Introduction

The pigmentations that color most flowers, fruits, and seeds, are anthocyanin secondary metabolites, and their biosynthesis has been well characterized in the flowers of petunia (Petunia hybrida) and snapdragon (Antirrhinum majus) and in the kernels of maize (Zea mays) (see reviews; Holton and Cornish 1995; Mol et al. 1998; Winkel-Shirley 2001; Koes et al. 2005). Among these, transposable elements greatly contribute to advances in the discovery and characterization of new flavonoid biosynthesis-related genes (van Houwelingen et al. 1998).

Transposable elements have been isolated from a number of species including higher plants, and can be divided into two classes according to whether their transposition intermediate is RNA or DNA, designated as retrotransposon (class 1) or DNA transposon (class 2), respectively (Feschotte et al. 2002; Graig 2002). Retrotransposons are the most abundant and widespread class of eukaryotic transposable elements, consisting of either a long terminal repeat (LTR) or non-LTR retrotransposons (Kumar and Bennetzen 1999). LTR-retrotransposons have LTRs as direct repeats and autonomous elements including at least two genes, so-called gag and pol. LTR-retrotransposons are further subclassified into Ty-1 copia and Ty3-gypsy groups according to their sequence similarity and order of encoded gene products. In contrast, non-LTR retrotransposons are divided into long interspersed repetitive elements (LINE) and short interspersed repetitive elements (SINE). Recently, terminal-repeat retrotransposon in miniature (TRIM) that belong to a new group of LTR-retrotransposons have been identified in several plant species including Arabidopsis and rice from the analysis of genome sequence data (Witte et al. 2001).

DNA transposable elements can be grouped into families based on both the sequence similarity of the homologue of terminal inverted repeats (TIR) and the specific number of nucleotides comprising the target site duplication (TSD) caused by insertion of the element into the host genomic DNA (Kunze and Weil 2002). Members of the hAT family, which include activator/dissociation (Ac/Ds) elements of maize (Kunze and Starlinger 1989) and Tam3 elements of snapdragon (Hehl et al. 1991), cause an 8 bp direct duplication upon insert into a new site, whereas members of the CACTA family, which include enhancer/suppressor-mutator (En/Spm) elements of maize (Pereira et al. 1986) and Psl of petunia (Snowden and Napoli 1998), cause a 3 bp of TSD. Besides, miniature inverted repeat transposable elements (MITEs) are found as small transposable elements that occur in high copy number in the host genomes, containing mPing elements of Oryza sativa activated in cells derived from anther culture (Kikuchi et al. 2003; Nakazaki et al. 2003; Jiang et al. 2003). Most gentian plants (Gentiana scabra, G. triflora and their hybrids) cultivated in Japan exhibit blue or purple flowers, which contain anthocyanin from delphinidin derivatives, named gentiodelphin (Goto et al. 1982) and albireodelphin (Hosokawa et al. 1997). Recently, pink-flowered gentian cultivars have been bred using mutants that have arisen from natural variations. A pink-flowered gentian is known to accumulate cyanidin derivatives, called gentiocyanin, lacking delphinidin derivatives in petals (Hosokawa et al. 1995). The structural genes involved in anthocyanin biosynthesis have been isolated and characterized from the flowers of G. triflora (Tanaka et al. 1996; Kobayashi et al. 1998; Fujiwara et al. 1998; Fukuchi-Mizutani et al. 2003; Nakatsuka et al. 2005). The key enzyme of delphinidin biosynthesis is known to be flavonoid 3′,5′-hydroxylase (F3′,5′H), which catalyzes the hydroxylation at the 5′ position of the B-ring of the anthocyanidin skeleton. F3′,5′H belongs to the cytochrome P450 superfamily, and its cDNAs have been isolated from petunia (Holton et al. 1993) and gentian (Tanaka et al. 1996). Therefore, we hypothesized that pink-flowered gentian plants might have an absence of F3′,5′H activity due to certain mutations in the F3′,5′H gene.

