowtie2 (V. 2.2.0 – (27)). Read counts for annotated genes were obtained using bedtools (V. 2.17.0 – (28)) in a strand-independent fashion. Intergenic blocks of 18–26 nt RNAs were defined using bedtools as follows: we computed intergenic coverage for each independent library, merged coverage files and extracted all segments with non-zero coverage to produce a unique list of intergenic blocks. Read counts were then obtained for each block and library. Differential expression of annotated genes was estimated using the DESeq package (V 1.16.0 – (29)). For each pairwise comparison, read counts were normalized across all replicates and significant differentially expressed features were determined by a P-adjusted value ≤ 0.05 and a log2 fold change >0 (upregulated) or <0 (downregulated). A locus produce siRNAs from both strands when the ratio of sense to antisense and antisense to sense reads was ≥0.25. Metagene representations were produced by a custom R script. All data were deposited to GEO under the reference GSE65056.

Nicotiana benthamiana and Arabidopsis thaliana agro-infiltration

Transient expression in tobacco or in 4-day-old Arabidopsis thaliana seedlings were performed as described before (19,30). See supplementary information online for more details.

Whole-mount indirect immunofluorescence

Whole-mount immunofluorescence was performed on 5-day-old seedlings according to (31), with some modifications. See supplemental material for detailed protocol.

Imaging and image analysis

See supplemental material for detailed protocol.

Plasmid and cloning

See supplementary material for details on the plasmids used and cloning procedures.

Viable alleles in the decapping machinery exhibit increased transgene PTGS

An enhancement of RDR6-dependent transgene PTGS by mutations in the P-body-resident enzyme DCP2 was previously described (23). We further examined the functional relationships between decapping and PTGS on transgenes and characterized the consequences of mutations in DCP1, DCP2 and VCS on PTGS using the well-characterized Arabidopsis reporter lines L1, Hc1 and 6b4, which all carry a 35S::GUS transgene (20,26,32). To circumvent the PTGS analysis limitations imposed by the seedling lethality of decapping null alleles (6), while simultaneously avoiding the potential for p35S-induced interference (33), we identified partial-loss-of-function 35S-free T-DNA insertion mutants in the promoters, introns or at the 3′ end of the coding sequence. New partial-loss-of-function mutants were identified for dcp1 (hereafter referred to as dcp1-3), and vcs (vcs-8 and vcs-9, Supplemental Figure S1). First, we tested the effect of dcp1-3, vcs-8 and vcs-9 on line 6b4, which naturally does not trigger PTGS spontaneously. The absence of GUS mRNA and the presence of GUS siRNAs indicated that 6b4 triggers PTGS in these mutant backgrounds (Supplemental Figure S2A and B). Secondly, the effect of dcp1-3, dcp2^tdt-1 (8), vcs-8 and vcs-9 on Hc1, a line that spontaneously triggers PTGS with 20% efficiency, was tested. Our analysis indicated that Hc1 PTGS was enhanced in dcp1-3, dcp2^tdt-1, vcs-8 and vcs-9 mutants (Supplemental Figure S2C and D). Lastly, we tested the effect of the vcs-6 seedling lethal allele (6) on L1. For this, we took advantage of the fact that although PTGS is triggered in 100% of adult L1 plants, it is possible to monitor an acceleration of L1 PTGS when endogenous PTGS suppressors are mutated (34). At five, seven and ten days after germination, silencing was enhanced in L1/vcs-6 mutant plants compared with L1 plants (Supplemental Figure S2E). Altogether, these results indicate that decapping prevents or suppresses transgene silencing.

