DNA Elements Reducing Transcriptional Gene Silencing Revealed by a Novel Screening Strategy

Transcriptional gene silencing (TGS)–a phenomenon observed in endogenous genes/transgenes in eukaryotes–is a huge hindrance to transgenic technology and occurs mainly when the genes involved share sequence homology in their promoter regions. TGS depends on chromosomal position, suggesting the existence of genomic elements that suppress TGS. However, no systematic approach to identify such DNA elements has yet been reported. Here, we developed a successful novel screening strategy to identify such elements (anti-silencing regions–ASRs), based on their ability to protect a flanked transgene from TGS. A silenced transgenic tobacco plant in which a subsequently introduced transgene undergoes obligatory promoter-homology dependent TGS in trans allowed the ability of DNA elements to prevent TGS to be used as the screening criterion. We also identified ASRs in a genomic library from a different plant species (Lotus japonicus: a perennial legume); the ASRs include portions of Ty1/copia retrotransposon-like and pararetrovirus-like sequences; the retrotransposon-like sequences also showed interspecies anti-TGS activity in a TGS-induction system in Arabidopsis. Anti-TGS elements could provide effective tools to reduce TGS and ensure proper regulation of transgene expression. Furthermore, the screening strategy described here will also facilitate the efficient identification of new classes of anti-TGS elements.


Introduction
Like that of endogenous genes, the expression of transgenes and activity of transposons/invasive nucleic acids can be influenced epigenetically [1,2]. Transgene silencing can depend on the sequence homology of a transgene [1,2], the degree of iteration of the transgene at the inserted location [3,4], the chromosomal environment into which it is inserted [5], or the coincidence of some/all of these factors.
In plant research, transgene silencing has been categorized generally into two classes: transcriptional gene silencing (TGS) and post-transcriptional gene silencing (PTGS). TGS acts through prevention of transcription, and occurs mainly when the genes involved share sequence homology in their promoter regions [1,2]; PTGS acts through sequence-specific degradation of transcripts and is dependent on homology within transcribed regions [1,2]. TGS is widely recognized as a major hindrance to transgenic technology [6] because TGS arises spontaneously in transgenic plants and can be inherited in subsequent generations [3,7,8]. Therefore, the development of strategies that can prevent TGS is essential to the success of transgenic technology.
The fact that some transgenes undergo TGS while others do not [4][5][6] makes it conceivable that endogenous DNA sequences exist that actively determine the epigenetic TGS/non-TGS state of genomic regions. Transgene insert(s) showing TGS often force unlinked homologous promoters to be silenced in trans (trans-TGS) [9,10]. Also, trans-TGS occurs on promoters in secondarytransfected (supertransformed) transgenes that are homologous to a pre-existing silenced promoter and inserted at a different locus/ loci [11][12][13]. However, DNA sequences that uncouple transgenes from trans-TGS are unknown.
In this study, we sought to explore such genomic elements (hereafter referred to as anti-silencing regions-ASRs), which we hypothesized would actively protect a flanked transgene from TGS. First, we developed a novel screening strategy using a tobacco transgenic plant causing obligatory TGS when a transgene driven by the same promoter is supertransformed. We then used this strategy to isolate ASRs, based on their ability to suppress trans-TGS. We confirmed that one of the ASRs isolated in this study repressed TGS in all three assay systems tested, i.e., (1) a tobacco plant containing multiple transgene insertions and exhibiting obligatory TGS; (2) a tobacco plant carrying a homozygous 35S promoter-driven transgene in a single position, and showing TGS by increasing copy number of the promoter; and (3) an Arabidopsis line inducing obligatory silencing by transformation of an FWA genomic clone. Our findings also suggest that this ASR would exert anti-silencing activity in various plant species.