In this study, we report on the isolation and characterization of two transposable elements, hAT transposable elements and TRIM within the F3′,5′H gene in two different pink-flowered gentian plants. The two transposable elements are thought to be nonautonomous and these homologues are abundant in the gentian genome. This is the first-known report on transposable elements in the Gentianaceae family containing approximately 1,200 species comprising many valuable pharmaceutical and ornamental plants.

Materials and methods

Plant materials

Ornamental gentian plants, Gentiana scabra cv. Momokorin (pink flower), a breeding line 13–98 (pink flower), cv. Alta (blue flower), and G. triflora cv. Maciry (blue flower), were provided by the Iwate Agriculture Research Center. The petals and leaves collected from each cultivar were stored at −80°C until used for analysis.

Anthocyanin extraction and HPLC analysis

For anthocyanidin analysis, frozen petal samples (250 mg) were ground in liquid nitrogen and then extracted using 1 ml EAA solution (ethanol:acetic acid:water=10:1:9) with gentle shaking at 4°C overnight. For hydrolysis of anthocyanin, 4.5 ml 3 N HCl was added and incubated at 100°C for 90 min. After the hydrolysis, the solution was extracted with 1 ml isoamylalchol and filtered through a 0.22 μm filter (Millipore); HPLC analysis was carried out by a high performance liquid chromatography (HPLC) system (D-7000 HPLC system manager, Hitachi). HPLC was carried out with a reverse phase column J’ sphere ODS M80 (4.6×150 mm, YMC) using 15% acetonitril and 3% acetic acid as the solvent for 20 min at 40°C at a flow rate of 1.0 ml/min. Anthocyanin was quantified by monitoring the peak area of absorbance at 500 nm, with pelargonidin, cyanidin, and delphinidin as the standard.

Northern analysis of F3′,5′H and F3′H genes

Total RNA was isolated from 100 mg fresh weight of gentian petals at flower development stage 3 as defined by Nakatsuka et al. (2005) using concert Plant RNA reagent (Invitrogen, CA, USA). The 5 μg of total RNA was separated on 1.25% MOPS-agarose gel, and transferred to a Hybond-N+ membrane (Amersham Pharmacia, Sweden). The membrane was hybridized and detected as described by Nakatsuka et al. (2005).

The characterization of F3′,5′H cDNA in pink-flowered gentian plants

The first strand cDNA was synthesized using the total RNA of the gentian petals by using RNA PCR Kit (AMV) version 2.1 (TaKaRa, Tokyo) according to the manufacturer’s protocol. The primers designed for amplification of the F3′,5′H open reading frame (ORF), were used as follows: forward primer 5′-ATG CCC ATA AAA ATG TCA CCC ATT-3′ (start codon indicated by the underline) and reverse primer 5′-TTA AGG AGC ATA AAC ATG AAG AGG -3′ (stop codon indicated by the underline). For 3′RACE analysis, the combination of the forward primer for the ORF amplification and M13-M4 primer was used. The reaction mixture in a total volume of 50 μl contained 1 μl first strand cDNA, 200 μM dNTPs, 2.5 mM MgCl2, 5 μM each primer, 2.5 U of LA Taq polymerase (TaKaRa, Tokyo), and 1× LA PCR buffer. The reaction conditions were preheating at 94°C/1.5 min, 35 cycles at 95°C/20 s, 55°C/40 s, and 72°C/2 min, and then extension at 72°C/10 min. The amplified fragments were purified by GENECLEAN II (BIO101, CA, USA) and then subcloned into a pCR4-TOPO cloning vector for sequencing (Invitrogen). Sequence analyses were carried out using Big-Dye Terminator cycle sequencing Kit version 1.1 and DNA sequencer model ABI PRISM 377 (Applied Biosystems Japan, Tokyo). Sequence analyses were carried out using GENETYX-MAC version 12.0 (GENETYX, Tokyo) and the BLAST network service from the National Center for Biotechnology Information (NCBI) (Altschul et al. 1990).

The F3′,5′H genomic sequence of pink-flowered gentian plants was amplified using the primer sets described earlier. The reaction mixture contained 100 ng genomic DNA as the template and was subjected to the same conditions as in the RT-PCR analysis. The obtained fragments were subcloned into a pCR4-TOPO cloning vector for sequencing (Invitrogen). Sequence analysis was performed as described earlier.

