• Open Access

Gene targeting in maize by somatic ectopic recombination

Authors


Correspondence (Tel +(0)4 73 67 88 40; fax +(0)4 73 67 88 99; email wyatt.paul@biogemma.com)

Summary

Low transformation efficiency and high background of non-targeted events are major constraints to gene targeting in plants. We demonstrate here applicability in maize of a system that reduces the constraint from transformation efficiency. The system requires regenerable transformants in which all of the following elements are stably integrated in the genome: (i) donor DNA with the gene of interest adjacent to sequence for repair of a defective selectable marker, (ii) sequence encoding a rare-cutting endonuclease such as I-SceI, (iii) a target locus (TL) comprising the defective selectable marker and I-SceI cleavage site. Typically, this requires additional markers for the integration of the donor and target sequences, which may be assembled through cross-pollination of separate transformants. Inducible expression of I-SceI then cleaves the TL and facilitates homologous recombination, which is assayed by selection for the repaired marker. We used bar and gfp markers to identify assembled transformants, a dexamethasone-inducible I-SceI::GR protein, and selection for recombination events that restored an intact nptII. Applying this strategy to callus permitted the selection of recombination into the TL at a frequency of 0.085% per extracted immature embryo (29% of recombinants). Our results also indicate that excision of the donor locus (DL) through the use of flanking I-SceI cleavage sites may be unnecessary, and a source of unwanted repair events at the DL. The system allows production, from each assembled transformant, of many cells that subsequently can be treated to induce gene targeting. This may facilitate gene targeting in plant species for which transformation efficiencies are otherwise limiting.

Introduction

Transgenesis offers the possibility to insert a known DNA sequence into the genome of an organism to introduce new heritable characters. It is commonly used in research to investigate gene function and in biotechnology to improve agronomic traits. However, random insertion of the transgene into the genome can result in mutations caused by the insertion into an endogenous gene (Krysan et al., 1999), potential production of unintended peptides or variable expression due to the genomic environment of the transgene (Matzke and Matzke, 1998). Thus, there are currently considerable efforts worldwide to develop efficient technologies for gene targeting (GT) to produce genetically modified (GM) crops with transgenes located at predetermined positions in the plant genome. Exploiting the cellular homologous recombination (HR) machinery, GT allows the exchange of genetic information between homologous DNA sequences and can be used to precisely modify the genome. The integration of transgenes flanked by sequences homologous to the desired genomic insertion site permits efficient and routine gene targeting in prokaryotes and fungi, but GT is very inefficient in higher plants with frequencies of the order of 10−4 per transformant (Cotsaftis and Guiderdoni, 2005; Hanin et al., 2001; Paszkowski et al., 1988).

HR and non-homologous end-joining (NHEJ) are triggered to repair double-strand breaks (DSBs) of DNA. These lesions are formed accidentally by genotoxic stresses (Hanin and Paszkowski, 2003; Khanna et al., 2001; Tuteja et al., 2009) or in a programmed manner, for example during meiosis by the Spo11 complex (Grelon et al., 2001). Repair through NHEJ links the two ends of the DSB and is frequently accompanied by the creation of mutations at the site of the repair. HR copies an endogenous (different allele or stably inserted transgene) or exogenous (non inserted transgene) sequence template with homology on either side of the break and allows a precise modification of the genome (Puchta et al., 1996). Transgene integration is believed to generally involve insertion via NHEJ into a pre-existing DSB (Tzfira et al., 2004) occurring randomly in the plant genome. A DSB at a precise genomic location presenting homologous sequence to the transgene significantly increases the recombination rate at this site (Puchta et al., 1993; Szostak et al., 1983). This has led to the development of tools such as meganucleases, zinc-finger nucleases and transcription activator-like effector nucleases for gene targeting (Christian et al., 2010; Shukla et al., 2009; Tzfira et al., 2012). These endonucleases create a DSB at the target locus (TL) and have been used to modify the TL by mutation using NHEJ (De Pater et al., 2009; Yang et al., 2009) or by precise sequence modification using HR (Tzfira and White, 2005). For example, the mitochondrial I-SceI meganuclease from Saccharomyces cerevisiae (Jacquier and Dujon, 1985) has been successfully used in plants to perform GT (D'Halluin et al., 2008; Puchta et al., 1996). In tobacco, cleavage of the TL containing an I-SceI restriction site by I-SceI increases recombination between the TL and the transforming T-DNA around 100-fold (Puchta et al., 1996). The enzyme required to produce the DSB can be introduced into the organism or cell via stable or transient transformation. For example, I-SceI has been introduced into plants via Agrobacterium-mediated retransformation of a TL line or by crossing lines stably expressing I-SceI to a TL line. In the latter case, the use of an inducible I-SceI can allow the creation of the DSB at a predetermined moment. For example, application of the glucocorticoid, dexamethasone, induced the activity of an I-SceI protein fused to the rat glucocorticoid receptor (GR) domain in Arabidopsis (Wehrkamp-Richter et al., 2009). The GR domain sequesters the I-SceI::GR complex in the cytosol. The addition of dexamethasone allows the dissolution of the complex (Aoyama and Chua, 1997), liberating the I-SceI::GR protein which can move to the nucleus and produce a DSB at the TL. In plants, an inducible I-SceI was used to enhance intrachromosomal recombination (Wehrkamp-Richter et al., 2009) and to perform targeted mutagenesis (Yang et al., 2009).

