Cereal crops have been the primary targets for improvement by genetic transformation because of their worldwide importance for human consumption. For a long time, many of these important cereals were difficult to genetically engineer, mainly as a result of their inherent limitations associated with the resistance to Agrobacterium infection and their recalcitrance to in vitro regeneration. The delivery of foreign genes to rice plants via Agrobacterium tumefaciens has now become a routine technique. However, there are still serious handicaps with Agrobacterium-mediated transformation of other major cereals. In this paper, we review the pioneering efforts, existing problems and future prospects of Agrobacterium-mediated genetic transformation of major cereal crops, such as rice, maize, wheat, barley, sorghum and sugarcane.
Cereal crops belong to the family Poaceae and account for two-thirds of the world's food supply (Borlaug, 1998). Initial improvement of cereals was achieved by conventional breeding. In particular, wide hybridization played a major role in developing numerous cultivars with improved agronomic performances (Brar and Khush, 1986; Xiao et al., 1996; Jauhar and Chibbar, 1999; Jauhar and Peterson, 2001). In recent years, genetic engineering has provided an opportunity for a faster and more targeted introduction of agronomically useful traits into major crop species (reviewed by Giri and Laxmi, 2000; Dahleen et al., 2001; Repellin et al., 2001; Sahrawat et al., 2003; Bajaj and Mohanty, 2005; Jones, 2005). Agrobacterium-based systems and biolistic transformation (direct gene transfer via microprojectile bombardment) have both been successfully used in the genetic transformation of cereals. Although the method of introducing DNA into cells by physical means (i.e. microprojectile bombardment) has revolutionized the field of genetic transformation of crop plants, a major drawback of this system is the considerable variation seen in stability, integration and expression of the introduced transgene (Kohli et al., 1999). The Agrobacterium-mediated transformation system, on the other hand, facilitates the precise integration of a small number of gene copies into the plant genome and shows a greater degree of stability for the transgene (Dai et al., 2001). A general scheme for Agrobacterium-mediated transformation of cereals is outlined in Figure 1.
The natural ability of Agrobacterium to deliver a discrete segment of DNA into the recipient genome has been exploited in Agrobacterium-mediated transformation of cereals (reviewed by Repellin et al., 2001; Cheng et al., 2004). Several factors influencing Agrobacterium-mediated transformation of monocotyledonous plants, including cereals, have been investigated and elucidated (reviewed by Cheng et al., 2004). These factors include the screening of the most responsive genotype and explant, Agrobacterium strain, binary vector, selectable marker gene and promoter, inoculation and co-culture conditions, and tissue culture and regeneration medium. However, the difference in the competence of Agrobacterium to infect a specific tissue, genotype or species is still a major drawback in utilizing Agrobacterium routinely for the introduction of the gene(s) of interest in major cereal crops, and will be the challenge of the future.
Several decades of extensive research have resulted in a relatively detailed understanding of the mechanism of transfer of T-DNA from Agrobacterium to the host plants with the aim of improving the Agrobacterium–plant interaction (reviewed by Gelvin, 2000; Zupan et al., 2000). Recently, the identification and molecular characterization of the plant genes involved in successful Agrobacterium-mediated transformation have opened up new avenues for a better understanding of the plant response to Agrobacterium infection (Veena et al., 2003). Such information may help to develop methods to enhance the transformation frequency of economically important plant species, including cereals. In addition, the in-depth studies and evaluation of the genes responsible for stimulating plant cell division and the competency of plant cells to Agrobacterium may increase not only the extension of transformation protocols to elite genotypes but also the transformation efficiency in cereals. In this review, we focus on recent advances made in obtaining transgenic cereals using Agrobacterium and highlight the existing problems encountered in transformation systems and challenges for the future.
Agrobacterium-mediated transformation of cereals
Agrobacterium tumefaciens naturally infects only dicotyledonous plants. Monocotyledonous plants remained inaccessible to genetic manipulation for many years. Between 1987 and 1993, several laboratories worldwide made enormous efforts to transform monocotyledonous plants, including economically important cereals such as rice, barley and wheat, with Agrobacterium (reviewed by Potrykus, 1991). However, these early studies were controversial, mainly because they were not based on large numbers of transgenic plants showing integration of the desired genes into the genome. In addition, Potrykus (1990) suggested that the researchers might have overlooked the possibility of gene expression by Agrobacterium cells, as well as the potential transformation of microorganisms that were silently infecting the host plant tissues. Chan et al. (1993) obtained a few transgenic rice plants by inoculating immature embryos with a strain of Agrobacterium tumefaciens. They demonstrated the inheritance of the transferred DNA to progeny plants by Southern hybridization, but analysed the progeny of only one transformed plant. In a landmark report, Hiei et al. (1994) resolved the controversy and provided unequivocal evidence for the stable transformation of Japonica rice with Agrobacterium after molecular and genetic analysis of large numbers of R0, R1 and R2 progeny. It is now understood that a number of factors are of critical importance in the Agrobacterium-mediated transformation of rice. These technical requirements explain why it was initially so difficult to apply this technique to rice. A super-binary vector containing an extra copy of each virB, virC and virG from Ti plasmid pTiBo 542 in Agrobacterium strain LBA4404 was demonstrated to be most effective for the transformation of rice.
In recent years, Agrobacterium-mediated transformation of rice has also emerged as a reliable and highly reproducible method for transferring genes of interest into the rice genome (reviewed by Giri and Laxmi, 2000; Ignacimuthu et al., 2000; Tyagi and Mohanty, 2000; Bajaj and Mohanty, 2005). These successes can be attributed to a number of factors, one of the most critical of which was the development of effective tissue culture regimes for the regeneration of fertile plants. Unfortunately, in comparison with rice, other major cereals are still far behind, and reports on the successful transfer of genes of interest via Agrobacterium-mediated transformation are sporadic and limited to specific cultivars.
The past 10 years have seen extensive efforts to develop an efficient Agrobacterium-mediated transformation system for maize. Ishida et al. (1996) took the initiative and reported stable transformation of maize cultivar A188 and its hybrids after co-cultivation of freshly isolated immature embryos with Agrobacterium harbouring a similar super-binary vector to that developed by Hiei et al. (1994). The developed system resulted in transgenic maize plants with transformation frequencies ranging from 5% to 30%. However, to the surprise of Ishida et al. (1996), non-A188-containing lines did not show transformation by the same method that worked for A188 and lines containing A188 as one of the parents. To extend the range of maize genotypes susceptible to Agrobacterium, Zhao et al. (2001) developed a system for Agrobacterium-mediated transformation of maize Hi-II, which contains an A188 and B73 genetic background. The optimized conditions included Agrobacterium inoculation, co-cultivation and resting and selection steps with N6 medium (Chu et al., 1975), an Agrobacterium concentration of 0.5 × 109 colony-forming units (c.f.u.)/mL for inoculation, and 3 days for co-cultivation and 4 days for resting. These conditions produced transgenic maize plants with a transformation frequency of 7.1%. Using phosphomannose isomerase (pmi) as a novel selectable marker gene, Negrotto et al. (2000) recovered transgenic maize plants with a frequency as high as 30%. Although Hi-II has been proven to be a very effective Agrobacterium host and performs well in culture, tissue culture-derived T0 plants have been reported to display serious agronomic problems in the glasshouse, resulting in poor seed set (Frame et al., 2002). In order to boost seed set in T0 plants, Horn et al. (2006) transformed hybrid embryos derived from crosses between Hi-II and elite inbred germplasm via Agrobacterium-mediated transformation. They observed a strong positive effect on the overall corn transformation system using Hi-II/elite crosses. Hybrid T0 plants exhibited excellent tassel development in the glasshouse and seed set was significantly (three- to fivefold) higher than with Hi-II transformants.
