Development of Selectable Marker-Free Transgenic Rice Plants with Enhanced Seed Tocopherol Content through FLP/FRT-Mediated Spontaneous Auto-Excision

Development of marker-free transgenic plants is a technical alternative for avoiding concerns about the safety of selectable marker genes used in genetically modified (GM) crops. Here, we describe the construction of a spontaneous self-excision binary vector using an oxidative stress-inducible modified FLP/FRT system and its successful application to produce marker-free transgenic rice plants with enhanced seed tocopherol content. To generate selectable marker-free transgenic rice plants, we constructed a binary vector using the hpt selectable marker gene and the rice codon-optimized FLP (mFLP) gene under the control of an oxidative stress-inducible promoter between two FRT sites, along with multiple cloning sites for convenient cloning of genes of interest. Using this pCMF binary vector with the NtTC gene, marker-free T1 transgenic rice plants expressing NtTC were produced by Agrobacterium-mediated stable transformation using hygromycin as a selective agent, followed by segregation of selectable marker genes. Furthermore, α-, γ-, and total tocopherol levels were significantly increased in seeds of the marker-free transgenic TC line compared with those of wild-type plants. Thus, this spontaneous auto-excision system, incorporating an oxidative stress-inducible mFLP/FRT system to eliminate the selectable marker gene, can be easily adopted and used to efficiently generate marker-free transgenic rice plants. Moreover, nutritional enhancement of rice seeds through elevation of tocopherol content coupled with this marker-free strategy may improve human health and public acceptance of GM rice.


Introduction
Rice is one of the most important food crops consumed worldwide. Rice is also a good source of vitamin E, an essential lipid-soluble nutrient that consists of four tocopherols and four tocotrienols. Each of these types of compounds has α-, β-, δ-, and γ-forms determined by the number of methyl groups on the chromanol ring. Tocopherols can efficiently quench singlet oxygen and scavenge various radicals, particularly lipid peroxy radicals, thereby terminating lipid peroxidation chain reactions [1,2]. Tocopherols are important constituents of the human diet and have been shown to aid in immune function [3] and to decrease the risk of a number of degenerative diseases, such as Parkinson's disease [4] and heart disease [5]. Additionally, recent reports have demonstrated that tocopherols can affect important physiological processes in plants, such as germination, photoassimilate export, growth, and leaf senescence; tocopherols also have antioxidant functions in photosynthetic membranes and play important roles in plant responses to abiotic stresses [6].
In addition to its importance as a source of vitamin E, rice is also an important model system for functional identification of genes in monocots. Several transformation systems have been developed in rice plants using different approaches, such as protoplast transformation, particle bombardment transformation, and Agrobacterium-mediated transformation methods [7][8][9]. The advantages of Agrobacterium-mediated gene transfer over other methods include the high efficiency of transformation, the ability to transfer relatively large segments of DNA, and the relatively short duration of transformation [10]. Thus, Agrobacterium-mediated methods have become popular in recent years.
In general, antibiotic-or herbicide-resistant genes linked to a transgene have been used as selectable marker genes for the preferential growth of transformed cells in transformation systems. However, once transformation is accomplished, the presence of these resistance genes is no longer necessary and can even be undesirable. Moreover, the presence of selectable marker genes in genetically modified (GM) crops has generated a number of environmental and consumer concerns, despite scientific evidence supporting the safety of GM crops [11]. However, regardless of scientific assurance of the safety of selectable marker genes, consumer acceptance is ultimately the most important element determining the successful commercialization of transgenic plants and products. Therefore, marker gene elimination is desirable in order to alleviate public concerns over the safety of food derived from transgenic crops.
