Molecular Characterization of a Silenced Transgenic Nicotiana Tabacum Line

Background RNA interference (RNAi), a natural regulatory mechanism involved in gene regulation, transgene silencing and virus defense in plants, has been widely used for improving the quality of crops and for conferring resistance to plant pests and pathogens. For the sustained application of RNAi-based genetically modied (GM) crops in agriculture, stable expression of the silencing-inducing hairpin transgene is important. As RNA interference may be accompanied by RNA-directed methylation of homologous DNA, including the RNAi transgene, gene body methylation may spread into the promoter and eventually lead to progressive transcriptional silencing of the RNAi inducing transgene. This would be detrimental for the effective use of RNAi crops and may have safety implications if food allergens or antinutrients are the target. Results repeat with transgene constructs separated by the bacterial transposon Tn5393 and to in two epigenetic to methylation. containing bacterial transposon be a possible molecular trigger for the observed silencing of transgene expression in this unstable GM line. epigenetic variant was subject to cross-hybridizations with homozygous F3 plants of three different transgenic Nicotiana tabacum lines. In addition backcrosses with non-transgenic N. tabacum var. W38 were performed. Young kanamycin resistant seedlings were transferred from the climate chamber to the greenhouse, potted into TKS 1 soil and grown at 24°C (14 h light, 10.000 lux). At owering time transgenic tobacco plants were emasculated and pollinated using pollen from the different transgenic lines or from non-transgenic N. tabacum var. W38. Mature seed pods were harvested and stored at room temperature. encompassing a bacterial Tn3 transposon and the chromatin status of GM line LX222GUS-A 9b may have been responsible for the initiation of DNA methylation independent of RdDM, while methylation at symmetrical C sites is then transmitted to successive generations during DNA replication by the action of maintenance methyltransferases.


Introduction
Genetically modi ed (GM) plants with new traits conferred by the expression of transgenes have been used in agriculture in many countries for more than two decades. Despite the rapid evolution and adoption of new plant breeding technologies allowing for targeted genome modi cations with increased precision [1], GM plants are still being developed for commercial applications. However, as has been frequently reported for GM plants, gene silencing might unexpectedly occur during the regeneration and testing phase, and sometimes only later during plant development or in transgenic progeny plants [2][3][4][5].
Several molecular features are known to act as possible triggers for transgene silencing, such as a suppressive chromatin structure at the integration site, the integration of multiple transgene copies, particularly in inverted-repeat (IR) orientation or transgene rearrangements [6][7].
IR integrations are often found associated with transcriptional gene silencing (TGS) as well as with post-transcriptional gene silencing (PTGS), both mediated by double-stranded (ds) RNA transcribed from the inverted repeat [7][8]. High transgene expression has been identi ed as another e cient trigger of PTGS through the action of a threshold-dependent RNA sensing mechanism [4,9]. Since PTGS and TGS are mechanistically linked [10], switches from one silencing state to the other are possible and have been reported to occur during long-term tissue culture and vegetative development of transgenic plants [11][12]. Switches from PTGS to TGS were shown to be correlated with progressive promoter methylation [11,13].