Plant GT strategies are generally based on the positive selection for GT events which repair a defective selectable marker. A DSB is induced at the TL inducing HR between a defective TL selectable marker gene and the repair DNA. For example in Zea mays (maize), D'Halluin et al. (2008) re-transformed TL lines with a repair DNA and a construct encoding I-SceI, either delivering the DNA via particle bombardment or Agrobacterium. The frequency of GT versus random insertion, measured by the acquisition of resistance to the herbicide BASTA, was up to 30% via particle bombardment and 3.7% using Agrobacterium. Shukla et al. (2009) have also reported efficient GT in maize using zinc-finger nucleases. Although these studies show that GT is now possible in a major crop plant, there is still the need to optimize GT to minimize the effort required to produce and identify GT events before GT becomes a routine tool for GM production. A major limiting factor is the need to deliver the repair DNA and nuclease-encoding sequence efficiently into a large number of cells, which in the case of maize transformation can involve the transformation of many thousands of immature embryos or calli to obtain a few GT events. An attractive alternative is to create a few transformation events where the repair DNA and I-SceI-encoding sequence are stably integrated into the genome. The repair DNA is then controllably excised from the genome and acts as a template for GT at the TL. This system has the advantage that every cell contains the repair template, and thus, a single transformed individual can yield a potentially unlimited population of cells for GT. Such a GT strategy has been successfully implemented in Drosophila, with the repair DNA being excised from the genome using the FLP recombinase and then linearized using I-SceI (Huang et al., 2008; Rong, 2002) and has recently been reported also in Arabidopsis (Fauser et al., 2012). The goal of the work presented here was to test a similar GT system in maize, using a dexamethasone-inducible I-SceI both to excise the repair DNA from the genome and to induce a DSB at the TL.

Results

The GT test system

Two plant transformation constructs, the TL construct and the donor locus (DL) construct, were developed to test the GT strategy. The T-DNA of the TL construct contains the plant transformation selectable marker phosphinothricin acyl transferase (bar) gene followed by an I-SceI restriction site and the 3′ part of the neomycin phosphotransferase II (nptII) gene (Figure 1b). The T-DNA of the DL construct contains a dexamethasone-activatable, maize codon-optimized I-SceI (I-SceI::GR) gene and an nptII repair region bordered by two I-SceI restriction sites (Figure 1a). The nptII repair region contains the bar gene, the green fluorescent protein (gfp) gene and a 5′ part of the nptII gene. The gfp gene here serves as a mock gene of interest to be inserted at the TL and additionally allows easy identification of DL-containing plants. The nptII repair region and the TL share common sequences of 2992 bp in the bar region and 1200 bp in the nptII region, provided largely by the insertion of a rice tubulin gene intron (intTubl) into the defective nptII genes. This homology should allow homologous recombination between these two sequences and the consequent repair of the nptII gene, resulting in kanamycin resistance (Figure 1c). The TL and the DL constructs were independently transformed into maize to generate TL and DL lines, respectively. Two intact TL (TL1 and TL2) lines and one DL line, each containing a single copy of the transgene, were selected by Southern blotting analysis (not shown) and their genomic flanking sequences isolated (Figure S1). The DL line expressed both gfp and the I-SceI::GR transcript. The two TL lines were then selfed in order to isolate homozygous descendants which were then crossed with the DL line (Figure 2a). The F1 progenies and their descendents were selfed (Figure 2b). The segregation of the TL and the DL indicates that the two constructions were not genetically linked.

Figure 1.

Maps of the transgenic loci (not to scale). Structures of the donor locus (a), the target locus (b) and the expected gene targeting event (c) with the position of the I-SceI cutting sites (black stars). For Southern blot analyses, the sizes of DNA fragments hybridized with intTubl (Tubl), intFad2 (Fad2) and terSac66 (Sac66) are shown for digestion with SacI (grey arrows) or NcoI (black dashed arrows). ‘+’ indicates that the size of the fragment will be greater than indicated as it will extend to the first relevant restriction site in the flanking genomic DNA. The left border of the T-DNA is adjacent to the bar gene. Fragment lengths are given as bp.

Figure 2.

Genealogy of maize Lines. The two homozygous target locus (TL) lines were crossed with a heterozygous donor locus (DL) line to obtain the F1 (a). The F1 progeny containing the DL and the TL were self-fertilized to obtain F2 descendants (b). ‘−’ represents the original wild-type loci. Only the genotypes of interest (containing TL and DL) are shown.

