Site-specific integration of transgenes in soybean via recombinase-mediated DNA cassette exchange.

A targeting method to insert genes at a previously characterized genetic locus to make plant transformation and transgene expression predictable is highly desirable for plant biotechnology. We report the successful targeting of transgenes to predefined soybean (Glycine max) genome sites using the yeast FLP-FRT recombination system. First, a target DNA containing a pair of incompatible FRT sites flanking a selection gene was introduced in soybean by standard biolistic transformation. Transgenic events containing a single copy of the target were retransformed with a donor DNA, which contained the same pair of FRT sites flanking a different selection gene, and a FLP expression DNA. Precise DNA cassette exchange was achieved between the target and donor DNA via recombinase-mediated cassette exchange, so that the donor DNA was introduced at the locus previously occupied by the target DNA. The introduced donor genes expressed normally and segregated according to Mendelian laws.

4 introduced as a circular DNA (Horn and Handler, 2005). A gene conversion approach involving Cre-lox and FLP-FRT mediated site-specific integration, RMCE, and homologous recombination was explored in maize (Djukanovic et al., 2006). RMCE using two oppositely oriented incompatible lox sites and transiently expressed Cre recombinase produced single copy RMCE plants in Arabidopsis (Louwerse et al., 2007).
To develop FLP-FRT mediated RMCE in soybean, we created transgenic target lines containing a hygromycin resistance gene flanked by two directly oriented incompatible FRT sites via biolistic transformation. Single copy target lines were selected and retransformed with a donor DNA containing a chlorsulfuron resistance gene flanked by the same pair of FRT sites. A FLP expression DNA was co-bombarded to transiently provide FLP recombinase. RMCE events were obtained from multiple target lines and confirmed by extensive molecular characterization.

Design of FLP-FRT Mediated RMCE
The target QC288A and donor QC329 DNA each contained a FRT1 site and a FRT87 site in the same orientation (Figs. 1A and 1B) (Tao et al., 2007) sites. FLP recombinase could then excise the DNA segment between the two FRT1 sites to restore the original target QC288A, or excise the DNA segment between the two FRT87 sites to form the recombined RMCE DNA QC288A329 (Fig. 1D). Since the FRT1 and FRT87 sites are not completely incompatible, all DNA between the outmost FRT1 and FRT87 sites could also be excised resulting in an excision event retaining only the scp1 promoter and a recombined FRT site (Fig. 1E). Depending on the DNA strands crossover position, excision between FRT1 and FRT87 sites could restore either the FRT87 or the FRT1 site (McLeod et al., 1986). If no excision occurred, the intermediate could remain as a simple integration containing all the QC288A and QC329 components. The FLP recombinase was provided transiently from the expression of the flp construct QC292 (Fig. 1C).
The QC288A DNA contained a selection gene hpt driven by a constitutive promoter scp1 and transgenic events were selected with hygromycin. The QC329 DNA contained a promoter-less selection gene als that would not be expressed unless a promoter was placed upstream of it. FLP mediated DNA recombination could bring the promoter-less als gene of QC329 downstream of the scp1 promoter of QC288A to form QC288A329 to enable retransformation events being selected with chlorsulfuron. The random integration events of QC329 would not survive the selection unless the promoter-less als gene inserted by chance downstream of a native promoter. The fluorescent protein genes yfp in QC288A and cfp in QC329 and QC288A329 were used to screen transgenic events (Figs. 1A,1B,and 1D). Suspension cultures initiated from the developing embryos of target lines A, B, and C homozygous T1 plants were retransformed with the donor DNA QC329 co-bombarded with the flp DNA QC292. Three putative retransformation events resistant to chlorsulfuron from target A, six from target B, and three from target C were screened at the callus stage for the reporter gene cfp expression followed by a common PCR to check DNA recombination at the FRT1 site (Table 2). Event B5 and B6 were derived from the retransformation of the original hemizygous target B callus that had never gone through plant regeneration. All events were then evaluated by four construct-specific qPCR analyses ( Fig. 1), to check for DNA recombination at the FRT1 site and the presence of the target, donor, and flp DNA (Table 2), followed by five border-specific PCR analyses specific to each target line using the 5' border, 3' border, and transgene-specific primers (Figs. 1A,1D,and 1E) to confirm DNA recombination at and between the FRT1 and FRT87 sites (Fig. 2).