Reproducible protocols for Agrobacterium-mediated maize transformation have used super-binary vectors (Ishida et al., 1996; Negrotto et al., 2000; Zhao et al., 2001). However, Frame et al. (2002) demonstrated that maize could also be transformed using Agrobacterium tumefaciens carrying a standard binary vector after producing fertile transgenic maize plants at a frequency of 5.5%. A relatively high transformation frequency resulted with the inclusion of the antioxidant cysteine in the co-culture medium. Antioxidants added to the medium during infection have been reported to favour stable transgenic event recovery in rice (Enríquez-Obregón et al., 1999). After the interaction of Agrobacterium with target tissues, an increase in the survival rate of infected cells and a corresponding increase in the stable transformation frequency may be a result of the fact that the antioxidant, such as cysteine, moderates the effect of the hypersensitive response. In maize, attempts have also been made to improve the transformation frequency with the modification of medium components, optimization of the co-culture and resting periods, and the addition of Agrobacterium growth-inhibiting agents, such as silver nitrate (Armstrong and Rout, 2001; Zhao et al., 2001; Zhang et al., 2003). By optimizing parameters such as vacuum infiltration, enzyme digestion and supersonication treatment and using embryogenic calli of multiple maize inbred lines, Quan et al. (2004b) produced transgenic maize plants with 8%−9% transformation frequency. In a recent study, Yang et al. (2006) evaluated the effects of the purine biosynthesis inhibitor mizoribine, purine and pyrimidine synthesis inhibitors azaserine and acivicin, and surfactant Silwet L-77 on Agrobacterium-mediated transformation of embryogenic calli from maize elite inbred lines Qi 319 and Ye 515. Treatment with suitable concentrations of these compounds during or before inoculation with Agrobacterium resulted in a transformation frequency of almost 12% in Qi 319. Studies on Agrobacterium-mediated transformation of tobacco and Arabidopsis have shown that the application of compounds that inhibit key enzymes in the de novo purine biosynthetic pathway improve the competency of plant cells for Agrobacterium transformation (Roberts et al., 2003). Surfactant Silwet L-77 has also been shown to improve the transformation frequency in wheat (Cheng et al., 1997; Wu et al., 2003).
In the Agrobacterium-mediated transformation of maize, the most advanced step towards enhancing the recovery of transformants was undertaken by Gordon-Kamm et al. (2002) by stimulating cell division with RepA protein (replication-associated protein). RepA-containing transgenic maize (Hi-II) calli remained embryogenic, were readily regenerable and produced fertile transgenic maize plants. The authors concluded that RepA protein stimulated cell division and callus growth in culture, which led to the recovery of enhanced maize transformants. Although most of the successful reports of the Agrobacterium-mediated transformation of maize have been restricted to cultivar A188 or the germplasm that contains this inbred as one of the parents, this study demonstrated that some elite maize cultivars could be transformed efficiently with a modified protocol of Ishida et al. (1996).
These studies indicate that further investigation, attempting to explore the broader significance of such novel approaches, would be very useful not only to extend the transformation efficiency of maize but also for other cereals.
Serious efforts have been made to develop an Agrobacterium-mediated transformation system of wheat since Cheng et al. (1997) documented the possibility of employing Agrobacterium tumefaciens to transform wheat (reviewed by Jones et al., 2005). Cheng et al. (1997) produced stable transgenic wheat plants within 3 months by co-cultivating freshly isolated immature embryos, pre-cultured immature embryos and embryogenic calli with Agrobacterium strain C58 harbouring a binary vector containing the nptII gene for selection and uidA as a reporter gene, both under the control of an enhanced cauliflower mosaic virus (CaMV) 35S promoter. In addition to acetosyringone, the presence of a surfactant during inoculation of the tissue with Agrobacterium was found to be an important factor for the efficient delivery of T-DNA into wheat. Although the transformation efficiency of 1.12% was 10-fold lower than that reported by the same laboratory using particle bombardment, molecular analysis of transgenic plants revealed that approximately 35% of the transgenic plants received a single copy of the T-DNA without rearrangement. Later, the same method as developed by Cheng et al. (1997) was also employed in other laboratories for the production of transgenic wheat (Guang-Min and Zhong-Ti, 1999; Weir et al., 2001) (Table 1).
Table 1. Genetic transformation of cereals using Agrobacterium tumefaciens
Khanna and Daggard (2002) have also demonstrated the superiority of a super-binary vector in Agrobacterium-mediated wheat transformation. Seventeen stable transformants were recovered after infecting 658 calli with LBA4404 harbouring the super-binary vector pHK21, whereas not a single transformant was obtained from 587 calli infected with LBA4404 carrying the binary vector pHK22. Wu et al. (2003) transformed several varieties of wheat with the Agrobacterium AGL1 strain harbouring the pGreen-based plasmid, containing the bar gene for selection and uidA as a reporter gene, both genes driven by the maize ubiquitin 1 promoter plus ubiquitin 1 intron. They produced transgenic wheat plants with transformation frequencies ranging from 0.3% to 3.3% by optimizing the size of the immature embryos, duration of pre-culture of the embryos, inoculation and co-cultivation conditions and the presence of acetosyringone and a surfactant such as Silwet L-77. Of these factors, the pre-culture of immature embryos proved to be very crucial in successful Agrobacterium-mediated wheat transformation. In order to improve the transformation efficiency, Hu et al. (2003) developed an Agrobacterium-mediated wheat transformation system which utilizes 4-day pre-cultured embryos and the aroA:CP4 gene conferring glyphosate resistance (as a selection agent) to transform spring wheat cv. Bobwhite. Glyphosate is an active gradient in the non-selective herbicide ‘Roundup’ that inhibits 5-enolpyruvylshikimate-3-phosphate synthase (EPSPS), a critical enzyme in the biosynthesis of aromatic amino acids. Glyphosate-based selection resulted in transgenic events within 50–80 days with an average transformation frequency of 4.4%.
Cheng et al. (2003) further exploited the fact that an explant such as an immature embryo with active cell division can enhance T-DNA delivery (Cheng et al., 1997; Hu et al., 2003; Wu et al., 2003) in order to increase the recovery of stable transgenic plants in wheat. Following desiccation of plant tissues post-Agrobacterium infection and the use of paromomycin and glyphosate selection, they produced stable transgenic wheat plants with frequencies ranging from 4.8% to 19%. In comparison with non-desiccated tissue of wheat, desiccated plant tissues showed far less browning after pathogen infection, an important factor that may be a useful tool for the development of transformation systems in recalcitrant species. The desiccation treatment has been found to increase the efficiency of T-DNA delivery into both monocot and dicot species and, subsequently, to be beneficial to plant cells/tissue recovery after Agrobacterium infection (Cheng and Fry, 2000). In this report, the authors speculated that the high transformation frequency might be a result of the desiccation treatment during co-culture, which significantly suppressed Agrobacterium growth, and subsequently facilitated better cell recovery after co-culture when compared with the control without desiccation. The stimulatory effect of partial desiccation on plant regeneration has also been reported in indica rice (Chand and Sahrawat, 2001).
These studies indicate that the co-culture conditions and the choice of Agrobacterium strain and vector can influence the successful transformation of wheat. The optimization of factors that directly or indirectly affect host–pathogen interaction should pave the way for the development of a reproducible method for the Agrobacterium-mediated transformation of wheat.
In comparison with wheat and maize, to date there are relatively few reports describing the successful regeneration of adult barley plants containing T-DNA insertions (Table 1). A major breakthrough in the Agrobacterium-mediated transformation of cereals was reported by Tingay et al. (1997), who used a non-super-virulent strain of Agrobacterium carrying a binary vector and produced transgenic plants with 4.2% transformation frequency. This was the first report to show that barley could be transformed using Agrobacterium tumefaciens. The major factors influencing the generation of transgenic barley plants included the wounding of immature embryos and the removal of the axis of immature embryos. In an attempt to extend the range of explant suitable for transformation, Wu et al. (1998) demonstrated that cultured microspores of barley, or cell suspension cultures derived from immature embryos, could also be used as an alternative to immature embryos for genetic transformation. This study described not only the use of visual-detectable marker genes (C1/Lc) as a reporter of transient T-DNA transfer, but also that the cultivars Dissa and Igri were susceptible to Agrobacterium and could be used for the genetic transformation of barley. Following the method of Tingay et al. (1997), which is based on the infection of immature embryos with Agrobacterium, a number of laboratories reported the successful production of transgenic barley plants with transformation frequencies ranging from 0.5% to 6.3% (Patel et al., 2000; Trifonova et al., 2001; Wang et al., 2001; Fang et al., 2002) (Table 1).