Detection of somatic repair of nptII in TL/DL leaves

For each TL line crossed with the DL line, the F1 progeny containing the TL and the DL were identified by PCR analysis and separated in two groups, seed of one group was treated with dexamethasone to induce I-SceI activity and the other not (see 'Experimental procedures'). To detect excision of the repair DNA from the DL, PCR was performed using primers positioned on either side of the DL I-SceI restriction sites. A total of 12 untreated and seven dexamethasone-treated plants were analysed and three excision events were detected for each condition, indicating a basal activity of I-SceI::GR and that dexamethasone treatment does not significantly induce I-SceI::GR in these conditions.

The analysed F1 plants were then selfed to identify kanamycin-resistant plants among the F2 descendants. To detect the presence of the TL and the DL, 176 F2 (42 for the TL1/DL line and 134 for the TL2/DL line) plants were analysed by PCR; 21 TL1/DL and 55 TL2/DL descendants contained both TL and DL. Kanamycin was applied to the apical meristematic region of wild-type (WT) and the F2 plants. On WT plants and F2 descendants containing only either the TL or the DL, this resulted in bleaching of the developing leaf (Figure 3b). However, leaves with green sectors within the kanamycin-bleached zones (Figure 3c) were observed in 60% of plants carrying both the TL and DL, corresponding to 38% of TL1/DL and 70% of TL2/DL plants (Figure 3d). PCR analysis performed on the DNA extracted from the green sectors permitted the amplification and sequencing of the repaired nptII gene, and this was not so for DNA extracted from bleached or untreated (Figure 3a) leaf sectors. Other batches of F2 seeds were sown, and none of the additional 504 descendants were fully kanamycin-resistant; however, green kanamycin-resistant sectors were again observed in TL/DL plants.

Figure 3.

Frequency of green sectors in kanamycin-treated plants. Leaves of F2 maize plants untreated with kanamycin (a), sensitive to kanamycin (b) and sensitive to kanamycin with kanamycin-resistant green sectors (c). The histogram (d) shows the percentage of sensitive plants with green kanamycin-resistant sectors for the different genotypes (n is the number of plants analysed). ‘Both’ indicates plants possessing a target locus (TL) and a donor locus (DL); ‘one of’ indicates plants possessing either only a TL or a DL.

Recovery of fully kanamycin-resistant plants via in vitro tissue culture

Notwithstanding the presence of kanamycin-resistant leaf sectors (and thus GT), no fully kanamycin-resistant progeny were identified in 680 F2 plantlets. We thus used a tissue culture approach. Plant regeneration from maize leaves has not been reported, but calli derived from immature maize embryos are routinely used to regenerate plants (Lu et al., 1983). Embryos isolated from immature kernels of selfed F2 plants containing the TL and the DL were placed on callus induction medium with dexamethasone at 0 μm (control), 30 or 50 μm (Figure 4). From 2356 extracted embryos (619 from the TL1/DL and 1737 from the TL2/DL plants), seven independent kanamycin-resistant GT events (Table 1) were recovered and shown to carry a repaired nptII gene which was amplified by PCR and sequenced. Two were obtained from the TL1/DL embryos (GT1 and GT2) and five from the TL2/DL embryos (GT3, GT4, GT5, GT6 and GT7). GT efficiencies calculated as the number of GT events per immature embryo range from 0.13% to 0.55% (Table 1).

Figure 4.

Strategy for in vitro culture and regeneration. Immature embryos were collected and separated into three groups, non-treated, treated with 30 μm and treated with 50 μm of dexamethasone. Kanamycin selection allowed the regeneration of kanamycin-resistant plants containing a repaired nptII gene (green plantlets).

Table 1. Summary of gene targeting experiments
LinesExtracted embryosDexamethasone concentration (μm)Number of kanamycin-resistant eventsName of kanamycin-resistant eventsGT frequency (kanamycin-resistant/total embryos)
  1. + indicates a sterile plant.

  2. GT, gene targeting; TL, target locus; DL, donor locus.

TL1 × DL18300/0%
255301GT10,39%
181501GT20,55%
TL2 × DL80001GT40,13%
563303GT3, GT5, GT60,53%
374501G7 (+)0,27%

Only one of the seven GT events was obtained from dexamethasone non-treated control embryos; thus, dexamethasone treatment appears to increase the number of GT events. The number of events are, however, low and as we observed somatic recombination during the development of F2 plants in the absence of dexamethasone treatment, some GT events probably come from a basal, leaky I-SceI::GR activity.

To clarify the question of the inducibility of I-SceI::GR activity, we tested the effect of dexamethasone treatment on TL DSB induction through the measurement of mutations in the TL I-SceI site. Embryos of F2 plants were extracted (14 from TL1/DL and 69 from TL2/DL lines) for somatic embryogenesis. A sample of the callus formed from each embryo was analysed by PCR to determine the presence of the DL and the TL. Each callus was divided into three parts, one part placed for 1 week on medium without dexamethasone, one on medium with 30 μm and one with 50 μm dexamethasone. Samples of TL/DL calli from each treatment and of the TL calli were pooled separately for DNA extraction (Table 2). A 400-bp region around the I-SceI TL restriction site was amplified from each pool and sequenced by 454 sequencing (Genome Sequencer FLX by Roche). Approximately 15 000 sequences were obtained and analysed to estimate the mutation rate at the I-SceI restriction site due to NHEJ repair. No mutations were detected in the absence of the DL (TL controls that do not carry the I-SceI gene). In the TL/DL lines, mutations of the TL I-SceI site were detected, with the number of independent mutations increasing 3.5 to 5-fold with 30 μm and 5 to 6-fold with 50 μm of dexamethasone. These data thus confirm a basal activity of I-SceI::GR in inducing mutations in the I-SceI target site and that dexamethasone treatment increases I-SceI::GR activity in calli (Table 2). The presence of mutations in the target in the absence of dexamethasone, however, confirms the leakiness in this system.