Post-transcriptional gene silencing triggered by small RNAs (siRNA or microRNA) may not only cause unintended transgene silencing, but is also the basis of RNA interference (RNAi), a natural eukaryotic gene regulatory mechanism. RNAi has been used for the production of GM crops with enhanced food quality traits such as the enrichment with essential amino acids or special fatty acids [14][15]. It has also been exploited to reduce food allergens and toxic metabolites [16][17] and to achieve resistance to plant pathogens and pests [18][19][20][21]. Targeted RNAi-mediated gene silencing is typically achieved by the introduction of gene-speci c antisense or IR constructs through stable plant transformation, resulting in the expression of silencing-inducing dsRNA or hairpin (hp) RNA. siRNAs derived from these bind to homologous target mRNA, leading to mRNA degradation and inhibition of translation. RNAi acting at the posttranscriptional level may also be associated with RNA-directed DNA methylation (RdDM) of homologous sequences [22] (Jones et al. 1999), and this may include the transcribed region of the silencing construct. If RNAi-induced DNA methylation spreads from transcribed sequences to the upstream transgene promoter due to the production of secondary siRNAs by an RNA-dependent RNA polymerase [23], a switch from PTGS to TGS will occur during plant development, and this will result in an undesirable shut-down of the expression of the silencing construct and phenotypic instability. Although it was found for two endogenous genes that virus-induced gene silencing (VIGS) was not accompanied by RdDM [22,24] and by target site spread through secondary siRNAs [23], unintended DNA methylation is seen as possible drawback of RNAi applications [25][26]. Loss of silencing has been reported during a eld experiment of T 3 generation Nicotiana attenuata plants carrying an acetyl-CoA-transferase 1 (acx1) silencing construct [13,27]. If shut-down of dsRNA production occurs, this will not only result in limited stability of the RNAi phenotype over generations [26], but may also raise biosafety issues in the case of RNAi plants with reduced allergens or toxins. With respect to the biosafety assessment of RNAi plants it is therefore important to get more insight into the mechanisms of RNAdirected DNA methylation in transgenic plants.
In previous work on gene expression and molecular analyses of GM Nicotiana tabacum plants transformed with a P35S-gus gene construct, we identi ed one line showing a high percentage of gene silencing in F1 progeny as well as progressive gene silencing during long-term tissue culture, accompanied by changes in the DNA methylation state [2]. In order to get more insight into the molecular mechanisms underlying these silencing phenomena, the structure of the T-DNA integration site and the DNA methylation patterns in epigenetic variants were analyzed. In addition, the potential of the silenced transgene locus to direct silencing of homologous active transgene alleles and DNA methylation in trans was investigated.

Greenhouse cultivation and cross-hybridization
To obtain hybrid progeny, a hemizygous F1 plant (F1-3) and a hemizygous T1 plant (T1-1) of the sh or tTC´respectivelythelong TC´ LX222GUS-A9b epigenetic variant was subject to cross-hybridizations with homozygous F3 plants of three different transgenic Nicotiana tabacum lines. In addition backcrosses with non-transgenic N. tabacum var. W38 were performed. Young kanamycin resistant seedlings were transferred from the climate chamber to the greenhouse, potted into TKS 1 soil and grown at 24°C (14 h light, 10.000 lux). At owering time transgenic tobacco plants were emasculated and pollinated using pollen from the different transgenic lines or from non-transgenic N. tabacum var. W38. Mature seed pods were harvested and stored at room temperature.
Genomic DNA isolation and molecular analyses Genomic DNA from plant leaves was extracted either according to [29] or using the NucleoSpin Plant II kit (Macherey-Nagel). Hybrid progeny plants were characterized with respect to the presence or absence of the LX222GUS-A9b transgene insert by Southern blot analysis or by differential polymerase chain reactions (PCR). For Southern blot analysis ten micrograms of plant DNA digested with the restriction enzyme HindIII were separated in a 0.8 % agarose gel and transferred to a Hybond-N (Amersham™/ GE Healthcare) nylon membrane [30]. Membrane blots were hybridized according to the manufacturer´s protocol with a digoxigenin (DIG)-labeled DNA probe (Roche Applied Science) speci c for nptII in order to detect left border fragments of individual LX222GUS transgenic lines. For differential PCR the primers 35S-2 (5´-GATAGTGGGATTGTGCG-3´) and Npt2 (5´-TAGCCAACGCTATGTCCT-3´) amplifying a 1920 base pair (bp) fragment were used to detect the LX222GUS-A9b insert, while GN151 (5´-TCAACTGACACCAAACTTAG-3´) and GN152 (5´-ACACAAACTGCTTATTATCT-3´) amplifying a 1096 bp fragment were used as primers for the detection of BI22SAR2E inserts. Ampli cation reactions containing 0.2 µM primers, 100 µM deoxynucleotides, 1.5 mM MgCl 2 , 1x GreenFlexi buffer and 1 unit GoTaq DNA polymerase (both Promega) were incubated after initial denaturation at 94°C in a thermocycler for 35 cycles with annealing temperatures between 52°C and 53°C and 5 min nal elongation at 72°C. PCR products were analyzed in 1.8 % agarose gels.

