Abstract
Defects in RNA maturation and RNA decay factors may generate substrates for the RNA interference machinery. This phenomenon was observed in plants where mutations in some RNA-related factors lead to the production of RNA-quality control small interfering RNAs and several mutants show enhanced silencing of reporter transgenes. To assess the potential of RNAi activation on endogenous transcripts, we sequenced small RNAs from a set of Arabidopsis thaliana mutants with defects in various RNA metabolism pathways. We observed a global production of siRNAs caused by inefficient pre-mRNA cleavage and polyadenylation leading to read-through transcription into downstream antisense genes. In addition, in the lsm1a lsm1b double mutant, we identified NIA1, SMXL5, and several miRNA-targeted mRNAs as producing siRNAs, a group of transcripts suggested being especially sensitive to deficiencies in RNA metabolism. However, in most cases, RNA metabolism perturbations do not lead to the widespread production of siRNA derived from mRNA molecules. This observation is contrary to multiple studies based on reporter transgenes and suggests that only a very high accumulation of defective mRNA species caused by specific mutations or substantial RNA processing defects trigger RNAi pathways.
Similar content being viewed by others
Introduction
Small RNA interference (RNAi) pathways are conserved mechanisms implicated in the defence against invading nucleic acids and in the regulation of gene expression. These processes are extensively studied in a model plant species Arabidopsis thaliana, where different small RNA pathways can be distinguished based on the nature of small RNA (sRNA) precursors, processing enzymes, and effector complexes. Major plant small RNA classes include microRNA (miRNA), small interfering RNA (siRNA), trans-acting siRNA (ta-siRNA), natural antisense transcripts siRNA (nat-siRNA), and heterochromatic siRNA (het-siRNA) (reviewed in1). We recently described two mutants involved in mRNA metabolism with elevated production of siRNAs from protein-coding genes2,3. Here, we tried to determine if such a phenomenon could also be observed in a broader set of RNA-related mutants.
Arabidopsis genome is relatively small and densely packed with genes, which makes regulation of Pol II transcription especially important. We and others showed that XRN3 5′–3′ exoribonuclease is a crucial factor for Pol II transcription termination and defects in this process give rise to read-through transcripts2,4. In the case these transcripts are antisense to mRNAs, they may trigger double-stranded RNA (dsRNA) formation and, consequently, siRNA synthesis2,5. Transcripts arising from defective maturation or impaired degradation are considered aberrant. Their features include, but are not limited to, lack of cap structure, as is the case for read-through transcripts in the xrn3 mutant, lack or shortened poly(A) tail, defective pre-mRNA 3′ end formation and incomplete splicing6. It has been postulated that mRNA surveillance and RNAi pathways compete in removing such faulty RNA molecules6. Several studies have demonstrated the existence of a new class of sRNAs called rqc-siRNA (RNA quality control small interfering RNA)7 or ct-siRNA (coding transcript-derived siRNAs)8, which are generated mainly due to defects in RNA degradation machinery from mRNAs that do not normally produce siRNAs. In line with this, our recent work demonstrated a strong accumulation of siRNAs in the mutant of the DXO1 gene involved in mRNA cap surveillance3.
Mutations in factors directly engaged in mRNA decapping3,7,9,10,11 or 5′–3′ and 3′-5′ RNA degradation8,12,13,14,15,16,17,18 initiate production of siRNAs from mRNAs mediated by RNA-dependent RNA polymerase (RDR) and Dicer-like proteins, mainly but not exclusively by RDR6 and DCL2/4 (reviewed in6,19). It has been demonstrated that severe phenotypes of decapping dcp2 and vcs mutants and the RNA degradation double mutant ski2 xrn4 are partially suppressed by elimination of siRNA production from several hundreds of mRNAs7,8,18, which underscores the significance of this new class of siRNAs. These RNA degradation factors and many others have also been identified as suppressors of transgene silencing mediated by siRNA7,16,20,21,22,23,24,25,26. However, it seems that transgene high expression renders them more sensitive to any disturbances in RNA metabolism than most endogenous mRNAs. Except for mRNA decapping, biogenesis of rqc-siRNA from endogenous mRNAs is strongly increased only when both 5′ and 3′ cytoplasmic RNA degradation pathways are defective, while a disturbance in either of these mechanisms results in an enhanced accumulation of small RNAs from a limited number of loci or reporter transgenes8,12,13,14,15. This is also true in the Ler accession of A. thaliana, which has an active form of SOV (SUPPRESSOR OF VCS) gene, encoding a homolog of DIS3L2 3′–5′ cytoplasmic exonuclease9,27. Notably, although a point mutation in the RRP4 gene of the exosome core complex subunit increases transgene silencing28, high-throughput sequencing of small RNAs from the RRP4 RNAi line showed a minor impact on the production of siRNAs from mRNAs29. Also, aberrant mRNA biogenesis may lead to the production of new siRNAs and the increase of transgene silencing. Enhancement of transgene silencing was observed in the case of defects in mRNA splicing30,31, cleavage and polyadenylation30,32,33, transcription regulation16 and termination5. A whole plethora of RNA maturation and degradation mutants showing stronger transgene silencing led to the hypothesis that defects in mRNA metabolism give rise to accumulation of aberrant mRNA species and production of unwanted siRNAs, with a potentially toxic effect on cellular functions6.
Here, we set out to validate this assumption by sequencing small RNA libraries from a set of mutants, which potentially are candidates for rqc-siRNA production. Our analysis of cstf64-2 transcription termination mutant showed a more robust siRNA accumulation from about a thousand mRNAs, mainly due to read-through transcription from downstream convergent genes. However, sequencing of small RNAs from mutants in other factors involved in pre-mRNA 3′ end formation (rsr1-2, fy-2), mRNA poly(A) tail control (ahg1-2, ccr4a, pab2 pab4), NMD (upf1-5, upf3-1), RNA 5′–3′ decay (lsm1a lsm1b) and splicing (ncb-4) revealed that the majority showed only weak or no signs of increased siRNA production from mRNAs. These results suggest that different RNA degradation pathways may substitute for each other or that the rqc-siRNA generating mechanisms are limited only to a specific group of aberrant transcripts.