In order to increase the transformation frequency, Matthews et al. (2001) transformed barley cultivar Golden Promise following the same strategy as described by Tingay et al. (1997), except that immature embryos were infected with Agrobacterium on the same day of isolation without prior wounding by biolistic gold particles, and transformed tissues were selected on hygromycin rather than bialaphos. Their method produced transgenic barley plants with average frequencies of 2%−12%. Recently, Murray et al. (2004) have transformed four cultivars of barley and have demonstrated that the Agrobacterium-mediated transformation method can be extended beyond the model genotype ‘Golden Promise’ to elite cultivars. However, in comparison with Golden Promise, in which an average transformation frequency of 4.4%−9.2% was achieved, a very low transformation frequency (0.6%) was obtained from the three other elite Australian cultivars: Schooner, Chebec and Sloop. Although stably transformed callus was produced at a high frequency in all the cultivars, only a few transgenic lines from the elite cultivars were produced. This study indicates that the poor transformation frequency in these elite cultivars may not be a result of Agrobacterium's inability to transform the barley scutellar tissues stably, but rather of its inability to generate transformed barley callus capable of regenerating plants. This study also confirms the importance of genotype in barley transformation with respect to regeneration. Recently, Kumlehn et al. (2006) have demonstrated that androgenetic pollen cultures are an excellent target tissue for Agrobacterium-mediated transformation of barley. By optimizing a number of factors, such as the pollen pre-culture time, choice of Agrobacterium strain and vector system, Agrobacterium population density, medium pH and the concentration of acetosyringone, CaCl2 and glutamine, these authors produced 2.2 fertile transgenic plants per spike. Interestingly, almost 60% of the primary transgenic plants set seed, indicating spontaneous genome doubling. This method has excellent potential for the creation of doubled haploid T1 seeds instantly homozygous for the transgene.
Despite successful reports on the Agrobacterium-mediated transformation of barley, including the stable integration of fungal and virus resistance genes into the barley genome, further efforts are required to establish general Agrobacterium-mediated transformation systems for barley.
Arencibia et al. (1998) were the first to produce morphologically normal transgenic sugarcane plants after co-cultivation of calli with Agrobacterium strains LBA4404 and EHA101 carrying a super-binary vector. Transgenic sugarcane plants were obtained with transformation frequencies ranging from 0.94% to 1.15%. The use of young regenerable calli as target explants, the induction and/or improvement of the Agrobacterium tumefaciens virulence system with sugarcane cell cultures and the pre-induction of organogenesis or somatic embryogenesis were found to be crucial parameters in order to increase the competence of the cells for T-DNA transfer. Necrosis and cell death of the tissues after Agrobacterium infection were found to be major limiting factors in developing efficient interactions of sugarcane tissue with Agrobacterium (Enríquez-Obregón et al., 1997).
In order to reduce the necrosis of the target tissues after pathogen infection, Enríquez-Obregón et al. (1998) used antioxidant compounds and produced herbicide-resistant sugarcane plants with transformation frequencies ranging from 10% to 35%. Although it has been shown that phenolic compounds, such as acetosyringone, activate vir genes on the Ti plasmid, the inclusion of phenolic compounds in the co-cultivation medium was not found to be essential in sugarcane to enhance the expression of the vir gene (Joubert et al., 2002). This study shows that cereal cells may be capable of producing a certain level of signal molecules. Elliott et al. (1998, 1999) have demonstrated the use of the green fluorescent protein gene (gfp) as a visual marker gene for Agrobacterium-mediated sugarcane transformation. Their protocol based on the use of gfp as a screenable marker aided the rapid segregation of individual transformation events and facilitated the visual selection of transgenic callus without herbicides, antibiotics or an assay. In order to extend the range of explants suitable for transformation, Manickavasagam et al. (2004) used axillary buds as an explant and produced herbicide-resistant sugarcane plants within 5 months with transformation efficiencies as high as 50%. Plant regeneration through callus cultures increases the chances of somaclonal variation (Larkin et al., 1984), and has been reported to be a particular problem in sugarcane (Lee, 1987; Arencibia et al., 1999). On the other hand, direct regeneration of transgenic plants from an explant, such as axillary buds, without an intervening callus phase, causes minimum genetic changes and therefore can be a method of choice for the production of true-to-type progenies in sugarcane. Recently, Carmona et al. (2005) reported the production of transgenic sugarcane plants after co-cultivation of callus cultures with Agrobacterium. After analysis of the genomic variability in three transgenic sugarcane populations using the amplified fragment length polymorphism technique, they verified that pre-existing DNA polymorphism into the donor genotype and in vitro culture steps during the transformation procedure are the main factors contributing to somaclonal variation in transgenic sugarcane plants.
Although work on the transformation of sorghum began about a decade ago, much less success has been achieved than with other crops, especially with Agrobacterium-mediated transformation methods (Table 1). Therefore, sorghum is considered to be one of the most difficult plant species to manipulate through tissue culture and transformation. Zhao et al. (2000) achieved success in the production of transgenic sorghum plants with an average transformation frequency of 2.1% after co-cultivation of immature embryos with Agrobacterium carrying a super-binary vector. It was found that the source of the embryos had a very significant impact on the transformation efficiency, with field-grown embryos producing a higher transformation frequency than glasshouse-grown embryos. Recently, after a long gap of 5 years, Gao et al. (2005a) used an Agrobacterium-mediated transformation protocol and produced stable transgenic sorghum plants in two inbreds (Tx 430 and C401) and one commercial hybrid (Pioneer 8505) with an average of 2.5% transformation frequency within 4–5 months. As a significant step in the production of transgenic plants, their protocol did not use any selection agent, such as bialaphos or antibiotic. Rather, they successfully used gfp screening for the production of stable sorghum plants with a rice tlp gene, which encodes thaumatin-like protein, to enhance resistance to fungal diseases and drought. This study further demonstrates that, for the genetic transformation of cereals, the ideal selectable marker gene is one that is visually detectable at various stages of plant growth, is non-detrimental to regeneration, and requires no addition of substrate or selective agent. In another report, Gao et al. (2005b) produced 167 transgenic plants with transformation frequencies ranging from 2.88% to 3.30% using the Escherichia coli phosphomannose isomerase gene, pmi, as the selectable marker gene and mannose as the selective agent.
The limited success on the Agrobacterium-mediated transformation of sorghum indicates that sorghum may be a difficult plant species to manipulate for tissue culture and transformation. Further optimization of parameters that are considered to be crucial for cereal transformation, such as the screening of highly regenerative tissue, the genotype and the development of an efficient plant tissue culture, the regeneration system and the use of a positive selection system, should broaden the scope for the genetic transformation of sorghum with genes of interest.
Novel selectable marker genes for the genetic transformation of cereals
A critical step in the regeneration of transgenic plants is the ability to distinguish between transformed plant cells with an integrated transgene and the bulk of non-transformed cells. The traditional way to achieve this goal is to use marker genes within the transgene and to select for their expression. A number of selectable marker genes have been used to generate transgenic cereal plants (Table 2). The most widely used selectable markers in cereal transformation are the genes encoding neomycin phosphotransferase (nptII), hygromycin phosphotransferase (hpt) and phosphinothricin acetyltransferase (bar) (Cheng et al., 2004). These genes confer resistance to kanamycin and some related aminoglycosides (such as G418 and paromomycin), hygromycin and l-phosphinothricin (PPT), respectively. Transformed cells in these systems are able to survive and non-transformed cells are killed by the selective agents. This type of selection is referred to as negative selection (Joersbo and Okkels, 1996; Hansen and Wright, 1999). It has been observed that, although kanamycin can be used as a selective agent during regeneration, it is not effective for the selection of transformed calli. In addition, callus cultures recovered after kanamycin selection are unable to regenerate green plants (Ayers and Park, 1996). G418 has been used in the Agrobacterium-mediated transformation of rice (Aldemita and Hodges, 1996). However, the authors also found that the exposure of cells to G418 for long periods inhibited regeneration. Another widely used and effective selectable marker is hpt, which confers resistance to the aminoglycoside antibiotic hygromycin. The bar gene confers resistance to PPT, glyphosinate (an ammonium salt of PPT) and bialaphos (a derivative of PPT). In the Agrobacterium-mediated transformation of rice, Wang et al. (1997) observed an improved transformation frequency after the insertion of introns into the coding region of hpt. In addition to an improved transformation frequency, these authors found reduced copy numbers of the marker gene. This modification in the selectable marker also enhanced the stable transformation in rice and barley cultivars (Upadhyaya et al., 2000; Wang et al., 2001).
