We describe a simple and efficient homology-based excision method to delete plastid genes. The procedure allows one or more adjacent plastid genes to be deleted without the retention of a marker gene. We used aadA-based transformation to duplicate a 649 bp region of plastid DNA corresponding to the atpB promoter region. Efficient recombination between atpB repeats deletes the intervening foreign genes and 1984 bp of plastid DNA (co-ordinates 57 424–59 317) containing the rbcL gene. Only five foreign bases are present in ΔrbcL plants illustrating the precision of homology-based excision. Sequence analysis of non-functional rbcL-related sequences in ΔrbcL plants indicated an extra-plastidic origin. Mutant ΔrbcL plants were heterotrophic, pale-green and contained round plastids with reduced amounts of thylakoids. Restoration of autotrophy and leaf pigmentation following aadA-based transformation with the wild-type rbcL gene ruled out mutations in other genes. Excision and re-use of aadA shows that, despite the multiplicity of plastid genomes, homology-based excision ensures complete removal of functional aadA genes. Rescue of the ΔrbcL mutation and autotrophic growth stabilizes transgenic plastids in heteroplasmic transformants following antibiotic withdrawal, enhancing the overall efficiency of plastid transformation. Unlike the available set of homoplasmic knockout mutants in 25 plastid genes, the rbcL deletion mutant isolated here is readily transformed with the efficient aadA marker gene. This improvement in deletion design facilitates advanced studies that require the isolation of double mutants in distant plastid genes and the replacement of the deleted locus with site-directed mutant alleles and is not easily achieved using other methods.
The simplicity and precision of homology-based excision, together with the predominance of homologous recombination in plastids, makes it a particularly attractive tool for manipulating plastid genomes. Use of native plastid enzymes avoids the extra transformation steps needed to introduce foreign site-specific recombinases, unintended site-specific recombinase-promoted rearrangements and the retention of site-specific recombinase target sites in engineered plastid genomes (Corneille et al., 2001, 2003; Hajdukiewicz et al., 2001; Lutz et al., 2004, 2006). Homology-based excision has allowed the removal of marker genes from transgenic plastids (Fischer et al., 1996; Iamtham and Day, 2000; Klaus et al., 2004). However, the potential of exploiting native plastid recombination pathways to manipulate plastid DNA has not been fully explored. To improve the design of deletion mutants to facilitate advanced studies on plastid genes, we examine whether homology-based excision can be used to make precise deletions in plastid DNA without a requirement for a marker gene to select mutant plastids.
Loss-of-function plastid mutants resulting from precise targeted deletion of plastid genes without the insertion of marker genes have not been described in flowering plants. All described knockout mutations in 29 plastid genes contain targeted insertions of the aadA marker gene used to select transformed plastids (reviewed by Maliga, 2005). Marker-free plastid mutants would provide new opportunities and advantages compared to the use of conventional aadA insertion mutants and wild-type (WT) plants in transplastomic research. They can be transformed with aadA, which is the most widely used plastid transformation marker gene (Svab and Maliga, 1993) due to the relative inefficiency of alternative markers (Carrer et al., 1993; Huang et al., 2002; Svab et al., 1990). When a modified plastid gene containing a site-directed mutation is transformed into a deletion mutant, all transformants will contain the modified allele whose phenotype will not be masked by the resident wild-type allele. The introduction of site-directed mutations into WT plastids can be problematic due to undesirable recombination events in the region between marker and mutation (Andrews and Whitney, 2003; Whitney et al., 1999). Marker-free mutants facilitate the stepwise introduction of mutations into the plastid genome to isolate plants with mutations in several plastid genes. Co-transformation with two plasmids provides a less predictable method for integrating changes at multiple sites within the plastid genome (Carrer and Maliga, 1995; Ye et al., 2003). Efficient aadA-based antibiotic selection combined with mutant rescue using the wild-type plastid allele (Klaus et al., 2003) enables rapid identification of plastid transformants and stabilizes recombinant genomes in transformed plants. This provides the basis for high-throughput aadA-based transformation of foreign trait genes into plastids. Pleiotropic effects of marker gene expression on transcription or protein accumulation within mutant plastids are removed, and marker-free plastid mutants are likely to satisfy regulatory approval for widespread dissemination and use of plastid mutants within the research community.
Here we target the rbcL gene, which encodes the large subunit (LS) of ribulose bisphosphate carboxylase/oxygenase (RuBisCO), the key enzyme in photosynthetic carbon assimilation (Andrews and Whitney, 2003; Parry et al., 2003). Our procedure is simple, can be used to delete one or more adjacent plastid genes, and is applicable to non-essential plastid genes, whose mutation allows growth on media containing sucrose.