Results and discussion
Strong defects in mRNA cleavage and polyadenylation induce siRNAs production
Our previous work showed that Pol II transcription termination defects in xrn3-8 mutant lead to small RNA production sourcing from convergent protein-coding gene pairs2. The crucial step of Pol II transcription termination is pre-mRNA 3′ end processing conducted by the multisubunit cleavage and polyadenylation complex (CPA). One of its components is the CstF64 protein responsible for the recognition of GU-rich elements in nascent RNAs34. We performed small RNA profiling of the cst64-2 mutant, which exhibits delayed Pol II release due to dysfunction of the cleavage and polyadenylation machinery. Mutant plants are sterile and show characteristic morphology35, allowing for the selection of 21 day-old cstf64-2 homozygous plants grown on soil along with Col-0 wild-type. Since widespread defects in Pol II transcription termination due to the cstf64-2 mutation have not been directly demonstrated in Arabidopsis, we checked the level of read-through transcripts downstream of several genes that we previously reported to have deficient termination in the xrn3-8 mutant2. All tested regions showed significant up-regulation of transcripts that was often unrelated to changes in the expression of upstream mRNAs (Fig. 1A). This analysis confirmed the role of CSTF64 in Pol II transcription termination and showed that the production of read-through transcripts in the mutant does not always affect the expression of parental genes.
Small RNA libraries were prepared in triplicates and sequenced to 10 million reads per sample (Table S1). Reads were mapped and counted in exons of genes from Araport11 annotation36, and counts were used for differential analysis using DESeq237. In such an approach, we focused on the identification of genes with siRNA production as a way to identify the role of their properties in this process. We identified 1172 genes with significantly elevated levels of small RNAs in the mutant (FDR < 0.05, log2FC > 1; Supplementary Dataset 1). Since the CPA complex is mainly involved in the 3′ end formation of pre-mRNAs, we excluded pseudogenes, transposons and other non-protein-coding genomic features, leaving 1102 genes for further analysis. As many as 76% of genes with up-regulated siRNA production in cstf64-2 plants have antisense convergent genes at a distance of less than 400 bp, which is twice as much as in the random sampling control (Fig. 1B). This observation supports our hypothesis that defects in transcription termination from downstream convergent genes create antisense read-through transcripts. Such RNA species may form double-stranded RNAs by pairing with mRNAs, which triggers siRNA production. This outcome mainly applies to highly expressed gene pairs since low-expression genes would not generate abundant read-through transcripts that could initiate siRNA synthesis. Examples of such gene pairs are shown in Fig. 1C–G. Consistently, genes with increased siRNAs show high overlap with those identified in the xrn3-8 transcription termination-deficient mutant2 and low with mutants involved in cytoplasmic mRNA degradation7,8 (Supplementary Fig. 1A–C). Contrary to our expectations, we do not observe an enrichment of small RNAs exclusively at 3′ ends of affected genes (Fig. 1H). This suggests that initial dsRNA formation close to the 3′ end of the transcript may trigger siRNA production from the whole molecule, which requires RNA-dependent RNA (RDR) polymerase activity. Another possibility is that defective mRNA 3′ end formation is a feature that recruits RDR independently of antisense transcripts pairing. Such a mechanism has been demonstrated earlier for several transgene reporters32,33. In our data, it is supported by identification of genes that lack downstream convergent partners but still generate read-through transcripts and produce siRNAs (Fig. 1B,I).
Albeit widespread among genes, the accumulation of novel siRNAs appears to be relatively moderate in terms of the absolute number of sRNAs (Fig. 1H). Small RNAs from genes showing an increase in their level in cstf64-2 represent 1% of mutant small RNA libraries and 0.15% of wild-type libraries (Fig. 1J). Such a limited increase may explain why siRNA accumulation does not lead to significant changes in the expression of their source genes, as assayed for ten mRNAs by RT-qPCR (Fig. 2A). However, we cannot exclude that some genes that have not been tested respond to these siRNAs. The length distribution of small RNAs from genes with siRNA increase in cstf64-2 plants has two prominent peaks—21 and 24 nt (Fig. 2B). Since it is assumed that biogenesis of 24 nt long siRNAs takes place in the nucleus and this class is involved in transcriptional gene silencing1, novel siRNAs in cstf64-2 plants could be produced mainly in this cellular compartment. It could also explain the lack or only weak impact of sRNA increase on the level of cytoplasmic mRNAs. Finally, GO term enrichment analysis for genes producing more siRNAs in the cstf64-2 mutant showed that a significant number are implicated in mRNA metabolism (Fig. 2C). This observation suggests a self-regulatory mechanism that affects gene expression when read-through transcripts accumulate in physiological conditions, as recently described for drought stress in Arabidopsis38. In plants, siRNAs derived from natural antisense transcripts have been described to be involved in the regulation of expression of their source genes39,40. It is possible that the controlled production of read-through transcripts followed by siRNA accumulation may affect the pattern of gene expression on the local scale or genome-wide.
The majority of tested mutants show limited production of mRNA-derived siRNAs
We chose the CSTF64 gene for our small RNA analysis as its mutation shows a clear developmental phenotype35, suggesting the importance of this gene for plant RNA metabolism. Two other components of the 3′ end processing machinery have been reported to act as suppressors of transgene silencing by siRNAs, CSTF64 paralog ESP1/RSR1 and CPSF100 component of the CPA complex30. Therefore, we decided to extend our analysis to include the viable rsr1-2 knock-out mutant41 and the hypomorphic mutant of the FY gene encoding a protein that interacts with CPSF100 in the CPA complex30. Since mRNA 3′ end formation and polyadenylation efficiency may affect its fate, we decided to add mutants of factors potentially involved in poly(A) tail length control. Importantly, two Arabidopsis putative deadenylating enzymes PARN/AHG2 and CCR4a act as suppressors of transgene silencing dependent on RDR6 and SGS3 (SUPPRESSOR OF GENE SILENCING3) and co-localise with siRNA-bodies24. However, it should be emphasised that data on the cellular localisation of AHG2 is not consistent with its function in mitochondria42. In turn, Poly(A)-binding (PAB) proteins regulate mRNA deadenylation and translation, contributing to mRNA cytoplasmic turnover and stability43. Out of eight Arabidopsis PAB proteins, PAB2, PAB4 and PAB8 are expressed in vegetative tissues. While triple mutants are not viable, the combination of pab2 and pab4 knock-out mutations shows the strongest phenotypic defects44.
