A pair of transposons coordinately suppresses gene expression, independent of pathways mediated by siRNA in Antirrhinum


Author for correspondence:

Yuji Kishima

Tel: +81 11 706 2439

Email: kishima@abs.agr.hokudai.ac.jp


  • Our knowledge is limited regarding mechanisms by which transposable elements control host gene expression. Two Antirrhinum lines, HAM2 and HAM5, show different petal colors, pale-red and white, respectively, although these lines contain the same insertion of transposon Tam3 in the promoter region of the nivea (niv) locus encoding chalcone synthase.
  • Among 1000 progeny from HAM5 grown under the preferred conditions for the Tam3 transposition, a few showed an intermediate petal color between HAM2 and HAM5. Transposon tagging using these progeny identified a causative insertion of Tam3 for the HAM5 type (white) petal color, which was found 1.6 kb downstream of the niv gene.
  • Insertion of Tam3 at the position 1.6 kb downstream of niv alone showed nearly wildtype petal pigmentation, and the niv expression reduced by only 50%. Severe suppression of niv observed in HAM5 required interaction of two Tam3 copies on either side of the niv coding sequence. DNA methylation and small interfering RNAs (siRNAs) were not associated with the suppression of niv expression in HAM5.
  • Insertion of a pair of transposons in close proximity can interfere with the expression of gene located between the two copies, and also provide evidence that this interference is not directly associated with pathways mediated by siRNAs.


Transposable elements (TEs) that proliferate in a disorderly manner result in repetitive sequences in the genome (Doolittle & Sapienza, 1980; Orgel & Crick, 1980). Host genomes have developed mechanisms to prevent TE activities, which disturb genomic integrity under an antagonistic relationship between TEs and the genome. Such antagonistic relationships may have been elaborated into various modification systems that influence gene expression (Fagard & Vaucheret, 2000; Uchiyama et al., 2009; Lisch & Bennetzen, 2011). Representative modification systems, including DNA methylation, RNA interference (RNAi), and chromatin modification, have been explored using phenomena involved in the suppression of TE activity in plants (Slotkin & Martienssen, 2007; Lisch & Bennetzen, 2011). These mechanisms, described as epigenetic modifications, are mostly mediated by small RNAs, which confer pleiotropic effects involving interference with different stages of transcription, or even translation (Brodersen & Voinnet, 2006; Xie & Qi, 2008; Liu & Paroo, 2010). However, there is limited knowledge regarding the alteration of gene expression caused by mechanisms based on the conflict between TE activity and the activity of the host genome.

The genome of Antirrhinum majus contains c. 50 copies of Tam3 (Kishima et al., 1999), which belongs to the hAT family of transposons (Calvi et al., 1991). Tam3 is considered an autonomous element. Among the Tam3 copies in the genome, active copies are known to have an identical structure of 3.6 kb of sequence encoding the transposase (TPase) enzyme (Kishima et al., 1997, 1999). The transposition of Tam3 is controlled by temperature; low temperatures of c. 15°C permit transposition, but high temperatures of c. 25°C strongly inhibit it (Harrison & Fincham, 1964; Hashida et al., 2003, 2006) (Fig. 1). We have studied the effect of Tam3 on the expression of the nivea (niv) locus, which encodes chalcone synthase (CHS), and is responsible for flower pigmentation (Sommer & Saedler, 1986). The nivearecurrens::Tam3 (nivrec) allele from the HAM2 line carries a Tam3 insertion 70 bp upstream of the niv transcription start site (29 bp upstream of the TATA box) and produces variegated spots because of Tam3 excision at low temperature (Sommer et al., 1985). The nivrec allele shows a pale-red background flower color that was described as a ‘flushed flower’ by Harrison & Carpenter (1973). This leaky phenotype caused by the so-called ‘Tam3-permissible allele’ is a common feature in Antirrhinum strains that contain Tam3 in their promoter regions (Uchiyama et al., 2009). These Tam3-permissible alleles share the same Tam3 orientation (the 3′ end of Tam3 adjacent to the 5′ region of the gene coding sequence), whereas the reverse orientation of Tam3 in the promoter region that is observed in the pallidarecurrens::Tam3 (palrec) allele in Antirrhinum does not permit gene expression (Uchiyama et al., 2009).

Figure 1.

Different flower pigmentation patterns in Antirrhinum majus lines HAM2 and HAM5. Both the lines exhibit petal variegation at low temperature (15°C) but not at high temperature (25°C). The variegation was attributed to Tam3 in the niv promoter, which exhibits low-temperature-dependent transposition (LTDT). HAM2 shows a pale-red background flower color in either condition, while HAM5 shows a white petal color at high temperature and faint pigmentation at low temperature.