Mutations in RDR6 rescue the lethality of vcs and dcp mutants

Given the antagonistic effect of decapping (dcp1, dcp2, vcs)(23 and our results) and PTGS (rdr6) (15,16) mutations on transgene PTGS, we postulated that the lethality of decapping mutants could result, at least in part, from the enhanced entry of endogenous mRNAs in the PTGS pathway due to the impairment of decapping activity in P-bodies. Indeed, when the RNA turnover machinery recognizes a dysfunctional transcript only this transcript is eliminated, while homologous functional transcripts remain unaffected. In contrast, PTGS is less specific because it involves the production of siRNAs that guide the degradation of both dysfunctional and functional homologous transcripts. Therefore, when P-body components are mutated, it is possible that dysfunctional endogenous mRNAs enter the PTGS pathway, resulting in the indiscriminate degradation of both dysfunctional and functional homologous mRNAs. On this ground, we anticipated that the lethality of null P-body component mutants could be, at least partly, rescued by null siRNA-body component mutants. To test this hypothesis, we crossed the vcs-6 null mutant (6), which is seedling lethal, to the rdr6^sgs2-1 null mutant (15,26), which exhibits very mild developmental defects. Whereas vcs-6 single mutants died at the two-cotyledon stage, vcs-6 rdr6^sgs2-1 double mutants developed leaves, stems and flowers (Figure 1A and B). However, they remained smaller than wildtype plants and were sterile, indicating that the rdr6 mutation does not suppress all of the developmental defects caused by the absence of a functional decapping complex. To extend this finding, we also crossed the dcp2-1 null mutant (6,8–9) to rdr6^sgs2-1. Whereas dcp2-1 single mutants died at the two-cotyledons stage like vcs-6 single mutants, dcp2-1 rdr6^sgs2-1 double mutants developed like vcs-6 rdr6^sgs2-1 (Figure 1C and D).

Impairing decapping leads to the production of rogue siRNAs from endogenous loci

The rdr6-mediated suppression of the lethality caused by the absence of a functional decapping complex, led us to postulate that, in absence of decapping in P-bodies, endogenous mRNAs could enter the PTGS pathway and produce siRNAs. To test this hypothesis, we profiled genome-wide the small (18–26 nt) RNAs present in plants defective in only P-body decapping or plants defective in both decapping and PTGS. We profiled by RNA-seq the small RNAs fractions of two biological replicates of 12-day-old wildtype plants, the null vcs-6 or dcp2-1 decapping mutant and the vcs-6 rdr6^sgs2-1 or dcp2-1 rdr6^sgs2-1 double mutant. For each genotype, the 18–26 nt reads were mapped to the Arabidopsis genome allowing no mismatches to identify transcription units that produce small RNAs (see Supplemental Table S1 and Figure 2). In support of our hypothesis, we identified 1250 transcription units that produced more small RNAs in the vcs-6 mutant than in wildtype plants (FC > 1 and P ≤ 0.05) and 1351 when comparing dcp2-1 to wildtype (Figure 2A, B and Supplementary Table S3). 816 of these transcription units produced more small RNAs in both dcp2-1 and vcs-6 mutants (Figure 2C), indicating that decapping generally prevents the production of siRNAs from hundreds of endogenous loci. We dubbed this new class of small RNAs, rqc-siRNAs because they appear in mutants impaired in RNA quality control. Rqc-siRNAs are predominantly 21nt long (Figure 2D, E and Supplemental Figure S3), and 40% start with a uradine (Supplemental Figure S4). More than 73% of the loci producing rqc-siRNAs correspond to protein-coding genes (See Supplemental Figure S5). As expected, loci producing rqc-siRNA presented reduced levels of corresponding mRNAs in vcs-6 mutant but near wildtype levels in the vcs/rdr6 double mutants (Supplemental Figure S6). For more than 80% of the rqc-siRNAs-producing loci, the rqc-siRNAs are derived from both the sense and antisense strands (Figure 3A, C, E–I). That rqc-siRNAs derive from the two strands of otherwise single stranded mRNAs, suggests that these endogenous mRNAs are transformed to dsRNAs in the vcs-6 and dcp2-1 mutants, by an RDR activity. In agreement with this, 20% of the loci producing rqc-siRNAs in vcs-6 mutant and 15% of those in dcp2, produced less rqc-siRNA when RDR6 is mutated (Figure 2A, B, F and Figure 3B, D, F–I), implicating RDR6 in the production of a subset of these rqc-siRNAs. Our genome wide analysis of small RNA abundance in the dcp2-1 and vcs-6 mutants revealed that unlike rqc-siRNAs that become more abundant upon impairment of decapping, the abundance of several miRNAs was reduced in vcs-6 and dcp2-1 mutants. 18 and 36 mature miRNAs were respectively reduced in vcs-6 and dcp2-1 mutants compared to wildtype (Supplemental Table S4), a result in agreement with prior observations (35).