Table 2. Selectable marker genes commonly used in cereal transformation
Unique ability to grow on xylose as a sole carbon source
As a result of increasing concern worldwide regarding the use of antibiotic or herbicide markers in transgenic crop plants, several positive selection systems have been developed in recent years and successfully used for the production of transgenic plants. In contrast with negative selection, positive selection does not kill the non-transgenic cells, but gives clear advantages to the transformed cells. Positive selection systems include benzyladenine N-3-glucuronide (Joersbo and Okkels, 1996), xylose (Haldrup et al., 1998a,b) and mannose (Miles and Guest, 1984). These selection systems allow transgenic plants to be produced without antibiotic or herbicide resistance genes. PMI catalyses the reversible interconversion of mannose-6-phosphate and fructose-6-phosphate. In comparison with a selection system in which transformed cells are selected on medium containing antibiotic or herbicide resistance markers, this type of selection utilizes the E. coli manA (pmi) gene encoding PMI as a selectable marker gene. pmi has been used as a positive selectable marker for particle bombardment and Agrobacterium-mediated transformation in many plant species, because expression of the E. coli manA (pmi) gene allows plant cells to utilize mannose as a carbon source and to survive on media containing mannose (Joersbo et al., 1998; Negrotto et al., 2000; Wang et al., 2000; Reed et al., 2001; Wright et al., 2001; Datta et al., 2003). In this system, cells transformed with pmi acquire a metabolic advantage, compared with non-transformed cells that remain unable to metabolize mannose-6-phosphate. This selection system has advantages over herbicide- and antibiotic-resistant selectable marker genes because of the potential risk of their transfer to the environment, and provides an efficient way to produce transgenic plants with high frequency (Negrotto et al., 2000; Reed et al., 2001). With the aim of developing maize transformation systems with better selection regimes, Negrotto et al. (2000) used pmi as a positive selectable marker gene and produced transgenic maize plants with high frequency (30%). Similarly, Reed et al. (2001) transformed maize, wheat and barley to express the E. coli manA gene (pmi). Transformation frequencies averaged 45% for maize transformation via projectile bombardment, 35% for maize Agrobacterium-mediated transformation, 20% for wheat, 3% for barley and 2% for watermelon transformation. The transformation frequencies were found to exceed those obtained for maize and wheat using the pat or bar gene with Basta selection. Moreover, the yield and nutritional composition of grain from pmi-transformed corn plants, compared with their non-transformed isogenic counterparts, were determined, and no statistically significant differences were found. Mannose selection has also been successfully used to produce transgenic rice plants via Agrobacterium-mediated transformation (Lucca et al., 2001; He et al., 2004). The establishment of this selection system in rice provided an efficient way of producing transgenic plants without using antibiotics or herbicides with a transformation frequency of up to 6% (He et al., 2004) and 41% (Lucca et al., 2001). In Agrobacterium-mediated transformation, Gao et al. (2005b) successfully produced transgenic sorghum plants using pmi as the selectable marker gene and mannose as the selective agent. It is believed that, in the pmi selection system, the arrest in cell growth of untransformed cells by starvation, rather than the necrosis induced in traditional selection systems, contributes to better survival and proliferation and, as a result, a high frequency of plant regeneration.
A new selectable marker system consisting of the herbicidal compound butafenacil (Tomlin, 2000) and its molecular target, protoporphyrinogen oxidase (PPO), has been utilized for maize Agrobacterium-mediated transformation (Li et al., 2003). PPO is a key enzyme in the chlorophyll/haem biosynthetic pathway, catalysing the oxidation of protoporphyrinogen IX to protoporphyrin IX (Smith et al., 1993). This system produced more than 2500 herbicide-tolerant maize plants. The bacterial CP4 gene was successfully used in the Agrobacterium-mediated transformation of wheat (Cheng et al., 2003; Hu et al., 2003). The development of a non-lethal selection system using the aminoglycoside-3′-adenyltransferase (aadA) marker gene for the efficient recovery of transgenic rice plants has been described by Oreifig et al. (2004). The use of aadA gene-mediated streptomycin resistance for the non-lethal selection of transgenic rice resulted in plant regeneration frequencies under selection pressure as high as those in non-transformed controls without selection. In this system, a vital pH indicator – chlorophenol red – was applied together with streptomycin to select transgenic calli by making use of the phenomenon that fast-growing cell lines lower the pH in the culture medium. Although non-transgenic plants showed retarded growth and bleaching in the presence of streptomycin, transgenic plants showed a dark green colour and more intense growth. Interestingly, the final transformation frequency was double that of transformations using PPT as a selective agent, and almost three times greater than that seen with transformations using hygromycin resistance. Streptomycin, unlike kanamycin, hygromycin and PPT, does not kill plant cells, and the differentiation of resistant cell lines is based on greening and faster growth (Jones et al., 1987). Another positive selection system based on xylose isomerase has been described in dicot plants (Haldrup et al., 2001). In this system, transgenic cells expressing xylA genes, encoding for the xylose isomerase gene from Streptomyces rubiginosus (Haldrup et al., 1998a) and Thermoanaerobacterium thermosulfurogenes (Haldrup et al., 1998b), can utilize xylose as a carbohydrate source and proliferate, whereas non-transgenic cells starve.
The gfp gene has also been utilized as a visual screenable marker to produce stably transformed fertile oat plants (Kaeppler et al., 2001). A synthetic gfp gene under the control of the maize ubiquitin promoter was delivered into embryogenic oat callus via microprojectile bombardment. Cell lines expressing gfp were visually identified using epifluoresence microscopy and physically isolated 3 weeks after bombardment. After transfer to regeneration medium, gfp-expressing plants were generated from 78% of the glowing cell sectors. Southern blot and polymerase chain reaction (PCR) analysis confirmed transgene integration and transmission to progeny. In addition, the transgene copy number and integration patterns were found to be similar to those in transgenic plants derived from a chemical-based selection system. Howe et al. (2002) have described the development of an efficient selectable marker system for the production of transgenic maize plants using genes that confer resistance to the herbicide glyphosate. Glyphosate, the active ingredient in the herbicide Roundup, inhibits the plastid enzyme EPSPS and thus prevents the synthesis of chorismate-derived aromatic amino acids and secondary metabolites in plants. Stably transformed callus and regenerated transgenic maize plants were obtained following the introduction of mutated maize EPSPS and GOX (glyphosate oxidoreductase, a bacterial gene that degrades glyphosate) genes into embryogenic callus or suspension cells by particle bombardment. Zhou et al. (1995) also reported the regeneration of transgenic wheat plants using the CP4 and GOX genes as selectable markers. However, the transformation frequencies for the CP4 and GOX genes were lower than those of nptII and bar under the same experimental conditions. The gene cah from the soil fungus Myrothecium verrucaria (Maier-Greiner et al., 1991), coding for the enzyme cyanamide hydratase, which converts cyanamide into urea, has also been used as a selectable marker gene for the genetic transformation of wheat via particle bombardment (Weeks et al., 2000). Cyanamide hydratase catalyses the hydration of the nitrile group of cyanamide to form urea, which can be used for plant growth as a useful nitrogen compound. Therefore, the cah gene may offer benefits to the plant that are not conferred by other selectable marker genes.
Removal of selectable marker genes from transgenic cereal plants
Selectable marker genes are required in virtually all plant transformation approaches, but are not generally required for the expression of the trait gene in the transgenic plants. The safety of each selectable marker gene is another potential problem in genetically engineered plants. Selectable markers are mostly based on genes conferring antibiotic or herbicide resistance. In recent years, as a result of increasing public and scientific concern regarding the presence of selectable marker genes (herbicide and antibiotic resistance genes) in transgenic plants, a number of techniques have been developed for the removal of selectable marker genes. These include the use of excision systems, such as Cre/lox (Dale and Ow, 1991), the generation of co-transformants in which the gene of interest and selectable marker gene segregate in later generations (Komari et al., 1996), and the use of a transposon system to separate the selectable marker gene and gene of interest (Goldsbrough et al., 1993).
In 1991, David Ow's group (Dale and Ow, 1991) demonstrated that a selectable marker placed in a transgene between two lox sites could be removed from the plant genome through the expression of the Cre recombinase. Cre is a site-specific recombinase that catalyses the circularization of the bacteriophage P1 of E. coli using lox sites at the ends of its genome unit. Recombinase-mediated DNA arrangements can include site-specific excision, integration, inversion and interchromosomal recombination. Zhang et al. (2003) used the Cre/lox method to remove the marker gene, and successfully produced transgenic maize plants with transformation frequencies ranging from 1% to 18.9%. They used immature embryos from a three-way cross of maize inbreds, (Pa91 × H99) × A188, for the stable transformation of maize. In wheat, a lox-containing transgenic line was crossed with a cre-expressing line, and extra copies of the transgene were deleted by site-specific recombination (Srivastava and Ow, 2003). This process included the removal of a lox-flanked selection marker gene, bar.
