• Open Access

Homologous recombination: a basis for targeted genome optimization in crop species such as maize

Authors


* Correspondence (fax +32 (0)9 383 67 31; e-mail: kathleen.d'halluin@bayercropscience.com)

Summary

In an attempt to understand the feasibility of future targeted genome optimization in agronomic crops, we tested the efficiency of homologous recombination-mediated sequence insertion upon induction of a targeted DNA double-strand break at the desired integration site in maize. By the development of an efficient tissue culture protocol, and with the use of an I-SceI gene optimized for expression in maize, large numbers of precisely engineered maize events were produced in which DNA integration occurred very accurately. In a subset of events examined in detail, no additional deletions and/or insertions of short filler DNA at the integration site were observed. In 30%–40% of the recovered events, no traces of random insertions were observed. This was true for DNA delivery by both Agrobacterium and particle bombardment. These data suggest that targeted double-strand break-induced homologous recombination is a superior method to generate specific desired changes in the maize genome, and suggest targeted genome optimization of agronomic crops to be feasible.

Introduction

The world-wide pressure to increase crop yields and the use of plants as factories for green products underline the growing importance of methods for the optimization of plant gene function. Although genomics is providing a rapidly increasing insight into pathways, regulatory networks, allelic variants, etc., the full realization of this knowledge awaits improved methods to obtain the desired allelic combinations in the desired genetic background.

Today, a broad repertoire of tools is available for the development and optimization of traits, including traditional mutation breeding, the use of advanced molecular breeding techniques and transgenesis. Furthermore, the analysis of wild (plant) species has revealed novel alleles as well as uncommon gene functions, which, together with in vivo and in vitro gene evolution approaches, provide a near-infinite source of biological diversity. Therefore, it is generally believed that, on arrival of efficient gene targeting technology, the targeted optimization of genes underlying traits is within reach.

Gene targeting by homologous recombination (HR) enables the precise integration of additional sequences at a specific genomic site, the replacement of an endogenous sequence by an exogenous introduced sequence and the correction of mutations. This recombination approach has several advantages over non-homology-based genome engineering. DNA integration by non-homologous recombination (NHR) is characterized by random integration, with no control over either the site of integration or copy number, increasing the risk of mutations by integration in host genes and of homology-dependent gene silencing by complex integration patterns (Matzke and Matzke, 1998; Kooter et al., 1999; Francis and Spiker, 2005).

The major drawback of HR is its inefficiency in most higher eukaryotes, including plants. The development of homology-based gene targeting technology for plants has been progressing step-wise over the past 10 years. Various attempts have been made to enhance gene targeting in plants such as the manipulation of the expression level of the genes involved in HR (Reiss et al., 2000) or chromatin remodelling (Shaked et al., 2005). Although some progress has been made, in general, gene targeting by these approaches has remained far from efficient. Recently, Terada et al. (2007) described a more efficient gene targeting procedure in rice based on the use of a strong positive/negative selection. Another way to increase the efficiency of gene targeting is by the targeted induction of DNA double-strand breaks (DSBs) via rare-cutting endonucleases, such as I-SceI. In tobacco, this technique has been shown to increase the gene targeting frequency by at least two orders of magnitude (Puchta et al., 1996). A similar approach using another existing rare-cutting endonuclease, I-CeuI, yielded a comparable increase in gene targeting frequency in tobacco (Chilton and Que, 2003).

The enhancement of gene targeting by the use of existing rare-cutting endonucleases is limited to loci with pre-engineered recognition sites. It does not allow gene targeting of just any locus in the genome. Directed genome optimization at predefined locations in the genome may, however, become a reality in the near future. Recent reports have shown that, upon the expression of zinc finger nucleases (ZFNs), designed to cleave a predefined sequence, specific modifications through HR can be induced in the mammalian genome (Bibikova et al., 2003; Porteus and Baltimore, 2003; Urnov et al., 2005). Also, with engineered meganucleases, targeted HR has been induced in yeast and mammalian cells (Nahon and Raveh, 1998; Epinat et al., 2003; Chames et al., 2005; Arnould et al., 2006). In plants, the feasibility of ZFN-mediated gene targeting, albeit at a pre-engineered site, has been demonstrated in Arabidopsis and tobacco (Lloyd et al., 2005; Wright et al., 2005).