Southern blot analysis of F3′,5′H and transposable elements

Total genomic DNA was isolated from a 1 g leaf sample using Nucleon PhytoPure (Amersham Biosciences) as described in the manufacturer’s instruction. The 10 μg of genomic DNA digested by EcoR I and Hind III (TaKaRa, Tokyo) was separated on 0.6% agarose gel and then transferred to a Nytran N membrane (Schleicher & Schuell, Germany). F3′,5′H fragments were amplified using the primer sets described earlier, the truncated transposase sequence of the dTgs1 fragment was amplified using primer sets as follows: dTgs1-U, 5′-CGA CGT GGA GTT GAT GGT AG-3′; dTgs1-L, 5′-CTC CTA GCA TTT TTA GAA GC-3′, and GsTRIM1 was amplified using M13-M4 and M13-RV attached within the pCR4-TOPO vector. Amplified fragments were labeled with a DIG-High Prime DNA labeling Kit (Roche Diagnostics, Germany) according to the manufacturer’s instructions. Each probe was hybridized and detected as described by Nakatsuka et al. (2005).

Results

Anthocyanidin accumulation in pink-flowered gentian plants

Pictures of three gentian plants used in this study are shown in Fig. 1. To measure the anthocyanidin compounds, HPLC samples were extracted from the petals of one blue- and two pink-flowered gentian plants. The HPLC profile of the blue-flowered gentian cv. Alta (Fig. 1a) showed two peaks, indicating co-accumulation of major delphinidin and minor cyanidin derivatives in the petals (Fig. 1d). On the other hand, the two pink-flowered gentian plants, line 13–98 (Fig. 1b) and cv. Momokorin (Fig. 1c), showed a single major peak, indicating an accumulation of cyanidin derivatives, but not delphinidin derivatives (Fig. 1e, f, respectively). The amount of cyanidin derivatives in the two pink-flowered gentian plants did not remarkably increase in comparison with that in the blue-flowered gentian.

Fig. 1
figure 1

Flower colors and HPLC profiles of anthocyanidin in blue- and pink-flowered gentian plants. Three gentian cultivars used in this study are shown, cv. Alta (a), line 13–98 (b), and cv. Momokorin (c). Anthocyanin was extracted from the petals and then hydrolyzed to convert it to anthocyanidin. Anthocyanidin extractions from each cultivar were subjected to HPLC analysis as described in Materials and methods. dp and cy indicate delphinidin and cyanidin, respectively

Full size image

The expression of flavonoid hydroxylase genes in pink-flowered gentian plants

Because pigment analysis confirmed that pink-flowered gentian plants accumulated only cyanidin derivatives, the difference in hydroxylation activity at the B-ring in the flavonoid skeleton was considered to cause this change. Therefore, one of the flavonoid biosynthetic structural genes, F3′,5′H, encoding an enzyme catalyzing the hydroxylation at the 3′- and 5′-positions of the flavonoid skeletons, was first analyzed with respect to gene expressions between the blue- and pink-flowered gentian plants. No transcript of the F3′,5′H gene in the pink-flowered gentian line 13–98 was observed by either northern blot (Fig. 2a) or RT-PCR analyses (Fig. 2b) using primer pairs for full length of F3′,5′H ORF, although minor transcripts of 1.4 kb in lengths were amplified by 3′RACE analysis (see subsequently). In contrast, the F3′,5′H transcript of another pink-flowered cv. Momokorin was detected at a higher molecular weight with a remarkably reduced amount compared with blue-flowered cv. Alta (Fig. 2a). By RT-PCR analysis using primers designed to amplify the full-length ORF of the F3′,5′H gene, cv. Momokorin showed the amplified fragments of about 1.3 and 2.1 kb, differing from the 1.6 kb single amplified fragment in blue-flowered gentian (Fig. 2b). These results show that neither of the pink-flowered gentian plants accumulated normal F3′,5′H transcripts in the petals. On the other hand, mRNA accumulation of F3′H, encoding an enzyme catalyzing the hydroxylation at 3′-position of the flavonoid B-ring, was similar among the petals of blue- and pink-flowered gentian plants (Fig. 2a), suggesting that the F3′H gene did not contribute to the inability to accumulate delphinidin derivatives in the pink flowers. All other flavonoid biosynthesis-related structural genes (CHS, CHI, F3H, DFR, ANS, 3GT, 5AT, and FSII) also exhibited similar expression among the petals of the pink- and blue-flowered gentian plants (data not shown).