Selection of an Obligatory trans-TGS Trigger Plant
We contrived a novel system to screen for ASRs based on promoter-homology causing obligatory TGS ( Figure 1A). We first isolated a transgenic plant causing obligatory promoter-homologydependent gene silencing of a ''secondly''-introduced promoter. We have previously identified several TGS plants in transgenic tobacco transformed with an enhanced cauliflower mosaic virus (CaMV) 35S promoter::bcl-x L /ced-9 construct [14] (hereafter P35S). These silencing plants had multiple copies of the transgene, transmitted the silencing status to the next generation, and exhibited reactivation of the transgene(s) upon treatment with 5azacytidine-a potent inhibitor of DNA methylation-suggesting that the silencing mechanism was TGS. We then used these TGS plants as a resource to select an obligatory trans-TGS trigger plant ( Figure S1), and identified M66-9 that could be used for ASR screening (Figures 1A).

Isolation of ASR Candidates
Use of the obligatory trans-TGS plant, M66-9, allows us to isolate ASR candidates from DNA libraries constructed with selectable marker-harboring vectors. To construct a library as a resource for isolating ASR, we used genomic DNA fragments of Lotus japonicus and a binary vector, pTH4, that contains another enhanced P35S::HPT (hygromycin phosphotransferase gene) cassette ( Figure 1B).
If P35S::HPT constructs containing genomic fragments with no ASR activity are supertransformed into explants of M66-9, the P35S in the construct would be silenced and no supertransformant would be obtained on hygromycin-containing medium ( Figure 1A, bottom left). In contrast, supertransformants of M66-9 explants can be regenerated on the selection medium either if the supertransformed P35S::HPT construct is protected from trans-TGS by an adjacent ASR ( Figure 1A, bottom middle) or if supertransformed cells lose their trans-TGS activity (revertants) ( Figure 1A, bottom right). Because loss of trans-TGS activity will lead to reactivation of the pre-existing transgene (P35S::ced-9), revertants can be screened by detection of Ced-9 protein as an indicator ( Figure 1C). When about 1000 leaf pieces of M66-9 were inoculated with Agrobacterium containing the genomic library, a total of 30 independent supertransformed shoots were obtained ( Figure 1B). To confirm that the pre-existing transgene (enhanced P35S::ced-9) still showed TGS, Ced-9 expression was examined in the 30 supertransformants by protein gel blot analysis. Ced-9 protein was detected in 27 out of the 30 plants, suggesting that these plants were epigenetic/genetic revertants from the silenced state ( Figure 1C). We excluded these plants from further screening for ASRs because the original transgenes in these 27 plants had lost the trans-TGS activity ( Figure 1A, bottom right, and 1C). The other three supertransformants showed no expression of the original transgene ( Figure 1A, bottom middle, and 1C). These results indicated that the original transgenes in these three plants were silenced and that the plants had kept their trans-TGS activity, while the supertransformed P35S::HPT selectable marker genes in these plants were not silenced, suggesting that ASR activity was conferred by each genomic fragment inserted in the supertransformed construct ( Figure 1A). Using a primer set designed within the vector sequence (Table S1) for PCR, we then isolated DNAs from three supertransformants as candidate ASRs ( Figure 1D).

ASR Candidates Include a Sequence Derived from Retrotransposons
The primary structures of the three ASR candidates were found to correspond to portions of the following sequences ( Figure S2

ASR Candidates have no Enhancer Activity
We next examined whether these ASR candidates have enhancer activity and can overcome TGS. Each of the three ASR candidates was inserted at the 59 edge of a CaMV 35S minimal promoter::GUS (b-glucuronidase gene) cassette ( Figure  S4). None of the three genomic fragments inserted into this construct led to any significant increase in GUS expression (Table  S2), indicating that these ASR candidates have no enhancer activity.

