Mobile gene silencing in Arabidopsis is regulated by hydrogen peroxide

EMS (ethyl methane sulfonate) mutagenesis and screening

As our starting material we used a transgenic Arabidopsis thaliana line, RtSS (Root to Shoot Silencing), in which 35S-GFP is expressed in all tissues. This line also expresses a root-specific promoter from tobacco, TobRb7, which controls the expression of LhG4-GR in the cell cytoplasm. When Dexamethasone (Dex) is present, the transcriptional activator, GR, enters the nucleus and binds to the 6xOp promoter, inducing bidirectional transcription of a GFP-silencing hairpin, hpGF, together with a GUS reporter, as described previously (Liang, White & Waterhouse, 2012). GFP silencing was then monitored using both fluorescence dissecting (Leica MZFLIII or Zeiss Stereo Lumar) and compound microscopes (Zeiss Axioimager) (Liang, White & Waterhouse, 2012), and all images were assembled and adjusted using Adobe Photoshop. EMS mutagenesis of RtSS was essentially performed as described by Kim, Schumaker & Zhu (2006). Initial trials showed that most of the potential mutants could not transmit the phenotype to the next generation. Hence, we used a higher concentration of EMS (2.5% ethanol and 0.6% EMS) to generate the M1 plants that formed the basis of this mutant screen. We pooled 200–300 M1 plants to provide M2 seeds. Around 100–150 M2 seeds were sown in each plate (150 cm²) onto 10 µM Dex-containing MS medium, grown for three weeks, then transferred to soil to screen for shoot silencing. In total, we screened 90,000 M2 seedlings derived from 110,000 M1 seeds, and identified 93 individual M2 plants showing root silencing but with no shoot silencing or very delayed shoot silencing. Potential mutants were allowed to self-fertilise for at least three generations to exclude environmental influences. Only one mutant, M397, showed stable genetic inheritance of the delayed silencing phenotype in 5 continuous self-fertilised generations.

Genome sequencing and genetic mapping

Both the M4 generation of M397 and the parent RtSS plants were subjected to Next Generation Sequencing using a protocol provided by the Australian Genome Research Facility (AGRF NGS Submission Guide). A total of 2.56 Gb and 5.60 Gb Paired-End 100-bp sequence was generated using the Illumina Hiseq-2000 platform for M397 and RtSS, respectively. Primers used are listed in Table S1. Bioinformatic analyses identified 4,522 and 7,470 potential SNPs from each genome, respectively, compared to the reference wild-type Columbia genome. Genetic tests showed M397 to be a single recessive mutant, therefore we reduced the total unique homozygous SNPs between M397 and the parent RtSS to 1,551 (File S1). We next used these SNPs to map the mutation in F3 families by Sanger sequencing, which showed that the mutation was strongly associated with Chr. 1 according to Wright’s fixation index with F_st > 0.1 (Wright, 1965) (Fig. S1). We then selected 20 SNPs across Chr.1 to genotype all 142 F3 families. A linkage group was constructed on Chr.1 by calculating the allele frequency (Fig. S1). Only one SNP (AT1G05260, Chr. 1 position: 1530461; G to A) showed 100% frequency. Further genetic complementation confirmed that this SNP was the causal mutation.

H₂O₂ detection and breakdown

The H₂O₂-specific fluorescent probes, Peroxy Orange 1 (PO1, SML0688-5MG; Sigma) (Dickinson, Huynh & Chang, 2010) and 2,7-dichlorodihydrofluorescein diacetate (H₂ DCFDA, D6883-250MG; Sigma) (Kristiansen et al., 2009) were used to detect H₂O₂ in RtSS and wild-type Columbia and in rci3-1 and rci3-2 mutants. Roots of 10 to 13-day old plants were placed into 1.5 cm diameter wells, made with a core-borer, in MS agar medium. Each well was then filled with 100 µl of 50 µM PO1 (diluted from 5 mM stock solution in DMSO into pH 6.4 perfusion solution Brauer et al., 1996), to submerge the roots. Plates were then sealed with parafilm and placed into growth rooms (16 h light/8 h dark at 22 °C) and stained overnight. After 16–18 h of staining, plants were briefly rinsed with water before beginning the longitudinal sectioning and observation using a Zeiss Axioimager fluorescence microscope with a DsRed filter to detect the PO1 signal, and an AF488 filter to detect GFP and chlorophyll autofluorescence. Identical camera and software settings were used for all images. Fluorescence intensity over each hypocotyl was measured using ImageJ (http://rsb.info.nih.gov/ij/), and statistical significance was assessed using a Student’s t test, as in Dickinson, Huynh & Chang (2010).