Generation of marker-free transgenic plants using the co-transformation strategy involves the introduction of two separate T-DNAs into the same plant cell; one carries the selectable marker and the other carries the gene of interest (Hohn et al., 2001). Unlinked integrations of the two T-DNAs lead to the segregation of the marker gene from the gene of interest in the T1 generation. Co-transformation can be performed using either two strains, or a single strain, of Agrobacterium tumefaciens. A mixture of two strains, each harbouring a binary vector (Komari et al., 1996), or a co-integrate and a binary vector (De Buck et al., 2000), have been used to study the factors that influence co-transformation frequencies. An alternative method for co-transformation using a single strain of Agrobacterium has been shown to yield higher co-transformation frequencies (Komari et al., 1996). In this technique, two T-DNAs can be harboured in the same Agrobacterium strain, either on different replicons (Daley et al., 1998), or on the same replicon (Komari et al., 1996). In maize, co-transformation with an octopine strain carrying a binary vector with two T-DNAs yielded co-transformation frequencies of 93% for the bar and gus (β-glucuronidase) genes in the R0 generation; 64% of the R1 progeny segregated as bar-free plants expressing gus (Miller et al., 2002). This contrasted dramatically with the 11.7% co-transformation frequency with mixed Agrobacterium strains (Miller et al., 2002). Matthews et al. (2001) produced transgenic barley plants free from selectable marker genes following co-transformation using a plasmid carrying two T-DNAs located adjacent to each other with no intervening region. Using this vector, they transferred three genes of interest into barley and observed co-insertion in 66% of the transformants. Only 24% of these segregated as marker-free plants, perhaps because nopaline strains were required for barley transformation. Huang et al. (2004) have also generated marker-free transgenic maize by regular two-border Agrobacterium transformation vectors. In comparison with a conventional single binary plasmid with multiple T-DNA borders, their protocol showed the possibility of producing marker-free transgenic plants by simply repositioning the selectable marker gene into the backbone of a regular two-border binary plasmid and leaving only the gene of interest in the T-DNA region. They found that T-DNA carrying only the gene of interest could be independently inserted and segregated from the T-DNA carrying the selectable marker to generate marker-free transgenic maize plants. Out of three modified vectors, the transformation frequencies (11.3% and 13.4%) of the two modified vectors were comparable with the efficiency of conventional vector (11%). Permingeat et al. (2003) have also obtained transgenic wheat plants without the selectable marker gene in co-transformation experiments using the gene-gun method with mild selective pressure during transformation. This method used a mixture of two plasmids: pGL2, containing the selectable marker gene hpt, and pAi1Gus, containing the uidA gene. Bombarded calli were subjected to the selective pressure of 25 mg/L of hygromycin and produced 19 transgenic wheat plants. Of the 19 transgenic plants, seven were co-transformed, nine plants contained only the hpt gene and three plants contained only the uidA gene. Southern hybridization confirmed that the uidA gene was transmitted to the progeny. This report suggests that it is possible to obtain transgenic wheat in co-transformation experiments with the gene of interest, but without introducing the selectable marker gene. However, the convenience of the method will largely depend on the suitability of the analytical method required. Moreover, Hohn et al. (2001) have suggested that the elimination of marker genes by co-transformation may be especially useful when using Agrobacterium-mediated transformation. In rice, Lu et al. (2001) produced marker-free transgenic plants using a double right-border binary vector, which carries two sets of T-DNA border sequences (twin T-DNA vectors) flanking a selectable marker gene, followed by a gene of interest and one copy of the left border sequence. Two sets of T-DNA inserts, one initiated from the first right border containing both the selectable gene and gene of interest, and the other from the second right border containing only the gene of interest, were integrated into the genome. In the subsequent generation, Lu et al. (2001) found that these inserts were segregated away from each other, allowing the selection of the progeny with only the gene of interest. A total of 36%−46% of the primary transformants yielded selection marker-free progeny. Thus, in combination with ‘twin T-DNA’ vectors, marker-free transgenic plants can be produced by carefully designing the transformation vectors.
Another method for producing marker-free transgenic plants was proposed by Vain et al. (2003) using a new dual binary vector system pGreen/pSoup (Hellens et al., 2000). pGreen is a small Ti binary vector unable to replicate in Agrobacterium without the presence of another binary plasmid, pSoup, in the same strain. Co-transformation with pGreen, carrying the gene of interest, and pSoup, carrying the selectable marker, may lead to the production of marker-free transgenic plants in subsequent progeny (Vain et al., 2003).
The maize Ac/Ds transposable element system has also been used to create novel T-DNA vectors for separating genes that are linked together on the same T-DNA after insertion into plants. The expression of the Ac transposase from within the T-DNA can induce the transposition of the gene of interest from the T-DNA to another chromosomal location. This results in the separation of the gene of interest from the T-DNA and selectable marker gene. In rice, this technique was used to create hpt-free rice plants that expressed the Bt endotoxin coded by the cry 1B gene (Cotsaftis et al., 2002). To generate marker-free plants, cry 1B was placed in the leader sequence of a gfp marker gene to monitor the transposition by the activation of gfp activity. Cotsaftis et al. (2002) found that excision and reinsertion occurred at high frequencies, and plants were recovered with high levels of resistance to striped stem borer. Transposon-mediated gene delivery has been reported to increase transgene expression stability in barley (Koprek et al., 2001).
The MAT vector system (for multiauto transformation) uses transposon or, more recently, recombinase-based excision to enable the production of marker-free transgenic plants (Sugita et al., 2000). An Agrobacterium isopentenyltransferase (ipt) gene provides a positive visual selectable marker for transformation by catalysing cytokinin synthesis and inducing a ‘shooty’ phenotype on hormone-free medium. After selection, subsequent excision via the R/RS system produces marker-free transgenic plants with a normal phenotype, allowing ipt and MAT to be used again for another round of transformation. Recent improvements to the method have increased its efficiency and have allowed it to be applied to species that do not regenerate through cytokinin-dependent organogenesis, but rather via somatic embryogenesis (Endo et al., 2002). Rice calli were transformed with the MAT vector containing the chimeric ipt gene from Agrobacterium and gfp genes fused with the CaMV 35S promoter. The site-specific recombination R/RS system was used to excise them after transformation together with the other genes (gusA, nptII and hpt) outside the R/RS cassette (Endo et al., 2002). Ebinuma and Komamine (2001) reported a single-step transformation method for generating marker-free transgenic rice plants using the ipt-type MAT vector. The MAT vector consisted of the nptII, gus and hpt genes outside of the RS-flanked 35S-ipt, 35S-R and gfp genes. Molecular analysis of transgenic plants revealed that 75% of the transgenic shoots contained the gus gene, but had lost the ipt gene (Ebinuma and Komamine, 2001).
These studies clearly indicate that marker-free transgenic cereal plants can be generated at varying efficiencies using different approaches and techniques, followed by segregation of the genes in the subsequent sexual generation. However, there are limitations associated with these techniques; for example, co-transformation technology is not suitable for all plant species and its efficiency is clearly dependent on a number of variables, including the Agrobacterium strain and the plant tissue being transformed. In addition, this technique is labour intensive, requiring the production of a large number of transgenic plants to isolate the plant of interest. Although site-specific recombinases hold the greatest promise for the excision of selectable marker genes, concerns also exist about pleiotropic effects induced by the action of recombinase on cryptic excision sites in the plant genomes.
Transgene integration and expression, and characteristics of transgenic plants
Transgenic plants often contain complex integration structures at an undetermined genomic location, which may cause variations in gene expression. It has been demonstrated that the precise integration of a transgene in a pre-determined genomic location can reduce the variation in transgene expression (Day et al., 2000). The integration of transgenes in a pre-determined genomic locus can be achieved by the use of site-specific recombinase systems, such as Cre/lox and FLP/frt (Ow, 2002). Integration by homologous recombination would favour the establishment of a simple integration pattern and allow the insertion of a transgene into a known and stable region of the genome. In the Cre/lox system, a target site, such as the lox site, is randomly introduced into the plant genome; the DNA construct is then integrated into this genomic target site via recombinase-mediated site-specific integration. The Cre/lox site-specific recombination system has been used successfully in wheat and rice to target single-copy insertion to lox sites placed on the genome (Srivastava et al., 1999, 2004). The FLP-FRT system involves using flippase (FLP) recombinase, derived from the yeast Saccharomyces cerevisiae (Sadowski, 1995). FLP recognizes a pair of FLP recombinase target (FRT) sequences that flank a genomic region of interest. A decrease in gene silencing problems, caused by position effects, can also be aided by the use of matrix attachment regions (MARs) placed on each side of the transgene expression cassette. MARs are DNA elements that are thought to influence gene expression by anchoring active chromatin domains to the nuclear matrix (Cockerill and Gerrard, 1986). Transgenes flanked by MARs are more strongly expressed and resist gene silencing (Allen et al., 2000). Recently, it has been shown that MARs can protect against gene silencing mediated by transcription of the transgene (Abranches et al., 2005). Gene cassettes containing MAR sequences have been reported to increase transgene expression levels in rice transformation (Vain et al., 1999; Van der Geest et al., 2004). This system thus appears to insulate the transgene from influences exerted by the surrounding chromatin, finally giving some protection against gene silencing.