Small RNA libraries were prepared in biological triplicates from 14 day-old Col-0 and ahg2-1, ccr4a, rsr1-2, fy-2 and pab2 pab4 mutant seedlings grown on MS solid medium, sequenced (Table S1) and analysed as described above. We identified only from 28 (crr4a and fy-2) to 273 (ahg2-1) genes with significantly elevated levels of small RNAs (FDR < 0.05, log2FC > 1; Fig. 3A, Supplementary Dataset 1). However, these numbers dropped to 19 and 102, respectively, when only protein-coding genes were considered (Fig. 3A). This was caused by a large fraction of affected genes belonging to pseudogenes and transposable elements. Since the analysed mutations are predicted to affect mRNAs, these effects are most likely indirect. Due to a small number of genes with changes in sRNA level, principal component analysis (PCA) showed poor separation of genotypes on the plot (Supplementary Fig. S2A).
For rsr1-2 and fy-2 mutants predicted to be involved in pre-mRNA cleavage and polyadenylation, we checked the level of read-through transcripts downstream of five genes with prominent defects in cstf64-2 and xrn3-8 mutants (Supplementary Fig. S2B). All showed a much weaker or no increase in read-through transcription. Lack of termination defect may explain the very limited accumulation of small RNAs in these mutants. Still, five genes with siRNA increase in the rsr1-2 mutant have antisense partners and were also found in the cstf64-2 analysis (Supplementary Fig. S2C–G), showing that effects in the rsr1-2 mutant are similar but more subtle or limited to a specific set of genes.
A very limited number of endogenous mRNAs producing siRNAs in mutants that have been shown to act as suppressors of transgene silencing was unexpected. Therefore, we set out to extend our analysis again by including mutants in factors affecting other aspects of mRNA metabolism. UPF1 and UPF3 proteins are key factors of Nonsense-mediated decay (NMD), which is an important translation-dependent mRNA surveillance mechanism that eliminates aberrant transcripts containing premature translation stop codons and also contributes to the regulation of gene expression (reviewed in45,46). Interestingly, upf1 and upf3 mutants have been shown to enhance transgene silencing through the production of transgene-derived siRNAs. Moreover, UPF1 protein co-localises with cytoplasmic siRNA-bodies, which represent sites of siRNA production24. COILIN is a conserved structural protein required for the assembly of peri-nucleolar organelles called Cajal bodies, which in different organisms are implicated in the formation and function of small nuclear (snRNA), small nucleolar (snoRNA) and small Cajal body-specific RNAs (scaRNA). Consequently, Cajal bodies have a role in mRNA splicing and rRNA maturation (reviewed in47). Although in plants composition and role of Cajal bodies are much less studied, they contain COILIN and several components of snRNP complexes including SmD3, SmB, U2A' and U2B''48,49,50. Interestingly, splicing defects have been shown to enhance siRNA production for transgene reporter transcripts30,51 and SmD1, which is involved in splicing and has been reported to play a role in transgene silencing31. Finally, Arabidopsis LSM1 protein is a subunit of the cytoplasmic heptameric LSM1-7 complex engaged in mRNA decapping and degradation52,53. In yeast and Metazoa, mRNA deadenylation and oligouridylation attract the LSM1-7 complex, which recruits decapping proteins that trigger XRN1-mediated RNA degradation from the 5′ end and at the same time protects RNA 3′ end against degradation by the exosome (reviewed in54). Although information concerning the role of LSM proteins in mechanisms engaging small RNAs is lacking, XRN4, a plant homologue of yeast Xrn1, acts as a suppressor of silencing by small RNAs, and its absence leads to the accumulation of decapped mRNAs that are targeted for siRNA production8,20,21,25.
Small RNA libraries from wild-type and lsm1a lsm1b, upf1-5, upf3-1 and ncb-4 (coilin) mutant 14 day-old MS-grown seedlings were prepared, sequenced (Table S1) and analysed as for the previous set of mutants. PCA plot showed that analysed libraries formed separate coherent groups (Supplementary Fig. S3A), however, the differences between genotypes may be modest as the variance in both PC1 and PC2 is small. Consistently, as with the previous set of mutants, we found a limited number of protein-coding genes with increased siRNA production (Fig. 3B). Several works described the accumulation of rqc-siRNAs in mutants with perturbed mRNA degradation. It was proposed that genes producing this class of siRNAs were direct substrates of the mutated factors. We were interested whether any of the protein-coding genes showing an increase in sRNA level in sequenced mutants had the properties of rqc-siRNA source, as it would identify them as the target of the analysed factors.
It has been proposed that small RNAs produced from mRNAs with a high expression and turnover represent degradation by-products as they are generated only from the sense strand and show neither the length distribution nor the first nucleotide bias observed for canonical sRNAs55. Synthesis of the sense-strand sRNAs does not involve RNA-dependent polymerases, which are required for rqc-siRNA production7,8, and such small RNAs cannot post-transcriptionally down-regulate the expression of their source mRNAs. Therefore, we tested the strandedness of small RNAs from protein-coding genes based on the criterion that the ratio of small RNAs mapping to both strands should be in the range of 0.25–4 for double-strand-derived siRNAs (as described in7). Most of the tested mutants showed a significant fraction of genes with small RNAs produced only from one strand (21–74%), except for the cstf64-2 mutant (Fig. 3C). We analysed changes in mRNA levels for ten of these genes using RT-qPCR, and most of them were up-regulated (Supplementary Fig. S3B). It is consistent with the notion that the accumulation of small RNAs produced exclusively from mRNA strand reflects changes in gene expression. Alternatively, in the case of genes with unaltered expression, mRNAs may be more efficiently converted into small RNAs due to their enhanced degradation. Both scenarios are probably true for several potential NMD substrates, including SMG7 as well as other mRNAs with an intron in the 3′ UTR (for example, AT3G03710, AT3G27906 and AT5G54390) or containing an upstream ORF (for example, AT1G36730, AT3G18000)46 as they were identified among the genes producing more sRNAs in upf3 or upf1 mutants but only from their coding strand (Supplementary Dataset 1).
