An endogene-resembling transgene is resistant to DNA methylation and systemic silencing

Abstract

In plants, endogenes are less prone to RNA silencing than transgenes. While both can be efficiently targeted by small RNAs for post-transcriptional gene silencing (PTGS), generally only transgene PTGS is accompanied by transitivity, RNA-directed DNA methylation (RdDM) and systemic silencing. In order to investigate whether a transgene could mimick an endogene and thus be less susceptible to RNA silencing, we generated an intron-containing, endogene-resembling GREEN FLUORESCENT PROTEIN (GFP) transgene (GFP^endo). Upon agroinfiltration of a hairpin GFP (hpF) construct, transgenic Nicotiana benthamiana plants harboring GFP^endo (Nb-GFP^endo) were susceptible to local PTGS. Yet, in the local area, PTGS was not accompanied by RdDM of the GFP^endo coding region. Importantly, hpF-agroinfiltrated Nb-GFP^endo plants were resistant to systemic silencing. For reasons of comparison, transgenic N. benthamiana plants (Nb-GFP^cDNA) carrying a GFP cDNA transgene (GFP^cDNA) were included in the analysis. HpF-agroinfiltrated Nb-GFP^cDNA plants exhibited local PTGS and RdDM. In addition, systemic silencing was established in Nb-GFP^cDNA plants. In agreement with previous reports using grafted scions, in systemically silenced tissue, siRNAs mapping to the 3′ of GFP were predominantly detectable by Northern blot analysis. Yet, in contrast to other reports, in systemically silenced leaves, PTGS was also accompanied by dense RdDM comprising the entire GFP^cDNA coding region. Overall, our analysis indicated that cDNA transgenes are prone to systemic PTGS and RdDM, while endogene-resembling ones are resistant to RNA silencing.

Introduction

RNA silencing refers to a network of small RNA (sRNA)-mediated pathways that are important for normal plant development, genome stability, defense against invading nucleic acids and response to abiotic/biotic stress.^1,2 In general, there are two major pathways of RNA silencing in plants; the small interfering RNA (siRNA) and the microRNA (miRNA)-mediated silencing pathways.^3,4 The siRNA pathway is induced by double stranded RNA (dsRNA) that is processed by the DICER-LIKE (DCL) endonucleases, DCL2, DCL3 and DCL4, into 22-, 24-, and 21-nucleotides (nt) long siRNAs, respectively.⁵ Mature siRNAs exhibit 3′ 2-nt overhangs and become stabilized though 3′ end methylation by HUE ENHANCER1 (HEN1).⁶ Depending on their 5′ terminal nucleotide, siRNAs are loaded onto specific ARGONAUTE (AGO) proteins.⁷ Only one siRNA strand, the guide strand, is finally selected.⁸ 21-nt siRNAs are loaded primarily onto AGO1, and to lower extent onto AGO2, leading to the formation of RNA-induced silencing complex (RISC).⁹ RISC mediates post-transcriptional gene silencing (PTGS) through targeting RNA molecules that are complementary to the guide strand. PTGS results in target RNA degradation and/or, in the case of mRNA targets, in translational inhibition.^10,11 24-nt siRNAs are loaded onto AGO4/AGO6 and are supposed to target homologous genomic DNA for de novo methylation, in a process termed RNA-directed DNA methylation (RdDM).^12,13