Altogether, these analyses show that mutations in the decapping complex provoke the production of siRNAs from endogenous mRNAs by the PTGS machinery. This suggests that in the absence of functional decapping enzymes, endogenous RNAs are more prone to becoming substrates for siRNAs production and triggering PTGS. These results also suggest that P-bodies act as a first layer of defense against defective RNAs, preventing or limiting the unintended entry of dysfunctional RNAs into the PTGS pathway, thus circumventing the production of rogue siRNAs that could potentially lead to unintended functional RNAs degradation.

P-bodies and siRNA-bodies form dynamically linked structures in the cytoplasm

The functional interplay between decapping and PTGS raises the question of where in the cell these two pathways are interacting. To this end, we examined the subcellular localization of decapping and PTGS proteins by expressing translational fusions to fluorescent reporters in Nicotiana benthamiana leaves. Coherent with previous studies (6,13,18–19), we observed that DCP1, DCP2, VCS, XRN4, RDR6 and SGS3 fusions accumulated in cytoplasmic foci of two types: RDR6 and SGS3 co-localized in siRNA-bodies, whereas DCP1, DCP2, VCS and XRN4 marked P-bodies (Figure 4A). The accumulation of DCP1, DCP2, XRN4, SGS3 and RDR6 in cytoplasmic foci was confirmed in the root meristem cells of stable Arabidopsis transgenic lines expressing GFP-tagged versions of these proteins (Supplemental Figure S8). It was previously reported that P-bodies, marked by DCP1, are distinct from siRNA-bodies, marked by SGS3 (18,19). We confirmed the existence of these separate foci, although we observed that 50% of the P-bodies and the siRNA-bodies were juxtaposed (Figure 4B and C). The majority of the clusters contained one spot of each class, but association of 3 foci also was observed (Figure 4B and C). We defined as juxtaposed, foci that had no measurable free space between them. The distance from center to center between two adjacent P- and siRNA-bodies was 2.5 ± 1.35 μm (average ± SD, 89 measurements in eight independent experiments). The high degree of variation in this measure is linked to the heterogeneity in the size of the foci varying from about 0.2 μm to 9 μm (mean ± SD of 1.87 ± 1.89 μm, n = 367). To rule out eventual artifacts due to expression in an heterologous system, we examined the expression of GFP:SGS3 and RFP:DCP1 fusion proteins in Arabidopsis cotyledons. We observed that, although distinct, the siRNA- and the P-bodies tended to juxtapose (Figure 4D). This association was further confirmed by whole-mount immunolocalization of endogenous (untagged) SGS3 in transgenic Arabidopsis lines expressing the RFP:DCP1 fusion protein (Figure 4E).

To determine if the observed spatial association between P- and siRNA-bodies reflected transient or more stable association between the two foci, we performed live imaging analysis in Nicotiana benthamiana leaves. We observed that the two bodies often remained associated as they moved through the cell (Figure 5A and Supplemental Movie S1), indicating an active spatial connection (velocity: 1.12 ± 0.8 μm.s⁻¹, mean ± SD). We observed several associations and dissociations taking place in short lapses of time between P- and siRNA-bodies (Supplemental Movie S2). To characterize this movement further, we monitored P- and siRNA-bodies displacement in the presence of oryzalin, a drug that depolymerizes microtubules, or latrunculinB (latB), a toxin that binds actin monomers and disrupts the actin filaments of the cytoskeleton (36). As previously published, samples were treated for 2 h treatment with 10 μM of oryzalin or latrunculinB (37,38). Although mobility of P-bodies has been shown to be dependent on microtubule in animal cells (39), we observed that oryzalin treatment had no effect on the mobility of P- and siRNA-bodies. In contrast, 10 μM latB treatment blocked the movement of both bodies (Figure 5B), indicating that in plants the movement of P- and siRNA-bodies depended on actin filaments but not on microtubules.