Another criterion for a genuine rqc-siRNA source is that siRNA accumulation should be attributed to defects in mRNA metabolism and should not spread from pre-existing genomic sRNA production hotspots (Supplementary Fig. S3C–F). Such hotspot genes usually have high levels of siRNA production also in the wild-type and consequently high levels of DNA methylation, which is maintained by sRNA-dependent pathways1. We observed that genes with accumulation of sRNAs in mutants, with the exception of cstf64-2, tend to produce more small RNAs also in their wild-type controls compared to the average value of all protein-coding genes (Supplementary Fig. 4A–C). Consequently, these small RNA hotspots may have elevated DNA methylation. We used Col-0 DNA methylation data56 to check the methylated cytosine fraction in the affected genes and noticed that it is often much higher than the average for all protein-coding genes. This effect was not observed for dcp2, vcs7 and ski2 xrn4 mutants8, which represent the golden standard of mutants producing rqc-siRNAs (Supplementary Fig. 4D). We cannot exclude that genes with high DNA methylation are the true source of rqc-siRNAs, but they would constitute atypical cases as they can be silenced even in wild-type plants. The presence of methylation and siRNA production hotspots in different parts of the gene can also be explained by spreading of small RNA synthesis57. Therefore, we decided to filter out genes with high level of DNA methylation using an arbitrarily chosen criterion of methylation level higher than 20% of all cytosines in a given gene, which is six times more than the average for protein-coding genes. Using both the strandedness and hotspot criteria, most libraries showed 53–88% dropouts, except for the cstf64-2 mutant with only 10% of the excluded genes (Fig. 3C).
Such filtering left a minimal number of mRNAs potentially generating rqc-siRNAs. It has been suggested that rqc-siRNAs are exclusively 21–22 nt long7,8. However, protein-coding genes with siRNA accumulation show a prominent peak for siRNAs of 24 nt in most of the tested mutants (Supplementary Fig. 5A). Assuming that restricting the analysis to this class of siRNAs will enhance our results, we repeated the differential analysis for 21–22 nt long small RNAs (Supplementary Dataset 2). As a control, we used 24 nt siRNAs that are implicated in transcriptional gene silencing (Supplementary Dataset 3). The results obtained after eliminating non-protein-coding genes, genes producing small RNAs from only one strand, and from genomic siRNA hotspots (Supplementary Fig. S5C–E) show high overlap with those for small RNAs before size selection (Supplementary Fig. S5F). Surprisingly, restricting the analysis to 21–22 nt small RNAs had only a minor effect on the identification of rqc-siRNA-producing genes.
It appears that there may be only a few examples of genuine rqc-siRNAs produced in the analysed mutants. However, each of the possible candidates requires careful examination to exclude the existence of adjacent siRNA production hotspots or having atypical small RNA profiles (Supplementary Fig. S3C–F). For example, the fy-2 mutation is characterised by elevated levels of siRNAs derived from the FY gene itself, but generated only from the gene fragment downstream of the T-DNA insertion, pointing to the possibility that the aberrant transcript originates from the insert (Supplementary Fig. S6A). Moreover, the fy-2 mutation causes accumulation of rqc-siRNAs from the DXO1 gene (Supplementary Fig. S6B) encoding the plant homolog of human DXO, which is involved in mRNA cap surveillance and mRNA degradation3,10,11. Lack of Arabidopsis DXO1 results in a strong up-regulation of rqc-siRNAs from several hundred genes3,11. However, the expression of DXO1 is not altered in fy-2 plants (Supplementary Fig. S6C), which accordingly do not show strong phenotypes observed for the dxo1-2 mutant. Therefore, the functional significance of this observation is currently unclear.
It is apparent that in most of the tested mutants only a small fraction of genes become the source of rqc-siRNAs. This can be considered surprising as selected mutants show an increase in transgene silencing or are involved in pathways whose dysfunctions have been suggested to induce rqc-siRNAs production. The enhanced generation of transgene-derived siRNAs is likely due to the particularly high expression of transgenes, making them susceptible to the production of a wide variety of defective transcripts. Their degradation relies on quality control and decay pathways, but they can easily avoid degradation and trigger siRNA synthesis even in wild-type plants. In turn, endogenous aberrant or superfluous transcripts, including those highly expressed, most likely evolved protective mechanisms to avoid such scenarios. They are quickly recognised by specialised quality control pathways and efficiently eliminated without activating RNA interference.
The limited production of rqc-siRNAs in analysed mutants can be explained by case-specific circumstances. The most obvious situation is functional redundancy due to the presence of closely related paralogs that replace the activities missing in individual mutants: PAB8 in the pab2 pab4 mutant, CCR4b-g in the ccr4a mutant and CSTF64 in the rsr1-2 mutant30,44,58. In the latter case, however, RSR1/ESP1 may have a specific role in the CPA complex since it lacks the RRM domain present in CSTF6430. Redundancy can also be provided by closely related proteins or complexes with similar functions, as is the case of CCR4-NOT, PAN2-PAN3 and PARN/AHG2 deadenylases. Lack of rqc-siRNA production in ahg2 mutant may also be attributed to the mitochondrial42 rather than cytoplasmic functions of AHG224. In addition, fy-2 and upf1-5 mutations are hypomorphic and were used since viable knockout lines are not available. Their remaining functionality may be sufficient to suppress rqc-siRNA production. Finally, specific features of transcripts generated as a result of defects in mRNA metabolism may not trigger rqc-siRNA production. This may be true for NMD substrates that accumulate in upf1-5 and upf3-1 mutants, and mRNAs with defective maturation in ncb-4 and fy-2 mutants involved in splicing and 3′ end processing, respectively. Nevertheless, we found that some genes may be the source of rqc-siRNA production in lsm1a lsm1b double mutant.