The stable inheritance and expression of foreign genes are of critical importance in the application of genetically engineered cereals to agriculture. The perfect transformation would contain a single copy of the transgene that would segregate in a Mendelian fashion, with uniform expression from one generation to the next. However, studies on transgene behaviour have reported non-Mendelian transgene segregation in transformed lines of cereals (Barro et al., 1998; Vain et al., 2003). This is a highly undesirable character when it occurs for transgenes encoding a useful trait. Many factors can contribute to variation in transgene expression, including tissue culture-induced variation or chimerism in the primary integration site (position effects), transgene copy number (dosage effects), transgene mutation and epigenetic gene silencing (Barro et al., 1998; Maqbool and Christou, 1999; Mohanty et al., 1999; Matzke et al., 2000; James et al., 2002). Gene silencing, the decline or loss of gene expression in subsequent generations of primary transformants, can occur at the transcriptional or post-transcriptional level, and the phenomenon has often been associated with a high transgene copy number (Matzke and Matzke, 1995; Matzke et al., 2000). In an analysis of transgenic rice lines, Chareonpornwattana et al. (1999) observed that gene silencing was more correlated with transgene expression level than with copy number. High promoter activity has also been found to be correlated with hypermethylation and abolishment of gene transcription in both monocot and dicot plant species (Matzke and Matzke, 1995; Morino et al., 1999). Thus, the choice of promoter to drive transgene expression may have a large impact on the occurrence of silencing. In our laboratory, we have also found occasional transgene instability in Agrobacterium-derived transgenic lines of barley and rice (Shrawat and Lörz, unpublished data). This may also be because backbone sequences from the binary vector are sometimes transferred and integrated into the plant genome, and the presence of such vector sequences may influence transgene expression (Matzke and Matzke, 1998). These studies indicate that the problem of transgene silencing raises serious concerns regarding the selection of transgenic lines for crop improvement with specific trait(s). Therefore, it now appears imperative that transgenic lines carrying gene(s) of economic importance need to be carefully tested for gene expression levels over many generations.
The patterns of integration, inheritance and expression of transgenes in plants after Agrobacterium-mediated and direct delivery-mediated transformation have been reported by many researchers (Kohli et al., 1999; Dai et al., 2001; Altpeter et al., 2004). Several excellent articles have been published which elaborate the mechanism of T-DNA transfer from Agrobacterium to the plant genome and favour the use of Agrobacterium-mediated transformation systems for the efficient delivery of foreign DNA into plant genomes (Gelvin, 2003; Tzfira and Citovsky, 2003; Valentine, 2003; Cheng et al., 2004). An excellent review article by Altpeter et al. (2005) also demonstrates the importance of particle bombardment in generating large numbers of transgenic plants with agronomically important traits. Both of these methods have advantages and disadvantages (Dai et al., 2001). However, as is the case with Agrobacterium-mediated transformation, the mechanism of transgene integration using particle bombardment is not well understood (Pawlowski and Somers, 1996; Kohli et al., 2003). This may be because such insertion events tend to be difficult to analyse, as they are often large and complex with multiple copies of transgene sequences at a single insertion site (Kohli et al., 1998; Maqbool and Christou, 1999; Vain et al., 2002; Kohli et al., 2003). The direct DNA delivery systems tend to result in integration of multiple copies of transgenes at single loci and the rearrangement of transgenes (Kohli et al., 1999; Dai et al., 2001). The Agrobacterium-mediated transformation system also results in small- or large-scale deletions, duplications or rearrangement of nearby plant DNA sequences, and in the integration of non-contiguous plant DNA at the site of T-DNA insertion (Dong et al., 1996; Tinland, 1996). T-DNA insertion can also cause integration of multiple elements in different patterns of inverted or tandem repeats (De Buck et al., 2001; Eamens et al., 2004). Agrobacterium-mediated transformation has also produced transgenic plants with insertions of incomplete T-DNAs, or insertions of more than one T-DNA (or T-DNA fragments), at one or more locations in the plant genome (Bhattacharyya et al., 1994; Cheng et al., 1997; Wolters et al., 1998; Dai et al., 2001; Dong et al., 2001; Forsbach et al., 2003; Kim et al., 2003; Kohli et al., 2003). Together with superfluous T-DNA insertions, insertion of superfluous plasmid DNA in the plant genome has been reported in the literature (Wenck et al., 1997; De Buck et al., 2000). Kim et al. (2003) reported that 45% of rice transformants contained integrated plasmid DNA. In Agrobacterium-mediated transformation, Wolters et al. (1998) found complex T-DNA insertion events, in which numerous stretches of plasmid DNA were interspersed with T-DNA and plant DNA. Vector backbone sequences have also been detected in transgenic rice plants generated through particle bombardment (Kohli et al., 1999). Irrespective of the transformation methods used for the production of transgenic plants, integration of vector backbone sequences into plant genomes has been found to have undesired effects on transgene stability and expression (Breitler et al., 2002). Therefore, efforts are being made to produce transgenic plants devoid of any backbone sequences (Fu et al., 2000; Breitler et al., 2002). By using minimal linear transgene constructs (promoter, coding region of the gene and terminator), Fu et al. (2000) produced transgenic plants with single-copy number integration events and fewer rearrangements than the transgenic plants produced with whole plasmid. This result suggests that using particle bombardment with a minimal gene cassette may be a favourable method for producing transgenic plants devoid of vector backbone sequences. However, irrespective of which method is used for transformation, the integration of transgenes in plants occurs through illegitimate recombination. Therefore, it is important to understand the process of integration in order to produce so-called ‘clean’ transgenic plants.
Despite the fact that the Agrobacterium-mediated gene transfer system also results in the integration of multiple elements in different patterns of inverted or tandem repeats, it is unlikely to be replaced by particle bombardment. In most cases, Agrobacterium-mediated transgenic plants show a lower copy number of transgenes and a more predictable pattern of integration (Cheng et al., 1997; Dai et al., 2001; Wu et al., 2003; Altpeter et al., 2004; Travella et al., 2005). Cheng et al. (1997) compared transgenic wheat lines produced by either particle bombardment or Agrobacterium, and found that 17% (13/77) of the bombarded lines contained a single copy compared with 35% (9/26) of those transformed by Agrobacterium. Another study from the same group found that 56% (40/77) of the bombarded lines showed a 3 : 1 segregation ratio, which is an indication of single-locus integration, compared with 47% (45/97) for Agrobacterium transformation; in addition, more of the Agrobacterium lines contained a single transgene copy (Hu et al., 2003). Later, Wu et al. (2003) also demonstrated that 22 of 44 Agrobacterium-derived wheat transgenic lines showed a segregation ratio of 3 : 1, with 12 of 20 with single copies. Dai et al. (2001) reported that transgenic rice plants obtained by Agrobacterium and direct delivery methods contained similar gene copy numbers, but the percentage of transgenic plants containing intact copies of foreign genes without rearrangement was significantly higher for Agrobacterium-mediated transformation. Moreover, Dai et al. (2001) reported that the fertility of transgenic plants obtained by Agrobacterium-mediated transformation was comparatively better than that of plants obtained after particle bombardment. Using Southern, Northern and real-time PCR techniques, Shou et al. (2004) compared transgene copy numbers and RNA expression levels in T1 and T2 generations of transgenic maize plants generated using particle bombardment and Agrobacterium-mediated transformation. They demonstrated that Agrobacterium-derived maize transformants contained lower transgene copies and higher and more stable gene expression than their bombardment-derived counterparts. Similarly, in barley, although 100% of the Agrobacterium-derived lines integrated between one and three copies of the transgene with minimal rearrangements, 60% of the transgenic barley lines derived by particle bombardment integrated more than eight copies of the transgene with extensive DNA rearrangements and multiple integration events (Travella et al., 2005). It was further demonstrated that the integrated DNA was stable and inherited in a Mendelian fashion in the majority of Agrobacterium-derived lines, whereas transgene silencing was frequently observed in the T1 populations of the bombarded-derived lines. Altpeter et al. (2004) generated large numbers of transgenic ryegrass plants using particle bombardment and Agrobacterium-mediated gene transformation methods. They found that the majority of the transgenic plants from both bombardment and Agrobacterium-mediated gene transfer had simple transgene integration patterns, with transgene copies ranging from one to four. None of the lines generated by Agrobacterium-mediated gene transfer contained more than five T-DNA inserts. In comparison, approximately 20% of the ryegrass transgenic lines generated by particle bombardment showed very complex integration patterns with between five and 20 transgene copies. Approximately 50% of these high copy number lines stably expressed the transgene. Interestingly, transgenic rice containing more than one copy of the transgene performed well under field conditions against a number of insect pests (Tu et al., 2000; Ye et al., 2001). These studies clearly indicate that both particle bombardment and Agrobacterium-mediated gene transfer techniques have a similar potential to produce fertile and stable transgenic plants.