Analysis of the molecular structure of T-DNA integration
For the determination of DNA sequences at left border (LB) and right border (RB) integration sites of LX222GUS-A 9b, inverse PCR assays were performed. T-DNA primers to be used in inverse PCR were pretested by PCRs to ensure the presence of binding sites in the transformed plant DNA. For RB analysis DNA of à short TC´ plant was cut with the restriction enzyme AseI which has a recognition site about 500 bp from the RB, followed by re-ligation. The resulting fragment consisting of known transgene sequences and of unknown anking sequences was cut with a second restriction enzyme yielding an RB-containing fragment which allowed ampli cation of anking DNA sequences by using the T-DNA primers iPCR1 (5´-CATAAAGTGTAAAGCCTGG-3´) and iPCR2 (5´-GACTCCCTTAATTCTCCG-3´). For LB analysis DNA was restricted with MaeI, cutting about 490 bp from the putative left border, re-ligated and cut with StuI, before anking DNA was ampli ed using primers LB1.1 (5´-GTCCGCAATGTGTTATTAAG-3´) and iPCRLB (5´-CGATTGCTCATCGTCATT-3´). Cycling conditions for inverse PCRs were: denaturation at 94°C for 2 min; 40 cycles of 94°C for 30 sec, 52°C for 30 sec, 72°C for 2 min + 2 sec/cycle; nal elongation at 72°C for Recombinant A. tumefaciens strain LBA4404 bacteria transformed with binary vectors pLX222GUS or pBI22SAR2E were grown overnight at 28°C and 200 rpm in 10 ml LB medium supplemented with 100 µg/ml rifampicin and 5 µg/ml tetracycline. Cells were harvested by centrifugation, washed with 20 ml 10 mM 2-[N-morpholino]ethanesulfonic acid (MES), pH 5.7; 10 mM MgCl 2 and re-suspended in MMA medium (10 mM MES, pH 5.7; 10 mM MgCl 2 , 200 µM acetosyringone) to an OD 600 of about 1.0. Subsequently, the cultures were incubated for 2-3 h at room temperature.
Leaves No. 4 to No. 6 from the top of different Nicotiana tabacum plants were in ltrated. The bacterial suspensions were applied to the lower side of the leaves using a 2 ml syringe without a needle. Agroin ltrations were done in replicate for each plant and bacterial strain. Leaf samples were taken before in ltration and from in ltrated areas 6 days post in ltration for further testing.
GUS enzyme assays b-glucuronidase (GUS) expression was analyzed by a spectrophotometric enzyme assay according to Jefferson [32]. Samples from the sixth leaf from the top were collected from stably transformed plants. 100 mg leaf tissue was homogenized in 100 µl extraction buffer (50 mM sodium phosphate, pH 7.0, 1 mM EDTA, 0.1% Triton X-100). After centrifugation for 5 min at 15 800 g, 4°C the concentration of soluble protein in the supernatant was quanti ed using the BioRad Protein Assay [33]. For the spectrophotometric GUS assay 35 µg of total protein was used with 1 mM p-nitrophenyl-ß-D-glucuronide as substrate.
Reactions were stopped at 0, 30 and 60 min by adding 0.4 volumes of 1 M 2-amino-2-methyl-2,3-propanediol. Absorbance of the reaction product pnitrophenol (p-NP) was measured at 415 nm and calibrated against a dilution series of p-nitrophenol (Sigma).
DNA methylation analysis through bisul te sequencing DNA methylation was analyzed through bisul te conversion of unmethylated cytosine residues and subsequent DNA sequencing. DNA was digested with an appropriate restriction enzyme to allow full denaturation during the conversion step. Bisul te treatments were carried out using the EZ DNA Methylation-Gold TM Kit (Zymo Research) following the manufacturer's instructions. Bisul te-treated DNA was subjected to PCR ampli cations as described above.