Several strategies have been developed to generate selectable marker-free transgenic plants genes using different transformation systems. These strategies include co-transformation, sitespecific recombination, transposon-based methods, and intrachromosomal homologous recombination [11][12][13][14]. The principle of the co-transformation strategy is the introduction of a selectable marker gene and a gene of interest from different T-DNAs. If the two genes are integrated into unlinked loci, subsequent segregation can separate the gene of interest from the selectable marker gene. Marker genes of transgenic plants can also be removed by site-specific recombination systems, such as Cre/lox, FLP/FRT, and R/RS. Several types of site-specific recombination systems function in plants [11]. Among these, the Cre/lox system from bacteriophage P1 has been most extensively used for the generation of marker-free plants. Moreover, strategies for generation of marker-free plants via site-specific recombination systems require either the transient expression of the recombinase gene, crossing with a recombinase-expressing line, or an inducible factor to turn on the expression of the recombinase gene. Among these methods, auto-excision using an inducible promoter has been developed due to the advantages of reduced time requirements of avoidance of crossing procedures. Several inducible systems responsive to external stimuli have been reported for plants. The heat-shock regulated system has been shown to be functional in Arabidopsis [15], tobacco [16], potato [17], maize [18], rice [19], and aspen [20]. The chemically regulated self-excision system, i.e., combination of the Cre gene with the XVE system, has been successfully applied in Arabidopsis [21], rice [22], and tomato [23,24]. In another FLP/FRT system from Saccharomyces cerevisiae, the selectable marker gene of transgenic tobacco was removed by a stress-inducible FLP/FRT site-specific recombination system [25]. In this system, an oxidative stress-inducible peroxidase (POD) promoter is fused to the recombinase gene FLP. Hydrogen peroxide is then used to induce expression of the recombinase gene, and successful excision of selectable marker genes via hydrogen peroxide-regulated site-specific recombination is achieved. Despite the advantages of temporally regulated recombinase expression, heat-shock or chemically regulated promoters require an external signal to be activated, and the recombination frequencies are greatly dependent on the penetration of the signal into plant cells [26]. Therefore, these systems may be needed further improvements for broad applications in the agricultural industry.
In our present study, we describe the construction of a spontaneous self-excision binary vector using the oxidative-stress inducible mFLP/FRT system for rice transformation. In a recent report, Woo et al. [27] showed that overexpression of NtTC could increase the tocopherol contents in leaves of rice plants. Therefore, for efficient generation of marker-free transgenic rice plants via the spontaneous auto-excision method with enhancement of the tocopherol content of rice seeds, we generated marker-free T 1 transgenic rice plants overexpressing NtTC by Agrobacterium-mediated transformation using hygromycin as a selective agent, followed by segregation of selectable marker genes and analysis of tocopherol content.

Vector construction and codon optimization
Molecular manipulation methods, such as plasmid DNA isolation, restriction enzyme analysis, ligation of DNA fragments, and transformation of Escherichia coli, were performed as described previously [28]. The binary vectors for this study were constructed as follows. The 6.6-kb backbone DNA fragment for binary vector construction was obtained by polymerase chain reaction (PCR) from the pCAMBIA2301 vector (CAMBIA, Canberra, Australia) with the following primers, incorporating the HindIII and SacI sites: forward primer, 5 0 -AAGAGCTCAACGT CCGCAATGTGTTATTAAGTTGTCTAAGCG-3 0 and reverse primer, 5 0 -AAAAGCTTGGT GACCAGCTCGAATTTCCCCGA-3 0 . The 1,817-bp DNA fragment containing the P35S-FRT m-hpt-T35S cassette and 2,964-bp Ppod-FLP-Tnos-MCS cassette were synthesized (Bio Basic Inc., Markham, Canada). The P35S-FRTm-hpt-T35S cassette was excised with SacI and SpeI and inserted into the same restriction enzyme sites of the pBluescript II KS vector (Stratagene, La Jolla, CA). The Ppod-FLP-Tnos-MCS fragment was excised with SpeI and HindIII and subcloned into the SpeI/HindIII site of the pBluescript-P35S-FRTm-hpt-T35S plasmid to generate pBluescript-P35S-FRTm-hpt-T35S-Ppod-FLP-Tnos-MCS. The insertion fragment was then excised from this construct with SacI and HindIII and ligated with the backbone fragment treated with SacI/HindIII restriction enzymes to generate the binary vector pHWMF.