The HAM5 line has the same Tam3 insertion as the nivrec allele, although this flower shows no background petal pigmentation in Tam3 inactive conditions. Despite the different background colors of the petals, both the HAM2 and HAM5 lines produce similar variegation in the form of full red-coloured sectors because of Tam3 excision from the niv promoter at low temperature (Fig. 1). We examined the mechanism responsible for the difference in the background colors of the HAM2 and HAM5 lines. During the course of this analysis, we isolated individuals with background petal pigmentation similar to HAM2 in the progeny of HAM5 plants grown at low temperature.

In this paper we report identification of another Tam3 copy present 1.6 kb downstream of the niv poly-A signal, insertion of which is responsible for the lack of background pigmentation in HAM5. The mechanisms by which insertion of the Tam3 element results in the different pigmentation patterns were investigated, and possibilities associated with DNA methylation and RNAi were ruled out. Our results suggest that two repeat sequences separated by a short distance interact with each other and disturb the expression of the gene between the repeats. Possible interactions between the repeats and the alteration of host gene expression are discussed.

Materials and Methods

Plant materials

The Antirrhinum majus L. line, HAM2, originated from John Innes Centre (Norwich, UK) stock JI: 98 line (nivrec), and the HAM5 line (nivbeni) was isolated from the progeny of the H101 line (JI: 98) carrying the nivrec allele crossed to a different niv line in the John Innes Centre collection (Supporting Information, Fig. S5). Flowers of HAM2 and HAM5 exhibited pale-red and white petals, respectively, when the plants were grown at 25°C, while both lines started to exhibit variegated spots on their petals when the plants were grown at 15°C (Fig. 1), because Tam3 excised from the niv locus. Plant material was grown initially for 2 months at 25°C, and subsequently transferred to a 15°C growth chamber or grown continuously at 25°C. To obtain Tam3-excised individuals, self-pollinated seeds from the HAM5 plants grown at 15°C for at least 2 months were collected, and c. 1000 self-pollinated seeds were sown to screen for variants with different petal pigmentations. These variants fell into two petal pigmentation phenotypes; one was an intermediate type between HAM2 and HAM5 (the plants showed variegated spots when the plant was grown at 15°C) and the other type showed the full pigmentation without variegation. The progeny from the former type segregated into HAM5 type, intermediate type and HAM2 type in the ratio 1 : 2 : 1 (Fig. 2), and these variants were subjected to transposon tagging analysis (Figs 3, S1). DNA was extracted from young leaves (3–4 cm in length) of Antirrhinum plants in accordance with the procedure described by Uchiyama et al. (2008). RNA was extracted from the young leaves in accordance with Martin et al. (1989). The HAM1 line (JI: 7) was employed throughout this study as a line carrying homozygote of the wildtype niv allele.

Figure 2.

Flower pigmentation phenotypes in a F2 population between Antirrhinum majus lines HAM2 and HAM5. The F2 progeny from a cross between HAM2 and HAM5 segregated intermediate phenotypes in addition to the parental types. The numbers of individuals obtained agreed with an expected ratio of 1 : 2 : 1 at < 5%.

Figure 3.

Segregation of progeny of germinal revertants from Antirrhinum majus HAM5 plants. (a) The germinal revertants obtained from HAM5 plants grown at low temperature had segregating progeny as observed in the F2 population from a cross between HAM2 and HAM5 (Fig. 2). (b) Hybridization experiment using Tam3 as a probe showed a 6 kb fragment (with a red dot) linked to the phenotypes in the segregating progeny.

Quantitative real-time PCR

Approximately 200 ng of DNase-treated total RNA from the flower lobes in the flower bud with three biological replicates (c. 2 cm in length) from wildtype, HAM2, HAM5 lines and F1 plants of a cross between HAM2 and HAM5 was reverse-transcribed using an oligo-dT primer with a SuperScript VILO cDNA synthesis kit for reverse transcription PCR (RT-PCR), as recommended by the manufacturer (Life Technologies, http://www.lifetechnologies.com). Quantitative RT-PCR was performed on an MX3000P (Agilent Technologies, http://www.home.agilent.com) using the primers CHS exon 2–3 F (5′-GAT GTA CCA ACA GGG TTG CTT TG-3′) and CHS 2740 R (5′-CAC CAA ACT ATC CAA GTG AGT G-3′). A 20 μl reaction mixture, containing 10 μl of SYBR GreenER-Supermix (Life Technologies), 0.2 μM of each primer and c. 30 ng of cDNA, was amplified using the following cycling parameters: 50°C for 2 min and 95°C for 10 min, followed by 45 cycles of 95°C for 15 s, 57°C for 30 s and 72°C for 20 s. Upon completion of the program, a melting curve analysis was performed from 40 to 95°C with steps of 0.5°C. The threshold cycle numbers (Ct), at which each sample reached the threshold fluorescence level for each type of PCR product, and the primer efficiency values were determined for all samples using MxPro QPCR software (Agilent Technologies). All Ct values were normalized to ubiquitin transcripts detected by a primer combination of UBIQ 876F (5′-CTC CGT GGT GGT TTC TGA AT-3′) and UBIQ 964R (5′-AAC GAA CCG AAC CAT CAG AC-3′). This gene is constitutively expressed in Antirrhinum (Uchiyama et al., 2009).