Fig. 2
figure 2

Expression analyses and genomic-PCR analysis of flavonoid hydroxylase genes in the petals of blue- and pink-flowered gentian plants. a Northern blot analysis was performed using total RNA (5 μg) from the petals of blue-flowered gentian cv. Alta, pink-flowered gentian line 13–98 and cv. Momokorin. The blot was hybridized by each probe for F3′,5′H or F3′H gene. The ribosomal RNA bands stained with ethidium bromide used as a control are also shown. b RT-PCR analysis was performed to amplify F3′,5′H ORF using total RNAs isolated from the petals. Numerals on the right indicate the sizes of the amplified fragments estimated by DNA marker. c PCR analysis was performed to amplify the genomic F3′,5′H sequence with the primer set used by RT-PCR analysis. Numerals on the right indicate the same as (b)

Full size image

Cloning of genomic F3′,5′H gene in pink-flowered gentian plants

To examine the genome structures of the F3′,5′H gene, the genomic DNAs of the blue- and pink-flowered gentian plants were used to amplify full-length fragments containing F3′,5′H ORF. Different length amplified fragments were obtained from each gentian line (Fig. 2c). Blue-flowered cv. Alta showed about a 2.8 kb amplified fragment, whereas, the amplified fragments of genomic F3′,5′H ORF in the pink-flowered gentian line 13–98 and cv. Momokorin were observed to be about 1.2 and 0.5 kb longer than that of blue-flowered gentian, respectively. Southern blot analysis of F3′,5′H also supported this result (see subsequently). Both the 4.0 and 3.3 kb of the genomic F3′,5′H fragments in the pink-flowered gentian plants were subcloned and their sequences were determined.

Transposable element inserted in the F3′,5′H gene of pink-flowered gentian line 13–98

4.0 kb of the F3′,5′H genomic region in line 13–98 contained an insertion of 1,187 bp in length at the 239 bp position from the start codon within the first exon, designated as dTgs1 (Fig. 3a). dTgs1 showed distinctive features of the hAT superfamily members of class 2 transposable elements, such as an 8 bp of target site duplication (TSD) and a 14 bp of terminal inverted repeats (TIR). The 14 bp of TIR had one substitution between 5′- and 3′-TIRs (Fig. 3b and Table 1). They showed a unique perfect or imperfect palindrome structure, and exhibiting partial similarity to that of dTph1 in petunia (Table 1). A sequence containing truncated ORF near the 3′ flank of this insertion has homology to similar regions in the A. thaliana hAT dimerization domain-containing protein (accession number NP_178689, 582 residues; 71% amino acid identity over 116 residues). The nucleotide sequences of the proximal region of both TIRs, recognized by Ac transposase, contained a CG-rich sequence, but not the conserved AAACGG motifs observed in Ac transposable elements in maize.

Fig. 3
figure 3

Schematic representation of structure, sequence, and transcripts of F3′,5′H gene in line 13–93. a Structure of genomic F3′,5′H gene. The F3′,5′H genomic structure of the pink-flowered gentian line 13–98 is 4.0 kb in length, containing 1.2 kb of dTgs1 transposable element (hAT family) within the first exon. b Nucleotide and deduced amino acid sequence of dTgs1. dTgs1 in the F3′,5′H gene of line 13–98 has 1,187 bp of hAT superfamily transposon. 8 bp of target sequence duplication (TSD) is shown double underlined. 5′ and 3′ inverted repeats (IR) are indicated by arrows, and CG-rich sequences of subterminal regions are boxed by lines. A truncated deduced amino acid sequence of transposase is shown in a lower sequence, and C-terminal conserved domain of hAT transposon, which is prospected by programs of CD-Search (Marchler-Bauer and Bryant 2004), is indicated by a closed gray box. c Minor transcripts of F3′,5′H gene. The primer positions with RT-PCR and genomic PCR analysis in Fig. 2 are shown by thick arrows. The numeral at the right of the illustration represents the fragment length obtained by genomic PCR or 3′-RACE analysis. The F3′,5′H gene in line 13–98 underwent alternative polyadenylation, resulting in short incomplete F3′,5′H transcripts