As an initial step towards an understanding of the feasibility of targeted genome optimization in agronomically relevant crops, we have studied the frequency and accuracy with which a target sequence can be converted into the desired sequence in the maize genome. In this report, we make use of targeted sequence insertion (TSI) at a pre-engineered I-SceI site in the maize genome to study the efficiency of DSB-induced gene targeting. Using an I-SceI gene optimized for expression in maize, we were able to produce a large number of plants with the target sequence converted precisely into the desired sequence, and without additional rearrangements or additional integrations elsewhere in the genome.

Results

Experimental design to study gene targeting in maize

To study DSB-induced DNA integration through HR, an assay was developed based on the restoration of a promoterless, defective bar gene by DSB-induced, HR-mediated TSI of a cauliflower mosaic virus (CaMV) 35S promoter (Figure 1). The target sequence in pTTAM78 consists of a promoterless bar gene (encoding phosphinothricin acetyltransferase), linked to a functional neomycin phosphotransferase gene (neo) driven by a nopaline synthase (nos) promoter. An I-SceI recognition site was cloned immediately upstream of the promoterless bar gene, and represents the targeted sequence integration site. To prevent any transcriptional interference, a 3′g7 transcription terminator was placed in between the nos promoter of the neo gene expression cassette and the I-SceI site. The homologous DNA pTTA82, designated ‘repair DNA’, contained a 35S promoter in between sequences homologous to the target locus, being the 5′ part of the bar gene downstream and the 3′g7 terminator and the neo cassette upstream (Figure 1). In order to obtain perfect homology between repair DNA and target DNA, the 35S promoter was further flanked by half I-SceI sites. On targeted DSB induction after the co-delivery of the I-SceI endonuclease-encoding construct (pCV78) and repair DNA (pTTA82), HR-mediated DSB repair should result in the targeted integration of the 35S promoter, leading to the activation of the bar gene, which can be selected for by resistance to phosphinothricin (PPT). The bar gene in the repair DNA contained a 3′ deletion to ensure that random integration of the repair DNA would not result in PPT resistance (PPTR). Thus, this assay offers the possibility to select directly for HR-mediated TSI events on the basis of resistance to the herbicide PPT.

Figure 1.

Assay for double-strand break (DSB)-induced, homologous recombination (HR)-mediated targeted sequence insertion (TSI). pTTAM78 represents the target DNA; between the neo gene and the promoterless bar gene, an I-SceI site is cloned next to a 3′g7 terminator. pTTA82 is the repair DNA; the neo gene, 3′g7 terminator and 3′ truncated bar gene function as the homologous regions (yellow and green boxes); between the last two, a 35S promoter flanked by half I-SceI sites is cloned. After DSB-induced HR between the target and repair DNA, TSI events are expected to have the structure depicted in the bottom part of the figure. The lengths of the PstI and PstI/I-SceI fragments are indicated for pTTAM78 and TSI events, and can be used as diagnostic in Southern blot analysis to determine the correct integration. The primers MDB521 and IB106 were designed to amplify the recombined target for sequence analysis. P35S, cauliflower mosaic virus 35S promoter; bar, phosphinothricin acetyltransferase; neo, neomycin phosphotransferase; 3′nos, 3′ocs and 3′g7, 3′ untranslated regions of nopaline synthase gene, octopine synthase gene and Agrobacterium gene 7, respectively.

For the study of TSI in maize, we made use of a derivative of maize genotype He89, a synthetic genotype with improved maintenance of its embryogenic capacity (Morocz et al., 1990). The target DNA pTTAM78 was introduced into the maize He89 derivative by either treatment of protoplasts (Dudits et al., 2001) or Agrobacterium transformation of type I callus (D’Halluin, 1998).

As the efficiency of TSI may depend on the position of the target locus in the genome, five independent single-copy transgenic lines were chosen. Lines 14-1 and 1-20 were produced by Agrobacterium-mediated transformation, whereas lines 5, 24 and 79 were produced by polyethylene glycol (PEG)-mediated transformation. The presence and cleavability of the genomic I-SceI site in these lines was demonstrated by Southern blotting. In all lines, hybridization with a 3′nos probe produced a single 1735-bp fragment when digested with PstI only. This band shifted to a fragment of 890 bp with a PstI/I-SceI digest (Figures 1, 2b,d).

Figure 2.