PTGS and RdDM are both initiated by dsRNA. Among others, sources of dsRNA are replication intermediates of viruses and viroids, transcription of inverted repeats, stress-induced overlapping antisense transcripts and RNA-DIRECTED RNA POLYMERASE (RDR) transcription of aberrant RNAs (abRNAs).^14-16 AbRNAs are most probably transcripts devoid of 5′ cap and/or 3′ polyadenylation tail.^17,18 Such transcripts may directly be generated through abnormal DNA-DEPENDENT RNA POLYMERASE II (POL II) transcription, e.g., producing truncated and/or read-through transcripts. In addition, abRNA may derive from RISC-mediated generation of 5′ and 3′ target RNA cleavage products. Thus, while initial cleavage is conducted by primary siRNAs, RDR6 transcription of the cleavage products generates secondary dsRNA that is processed into secondary siRNAs.^19,20 In contrast to primary siRNAs, that match only to a sub-region of the target RNA, secondary siRNAs may map upstream and downstream from the binding sites of the primary siRNAs. This spreading of silencing beyond initial target sites is termed transitivity.^15,21 The exact nature of secondary siRNAs is not known, but it seems that at least DCL2 is involved in their biogenesis.²² Interestingly, while secondary siRNAs are cleavage-competent, they cannot further support the biogenesis of tertiary siRNAs. Thus, transitivity may spread to 5′ and 3′ direction, but not indefinitely.^21,23 Nevertheless, transitivity results in the amplification of siRNAs constituting a self-reinforcing feedback loop of silencing. Silencing may spread through plasmodesmata cell-to-cell and through the vasculare system to systemic parts of the plant.^24,25 While cell-to-cell signals are most probably 21-nt duplex siRNAs, the nature of systemic silencing signals is still elusive.^26-29 It has been proposed that 24-nt siRNAs^30-32 or longer RNAs³³ may be the molecules traveling through the phloem to metabolic sink tissues, but solid experimental evidence for either scenario is lacking.

In general, the response of transgenes and endogenes against RNA silencing is diverse. Both, endogenes and transgenes are equally susceptible to sRNA-mediated trans-PTGS. However, with a few exceptions, including TAS genes in Arabidopsis and BETA-1,3-GLUCANASE genes in tobacco, (1) endogenes are generally not prone to transitivity.^34,35 Likewise, with the exception of two endogenous INVERTED-REPEAT (IR) loci in Arabidopsis, (2) endogenes are not prone to systemic silencing.^26,36,37 In contrast to transgenes, (3) endogenes appear to be resistant to RdDM in N. benthamiana.³⁸ Finally, in contrast to transgenes, (4) endogenes are not prone to spontaneous silencing (cis-silencing).^39,40 In order to analyze if a transgene can mimick an endogene with introns and thus, become less prone to RNA silencing, we constructed an intron-containing endogene-resembling GFP transgene (GFP^endo). We have previously shown that upon agroinfiltration, GFP^endo was poorly processed into siRNAs.⁴¹ In this study, we generated transgenic N. benthamiana GFP^endo plants (Nb-GFP^endo) and we explored the response of GFP^endo against RNA silencing. In summary, our data indicated that, similar to most endogenes, GFP^endo was no target for RdDM and systemic silencing.

Results

Transgenic Nb-GFP^endo and Nb-GFP^cDNA plants

The GFP^endo construct was designed to resemble the structure of one of the highly transcribed small subunit of RIBULOSE-1,5-BIPHOSPHATE CARBOXYLASE OXYGENASE (RuBisCO) genes.⁴¹ The GFP^endo consists of the RuBisCO promoter/5′UTR/two introns/3′ UTR/terminator and three (arbitrarily chosen) GFP exons (Fig. 1). RuBisCO genes are highly transcribed endogenes that are not prone to RDR6-mediated processing.¹⁵ Previously, we could show that upon agroinfiltration, GFP^endo transcripts were inefficiently processed by RDR6-mediated pathway.⁴¹ In addition to transitivity, RDR6 is involved in the establishment of RdDM and systemic silencing.^15,42,43 We assumed that reduced RDR6-processivity of the agroinfiltrated GFP^endo construct could also have an impact on the efficiency of RdDM and systemic silencing of a genome-integrated GFP^endo construct. Thus, transgenic N. benthamiana GFP^endo plants (Nb-GFP^endo) were generated. For reasons of comparison, the N. benthamiana plant line 16C carrying a GFP cDNA transgene⁴⁴ was used. Line 16C (hereafter referred to as Nb-GFP^cDNA) is exceptional in that it shows remarkably high GFP expression without being prone to spontaneous cis-silencing.^36,39,40 Importantly, Nb-GFP^endo and Nb-GFP^cDNA plants exhibited almost similar GFP expression (Fig. 2, 3 and S1). Thus, any putative divergence of responses against RNA silencing of the two transgenes should reflect qualitative rather than quantitative effects.