Full size image
Table 1 Comparison of plant hAT element terminal inverted repeats (TIRs)
Full size table

Transposable element inserted in the F3′,5′H gene of pink-flowered gentian cv. Momokorin

A 3.3 kb of genomic F3′,5′H in cv. Momokorin contained an insertion of 513 bp in length at the 87 bp position from the start codon within the first exon, designated as GsTRIM1 (Fig. 4a). GsTRIM1showed the feature of TRIM belonging to class 1 transposable elements, such as a 5 bp of TSD and a 76 bp of terminal direct repeats (TDR) containing 8 bp inverted repeats (TGTTAAAG) at both ends (Fig. 4b). However, unlike the usual TRIM elements, GsTRIM1 contained three TDRs the second of which had an incomplete structure in comparison with the TDRs of both ends. Further examination of this insertion revealed the presence of primer binding sites (PBS) and polypurine tracts (PPT) between each TDR. This insertion did not contain sequences encoding any proteins, indicating that this is a nonautonomous transposable element like other TRIM elements. Sequence analysis also showed that the GsTRIM1 transposable element was similar to some sequences in the Lotus japonicus genomic library (data not shown). In particular the 3′-end of TDR of GsTRIM1 was similar to that of the L. japonicus GsTRIM1 homologue, designated as LjTRIM. LjTRIM also contained some copies at independent loci within the L. japonicus genome and had the same feature of 138–141 bp of TDR, PBS, and PPT motifs as GsTRIM1 (Fig. 5a). Because TRIM elements were first identified by database analysis and have the striking peculiarity to be quite conserved among plant species, we performed a phylogenic tree analysis (Fig. 5b) using TDR sequences. The result clearly showed that GsTRIM1 and LjTRIM positioned apart from the previous reported TRIMs, indicating that these two are evolutionarily different from other plant TRIM elements. Notably, the TDR sequences of GsTRIM1 had an incomplete palindrome structure (Fig. 5c).

Fig. 4
figure 4

Schematic representation of structure, sequence, and transcripts of F3′,5′H gene in cv. Momokorin. a Structure of genomic F3′,5′H gene. The F3′,5′H genomic structure of the pink-flowered gentian cv. Momokorin is 3.3 kb in length, containing 520 bp of GsTRIM1 transposable element within the first exon. b Nucleotide sequence of GsTRIM1. GsTRIM1 has the features of terminal-repeat retrotransposons in miniature (TRIM), containing 5 bp of target sequence duplication (TSD, double underlined) and triple 76 bp of terminal direct repeats (TDR, boxed lines) with 8 bp of inverted repeats (IR, thick arrows). Primer binding sites (PBS) and polypurine tract (PPT) are indicated by solid and dashed lines, respectively. A 5′-splicing recognition site of alternative splicing as observed in Fig. 2b is boxed by thick lines. c Transcripts of F3′,5′H gene. The primer positions with RT-PCR and genomic PCR analysis in Fig. 2 are shown by thick arrows. The numeral at the right of the illustration represents the fragment length obtained by Genomic PCR or RT-PCR analysis. Two abnormal F3′,5′H transcripts accumulated in pink-flowered gentian cv. Momokorin by the insertion of GsTRIM1 or alternative splicing