Southern analysis of target lines and targeted sequence insertion (TSI) recombinants. Southern analysis of the target lines and events obtained through TSI of the 35S promoter on particle bombardment-mediated repair DNA delivery (a, b) and Agrobacterium-mediated repair DNA delivery (c, d). Each lane shows a PstI (–) and PstI/I-SceI (+) digest hybridized with either a 35S-specific probe (a, c) or 3′nos-specific probe (b, d). (a) Lanes 1–2, TSI events for target line T24; lane 3, target line T24; lanes 4–9, TSI events for target line 14-1; lane 10, target line 14-1. (b) Blot a after reprobing with a 3′nos-specific probe. (c) Lanes 1–2, TSI events obtained for target line T5; lane 3, target line T5; lanes 4–14, TSI events obtained for target line 1-20; lane 15, target line 1-20. (d) Blot c after reprobing with a 3′nos-specific probe. The fragments corresponding to the detected bands of 2530, 890 and 1735 bp are shown in Figure 1. The additional bands obtained after hybridization with the 3′nos probe are specific for the target lines transformed with the epsps chimeric gene construct containing a 3′nos terminator; these bands are indicated by EPSPS. Non-indicated bands refer to extra inserts of repair DNA at random positions in the genome.

Because a meaningful evaluation of homology-based TSI requires efficient cell and tissue culture procedures, we set up a system that allows the handling of large numbers of cells for the selection of TSI events. We established an efficient Agrobacterium- and particle bombardment-mediated transformation protocol for maize He89 derivative suspension cells, using a functional bar gene cassette and a filter-based cell transfer system for the selection of PPTR events. From a 50-mL suspension culture with a packed cell volume of about 10 mL, we routinely obtained between 500 and 2000 independent transformed calli by both DNA delivery methods. For the study of homology-based TSI in maize, we established suspension cultures for the independent target lines, and used these suspension cells as target cells for the delivery of repair DNA by both Agrobacterium and particle bombardment.

Restoration of PPT resistance by TSI

A prerequisite for studying TSI is the appropriate expression of the endonuclease for the induction of DSBs. The I-SceI sequence originally used was the universal code version, described by Perrin et al. (1993), equipped with an N-terminal nuclear localization signal (NLS) of the SV40 large T-antigen (Kalderon et al., 1984). On co-delivery of the repair DNA and this I-SceI gene to allow the transient expression of I-SceI, in target lines transformed with the target DNA pTTAM78, no PPT-resistant events could be recovered. This suggests that no DSBs were induced in maize. To evaluate whether this was caused by an expression problem at the level of I-SceI, we produced stable transformants with only the I-SceI expression cassette, and tested I-SceI expression in these transformants by RNA gel blot analysis. Full-length I-SceI transcripts were not observed (data not shown). Further analysis by reverse transcriptase-polymerase chain reaction (RT-PCR) showed the presence of truncated I-SceI transcripts (data not shown), which could possibly be explained by both cryptic intron splicing and premature 3′ end formation. We therefore decided to improve I-SceI expression in maize by optimizing the I-SceI coding sequence for maize, taking care to avoid possible premature polyadenylation signals, sequences involved in the regulation of gene expression by methylation, and sequences prone to secondary RNA structure formation that can affect pre-mRNA splicing. Using the maize optimized I-SceI gene (pCV78), we investigated the frequency of targeted integration. Repair DNA in the presence and absence of the synthetic I-SceI gene expression cassette was introduced in suspension cells of the different target lines by either particle bombardment or Agrobacterium-mediated delivery. PPTR events were obtained with both delivery methods, but only in the presence of the synthetic I-SceI endonuclease. Suspension cells were routinely pre-treated with the phenolic compound acetosyringone for Agrobacterium- and even particle bombardment-mediated DNA delivery, as, in pilot experiments, at least twice as many PPTR events were observed per bombarded plate when suspension cells were pre-incubated with acetosyringone relative to non-treated cells (data not shown).