Figure 1. Schematic representation of GFP^endo, hpF and GFP^cDNA constructs. P35S: CaMV 35S promoter; TNOS: nopaline synthase (NOS) terminator; GFP: green fluorescent protein; PRbc: tobacco RuBisCO promoter including 5′ UTR; TRbc: tobacco RuBisCO terminator including 3′ UTR; Intron 1 and 2: tobacco RuBisCO introns 1 and 2. The hybridization probes used for northern blot analysis (Fig. 3) and the 182 bp and 134 bp regions analyzed by bisulfite sequencing (Fig. 4, S2 and S3) are indicated by black bars.

Figure 2. GFP monitoring of hpF-agroinfiltrated GFP^endo and GFP^cDNAN. benthamiana plants. The development of RNA silencing was monitored at local levels (4 and 8 dpi) and systemic levels (21 and 46 dpi).

Figure 3. Northern blot analysis of RNA samples of hpF-agroinfiltrated GFP^endo and GFP^cDNAN. benthamiana plants. Large and small RNA analysis was performed at local levels (4 and 8 dpi) and systemic levels (21 dpi). The regions corresponding to hybridization probes are indicated in Figure 1. Ethidium bromide staining and miR159 accumulation served as loading controls. 'Mock' corresponds to 10 mM MgCl₂ infiltration.

Analysis of PTGS in locally silenced leaves

Endogenes and transgenes are prone to RNA silencing when an exogenous trigger e.g., hairpin RNA (hpRNA) is present and processed into siRNAs (trans-silencing).^45,46 In order to compare trans-silencing of GFP^endo and GFP^cDNA, a hairpin GFP (hpF) construct covering 139 bp of the GFP middle region (F) (Fig. 1), was agroinfiltrated into Nb-GFP^endo and Nb-GFP^cDNA plants. For each treatment, three plants were agroinfiltrated (three leaves for each plant) and the pooled tissue was used for subsequent analysis. Both, Nb-GFP^endo and Nb-GFP^cDNA plants revealed local PTGS already 4 days post infiltration (dpi) and displayed full local GFP silencing at 8 dpi (Fig. 2). SRNA-mediated silencing in plants results in mRNA degradation and/or translational inhibition.^10,39 In order to discriminate between the two processes, RNA was extracted from a pool of agroinfiltrated leaves and subjected to Northern blot analysis. Our data revealed that primary GFP siRNAs, originating from the agroinfiltrated hpF construct, targeted GFP^endo and GFP^cDNA transcripts for degradation (Fig. 3, probe F). In addition to the production of primary siRNAs, transitivity-derived secondary siRNAs may also be generated by RDR6/DCLs processing of abRNA. However, in samples of locally silenced tissue, no secondary 5′ and 3′ siRNAs were detectable under the applied conditions (Fig. 3, probes G and P). Secondary siRNAs production may be slow²³ and may take longer than 8 dpi. However, progressive physiological degeneration of agroinfiltrated tissue 8 dpi did not allow analysis at later stages.

Analysis of RdDM in locally silenced leaves

PTGS is tightly related to RdDM. In the case of transgenes, silencing of mRNA is generally associated with RdDM of its coding region, which in turn, may contribute to maintain PTGS.⁴⁶ Endogenes and transgenes are efficiently targeted for trans-PTGS. However, why only transgenes are prone to RdDM⁴⁶ is still elusive. In order to investigate if the endogene-resembling GFP^endo differs from GFP^cDNA in view of being a target for RdDM, the methylation status of the transgene constructs was analyzed by bisulfite sequencing using pooled samples from locally silenced leaf tissue (8 dpi) (Fig. 2). We found that GFP^cDNA was heavily methylated, not only in the region targeted by the hpF silencing trigger, but also in the 3′ downstream sequence (Figs. 4 and S2). This finding was in agreement with another study reporting on RdDM spreading.⁴⁷ Notably, the observed 3′ spreading of RdDM was not correlated with the accumulation of detectable amounts of 3′ siRNAs (Fig. 3, probe P). In contrast to the GFP^cDNA, GFP^endo was poorly methylated in locally silenced tissue (Figs. 4 and S3). This finding was striking, since the silencing triggers of Nb-GFP^endo and Nb-GFP^cDNA were identical (hpF). In addition, abundant primary hpF-derived siRNAs were detectable in Nb-GFP^endo and Nb-GFP^cDNA samples (Fig. 3, probe F). Thus, one would have expected RdDM of GFP^endo, at least, in the region targeted by hpF. These findings suggest that the mere presence of siRNAs was not sufficient to efficiently trigger RdDM.