Results
Switch from post-transcriptional gene silencing (PTGS) to transcriptional gene silencing (TGS) during prolonged tissue culture propagation As previously reported [2], about 60% of the transgenic progeny plants obtained from the transformant LX222GUS-A9b after short-term tissue culture (18 months) lost expression of the gus gene, while 40 % showed reduced expression under tissue culture conditions. When the LX222GUS-A9b primary transformant was vegetatively propagated over a period of 10 years, GUS expression was completely and irreversibly shut off, and this complete silencing was also found in all of the T1 progeny plants. In order to con rm the presumed switch from incomplete PTGS-based gene silencing to TGS we investigated the potential for trans-silencing of homologous transgenes. Trans-silencing mediated by small interfering RNA (siRNA) has been reported by Fojtova et al.
[36] for a post-transcriptionally silenced transgene locus, but not for the respective TGS variant. We therefore introduced homologous transgene constructs by agroin ltration into plants of both tissue culture (TC) variants. As shown in Fig. 1a, in the silenced `short TC´ plant F1-3 there was no expression from two different agroin ltrated homologous P35S-gus-Tnos gene constructs, con rming PTGS in trans. In contrast, no trans-effect on gus gene expression was found for the `long TC´ variant, similar to non-transformed W38 tobacco plants.
In order to investigate the in uence of sh or tTC´and long TC´ LX222GUS-A 9b variants on the expression of stably integrated homologous transgenes, cross-hybridizations of hemizygous plants with different non-silenced transgenic N. tabacum lines were carried out. As expected, there was a clear correlation between the presence of the `short TC´ LX222GUS-A9b epiallele and silencing of the gus gene (Fig. 1b). Silencing was found to be also correlated with the detection of siRNA homologous to the 3´-end of the gus gene (Fig. 1c). Unexpectedly and in contrast to the agroin ltration experiments, the presence of the LX222GUS-A9b `long TC´ epiallele affected the expression of a homologous transgene in hybrids. Following a temporary suppression, a general increase in expression usually occurred after transfer of cuttings to the greenhouse (Fig. 1d).
Switch from PTGS to TGS accompanied by progressive DNA methylation It has been previously shown that the 35S promoter and the 5´gus region of silenced `short TC´ LX222GUS-A 9b F1 plants were not methylated [2]. However, upon further analysis, we found the 3´gus-Tnos region of a silenced (F1-3) as well as of a GUS expressing (F1-1) F1 plant to be strongly methylated, in contrast to plants of stable GUS-expressing GM lines LX222GUS-A 6b and BI22SAR2E 16b (Fig. 2b). The DNA methylation status was inherited to the analyzed LX222GUS-A 9b F2 plant derived from F1-3. Trans-generational spread of methylation into the 5´-part of the gus gene was not observed for this epigenetic variant (data not shown). In contrast to `short TC´ plants and in accordance with the observed switch to transcriptional silencing, strong gus gene methylation and varying levels of methylation of the 35S promoter were detected in `long TC´ T 0 plants, and these epigenetic DNA modi cations were conserved in a leaf regenerant and inherited to its T1 progeny plants (Fig. 2c-d).

Transposon sequence identi ed at T-DNA integration site in line LX222GUS-A 9b
The progressively silenced transformant LX222GUS-A 9b contains two T-DNA copies that are tightly linked, since they could not be separated by outcrossing. However, no direct tandem or inverted repeat integration could be detected, when using conventional PCR analysis [2]. We therefore aimed to identify sequences at the integration sites to get clues on molecular features that may possibly cause unstable expression and silencing of the gus gene. Inverse PCR yielded a right border (RB) integration site fragment that had no sequence homology to the N. tabacum genome. However, BLASTn analysis revealed a 566 bp sequence adjacent to the right border that has 99% identity with Ti plasmid sequences in Agrobacterium tumefaciens strain LBA4404 encoding part of a Tn3 family transposase (Sequence ID: KY000037.1). It was previously shown that A. tumefaciens strain LBA4404 carries the Tn3 family transposon Tn5393 which was found to be introduced into the T-DNA of binary vectors, resulting in the presence of Tn5393 sequences in transgenic rice plants [31]. The 3´ end of the transposase gene (tnpA) located on the transposable element Tn5393, including the tnpA stop codon as well as the 81 bp 5´-terminal inverted repeat (5´-IR; IR-L) of Tn5393, was identi ed within the 566 bp sequence. PCR analysis using speci c primers from Tn5393 revealed that the transposon was completely integrated close to the right border of the second T-DNA copy. No second RB fragment could be identi ed by inverse PCR, however a 21 bp inverted repeat of RB sequences separated by 7 nucleotides was created during plant transformation. Adjacent to the 21 bp RB sequences of the second T-DNA is the entire Tn5393, terminated by the 81 bp 3´-terminal inverted repeat (3´-IR; IR-R), followed by continuing sequences from the right border of the second T-DNA. Figure 3 shows the molecular structure of the T-DNA insertion as well as details of the DNA sequences at the borders of Tn5393 and integrated T-DNA copies.