Anti-silencing Activities of ASR Candidates in a P35Sdosage-dependent TGS System
To confirm the anti-silencing activities of the ASR candidates on different transgenes, we established a TGS-inducing system based on an increasing copy number of P35S regions (an inverse relationship between transgene copy number and expression level is often observed in homology-dependent TGS [3,4,9]). We had previously produced a transgenic tobacco plant (Figure 2A, NW7-24-4) harboring an enhanced P35S::LUC (luciferase gene) that exhibits a markedly high level of LUC expression [15]. Since this transgenic plant is a single-copy-inserted homozygote of the enhanced P35S::LUC construct, which itself carries two copies of the 35S enhancer region and one copy of the 35S promoter region, this plant has four copies of the 35S enhancer region and two copies of the 35S promoter region per diploid (Figure 2A). We supertransformed this enhanced P35S::LUC plant with pMLH2113-GUS (Figures 2A and S4). If this enhanced P35S::GUS construct is introduced into the enhanced P35S::LUC plant, the resulting supertransformants should have at least eight copies of the 35S enhancer region and four copies of the 35S promoter region (Figure 2A, CST, control supertransformant). When the enhanced P35S::LUC plant was supertransformed, about half of the resulting supertransformants showed an extremely low level of GUS expression ( Figure 2B, CST). In contrast, when wild type tobacco was transformed with pMLH2113-GUS, most of the transformants exhibited relatively high GUS expression ( Figure 2B, CT, control transformant), as reported with the same GUS cassette in a different vector [16]. The higher frequency of plants showing no/low GUS expression in the CST plants can be explained by TGS due to the increasing copy number of the 35S promoter in the supertransformants (Figure 2A and 2B, CST). Using this TGSinducing system (hereafter, P35S-dosage-dependent TGS system), we next examined whether the three ASR candidates show genuine anti-silencing activities. ASR-containing pMLH2113-GUS constructs (Figure 2A, right) were introduced into the enhanced P35S::LUC plant. In supertransformants containing an ASR candidate ( Figure 2B, ASR ST), the frequency of plants showing suppressed GUS activity was lower than that of supertransformed plants without ASR ( Figure 2B, CST), suggesting that these ASR candidates exhibit anti-silencing activity also in this TGS-inducing system. ASR102 ST and ASR602 ST clearly showed higher medians of GUS activity than the control transformants/supertransformants ( Figure 2B, CT and CST). The shortest sequence, ASR602 (171 bp), was chosen for further characterization.

DNA Methylation Status of Enhanced P35S Promoters Flanked (or not) with ASR602 in the GUS Construct in a P35S-dosage-dependent TGS System
DNA methylation in promoter regions is often associated with TGS [17]. We analyzed the DNA methylation status of the 35S promoter region fused to the GUS gene in supertransformants with or without ASR602 (Figure 2A and 2B, CST and ASR602 ST) using a PCR-mediated methylation assay. The 35S promoter region fused with the GUS gene (yellow bar of pMLH2113-GUS in Figure S4) was digested using methylation-sensitive restriction endonucleases followed by PCR amplification (Table S1). If the promoter region is heavily methylated, the DNA in this region should be resistant to these enzymes, and PCR products from the region will be amplified. In most of the supertransformants without ASR602, amplified PCR products were detected after digestion with methylation-sensitive restriction enzymes, suggesting that the 35S promoter regions fused with GUS are heavily methylated in the supertransformants without ASR602 ( Figure 2C, CST). On the contrary, amplified DNAs were barely detected in ASR602containing supertransformants ( Figure 2C, ASR602 ST). The results suggested that ASR602 might protect the adjacent promoter region from DNA methylation.

Anti-trans-TGS Activity of ASR602 in the P35S-dosagedependent TGS System
We next examined whether TGS is also induced at the preexisting P35S::LUC locus in the same supertransformants ( Figure 2D). Since supertransformation of the P35S::GUS should not alter the pre-existing P35S::LUC sequence, all the supertransformants should exhibit high levels of LUC expression. However, if the TGS induced by P35S::GUS supertransformation of the P35S::LUC plant is trans-TGS, expression of the pre-existing P35S::LUC would be affected ( Figure 3A, CST).
We found a considerable number of supertransformed plants showing low levels of LUC expression ( Figure 2D, left). In supertransformants without ASR602 ( Figure 2D, top left, CST/ Total), there was a significant positive correlation between P35S::GUS and P35S::LUC expression levels ( Table S3); most of the supertransformants with low GUS expression also showed extremely low LUC expression, indicating that supertransformation with P35S::GUS induced trans-TGS of P35S::LUC ( Figure 2D Plants harboring multiple copies of a transgene tend to have complicated transgenic loci [18] and/or undergo PTGS [2,15]. We measured the copy number of P35S::GUS to select supertransformants containing only a single copy of P35S::GUS ( Figure 2D, right, 1 copy) from all the plants examined ( Figure 2D, left, Total). To determine the copy number of P35S::GUS in the supertransformants, gene-dosage ratios of GUS/ LUC were measured with real-time PCR using genomic DNAs isolated from the supertransformants (Table S1). We re-plotted GUS and LUC expression data of the selected supertransformants ( Figure 2D, right, 1 copy). In supertransformants without ASR602 ( Figure 2D, top right, CST/1 copy), about half the plants (7/13) showed low expression levels of the pre-existing P35S::LUC and/ or the supertransformed P35S::GUS. In supertransformants with ASR602 ( Figure 2D, bottom right, ASR602 ST/1 copy), no plants (0/12) expressed low levels of the P35S::LUC and/or P35S::GUS transgenes. These results show that ASR602 effectively repressed