Lack of LSM1 cause accumulation of rqc-siRNAs from known mRNA sources
We observed a very high accumulation of siRNAs generated from the NIA1 locus in the lsm1a lsm1b mutant (Fig. 4A). Similar siRNAs were detected in the ski2 xrn4 double mutant and reported to be responsible for down-regulation of NIA1 and NIA2 (NIA1 paralog) mRNAs and some of the observed mutant phenotypes8,18. In contrast, in the lsm1a lsm1b mutant, the NIA2 gene produced only moderately more siRNAs (Fig. 4B), and the change in their level was not statistically significant when small RNAs of all lengths were taken into account, but was prominent for 21–22 nt small RNAs (Supplementary Dataset 2). As expected from the increase of small RNAs in lsm1a lsm1b plants, we observed down-regulation of NIA1 and NIA2 mRNAs (Fig. 4D), but these plants did not show such severe morphological phenotypes as the xrn4 ski2 double mutant8,53. Another gene reported to accumulate siRNAs with an important outcome for plant physiology is SMXL5. High production of siRNAs from SMXL4 and SMXL5 induces their silencing and causes over-accumulation of carbohydrates and defects in phloem transport18,59. We observed a moderate accumulation of small RNAs from SMXL5, but not from SMXL4, in the lsm1a lsm1b mutant, consequently with no impact on SMXL5 expression (Fig. 4C,D). The different outcomes of rqc-siRNA production on the level of their source mRNA can be explained by the extent of their accumulation. A high amount of siRNA would lead to reduced gene expression (NIA1), while a moderate increase in siRNA would not affect mRNA levels (SMXL5). Interestingly, when used as a transgenic reporter, NIA2 induces silencing of its endo- and exogenous copies60, whereas SMXL5 is one of the six protein-coding genes which produce more siRNAs when DCL4 is mutated59. These observations suggest that a low number of genes accumulating rqc-siRNAs in the lsm1a lsm1b mutant may represent those especially sensitive to induction of rqc-siRNA biogenesis. Recent work has shown the potential physiological role of 22 nt siRNAs produced from NIA1/2 mRNAs as they are induced not only by xrn4 ski2 mutations but also by nitrogen starvation and ABA treatment18.
In the lsm1a lsm1b mutant, we also observed a modest siRNAs increase for AGO1 and several other miRNA targets that produce siRNAs also in wild-type plants (Fig. 4E–K), but the level of these mRNAs was unaffected by the enhanced siRNA production (Supplementary Fig. S7A). On the other hand, miR168 and miR472, but not miR393 and miR399, which are predicted to target these mRNAs, accumulated in the lsm1a lsm1b mutant (Supplementary Fig. S7B). As LSM1 potentially cooperates with XRN4 in plant 5′–3′ mRNA decay, siRNAs increase from these genes may reflect the role of XRN4 in removing 3′ fragments of mRNAs targeted by miRNAs21 and is consistent with siRNA accumulation from miRNA targets in the ski2 xrn4 double mutant8. However, it is not clear why lack of LSM1 affects a very limited number of miRNA targets and why the levels of only some miRNAs change in the mutant.
Despite the low number of genes generating siRNAs in the absence of the decapping activator LSM1, some of them seem to match the requirements of the rqc-siRNA source, including the length of 21–22 nt of produced sRNAs (Supplementary Fig. S7C). A limited accumulation of rqc-siRNAs resembles single xrn4, cer7 and ski2 mutants12,13,14, suggesting that only mutations with a strong effect on RNA decay, i.e. decapping or simultaneous inhibition of 5′–3′ and 3′–5′ degradation, are capable of triggering rqc-siRNA production. Moreover, these rqc-siRNAs are mostly limited to known rqc-siRNA hotspots.
Conclusions
Shortly after the discovery of plant small interfering RNAs61, their accumulation under perturbed RNA degradation conditions was reported20. Moreover, it has been observed that many defects in mRNA maturation or degradation can lead to the accumulation of aberrant endogenous or transgenic transcripts that are recognised by RNAi pathways23,24,25,30,31,62. More recently, a new class called RNA quality control siRNA7 or coding transcript-derived siRNA8 was described. However, other experiments have shown that in most cases the number of endogenous protein-coding genes that accumulate rqc-siRNAs is surprisingly low12,13,14. Our work supports these studies by providing evidence based on additional nine mutants with defects in different RNA metabolism pathways. With the exception of the cstf64 mutation, the remaining mutants showed no or only limited rqc-siRNA synthesis. This strongly suggests that results from transgene-based studies are incompatible with quality control processes that affect endogenous mRNAs. From the set of mutants analysed in this study, only strong impairment of pre-mRNA 3′ end processing and Pol II transcription termination in the cstf64-2 mutant led to genome-wide production of rqc-siRNAs. Notably, these rqc-siRNAs may also be produced in some physiological conditions as the accumulation of read-through transcripts was reported to be strongly enhanced under drought stress38. It was previously proposed that plants evolved multiple RNA quality control pathways, protecting them from the production of unwanted siRNAs. Such mechanisms would allow the distinction between endogenous and exogenous RNA species. The limited production of rqc-siRNAs in different RNA-related mutants supports this hypothesis.