Figure 4. Histogram representing bisulfite sequencing data from GFP^cDNA and GFP^endo. Non-agroinfiltrated, 8 dpi local and 21 dpi systemic tissue was used for analysis. 10–15 clones were sequenced and data were analyzed with CyMate software.⁶⁸ Histogram represents the data presented in Fig. S2 and S3. In non-agroinfiltrated GFP^cDNA, transgenerationally maintained CG methylation was detected, in agreement with previous observations.^41,70

Systemic PTGS analysis

Endogenes and transgenes are efficiently targeted for local PTGS, but only transgenes appeared to be prone to systemic PTGS.^25,37,48 Thus, we were interested to investigate if locally silenced Nb-GFP^endo plants also become systemically silenced. For this reason, 12 Nb-GFP^cDNA and Nb-GFP^endo plants displaying local silencing post hpF-agroinfiltration were monitored up to 46 dpi for systemic silencing. All Nb-GFP^cDNA plants exhibited systemic PTGS already 21 dpi (Fig. 2). Importantly, and in agreement with previous observations, only 3′ GFP siRNAs were detectable in systemically silenced leaves by Northern blot analysis (Fig. 3, probe P).^33,49 In contrast, 21 dpi and even 46 dpi no systemic GFP silencing was observed in Nb-GFP^endo plants. At 46 dpi Nb-GFP^cDNA plants were fully silenced (Fig. 2). We thus conclude that, similar to most endogenes, GFP^endo failed to induce the production of and/or failed to respond to systemic silencing signals.

Analysis of RdDM in systemically silenced leaves

In grafting experiments employing transgenic plants, silenced scions induced systemic PTGS and RdDM in the rootstock.^30,32,50 In a reverse experiment using a silenced rootstock, the scion became systemically silenced, but de novo methylation of the transgene was not detectable suggesting that the signals for induction of PTGS and RdDM are not identical.^33,51,52 In contrast to grafting experiments, we induced local silencing in leaves of individual Nb-GFP^cDNA and Nb-GFP^endo plants. DNA was extracted from a pool of newly emerged Nb-GFP^cDNA (almost fully silenced) and Nb-GFP^endo (non-silenced) leaves 21 dpi, respectively (Fig. 2). The DNA samples were subjected to bisulfite sequencing. The sequencing data showed that, as expected, the GFP^endo was not methylated (Figs. 4 and S3). In contrast, the coding region of the GFP^cDNA construct was heavily methylated (Figs. 4 and S2). Importantly, RdDM patterns were not restricted to the siRNA-mapping 3′ region but were established along the entire coding region (Fig. S2). Bisulfite sequencing data were validated by Southern blot analysis using the methylation sensitive restriction endonucleases, NcoI, DdeI and PvuII (Fig. S4). In summary, our data suggested that in a non-grafting system, systemic PTGS of GFP^cDNA is associated with efficient RdDM of the silenced transgene.

Discussion

The RdDM machinery is considered to be guided by 24-nt siRNAs.^53-55 In our experiments, primary GFP siRNAs were detected in samples of agroinfiltrated Nb-GFP^endo and Nb-GFP^cDNA leaves. However, GFP^endo was inefficiently targeted for RdDM. Thus, our data suggests that either the mere presence of siRNAs is not sufficient to trigger RdDM, or molecules, other than siRNAs may target genomic DNA for de novo methylation.^56,57 Establishment of RdDM is not only based on the presence of RdDM trigger molecules, but also on the transcription of RdDM targets (DNA and/or nascent transcripts). Typically, DNA-DEPENDENT RNA POLYMERASES IV and V (POLIV and POLV, respectively) are thought to transcribe RdDM targets, thus producing scaffold transcripts for the binding of trigger molecules.^13,58,59 In addition, RDR2 and RDR6 transcription of target transcripts appeared to be essential for the establishment and/or amplification of RdDM.^43,47 Differential POLIV/POLV/RDR2/RDR6 processing of GFP^endo and GFP^cDNA may also account for their differential methylation patterns we observed.