DNA methylation in hybrids of LX222GUS-A 9b
It was investigated whether siRNA-directed DNA methylation occurred in trans at homologous loci in hybrid plants containing the LX222GUS-A9b `short TCé piallele. Due to sequence differences between LX22GUS-A and BI22SAR2E transgenic plants in the region downstream of the nos terminator, differential methylation analysis was possible for hybrids obtained from crossings with BI22SAR2E 13a and BI22SAR2E 16b. As is shown in Fig. 4, the 3´gus-Tnos region remained methylated at the LX222GUS-A9b `short TC´ epiallele, mainly at symmetrical C positions, while no or very little DNA methylation was found for the BI22SAR2E transgene loci.

Discussion
Shut-down of GUS reporter gene expression during prolonged tissue culture accompanied by changes in DNA methylation Epigenetic changes within the transgene construct of the GM plant LX222GUS-A9b which is progressively silenced during tissue culture propagation have been investigated. Methylation predominantly at symmetrical C positions (Fig. 2b) that was found for the `short TC´ epigenetic variant in the most distal part of the gus gene (starting 230 bp upstream of the stop codon) and within the nos terminator suggests DNA methylation mechanisms, which at least partly are independent of siRNA. As is stated in a review by Matzke et al. [37], once established, methylation at symmetrical CG and CHG sites is copied during DNA replication by maintenance methyltransferases. One of these, chromomethylase 3 (CMT3) which catalyzes CHG methylation, interacts with repressive histone modi cations [38][39]. In line with this, Dalakouras et al. [40] reported that in stably integrated transgenes CG methylation was most prominent, while it has been proposed that establishment and maintenance of CHG methylation probably depend on structural features of the transgene insertion [41][42]. Philips et al. [43] discovered recently in Nicotiana benthamiana that progressive T-DNA methylation at promoter and gene body sequences in transiently agroin ltrated leaves occurred primarily in the CHH context and less at CHG and CG sites. These authors suggest that early T-DNA methylation prior to integration is one of the factors responsible for variable and unstable expression in GM plants. For LX222GUS-A9b there is no evidence for early T-DNA methylation at P35S and 5 gus gene sequences, however early methylation may have occurred in T-DNA sequences downstream of the gus gene. The methylation status of the 3´gus-Tnos region in the young primary LX222GUS-A9b transformant, which shows stable GUS expression during development, has not been determined. However, in the `short TC´ progeny plant F1-1 GUS expression is not silenced despite methylation in the 3´gus-Tnos region, and it is therefore unlikely that this methylation was triggered by siRNAs homologous to the 3´-part of the gus gene. It is also interesting to note that in plant F1-1 methylation in the 3´gus-Tnos region is largely restricted to CG sites. Methylation con ned to CG contexts can also be found within active genes and has been shown to be independent of RNA and to be mediated by maintenance methyltransferase MET1 [44]. In contrast to the young LX222GUS-A9b transformant and to F1-1, plant F1-3 expresses gus at a very low level, and this silencing is probably due to siRNA homologous to the 3´-part of the gus gene which may also trigger the initiation of additional DNA methylation in the 3´gus-Tnos region in CG, CHG and CHH sequence contexts. C-methylation in all sequence contexts is typical for RdDM [44][45]. PTGS through gus-speci c siRNAs in F1-3 is con rmed by the observed trans-silencing of homologous gus gene constructs (Fig. 1a-c).