The FLP recombinase gene originated from a 2-μm plasmid in S. cerevisiae. To optimize codon usage, the relative synonymous codon usage (RSUC) values of Oryza sativa and the native FLP gene were calculated using the information obtained from the codon usage database (http://www.kazusa.or.jp/codon/). RSUC values were calculated by dividing the observed codon usage by that expected when all codons for the same amino acids were used equally [29]. The entire mFLP gene was synthesized based on the RSUC values of O. sativa and the codon usage of the rice high-GC gene [30]. The synthetic mFLP gene contained the BglII sites at the 5 0 and 3 0 ends to facilitate gene exchange. The mFLP gene, excised from the pUC57 vector with BglII, was inserted into the corresponding FLP gene site of the pHWMF vector. The resulting binary expression vector, designated as pCMF, was verified by DNA sequencing and restriction enzyme analysis. The full-length NtTC gene (Genbank accession number KJ645980) was obtained from pMJ102TC [27] by PCR with the following primers incorporating BamHI and HindIII sites: forward primer, 5 0 -TTGGATCCATGGAGAACATATACGATTTTTCCACCA TCTCTTC-3 0 and reverse primer, 5 0 -TTAAGCTTACAGGCCAGGAGGTTTGAGAAA TGGAG-3 0 . The gene was then inserted into MCS of the pCMF vector with BamHI and Hin-dIII, yielding pCMF-TC.

Rice transformation and growth conditions
pCMF-TC was introduced into A. tumefaciens strain LBA4404 for genetic transformation of rice (O. sativa subsp. japonica cv. Dongjin). Rice transgenic plants were generated by the Agrobacterium-mediated cocultivation method [31], and transformants were selected based on hygromycin resistance. The regenerated shoots were transferred to Murashige and Skoog (MS) solid medium for root induction for 2 weeks and then transferred to 0.1% Hyponex solution (Hyponex, Imlay, MI, USA) for 1 week in a culture room for acclimatization. Regenerated plantlets were subsequently transplanted into the soil in pots and grown in a greenhouse. Transgenic rice plants and seeds were sown in a vermiculite and potting soil mixture soaked with water. All plants were grown under white fluorescent light (600 μmol m -2 s -1 , 12 h photoperiod) and 75% relative humidity at 28°C in a greenhouse.

Southern blot analysis
Genomic DNA was isolated from mature leaves of rice using the CTAB DNA extraction method [32]. About 20 μg of total DNA was digested overnight with EcoRV. The digested DNA was separated on 0.8% agarose gels and transferred onto a Hybond-N+ membrane (Amersham Biosciences, Piscataway, NJ, USA). The blot was hybridized with the DIG-labeled probe, and the signal was detected with the Fusion FX imaging system (Vilber Lourmat, Eberhardzell, Germany). DNA probes for hygromycin and the tocopherol cyclase gene were obtained by PCR using the following primers: hpt, 5 0 -TTCTGATCGAAAAGTTCGACAGC GTCTCC-3 0 and 5 0 -AATTGCCGTCAACCAAGCTCTGATAGAGTTG-3 0 ; and NtTC, 5 0 -AATTTGAGCACTTTTCCGTCTACCTTGAAGCTG-3 0 and 5 0 -CAGCAGTCAAAGC AACATCTCCAATCGC-3 0 . DNA probe preparation, hybridization, and membrane washing were performed using a DIG High Prime DNA Labeling and Detection Starter Kit II according to the manufacturer's instructions (Roche Inc., Mannheim, Germany).

Seed germination assay
Self-pollinated seeds were harvested from wild-type and T 0 transgenic rice plants, surface-sterilized for 10 min in 1.5% sodium hypochloride solution, and rinsed thoroughly (3-4 times) in sterile water. Surface-sterilized seeds were grown on germination medium (MS medium 2% [w/v] sucrose and 0.7% [w/v] agar) containing hygromycin (30 mg L -1 ) or without any antibiotics. Ten days later, seedlings grown on MS media were evaluated for resistance.

Tocopherol extraction and analysis
Mature seeds harvested from T 1 transgenic rice plants and wild-type rice seeds were husked and ground to a fine powder using a stainless-steel planetary ball mill (Pulverisette 6, Fritsch GMBH, Germany). Alpha-and gamma-tocopherol in rice seeds were quantified by gas chromatography (GC) coupled with time-of-flight mass spectrometry (TOF-MS) as described by Park et al. [33]. Quantitative calculations were based on the corrected peak area ratios relative to the peak area of the internal standard (IS). Linearity was tested by least-squares regression analysis of the corrected peak area ratios against increasing weight ratios (relative to IS).