Plastid transformation vectors and experimental design
The approach we used to delete the rbcL gene is illustrated in Figure 1(a). In vector pUM83, the aadA marker gene is placed downstream of the rbcL gene. Sequences upstream of rbcL are duplicated and placed downstream of rbcL and aadA. This creates a direct repeat flanking rbcL and aadA. Recombination between these direct repeats excises the rbcL and aadA genes in one event. The advantage of this strategy is that all mutant ΔrbcL cells will also be free of the aadA marker gene. This means that a visual screen for plastid mutants, which often have a pigment-deficient phenotype (Kanevski and Maliga, 1994; Klaus et al., 2003; Swiatek et al., 2003), will also identify aadA-free plants.
Vector pUM83 (Figure 1a) contains converging expression cassettes with the aadA marker (Goldschmidt-Clermont, 1991) and uidA reporter genes (Jefferson, 1987) inserted into the AocI site located in the intergenic region between the rbcL and accD genes (Wakasugi et al., 2001). The 5′atpB regulatory region drives expression of uidA and is comprised of 649 bases including the ATG initiation codon. This creates two 649 bp direct repeats, DR1 and DR2 (Figure 1a), comprised of 5′atpB regulatory sequences that flank the rbcL gene. The 5.7 kbp left and 2.1 kbp right plastid (pt) DNA sequences bordering aadA and uidA target integration by homologous recombination following plastid transformation. Transgenic plastid genomes are then selected with antibiotics to replace WT plastids. A recombination event between the 649 bp direct repeats will excise the rbcL, aadA and uidA genes as a 6.1 kbp circle with a single HindIII site, leaving a 153 kbp marker-free ΔrbcL plastid genome (Figure 1a). The 6.1 kbp circle lacks sequences necessary for stable maintenance as an episome and is lost. In the absence of selection, marker-free ΔrbcL plastid genomes will accumulate, leading to the isolation of marker-free mutant plants.
Isolation of pUM83 (T0) plastid transformants selected with antibiotics
Plastid transformation was carried out using particle bombardment on leaf explants (Kode et al., 2005). Green resistant shoots were taken through three rounds of regeneration on media containing spectinomycin and streptomycin. Of seven antibiotic-resistant clones isolated, three stained positive for GUS using X-gluc (not shown), indicating the presence of an active uidA gene. Two of these GUS-positive clones (8a and 8c) were taken forward for further analysis. DNA blot analysis indicated replacement of WT ptDNA with transgenic plastid genomes (see below). Seeds were collected from selfed T0 pUM83 transplastomic plants.
Isolation of ΔrbcL mutant plants (T1 generation)
Seeds from pUM83-8a and -8c transplastomic parents were germinated on MS medium containing spectinomycin (500 mg l−1) or MS medium without antibiotics. All of the 30 pUM83-8a (T1) seedlings were green and resistant to spectinomycin (not shown). Two white sensitive seedlings were found amongst the 30 T1 progeny of pUM83-8c grown on spectinomycin-containing medium (Figure 2a) and resembled spectinomycin-bleached WT seedlings. On antibiotic-free medium, three pale-green pUM83-8c seedlings (Figure 2b) were found amongst 30 T1 seedlings and were propagated in vitro for further analysis. On further growth on MS medium without antibiotics, half of the T1 seedlings from pUM83-8a and -8c plastid transformants gave rise to plants with one or more true leaves containing pale-green sectors (Figure 2c). WT plants do not contain pale-green sectors (Figure 2d). T2 seeds were collected from variegated T1 plants, allowing long-term storage of the mutant (Figure 2e,f).
A pale-green phenotype is characteristic of mutations in some photosynthesis-related genes (Klaus et al., 2003; Swiatek et al., 2003) including rbcL (Kanevski and Maliga, 1994). Sensitivity to spectinomycin and a pale-green phenotype would result from excision of the aadA and rbcL genes (Figure 1). To purify pale-green mutant plants, three rounds of regeneration were performed on media lacking antibiotics using either leaves from pale-green plants (clone 8c) or pale-green sectors from variegated plants (clone 8a). Pale-green mutant plants (Figure 2g) were propagated by sub-culturing shoots on MS medium supplemented with 3% w/v sucrose.
PCR analysis of ΔrbcL mutant plants and sequencing confirms the precision of homologous recombination
Recombination within the duplicated atpB promoter region excises rbcL and the foreign genes. The precision of the recombination event was analysed by PCR using primers pt F and pt R (Figure 3c). The primers amplified a 3.2 kbp band from WT DNA (Figure 3a) and a 1.3 kbp band from pale-green ΔrbcL-8a plant DNA (Figure 3a). The absence of a WT 3.2 kbp band in ΔrbcL lanes is consistent with the absence of WT ptDNA but might also reflect preferential amplification of the smaller 1.3 kbp band. The 1.3 kbp amplification product from ΔrbcL plants was purified and sequenced. The sequence (Figure 3c) confirmed that the 1.3 kbp PCR product was the result of recombination between the 649 bp direct repeats (Figure 1a). No changes were found in the recombined copy of the 649 bp atpB 5′ region remaining in ΔrbcL plants, demonstrating a perfect homologous recombination event between the two repeats. An introduced NotI site spans the junction between the deletion end points located at bases 57 423 and 59 318. The ΔrbcL plastid genome contains only 5 bp of foreign sequence (GGCCG) corresponding to internal bases in the NotI site (Figure 3c). The ΔrbcL plants should lack the aadA marker gene. Primers directed against aadA (see locations in Figure 1) amplified a band of the expected size from a uniformly dark-green pUM83-8a (T0) plant DNA but not from WT or pale-green ΔrbcL-8a extracts (Figure 3b).