Production of systemic silencing signals does not require DCL1, DCL2, DCL3, DCL4, RDR2, RDR6 and POLIV,²⁹ indicating that systemic silencing is not triggered by a typical siRNA. However, upon reception, the response to systemic signals depends on RDR6, RDR2, DCL3 and POLIV.²⁹ Our previous data showed that, when agroinfiltrated into Nb-GFP^cDNA leaves, GFP^endo induced diminished local and delayed systemic silencing of GFP^cDNA.⁴¹ Here, we found that upon agroinfiltration of hpF, GFP^endo did not respond to systemic silencing signals. Since systemic silencing requires RDR6, one may speculate that GFP^endo does not produce RNA molecules that serve as a substrates for RDR6. Thus, similar to typical endogenes,³⁶ transgenes resembling endogenes seem not to respond to systemic silencing signals.

The diverse response of GFP^endo and GFP^cDNA to systemic silencing and RdDM indicated that the availability of RDR6 substrates, probably abRNA, is a critical step in these processes. RDR6 seems to prefer abRNAs that lack either 5′ cap and/or 3′ polyadenylation tail.^17,18 Most likely, both GFP^endo and GFP^cDNA produce abRNAs. In the case of GFP^cDNA, abRNAs may refer to truncated and/or read-through transcripts, lacking 5′ cap and/or 3′ polyadenylation tail, thus serving as optimal RDR6 substrates.^17,18 In the case of GFP^endo, abRNAs may in addition, include intron-containing mispliced and/or unspliced transcripts that were 'overlooked' by the spliceosome. We suggest that, while GFP^cDNA abRNAs are predominantly copied by RDR6 and induce RNA silencing, GFP^endo abRNAs are not processed by RDR6 but are instead eliminated via exonucleolytic RNA decay mechanism.^17,60,61 GFP^endo differed in several aspects from GFP^cDNA (promoter, UTRs, introns, terminator, integration site). In support of our suggestion, while the presence of UTRs and introns is positively correlated with RNA decay, it is negatively correlated with RNA silencing.^23,60,62-64 Spliceosome association likely excludes RDR6 recruitment.¹⁶ Thus, endogenous siRNAs matching the transcripts of intron-containing genes are underrepresented in sRNA libraries of Arabidopsis, while siRNAs derived from intronless gene transcripts are highly abundant.⁶³ Nevertheless, a couple of endogenous transcripts are targeted by RDR6 leading to the production of endogenous siRNAs.^34,65

In conclusion, we have shown that an endogene-resembling transgene is resistant to systemic silencing and DNA methylation, most probably due to its poor RDR6-mediated processing. The presence of each endogenous element (promoter, 5′ UTR, introns, 3′ UTR, terminator) may have its own contribution to this phenomenon, but we find of particular importance the presence of introns. Yet, investigating the role of each parameter was beyond the scope of our analysis and would have required the production of an enormous amount of independent transgene constructs comprising all free combinations of the individual RuBisCO gene elements. Moreover, such work would not include the introduction of the introns at different GFP coding region sides or the analysis of an alternative reporter gene. Our data suggest that endogene-resembling transgenes may point toward a new generation of transgenes, where high transcription rates would not be compromised by concomitant susceptibility to RNA silencing.

Materials and Methods

Generation of transgenic plants

The generation of pG104-GFP^endo construct has been previously described.⁴¹ Agrobacterium tumefaciens (strain ATHV) containing pG104-GFP^endo was used to transform N. benthamiana via leaf disc transformation, as previously described.⁶⁶ Since pG104-GFP^endo did not contain a selection marker, co-transformation with A. tumefaciens (strain GV3101) containing the binary vector pPCV702SM-Δ⁶⁷ was conducted for kanamycin selection. Independent T1 lines were characterized by Southern and northern blot analysis for copy number and expression levels of the transgene (Fig. S5). At least two independent GFP^endo lines (18 and 25) containing multiple transgene copies, displayed GFP expression levels and GFP fluorescence similar that of GFP^cDNA line 16C. Segregants of the GFP^endo line 25, containing 2–5 transgene copies (Fig. S5) were used for subsequent analysis.