Construction and characteristics of pHWMF, pCMF, and pCMF-TC vectors
The region between the left and right T-DNA borders in the binary vector pHWMF is depicted schematically in Fig 1, in which the hygromycin-selectable marker gene (hpt) and the recombinase gene FLP under the control of the oxidative stress-inducible promoter of the sweet potato were placed between two opposite FRT sites.
The multiple cloning sites contained unique restriction sites, where the gene of interest could be inserted, located between the FRT site and nopaline synthase (nos) gene terminator. The pHWMF vector was constructed with modular components having unique pairs of restriction sites flanking each genetic element within the T-DNA region, enabling the replacement of each of genetic element by equivalent counterparts with ease. Additionally, we used the modified FRT site (FRTm) and one full-length FRT site in the pHWMF vector. Although FRTm contained only two symmetry elements for the FLP protein compared with the full-length FRT sequence having three FLP binding sites, this modification did not substantially affect the function of the FLP recombinase and provided a means to clearly distinguish site-specific recombination products from possible artifacts generated by other genomic DNA modifications [34].
Genes with a high GC content are generally expressed at high levels in rice [30]. Furthermore, codon bias has a significant influence on protein yields. When we compared the codon bias of the native FLP recombinase gene from yeast with the codon usage in coding sequences of O. sativa using the codon usage database, we found distinct differences (Table 1).
For example, proline can be encoded by the codons CCU, CCA, CCC, and CCG. However, the amino acid proline in the native FLP gene was mostly encoded by the triplets CCU and CCA, while CCC and CCG were rare codons. In contrast, CCG was the favored codon for proline in O. sativa and was used significantly more often in high-GC genes than in low-GC genes (GC content < 60%) in the rice genome. Furthermore, the S. cerevisiae FLP gene had a GC content of 37%. In contrast, the nuclear genome of O. sativa has an overall GC content of 43% [35], with coding regions having a GC content of 55% (as determined using the codon usage database). Due to these differences between S. cerevisiae and O. sativa, we synthesized the modified FLP recombinase (mFLP) gene with a bias for preference in high-GC genes of rice. The resulting sequence had a GC content of 61% with 70% DNA homology to the native FRT Tocopherol cyclase (TC/VTE1) catalyzes the conversion of 2,3-dimethyl-5-phytyl-1,4-benzoquinone (DMPBQ) to γ-tocopherol (S2 Fig). Previous studies have reported the characterization of NtTC, a tocopherol cyclase ortholog originating from tobacco, and have demonstrated a relative increase in tocopherol content in the leaves of transgenic rice plants upon overexpression of NtTC [27]. To enhance the tocopherol contents of rice seed, the plasmid binary pCMF-TC was constructed as shown in Fig 1 and introduced into A. tumefaciens strain LBA4404 for rice transformation. An asterisk (*) after the codon indicates a favored codon (RCSU > 1) in genes of coding DNA sequences in the rice genome, and a plus sign (+) indicates a favored codon (RCSU > 1) in the native FLP gene. Codons used for synthesis of the mFLP gene are shown as bold text and were favored codons in high-GC genes within the rice genome [30]. Relative synonymous codon usage (RSCU) was defined as the ratio of the observed frequency of codons to the expected frequency given that all synonymous codons for the same amino acids were used equally.  feasible by segregation in the T 1 generation. Therefore, some of T 1 progenies from the TC2 and TC3 line were analyzed for the presence of the hph and NtTC transgene by Southern blotting. All five T 1 progenies of the TC2 line and three T 1 progenies of the TC3 line were randomly chosen, and nontransformed wild-type Dongjin was used as control for Southern blotting. All of these T 1 plants exhibited multiple NtTC bands, and progenies of crosses between the TC2 and TC3 lines showed different band patterns representing independent transgenic lines. In addition, a weak 9.4-kb band also appeared in the wild-type strain; this band may have been associated with the endogenous rice tocopherol cyclase gene, which exhibited high similarity with the NtTC probe. In the case of hybridization with the hpt probe, two T 1 progeny of the TC2 line (TC2-2 and TC2-3) and one T 1 progeny of the TC3 line (TC3-3) did not exhibit a signal, thus confirming complete excision (Fig 3).
