Suppression of endogenous gene silencing by bidirectional cytoplasmic RNA decay in Arabidopsis

See allHide authors and affiliations

Science  03 Apr 2015:
Vol. 348, Issue 6230, pp. 120-123
DOI: 10.1126/science.aaa2618

Protecting against runaway defense systems

RNA interference defends cells against invading genetic elements, such as viruses or transgenes. But while the invasive RNAs are under attack, what protects the normal endogenous RNAs? Zhang et al. identified, in the small plant Arabidopsis, a surveillance system to do just that: Preserve the normal transcriptome and keep the attack focused on invasive transcripts.

Science, this issue p. 120


Plant immunity against foreign gene invasion takes advantage of posttranscriptional gene silencing (PTGS). How plants elaborately avert inappropriate PTGS of endogenous coding genes remains unclear. We demonstrate in Arabidopsis that both 5′-3′ and 3′-5′ cytoplasmic RNA decay pathways act as repressors of transgene and endogenous PTGS. Disruption of bidirectional cytoplasmic RNA decay leads to pleiotropic developmental defects and drastic transcriptomic alterations, which are substantially rescued by PTGS mutants. Upon dysfunction of bidirectional RNA decay, a large number of 21- to 22-nucleotide endogenous small interfering RNAs are produced from coding transcripts, including multiple microRNA targets, which could interfere with their cognate gene expression and functions. This study highlights the risk of unwanted PTGS and identifies cytoplasmic RNA decay pathways as safeguards of plant transcriptome and development.

Gene expression and silencing establish the proper transcriptome of eukaryotic cells. Posttranscriptional gene silencing (PTGS), also known as RNA interference, serves as an RNA-based immune system against foreign gene invasion and is engaged in silencing a subset of endogenous genes, mainly transposons (1, 2). PTGS is triggered by cellular double-stranded RNAs, which are subsequently cleaved into 20- to 24-nucleotide (nt) small RNA duplexes by Dicer family proteins (2). One strand of the small RNA duplex can be loaded into an Argonaute (AGO)–containing silencing complex, which targets homologous RNAs for degradation and/or translation inhibition (2). Plants, fungi, and worms possess RNA-dependent RNA polymerases (RDRPs) that could produce secondary small interfering RNA (siRNA) molecules after the primary siRNA targeting, thus amplifying the effects of PTGS (3).

PTGS is an elaborately regulated process. Various plant viruses carry repressors of PTGS, combating plant immunity against infection (4, 5). Several components that participate in mRNA quality control and processing function as repressors of transgene PTGS (6, 7). In this work, we use the model plant Arabidopsis to investigate how endogenous protein-coding transcripts avoid being targeted by PTGS machineries that efficiently silence the expression of viral genomes or transgenes.

In Arabidopsis, overexpression of ETHYLENE-INSENSITIVE3 (EIN3), a transcriptional activator of ethylene signaling (8), resulted in a small-cotyledon phenotype characteristic of enhanced ethylene response (9) (Fig. 1A) (see also supplementary materials and methods). In a genetic screen searching for suppressors of this stock, we isolated four large-cotyledon mutants that phenocopied the ein3 loss-of-function mutant (8) (Fig. 1A and fig. S1A). One mutant affected ETHYLENE-INSENSITIVE5 (EIN5), encoding a cytoplasmic 5′-3′ exoribonuclease known in transgene PTGS repression (7, 10) (fig. S1). The other three mutants (s28, s37, and s40) showed reduced expression of EIN3 and one of its targets, ETHYLENE-RESPONSE-FACTOR1 (ERF1) (11) (Fig. 1B). s28 carried a mutation in the coding sequence of a DExH-box RNA helicase (AT3G46960) of the SKI2 (Super-Killer2) type (12) (Fig. 1C and fig. S2A). Green fluorescent protein (GFP)–tagged AtSKI2, expressed exclusively in the cytoplasm (Fig. 1D), restored normal cotyledon development and expression of the EIN3 transgene in the s28 line (fig. S3). Two additional Arabidopsis insertional ski2 alleles, ski2-2 and ski2-3 (figs. S2A and S4), displayed normal responses to ethylene (fig. S5), excluding AtSKI2 functions in the ethylene signaling pathway. s37 and s40, allelic to each other, both carried nonsense mutations in AtSKI3 (AT1G76630) (12) (Fig. 1C and fig. S2A). In yeast, ScSki2, ScSki3, and ScSki8 function as a cytoplasmic SKI complex to unwind and thread RNAs into the 3′-5′ exoribonuclease complex (exosome) for decay (13) (fig. S2B). In Arabidopsis, we found peptides characteristic of AtSKI3, AtSKI8, and AtSKI7 and several other proteins associated with AtSKI2 (fig. S6). Pull-down experiments confirmed the direct interaction between AtSKI2 and AtSKI3 (Fig. 1E).