Creation and Characterization of Target Events
For example, event A1 was positive for both CFP expression and the FRT1 sitespecific PCR. The construct-specific qPCR analyses revealed that event A1 had one copy of RMCE, contained two copies of the donor, and was free of either the target or flp (Table 2). The border-specific PCR analyses revealed that event A1 was positive for both the 5' end and 3' end assays specific to RMCE (Figs. 2A and 2B), negative for either the 5' end or the 3' end assays specific to the target (Figs. 2C and 2D), and positive for a small excision-specific band amplified by the full length PCR (Fig. 2E). Since one copy of RMCE was simultaneously detected with the excision in the homozygous target derived event A1, the event had to be a RMCE-excision with one target converted to RMCE and the other converted to excision. The expected large RMCE-specific band

8
(6652 bp) of event A1 failed to be amplified by the same full length PCR due to its competitive disadvantage to the small excision-specific band (1307 bp).
Based on similar qPCR (Table 2) and border-specific PCR analyses (Fig. 2), event A2 was an RMCE-excision event containing a copy of the donor and flp DNA. Event A3 was a homozygous target escape containing 5 copies of the donor. The target-specific band of A3 was amplified by the full length PCR (Fig. 2E). Events B1, B2, and B4 were clean RMCE-excision events containing no additional donor or flp DNA insertions.
Event B3 was an RMCE-excision containing a donor. Event B5 was an RMCE-wt (wild type) since it was derived from the retransformation of the original hemizygous target B callus. Accordingly, no excision band was detected in event B5 by the full length PCR even though the large RMCE specific-specific band was amplified (Fig. 2E). The 1.0 copy of the target in event B5 probably was a partial copy since the target border-specific PCR did not detect it (Figs. 2C and 2D). Event B6 was a hemizygous target escape, also containing the partial target since 1.8 copies of the target were detected. The detection of the partial target in events B5 and B6 suggested that the original target B callus was chimeric. Event C1 was a RMCE-excision with some cells still containing the target detected as 0.01 copy, which was confirmed by the border-specific PCR (Figs. 2C and 2D). Events C2 and C3 were homozygous RMCE-RMCE events containing two copies of RMCE and one copy of the donor. Accordingly, the border-specific PCR failed to detect any target or excision-specific bands but amplified the large RMCE-specific band ( Fig. 2C, 2D, and 2E).
To summarize, 2, 5, and 3 RMCE events were obtained from the retransformation of 5, 5, and 6 plates of target A, B, and C cultures, the RMCE retransformation frequencies were thus calculated as 0.4, 1, or 0.5 event/plate ( Table 2). The average of these frequencies is ~10 times lower than the average 5 events/plate frequency for standard soybean biolistic transformation but high enough for routine RMCE event production.
The 1.0 copy of target detected in B5 callus was no longer observed in T0 plants B5-1, B5-2, or B5-3. The same border-specific PCR analyses also confirmed that the T0 plants were the same as their respective callus parents (Fig. 3).
Since the target QC288A and the RMCE QC288A329 sequences diverge downstream of the FRT1 site with hpt in QC288A and als in QC288A329, and upstream of the FRT87 site with yfp:nos in QC288A and cfp:nos in QC288A329 (Figs. 1A and 1D), the alignment of the target and RMCE transgene sequences with their map sequences should confirm RMCE at sequence level. The 30 bands, marked with "x" in
Scp1 only bands were specific to excision that contained only the scp1 promoter. The excision-specific fragment was produced by digestion at the same NdeI site in the 5' genomic DNA border and another NdeI site in the 3' genomic DNA border (Fig. 1E). The Southern hybridization results were consistent with previous qPCR and PCR results except for a large scp1 band detected in C3-1, and C3-2, and extra ubq bands detected in target A derived plants (Figs. 4B and 4C). The large scp1 bands of C3-1 and C3-2 were not from RMCE or excision that otherwise would be detected by RMCEspecific qPCR or full length border-specific PCR. Since the band also hybridized to the ubq probe and disappeared with the donor in C3-1-1 and C3-1-2, it was considered as a partial scp1 promoter mingled with the donor at an unlinked random insertion site. Four of the five ubq only bands below the ~6 kb wt band detected in A2-3 and A2-4 ( Fig. 4D) were likely partial copies of the donor since qPCR detected only one donor insert ( Table   2). One of the ubq only bands, not detected by the donor-specific qPCR, remained in RMCE-excision plants A2-3-1 and A2-3-2 and two remained in excision-excision plants A2-3-3 and A2-3-4.

DISCUSSION
Single site SSI creates two directly oriented recognition sites vulnerable to excision that makes the recombination events unstable. Mutant lox sites (Albert et al., 1995;Srivastava and Ow, 2001), cre gene displacement, and transient cre expression (Albert et al., 1995;Vergunst et al., 1998), can be used to prevent the excision. A donor DNA can be circularized prior to integration by a recombinase to remove any unwanted components such as the vector backbone to prevent them from being integrated  (Vergunst et al., 1998;Srivastava and Ow, 2001). To achieve marker-free SSI, a two-step approach was proposed to combine gene integration using one recombinase system such as Cre-lox with gene excision using another system such as FLP-FRT conditionally controlled by an inducible promoter (Srivastava and Ow, 2004). RMCE using two recognition sites provides a flexible way for gene targeting. If two identical sites are used, they must be in opposite orientations to prevent excision, though the DNA segment between the two sites can flip (Nanto et al., 2005). Two incompatible sites such as loxP and lox5171 can also be arranged in opposite orientations for successful RMCE (Louwerse et al., 2007). Preferably, RMCE using two directly oriented incompatible sites can avoid the excision or flipping of the flanked DNA segment and has succeeded to some extents in animal systems with a loxP and a mutant loxP511 14 RMCE-excision and the lack of RMCE-target in our experiments indicate that the FRT1 and FRT87 sites are not completely incompatible and that the FLP mediated DNA recombination is highly effective. RMCE can even occur simultaneously on two homologous chromosomes such as in the case of events C2 and C3. More likely, a homozygous RMCE has to be obtained at the T1 generation via segregation. Any donor or flp DNA integrated randomly at a separate genomic site in an RMCE event can be removed by segregation.
The effective RMCE described here opens new ways for transgenic products development and transgene expression studies. Large DNA fragments can be integrated via RMCE which seems to rely only on FLP catalyzed interactions between FRT sites.
Various target lines can be produced in advance and maintained as production lines to accept genes with various preferences for gene silencing, tissue-specific expressions, agronomic performances, etc. Multiple genes can be stacked reversibly at the same genetic locus by multiple rounds of RMCE using different recombination sites.