DNA blot analyses of ΔrbcL mutant plants
Total DNA was extracted from WT plants, a uniformly dark-green pUM83-8a (T0) plant and two ΔrbcL mutant clones (8a, 8c) from independent pUM83 transformants 8a and 8c. The locations of rbcL, accD and aadA hybridization probes are shown in Figure 1(a). An rbcL probe hybridizes to an 11.5 kbp HindIII band in digests of WT DNA (Figure 4a, lane 1). This is replaced by two bands of 7.0 and 6.1 kbp in digests of leaf DNA from a T0 pUM83-8a plant (Figure 4a, lane 2). The 7.0 kbp band corresponds to the recombinant plastid genome resulting from integration of aadA and uidA expression cassettes (Figure 1a). The less intense 6.1 kbp HindIII band corresponds in size to the predicted excision product containing foreign genes and rbcL that has a single HindIII site (Figure 1a) and also hybridizes to aadA (Figure 4d, lane 2). A discrete low molecular weight band of uncut DNA from a T0 pUM83-8a plant (Figure 4b, lane 2), located well below the shear size of the bulk of DNA (>20 kbp), might correspond to the 6.1 kbp excision product. No low molecular weight species were found in other lanes (Figure 4b).
No rbcL hybridization was expected to digests of ΔrbcL plant DNA, but we found a weakly hybridizing 11.5 HindIII band (Figure 4a, lanes 3 and 4) that co-migrates with the strong WT band (Figure 4, lane 1). This faint 11.5 kbp band is unlikely to represent residual copies of WT plastid genomes because WT ptDNA in mutant plants would be expected to give rise to green sectors. No dark-green sectors have been observed in ΔrbcL mutant plants propagated for three years on MS media. This indicates that the rbcL related sequences in ΔrbcL plants are not functional. Primers pt F1 and pt R (locations shown in Figure 3c) should amplify residual copies of any WT DNA present in ΔrbcL plants because primer pt F1 binds to the end of the rbcL gene, which is deleted in ΔrbcL plants. A 0.8 kbp PCR product amplified from ΔrbcL DNA contained seven base substitutions and six insertion/deletion events in a 386 bp region when compared to WT ptDNA (Figure 5). This sequence divergence is typical of plastid DNA sequences transposed to the nucleus (Huang et al., 2005). No WT rbcL sequences were detected in ΔrbcL plants.
Hybridization of the accD probe to HindIII digests revealed an 11.5 kbp WT band (Figure 4c, lane 1), an 8.7 kbp band in pUM83 transformants (Figure 4c, lane 2), and a 9.6 kbp band in ΔrbcL plants (Figure 4c, lanes 3 and 4) as expected (see Figure 1a). An aadA probe hybridizes to expected bands of 8.7 and 6.1 kbp in pUM83 transformants (Figure 4d, lane 2) but did not hybridize to digests of WT and ΔrbcL DNA (Figure 4d, lanes 1, 3, 4); refer to maps in Figure 1(a) for sizes. Blot analyses confirm the isolation of aadA-free ΔrbcL mutant plants.
Phenotype of ΔrbcL plants
Sections of pale-green ΔrbcL leaves contained a predominance of circular plastids (pt) in which stroma lamellae (L) and grana (Gr) were clearly visible (Figure 2h). Much of the stroma was devoid of grana and stroma lamellae, unlike the typical lens-shaped chloroplasts found in leaf mesophyll cells in WT plants (see examples of WT plastids in Kode et al., 2005; Swiatek et al., 2003). Shoots formed on RMOP regeneration medium supplemented with 3% w/v sucrose were transferred to RMOP medium lacking sucrose. WT shoots continued to grow on RMOP medium lacking sucrose while ΔrbcL shoots from clone 8a died (Figure 2i). This indicates that mutant ΔrbcL plants are heterotrophic and unable to grow on media lacking sucrose. The large subunit (LS) of RubisCO was not detected in ΔrbcL mutant plants by Western blot analysis (not shown).