Transposon tagging

Plants exhibiting an intermediate petal color between HAM5 and HAM2 were obtained from the progeny of HAM5 plants grown at 15°C. An F2 population of plants with the intermediate petal color was employed for transposon tagging. A probable reason for the phenotypic change from HAM5 type to HAM2 type was assumed to be germinal excision of Tam3 from another DNA location in the HAM5 genome owing to the low temperature. First, to examine the link between Tam3 and the petal phenotypes, Southern hybridization was performed using the Tam3 sequence as a probe (Fig. 3b). In Southern hybridization, EcoRI-digested genomic DNA was transferred to Hybond N membranes (Pall, http://www.pall.com); all the procedures were carried out using an ECL Direct Nucleic Acid Labeling and Detection System in accordance with the manufacturer's instructions (GE Healthcare, http://www.gelifesciences.co).

To isolate the causative DNA sequence, transposon display using Tam3 was performed with HAM5 type and HAM2 type individuals in the F2 population. The procedures for transposon display were detailed in Takagi et al. (2003) with modification of the primers to match the Tam3 sequence. In transposon display, a 600 bp DNA band specific for HAM5 type was visualized and then isolated from the polyacrylamide gel (Fig. S1). The 600 bp band was confirmed to be associated with the three petal color phenotypes in the F2 population by Southern hybridization (Fig. S1). The 600 bp band obtained from transposon display was cloned into pBluescriptSK+ (Agilent Technologies), and Southern hybridization was carried out as described earlier.

Screening of a genomic library and sequencing

The 600 bp fragment was used as a probe for screening of the Antirrhinum genomic library in λEMBL4 provided by Z. Schwarz-Sommer (Max-Planck-Institut für Züchtungsforschung, Germany). Plaque hybridization was performed using an ECL Direct Nucleic Acid Labeling and Detection System (GE Healthcare). Consequently, five independent λEMBL4 plaques were selected from the library. A 14 kb contiguous segment was provided by DNA extracted from the five clones using a standard technique (Sambrook et al., 1989). Antirrhinum genomic DNA in each phage clone was digested randomly with DNase I, and the resultant short DNA segments were cloned into pBluescript SK+, followed by sequencing (Anderson, 1981). For DNA sequencing, a d-Rhodamine Terminator Cycle Sequencing Ready Reaction-Sequencing Kit (Life Technologies) and an ABI377 Automated DNA Sequencer (Life Technologies) were used. The 3′ region of the nivea sequence (accession number: X03710) was extended up to 8000 bp in total length. The extended sequences of the wildtype niv and nivrec alleles were deposited in DDBJ, GenBank and EMBL as accession numbers, AB691773, AB691774, respectively, and the nivbeni allele sequence containing Tam3 insertion was registered as AB691775.

Sodium bisulfite sequencing to detect methylcytosine

Before the sodium bisulfite reaction, leaf genomic DNA from the three niv lines (wildtype; niv, HAM2; nivrec and HAM5; nivbeni) was purified with the repeated use of phenol/chloroform extraction. Sodium bisulfite modification was carried out using a MethylEasy™ Xceed Rapid DNA Bisulfite Modification Kit (Diagenode, http://www.diagenode.com) in accordance with the manufacturer's instructions. Bisulfite-modified DNA was amplified with specific primers for nested PCR. The regions analyzed were a range from 300 bp upstream from the 5′ coding region of the niv gene. The wildtype allele does not contain Tam3, but the other niv alleles do. The sequences of the primers used are listed in Table S1. Each of the nested PCRs was performed as follows: 1 cycle of 95°C for 3 min, 30 cycles of 95°C for 1 min, 50°C for 2 min, 72°C for 1 min, and terminated using 72°C for 10 min. The amplified PCR products were cloned into a pT7Blue T-Vector (Novagen, http://www.merck-chemicals.com). The clones were sequenced using an ABI377 Automated DNA Sequencer (Life Technologies). Among the sequences obtained from each allele, two or more clones that coincided with cytosine methylation sites in the overall sequence were evaluated as a single clone.