Table 2. Quantification of mutations at the I-SceI site of target locus (TL)
LinesNumber of EmbryosGenotypeNumber of readsDexamethasone concentration (μm)Mutations in TL I-SceI site
TL1 × DL14 TL1 15370/0
TL1/DL (I-SceI::GR)11575015
176853076
232795076
TL2 × DL69 TL2 15597/0
TL2/DL (I-SceI::GR)1629306
195833015
179315036

Analysis of GT obtained from TL1/DL plants

Two GT events were identified from calli from TL1/DL plants. These were regenerated to give plants GT1 and GT2. Southern blot analysis was carried out on SacI-digested genomic DNA of these plants, the parent line (TL1/DL) and control lines carrying only the target locus (TL1) or the DL. Three different probes were used: Arabidopsis AtFad2 gene intron (intFad2, present in the DL and predicted to be present in a GT locus), intTubl (common to the DL, TL and predicted to be in the GT locus) and Arabidopsis AtSac66 terminator (terSac66, present in the TL and predicted to be in the GT locus). The Southern blot results with the intTubl probe are shown in Figure 5b. A band of 3.7 kbp was detected in the DL lane and a 4.1-kbp band in the TL1 lane; both bands were observed in the DL/TL1 control lane. As expected, in GT1 and GT2 lanes, the TL1 band disappeared and a new 5.5-kbp band, also observed with intFad2 and terSac66 probes (Figure S2), was detected, confirming nptII repair at the TL. The non-excised DL band was observed at 3.7 kbp with an intensity consistent with a homozygous state.

Figure 5.

Southern blot analysis of gene targeting (GT)1 and GT2 events. (a) Schema of the GT event occurring in the GT1 and GT2 events. The black stars show I-SceI restriction sites, and cleavage of the site in target locus (TL)1 is indicated by the separation of the two halves of the locus. Grey arrows indicate SacI restriction sites and the size of DNA fragments detected by the intTubl probe. (b) Southern blot analysis of control parental plants (TL1, donor locus (DL), TL1/DL plants) and of recombinant GT plants from TL1/DL tissue culture (GT1 and GT2 lines) with SacI-digested DNA and the intTubl probe. A fragment around 3.7 kbp, indicative of the original DL, was detected in the GT1 and GT2 events. A 4.1-kbp fragment, indicative of TL1, was not observed in the GT1 and GT2 events. The expected fragment of 5.5 kbp, indicative of npII reconstitution due to HR between the DL and TL1, was detected in the GT1 and GT2 events. The DL band is more intense than the TL1 band for the lanes GT1 and GT2, indicating the homozygous state of the DL and the non-excision of the DL. Fragment lengths are given as kbp.

The GT1 and GT2 plants were backcrossed twice to wild-type plants, and kanamycin resistance was inherited as a single Mendelian locus. In the first backcross, 53% and 55% of GT1 and GT2 descendants were kanamycin resistant and all presented the non-excised DL, confirming that GT1 and GT2 are heterozygous for the reconstructed (by GT) nptII gene at the TL. For the second backcross, 35% and 36% of GT1 and GT2 descendants were kanamycin resistant. All resistant plants expressed gfp, and PCR amplification confirmed both the presence of the intFad2 and terSac66 regions and the absence of the TL1-specific fragment containing the I-SceI TL1 site (Figure S3), which together confirm the expected reconstitution of nptII. Among the kanamycin-resistant descendants, 61% of GT1 and 55% of GT2 also contained all sequences specific to the DL [left border (LB) I-SceI site, right border (RB) I-SceI site and I-SceI::GR].

Finally, to confirm that the modified TL in plants GT1 and GT2 are the result of homologous recombination on both sides of the break in the TL with the donor, we also sequenced the junction fragments amplified by PCR with primers to the terSac66 and the genomic flanking sequence of the TL1 LB. PCR fragments were amplified from kanamycin-resistant GT1 and GT2 plants containing only the GT locus and sequenced. The sequence of the amplified fragment obtained (Figure S4) was identical to that predicted for HR between the DL and TL1 resulting in the repair of nptII at the TL1.

Analyses of the GT1 and GT2 events thus showed that they are true GT events at the TL and that the GT was not associated with excision of the donor sequence from the DL in either case, suggesting that they arose through ectopic recombination (Puchta, 1999) between the TL and DL (Figure 5a).