Fig. 1 RDR6-dependent transgene PTGS upon the dysfunction of 3′-5′ RNA decay.

(A) Cotyledon phenotype of 6-day-old seedlings grown on Murashige and Skoog (MS) medium supplemented with or without 10 μM ACC, an ethylene biosynthetic precursor. s28, s37, and s40 mutants were generated in the EIN3ox (EIN3-overexpressor) background. The s28 rdr6 double mutant was generated by crossing. Scale bar, 1 mm. (B) Relative expression levels of EIN3 and its target gene ERF1 in 6-day-old seedlings. Error bars indicate SD; number of replicates (n) = 3. (C) Identification of the causal genes in the mutants by positional cloning. Mutations of amino acid (aa) residues are indicated. T, Thr; I, Ile; Q, Gln; W, Trp. (D) Subcellular localization of AtSKI2 fused with GFP. 4′,6-diamidino-2-phenylindole (DAPI) staining indicates the nuclei. Scale bar, 5 μm. (E) In vitro glutathione S-transferase (GST) pull-down assay between the C terminus of AtSKI3 (AtSKI3C, aa 930 to 1169) and the N terminus of AtSKI2 (AtSKI2N, aa 1 to 344). WB, Western blotting. (F) Northern blot detection of small RNAs derived from EIN3. (G) Relative expression levels of EIN3 in 6-day-old seedlings. Error bars indicate SD; n = 3.

Evidence that AtSKI2 is a general repressor of transgene PTGS came from two experiments: (i) When we introduced an unrelated transgene [ADENINE PHOSPHORIBOSYLTRANSFERASE 1 (APT1)] to the ski2-2 line, the phenotype mimicked the apt1 mutant (14) (fig. S7), suggesting cosuppression of both the APT1 transgene and endogenous APT1. (ii) Mutation of RDR6, an RDRP essential for transgene PTGS in plants (15), caused the small-cotyledon phenotype and reactivated expression of the EIN3 transgene in s28 (Fig. 1, A and G). Furthermore, the marked accumulation of EIN3 transgene–derived siRNAs in s28 was eliminated by rdr6 mutation (Fig. 1F). Thus, the cytoplasmic 3′-5′ and 5′-3′ RNA decay pathways mediated by the SKI complex and EIN5, respectively, are required for suppressing transgene PTGS.

Neither cytoplasmic RNA decay pathway alone was essential for plant development, as the ein5 and ski2 single mutants were morphologically normal except for mild leaf serrations in ein5 (10) (Fig. 2A). However, disruption of both pathways (ein5-1 ski2-2) proved to be lethal at the embryonic stage (fig. S8A). Hypomorphic ein5-1 ski2-3 homozygotes (figs. S2A and S4B) were viable but arrested at the early vegetative stage with growth disorders, including defects in meristem, rosette leaves, and leaf coloration (Fig. 2A). rdr6 rescued the ein5 ski2 double mutants, and the resultant triple mutants showed normal vegetative growth and fertility (Fig. 2A and fig. S8). Thus, the defects caused by ein5 ski2 are RDR6-dependent.