Detection of siRNA of CHS genes in Antirrhinum and soybean

Extraction of small RNA and detection of CHS siRNAs were performed essentially in accordance with the method described previously (Senda et al., 2004). Sense- and antisense-specific riboprobes for detecting CHS siRNAs were synthesized from the pBluescript SK+ plasmids containing the Antirrhinum niv coding region (Uchiyama et al., 2009) or soybean CHS7 cDNA (GmCHS7) (Senda et al., 2004). Detection of siRNAs was performed essentially as described by Dalmay et al. (2000).


F2 segregation in HAM2 and HAM5

It was thought that the HAM2 and HAM5 lines possess the same Tam3 insertion 70 bp upstream of the transcription start site of the niv locus that encodes CHS in Antirrhinum (Sommer & Saedler, 1986). These lines show differences in background petal pigmentation phenotypes at high temperature (Tam3 inactive state) (Fig. 1a, b). We confirmed that there were no differences in the two niv sequences (2200 bp of coding region, 1000 bp 5′ flanking region from the transcription start site, and 300 bp 3′ untranslated region) in accordance with the sequence reported by Sommer & Saedler (1986) nor in the 3633 bp Tam3 sequences. When HAM2 was crossed to HAM5, all F1 plants had a background petal pigmentation phenotype intermediate in intensity between HAM2 and HAM5, and the F2 population segregated HAM2, HAM5, and intermediate phenotypes in a ratio of 1 : 1 : 2, respectively (Fig. 2). Together these results suggested that the difference in HAM2 and HAM5 with regard to flower pigmentation was not attributable to the niv coding sequence but was attributable to a change of the niv expression responsible for a single semi-dominant phenotype.

Isolation of revertants from HAM5 and transposon tagging

When HAM5 plants were grown at low temperature, the progeny occasionally gave rise to variants carrying an intermediate background petal pigmentation phenotype (Fig. 3). When selfed, these variants gave progeny of HAM2, HAM5 and intermediate phenotypes in a 1 : 1 : 2 ratio (Fig. 3a). These variants likely originated from germinal reversion as a result of transposable element excision from the causative genomic site in the HAM5 lines. We undertook transposon tagging to identify Tam3 copies associated with the loss of background pigmentation in HAM5 homozygotes and heterozygotes that was absent in HAM2 lines. The Tam3 probe detected a 6 kb fragment present in the HAM5 and intermediate phenotype plants (Fig. 3b). Transposon display successfully extracted a 600 bp fragment that distinguished the homozygous HAM2, HAM5, and heterozygous intermediate types (Fig. S1a). This 600 bp fragment was used as a probe to confirm its association with the phenotypes in Southern blotting analysis (Fig. S1b). Based on the sequence around the Tam3 target site, we examined empty donor sites of the revertants with intermediate phenotypes from HAM5, and we found features typical of Tam3 excision footprints in five independent revertants (Fig. S2). Therefore, it was judged that this DNA fragment contained the sequences causing the difference in petal pigmentation.

Structure of the niv allele in HAM5

Using the sequence flanking Tam3 in the 600 bp fragment from HAM5, we screened a lambda genomic library constructed using DNA from a wildtype Antirrhinum line and obtained five overlapping clones that aligned to form a 14 kb contiguous sequence. Sequence analysis indicated that the 14 kb region contained the whole niv sequence and c. 10 kb of its 3′ sequence downstream from the niv stop codon (Fig. 4). A Tam3 insertion was found at a position 1.6 kb downstream of the stop codon of niv in HAM5, and the orientation of the insertion was inverted relative to the Tam3 copy inserted in the promoter region. We failed to find this Tam3 copy in the sequence downstream of niv in the wildtype or in HAM2. Therefore, HAM5 has an additional Tam3 insertion 1.6 kb downstream of the niv coding region (downstream Tam3), insertion of which gave rise to the loss of background pigmentation in the petals in HAM5. To distinguish this from the nivrec allele of HAM2, we designated the HAM5 niv allele as niveabeni::Tam3 (nivbeni: ‘beni’ means rouge in Japanese).

Figure 4.

Structures of the two nivea alleles from Antirrhinum majus lines HAM2 and HAM5. The nivrec allele of HAM2 contains Tam3 29 bp upstream of the TATA box. The nivbeni allele of HAM5 contains two Tam3 sequences: one element located in the promoter sequence at the same site as in the nivrec allele and the other element 1.6 kb downstream of the poly-A attachment site. The two Tam3 copies are aligned head-to-head in a reverse orientation. Pink boxes indicate the coding regions of the niv gene.