Methods
Wild-type Col-0 and mutant lines: ahg2-163, ccr4a (SAIL_784_A07)24, pab2 pab4 (SALK_026293 SALK_113383)44 (these three lines were a kind gift of Hervé Vaucheret, INRA Centre de Versailles-Grignon, France), cstf64-2 (SAIL_794_G11, kind gift of Caroline Dean, John Innes Centre, UK)35, fy-264 (kind gift of Szymon Swiezewski, IBB, Poland), lsm1a lsm1b (SALK_106536 SAIL_756_305, obtained in our earlier work)53, ncb-449 (kind gift of Peter Shaw, John Innes Centre, UK), rsr1-2 (SALK_078793, kind gift of Dietmar Funck, University of Konstanz, Germany)41, upf1-5 and upf3-1 (SALK_112922, SALK_025175, kind gift of Brendan Davies, University of Leeds, UK)65 were used in this study. We had obtained permission to grow genetically modified plants and we handled them in accordance with the institutional, and national guidelines and legislation. Seeds were sown on MS plates and stratified for 2 days at 4 °C. Plants were grown at constant 21 °C under long-day conditions and harvested at the age of 14 days. The cstf64-2 mutant was grown in soil at constant 21 °C under long-day conditions, and homozygous 21-day-old plants were harvested based on their phenotype. Wild-type Col-0 plants for this experimental set were grown together at the same time and condition but their small RNA sequencing analysis results are part of the earlier GEO submission (GSE99600). RNA was isolated using the Tri Reagent method and analysed with Bioanalyzer. All library preparation steps using NEB Next Small RNA Library Prep Set for Illumina, including PAGE selection of small RNAs and sequencing with Illumina HiSeq4000 in 50 bp single-end mode, were conducted by BGI Sequencing Services. Obtained fastq files were quality checked using FastQC (v0.10.1; http://www.bioinformatics.babraham.ac.uk/projects/fastqc/) with all the replicates showing high-quality RNA-seq data. Adaptor sequences (Illumina Small RNA Adapter2 TCGTATGCCGTCTTCTGCTTGT) were removed using cutadapt (v1.9.dev6; http://cutadapt.readthedocs.io/en/stable/guide.html), and reads were quality trimmed with sickle se (v0.940; https://github.com/najoshi/sickle) with the following command-line parameters: -t illumina -q 20 -l 15. Reads were mapped to the TAIR10 A. thaliana genome from Ensembl66 (release v29) using bowtie67 (v1.0.0) with the following command-line parameters: -phred33 -v 0 -k 10 -m 10. Mapped reads were sorted using samtools sort (v1.1)68, counted with HTseq-count69 (v0.6.0) with or without respect to the strand and using exon features from Araport11 gene annotation release 20160436. In that annotation, exon stands for any gene fragment (both protein-coding and non-protein-coding) included in a mature RNA molecule. Differential expression and PCA was performed using DESeq237 (v1.8.2) R (v3.2.2) package with parameter alpha = 0.05. Genes with FDR < 0.05 and absolute log2FC > 1 were considered significantly changed. Analysis of siRNAs of different lengths was performed on reads grouped according to their length using reformat.sh with parameters minlength and maxlength from bbmap v35.x (https://sourceforge.net/projects/bbmap/). For DNA methylation analysis, we extracted the number of methylated cytosines for each gene using an R script and published data for Col-0 ecotype56; GSM1085222). Alignment coverage graphs were calculated with genomeCoverageBed from bedtools70 (v2.17.0) for all the alignment files with normalisation to the number of reads and were converted to bigwig format with bedGraphToBigWig (v4) from the UCSC Genome Browser application binaries collection (http://hgdownload.cse.ucsc.edu/admin/exe/linux.x86_64/) for visualisation in Integrated Genome Browser71. GO enrichment analysis was performed using g:Profiler (https://biit.cs.ut.ee/gprofiler/gost) with FDR < 0.001 and enrichment > 1.5. RT-qPCR was carried out on 2 µg of total RNA following DNase I digestion (TURBO DNase, Thermo Fisher Scientific) with Random Primers and SuperScript III Reverse Transcriptase (Thermo Fisher Scientific). Quantitative PCR was performed using SYBR Green I Master Mix (Roche) using the Roche qPCR platform (LightCycler 480). Results were normalised to UBC9 or ACT2 mRNA. Primers sequences used for qPCR are listed in Supplementary Table S2.
Data availability
The data presented in this study are openly available in the GEO database, reference numbers GSE169171 and GSE99600 (Col-0 wild-type for cstf64-2 mutant).
References
Borges, F. & Martienssen, R. A. The expanding world of small RNAs in plants. Nat. Rev. Mol. Cell Biol. 16, 727–741 (2015).
Krzyszton, M. et al. Defective XRN 3-mediated transcription termination in Arabidopsis affects the expression of protein-coding genes. Plant J. 93, 1017–1031 (2018).
Kwasnik, A. et al. Arabidopsis DXO1 links RNA turnover and chloroplast function independently of its enzymatic activity. Nucleic Acids Res. 47, 4751–4764 (2019).
Crisp, P. A. et al. RNA polymerase II read-through promotes expression of neighboring genes in SAL1-PAP-XRN retrograde signaling. Plant Physiol. 178, 1614–1630 (2018).
Parent, J.-S. et al. Post-transcriptional gene silencing triggered by sense transgenes involves uncapped antisense RNA and differs from silencing intentionally triggered by antisense transgenes. Nucleic Acids Res. 43, 8464–8475 (2015).
Liu, L. & Chen, X. RNA quality control as a key to suppressing RNA silencing of endogenous genes in plants. Mol. Plant 9, 826–836 (2016).
Martínez de Alba, A. E. et al. In plants, decapping prevents RDR6-dependent production of small interfering RNAs from endogenous mRNAs. Nucleic Acids Res. 43, 2902–2913 (2015).
Zhang, X. et al. Suppression of endogenous gene silencing by bidirectional cytoplasmic RNA decay in Arabidopsis. Science 348, 120–123 (2015).
Sorenson, R. S., Deshotel, M. J., Johnson, K., Adler, F. R. & Sieburth, L. E. Arabidopsis mRNA decay landscape arises from specialized RNA decay substrates, decapping-mediated feedback, and redundancy. Proc. Natl. Acad. Sci. USA 115, E1485–E1494 (2018).
Pan, S. et al. Arabidopsis DXO1 possesses deNADding and exonuclease activities and its mutation affects defense-related and photosynthetic gene expression. J. Integr. Plant Biol. 62, 967–983 (2020).
Yu, X. et al. Messenger RNA 5′ NAD+ capping is a dynamic regulatory epitranscriptome mark that is required for proper response to abscisic acid in Arabidopsis. Dev. Cell 56, 125-140.e6 (2021).
Gregory, B. D. et al. A link between RNA metabolism and silencing affecting Arabidopsis development. Dev. Cell 14, 854–866 (2008).
Branscheid, A. et al. SKI2 mediates degradation of RISC 5′-cleavage fragments and prevents secondary siRNA production from miRNA targets in Arabidopsis. Nucleic Acids Res. 43, 10975–10988 (2015).
Lam, P. et al. The exosome and trans-acting small interfering RNAs regulate cuticular wax biosynthesis during Arabidopsis inflorescence stem development. Plant Physiol. 167, 323–336 (2015).
Lange, H. et al. RST1 and RIPR connect the cytosolic RNA exosome to the Ski complex in Arabidopsis. Nat. Commun. 10, 3871 (2019).