Analysis of GT from TL2/DL plants

Five GT events were identified from calli from TL2/DL plants. These were regenerated to give plants GT3, GT4, GT5, GT6 and GT7. Analysis of these plants revealed a second class of GT events involving reconstitution of nptII at the DL, rather than at the TL (Figure 6a).

Figure 6.

Southern blot results of gene targeting (GT)3-7 events. (a) Schema of the GT events occurring in GT3-7. The black stars show I-SceI restriction sites, and cleavage of the site in donor locus (DL) is indicated by the separation of the two halves of the locus. The grey arrows and black dashed arrows indicate respectively the SacI and NcoI restriction site positions and the size of DNA fragments detected by the probes used. (b) Southern blot analyses of control parental plants [target locus (TL)2, DL, TL2/DL] and of kanamycin-resistant events derived from the TL2/DL (GT3-7) with SacI-digested genomic DNA and the intTubl probe. A 3.7-kbp band, indicative of the non-excised DL, was detected in the GT4-7 events, but not in the GT3 event. The 5.7-kbp fragment indicative of native TL2 was observed in the GT3–7 events. All putative GT events show an additional band around 5.5 kbp for GT3 and around 4.6 kbp for GT4–7 events (but not at the expected size of 7.1 kbp). (c) Southern blot analysis of the same events with NcoI-digested DNA blotted with a terSac66-specific probe. The signal at 2.0 kbp was observed for the TL2 and DL/TL2 controls. For the GT3, GT5 and GT6 events, unexpected bands appeared at respectively 2.5, 2.4 and 2.1 kbp. Fragment lengths are given as kbp.

Southern blot analysis was carried out on SacI-digested genomic DNA of these plants, the parent line (TL2/DL) and control lines carrying only the target locus (TL2) or the DL, hybridized with intFad2, intTubl and terSac66 probes. The results with the intTubl probe are presented in Figure 6b. The original DL and TL2 bands of 3.7 and 5.7 kbp respectively were detected in the GT samples, except for GT3 that lacked the original DL. However, the predicted GT-specific band of 7.1 kbp for GT at the TL (4.5-kbp plus 2.6-kbp TL2 flanking sequence) was not detected in the GT lanes. Instead, a band was observed at 5.5 kbp for GT3 and around 4.6 kbp for GT4, GT5, GT6 and GT7, indicating a different mechanism of nptII repair. This band was also observed with the intFad2 and terSac66 probes (Figure S2). The SacI digestion results were confirmed by Southern blotting of NcoI-digested DNA hybridized with the terSac66 probe (Figure 6c). The TL band of 2.0 kbp was found unchanged in all GT lanes, and an additional band was observed in GT3, GT5 and GT6 lanes. The presence of the terSac66 on two different DNA fragments indicates that either one copy of a potentially homozygous TL2 was modified, but not via the expected double crossover, or that the TL2 was used as template by HR to repair an I-SceI::GR-induced DSB in the DL (Figure 6a). The fact that PCR of the GT3 line could not detect a DL lacking the nptII repair fragment (data not shown) and that GT3 lacks a band specific to the original DL supports the idea that in GT3 at least, the DL has been modified.

To resolve this question, the GT events were backcrossed twice to the wild type and analysed by PCR (Figure S3). Kanamycin resistance was inherited as a single locus and was not correlated with the presence of the LB TL amplicon, which is specific to the TL and predicted to be present in a true GT event at the TL. Amplification of the LB TL and TL I-SceI amplicons in 46% of the GT4, 44% of the GT5 plants and about 17% of the GT6 events can thus be attributed to the presence of a segregating unmodified TL in these plants. Kanamycin resistance was strictly correlated with the presence of DL sequences on either side of the I-SceI restriction site next to the defective nptII in the DL. However, a PCR fragment of the expected size across this I-SceI site could not be amplified. This suggests either deletion around this I-SceI site or the insertion of a sequence including the TL terSac66 into this I-SceI site. This latter hypothesis was confirmed by amplification and sequencing of the GT loci using primers located in the intFad2 and in the I-SceI::GR gene. Analysis of the amplified sequence (Figure S4) showed an HR event restoring the nptII gene on the one side and a NHEJ or microhomology-mediated end-joining (MMEJ) event copying and linking a part of the TL2 flanking sequence to the I-SceI::GR promoter on the other side. For the GT3 event, after the region of homology in the nptII gene, 909 bp of the TL2 corresponding to the missing part of the defective nptII (including the terSac66) and 502 bp of the genomic flanking sequence of the TL2 RB were linked by non-homologous recombination to the other side of the break, which had lost 52 bp of DL sequence. For GT5, the event is similar to the GT3 event, but only 82 bp of the TL2 flanking sequence was copied and 9 bp of the break was deleted to repair this side by NHEJ including 115 bp of mitochondrial DNA in the junction. In the GT4 event, a microhomology of 4 bp is present at the junction, and 882 bp of the TL2 comprising the missing part of nptII (including the terSac66) was copied into the repair sequence with 55 bp deleted from the DL. The lengths of these sequences of these GT events correspond to the observed sizes of bands seen on the Southern blots.