Full size image
Fig. 5
figure 5

Comparison of TRIM elements among higher plants. a Comparison of overall structure of TRIM elements between G. scabra (GsTRIM) and L. japonicus (LjTRIM). GsTRIM and LjTRIM were presumed from GsTRIM1 and 15 independent GsTRIM1 homologues of L. japonicus, respectively. b Phylogenic analysis of TDR of GsTRIM1 with those of other higher plants. The phylogenic tree was generated using CLUSTALW (Thompson et al. 1994) and TREEVIEW (Page 1996) programs. Numerals indicate bootstrap values from 1,000 replicates. The bar indicates an evolutionary distance of 0.1%. Accession numbers in the GenBank/DDBJ databases are as follows: A. thaliana At1 (gi 6598490, position 8,088–8,204), A. thaliana At2 (gi 7649355, position 68,877–68,994), G. scabra GsTRIM1 (the present study, AB222606), Lycopersicon esculentum (gi 4220970, position 38–157), L. japonicus LjTRIM (AJ580824, position 45,546–45,170), Medicago sativa (gi 19642, position 8,088–8,204), Nicotiana tabacum (gi 9392606, position 3,821–3,947), O. sativa (gi 10800055, position 68,877–68,994), Phaseolus vulgaris (gi 2576326, position 1,878–1,761), Solanum tuberosum (gi 14599414, position 1,522–1,651). c Predicted secondary structure of TDR of GsTRIM1. The potential secondary structure is predicted with GENETYX-Mac software

Full size image

Effect of transposable element insertions on the F3′,5′H transcription in pink-flowered gentian plants

Since two pink-flowered gentian plants had insertions of independent transposable elements in the F3′,5′H gene, functional F3′,5′H mRNA could not be transcribed. Namely, F3′,5′H transcripts of line 13–98 were suppressed by the insertion of dTgs1, while two different sizes of transcripts were observed by the insertion of GsTRIM1 in cv. Momokorin (Fig. 2a, b). However, when 3′RACE analysis was performed using forward primer within exon 1, minor transcripts of 1.4 kb in length were also detected in line 13–98. Sequence analysis revealed that the 1.4 kb of the minor F3′,5′H transcripts were derived from alternative polyadenylation after the stop codon of putative transposase of dTgs1, resulting in short incomplete F3′,5′H transcripts that could not translate functional F3′,5′H enzyme (Fig. 3c). To investigate alternative F3′,5′H transcripts of cv. Momokorin, the RT-PCR products, 1.3 and 2.1 kb fragments (Fig. 2b), were also subcloned and subjected to sequence analysis. The 2.1 kb of the F3′,5′H transcript contained the sequence of 513 bp GsTRIM1 within the F3′,5′H mRNA, whereas the 1.3 kb fragment of F3′,5′H was shorter than the transcripts of blue-flowered gentian, which might be due to an alternative splicing within the first TDR of GsTRIM1 (Fig. 4b, c). Neither F3’,5’H transcripts of cv. Momokorin could translate any functional F3′,5′H enzymes by the nonsense codon within the first TDR of GsTRIM1. This was also confirmed by 3′RACE analysis of cv. Momokorin transcripts.

Existence of dTgs1 and GsTRIM1 transposable elements in the Gentiana genome

The copy numbers of F3′,5′H, dTgs1, and GsTRIM1 were determined in the three G. scabra lines by Southern blot analysis (Fig. 6). Moreover, to investigate whether both transposable elements were present universally in the genus Gentiana, G. triflora that is a related species to G. scabra was also analyzed. Because the restriction enzymes that did not cut the F3′,5′H gene were used, the analysis suggested that G. scabra cv. Alta, line 13–98 and cv. Momokorin contain at least two copies of F3′,5′H in the genome, while G. triflora cv. Maciry has four copies. Both F3′,5′H loci of cv. Momokorin observed an increase of about 0.5 kb in length by the insertion of the GsTRIM1 transposable element compared with cv. Alta. About a 1.2 kb shift was also observed in EcoR I-digested DNAs in line 13–98 by the insertion of the dTgs1 transposable element, while three Hind III fragments that differ from cv. Alta were observed. In addition, about a 2.5 kb EcoR I fragment was newly observed compared with cv. Alta, suggesting rearrangement or base substitution by the insertion of a transposable element occurring in line 13–98. A number of both dTgs1 and GsTRIM1 transposable elements were detected in the genome of all samples, suggesting that they are members of a relatively high copy number family, the existence of which is widespread in the genome of the genus Gentiana.