To validate these Southern blotting results, the excision status of T 1 rice progeny from transgenic lines TC2 and TC3 was also analyzed by PCR using the P35SF/HTPR and P35SF/TCR primer sets (Fig 4A). PCR analyses indicated the presence of the expected P35S-NtTC fragment and the absence of the P35S-hpt gene in the TC2-2, TC2-3, and TC3-3 lines, consistent with Southern blot analysis data. Moreover, these marker-free rice lines did not exhibit expression of the hpt gene but expressed high levels of NtTC, as shown by RT-PCR analysis ( Fig 4B).
Next, we determined the hygromycin resistance of T 1 seeds of transgenic rice lines. The four transgenic lines confirmed above, in which the hpt gene was excised or existed, were chosen for resistance analysis. Rice seeds derived from the TC2 and TC3 lines that had been selfed were sown on hygromycin-supplemented medium. As shown in Fig 5, the TC2-1 and TC3-1 lines, which were chimeric with excised and unexcised T-DNA loci coexisting in the chromosomal DNA, grew on hygromycin-supplemented medium, while TC2-2 and TC3-3 did not, indicating that the hpt gene was excised successfully by mFLP/FRT-mediated marker gene excision. In

Tocopherol contents of transgenic rice grains
The TC3-1 transgenic rice line, carrying a single copy of the transgene in its genome, and the untransformed controls were grown in the same field under normal culture conditions, and  mature seeds were harvested for determination of tocopherol content using GC coupled with TOF-MS. In rice, βand δ-tocopherol usually appear in trace amounts, while α-tocopherol and γ-tocotrienol are the most abundant isomers [36,37]. Therefore, these two isomers were not included, and the total tocopherol content is given here as the sum of αand γ-tocopherol.
The contents of α-, γ-, and total tocopherol in the seeds of wild-type rice plants were 6.57 ± 1.05, 0.39 ± 0.03, and 6.96 ± 1.06 μg/g dry weight, respectively. In the case of the transgenic TC3-1 line, α-, γ-, and total tocopherol levels were approximately 53%, 194%, and 61% higher than those in seeds of wild-type plants (Fig 6; p < 0.01). These results demonstrated that overexpression of NtTC could increase tocopherol levels in rice seeds.

Discussion
Selectable marker genes, such as antibiotic-or herbicide-resistant gene, are powerful selection tools for use in plant transformation processes. However, once transformation is accomplished, the presence of these resistance genes is no longer necessary and can even be undesirable. Therefore, various genetic approaches have been developed to produce marker-free transgenic plants, and several techniques, including cotransformation and auto-excision using site-specific recombination, have been successfully established for the elimination of selectable marker genes [11,14,38]. However, these methods still require additional steps or more labor to obtain marker-free transgenic plants, and further improvements are needed for practical application and conventional use. Accordingly, in this study, we developed a spontaneous self-excision binary vector using an oxidative stress-inducible modified FLP/FRT system. This system was successfully applied to produce marker-free transgenic rice plants with enhanced seed tocopherol content.
The oxidative-stress inducible POD promoter from sweet potato was not expressed in any tissues in differentiated plants, but was strongly expressed and induced by environmental stress conditions, including hydrogen peroxide and wounding in cultured cells [39]. Also, spontaneous excision of a marker gene in an oxidative stress-inducible FLP/FRT system was reported in tobacco plants without chemical induction, presumably resulting from activity of the POD promoter in transgenic cells during the transformation process [39].