Analyses of the GT3-7 events thus showed that they are GT events, but involve the modification of the DL using the TL as template. Cleavage of the I-SceI site of the DL adjacent to the nptII sequence, followed by recombination of the nptII side of the break with the homologous TL2 as donor, creates a functional nptII at the DL. The other side of the break in the DL does not carry homology to the TL2, and thus must be repaired by NHEJ or MMEJ, resulting in variable lengths of TL2 sequence integrated into the DL (Figure 6a).

Discussion

The goal of this study was to develop a tool for precise remobilization of a transgene randomly inserted into the maize genome by its excision and insertion into a defined genomic site using homologous recombination. This strategy was tested by crossing of stably transformed TL and DL maize lines containing 3′ and 5′ overlapping regions of an nptII gene, respectively. Induction of I-SceI activity in these lines with dexamethasone was used both to create a DSB at the TL and also to release the nptII repair DNA from the DL. HR of the liberated nptII repair DNA with the TL would then reconstitute the nptII gene and also mobilize a gfp gene into the TL. Kanamycin selection allows the selection of putative GT events.

Testing of 680 F2 progeny carrying the TL and DL did not permit the identification of any kanamycin-resistant plants, suggesting that germinal or early meristematic GT events are very rare under the conditions tested. However, in the course of testing these plants for kanamycin resistance, we noted the presence of green kanamycin-resistant sectors on the kanamycin-bleached leaves, suggesting the presence of somatic HR events between the TL and DL. DNA extracted from these green sectors, but not bleached leaf regions, could be used to amplify a restored functional nptII gene. Such green kanamycin-resistant sectors on bleached plants have previously been described in tobacco plants carrying an intrachromosomal HR reporter based on nptII reconstitution (Peterhans et al., 1990) and also in Arabidopsis (Assaad and Signer, 1992). Other studies of GT based on nptII restoration and selection of resistant plants through the addition of kanamycin to the culture medium in tobacco (Puchta, 1999) and Arabidopsis (Vergunst et al., 1998) did not, however, report green kanamycin-resistant sectors. In maize, we show here that application of kanamycin to the apex permits the detection and quantification of somatic GT events in leaves without affecting the survival of the sensitive plants. Multiple kanamycin treatments are possible and progeny can be obtained from treated plants. This assay, which should be applicable to other plant species, is currently being used to test and optimize GT frequencies.

In tobacco lines containing the equivalent of our TL, retransformed with a repair sequence and constitutive I-SceI, the observed GT frequency increased proportionally with the expression level of I-SceI (Puchta et al., 1996). Similarly, in our maize plants, the frequency of green kanamycin-resistant sectors gives direct information about I-SceI::GR activity. Given that we observed nptII repair sequence excision from DL in equivalent proportions from maize plants grown in the absence or the presence of dexamethasone treatment, there is clearly basal activity of I-SceI::GR in the maize leaves and dexamethasone does not further induce I-SceI::GR in the tested conditions. In our previous study with I-SceI::GR in Arabidopsis, basal activity was found, but the expression could be induced around 25- to 200-fold when dexamethasone was supplied in the growth medium (Wehrkamp-Richter et al., 2009). We speculate that the dexamethasone applied to maize germinating seed does not penetrate into the seed in sufficient quantities to further induce I-SceI::GR activity. Dexamethasone treatment does, however, induce I-SceI::GR activity when added to the callus growth medium, where a 3.5 to 6.0-fold increase in the number of mutations at the TL was observed with dexamethasone (Table 2).

Notwithstanding the GT observed in somatic tissues, no kanamycin-resistant plants were found in the 680 tested F2 progeny of the TL/DL lines. We thus tested a strategy based on tissue culture selection and regeneration of kanamycin-resistant plants from TL/DL calli. This approach permitted the selection of seven independent GT events in two separate experiments involving a total of 2356 embryos (Table 1). Two of these, GT1 and GT2, were generated from embryos from TL1/DL plants; molecular and genetic analyses confirmed that they are true GT in which the TL has been modified by ectopic recombination using the DL as template on both sides. The overall frequency of obtaining true GT events at the TL from the two experiments was therefore 0.085% (29% of recombinants). The remaining five events (GT3-7) were generated from TL2/DL line embryos, and Southern blot and sequence analyses showed that they result from the modification of the DL, using the TL as template. The mechanism appears to be the creation of a DSB by I-SceI::GR in the DL I-SceI site next to the 5′ nptII region. Recombination of the nptII side of the break with the homologous TL2 region as donor creates a functional nptII at the DL. However, the other side of the break does not carry homology to the TL2, and thus must be repaired by NHEJ or MMEJ (GT4), resulting in variable lengths of TL2 sequence integrated into the DL. Such HR + NHEJ gene conversion events have been previously reported in plants (Puchta, 1999).