DNA Construction
The 15 molecular cloning procedures through multiple steps using components from existing DNA constructs (Li et al., 2007).

Plant Transformation
The target DNA scp1-FRT1:hpt:nos+ubiq10:yfp:nos-FRT87 was released as a 4544 bp fragment QC288A with AscI digestion from QC288, resolved by agarose gel electrophoresis, and purified using the gel extraction kit (Qiagen). Soybean embryogenic cultures were transformed with QC288A following the biolistic transformation protocol using 30 µg/ml hygromycin for selection (Li et al., 2007). Transgenic plants (T0) were regenerated from single copy events identified by qPCR and Southern. The original target line hemizygous cultures or cultures initiated from the developing embryos of target homozygous T1 plants were co-bombarded with the donor QC329 and flp QC292 plasmid DNA at 9:3 ratios following the same biolistic transformation protocol except using 90 ng/ml chlorsulfuron (DuPont) to select retransformation events.

Transgene Cloning and Sequencing
Genomic DNA bordering the QC288A transgene was acquired using the GenomeWalker kit (ClonTech). Genomic DNA digested with EcoRV, DraI, HpaI, or StuI was ligated to adaptors and amplified by two rounds of PCR. The first PCR used adaptor-specific primer AP1 gtaatacgactcactatagggcacg and QC288A-specific primers Scp1-A ctactgtccttttgatgaagtgacag for the 5' end border and Vec-S1 gatcgggaattctagtggccgg for the 3' border. were done using Vector NTI programs (Invitrogen). Sequence searches were done using the NCBI (www.ncbi.nlm.nih.gov) advanced BLAST algorithm.

PCR Analysis
PCR was done on leaf or somatic embryo DNA samples following the same protocol 18 C. The target 5' border PCR used the same target line 5' border-specific primers and a common target-specific primer Hygro-A. The target 3' border PCR used the same target line 3' border-specific primers and a common target-specific primer Yfp-3. The full length PCR used the same target line 5' and 3' border-specific primers to simultaneously amplify the full length RMCE, target, and excision. The expected sizes of all PCR bands are described in the figure 2 legend.

Quantitative PCR Analysis
qPCR analyses were done on genomic DNA samples using the Taqman DNA polymerase kit with a 7500 real time PCR system (Applied Biosystems). The relative quantification methodology and single tube duplex PCR reactions, one for a target gene and the other for an endogenous control gene to normalize reactions, were used. After 2 minutes incubation at 50 °C to activate the Taq DNA polymerase and 10 minutes incubation at 95 °C to denature the DNA templates, 40 cycles of 15 seconds at 95 °C and 1 minute at 60 °C were used. A soybean heat shock protein (hsp) gene was used as the endogenous control for all assays. Primers Hsp-F1 caaacttgacaaagccacaactct, Hsp-R1 ggagaaattggtgtcgtggaa, and probe Hsp-T1 VIC-ctctcatctcatataaatac-MGB (Applied Biosystems) were used for the hsp control. A DNA sample known containing one copy of the transgene to be analyzed was included as the calibrator for each qPCR assay. Primers Ucp3-57F tcgagcggctataaatacgtacct, Flp-A gtcttgcagaggatgtcgaactgg and probe FAM-cctgcgctaccatccctagagctgc-BHQ1 were used for the flp QC292-specific qPCR.
Novel materials described in this publication may be available for non-commercial research purposes upon acceptance and signing of a material transfer agreement. In some cases such materials may contain or be derived from materials obtained from a third party. In such cases, distribution of material will be subject to the requisite permission from any third-party owners, licensors or controllers of all or parts of the material.
Obtaining any permission will be the sole responsibility of the requestor. Plant    border, 3' border, and full length of RMCE, target, and excision were done using various combinations of the 5' border, 3' border, and transgene-specific primers (Fig. 1).
The 5063, 4742, and 4907 bp target bands "t" were amplified for target A, B, and C.
The thirty bands marked with "x" were cloned and sequenced.