Fig. 2 Disruption of bidirectional RNA decay exhibits RDR6-dependent developmental defects and transcriptomic alterations.

(A) Rosette morphology of 3-week-old plants of indicated genotypes. Scale bar, 1 cm. (B) Clustering of transcriptome profiles of ein5-1, ski2-3, rdr6-11, and their combinations. The clustering results are based on expression levels of 15,663 genes that have the expression levels with reads per kilobase per million > 1 in the Col-0 background. (C) Relative expression of genes that regulate anthocyanin biosynthesis in 16-day-old plants. Error bars indicate SD; n = 3. (D) Quantification of anthocyanin accumulation in 16-day-old plants. Error bars indicate SD, n = 6.

We profiled the transcriptomes in the various genotypes. The profile of rdr6-11 ein5-1 ski2-3 resembled that of rdr6-11 but differed from that of ein5-1 ski2-3 (Fig. 2B). We identified 596 differentially expressed genes (111 up-regulated and 485 down-regulated) when both EIN5 and AtSKI2 were mutated (fig. S9A). Of the up- and down-regulated genes, 85.6 and 73.0%, respectively, were restored by rdr6 mutation (fig. S9B), indicating that most of the transcriptomic changes in ein5-1 ski2-3 are dependent on RDR6.

The genes regulated by the combination of EIN5 and AtSKI2 are enriched in functional categories such as flavonoid biosynthesis, ribosome, and photosynthesis (table S1). For instance, expression of several anthocyanin biosynthesis enzymes (DFR, TT8) and their regulatory transcription factors (PAP1, PAP2) (16) was up-regulated in ein5-1 ski2-3 but reduced in rdr6-11 ein5-1 ski2-3 (Fig. 2C). Accordingly, anthocyanin highly accumulated in ein5-1 ski2-3 but not in rdr6-11 ein5-1 ski2-3 (Fig. 2D), manifesting as purple pigmentation in rosette leaves of ein5-1 ski2-3 plants (Fig. 2A).

The defects of ein5 ski2 were rescued by other mutations that affect the 21- to 22-nt siRNA pathway, such as ago1, suppressor of gene silencing3 (sgs3), and the dicer-like4 (dcl4) dcl2 double mutant (17) (Fig. 3, A and B, and figs. S10 and S11). In Arabidopsis, DCL4 generates 21-nt siRNAs, whereas DCL2 generates 22-nt siRNAs, which could efficiently trigger secondary siRNA biogenesis (18). Neither dcl2 nor dcl4 rescued the lethality of ein5-1 ski2-2 (Fig. 3B); however, dcl2 evidently recovered the developmental disorders of the hypomorphic ein5-1 ski2-3 (Fig. 3A). Thus, both 21- and 22-nt siRNA pathways are deleterious to plant survival, and the 22-nt siRNA pathway is particularly detrimental to plant development when bidirectional RNA decay is disrupted. In contrast to normal morphology of ein5 dcl2 and ski2 dcl2, both ein5 dcl4 and ski2 dcl4 plants displayed defects in rosette growth, meristem function, and leaf pigmentation. Further loss of DCL2 function rescued these defects (Fig. 3, C and D). Thus, upon deficiency of either 5′ (EIN5) or 3′ (AtSKI2) RNA decay, the DCL4-mediated pathway could compete with the DCL2-mediated pathway to minimize the risk that 22-nt siRNAs might trigger amplified PTGS harmful to plant development.

Fig. 3 Genetic interaction between the PTGS and RNA decay pathways.