Through prolonged tissue culture DNA methylation spread upstream into the 5´-part of the gus coding sequence and into the 35S promoter, causing a total shut-down of GUS expression. Progressive promoter methylation and silencing during prolonged in vitro cultivation or vegetative propagation of GM plants has been previously observed in potato and tobacco [11][12]. In the `long TC´ LX222GUS-A 9b primary transformant all analyzed copies of the 5´-gus gene region were extensively methylated at symmetrical and non-symmetrical C positions, and this methylation pattern was transmitted to a leaf regenerant and its T1 progeny pants (Fig. 2c). Weaker and more variable methylation was detected in the 35S promoter of the `long TC´ primary transformant and of its descendants obtained from secondary regeneration and sexual propagation (Fig. 2d). A spread of methylation from the 3´-part of a transcribed gene region into the 5´-gene region and the 35S promoter was reported in the study of Fojtova et al. [11] where a post-transcriptionally silenced N. tabacum line containing a T-DNA inverted repeat gave rise to an epigenetic variant after 24 months of in vitro cultivation. In the present study de novo methylation of 5´-gus sequences may have been caused by primary and secondary siRNAs, produced from aberrant transgene RNAs (particularly transcripts lacking 3´-polyadenylation) by RNA-dependent RNA polymerase 6 (RDR6) during a temporary PTGS phase, similar to what has been suggested by Bond and Baulcombe [46] for the new establishment of transposable element methylation. Aberrant transcripts have been shown to be strong inducers of PTGS of transgenes in plants [47]. Alternatively, as suggested by Fojtova et al. [12], spread of methylation may have been triggered by chromatin modi cation. Zemach et al. [39] found that in Arabidopsis thaliana most non-symmetric methylation of plant transposable elements is mediated by chromomethylase CMT2 and is facilitated by the nucleosome remodeler DDM1 independent of RdDM. The initial chromatin and methylation status of the integrated transgene DNA, probably determined by speci c features at the integration site, might have affected the progression of silencing observed in GM tobacco line LX222GUS-A 9b.
While GUS expression was silenced in progeny plants of the `short TC´ LX222GUS-A 9b transformant, silencing of the selectable marker gene nptII in these plants was not observed. In contrast, for `long TC´ progeny the ratio of kanamycin resistant and sensitive plants was 30:17 and deviated from the expected 3:1 ratio of progeny containing and not-containing the nptII gene, that was con rmed by PCR analysis. It is therefore likely that silencing eventually proceeds beyond the gus gene. In line with this, we found evidence for enhanced methylation of Pnos regulatory sequences in the `long TC´ primary transformant (data not shown).

Bacterial transposon as possible target and trigger for PTGS and TGS
In the present study, the bacterial tnpA gene as well as the complete bacterial transposon Tn5393 were identi ed in the RB anking region of the rst of two LX222GUS-A 9b T-DNA copies (Fig. 3) integrated into the plant genome. Tn5393 most likely originated from the A. tumefaciens strain LBA4404 that was used for plant transformation [31]. From PCR and DNA sequence analysis it is evident that Tn5393 was initially present within the second T-DNA, close to its right border. Upon inverted repeat T-DNA integration Tn5393 sequences were placed between right border sequences of the two T-DNA copies. There is no indication for additional transposon sequences integrated within the rst T-DNA copy. This is in line with ndings of Kim and An [31] who often detected normal T-DNA and Tn5393-carrying T-DNA in the same transgenic plant. These authors also found that in about 26% of 331 randomly selected transgenic rice lines transformed with binary vectors using LBA4404 Tn5393 was co-integrated. No bacterial transposon sequences were detected in the other GM lines in our study using different PCR primer combinations from Tn5393. When analyzing the widely used Nicotiana benthamiana line 16c, Philips et al. [48] found the 5Í R, the complete tnpA gene, the recombination region (res) and part of the resolvase (tnpR) gene of Tn5393 co-integrated with the T-DNA. These authors speculate that the bacterial transposon sequence adjacent to the reporter transgene may have an enhancing effect on the plant´s response to silencing inducers, making that line a sensitive