Expression of niv alleles

The two flower pigmentation phenotypes resulted from structural differences between the niv alleles. We examined whether niv transcription was influenced by the structural difference of the two niv alleles (Fig. 5). RNA was prepared from the flower lobes grown at high temperatures, c. 25°C, and RT-PCR was performed with the primers designed within the niv coding sequence. Abundant accumulation of the niv transcript was observed in the HAM1 strain, which has a wildtype niv allele, whereas HAM2 with a homozygous nivrec allele exhibited a lower accumulation of niv transcript (Fig. 5a). A heterozygote of nivrec and nivbeni showed much lower accumulation of the niv transcript (Fig. 5a). Transcription was not detectable from the nivbeni allele in HAM5 (Fig. 5a). Quantitative analysis using real-time PCR also showed the positive correlation between niv transcript abundances and the degree of background petal pigmentation (Fig. 5b). An abundance of the transcript from the wildtype niv was as much as seven times that from nivrec, and the heterozygote of nivrec/nivbeni gave rise to approximately half of the abundance of the transcript from nivrec. The nivbeni allele accumulated the transcript 360 times less than the wildtype as observed in the petal phenotype.

Figure 5.

Transcription of Antirrhinum majus niv alleles. (a) RNA gel blotting analysis was performed using a 657 bp PCR product from the niv coding region as a probe. RNA was prepared from the flower lobes of HAM1 (lane 1), HAM2 (lane 2), F1 hybrid of HAM2 and HAM5 (lane 3) and HAM5 (lane 4) lines. All the plants were grown at 25°C, which immobilized Tam3. HAM1 harbors a wildtype (WT) allele of niv, and HAM2 and HAM5 contain the nivrec and nivbeni alleles, respectively. Each lane shows a c. 3 kb transcript. The ethidium bromide (EtBr) staining patterns are included to show equal loading of rRNA. (b) Quantitative real-time PCR verified the different accumulation levels of niv transcripts in the four lines used in the RNA gel blotting analysis. The relative value was calculated by dividing the value obtained from chalcone synthase (CHS) primers with the value from ubiquitin primers. Error bars indicate each average.

Influence of downstream Tam3 on niv expression

Plants lacking either or both of the two Tam3 copies at the niv locus were isolated from HAM2 and HAM5 to investigate the influence of the two Tam3 insertions on niv expression. We isolated a variant plant carrying only the downstream Tam3 from a HAM5 plant grown at 15°C (Figs 6, S3). A variant with a niv allele lacking the Tam3 copy in the niv promoter (upstream Tam3) from HAM5 revealed dark petal pigmentation similar to the wildtype (Fig. 6). The homozygous plants of this variant allele, screened from the selfed population, showed nearly full red petal pigmentation similar to the wildtype allele (Fig. 6). The expression of niv was compared with that of the wildtype allele. Quantitative real-time PCR of the petal RNA revealed that the valiant allele expressed half the level of the wildtype niv transcript, that is, the insertion of the downstream Tam3 reduced the wildtype niv expression by 50% (Fig. 6), when the effect of the promoter mutation (footprint) occurred by Tam3 excision was ignored. Accordingly, the expression level of the nivbeni allele, where both Tam3 copies are simultaneously present, should be 50% of the nivrec allele (i.e. 1/14 of the wildtype niv expression), which was expected to be close to that of heterozygote of nivrec/nivbeni for the purpose of the calculation (Fig. 5). However, the expression level of the nivbeni allele was actually 51 times lower than that of the nivrec allele and 360 times lower than that of the wild type allele (Fig. 7). Hence, the nivbeni allele gave rise to a further reduction of niv expression of c. 26-fold (51/2 or 360/14) in addition to the reducing effect of both independent insertions of Tam3 on the niv expression (Fig. 7). The synergistic effect lowering of niv expression is therefore likely brought about by interaction of the two Tam3 insertions across the niv gene.

Figure 6.

The flower phenotype, niv expression and structure of a variant carrying only the downstream Tam3 copy were compared with the wildtype (WT) niv allele in Antirrhinum. (a) The variant (Var) that lost the upstream Tam3 owing to low-temperature-dependent transposition exhibited nearly the WT phenotype of niv (Fig. S3). (b) The structures of the niv alleles in the variant and WT niv are depicted. (c) The niv expression was compared between the variant and WT by means of quantitative real-time PCR. Error bars indicate each average.