The frequency of PPTR events on co-delivery of the repair DNA in the presence of the synthetic I-SceI endonuclease varied between the five target lines. On particle bombardment-mediated DNA delivery, PPTR events were obtained at frequencies of 7.6, 2.2 and 0.13 events per bombarded plate for target lines 1-20, 14-1 and 24, respectively (Table 1). We were able to estimate the HR vs. random or NHR integration ratio by comparing the mean number of PPTR events obtained per plate bombarded with repair DNA pTTA82 and the synthetic I-SceI endonuclease pCV78, with the mean number of transformants per plate bombarded with a functional bar gene expression cassette pRVA52. This showed that the HR vs. NHR ratios were about 30%, 18% and 1% for target lines 1-20, 14-1 and 24, respectively (Table 1). Agrobacterium-mediated repair DNA delivery resulted in a lower frequency of PPTR events relative to particle bombardment-mediated repair DNA delivery. In a similar approach to that used for bombardment-mediated repair DNA delivery, we estimated the HR vs. NHR ratios for Agrobacterium-mediated repair DNA delivery to be ~1%–3.7%, ~0.14% and ~1.25% for target lines 1-20, 5 and 79, respectively (Table 2).

Table 1.  Numbers of homologous recombination (HR)-mediated targeted sequence insertion (TSI) events upon particle bombardment-mediated repair DNA delivery in the presence of the synthetic I-SceI gene
Target lineTransformation pRVA52Targeted sequence integration pTTA82 + pCV78HR/NHR (%)
Number of bombarded platesNumber of PPTR eventsMean number of PPTR events/plateNumber of bombarded platesNumber of PPTR eventsMean number of PPTR events/plate
  1. NHR, non-homologous recombination; PPT, phosphinothricin.

1-2037525151157.630
14-133712.317 382.218
2434013.315  20.13 1
Table 2.  Numbers of homologous recombination (HR)-mediated targeted sequence insertion (TSI) events upon Agrobacterium-mediated repair DNA delivery in the presence of the synthetic I-SceI gene
Target lineTransformation LBA4404(pSB131)Targeted sequence integration LBA4404(pTCV87)* or LBA4404(pTCV83)HR/NHR (%)
Number of platingsNumber of PPTR events§Mean number of PPTR events/platingNumber of platingsNumber of PPTR eventsMean number of PPTR events/plating
  • NHR, non-homologous recombination; PPT, phosphinothricin.

  • *†Co-cultivation with LBA4404(pTCV87)* and LBA4404(pTCV83)†, respectively.

  • After co-cultivation, the maize cells were resuspended and plated over several filter papers at about the same density. The number of platings refers to the number of filter papers over which the co-cultivated cells were plated.

  • §

    The transformation values are approximate, as clusters of transformation events often occurred close to each other, complicating the exact counting of independent events.

1-2018~400~2240* 9*0.225*~1*
1-2018~200~1127110.41~3.7
512~600~5028* 2*0.07*~0.14*
7924~480~2024 60.25~1.25

These results show that the delivery of repair DNA in the presence of the maize optimized I-SceI gene allows for the selection of PPTR events for both Agrobacterium- and particle bombardment-mediated DNA delivery, albeit at different frequencies.

Characterization of selected targeted sequence integration events

To verify whether the PPT resistance of the selected TSI lines was the result of the targeted integration of the 35S promoter, Southern blot analysis was performed on a number of selected PPTR events. Hybridization of PstI-digested genomic DNA with a 3′nos probe showed that the 1735-bp fragment, originally present in the target lines, shifted to a 2530-bp fragment (Figures 1, 2b,d). This indicates that an insertion has occurred at the target locus. The additional fragments obtained after hybridization with the 3′nos probe were the result of the fact that, for the production of the target lines, a co-delivery was performed of the target locus construct pTTAM78 with an epsps chimeric gene construct containing a 3′nos terminator (these fragments are indicated by EPSPS in Figure 2b,d). This co-delivery of the target locus construct pTTAM78 with an epsps gene expression cassette was performed to allow for the selection of glyphosate resistance, as we were unable to select transformants using the neo gene of plasmid pTTAM78 as selectable marker gene, probably because of the low expression levels of the pnos-driven neo gene.

By hybridizing the PstI-digested DNA with a 35S probe, we showed that the 35S promoter from the repair DNA had inserted at the target locus, as the same band of 2530 bp was identified (Figure 2a,c). Hybridization with 3′nos and 35S on PstI/I-SceI-digested DNA again revealed this band of 2530 bp, implying the loss of the I-SceI site at the target locus as a consequence of cleavage of the I-SceI site by the synthetic I-SceI endonuclease and repair of the DSB by HR-mediated TSI of the 35S promoter (Figure 2b,d). Southern blot analysis of 43 events revealed that nearly all TSI events (95%; 41/43) were perfect TSI events based on the size of the hybridizing bands.