tool for epigenetics research. Eukaryotic transposons are often targets for DNA methylation and silencing [49][50][51], and similarly bacterial transposon sequences transferred to plants could be nucleation centers for DNA methylation. It is therefore conceivable that in the case of the GM plant line LX222GUS-A 9b the Tn5393 sequences adjacent to the right borders of the T-DNA copies had an effect on the chromatin structure and on the methylation status of the integrated T-DNA. Unlike N. benthamiana line 16c which is a non-silenced line expressing green uorescent protein (GFP), GM tobacco line LX222GUS-A 9b contains two T-DNA copies and undergoes progressive gus gene silencing without an external silencing inducer. Silencing, however, does not occur immediately upon transgene integration, but only in the next generation or during extended tissue culture. The exact trigger for siRNA production and initiation of PTGS in this line is not known. It is likely that Tn5393 with its terminal IR sequences in combination with the inverted repeat T-DNA integration structure was involved in siRNA production and establishment of DNA methylation and presumably had an enhancing effect on the progression of silencing, as has been suggested by Philips et al. [47].

Effects on the expression and methylation of homologous transgenes
In the agroin ltration assays two different homologous transgene constructs introduced for transient expression of the gus gene were subject to gene silencing in a non-expressing `short TC´ F1 plant, con rming our suspicion that this variant conferred trans-PTGS via small interfering RNA (siRNA), similar to the trans-silencing described by Fojtova et al. [36]. The suppressive effect of the `short TC´ epigenetic variant was also found in hybrids carrying the LX222GUS-A 9b F1-3 locus obtained after cross-hybridization of three GM lines with homologous transgene sequences. The fact that in two of these lines the P35S-gus-Tnos gene construct was anked by a scaffold/matrix attachment region had no effect on the trans-silencing e ciency of the silencing-triggering locus. As is shown in Fig. 1b-c, there was a strict correlation between presence of the LX222GUS-A 9b epiallele, gene silencing, and the detection of siRNA. These observations con rm the non-heritable nature of PTGS in the absence of the silencing triggering locus. PTGS relies on siRNA as trigger for mRNA decay or translational inhibition, and this signal has to be produced via dsRNA from a silencing-inducing locus.
When the `long TC´ epigenetic variant was used as recipient for agroin ltration of homologous transgene constructs, no suppressive effect on the transient expression of the newly introduced gus gene was found, in accordance with the transcriptional silencing state of this epiallele. However, in hybrid progeny plants of this epigenetic variant variable trans effects on gene expression were observed (Fig. 1c), and this may be due to low levels of siRNA homologous to the gus gene which were detected in these hybrids (data not shown). 30-50% reduced expression of homologous transgenes conferred by a TGS epiallele has also been reported by Fojtova et al. [36].
Despite the presence of siRNA, mainly methylation at symmetrical CG and CHG sites of the 3´gus-Tnos region was detected in progeny plants derived from thè short TC´ plant F1-3 ( Fig. 2b and Fig. 4). This methylation pattern con rms the presumed predominance of maintenance methylation by MET1 and CMT3 in this epigenetic variant and the assumption that methylation in the 3´gus-Tnos region is not primarily mediated by siRNAs. In accordance with this, no or only negligible DNA methylation was found in the 3´gus-Tnos region of homologous transgene loci in hybrids. In contrast to these ndings, Fojtova et al. [36] observed that the PTGS variant of an inverted repeat transgene locus did not only confer silencing in trans to homologous transgene loci, but conferred transmethylation at both symmetrical and non-symmetrical C positions as well. As already mentioned above, the speci c integration structure encompassing a bacterial Tn3 transposon and the chromatin status of GM line LX222GUS-A 9b may have been responsible for the initiation of DNA methylation independent of RdDM, while methylation at symmetrical C sites is then transmitted to successive generations during DNA replication by the action of maintenance methyltransferases.

Statements And Declarations
Compliance with ethical standards Potential con ict of interests