This surprising difference in the nature of the GT events identified in the calli from the two parent lines led us to resequence the TL and DL of these lines. This analysis identified a mutation which eliminates the right side I-SceI cut site of the DL in the F1 TL1/DL plant (between nptII and I-SceI::GR – see Figure 7a). In the TL1/DL calli therefore, and in contrast to the TL2/DL calli, I-SceI::GR can only cleave the DL once (to the left of the bar marker). Although the numbers of GT events analysed are low, this difference very probably explains the different types of GT events identified in calli from the two lines. In the TL1/DL calli, recombination initiated by I-SceI cleavage of the DL would not generate a functional nptII gene and so only events initiated by cleavage in the TL would be selected. In the TL2/DL calli however, recombination initiation through cleavage adjacent to the nptII sequences in either the TL or the DL would result in the reconstruction of nptII (Figure 7b). In the TL2/DL calli, identification of recombination events in which only the DL was recipient clearly shows that single, incomplete I-SceI cleavage of the DL is frequent in these cells.

Figure 7.

Model. (a) For the target locus (TL)1/donor locus (DL)-derived embryos, I-SceI cleavage in the TL initiated recombination with the DL, using homology in both bar and nptII regions. This resulted in gene targeting (GT) at the TL, using the DL as template. (b) For the TL2/DL-derived embryos, I-SceI cleavage in the DL (right I-SceI site in the DL between the nptII and I-SceI::GR cassette) initiated recombination with the TL, using homology in only the npII region. The other side of the break was repaired by non-homologous recombination. This resulted in GT at the DL using the TL as template. The I-SceI sites are shown as black stars. The rightmost I-SceI in the DL is mutated in the TL1/DL line, as indicated by the absence of the corresponding star.

These data thus show that only cleavage of the TL is needed for successful GT in these plants, as well as providing a clear illustration of the risk of including multiple I-SceI restriction sites in plants in which I-SceI expression or activity is limiting. The basal level of I-SceI cleavage in the absence of dexamethasone induction further compounds this risk, through increasing levels of mutation in the I-SceI sites of the DL. The dependence of this problem on limited I-SceI activity would thus explain the difference with the recent study in Arabidopsis using a comparable strategy, in which only clean GT events were found (Fauser et al., 2012). They observed efficient repair DNA excision, probably due to efficient activity of the I-SceI, and GT was observed in up to 1% of the progeny. Limiting endonuclease activity is, however, a common problem in experiments of this type (Puchta et al., 1996). In Drosophila, Gong et al., (Gong and Golic, 2003) also reported low I-SceI-mediated repair fragment excision and estimated that excision occurred in 7% of cells. They thus used the FLP recombinase to excise a circular repair DNA from the genome that was subsequently linearized by I-SceI. This system gave good GT rates in Drosophila but has not thus far been shown to work in maize (Yang et al., 2009).

The goal of this work was to test in maize a GT system based on I-SceI-mediated cleavage of the target, and excision of the nptII repair region from the genome. In previously described GT systems in maize, GT can occur only within a limited period after transformation of donor sequences and site-specific nucleases (D'Halluin et al., 2008; Shukla et al., 2009). In contrast, in our GT system, transformation of components required for GT is uncoupled from recombination and GT. This allows multiplication of the cells carrying the TL, the DL and I-SceI-encoding sequences and permits the selection of rare ectopic recombination events. With this approach, a few stable transformation events can be used to generate a large population of cells from which to select GT events, of particular interest in cases where plant transformation frequencies are a limiting factor. In the work reported here, GT efficiencies range from 0.13% to 0.55% GT events per immature embryo. The actual number of cells screened is of course much higher than the number of calli, but the calculation with respect to calli expresses best the employed human effort. Our results furthermore show that cleavage of the DL is both unnecessary for targeted recombination, and a source of unwanted events when endonuclease cleavage is limiting.

Experimental procedures

Production of GT constructs and maize transgenic lines

The binary vectors for the creation of the target locus, pBIOS-TL, and donor locus, pBIOS-DL, were constructed in the following manner. First in order to extend the region of homology between the truncated nptII genes in the TL and DL lines, an 886-bp rice tubulin intron (GenBank, AJ488063) was introduced into the coding sequence of the nptII gene at position 204 bp downstream of the ATG. A 5′ truncated nptII-intTubl fragment lacking the first 150 bp of the nptII coding region was cloned between an I-SceI site and in front of the Arabidopsis Sac66 polyadenylation sequence (GenBank, AJ002532). The I-SceI-3′nptII-intTubl-terSac66 fragment was then cloned into an SB11-based plant binary vector (Komari et al., 1996) containing a rice actin promoter (pAct) (McElroy et al., 1991) linked to the bar selectable marker gene (White et al., 1990) and a nopaline synthase terminator (terNos), forming pBIOS-TL. To produce pBIOS-DL, a 3′ truncated nptII-intTubl fragment lacking the last 227 bp of the nptII coding region was cloned behind the constitutive SC4 promoter (pSC4) (Schünmann et al., 2003). A pSB11-based binary vector was created that contained the pAct-bar-terNos gene cassette and a cassava vein mosaic virus (CsVMV) promoter (Verdaguer et al., 1998) linked to gfp, with both gene cassettes flanked by I-SceI restriction sites. The pSC4-5′nptII-intTubl fragment was then cloned between the terminator of the gfp gene and the 3′ I-SceI site to complete the nptII repair region. Next, the NLS::I-SceI::GR gene (Wehrkamp-Richter et al., 2009), codon optimized for maize expression, was cloned between a CsVMV promoter and 35S cauliflower mosaic virus terminator. This cassette was then cloned between the nptII repair region 3′ I-SceI site and the RB to form pBIOS-DL. Agrobacterium tumefaciens strain LBA 4404 (pSB1) (Hoekema et al., 1983) was transformed with pBIOS-TL and pBIOS-DL. For each construction, a clone containing the recombinant plasmid was selected. Embryos of maize inbred A188 were transformed with each construction and transformed plants were regenerated according to Ishida et al. (Ishida et al., 1996) using glufosinate selection.