Li, T. et al. A genetics screen highlights emerging roles for CPL3, RST1 and URT1 in RNA metabolism and silencing. Nat. Plants 5, 539–550 (2019).
Scheer, H. et al. The TUTase URT1 connects decapping activators and prevents the accumulation of excessively deadenylated mRNAs to avoid siRNA biogenesis. Nat. Commun. 12, 1298 (2021).
Wu, H. et al. Plant 22-nt siRNAs mediate translational repression and stress adaptation. Nature 581, 89–93 (2020).
Tsuzuki, M., Motomura, K., Kumakura, N. & Takeda, A. Interconnections between mRNA degradation and RDR-dependent siRNA production in mRNA turnover in plants. J. Plant Res. 130, 211–226 (2017).
Gazzani, S., Lawrenson, T., Woodward, C., Headon, D. & Sablowski, R. A link between mRNA turnover and RNA interference in Arabidopsis. Science 306, 1046–1048 (2004).
Gy, I. et al. Arabidopsis FIERY1, XRN2, and XRN3 are endogenous RNA silencing suppressors. Plant Cell 19, 3451–3461 (2007).
Vogel, F., Hofius, D., Paulus, K. E., Jungkunz, I. & Sonnewald, U. The second face of a known player: Arabidopsis silencing suppressor AtXRN4 acts organ-specifically. New Phytol. 189, 484–493 (2011).
Thran, M., Link, K. & Sonnewald, U. The Arabidopsis DCP2 gene is required for proper mRNA turnover and prevents transgene silencing in Arabidopsis. Plant J. 72, 368–377 (2012).
Moreno, A. B. et al. Cytoplasmic and nuclear quality control and turnover of single-stranded RNA modulate post-transcriptional gene silencing in plants. Nucleic Acids Res. 41, 4699–4708 (2013).
Yu, A. et al. Second-site mutagenesis of a hypomorphic argonaute1 allele identifies SUPERKILLER3 as an endogenous suppressor of transgene posttranscriptional gene silencing. Plant Physiol. 169, 1266–1274 (2015).
Kim, M.-H. et al. Proteasome subunit RPT2a promotes PTGS through repressing RNA quality control in Arabidopsis. Nat. Plants 5, 1273–1282 (2019).
Zhang, W., Murphy, C. & Sieburth, L. E. Conserved RNaseII domain protein functions in cytoplasmic mRNA decay and suppresses Arabidopsis decapping mutant phenotypes. Proc. Natl. Acad. Sci. USA 107, 15981–15985 (2010).
Hématy, K. et al. The zinc-finger protein SOP1 is required for a subset of the nuclear exosome functions in Arabidopsis. PLoS Genet 12, e1005817 (2016).
Shin, J.-H. et al. The role of the Arabidopsis exosome in siRNA–independent silencing of heterochromatic loci. PLoS Genet. 9, e1003411 (2013).
Herr, A. J., Molnar, A., Jones, A. & Baulcombe, D. C. Defective RNA processing enhances RNA silencing and influences flowering of Arabidopsis. Proc. Natl. Acad. Sci. USA 103, 14994–15001 (2006).
Elvira-Matelot, E. et al. The nuclear ribonucleoprotein SmD1 interplays with splicing, RNA quality control, and posttranscriptional gene silencing in Arabidopsis. Plant Cell 28, 426–438 (2016).
Luo, Z. & Chen, Z. Improperly terminated, unpolyadenylated mRNA of sense transgenes is targeted by RDR6-mediated RNA silencing in Arabidopsis. Plant Cell 19, 943–958 (2007).
Nicholson, S. J. & Srivastava, V. Transgene constructs lacking transcription termination signal induce efficient silencing of endogenous targets in Arabidopsis. Mol. Genet. Genomics 282, 319–328 (2009).
Shi, Y. & Manley, J. L. The end of the message: Multiple protein–RNA interactions define the mRNA polyadenylation site. Genes Dev. 29, 889–897 (2015).
Liu, F., Marquardt, S., Lister, C., Swiezewski, S. & Dean, C. Targeted 3’ processing of antisense transcripts triggers Arabidopsis FLC chromatin silencing. Science 327, 94–97 (2010).
Cheng, C. et al. Araport11: A complete reannotation of the Arabidopsis thaliana reference genome. Plant J. 89, 789–804 (2017).
Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and dispersion for RNA-seq data with DESeq2. Genome Biol. 15, 550 (2014).
Sun, H.-X., Li, Y., Niu, Q.-W. & Chua, N.-H. Dehydration stress extends mRNA 3′ untranslated regions with noncoding RNA functions in Arabidopsis. Genome Res. 27, 1427–1436 (2017).
Borsani, O., Zhu, J., Verslues, P. E., Sunkar, R. & Zhu, J.-K. Endogenous siRNAs derived from a pair of natural cis-antisense transcripts regulate salt tolerance in Arabidopsis. Cell 123, 1279–1291 (2005).
Jin, H., Vacic, V., Girke, T., Lonardi, S. & Zhu, J.-K. Small RNAs and the regulation of cis-natural antisense transcripts in Arabidopsis. BMC Mol. Biol. 9, 6 (2008).
Funck, D., Clauß, K., Frommer, W. B. & Hellmann, H. A. The Arabidopsis CstF64-like RSR1/ESP1 protein participates in glucose signaling and flowering time control. Front Plant Sci. 3, 80 (2012).
Hirayama, T. et al. A poly(A)-specific ribonuclease directly regulates the poly(A) status of mitochondrial mRNA in Arabidopsis. Nat. Commun. 4, 2247 (2013).
Zhao, T. et al. Impact of poly(A)-tail G-content on Arabidopsis PAB binding and their role in enhancing translational efficiency. Genome Biol. 20, 189 (2019).
Gallie, D. R. Class II members of the poly(A) binding protein family exhibit distinct functions during Arabidopsis growth and development. Translation 5, e1295129 (2017).
Lykke-Andersen, S. & Jensen, T. H. Nonsense-mediated mRNA decay: An intricate machinery that shapes transcriptomes. Nat. Rev. Mol. Cell Biol. 16, 665–677 (2015).
Shaul, O. Unique aspects of plant nonsense-mediated mRNA decay. Trends Plant Sci. 20, 767–779 (2015).