Anti-silencing Activity of ASR602 on TGS of the FWA Gene Promoter in Arabidopsis thaliana
To test whether ASR602 shows anti-silencing activity in different plant species and with different promoters, we transformed wild type Arabidopsis thaliana, ecotype Colombia (Col-0) using a genomic clone of the Arabidopsis flowering gene FWA with/ without ASR602 ( Figure S4). The endogenous gene is heritably silenced, and direct repeats at the 59-end of its transcribed region are methylated in most tissues; a hypomethylated fwa mutant shows ectopic overexpression of FWA, resulting in late flowering [19]. Wild-type plants transformed with an FWA genomic clone, including the direct repeats, silence the FWA transgene efficiently and do not alter flowering time [19,20].
If ASR602 has an anti-silencing activity, wild-type plants transformed with an ASR602-fused FWA gene should exhibit late flowering. Plants transformed with an FWA genomic clone without ASR602 (Figure 3, FWA) showed the same flowering time as vector control transformants ( Figure 3B, Vector) and wt Col-0 plants, indicating FWA transgene silencing. Plants transformed with an FWA genomic clone flanked with ASR602 (Figure 3, ASR+FWA) showed significantly later flowering compared with both transformants of FWA without ASR (Figure 3, FWA) and vector control transformants ( Figure 3B, Vector), indicating that ASR602 inhibits silencing of the FWA transgene. It would also appear that ASR602 might exert anti-TGS activity in the next generation ( Figure S5). Taken together, these results suggest that ASR602 exhibits cross-species TGS suppression activity.