RNA extraction and northern blot analysis

RNA was isolated with NucleoSpin miRNA kit (Macherey-Nagel, www.mn-net.com) according to the manufacturer's instructions. Large and small RNA fractions were separately isolated and used for mRNA and siRNA analysis. For mRNA analysis, 5 µg of large RNA fraction was analyzed in 1.2% agarose/formaldehyde gels, and blotted onto positively charged nylon membranes BioBond Plus (Sigma-Aldrich, www.sigmaaldrich.com) and UV_312nm-cross-linked (300 mJ/cm²). The PerfectHyb Plus 1x (Sigma-Aldrich) was used for overnight hybridization at 64 °C. Membranes were washed at 64 °C with buffer 1 (2x SSC, 0.1% SDS (w/v)) for 30 min and with buffer 2 (0.5x SSC, 0.1% SDS (w/v)) for 15 min. For siRNA analysis, 5 µg of small RNA fraction was analyzed in 15% TBE-Urea gels and blotted onto positively charged nylon membranes Bright Star (Life Technologies, www.lifetechnologies.com) by electro-blotting at 300mA for 1 h and UV_312nm-cross-linked (300 mJ/cm²). The PerfectHyb Plus 1x (Sigma-Aldrich) was used for overnight hybridization at 42 °C and membranes were washed twice with buffer 1 (2x SSC, 0.1% SDS (w/v)) at 42 °C for 30 min. For the generation of probes, templates obtained by PCR on GFP sequence were used in a random labeling reaction with DecaLabel DNA labeling kit (ThermoScientific, www.thermoscientific.com) with a-³²P dATP. For probe G (291 bp, 62% A/T context), the primers 5′-ATG AAG ACT AAT CTT TTT CT-3′ and 5′-ATC TGG GTA TCT TGA AAA GC-3′ were used. For probe F (144 bp, 47% A/T context), the primers 5′-CAT ATG AAG CGG CAC GAC TT-3′ and 5′-CTC GAT CCT GTT GAC GAG GG-3′ were used. For probe P (336 bp, 53% A/T context), the primers 5′-CTT AAG GGA ATC GAT TTC AA-3′ and 5′-TAG TTC ATC CAT GCC ATG TG-3′ were used. Probes G, F and P are continuous and cover the full GFP sequence. For the detection of miR159, the oligonucleotide 5′-AAG AGC TCC CTT CAA TCC AAA-3′ was used in an end-labeling reaction using T4 polynucleotide kinase (New England Biolabs, www.neb.com) and γ-³²P ATP.