Love, A. J. et al. Cajal bodies and their role in plant stress and disease responses. RNA Biol. 14, 779–790 (2017).
Lorkovic, Z. J., Hilscher, J. & Barta, A. Use of fluorescent protein tags to study nuclear organization of the spliceosomal machinery in transiently transformed living plant cells. Mol. Biol. Cell 15, 11 (2004).
Collier, S. et al. A distant coilin homologue is required for the formation of Cajal bodies in Arabidopsis. Mol. Biol. Cell 17, 10 (2006).
Li, C. F. et al. An ARGONAUTE4-containing nuclear processing center colocalized with Cajal bodies in Arabidopsis thaliana. Cell 126, 93–106 (2006).
Christie, M., Croft, L. J. & Carroll, B. J. Intron splicing suppresses RNA silencing in Arabidopsis. Plant J. 68, 159–167 (2011).
Perea-Resa, C., Hernández-Verdeja, T., López-Cobollo, R., del Castellano, M. M. & Salinas, J. LSM proteins provide accurate splicing and decay of selected transcripts to ensure normal Arabidopsis development. Plant Cell 24, 4930–4947 (2013).
Golisz, A., Sikorski, P. J., Kruszka, K. & Kufel, J. Arabidopsis thaliana LSM proteins function in mRNA splicing and degradation. Nucleic Acids Res. 41, 6232–6249 (2013).
Mugridge, J. S., Coller, J. & Gross, J. D. Structural and molecular mechanisms for the control of eukaryotic 5′–3′ mRNA decay. Nat. Struct. Mol. Biol. 25, 1077–1085 (2018).
Li, Q., Li, Y., Moose, S. P. & Hudson, M. E. Transposable elements, mRNA expression level and strand-specificity of small RNAs are associated with non-additive inheritance of gene expression in hybrid plants. BMC Plant Biol. 15, 168 (2015).
Schmitz, R. J. et al. Patterns of population epigenomic diversity. Nature 495, 193–198 (2013).
de Felippes, F. F. & Waterhouse, P. M. The whys and wherefores of transitivity in plants. Front Plant Sci. 11, 579376 (2020).
Suzuki, Y., Arae, T., Green, P. J., Yamaguchi, J. & Chiba, Y. AtCCR4a and AtCCR4b are involved in determining the poly(A) length of granule-bound starch synthase 1 transcript and modulating sucrose and starch metabolism in Arabidopsis thaliana. Plant Cell Physiol. 56, 863–874 (2015).
Wu, Y.-Y. et al. DCL2- and RDR6-dependent transitive silencing of SMXL4 and SMXL5 in Arabidopsis dcl4 mutants causes defective phloem transport and carbohydrate over-accumulation. Plant J. 90, 1064–1078 (2017).
Elmayan, T. et al. Arabidopsis mutants impaired in cosuppression. Plant Cell 10, 1747–1757 (1998).
Hamilton, A. Two classes of short interfering RNA in RNA silencing. EMBO J. 21, 4671–4679 (2002).
Lange, H. et al. The RNA helicases AtMTR4 and HEN2 target specific subsets of nuclear transcripts for degradation by the nuclear exosome in Arabidopsis thaliana. PLoS Genet. 10, e1004564 (2014).
Nishimura, N. et al. Analysis of ABA Hypersensitive Germination2 revealed the pivotal functions of PARN in stress response in Arabidopsis. Plant J. 44, 972–984 (2005).
Henderson, I. R., Liu, F., Drea, S., Simpson, G. G. & Dean, C. An allelic series reveals essential roles for FY in plant development in addition to flowering-time control. Development 132, 3597–3607 (2005).
Arciga-Reyes, L., Wootton, L., Kieffer, M. & Davies, B. UPF1 is required for nonsense-mediated mRNA decay (NMD) and RNAi in Arabidopsis. Plant J. 47, 480–489 (2006).
Kersey, P. J. et al. Ensembl Genomes 2016: More genomes, more complexity. Nucleic Acids Res. 44, D574–D580 (2016).
Langmead, B., Trapnell, C., Pop, M. & Salzberg, S. L. Ultrafast and memory-efficient alignment of short DNA sequences to the human genome. Genome Biol. 10, R25 (2009).
Li, H. et al. The Sequence Alignment/Map format and SAMtools. Bioinformatics 25, 2078–2079 (2009).
Anders, S., Pyl, P. T. & Huber, W. HTSeq–a Python framework to work with high-throughput sequencing data. Bioinformatics 31, 166–169 (2015).
Quinlan, A. R. & Hall, I. M. BEDTools: A flexible suite of utilities for comparing genomic features. Bioinformatics 26, 841–842 (2010).
Freese, N. H., Norris, D. C. & Loraine, A. E. Integrated genome browser: Visual analytics platform for genomics. Bioinformatics 32, 2089–2095 (2016).
Acknowledgements
We thank Brendan Davies (University of Leeds, UK) for the upf1-5 and upf3-1 mutants, Peter Shaw (John Innes Centre, UK) for ncb-4 mutant, Caroline Dean (John Innes Centre, UK) for cstf64-2 mutant, Dietmar Funck (University of Konstanz, Germany) for rsr1-2 mutant and Hervé Vaucheret (INRA Centre de Versailles-Grignon, France) for ahg2-1, ccr4a and pab2 pab4 mutants as well as advice regarding the project. We thank Vladyslava Liudkovska for critical reading of the manuscript.
Funding
This research was funded by the National Science Centre Grant UMO-2013/08/M/NZ1/00931 and UMO-2018/29/B/NZ3/01980 to J.K.
Author information
Authors and Affiliations
Contributions
M.K. and J.K. wrote the manuscript; M.K. performed experiments.
Corresponding authors
Ethics declarations
Competing interests
The authors declare no competing interests.
Additional information
Publisher's note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Krzyszton, M., Kufel, J. Analysis of mRNA-derived siRNAs in mutants of mRNA maturation and surveillance pathways in Arabidopsis thaliana. Sci Rep 12, 1474 (2022). https://doi.org/10.1038/s41598-022-05574-4
Received:
Accepted:
Published:
DOI: https://doi.org/10.1038/s41598-022-05574-4
- Springer Nature Limited