Studies on cereal transformation indicate that particle bombardment-mediated delivery facilitates a broad variety of transformation strategies with a wide range of gene expression, has no biological constraints or host limitations, and diverse cell types can be targeted efficiently for foreign DNA delivery (Altpeter et al., 2005). Nevertheless, Agrobacterium-mediated transformation will remain a method of choice for obtaining transgenic plants with a lower copy number, intact foreign genes and stable gene expression. In addition, recent advances in cereal transformation have indicated that Agrobacterium-mediated transformation systems not only promise to intensify attention on the production of cereals with genes of interest, but also to concurrently test and capture the value of the introduced traits (Table 3). Therefore, several laboratories are making efforts to establish Agrobacterium-mediated transformation systems to achieve stable expression of the transgene at an expected level, rather than that imparted by a random site of integration. We strongly believe that, irrespective of the transformation method used for foreign DNA delivery, in the long run, the emphasis on the production of transgenic plants should be on quality rather than quantity, as the handling, maintenance and genetic and molecular analysis of transgenic plants are costly and time consuming.
Table 3. Agronomically important genes transferred to cereals using Agrobacterium tumefaciens
In recent years, the transformation of rice has progressed to a level at which agronomically useful genes can now be introduced into the rice genome employing Agrobacterium as a vehicle of DNA delivery. Unfortunately, in comparison with rice, Agrobacterium-mediated transformation of other major cereal crops lags significantly behind (Table 1). Tissue browning and necrosis after Agrobacterium infection are still major obstacles in the genetic transformation of cereals. In the Agrobacterium-mediated transformation of barley, we have found that transformed tissues become severely necrotic even 1 day after co-cultivation. However, despite severe necrosis, transformed tissues often contain areas of cells still apparently alive, with the rest of the area becoming necrotic. Indeed, these cells, when cultured on a suitable medium containing selective agent, further proliferate into embryogenic callus and finally develop into transgenic plants (Figure 2).
On pathogen infection, one of the earliest defence mechanisms activated is the production of reactive oxygen species, referred to as an oxidative burst. The reactive oxygen intermediates produced during the oxidative burst are responsible for activating programmed cell death (Parrott et al., 2002). The co-cultivation of Agrobacterium with maize or wheat tissues has been reported to result in a process closely analogous to apoptosis in animal cells, wherein cell death is characterized by DNA cleavage into oligonucleosomal fragments and defined morphological changes (Hansen, 2000). Parrott et al. (2002) reported that, after Agrobacterium infection, wheat embryos and root cells rapidly produced hydrogen peroxide (H2O2), displayed altered cell wall composition and resulted in higher levels of cellular necrosis and subsequent cell death. A correlation between the reduction in cell death and the improved transformation frequency has been demonstrated in rice (Enríquez-Obregón et al., 1998), sugarcane (Enríquez-Obregón et al., 1997), sorghum (Zhao et al., 2000) and maize (Ishida et al., 1996). Parrott et al. (2002) also observed that lowering the O2 tension from 7.4 to 2.1 mm significantly reduced the extent of embryo and root cell death in wheat after Agrobacterium transformation. It has also been found that Agrobacterium-induced necrosis observed in Poaceae can be inhibited by the use of necrosis-inhibiting agents, such as silver nitrate (Hansen and Durham, 2000). Silver nitrate has been found to suppress Agrobacterium growth during co-culture without influencing T-DNA delivery and subsequent T-DNA integration (Cheng et al., 2003). Anti-necrotic treatment of the target tissues may provide an adequate environment for the interaction of Agrobacterium with the plant cells by inhibiting necrosis, and may result in increased efficiency of transformation (Enríquez-Obregón et al., 1997). Furthermore, a molecular understanding of cell death via apoptosis-like processes may lead to the minimization of Agrobacterium-induced cell death. The modification of the potential Agrobacterium elicitors that signal plant cell death should be studied in more depth, in addition to the prevalent use of antioxidants, to improve the transformation frequency of cereals.
In Agrobacterium-mediated transformation of cereals, apart from necrosis, a number of factors, including genotype, explant, Agrobacterium strain, binary vector, selectable marker, inoculation and co-cultivation medium, and co-culture conditions, have been found to influence the recovery of stable plant cells after Agrobacterium infection (reviewed by Cheng et al., 2004; Jones et al., 2005). Of these factors, genotype and explant are considered to be the major limiting factors in Agrobacterium-mediated transformation of cereals, especially in extending the host range to commercial cultivated cultivars. Amongst the cereals transformed to date, rice appears to be the least genotype dependent, as more than 40 genotypes of japonica, indica and javonica have been transformed. In comparison, only a few model genotypes have been successfully used in the Agrobacterium-mediated transformation of other major cereal crops: for example, cultivar A188 or its hybrids in maize, cultivar Bobwhite in wheat, Golden Promise and Igri in barley, and cultivar Ja 60-5 in sugarcane. Although transgenic plants have recently been recovered from elite cultivars or lines of sorghum (Zhao et al., 2000), maize (Gordon-Kamm et al., 2002) and barley (Wang et al., 2001), the overall transformation frequency is lower than that with model cultivars. These studies indicate that the genotype-dependent response in Agrobacterium-mediated transformation of cereals is a major drawback in extending Agrobacterium-mediated transformation systems to elite cultivars of economically important cereals. Therefore, it becomes important to make elite cultivars amenable to tissue culture and to improve their regenerability by manipulating existing tissue culture medium.
The difference in the susceptibility of genotypes to Agrobacterium may be a result of the presence of inhibitors of the Agrobacterium sensory machinery. 2,4-Dihydroxy-7-methoxy-2H-1,4-benzoaxzin-3(4H-one) (DIMBOA), the major organic exudate of maize seedling roots which is present in varying amounts in different genotypes of maize, specifically inhibits the induction of vir gene expression (Zhang et al., 2000). The ability of Agrobacterium strains to infect recalcitrant plants has also been shown to be determined by the VirA locus (Heath et al., 1997). The relative difference in the resistance of agronomically important plant species, such as maize, to Agrobacterium-mediated genetic transformation may be a result of the presence of such inhibitors, rather than to insufficient activation of the Agrobacterium virulence machinery by host cell exudates. It seems that naturally occurring inhibitors directed against signal perception by the VirA/VirG two-component regulatory system may play an important role in host defence (Zhang et al., 2000). In addition, DIMBOA-resistant Agrobacterium strains, such as Chry 9 and K289, should be used for the efficient transfer of desirable genes to maize or other cereals (Mohamalawari et al., 2002).
The explant type, explant quality and source of the explant have been found to be correlated with successful reports on the Agrobacterium-mediated genetic transformation of cereals (reviewed by Repellin et al., 2001). For example, freshly isolated immature embryos with and without pre-treatment have been part of the majority of successful reports on the genetic transformation of cereals and are considered to be the best explant type (Table 1). Embryogenic callus derived from mature seeds has been reported to be the best explant for Agrobacterium-mediated transformation of rice as a result of active cell division (Hiei et al., 1994). The transformation of embryogenic callus was later extended to other species, such as wheat (Cheng et al., 1997, 2003). In sorghum, the source of the explant had a significant effect on the transformation rate (Zhao et al., 2000). Immature embryos from field-grown plants showed a higher transformation frequency than immature embryos from glasshouse-grown plants. In our laboratory, we have also found that immature embryos of barley collected during the summer had a significant positive effect in tissue culture, regeneration and transformation frequency. In comparison, immature embryos collected during the winter yielded a lower transformation frequency (Shrawat and Lörz, unpublished data).