Discussion
We have developed a novel screening system based on antisilencing ability in order to isolate regulatory DNAs reducing TGS. ASR602 is the first DNA element with interspecies-wide anti-silencing activity to be identified using this assay system. By employing three different trans-TGS trigger systems using different constructs and promoters, we showed that transgenes linked to ASR602 were able to circumvent trans-TGS. We identified ASR602, based on its activity to prevent a supertransformed P35S-driven construct from trans-TGS, as screening criterion ( Figures 1A) and confirmed the anti-TGS activity by the different TGS-inducing system using ''NW7-24-4'', the highly-expressed P35S::LUC tobacco plant (Figure 2A). It is not until the LUCtransgenic plant is supertransformed with a P35S-driven construct (e.g. the P35S::GUS construct, Figure 2A) that TGS occurs in the LUC plant. In this respect, the TGS-inducing system using the LUC plant is different from the ASR screening system using the trans-TGS plant, ''M66-9'', in which TGS already occurs at the pre-existing P35S::Ced-9 transgene loci (Figures 1 and S1B). In the TGS-inducing system (Figure 2A), it would appear that the supertransformed P35S::GUS construct become a TGS trigger locus(loci), causing not only TGS on the supertransformed construct itself but also trans-TGS on the pre-existing LUC locus after supertransformation, although the mechanisms that make the supertransformed construct a TGS trigger locus(loci) remain unknown. We speculate that ASR602 would first hamper TGS initiation on the supertransformed ASR602::P35S::GUS construct, simultaneously preventing the supertransformed construct from becoming a TGS trigger locus(loci), and consequently TGS would not occur on the pre-existing LUC locus.
ASR602, a species-specific sequence obtained from a genomic DNA library of L. japonicus, had no sequence characteristics similar to known scaffold/matrix attachment regions (S/MARs) using three prediction tools (Text S1). Because we have not assessed experimentally whether ASRs can influence chromatin structure, our results at present do not rule out the possibility that ASRs have S/MAR-like or insulator-like activities. S/MARs and barrier insulators have been identified as boundary elements in the genomes of several species, and can stabilize expression when used to flank transgenes. Such elements uncouple (trans)genes from position effects such as spreading of a heterochromatic region into adjacent euchromatic regions [21,22]. On the other hand, when fused to a transgene, ASR602 suppressed TGS caused by a silencer locus (loci) located on a separate chromosome(s)/region(s). This suggests that ASR602 acts as a ''cue'' to determine the active status of chromatin as controlled by the epigenetic regulation systems of the genome. Regulatory factors that directly bind to, or indirectly associate with, ASR602-like sequences could recruit activation factors and/or ward off repression factors [23]. The epigenetic status of the genome would be controlled by both boundary elements, which compartmentalize genomic regions to be regulated epigenetically, and ''cue elements'', like ASR602, that keep genomic regions active epigenetically (or prevent such regions from being inactive). Genomic elements with anti-silencing activity, like ASR602, might be abundant in the genomes of eukaryotic organisms. Further larger-scale screening could lead to the isolation of many other anti-silencing elements with various characteristics/structures.
An intriguing possibility is that ASR602-the first reported ''cue element'' to determine the epigenetic active status of chromatinhas functions similar to boundary elements. Although two DNA sequences with barrier insulator activity have been identified previously from Drosophila Ty3/gypsy retrotransposons [24,25], their anti-trans-TGS activities have not been determined. Although functions of ASR602, which originated from a Ty1/copia retrotransposon, as a boundary element have not been determined experimentally, supertransformed P35S::LUC tobacco plants harboring single copy of the ASR602::P35S::GUS construct ( Figure 2D, bottom right, ASR602 ST/1 copy) included no supertransformants showing TGS, suggesting that no position effect on the GUS transgene occurs in these supertransformants (ASR602 ST/1 copy), and that the function of boundary elements in preventing position effects and those of ''cue elements'' like ASR602 could overlap rather than be mutually exclusive. The screening strategy described here will facilitate the efficient identification of novel anti-TGS elements.
The mechanisms underlying the anti-silencing activity of the ASRs isolated in this study remains to be determined in further investigations, and will certainly be intriguing. Protection from DNA methylation of promoters linked with ASRs might be crucial to the underlying mechanisms ( Figure 2C).
We speculate that some retrotransposons/retroviruses have evolved specific sequences/activities to break down host defense systems. PTGS forms a major part of the defense system of plants against RNA viruses, and many plant RNA viruses can combat defense systems using viral PTGS suppressors [1,2]. Some plant DNA viruses have also been suggested to have the ability to counteract the TGS responses of their hosts [26]. Since TGS forms part of the genome surveillance system that prevents multiplication of parasitic DNAs/transposons [1,9,27], it is conceivable that TGS poses a threat to retrotransposons in host genomes, and that, in response, some retrotransposons have evolved counter-defense activities.
The concepts developed for this screening strategy could be applied to the isolation of ASRs from a wide range of organisms. ASRs isolated by this method could lead to significant breakthroughs in transgenic technology, and analyses of ASRs will open up new research topics in the field of epigenetic mechanisms of genome regulation.

Plant Materials
Plants and cultured cells of wild type (Nicotiana tabacum cv. Samsun NN) and transgenic tobacco were grown as described [15]. To select a plant showing trans-TGS activity, plants from transgenic tobacco lines M65 and M66 were used [14]. The LUCtransgenic tobacco plant, ''NW7-24-4'', is described elsewhere [15]. Agrobacterium-mediated transformation of tobacco and Arabidopsis thaliana was carried out as described elsewhere [15,28]. Flowering time of Arabidopsis ecotype Col-0 and the transformants was measured by counting the number of rosette leaves of plants grown under long-day conditions (16 hr light/8hr dark). Flowering time data shown in Figure 3 were compared with Dunn's multiple comparison procedure [29]. The other statistical methods were carried out according to a standard textbook of biostatistics [30], except as otherwise noted.