Figure 7.

Different expression activities of the niv alleles resulted from Tam3 insertions in Antirrhinum. (a) The flower pigmentations and niv structures are shown for the four alleles, the wildtype (WT), variant, nivrec and nivbeni. Relative ratios of the niv expression are estimated from the values based on the quantitative real-time PCR, when the value of the WT is regarded as 1. (b) The upstream and downstream Tam3 insertions reduce the WT expression by seven times or half, respectively. Besides the independent reductions, 26-fold reduction was induced by simultaneous insertions of the two Tam3, which should result from interaction of the pair of Tam3 copies.

Investigation of epigenetic silencing as a possible cause of the nivbeni phenotype

The causes for the reducing effect resulting from the insertion of a pair of Tam3 elements on the expression of the nivbeni allele were considered likely to be associated with epigenetic silencing. The DNA methylation status of the promoter regions from the three niv alleles (wildtype, nivrec, and nivbeni) was analyzed by sodium bisulfate sequencing. Compared with the promoter regions in the wildtype niv and nivrec alleles, the nivbeni allele possessed a lower methylation state for all cytosine contexts in the promoter regions of 300 bp ranging from the 3′ sequence of the upstream niv Tam3 to the 5′ sequence of the niv coding sequence (Figs 8, S4). This implies that the insertion of the downstream copy of Tam3 induced a decrease in methyl-cytosine concentrations in the niv promoter region (Fig. 8).

Figure 8.

DNA methylation patterns of the promoter sequences from the three Antirrhinum niv alleles: wildtype (WT), nivrec and nivbeni. (a) Independent sequences were evaluated from a total of 80 clones (26 clones of WT niv, 15 clones of nivrec and 23 clones of nivbeni). The distribution and frequency of methylated cytosines in the sense strands of the 300 bp promoter sequences in the three alleles were analyzed by bisulfite sequencing. The positions of cytosine residues in each genomic sequence are indicated on the horizontal axis (white box, transcribed region; black box, TATA box; gray box, Tam3 sequences; triangle, Tam3-TIR). (b) The frequency of methylated cytosines in the three contexts is indicated on the vertical bars (black bar, CpG; blue bar, CpNpG; red bar, CpHpH). These results coincided with the results obtained from the Southern blotting analysis using methylation-sensitive restriction enzymes (Supporting Information Fig. S4). The methylation profile for the promoter in nivrec was utilized from our previous result referred from Uchiyama et al. (2009). The overall proportions of methylated cytosines in every context are summarized.

Because all the nivrec/nivbeni heterozygotes (F1 plants of HAM2 and HAM5) showed stable flower color with intermediate pigmentation when grown at higher temperature, we considered that a gene-silencing or RNAi-like event was unlikely to be induced by the insertion of the downstream copy of Tam3. An attempt was made to detect small interfering RNA (siRNA) molecules derived from the niv coding sequence. Soybean (Glycine max) was employed as a positive control where CHS genes are known to be silenced by RNAi and to generate a corresponding siRNA (Senda et al., 2004). A nucleotide identity of 72% was estimated between both the coding regions of the Antirrhinum niv gene and soybean CHS gene (GmCHS7), and this value allowed us to predict that the detection of siRNAs would be difficult in northern blotting analysis between these two species. Four blotting results in Fig. 9 revealed that four RNAs from soybean, HAM1 (niv: wildtype CHS gene), HAM2 (nivrec), and HAM5 (nivbeni) hybridized to the sense and antisense probes prepared from both the Antirrhinum and soybean CHS coding regions. The blotting patterns with the soybean RNA were distinguished from the Antirrhinum RNAs, for which blotting patterns were similar among the three lines. In terms of the detection of siRNAs, the two soybean probes clearly detected siRNAs (21–22 mers) in soybean RNA but not in Antirrhinum RNA (Fig. 9a). The two Antirrhinum probes failed to detect siRNAs in RNA from the three Antirrhinum lines (Fig. 9b). The soybean RNA did not detect siRNAs with the Antirrhinum probes either (Fig. 9b); this might be because the homology between the soybean CHS gene and the niv sequences was not enough for the detection of signals. A reasonable explanation of the results was that these Antirrhinum lines produced few siRNAs complementary to the niv transcript in contrast to soybean, and this suggested that RNA silencing was not associated with transcriptional impairment of the nivbeni allele. From these results, we considered DNA methylation and RNAi were unlikely to be responsible for the suppression of transcript abundance with the nivbeni allele where two Tam3 copies were located either side of the niv gene.

Figure 9.