DNA extraction and Southern blot analysis

To 1 g of grinded fresh leaf material, 4 ml extraction buffer (2% (w/v) CTAB, 20 mM EDTA, 1.4 M NaCl, 100 mM TRIS-HCl, pH = 8), 40 µl proteinase K solution and 8 µl RNase A were added, the mixture was incubated at 65 °C for 1 h then briefly cooled on ice. Four ml of chloroform:octanol (24:1) was added and the sample was centrifuged (6000 rpm, 15 min, 4 °C) on Rotanda 360R. To the upper phase, 400 µl CTAB/NaCl solution (10% (w/v) CTAB, 0.7 M NaCl) and 4 ml chloroform:octanol (24:1) were added and the sample was centrifuged (6000 rpm, 15 min, 4 °C). To the upper phase 4 ml precipitation buffer (1% (w/v) CTAB, 10 mM EDTA, 50 mM TRIS-HCl, pH = 8) were added and the mixture was kept overnight at 37 °C. The DNA was precipitated (6000 rpm, 15 min, 4 °C) and the pellet was resuspended in 1 ml high salt TE solution (1 M NaCl, 0.1 mM EDTA, 10 mM TRIS-HCl, pH = 8) at 65 °C for 30 min. The DNA was re-precipitated (6000 rpm, 15 min, 4 °C) with 1 ml isoporopanol and washed twice (centrifugation at 6000 rpm, 15 min, 4 °C) with 1 ml ethanol 70% (v/v). Finally, the DNA pellet was air-dried for 5 min and resuspended in water. For Southern blot analysis, 10 µg DNA were digested with the corresponding restriction endonucleases, separated onto 1% agarose gels and treated for 30 min with denaturation buffer (1 M NaCl, 0.4 N NaOH). DNA was capillary transferred overnight onto positively charged nylon membranes (BioBond Plus, Sigma-Aldrich, www.sigmaaldrich.com) with denaturation buffer. After transfer, the membranes were treated with equilibration buffer (1 M NaCl, 1 M TRIS-HCl, pH = 7), air-dried and UV_312nm-cross-linked (100 mJ/cm²). For the generation of probe 35S, templates obtained by PCR (primers 5′-CCG GAA ACC TCC TCG GAT TCC-3′ and 5′-GAT GAT CCC CCT CTC CAA ATG-3′) on 35S promoter sequence were used in a random labeling reaction with DecaLabel DNA labeling kit (ThermoScientific, www.thermoscientific.com) and a-³²P dATP. The PerfectHyb Plus 1x (Sigma-Aldrich, www.sigmaaldrich.com) was used for overnight hybridization at 64 °C. Membranes were washed at 64 °C with buffer 1 (2x SSC, 0.1% (w/v) SDS) for 30 min and with buffer 2 (1x SSC, 0.1% (w/v) SDS) for 15 min. Membranes were exposed to FujiFilm Imaging Plates (Fuji, www.fujifilm.com) for 2 d and scanned using PharosFX Plus PhosphorImager (BioRad, www.bio-rad.com).

Western blot analysis

Protein was extracted from Nb-GFP^endo and Nb-GFP^cDNA and subjected to western blot analysis with GFP antibody as previously described.⁶⁷

Bisulfite sequencing

For bisulfite sequencing, 500 ng DNA were pre-digested with DraI and subjected to bisulfite treatment using the EZ DNA Methylation-Lightning Kit (ZYMO Research, www.zymoresearch.com) according to the manufacturer's instructions. The recovered material was amplified by PCR with Zymo Taq DNA Polymerase (ZYMO Research, www.zymoresearch.com) with the primer pair 5′-TTT GTT YYA AAA AAA AAA GAA GAA GAA G-3′ and 5′-RRA RTT RTA RTT RTA TTC CAA CTT-3′ for GFP^endo and the primer pair 5′-CTT TRA TRC CRT TCT TTT RCT TRT C-3′ and 5′-TTY AAG GAY GAY GGG AAY TAY AAG A-3′ for GFP^cDNA at the following conditions: 10 min at 95 °C, 44 amplification cycles (95 °C for 30 s, gradient 45–47–51 °C for 30 s and 72 °C for 30 s), and 72 °C for 10 min. Gradient amplicons were gel excised with Qiagen gel extraction kit (Qiagen, www.qiagen.com), pooled and ligated into pGEM T Easy (Promega, www.promega.com). 10–15 clones were sequenced for each experiment. For interpretation of bisulfite sequencing data the CyMate software was used.⁶⁸

Agroinfiltration assay

Agrobacterium tumefaciens (strain GV3101) containing the binary vector pPCV702SM-GpG (hpF)⁶⁷ was used for N. benthamiana leaf agroinfiltration at an optical density (OD) of 1.0, as previously described.⁶⁹

GFP monitoring

Pharos FX molecular scanning (Biorad, www.bio-rad-com) was used for scanning of GFP fluorescence, as recommended by the manufacturer‘s instructions. Leaves were scanned by CY3 and FITC filter channels and the merged picture depicted GFP fluorescence.

Acknowledgments

This project was supported by the German Research Foundation (DFG) (grants: Wa1019/8–1). We thank Dr. Mirko Moser and Nora Schwind for their contribution in the production of transgenic plants.

10.4161/rna.29623