Treatments to modify PD or peroxide levels

Treatment media were prepared by adding chemicals directly to warm liquid MS agar medium. For H₂O₂ treatment, aliquots of 35% stock solution (Sigma, 349887-500ML) were added to Dex-containing MS agar medium just prior to setting. Catalase was filter-sterilized and added to near room temperature MS medium from 100 mg/ml stock in 50 mM potassium phosphate to give a final concentration of 1 mg/ml. For MnO₂treatment, approximately 1 g autoclaved MnO₂ (529664-5G) powder was sprinkled onto the surface of Dex-containing MS medium before seeds were sown. For n-ethyl maleimide (NEM) treatment, plants germinated on medium with or without Dex for 4 days were transferred to 50 µM or 100 µM NEM-containing MS medium, with corresponding Dex treatment. Control plants were transferred to normal MS medium, and other controls were never treated with Dex. The number of plants showing silencing were recorded, providing binary data (presence or absence of silencing) which were subjected to logistic regression analysis. Odds ratios greater than 1.0 indicate that a treatment is more likely to induce silencing than in controls, while odds ratios less than 1.0 indicate that a treatment is less likely to induce silencing, at calculated levels of significance (p value). In cases where percentage data were used to compare two treatments, the Wilcoxon signed rank test was applied to assess the significance of observed differences.

Implementation and spread of RNAi can be genetically uncoupled

To investigate genetic factors involved in regulating silencing mobility, we generated a Root-to-Shoot mobile Silencing System (RtSS) in which a green fluorescent protein (GFP) reporter transgene is expressed throughout the plant but is specifically silenced in root tissue following the application of Dex. The system, triggered by localized hairpin RNA expression in the roots, generates a signal that moves cell-to-cell from the root, through the hypocotyl and into the shoot, visibly silencing the green fluorescence (Liang, White & Waterhouse, 2012). Treating seed from our RtSS line (T5 generation) with a chemical mutagen (EMS) and screening the Dex-induced progeny yielded plant lines with accelerated, retarded or abolished spread of silencing into the shoot tissue. From 110,000 seeds, 93 lines displayed root silencing that failed to spread into the shoots. The silencing in the root tissue demonstrated that components of the RNAi generation mechanism had not been compromised in these mutants and that they were likely to be defective in either signal mobility or the ability to respond to the signal. Of these lines, one showed stable inheritance of the trait for 5 generations. Longitudinal sections (Fig. 1A) revealed that the induction and initial shootward spread of silencing at the base of the hypocotyl in the mutant and wild-type RtSS plants were almost indistinguishable up to 5 days post induction (dpi). However, at 11 dpi there was a marked difference in spread, and by 21 dpi all of the wild-type RtSS plants but none of the mutant plants displayed silencing in the rosette leaves (Fig. 1 and Figs. S2E, S2F). In the wild-type, the silencing front migrated from the root-hypocotyl junction to the hypocotyl-epicotyl junction at a rate of 373 ± 65 µm per day (N = 9, similar to the rate of 377 ± 96, N = 12, observed in Liang, White & Waterhouse (2012)), whereas in the mutant this was reduced to 110 ± 20 µm (N = 9) per day. Once the silencing front reached the hypocotyl-epicotyl junction, it migrated slowly towards the meristem at a rate of 56 ± 22 µm per day (N = 14) in the wild-type (Liang, White & Waterhouse, 2012), which was reduced to 20 ± 7 µm per day (N = 9) in the mutant. The rate of spread in the epicotyl of the mutant was so slow that the silencing front never reached the shoot apical meristem, and as a result neither the rosette leaves nor the floral bolt were silenced.

In Arabidopsis grafts or RtSS constructs, the silencing hairpin construct in the roots, which is against the first 400-bp fragment (the GF fragment) of the GFP gene, shows transitivity, such that silencing siRNAs in the graft scion or RtSS shoot are mostly against the 3′ region of the gene (the P fragment) (Brosnan et al., 2007; Liang, White & Waterhouse, 2012). The roots contain both GF- and P-derived siRNAs (Liang, White & Waterhouse, 2012). The abundant P region-derived siRNAs in both wild-type RtSS and mutant roots, and their presence in RtSS but not in the mutant shoots, confirmed that the mutants could generate but not mobilise these silencing signals (Fig. 1B). Apart from the reduced spread of Dex-induced silencing, mutant plants were phenotypically almost indistinguishable from wild-type plants, and showed normal induction of GUS expression by Dex (Fig. S3).