Attempted detection of small interfering RNAs (siRNAs) corresponding to the soybean CHS and niv transcripts. RNA blots were prepared from soybean, HAM1 carrying the wildtype niv allele (niv WT), HAM2 carrying the nivrec allele and HAM5 carrying the nivbeni allele. (a) Profiles obtained using antisense or sense probes from soybean CHS gene (GmCHS7). (b) Profiles obtained using antisense or sense probes from the WT niv gene. The position of the 22-nucleotide (nt) oligomer is indicated on the left.


The different background petal pigmentation phenotypes of HAM2 and HAM5 were the result of different abundances of niv (CHS gene) transcripts (Figs 1,5). Both lines shared the same promoter sequence containing Tam3 (upstream Tam3), while transposon tagging identified a structural difference 1.6 kb downstream region of the 3′ end of the niv gene, where HAM5 had an additional Tam3 insertion (downstream Tam3) (Fig. 4). The insertion of the upstream and downstream Tam3 copies both resulted in the reduction of niv expression to one-seventh and one-half, respectively (Fig. 7). The simultaneous insertion of both the upstream and downstream Tam3 copies induced a synergistic fall in the level of niv expression, by c. 360-fold, reflected by a petal phenotype with no background pigmentation observed in HAM5. Therefore, it was certain that the two Tam3 copies spanning the niv gene coordinately suppressed niv gene expression.

The nivrec allele carrying the upstream Tam3 exhibited a leaky phenotype, because the 3′ end region of Tam3 adapts any promoter sequence for limited expression of the gene (Uchiyama et al., 2009). In fact, leaky phenotypes in Antirrhinum have been described in several loci where Tam3 was inserted in the same orientation as the downstream gene; a reporter gene with Tam3 in the promoter sequence validated this finding in yeast (Uchiyama et al., 2009). These alleles, called ‘Tam3-permissible alleles’, seem to be ubiquitous for the expression of any genes (Uchiyama et al., 2009).

In the nivbeni allele, where the presence of two Tam3 copies reduced with niv transcript abundance, epigenetic controls were examined. However, we found no evidence that differences in DNA methylation or RNAi were associated with the reduction in the niv expression between HAM2 and HAM5 (Figs 8, 9). The tail-to-tail orientation of the two Tam3 copies either side of the niv gene is a structural feature that typically can induce production of double-stranded RNA if both Tam3 sequences were transcribed through the niv gene. Double-stranded RNA is eventually digested to generate siRNA that triggers DNA methylation or RNAi (Brodersen & Voinnet, 2006; Liu & Paroo, 2010). As a possible explanation, niv transcription might be suppressed by DNA methylation or the transcript might be degraded by RNAi. However, this scenario was not supported by our analysis: a relatively low concentration of methylated cytosine was detected in the nivbeni promoter region compared with that in other niv alleles, and no corresponding siRNA was detected using the niv coding sequence as a probe (Figs 8, 9). This interpretation was supported by the fact that F1 plants between HAM2 and HAM5 gave rise to intermediate background petal pigmentation (Fig. 2), indicating that the silencing effect of nivbeni was not trans-acting. The intermediate phenotype in the F1 plants suggested that transcription of the nivrec allele was independent of interference by the nivbeni allele, and that the reduced pigmentation and silencing of nivbeni was cis-acting. The significance of the inhibition of gene expression might be related to homology between the Tam3 sequences in inverted orientation. The situation in the nivbeni allele could be likened to being trapped between the two Tam3 copies.

Insulators are chromatin boundary elements that can block communication between enhancer and promoter elements, resulting in the suppression of gene expression (Valenzuela & Kamakaka, 2006). The enhancer-blocking activity of the insulator prevents enhancer action on a promoter when the insulator is placed between the enhancer and the promoter (Wei et al., 2005). We do not know whether the downstream Tam3 exerted an insulator-like function to inhibit niv expression, although the presence of only the downstream Tam3 copy (i.e. in the absence of the upstream Tam3 copy in the nivbeni allele) reduced the transcript abundance of the wildtype allele by half, resulting in nearly full red petal pigmentation (Fig. 6). The further inhibition of niv expression could occur when two Tam3 sequences are located either side of the gene. In the nivbeni allele, the downstream Tam3 negated the function of permissible expression in the nivrec allele (Figs 4, 5).