Fig. 6
figure 6

Southern blot analysis of F3′,5′H, dTgs1 and GsTRIM1 elements in Gentiana sp. Five microgram genomic DNA of G. scabra cv. Alta (line 1, blue flower), line 13–98 (line 2, pink flower), cv. Momokorin (line 3, pink flower) and G. triflora cv. Maciry (line 4, blue flower) digested by EcoR I (E) or Hind III (H) was separated and hybridized by the probe of F3′,5′H, truncated putative transposase of dTgs1 and GsTRIM1. Numbers on the left of the figure indicate molecular size markers

Full size image

Discussion

In this study, we showed that two pink-flowered gentian plant lines are caused by independent insertions of transposable elements into the first exon of the F3′,5′H gene. Two transposable elements, dTgs1 and GsTRIM1, were classified as class 2 and class 1 transposable elements, respectively, and are interesting cases in which such different types of transposable elements cause the same phenotypic change of flower color. In petunia, corollas with a small number of purple sectors on a predominantly pink background have been identified as the result of inserting a petunia Spm-like (Psl) transposable element into the Hf1 allele, encoding F3′,5′H (Snowden and Napoli 1998).

dTgs1 showed the feature of the hAT superfamily, containing an 8 bp direct duplication of the target insertion site and truncated sequences of putative hAT transposase (Fig. 3, Table 1). The hAT transposon family, such as those containing Ac/Ds of maize (Kunze and Sterlinger 1989) and Tam3 of snapdragons (Hehl et al. 1991), was one of the largest families in the plant mobile elements. dTgs1 contains imperfect inverted repeats of 14 bp that share homology with a number of other hAT transposable elements, with the most similarity to the dTph1 of petunia (Gerats et al. 1990). Furthermore, the terminal 5 nucleotides of dTgs1 were completely consistent with dTph1, Ac/Ds (Muller-Neumann et al. 1984), and Bg of maize (Brown et al. 1989). dTgs1 contained only a truncated putative transposase gene and is probably a nonautonomous transposable element, depending on other transposable elements having mobility components. Insertion of dTgs1 in F3′,5′H gene inhibited normal F3′,5′H transcriptions (Fig. 2a, b) and caused low levels of incomplete F3′,5′H transcripts (Fig. 3c). These suggest that the 3′ region of putative transposase has a functional structure in relation to polyadenylation. Southern blot analysis showed that dTgs1 is abundant and widespread in the genomes of G. scabra and G. triflora (Fig. 6), perhaps a few of them might also contain Tgs1 autonomous transposable elements. However, the ‘revertant’ phenotype, exhibiting blue flowers, has not yet been obtained from line 13 to 98, suggesting that dTgs1 inserted in the F3′,5′H gene has not been activated under normal cultivated conditions. It should be noted that Tam3 of snapdragons was known to be strictly controlled by low temperature (Carpenter et al. 1987), and transposition of transposable elements could also be controlled by DNA methylation (Miura et al. 2001; Hashida et al. 2003). The subterminal sequence of dTgs1 contains a CG-rich sequence (Fig. 3) and it might possibly be silenced by DNA methylation. There is also another possibility to explain why no excision event has been observed in this line. The nucleotide sequences of the terminal ends of dTgs1 were imperfect inverted repeats, namely one nucleotide difference was present between TIR sequences. Such an imperfect TIR sequence could not be recognized by the transposase of dTgs1, so that dTgs1 could not be excised anymore. Further studies will be needed to elucidate the possible factor involved in dTgs1 mobility and to identify autonomous Tgs1 with complete transposase controlling its translocation.

