A Link Between mRNA Turnover and RNA Interference in Arabidopsis

RNA interference is a conserved posttranscriptional control mechanism that is initiated by dsRNA and causes degradation of mRNAs with homology to the dsRNA trigger. The dsRNA is cut by the Dicer RNase into 21- to 25-nucleotide (nt) fragments, called small interfering RNAs (siRNAs). The siRNAs are incorporated into the RNA-induced silencing complex (RISC), which uses them as a guide to identify and to degrade matching mRNAs (1).

RNA interference can be triggered by viral dsRNA, by self-complementary transcripts, or by sense transgenes, a phenomenon also called cosuppression (2, 3). Cosuppression can be triggered by complex transgene insertions, which may produce antisense or self-complementary transcripts, or by single transgenes with high expression levels (4). In the case of single, sense transgenes, it has been proposed that aberrant transcripts (e.g., caused by premature transcriptional termination) could be converted to dsRNA by RNA-dependent RNA-polymerases (RdRps), which are required for RNAi in plants, fungi, and Caenorhabditis elegans (5–9). The RdRps are also believed to amplify and to maintain RNAi, by using siRNAs as primers to synthesize new dsRNA (10). In both the initiation and amplification of RNAi, however, the template used by RdRp in vivo is unknown.

Here, we have isolated an Arabidopsis mutant that promotes RdRp-dependent cosuppression and show that the corresponding wild-type gene encodes an RNA exonuclease that likely degrades the template for RdRp. The mutant was isolated (11) in a screen for suppressors of a single-copy transgene expressing a fusion between SHOOT MERISTEMLESS (STM, a regulator of meristem development) (12), and the rat glucocorticoid receptor (GR), under the widely expressed 35S promoter. Activation of STM-GR with dexamethasone (DEX) activates meristem genes and inhibits cotyledon and leaf development (Fig. 1, A and B) (13). The recessive xrn4-1 mutation suppressed this phenotype (Fig. 1, C and D). When xrn4-1, STM-GR was crossed with the wild type lacking STM-GR, all progeny showed the DEX-induced STM-GR phenotype, so the transgene was intact. Suppression in trans was also confirmed by segregating xrn4-1 from the transgene and, after three backcrosses with the wild type, crossing again with STM-GR.

• Download high-res image
• Open in new tab
• Download Powerpoint

In a homozygous STM-GR background, xrn4-1 segregated as a single locus. In xrn4-1 plants that were hemizygous for STM-GR, however, the phenotype was variable and rarely reached the full suppression seen in xrn4-1, STM-GR homozygous plants. This dependence on transgene dosage is typical of cosuppression (4). Accordingly, STM-GR mRNA levels were lower in xrn4-1, STM-GR plants (Fig. 2A). siRNAs corresponding to STM sequences (Fig. 2B) and to GR sequences (fig. S1) were detected in xrn4-1, STM-GR, but were eliminated by the sde1-1 mutation, which disrupts the RdRp required for transgene silencing in Arabidopsis (5, 6). Suppression of the STM-GR phenotype by xrn4-1 also depended on SDE1 (Fig. 2, D to G). The accumulation of siRNAs and the requirement of SDE1 showed that the suppression of STM-GR by xrn4-1 was caused by RdRp-dependent RNAi.

• Download high-res image
• Open in new tab
• Download Powerpoint

Silencing was not specific to the STM-GR locus used in the mutant screen: Similar suppression was seen after crossing xrn4-1 with an independent STM-GR line (fig. S2). xrn4-1 also suppressed WUS-GR (expressing a fusion between GR and the meristem regulator WUSCHEL) (14), but not AG-GR (with GR fused to the floral homeotic protein AGAMOUS) (15) (Fig. 1, E to H). Silencing was, however, specific to ectopically expressed STM and WUS: Meristem activity, which requires endogenous STM and WUS, was unaffected.

Positional cloning, complementation, and isolation of independent alleles (fig. S3) showed that the mutation disrupted XRN4, which is likely the functional homolog of Xrn1p, a 5′-3′ exonuclease that degrades decapped mRNAs in yeast (16, 17). Consistent with a role in mRNA turnover, XRN4 is cytoplasmic (17) and degrades the 3′ fragment of microRNA-cleaved mRNAs (18).