Biochemical Analysis
Genomic DNA extraction from tobacco and Arabidopsis was carried out as described [15]. Genomic DNA of Lotus japonicus (accession Miyakojima MG-20) was provided by Y. Umehara (NIAS). A polyclonal antibody to Ced-9 was used for immunoblotting [31].

Plasmid Construction
pTH1, which contains a P35S::HPT (hygromycin phosphotransferase gene) as a selectable marker, has the same features as the pTH2 vector [32] except that the direction of the HPT expression cassette is reversed. pMLH7133-GUS-a derivative of pMLH7133-mwti1b (which contains HPT) [33]-was constructed by replacing the mwti1b gene with the GUS reporter gene of the promoter-GUS cassette in pE7133-GUS [16]. To generate pTH4, the HPT expression cassette of pMLH7133-mwti1b (SmaI-SmaI region) [33] was inserted into the XhoI site (flush-ended with Klenow fragment) of pTRA415(R)-delNPT [34]. To create the Lotus genomic library, genomic DNA was digested with HindIII and ligated into pTH4 ( Figure S4).
To generate pP35Sm-GUS, the CaMV promoter region of pBI121 [35] was replaced by the -46 CaMV 35S minimal promoter (P35Sm) [36]. ASR candidates were inserted into the HindIII site located immediately upstream of P35Sm in this construct and tested for enhancer activity ( Figure S4). pMLH2113-GUS was derived from pMLH7133-GUS by replacing the promoter-GUS cassette with the promoter-GUS cassette of pE2113-GUS (HindIII-BamHI fragment) [16].
To generate a 'control' construct for FWA transformation (pBI-FWA, Figure S4), a 5.5-Kb fragment of the FWA gene, which corresponds to positions 13037026 to 13042519 in GenBank accession NC_003075, was PCR-amplified (with primer set 'UP FWA51' bearing an introduced HindIII site, and 'FWA32' bearing an endogenous EcoRI site; Table S1) from genomic DNA of Arabidopsis (ecotype Col-0). The promoter-GUS reporter cassette of pBI121 (HindIII-EcoRI fragment) [35] was replaced with the PCR product digested with HindIII and EcoRI. To generate the ASR602-containing construct for FWA transformation, a 172-bp region between the introduced HindIII site and an endogenous XbaI site of pBI-FWA was replaced with an ASR602 fragment with an introduced XbaI site to which the 39 end of the original ASR602 (171 bp HindIII fragment) was converted by PCR.

Identification of Eight ASR602-containing Retrotransposon-like Sequences (ASLs)
First, we used the DNA sequence of ASR602 (171 bp) as a query sequence for BLASTN analysis on the GenBank web site (http://blast.ncbi.nlm.nih.gov/Blast.cgi) with default conditions, except using 'Database' = 'Others (nr etc.)', resulting in a list of accession numbers including ASLs based on E-values. Second, we used sequence data of the 'top 20' accession numbers to find all ASLs with the 20th E-value or with an E-value less than the 20th's. Third, in such ASLs (E-values#the 20th's E-value), we looked for ASLs that were sandwiched between a pair of LTRs with target site duplication. Finally, we identified the eight ASLs shown in Figure S2C. Eight ASLs (ASL1-ASL8) were found in sequence data of following accession numbers in order, respectively: AP010575, AP006141, AP009783, AP009654, AP009679, AP010505, AP006114 and AP004952.

GUS and LUC Assays
Reporter gene assays were performed as described [36]. An aliquot of cell extract was incubated in buffer containing 4methylumbelliferyl-b-D-glucuronide at 37uC for 30 min. To measure GUS activity, the quantity of 4-methylumbelliferone formed was determined using a fluorescence spectrophotometer (F-2500; Hitachi High-Tech Co., Ltd.). A 20-ml aliquot of the supernatant was incubated in 50 ml of PicaGene TM (a solution for LUC assay; Toyo B-Net Co., Ltd.). LUC activity was determined for 10 sec using a Microtiter Plate Luminometer MLX TM (Dynex).