(A to D) Rosette morphology of 19-day-old plants was shown. No ein5-1 ski2-3 dcl4-2 plant was viable based on genotyping the segregating population derived from the ein5-1 ski2-3 hemizygote and dcl4-2 dcl2-1 cross (A). Similarly, the ein5-1 ski2-2 double mutants were not viable, and by genotyping the F2 and F3 plants propagated from the ein5-1 ski2-2 hemizygote and dcl4-2 dcl2-1 cross, no ein5-1 ski2-2 dcl2-1 or ein5-1 ski2-2 dcl4-2 triple mutants were viable (B). Scale bars, 1 cm.

We used small RNA sequencing to profile small RNA production in wild-type (Col-0), rdr6-11, ein5-1 ski2-3, and rdr6-11 ein5-1 ski2-3 (fig. S12A). We found that few reads mapped to all eight Arabidopsis TAS genes in rdr6-11 and rdr6-11 ein5-1 ski2-3, consistent with the requirement of RDR6 for trans-acting siRNA (tasiRNA) biogenesis (19). Abundant and comparable levels of tasiRNA were detected in Col and ein5-1 ski2-3 for 7 TAS genes (fig. S12, B and C), except for TAS4 tasiRNAs, which may have accumulated due to the elevated expression of TAS4 in ein5-1 ski2-3 (fig. S13). Therefore, defects in cytoplasmic RNA decay do not affect tasiRNA biogenesis.

To further explore the genetic cause of detrimental PTGS, we filtered the genes that generate 21- to 22-nt siRNAs using two criteria: (i) loci with more mapping reads of siRNAs in ein5-1 ski2-3 than in Col [twofold cutoff, P < 0.005, false discovery rate (FDR) < 0.1], and (ii) loci with fewer mapping reads of siRNAs in rdr6-11 ein5-1 ski2-3 than in ein5-1 ski2-3 (twofold cutoff, P < 0.005, FDR < 0.1). Of 456 genes that met these criteria, 441 encoded proteins (Fig. 4, A and B, and table S2). We named the siRNAs derived from the 441 loci as coding transcript–derived siRNAs (ct-siRNAs). Among these 441 loci were 39 microRNA (miRNA)–targeted protein-coding genes (table S3), including miR167 targets (ARF6, ARF8) and miR165/166 targets (five HD-ZIPIII genes) (Fig. 4A and fig. S14). ct-siRNAs arise from either 5′ miRNA-cleavage fragments or 5′ plus 3′ fragments of these miRNA targets (Fig. 4A and fig. S15), distinct from the biogenesis of most tasiRNAs within 3′ fragments of TAS loci (20).

Fig. 4 Identification and functional analysis of a class of siRNAs accumulated in ein5 ski2.

(A) Accumulation of 21- to 22-nt siRNAs derived from representative coding genes in ein5 ski2, but barely in other genotypes. Arrows indicate the miRNA cleavage sites in the selected miRNA-target genes. (B) Overall abundance of ct-siRNAs generated from 441 protein-coding loci in each genotype. Error bars indicate SD; n = 3. (C) Relative expression levels of select miRNA target genes (ARF and HD-ZIPIII family genes). Error bars indicate SD; n = 3. (D) Relative expression levels of NIA1/2 and three nitrogen-responsive genes (LBD37, -38, and -39). Error bars indicate SD; n = 3. (E) Two-week-old plants grown in soil irrigated with water (Mock) or 1.65 g/liter ammonium nitrate solution (NH4NO3). Scale bar, 1 cm.

The ct-siRNA loci were overrepresented in the down-regulated genes of ein5-1 ski2-3 (fig. S16). For instance, expression of the above-mentioned ARF6/8 and HD-ZIP III genes was reduced in ein5-1 ski2-3 but not in rdr6-11 ein5-1 ski2-3 (Fig. 4C), suggesting that the RDR6-dependent ct-siRNA biogenesis intensifies the repression of gene expression triggered by miRNAs. We identified NIA1 and NIA2, which encode two nitrate reductases required for nitrate assimilation in Arabidopsis (21), from the top 20 loci that show the most notable ct-siRNA production in ein5 ski2 (Fig. 4A and fig. S17). Expression of NIA1 and NIA2, as well as three nitrogen-responsive genes (LBD37, -38, and -39) (22), was reduced in ein5-1 ski2-3 and was restored in rdr6-11 ein5-1 ski2-3 (Fig. 4D). Thus, ct-siRNA–mediated silencing of NIA1 and NIA2 genes disturbed nitrogen metabolism, which activates anthocyanin biosynthesis (23). In agreement, the purple pigmentation of ein5-1 ski2-3 leaves at the early stage was normalized when the plants were irrigated with ammonium nitrate, which compensates for the compromised nitrate assimilation (Fig. 4E). Thus, biogenesis of ct-siRNAs in many cases could reduce the expression of their cognate genes and compromise gene functions.