The difference in the competence of Agrobacterium to infect a particular tissue, genotype or species has also been a major drawback in the genetic transformation of elite cultivars of cereals. For example, when McCormac et al. (1998) compared the T-DNA transfer efficiency in wheat between two Agrobacterium rhizogenes strains (LBA9402 and Ar2626) and two Agrobacterium tumefaciens strains (LBA4404 and EHA101), they found that only EHA101 facilitated T-DNA delivery successfully into wheat. In the majority of cereals, genetic transformation has been achieved using Agrobacterium strains LBA4404, EHA101 and their derivatives (EHA105 from EHA101, AGL0 and AGL1 from EHA101) (Cheng et al., 2004). The combination of a standard binary vector in a super-virulent strain (e.g. pIG121Hm in EHA101) and a super-binary vector in a regular strain (e.g. pTOK233 in LBA4404) has resulted in the successful transformation of rice (Hiei et al., 1994). As a result of the success in rice, identical or similar combinations were used for the genetic transformation of maize (Ishida et al., 1996; Negrotto et al., 2000; Zhao et al., 2001), barley (Tingay et al., 1997; Wu et al., 1998), sorghum (Zhao et al., 2000), wheat (Khanna and Daggard, 2002) and sugarcane (Arencibia et al., 1998). With some crops, such as maize and sorghum, an efficient transformation system was established only with super-binary vectors in LBA4404, whereas a standard binary vector in a super-virulent strain showed low transformation frequency even with improved co-culture conditions in maize (Frame et al., 2002). However, in some cases, successful transformation was achieved without these combinations (Enríquez-Obregón et al., 1999; Cheng et al., 2003). It has also been shown that the inclusion of the constitutively active virG mutant gene in a binary vector increases T-DNA delivery in both monocot and dicot species (Hansen et al., 1994; Ke et al., 2001). These studies indicate that the use of other strains with a combination of the super-binary or binary vectors containing a constitutively active virG gene may further improve the transformation efficiency in cereal crops.
In recent years, efforts have also been made to understand the interactions of host plants with Agrobacterium at the molecular level (Ditt et al., 2001; Hongmei Jiang et al., 2003; Veena et al., 2003). To identify the plant genes involved in Agrobacterium infection, Nam et al. (1999) screened T-DNA insertion mutant lines of Arabidopsis for recalcitrance to transformation following bacterial infection, and found a large degree of variation in transformation amongst ecotypes. This study suggested that many plant genes may be involved in this process, and therefore screening of such mutagenized lines may be an important tool to understand the role of host genes during interaction with Agrobacterium. Agrobacterium infection triggers changes in the gene expression pattern of host cells, inducing or repressing specific sets of plant genes (Ditt et al., 2001; Veena et al., 2003). Ditt et al. (2001) found the altered expression of a number of plant transcripts within host cells after 24 and 48 h of interaction with Agrobacterium. They also showed that proteins encoded by these genes had a putative role in plant signal transduction and the defence response. Using suppressive subtractive hybridization and DNA microarrays, Veena et al. (2003) identified numerous plant genes that were differentially expressed during the early stages of Agrobacterium-mediated transformation. The genes identified in this study included those involved in defence responses, cell division and growth, chaperones, and primary and secondary metabolism. The majority of these genes showed induced expression during the early stages of infection with various strains of Agrobacterium. This study also demonstrated the involvement of T-DNA and/or Vir proteins as factors that resulted in the differential expression of these genes during Agrobacterium infection. However, in comparison with Ditt et al. (2001), who investigated late (24 and 48 h) times of infection and utilized only a transfer-competent Agrobacterium strain, Veena et al. (2003) showed that the expression of defence response genes was significantly decreased during the later stages of infection in plant cells infected with transfer-competent strains of Agrobacterium. Interestingly, cells infected with transfer-deficient Agrobacterium strains showed significant re-induction of these genes during the later stages of transformation. These results suggest that the suppression of the host defence response is a prerequisite to successful plant transformation.
Recently, Hwang and Gelvin (2004) have identified four Arabidopsis proteins that interact with the main T-pilus protein, VirB2, and have shown that the presence of these proteins is required for efficient transformation. Increased plant susceptibility to Agrobacterium infection by over-expression of the Arabidopsis nuclear protein VIP1 was demonstrated by Tzfira et al. (2002). Their results indicated that VIP1, which specifically interacts with VirE2 (Tzfira et al., 2001), renders the host plants significantly more susceptible to transient and stable genetic transformation by Agrobacterium, probably as a result of increased nuclear import of the transferred DNA. Therefore, to better understand the underlying process accounting for the host range and susceptibility of plant cells to Agrobacterium infection, it is necessary to identify more plant factors participating in T-DNA transformation. The molecular events that occur within Agrobacterium during interactions with host plants have been studied extensively, and we now have a reasonable understanding of the key factors involved in the regulation of T-DNA nuclear import and genomic integration (Zupan et al., 2000; Gelvin, 2003). By contrast, very little is known about the events that take place in the host cells during genetic transformation by Agrobacterium (Tzfira and Citovsky, 2002). Recent studies have suggested that Agrobacterium infection induces the plant genes necessary for the transformation process (Ditt et al., 2001; Veena et al., 2003). However, a number of questions remain: (i) how do host cells recognize and permit the integration of T-DNA into their genome?; (ii) how does the T-DNA complex reach and recognize the integration site?; (iii) what are the roles of host proteins that uncoat the T-DNA complex?; and (iv) why do some species or some cultivars of a species become infected with Agrobacterium and others do not?
Recent studies have also shown that virus-based vectors can be efficiently used for high levels of transient expression of foreign proteins in transfected plants, and permit non-Agrobacterium bacterial species to be employed for the production of transgenic plants (reviewed by Chung et al., 2006). Viral vectors hold great promise as efficient tools for transient recombinant protein expression in plant cells because of their ability to replicate in host cells autonomously (Marillonnet et al., 2004, 2005). These viral vectors are built on the backbones of plus-sense RNA viruses, such as tobacco mosaic virus (TMV) or potato virus, and have been used for the expression of foreign sequences in plants (Porta and Lomonossoff, 2002; Gleba et al., 2004). Marillonnet et al. (2004) demonstrated that TMV-based vectors can also be delivered to plant tissues using Agrobacterium tumefaciens. Interestingly, in this system, infection of plant cells with recombinant plus-sense RNA viral vectors occurs through a recombinant cDNA molecule encoded by Agrobacterium T-DNAs, which are delivered to plant cells by agroinfiltration. However, this system resulted in the low infectivity of cells. With the aim of improving the infectivity of the virus, Marillonnet et al. (2005) genetically modified the TMV-based vector by removing some of the putative cryptic splice sites and increasing the GC content of the TMV viral vector sequence through silent mutations. These genetic modifications not only resulted in efficient processing of the DNA into active replicons in almost all cells (as high as 94%) of Nicotiana benthamiana, but also optimized the viral vector such that it was delivered by the T-DNA of Agrobacterium more efficiently. This report suggests that Agrobacterium can be used for the transient production of recombinant proteins in plant cells. In a landmark report, Broothaerts et al. (2005) demonstrated that several species of bacteria outside of the Agrobacterium genus could be modified to mediate gene transfer to a number of diverse plants including rice. Plant-associated symbiotic bacteria (Rhizobium sp. NGR234, the alfalfa symbiont Sinorhizobium meliloti and Mesorhizobium loti of Phyllobacteriaceae) were made competent by the acquisition of both a disarmed Ti plasmid (pEHA105) from a hypervirulent Agrobacterium strain (Hood et al., 1993) and a binary vector containing the hygromycin resistance gene (aph) for plant selection. Regenerated plants showed GUS activity in their leaves and Southern blot analysis of transgenic plants revealed the integration of T-DNA at one to three sites per plant. Sequence analysis also showed integration at independent loci. Moreover, the analysis of the seed progeny of all three plant species transformed with Sinorhizobium meliloti showed stable inheritance of the transgenic GUS and hygromycin resistance phenotypes. This study suggests that a number of diverse plant-associated bacteria, when harbouring a Ti plasmid and binary vector, can be successfully used for gene transfer to plants.
We believe that the extension of the range of transformable genotypes and explants with a better understanding of host–pathogen interactions, the development of methods for checking and/or minimizing necrosis in transformed tissue and the identification of plant proteins involved in the facilitation of T-DNA delivery into the host genome may improve the Agrobacterium-mediated transformation of cereals and provide further scope for its applicability and general efficiency. In addition, in-depth analysis of the plant genes involved in Agrobacterium-mediated transformation will contribute to a better understanding of molecular events, such as cell communication, intracellular molecular transport, DNA repair and recombination, that occur during the process of Agrobacterium-mediated transformation, and may help to broaden the range of elite cultivars with improved transformation efficiency of economically important crops. The development of new technologies for the transient and stable expression of recombinant proteins in plants using viruses and other bacteria outside of the Agrobacterium host range will be of great value in plant science and will promote further research and development in this area.
A.K. Shrawat wishes to thank the Alexander von Humboldt Foundation of Germany for the award of a long-term Alexander von Humboldt Research Fellowship. The authors also wish to thank Drs Dirk Becker, Stephanie Luetticke, Melissa Hills and Mary Depauw for valuable suggestions and critical reading of the manuscript. We apologize to colleagues whose original works might not have been cited owing to a lack of space.