Our results allowed us to consider an interaction between the two Tam3 sequences. An analogy for the ‘trapping of a gene’ by paired sequences is found in the case of the insulator suHW, which is a 340 bp element of the Drosophila gypsy retrotransposon (Gerasimova & Corces, 2001). When two suHW insulators were inserted between the enhancer and promoter of a gene in Drosophila, enhancer-blocking activity arising from one of the two suHW insulators was modulated probably via interaction of the two insulators (Cai & Shen, 2001; Muravyova et al., 2001), suggesting that bound protein complexes formed a mini-loop (Byrd & Corces, 2003). Insulators are known to facilitate loop conformation mediated with insulator-binding protein (Gaszner & Felsenfeld, 2006; Valenzuela & Kamakaka, 2006). The two identical Tam3 sequences in this study might come in to close proximity or bind to each other to form a loop and prevent the access of transcription factors to the niv promoter, resulting in the loss of transcription (Fig. 10). In the nivbeni allele, the low degree of DNA methylation in the promoter region (Figs 8, S4) might reflect the prevention of access for DNA binding proteins.

Figure 10.

Models of transcriptional regulation of the two Antirrhinum niv alleles. Transcription factors can access the promoter containing the Tam3 sequence in the nivrec allele, resulting in a pale-red color in the flower lobe, whereas two Tam3 copies in the nivbeni allele do not permit transcription factors to activate the niv allele, resulting in white flowers.

Several protein complexes have been identified as insulator-binding proteins required for loop conformations (Wei et al., 2005; Gaszner & Felsenfeld, 2006). A mammalian protein, CTCF, is one of the most extensively studied insulator-binding proteins, which demarcate boundaries between euchromatin and heterochromatin (Bell et al., 1999; Yusufzai et al., 2004). Genome-wide analysis identified CTCF interactions at 1480 intrachromosomal and 336 interchromosomal loci in the mouse genome and showed that chromatin loops organize the boundaries of chromatin domains to demarcate different transcriptional control units (Handoko et al., 2011). In addition to suHW, other TEs and/or repetitive sequences that exist in eukaryote genomes could be employed widely as target sites for insulator-binding proteins. These elements have the potential to form loops that participate in genome architecture or influence gene expression in plant genomes (Louwers et al., 2009). In particular, the loop structures of TEs tend to generate alternative gene expression patterns (West et al., 2002; Sexton et al., 2009). The Tam3 TPase in Antirrhinum does not participate in loop conformation because it does not enter the nucleus when plants are grown at high temperatures (Fujino et al., 2011). The reduced pigmentation phenotype of the nivbeni allele is temperature-independent, so if loop formation between the paired Tam3 copies occurs, proteins other than the TPase must be involved. Alternatively, the two Tam3 copies might interact autonomously in a homology-dependent manner.

Suppressor-mutator (Spm) and Mutator (Mu) elements are the well-known maize DNA-type TEs that can also control the expression of the proximal host genes (Craig, 2002). The expression of maize genes with the insertion of a defective Spm element can be under either positive (Spm-dependent) or negative (Spm-suppressive) control in the presence of an autonomous Spm element (McClintock, 1958, 1961; Masson et al., 1987; Grant et al., 1990). In several maize loci, Mu insertion in the promoter regions interfered with transcription when Mu was active (Martienssen et al., 1990; Martienssen & Baron, 1994; Cui et al., 2003), but the genes were transcribed when Mu was inactive (Greene et al., 1994). This Mu-suppressive system is correlated with the epigenetic state of the Mu element (Barkan & Martienssen, 1991; Robbins et al., 2008).

The occurrence of the different phenotypes in suppressible alleles described for Spm and Mu elements in maize depends on the presence/absence of their TPases (autonomous elements), although the background petal pigmentation driven by the nivrec allele in HAM2 was not related to transcription and translation of the Tam3 TPase gene (Uchiyama et al., 2008, 2009). Thus, niv regulation as a result of Tam3 is apparently distinct from that occurring in the suppressible alleles of Spm and Mu in maize. The nivbeni allele in HAM5 showed no background pigmentation in petals owing to an interaction of the paired Tam3 copies, a phenomenon dissimilar to those described for the modulation of gene expression associated with TEs. This example demonstrates another mechanism whereby TEs may influence gene expression, a mechanism that could be significant in times of ‘genome shock’ (McClintock, 1984), when transposition is frequent and elements may insert in close proximity to each other on the genome.


We are especially grateful to Yoshio Sano (Laboratory of Plant Breeding, Hokkaido University) for comments on the manuscript and Yuji Noro (Laboratory of Plant Breeding, Hokkaido University) for technical assistance. This research was supported by the Program for Promotion of Basic and Applied Researches for Innovations in Bio-oriented Industry (BRAIN) to Y.K. and a Grant-in-Aid for Young Scientists from MEXT to T.U. C.M. was supported by the core strategic grant of the Biological and Biotechnological Science Research Council (BBSRC).