Furthermore, it was remarkable that the frequency of perfect TSI events, which do not contain extra integrations of repair DNA, was not lower with particle bombardment-mediated DNA delivery than with Agrobacterium-mediated delivery, as biolistics usually leads to far more multiple integrations. In eight of 20 (40%) and six of 21 (30%) TSI events obtained by particle bombardment- and Agrobacterium-mediated DNA delivery, respectively, no additional bands were observed after hybridization of PstI- or PstI/I-SceI-digested DNA with a 35S probe. This indicates that these TSI events do not contain extra copies of the repair DNA or the I-SceI endonuclease.

Sequence analysis of a 1900-bp fragment, PCR-amplified from the target locus of eight randomly chosen independent PPTR events, confirmed that, in all of these events, the integration of the 35S promoter occurred by HR and these events were indeed perfect up to the nucleotide (Figure 3).

Figure 3.

Sequence analysis. DNA sequence at target locus of targeted sequence insertion (TSI) events, illustrated by only one TSI event, as all eight analysed TSI events show exactly the same sequence. The 35S promoter sequence is highlighted in red, the I-SceI sequence in blue, the bar coding sequence in green and the 3′nos terminator in purple. Target DNA and repair DNA are illustrated at the top of the figure. Primers MDB521 and IB106, used to amplify the fragment for sequence analysis, are indicated in Figure 1. In all TSI events, the 35S promoter insert shows junctions as expected for faithful homologous recombination.

Plants were regenerated from these TSI events, and it was shown that the restored bar gene was stably transmitted to the next generation based on the absence of necrotic lesions on localized application of the herbicide solution BASTA® (data not shown).

From these results, it can be concluded that homology-based TSI at a pre-engineered I-SceI site in the maize genome occurs with high precision. This observation was made for both Agrobacterium- and particle bombardment-mediated DNA delivery. Together, these data show that HR-mediated repair of targeted induced DSBs occurs with high efficiency. This shows the feasibility of directed genome modification at pre-engineered DSB sites, and suggests the feasibility of future targeted genome optimization in agronomically relevant crops.

Discussion

In this paper, we have described the feasibility of targeted genome modification in maize. Upon DSB induction at a pre-engineered I-SceI site, DNA integration through HR occurred with great accuracy. Our observations, together with the exciting developments in the fields of ZFNs and meganucleases with tailor-made specificities, suggest that gene targeting may be feasible at any desired genomic position. This opens up new perspectives for future genome optimization in agronomically relevant crops.

As a result of the low frequency of HR, targeted genome optimization requires, in the first instance, robust cell and tissue culture procedures, allowing the efficient manipulation of large numbers of cells, as described by Terada et al. (2002). In this study, we first developed an efficient transformation protocol using suspension cells, and a filter-based cell transfer system for the selection of transformants. In previous work, we have shown that stable transformation of maize type I callus by Agrobacterium can be improved by pre-treatment of the callus with acetosyringone, or a related plant phenolic compound, for approximately 5 days prior to co-cultivation with Agrobacterium (D’Halluin, 1998). Moreover, in this study of homology-based TSI at a pre-engineered I-SceI site, we pre-treated the suspension cells with acetosyringone for both Agrobacterium- and particle bombardment-mediated DNA delivery. We routinely used acetosyringone pre-treated cells for particle bombardment experiments, as an initial particle bombardment experiment with cells pre-treated or not with acetosyringone yielded more TSI events with pre-treated cells. One of the reasons for the beneficial effect of pre-incubation with acetosyringone may be the induction of cell division, as described by Guivarc’h et al. (1993). Of course, the stimulating effect of acetosyringone may also be the result of enhanced competence for homology-based TSI in a manner independent of cell division, e.g. through the stimulation of certain DNA repair enzymes.

For the Agrobacterium-mediated repair DNA delivery experiments, we made use of superbinary vectors, as it has been shown by Ishida et al. (1996) that these vectors are very useful for maize transformation. These vectors carry extra virulence genes of the supervirulent A281 strain, a highly efficient strain in the transformation of higher plants (Komari, 1990). Both T-DNA vectors pTCV83 and pTCV87 carry exactly the same repair DNA. The only difference between these superbinary vectors is the size of the extra vir fragment of pTiBo542, which is a 1.3-kb BglII-SphI fragment and a 15.2-kb KpnI fragment from the virulence region of pTiBo542 for pTCV83 and pTCV87, respectively. With both vectors, TSI events can be selected at about the same frequency.