Plant analysis

Genomic DNA was extracted from the leaves by using the DNeasy 96 plant kit (Qiagen, Valencia, CA). Genomic DNA (10 μg) was digested, separated on 1% agarose gel by electrophoresis, transferred to nylon membrane and hybridized to 32P-marked probes following standard procedures (Sambrook and Russell, 2006). The genomic sequences flanking the transgenes were amplified using an adapter-anchor PCR method according to the method of Balzergue et al. (Balzergue et al., 2001), with previously described modifications (Sallaud et al., 2003), using DNA digested with SspI or PvuII. Plant genotyping was performed by PCR. To amplify the fragments longer than 2.0 kbp, a Takara La Taq kit (Takara Bio, Shiga, Japan) was used. GFP fluorescence of sampled plant leaves was visualized under a fluorescence stereomicroscope (Leica MZ16F) using a GFP2 (Leica, Bannockburn, IL) filter.

Crossing, culture and treatment of transformed plants

Plants were grown in the glasshouse with a 16-h day at 26 °C, 400 μE/m2/s and an 8-h night at 18 °C. Dexamethasone treatments on seed were performed by immersion of the seed during 2 days in an aqueous solution of 30 μm dexamethasone. Kanamycin treatments were performed by the application of 50 μL of a solution at 200 mg/L kanamycin and 1% (v/v) Tween-20 on the apical region of 2-week-old plants.

Somatic embryogenesis

Embryos isolated from selfed plants containing the TL and DL were placed onto medium according to Ishida et al. (Ishida et al., 1996), lacking the kanamycin selective agent. For the first experiment, LS-AS medium was complemented by 0, 30 or 50 μm dexamethasone, and after 1 week, plantlets were transferred sequentially to LSD1.5, LSZ and 1/2LSF media lacking dexamethasone and containing 50 mg/L of kanamycin. For the second experiment, callus was developed for 3 days on LS-AS, 1 week on LSD1.5 and 3 weeks on LSZ medium. Then, callus was cultivated 1 week on LSZ medium containing 0, 30 or 50 μm of dexamethasone. GFP fluorescence of calli was visualized under the fluorescence stereomicroscope (Leica MZ16F) with a GFP2 filter.

Callus analysis

PCR was performed directly on callus tissues using Terra direct PCR polymerase (Clontech Inc., Palo Alto, CA) in 20 μL with specific TL (forward: GTGGCGGACCGCTATCAG and reverse: ACATGTATTAAGAAGCAATGCATGTAGTAC) and DL (forward: TGGCAATCCCTTTCACAACC and reverse: CCCAGTCATAGCCGAATAGCC) primers. Genomic DNA was extracted from pooled calli, of the same genotype and dexamethasone treatment, with the DNeasy 96 plant kit (Qiagen). Primers were designed according to GS FLX Titanium emPCR LIBL kit (Roche Applied Science, Mannheim, Germany) with a specific TAG for each condition (forward: CCATCTCATCCCTGCGTGTCTCCGACTCAG-X-TCATCCCTACCCGTTCGTT and reverse: CCTATCCCCTGTGTGCCTTGGCAGTCTCAG-X-ATCACCCAGATCCACCCA, X represents the specific TAG of 10 bp for each condition). PCR was performed using Platinum Taq DNA Polymerase High Fidelity (Invitrogen, Carlsbad, CA). Emulsion PCR was realized on the obtained PCR products with the emPCR Emulsion kit (Roche). PCR products were sequenced with the emPCR sequencing kit (Roche) with Genome Sequencer FLX (Roche). Sequences from each condition were independently assembled and aligned to the reference sequence containing the non-mutated I-SceI site. Sequence differences of 1 or 2 bp with the reference sequence found in the TL genotype, lacking I-SceI::GR, were discarded as these are likely to be sequencing errors. Sequence differences of three or more base pairs encompassing the I-SceI site were identified and manually regrouped per condition to identify the number of independent mutations per condition.

Acknowledgements

This work was supported by the EU FP6 project TAGIP; by the EU FP7 project RECBREED; and by an ANRT CIFRE grant to AA. We thank Friedrich Fauser and Holger Puchta for sending us their manuscript prior to publication. The authors are grateful for the support of the Biogemma Cloning, Transformation, Analysis, Greenhouse and Upstream Genomic teams and to Dr. O. Da Ines and Dr. S. Amiard for reading of the manuscript.

Ancillary