Qualitative PCR Assay to Detect DNA Methylation Using Methylation-sensitive Enzymes
The basic concept of this assay is described elsewhere [37]. Tobacco plants analyzed for methylation status were chosen randomly from among supertransformed tobacco plants with/ without ASR602 ( Figure 2B and 2D, CST and ASR602 ST). Genomic DNAs were isolated individually and digested with a methylation-sensitive restriction enzyme: AluI, HgaI or XmnI. Each DNA sample (100 ng) was digested with 0.5 units of each restriction enzyme for 2 hr according to the manufacturer's instructions, followed by ethanol precipitation. For PCR amplification, 75 ng of each digested DNA was used as template with AmpliTaq Gold DNA Polymerase (Applied Biosystems) in 10 ml reactions, according to the manufacturer's recommendations. The cycling conditions were as follows: 95uC for 9 min; 40 cycles of 30s annealing at 55uC, 1 min elongation at 72uC, and 30 s denaturation at 95uC; 72uC for 5 min. The 35S promoter region fused with the GUS gene (yellow bar in pMLH2113-GUS in Figure  S4) was PCR amplified with the primer set 'pBI-Hind3-51' and 'GUSI3', which correspond to a sequence 59 upstream of the promoter region and a sequence at the 59 end region of GUS, respectively (Table S1). Half of the volume of each PCR product was subjected to gel-electrophoresis.
Quantitative PCR Assay to Determine Transgene Copy Number in Supertransformants pMLH2113-GUS construct copy number was quantified by measuring gene-dosage ratios of GUS/LUC in each supertransformant, using real time PCR as described [38]. DNAs were isolated individually from all of the transformed and supertransformed plants whose GUS and LUC activities were measured. The GUS and LUC genes were PCR amplified with two primer pairs (GUS: 'GUSI51' and 'Tnos Yomeru'; LUC: 'LUCI51' and 'Tnos Yomeru'), respectively (Table S1).