To understand the feasibility of targeted genome optimization in agronomic crops, we studied the frequency and accuracy with which homology-based TSI at a pre-engineered I-SceI site occurred within the maize genome. TSI occurred, although the frequency of recovered events varied between target lines. From the microprojectile bombardment data (Table 1), there was only a twofold difference in transformation frequency between target lines 1-20 and 24, but a 58-fold difference in targeting frequency. This variation in frequency of TSI between target lines may be explained by the fact that the I-SceI sites in different target lines are at different genomic positions, and therefore are not equally accessible to the I-SceI endonuclease and/or repair DNA.

As many of the characterized events are identical, which is expected for gene targeting events, it can be questioned whether all of these events are independent. However, we believe that the reported frequencies of TSI from our targeting experiments are most probably the actual number of independent recombination events. Although suspension cells were used as target cells for recombination, the separation of daughter cells from aggregates was probably not an issue as the selection of recombinants was performed on a solid substrate. After bombardment, the filters with bombarded cells were transferred to a solid selective substrate without being first resuspended in liquid substrate. In the case of Agrobacterium-mediated repair DNA delivery, the cells were co-cultivated on solid substrate for 3 days. After co-cultivation, the cells were resuspended in liquid substrate and grown as a suspension for about 5 days only, prior to plating onto filter papers on top of a solid selective substrate. Moreover, the fact that during this short suspension period of only a few days the suspension remained rather clumpy, and did not develop into a finely dispersed suspension, also minimizes the chance of the selection of identical recombinants in the Agrobacterium delivery procedure.

The precision of homology-based TSI in maize was very high. In a large fraction of the TSI events, no rearrangements at the target locus, nor any additional integrations at other positions in the maize genome, were detected. This was true for DNA delivery by both Agrobacterium and particle bombardment. This is remarkable, as DNA delivery by particle bombardment is usually associated with the integration of fragmented and rearranged DNA sequences at high copy number (Hansen and Chilton, 1996; Travella et al., 2005).

To our knowledge, this is the first report on successful gene targeting by particle bombardment-mediated repair DNA delivery. Agrobacterium-mediated gene targeting in plants has already been reported several times (Miao and Lam, 1995; Hanin et al., 2001; Terada et al., 2002, 2007). Puchta et al. (1996) showed that T-DNA could be used as template to repair an I-SceI-induced DSB in the tobacco genome. In our gene targeting experiments with the target line 1-20, which was used in both Agrobacterium and microprojectile bombardment experiments, the frequency of TSI was higher for microprojectile bombardment than for Agrobacterium. As the other target lines used in microprojectile bombardment and Agrobacterium experiments were different, we cannot draw general conclusions about the preference of particle bombardment- over Agrobacterium-mediated DNA delivery for gene targeting purposes.

The observation that, even on particle bombardment, very precise targeting events were obtained suggests that homology-based DNA integration at an induced DSB could be a useful approach to improve the quality of transgenic plants produced by direct DNA transfer methods, such as particle bombardment. When using standard transformation protocols, it is necessary to produce a large population of independent transformants in order to allow the selection of single-copy transformants with optimal expression of the transgenes and with no side-effects on the overall phenotype of the transgenic plant. It would be advantageous if this trial-and-error process could be replaced by a more directed transformation approach with a more predictable outcome. This would also reduce the high costs associated with repeated field trials required for the elimination of undesired transgenic events. Further progression in the fields of meganucleases and ZFNs may also allow the development of improved transformation procedures, with gene integration at will at any desired position in the genome. Such a gene addition procedure would avoid the risks of mutations arising from random integration and gene silencing by multiple copy integration. Any reduction in these risk factors could be considered positive from a biosafety/acceptability viewpoint.

With the work described in this paper, it was our aim to demonstrate proof-of-concept of targeted genome modification in maize using any line or any DNA delivery method. There are indications that particle bombardment-mediated repair DNA delivery may be the preferred DNA delivery method for gene targeting. However, additional experiments are needed to compare precisely the efficiency of gene targeting between the two DNA delivery methods, allowing sound statistical conclusions to be drawn.

This is a first step towards targeted genome optimization in agronomically relevant crops. It has been shown that precision engineering in maize is feasible through targeted DSB-induced HR at a pre-engineered I-SceI site. These observations, together with developments in the fields of both targeted DSB induction and robust tissue culture techniques for commercially relevant genotypes, may open up perspectives for genome optimization at will at any desired position in the genome of agronomic crops.