In this study, in order to develop a simple and convenient method of producing marker-free transgenic rice plants using the oxidative-stress inducible FLP/FRT system, we synthesized the pHWMF vector, in which the site-specific recombinase gene FLP and selectable marker gene hpt were inserted between two FRT sites. The FLP gene was under the control of oxidativestress inducible POD promoter; thus, both the selectable marker and mFLP genes could be excised spontaneously in the rice callus stage of transformation. However, when the pHWMF vector was introduced into E. coli or A. tumefaciens, weak gene excision was observed in our experiments, possibly due to the expression of the FLP gene driven by leakiness of POD promoter activity. This phenomenon has been reported by other laboratories using chemical-or heat shock-inducible promoters for recombinase gene expression [40] and could be associated with the maintenance of plasmid constructs and transformation efficiency. To solve this problem, the FLP gene was modified based on the codon usage criteria for rice and was altered to have elevated GC content. As a result, pCMF containing the mFLP gene controlled by the POD promoter was stably maintained in E. coli and Agrobacterium (data not shown). Additionally, this modification may contribute to the efficient expression of the construct in rice and other cereals with the same GC content preference.
In Agrobacterium-mediated transformants, transformed T-DNAs are often integrated into different loci in the plant genome, and more than one-third contain a single T-DNA insert, while half contain 2-3 copies and the remainder (about 15%) contain 4-5 copies [41]. The principle of the spontaneous auto-excision strategy doe generation of marker-free transgenic rice plants is the introduction of multiple T-DNAs. If the T-DNAs were integrated into unlinked loci, and gene excision occurred in some of the T-DNAs resulting from stress stimuli during the transformation procedure, excised T-DNA was inherited independently, consequently producing marker-free transgenic plants in the subsequent T 1 generation (Fig 7). Although we did not attempt to analyze the excision rates of all regenerated transgenic plants, about half of regenerated plants obtained by hygromycin selection presented a P35S-NtTC fragment, representing the expected PCR amplification band of an excised T-DNA. Moreover, marker-free T 1 transgenic rice plants were generated from two T 0 transgenic lines among the five transgenic lines analyzed in this study. Thus, this method was effective for production of marker-free transgenic rice plants. Compared with other methods for developing marker-free transgenic plants, this spontaneous auto-excision strategy has various unique advantages. For example, this strategy does not require additional procedures for eliminating the selectable marker gene, such as crossing of the recombinase gene or chemical treatment, and can therefore be applied with general rice transformation methods to obtain marker-free rice plants.
We attempted to produce the marker-free transgenic rice plants overexpressing NtTC using the pCMF-TC vector in order to verify efficient generation of marker-free transgenic rice plants via the spontaneous auto-excision method and improve the tocopherol content of rice seeds. Tocopherols, which are collectively known as vitamin E, are essential antioxidant components of both human and animal diets, and the main source for the dietary uptake of tocopherols is plants, consumed as seeds, vegetables, fruits, and plant seed oils. Hence, increased levels of tocopherols in rice grains would be greatly beneficial for human health.
Here, we demonstrated that overexpression of NtTC under the CaMV 35s promoter resulted in increases levels of total tocopherols by approximately 1.6-fold in T 2 seeds of marker-free transgenic rice plants. In addition, γ-tocopherol content was significantly increased by 194% compared to that in wild-type plants. Similar results have also been reported in other transgenic rapeseed plants expressing the TC gene from Arabidopsis or maize; indeed, the seeds of transgenic B. napus plants exhibit significantly enhanced tocopherol content by a factor of 1.6 [42]. Thus, these results implied that TC activity may be the limiting factor for tocopherol biosynthesis, and overexpression of TC influences the total tocopherol content of plants.
In conclusion, the results reported herein demonstrate that the spontaneous auto-excision method using a FLP/FRT recombination system is able to produce marker-free transgenic rice plants with enhanced seed tocopherol content by Agrobacterium-mediated transformation with antibiotic selection. This method is simple and does not require additional procedures to eliminate the selectable marker gene. Hence, this method provides a useful tool for production of marker-free transgenic rice plants. Importantly, this system may also be applicable to other cereals. Additionally, the nutritional enhancement of rice seeds through elevation of tocopherol content coupled with this marker-free strategy may improve human health and public acceptance of GM rice.
Supporting Information S1 Fig. DNA sequence alignment of the native and modified FLP genes. The alignment was created using ClustalW2 software with sequences available in the EMBL database. DNA