Accession Number
The sequence of ASR602 reported in this paper has been deposited with the DDBJ database under accession number AB632368. Figure S1 Identification of an obligatory trans-TGS trigger plant. (A-C) Identification of an obligatory trans-TGS trigger plant, M66-9, by supertransformation with a P35S-driven HPT gene. To select plants showing trans-TGS activity, plants from transgenic tobacco lines M65 and M66 were used [14] (A). M65 and M66 lines harbor the open reading frames bcl-xL, and ced-9, respectively [31], inserted into the expression vector pBE2113 [16]. pBE2113 has two copies of the enhancer region (El; -419 to -90) of the CaMV 35S promoter (P35S), and one copy of the core promoter (-90 to -1) of P35S followed by a gene of interest to be expressed. We have previously identified several TGS plants in M65 and M66 [14]. To select a plant showing obligatory trans-TGS activity, we supertransformed explants of these TGS plants with the P35S::HPT construct pTH1 ( Figure S4) (A and B). If any of these TGS plants had a potent trans-TGS activity, pTH1 would be silenced and supertransformation would lead to neither callus induction nor shoot regeneration from the explants on selective medium. Indeed, such a line was found: no callus or shoots were obtained from TGS plant M66-9 upon supertransformation with pTH1 (C), suggesting that M66-9 has trans-TGS activity. The red circles in C indicate regenerating shoots. (D) M66-9 becomes infected with Agrobacterium and confers obligatory trans-TGS on another 35S promoter-driven construct. We supertransformed M66-9 with another P35S-driven construct, pMLH7133-GUS, which contains the HPT and GUS genes, each driven by an enhanced P35S ( Figure S4). At two days after infection with Agrobacterium containing this construct, M66-9 explants show GUS staining spots. However, the explants no longer showed any GUS spots at seven days after infection, nor did they regenerate any supertransformed shoots, indicating that the GUS construct is introduced into M66-9 explants but subjected to trans-TGS thereafter. M66-9 was thus identified as an obligatory trans-TGS triggering plant. (TIF) Figure S2 Primary structure of three ASR candidates (see also Text S1). Amino acid positions of Tnt1 and copia (in parentheses) refer to those used in the accession numbers. (TIF) Figure S3 Chromosomal distribution of Lotus DNA sequences similar to ASR602. (A) Ratios of accession numbers containing ASR602-like sequences to accession numbers at http://www.kazusa.or.jp/lotus/clonelist.html. ''Contigs harboring ASR602-like sequences'' (in red) consist of accession numbers including one or more ASR602-like sequences. ''Contigs without ASR602-like sequences'' (in blue) consist of accession numbers that do not contain ASR602-like sequence. n, number of nonredundant accession numbers assigned to each chromosome or categorized as unmapped contigs. (B) Sequences similar to ASR602 were mapped on a genetic map of L. japonicus (http:// www.kazusa.or.jp/lotus/clonelist.html) at 1 centimorgan (cM) intervals. To find similar sequences, the ''bl2seq'' search (a specialized BLAST to align two sequences on http://blast.ncbi. nlm.nih.gov/Blast.cgi) was used with default parameters with the ASR602 sequence as the query and with each non-redundant accession number on http://www.kazusa.or.jp/lotus/clonelist. html as the subject. The identified ASR602-like sequences were plotted at the map position (cM) (http://www.kazusa.or.jp/lotus/ clonelist.html) of the accession number used as the subject. Filled circles, E-value ,10 -20 ; Open circles, E-value .10 -20 . Horizontal lines and vertical bars under the x-axes indicates regions containing centromeres and the genetic markers used as in situ hybridization probes [a marker on chromosome 1 (the dashed part) is not mapped on the genetic map], respectively [39]. (TIF) Figure S4 Binary vector constructs used for this study (see also Materials and Methods). pTH1 was used to supertransform TGS plants for selection of plants showing trans-TGS activity. This vector contains the CaMV 35S enhancer region (E35S) and promoter region (P35S) of pBI121, followed by HPT. RB and LB, right and left borders of T-DNA of Agrobacterium tumefaciens Ti plasmid, respectively. E35S, 59-upstream sequence of CaMV 35S promoter (-940 to -90). P35S, 59-upstream sequence of CaMV 35S promoter (-90 to -1). HPT, hygromycin phosphotransferase gene (a selectable marker). Tnos, polyadenylation signal of the nopaline synthase gene (nos) in the Ti plasmid. Tet R , a tetracycline-resistance marker gene. pMLH7133-GUS was used as a second CaMV 35S promoter-driven construct to confirm trans-TGS activity of the trans-TGS plant M66-9. The GUS expression cassette in this construct contains seven copies of the CaMV 35S enhancer (E7) and the P35S. Pnos::NPTII, nos promoter-driven neomycin phosphotransferase gene (a selectable marker) that confers resistance to kanamycin. E7, 59-upstream sequence of CaMV 35S promoter (-940 to -290) and (-290 to -90) x 7. V, 59-untranslated sequence of tobacco mosaic virus. In, first intron of a phaseolin gene. GUS, b-glucuronidase gene (a reporter gene). T35S, polyadenylation signal of the CaMV 35S transcript. pTH4 was used to construct the genomic library of L. japonicus for ASR screening. This vector contains two copies of the CaMV 35S enhancer (El) and the P35S. El, 59-upstream sequence of CaMV 35S promoter (-419 to -90). pP35Sm-GUS was used to examine whether an ASR candidate has enhancer activity. This construct harbors the -46 CaMV 35S minimal promoter region. pMLH2113-GUS was used to supertransform the LUC tobacco plant ( Figure 3A, NW7-24-4). This construct carries two copies of El and the P35S. The yellow bar depicts the region where the methylation status was analyzed by methylation-sensitive restriction enzyme-coupled PCR assay ( Figure 3C). Two arrows flanking the yellow bar (pBI-Hind3-51 and GUSI3) indicate the primer set used for the PCR assay (Table S1). pBI-FWA and an ASRcontaining pBI-FWA were used to examine if ASRs prevent silencing of the FWA transgene in Arabidopsis. A 59 end portion of the FWA region cloned in pBI-FWA was replaced with the ASR602 sequence, resulting in ASR602-containing pBI-FWA. Both FWA constructs contain the SINE-related tandem repeats of FWA gene (TR), which are sufficient to trigger de novo DNA methylation [40,41].    Table S3 Pearson's correlation between the P35S::GUS and the P35S::LUC expression levels in Figure 2D.

(DOC)
Text S1 Primary structure of three ASR candidates. (References for supporting information are compiled as ''Supporting Information References'' on the last page of the.doc file Text S1. (DOC)