The ct-siRNA loci were also overrepresented in the up-regulated genes of ein5-1 ski2-3 (fig. S16). Furthermore, the ct-siRNA loci, on average, were expressed at markedly higher levels than the rest of the genome (fig. S18). Transgenes expressed at higher levels are more likely to undergo PTGS to produce siRNA (24). It is thus likely that accumulation of a subset of ct-siRNAs reflects the disturbed expression of their cognate genes upon dysfunction of RNA decay pathways. In line with this, TAS4-tasiRNA accumulated to high levels due to elevated expression of the TAS4 gene in ein5 ski2 (figs. S12C and S13).

Here we describe a class of endogenous siRNAs (ct-siRNAs) with these characteristics: (i) 21 or 22 nt in length; (ii) derived from coding transcripts; (iii) produced upon the dysfunction of EIN5- and AtSKI-mediated RNA decay; (iv) dependent on RDR6, SGS3, DCL2, and DCL4 for biogenesis; and (v) partially dependent on AGO1 for action (fig. S19). Our study reveals that both transgene and endogenous PTGS occurs upon dysfunction of bidirectional cytoplasmic RNA decay, which normally eliminates aberrant transcripts arising from RNA degradation, end processing, or endo-cleavage (25). When RNA decay pathways are disrupted by genetic mutation or overloaded by overexpression of foreign genes, aberrant transcripts are selected and channeled into PTGS pathways to produce 21- to 22-nt ct-siRNAs. The DCL4-mediated pathway could serve as a decoy to antagonize the more destructive DCL2-mediated pathway, protecting endogenous mRNAs from undesirable clearance (fig. S19). Thus, cytoplasmic RNA decay sets a silencing threshold such that those invading or vastly expressed genes with high levels of aberrant RNA production probably undergo PTGS, whereas the majority of endogenous genes hardly do. As such, the siRNA-based defense system in plants that originally evolved against genome invasion carries risks of triggering adverse endogenous PTGS not typical of the protein-based immune systems found in vertebrates (3). We also propose that cytoplasmic RNA decay and siRNA pathways act as crucial components in the miRNA regulatory network. After an initial miRNA-directed cleavage, cytoplasmic RNA decay could preclude sustained or exacerbated siRNA-triggered silencing of target gene expression. Notably, miRNA-triggered tasiRNA biogenesis seems not to be affected by cytoplasmic RNA decay, raising the question as to how aberrant transcripts are sorted and targeted between cytoplasmic RNA decay and PTGS pathways.

Supplementary Materials

Materials and Methods

Figs. S1 to S19

Tables S1 to S4

References (2645)

References and Notes

  1. Acknowledgments: We thank Y. Qi (Tsinghua University) and F. Tang and Y. Li (Peking University) for assistance in RNA sequencing experiments. We also thank J. Olson (Peking University) for assistance in English editing. This work is supported by the National Basic Research Program of China (973 Program; grant 2012CB910902) and the National Natural Science Foundation of China (grants 91217305 and 91017010 to H.G.). The publication fee is covered by the 111 Project of Peking University. Data have been deposited in the Gene Expression Omnibus under accession numbers GSE52407 and GSE57936. The supplementary materials contain additional data.
View Abstract

Navigate This Article