Experimental procedures

Bacterial strains and plasmid

The T-DNA vector pTTAM78 carrying the target DNA was constructed as follows. A construct was prepared in which a pnos-neo-3′ocs cassette was linked to a 35S-bar-3′nos cassette, followed by the replacement of the 35S promoter by an I-SceI recognition site and a 3′g7 terminator.

To generate the T-DNA vector pTTA82 carrying the repair DNA, we deleted the 3′nos terminator and C-terminal 18 amino acids from the bar gene in pTTAM78, and replaced the I-SceI site by a 35S promoter flanked by half I-SceI sites in front of the truncated bar gene.

Plasmid pCV78 carries a CaMV35S-synth I-SceI-3′35S gene, with the four amino-terminal amino acids of the coding region of I-SceI being replaced by an NLS of SV40 large T-antigen (Kalderon et al., 1984).

Agrobacterium tumefaciens LBA4404(pSB131) has been described previously (Ishida et al., 1996).

The T-DNA vector pTCV83 carrying the repair DNA linked to a synthetic I-SceI endonuclease was constructed as follows. The synthetic I-SceI gene from plasmid pCV78 was transferred as a 1143-bp EcoRV-XbaI fragment to a plant expression vector pTIB38 carrying the repair DNA fragment and a 1.3-kb BglII-SphI fragment from the virulence region of pTiBo542 (Liu et al., 1992) to give pTCV83.

The T-DNA vector pTCV87 is essentially the same as pTCV83, as it carries exactly the same repair DNA linked to the synthetic I-SceI gene as in pTCV83; however, outside the T-DNA borders, pTCV87 contains an additional 15.2-kb KpnI fragment from the virulence region of pTiBo542 (Ishida et al., 1996).

Plasmid pRVA52 contains a chimeric 35S-cab22L-bar-3′nos gene, which is a derivative from pDE110 (Denecke et al., 1989).

The T-DNA vector ptKA19 carries the rice actin promoter with its intron 1 (McElory et al., 1991), an optimized chloroplast transit peptide, a glyphosate-resistant epsps gene (Lebrun et al., 1997) and a 3′nos terminator.

Plant material, transformation, repair DNA delivery and selection of TSI events

Embryogenic callus was initiated from immature embryos of the hybrid (Pa91 × H99) as female pollinated with the synthetic genotype He89 (Morocz et al., 1990), and maintained on Mahi1VII substrate (D’Halluin et al., 1992).

For the production of target lines with the target locus construct pTTAM78, we performed co-delivery of the target locus construct pTTAM78 with an epsps gene expression cassette ptKA19, allowing selection for glyphosate resistance, as we were unable to select transformants using the neo gene of plasmid pTTAM78 as the selectable marker gene, probably because of the low expression levels of the pnos-driven neo gene. Events resistant for glyphosate were molecularly screened for the presence of an intact single copy of the pTTAM78 target locus DNA. The production of target lines by Agrobacterium was performed essentially as described by D’Halluin (1998). Briefly, embryogenic callus pieces (about 1.5 mm) were pre-induced with 200 µm acetosyringone for 5 days on medium A, which is a slightly modified LS-AS medium (Ishida et al., 1996) [Murashinge and Skoog salts, 0.5 mg/L nicotinic acid, 0.5 mg/L pyridoxine HCI, 1 mg/L thiamin HCI, 100 mg/L myo-inositol, 6 mm l-proline, 0.5 g/L 2-(N-morpholino)ethanesulphonic acid (MES), 2% sucrose, 1% glucose, 1.5 mg/L 2,4-dichlorophenoxyacetic acid (2.4-D), 200 µm acetosyringone, 2.5 g/L Gelrite, pH 5.2]. Pre-induced callus pieces were immersed in a bacterial suspension of EHA101(ptEM3) and EHA101(pTTAM78), as described by Ishida et al. (1996), and co-cultivated for 6 days on medium B, which was identical to medium A with the exception that 6 mm l-proline was replaced by 1 g/L casamino acids and 2.5 g/L Gelrite was replaced by 0.5% agarose BRL Ultra Pure. After co-cultivation, the tissue was transferred to substrate C supplemented with 250 mg/L triacillin and 150 mg/L glyphosate (M & S salts and vitamins, 1 g/L inositol, 0.5 g/L MES, 30 g/L sucrose, 10 g/L glucose, 1.5 mg/L 2.4-D, 2.5 g/L Gelrite, pH 5.8). Proliferating calli were subcultured on Mh1VII substrate (= Mah1VII substrate without caseine hydrolysate and without proline) supplemented with 150 mg/L glyphosate. The production of target lines by PEG with plasmids pTTAM78 and ptKA19 was performed essentially as described by Dudits et al. (2001), using 150 mg/L glyphosate for the selection of transformants.