RCI3 (Rare Cold Inducible 3), a type III peroxidase, is required for mobile RNAi

In order to identify the mutated gene responsible for this loss of silencing spread into aerial tissues, we backcrossed our homozygous mutant line with its wild-type RtSS parent, deep sequenced the parental lines, and deep sequenced a pool of 44 F2 plants expressing the mutant phenotype. We identified 1,551 single nucleotide polymorphisms (SNPs) between the parental wild-type RtSS and the mutant (Table S1, Fig. S1), and mapped the mutation to the top arm of chromosome 1 between nucleotide position 863625 and 17192682. Further Sanger sequencing analysis using 20 SNPs and 12 F3 families refined this to a zone between SNP markers 1g001 and 1g012 (Table S1, Fig. S1) and ultimately to one mutation, at 1530461, in the RCI3 gene (At1g05260), which encodes a type III peroxidase. The mutation causes an Arg¹⁴⁵ to Lys substitution in a motif of the protein that is highly conserved from bryophytes to eudicots (Attwood et al., 1994) (Fig. 2A) and substitution of this amino acid within the invariant GRRDG sequence seemed highly likely to compromise the function of the enzyme (Welinder et al., 2002). To confirm that this was the cause of the significantly slowed silencing spread, we re-introduced a functional copy of RCI3 into the mutant background. A 4.5 kb genomic fragment containing the wild-type promoter, coding region and terminator sequence was used and restored root-to-shoot mobile silencing in 14 out of 20 transformants (Fig. S4). This demonstrates that RCI3 is required for silencing mobility in Arabidopsis.

H₂O₂ is an endogenous signal that regulates mobile RNAi

RCI3 is involved in the production of reactive oxygen species (ROS) (Llorente et al., 2002; Kim, Ciani & Schachtman, 2010), so we analyzed our mutant (now termed rci3-2) and rci3-1 (Kim, Ciani & Schachtman, 2010), a T-DNA insertion mutant, with H₂O₂-detecting dyes (Dickinson, Huynh & Chang, 2010). This revealed that H₂O₂ is endogenously produced and readily detectable in the wild-type RtSS parental line, whereas it is barely detectable in rci3-1 and almost undetectable in rci3-2 plants (Fig. 2B and Fig. S5). Overexpression of RCI3 increases ROS production (Kim, Ciani & Schachtman, 2010); therefore, we wondered whether treatment with H₂O₂ could restore silencing spread in rci3-2 plants. Although slower to reach the shoots than in wild-type RtSS plants, 42 of 46 Dex-induced rci3-2 plants treated with 1.5 mM H₂O₂ displayed shoot silencing by 36 dpi (Figs. 3A, 3B and 3E). Furthermore, H₂O₂ treatment accelerated the rate of silencing spread in wild-type RtSS plants (Figs. 3A, 3B and 3E). A logistic regression analysis of the data in Fig. 3E showed that increasing H₂O₂ concentration was significantly associated with shoot silencing (odds ratio = 1.78, 95% confidence interval = 1.51–2.10; p < 0.001), although the highest concentrations were sub-optimal. Peroxide treatment also increased the distance of spread from a vascularly-expressed silencing signal targeting phytoene desaturase (AtSuc2:PDS) (Smith et al., 2007), which results in leaf bleaching (Figs. 3C and 3D). Indeed, all 66 Dex-induced RtSS plants growing on H₂O₂-containing medium showed accelerated silencing into the shoots 13 days later (Fig. 3F). However, if catalase was added to this medium to eliminate H₂O₂, none of 74 plants displayed any shoot silencing after the same time period (Fig. 3F). Furthermore, medium containing catalase also delayed silencing spread in Dex-induced wild-type-RtSS (Fig. 3F and Fig. S6). In this case, a logistic regression analysis of the data in Fig. 3F showed that the presence of catalase was significantly associated with reduced shoot silencing (odds ratio = 0.01, 95% confidence interval = 0.005–0.029; p < 0.001).

We interpreted this to be a consequence of the catalase accelerating the breakdown of endogenous pools of H₂O₂ in the plants. Applying a different H₂O₂-breakdown catalyst (MnO₂) also caused slower silencing spread in both Dex-induced wild-type-RtSS (Fig. 3F; logistic regression; odds ratio <0.001; p < 0.001) and in plants with vascular-pattern PDS silencing, both catalase and MnO₂ reduced the distance of silencing spread from veins (Fig. S6). Collectively, the results demonstrate that H₂O₂ is an endogenous signal that regulates the rate of silencing mobility in plants.

Increased PD permeability requires RCI3

Previous work has shown that H₂O₂ enhances PD permeability to a symplastic dye (Rutschow, Baskin & Kramer, 2011). Therefore, the restoration of silencing in rci3-2 by H₂O₂ treatment might be a result of enlarged PD. To test this idea, we applied n-ethyl maleimide (NEM), which can increase transport via PD (White & Barton, 2011; Liang, White & Waterhouse, 2012). As previously reported (Liang, White & Waterhouse, 2012), 100 µM NEM enhanced the rate of spread of gene silencing in RtSS plants (Fig. S7), but did not restore silencing spread into the shoots of rci3-2 plants (Fig. S7). This suggests that the rescue of silencing spread in rci3-2 plants by H₂O₂ treatment is due to a specific effect of H₂O₂ and that NEM cannot cause PD enlargement without the enzymatic activity of RCI3.