Cosuppression caused by mutation of an RNA exonuclease suggests that XRN4 degrades an RNA required to initiate or maintain silencing. A role in RNAi has also been suggested for xrn1 in C. elegans, but reduced xrn1 activity was lethal, which complicated further analysis (19). Still in C. elegans, the RNA exonuclease ERI1 antagonizes RNAi by turning over small RNAs (20). The xrn4-1 mutation, however, did not affect the levels of microRNAs (which are comparable to siRNAs in chemical structure) (Fig. 2C and fig. S1). It also seems unlikely that XRN4 could antagonize silencing by degrading dsRNA, because XRN exonucleases stall on RNA with strong secondary structure (17, 21). In addition, xrn4-1 did not change susceptibility to the turnip mosaic virus (TuMV) and cucumber mosaic virus (CMV), which replicate through a dsRNA intermediate (22).

Other XRN4 substrates could be aberrant STM-GR and WUS-GR transcripts. On the basis of the properties of XRN proteins, these transcripts could be predicted to be cytoplasmic and to have exposed 5′ ends. Obvious candidates would be decapped mRNA or products of small RNA-directed cleavage. To detect accumulation of aberrant STM-GR RNA in xrn4-1, we used the sde1-1 mutation to prevent silencing and the consequent destruction of RNA by RISC. No cleavage of STM-GR mRNA was detected by Northern blot, but the levels of full-length mRNA were slightly increased (Fig. 3A). To detect exposed 5′ RNA ends, we used a modified RACE (rapid amplification of cDNA ends) protocol (23). Full-length, decapped STM-GR mRNA (verified by cloning and sequencing of the cDNA) consistently accumulated in xrn4-1 (Fig. 3C), with an estimated increase of 2.4-, 4.1-, and 12.9-fold in three independent experiments (fig. S4). Shorter cDNAs were amplified, but Southern blotting showed that these did not contain STM-GR sequences (Fig. 3D). Although discrete cleavage products were not seen, it is difficult to exclude that low levels of heterogeneous cleavage by siRNAs might escape detection.

• Download high-res image
• Open in new tab
• Download Powerpoint

The accumulation of decapped STM-GR mRNA in xrn4-1 confirmed that XRN4, like yeast xrn1p, functions in mRNA turnover, by means of decapping, and 5′-3′ exonucleolysis. XRN4 is broadly expressed (17) (fig. S5), which suggests that its role is not specific for certain tissues or developmental stages. However, XRN4 is probably not essential for general mRNA turnover, because even severe mutations caused no obvious growth defects (18, 24)]. In yeast, xrn1 mutants are also viable, unless they are combined with mutations affecting the 3′-5′ exonucleolytic pathway (25). It is possible that in plants, as in yeast, mRNA turnover occurs through redundant 5′-3′ and 3′-5′ degradation pathways.

Our data also indicate that XRN4 antagonizes RNAi, possibly by degrading the template for RdRp. Tomato and Neurospora RdRps synthesize dsRNA in vitro from both primed and unprimed single-stranded RNA (26, 27), so decapped mRNA could serve as the template either to initiate or to maintain silencing. This would also agree with the cosuppression triggered in tobacco cells by direct introduction of single-stranded, sense, uncapped RNA (28). Alternatively, or in addition to the role of decapped mRNA, XRN4 might degrade RISC-cleaved mRNA that would otherwise serve as the RdRp template during RNAi amplification.

A simple reason why accumulation of mRNA lacking a cap structure (owing to either decapping or cleavage) could promote RNAi is that absence of ribosomes exposes the RNA as a substrate for RdRp. The decapping machinery competes with translation for access to mRNAs (16). In the case of STM, translation depends on other genes that are expressed in the meristem and vasculature (29), so ectopic STM-GR mRNA may be inefficiently translated and targeted for decapping (or silencing, in xrn4-1). Conversely, efficient translation in the meristem may explain why endogenous STM was not affected, although there is also evidence that the shoot meristem may be protected from RNAi (30). mRNA decapping, for example due to limiting levels of translation cofactors, could be a general reason why cosuppression correlates with high levels of transgene expression.