Suspensions from these target lines were established in N6M cell suspension medium (Morocz et al., 1990), and grown on a shaker (120 r.p.m.) at 25 °C. Suspensions were subcultured every week. For the TSI experiments, the repair DNA and the I-SceI endonuclease were introduced into the target lines by either particle bombardment or Agrobacterium. About 1 week before repair DNA delivery, the suspension cells were diluted in medium A. A few hours before bombardment, the suspension cells were plated as a thin layer on a filter paper on Mah1VIIq substrate (= Mah1VII substrate supplemented with 0.2 m mannitol and 0.2 m sorbitol). The DNA was bombarded into the cells using a PDS-1000/He biolistics device. The particle bombardment parameters were as follows: target distance, 9 cm; bombardment pressure, 9301.5 k Pa; gap distance, 6.4 mm; macrocarrier flight distance, 11 mm. Immediately after bombardment, the tissue was transferred on to Mahq1VII substrate. Four days after bombardment, the filters were transferred on to medium C with 25–50 mg/L PPT. With 14-day intervals, the filters were transferred on to fresh Mh1VII substrate with 10–25 mg/L PPT. The proliferating callus was transferred to MS medium with 6% sucrose, 2 mg/L benzylaminopurine (BAP) and 10 mg/L PPT for shoot regeneration. For Agrobacterium-mediated repair DNA delivery, the cells were imbibed in the Agrobacterium suspension LBA4404(pTCV83) or LBA4404(pTCV87), and co-cultivated as described for the Agrobacterium transformation of callus, but co-cultivation with the LBA strain was only for 3 days. After co-cultivation, the cells were resuspended in substrate C supplemented with 500 mg/L triacillin, and grown as a suspension for about 5 days, prior to plating on to filter papers on top of substrate C with 25–50 mg/L PPT and 500 mg/L triacillin, followed by successive transfers on selective Mh1VII substrate, as for particle bombardment-mediated repair DNA delivery.

Test of the progeny for resistance to PPT

A 0.2% BASTA® (Bayer CropScience, Monheim, Germany) solution was locally applied to the leaves of 8-day-old seedlings from the crossed progeny of TSI events with H99.

Nucleic acid procedures

Genomic DNA was extracted from maize callus or leaf tissue using the cetyltrimethylammonium bromide (CTAB) method (Doyle and Doyle, 1987). The DNA was digested, separated by electrophoresis on a 1% agarose gel, transferred to nylon membranes and hybridized with 32P-radioactive probes following standard procedures (Sambrook and Russell, 2001).

Sequence analysis was performed on a 1932-bp fragment covering the inserted 35S promoter and the entire bar coding region, amplified with primers IB106 in the pnos promoter (5′-GCGGTTCTGTCAGTTCCAAACG-3′) and MDB521 in the 3′nos terminator (5′-GATAATCATCGCAAGACCGG-3′). The fragment was amplified from 40 ng of genomic DNA during 35 cycles of 30 s at 95 °C, 3 min at 64 °C (extended with 7 s per cycle) and 2 min at 72 °C, using the Expand enzyme mix (mixture of thermostable Taq DNA polymerase and thermostable Tgo DNA polymerase). The amplified fragments were subcloned into the pGEM®-T Easy Vector System (Promega, Madison, WI, USA) and sequenced by GENOME Express (Meylan, France).

Acknowledgements

We are grateful to Günter Donn, Miriam Hauben, Els Bonne, Martine Bossut and Rosita Le Page for the production of target lines, and Tanya Moens for the development of the plasmids pTTA82 and pTTAM78. Günter Donn and Koen Weterings are kindly acknowledged for their critical reading of the manuscript, and Kristel D’Hont for technical support. We thank Ilse Van den Brande and Rosalinde van Lipzig for scientific discussions, and Malika De Beir for help in preparation of the manuscript.

Ancillary