GsTRIM1 showed the feature of TRIM, containing a 5 bp direct duplication of the target insertion site and the triple 76 bp direct repeats of TDR (Fig. 4). Interestingly, the TDR of GsTRIM1 is the shortest among those of the previously reported plant TRIM and LTR-retrotransposons. For example, the TDR of TRIMs presented in both monocotyledonous and dicotyledonous plants was 100–250 bp in length (Witte et al. 2001, 2005), whereas the LTR of Tos17 in O. sativa was 138 bp (Hirochika et al. 1992), and Tnt1 in tobacco was 610 bp (Grandbastien et al. 1989). Because the TDR of GsTRIM1 was an incomplete palindrome sequence (Fig. 5c), its structural features might have not been developed from other LTR-retrotransposons previously reported. Witte et al. (2001) has reported that TRIM was derived from LTR-retrotransposons and also mobilized in trans by other retroelements. Although GsTRIM1 was possibly mobilized by functional autonomous element counterparts, a full-length LTR-retrotransposon with mobility-related polyproteins has not yet been discovered in gentian. L. japonicus, the model plant of the Fabaceae family, was found to contain some GsTRIM1 homologues by a search of genomic data base sequences, but their homologue could not be found in the genomic sequences of other model plants, Arabidopsis thaliana and O. sativa. GsTRIM1 homologues (LjTRIM) are different from the TRIMs reported previously in L. japonicus (Witte et al. 2001). It is probable that the horizontal transmission of retrotransposons between Gentiana and Lotus occurred in the distant past. This was in part supported by the fact that GsTRIM and LjTRIM resemble each other apart from the previously reported other plant TRIMs by a phylogenic analysis (Fig. 5b). Southern blot analysis showed that the genome of the Gentiana species has high copy numbers of GsTRIM1. Because Japanese ornamental gentian species contain a comparatively higher genome size than other closely related gentian species (unpublished data), retrotransposons like GsTRIM1 might contribute to increasing their genome size. In the case of maize, the large genome size might result from amplification of retrotransposon, comprising about 50–80% of the maize nuclear genome (SanMiguel et al. 1996; SanMiguel and Bennetzen 1998). On the contrary, in the light of the fact that the structure of GsTRIM1 consists of triple TDRs, homologous recombination between neighboring TDRs might occur as proposed in LTR-retrotransposons (SanMiguel et al. 1996; Vicient et al. 1999). Such unusual TRIM elements are also reported to be present in Arabidopsis, though the precise formation mechanism is unknown (Witte et al. 2001). Formation of nested GsTRIM1 might be involved in genome size reduction and deactivating retrotransposon (Vicient et al. 1999).

Alternative splicing of F3′,5′H by the insertion of a GsTRIM1 transposable element was observed by RT-PCR analysis (Fig. 2b). This 1.3 kb of transcripts was generated de novo by splicing a site within the TDR of GsTRIM1, and the first exon of F3′,5′H. TRIM (Katydid-At1) in Arabidopsis has also been hypothesized to create a new intron within a gene (Witte et al. 2001). Waxy gene in maize generated multiple alternative splicings but not a newly spliced site by insertion of Stonor, classifying into LTR-retrotransposon, because these insertions disrupted long-range splice site recognition leading to novel transcripts (Varragona et al. 1992). In contrast, ca. 5% of all alternative spliced internal exons in the human genome are derived from Alu elements (Sorek et al. 2002; Kreahling and Graveley 2004). The insertion of a GsTRIM1 transposable element in F3′,5′H generated a de novo 5′ splice site (Fig. 4); thus, the GsTRIM1 transposable element might have the capacity of inducing alternative splicing like Alu elements. Moreover, the insertion of GsTRIM1 also caused reduction of F3′,5′H transcripts (Fig. 2a). A nonsense codon generated by the point mutations of F3′H in Ipomoea has been reported to cause nonsense-mediated decay (NMD), affecting the stability of mature mRNA (Hoshino et al. 2003). The stop codon generated in the GsTRIM1 of cv. Momokorin might also induce the reduction of mature F3′,5′H mRNA by NMD. Retrotransposons have been known to be activated by biotic and abiotic stresses, e.g., salicylic acid, methyl jasmonate, wounding, and cell culture (Kumar and Bennetzen 1999). Since it is unknown what stimuli activate TRIMs, activation factor(s) regulating GsTRIM1 transposition must be further studied.

Taken together, we identified two different transposable elements isolated from pink-flowered gentian plant lines in which their independent insertion events caused the same flower color mutation. There is almost no doubt that transposable elements have influenced genome evolution in gentian plants. Further characterization should lead to a deeper understanding of the roles of these transposable elements during the evolution of gentians.