Repair of the lesion in ΔrbcL mutant plants by transformation with the WTgene
Transformation of ΔrbcL plants with plastid transformation vector pUM79 introduces aadA and the WT rbcL allele into mutant plastids (Figure 1b). The HindIII restriction patterns of recombinant ptDNA in pUM79 and pUM83 transformants are different (see Figure 1a,b). This allows us to distinguish pUM79 transformants from the primary pUM83 transformants from which the ΔrbcL plants, used as recipients of transformation, were derived. Spectinomycin-resistant shoots with either dark-green or pale-green leaves were isolated 4–8 weeks after transformation. PCR analyses with aadA primers indicated the presence of aadA in dark-green leaves and its absence in pale-green leaves (not shown). The pale-green resistant plants resembled the colour of the ΔrbcL recipient used for transformation and were likely to be spontaneous spectinomycin-resistant mutants, a common problem with spectinomycin-based selection (Klaus et al., 2003). Shoots were transferred to Magenta jars for rooting without further cycles of regeneration. Dark-green clones (Figure 2k) stained blue with X-Gluc (Figure 2l), indicative of the presence of the uidA gene, whereas the pale-green resistant shoots and recipient ΔrbcL-8a plant (Figure 2g) used for transformation were GUS-negative (Figure 2j). Leaves from dark-green pUM79 transformants contained normal levels of LS RubisCO (not shown). HindIII digests of pUM79 transformant DNA gave 7.0 kbp rbcL (Figure 4a, lanes 5 and 6) and 4.6 kbp accD bands (Figure 4c, lanes 5 and6), compatible with the integration pattern of pUM79 (Figure 1b). The absence of a 9.6 kbp HindIII accD band, diagnostic of ΔrbcL plants (Figure 4c, lanes 3 and 4) in pUM79 transformants (Figure 4c, lanes 5 and 6) indicates that the ΔrbcL plastid genomes have been replaced with a homoplasmic population of transgenic pUM79 plastid genomes. An aadA probe hybridized to the expected 4.1 kbp HindIII band in digests of DNA from pUM79 transformant clone 1 (Figure 4d, lane 5).
We have described a precise and simple method for deleting the plastid rbcL gene following a single transformation step. The precision of homology-based excision is demonstrated by the presence of only five foreign bases (GGCCG) in the ΔrbcL plastid genome described here. The deletion (co-ordinates 57 424–59 317; Wakasugi et al., 2001) removes the 1434 bp rbcL coding region, 171 bp of upstream and 289 bp of downstream sequences. This makes the mutant suitable for plastid transformation with vectors containing changes in the rbcL coding sequence for structure–function studies (Andrews and Whitney, 2003; Parry et al., 2003; Spreitzer et al., 2005) and in the regulatory region (Shiina et al., 1998) to modify rbcL expression levels. Rescue of the mutant phenotype by the WT rbcL allele following aadA-based plastid transformation rules out mutations at other sites.
A relatively long duplication of 649 bp separated by 5.4 kbp was used because excision was barely detected between two copies of a 418 bp psbA 3′ regulatory region spaced 1 kbp apart (Iamtham and Day, 2000). The influence of sequence composition on rates of plastid recombination is not known. High levels of sectoring in leaves of T1 plants indicated efficient excision of the rbcL gene. Excision products were also detected in the T0 generation by DNA blot analyses. This indicates that it should be possible to visualize mutant sectors in leaves of T0 plants following vegetative propagation and additional cycles of regeneration on non-selective medium. Cytoplasmic sorting is an essential component of the method because it drives the segregation of different plastid types and leads to the isolation of homoplasmic mutants in the absence of selection. Use of site-specific recombinases is less reliant on cytoplasmic sorting because, in principle, they allow simultaneous excision of genes from multiple plastids (Kuroda and Maliga, 2003).
Deletion of native plastid genes and foreign genes in one recombination event simplifies the isolation of plastid mutants lacking foreign genes. The method has the added advantage of allowing any two regions of ptDNA to be spliced together by recombination and reduces the reliance on the proximity of convenient restriction sites to delete plastid genes. The approach also allows the deletion of multiple adjacent plastid genes. A duplication of the region downstream of the atpE gene would delete the rbcL, atpB and atpE genes in one step. Deletion of plastid genes that do not confer a pigment-deficient phenotype, such as the ndh genes (Burrows et al., 1998; Kofer et al., 1998) would require PCR-based screening of seedlings or shoots from vegetatively propagated plants. Only regions of the plastid genome lacking essential genes can be deleted by this approach. In tobacco, essential genes include clpP1 (Kuroda and Maliga, 2003; Shikanai et al., 2001), ycf1, ycf2 (Drescher et al., 2000) and accD (Kode et al., 2005) as well as plastid-encoded components of the plastid translation machinery such as rps14 (Ahlert et al., 2003) that are needed to express these genes.
Restoration of photosynthesis following aadA-based plastid transformation with the WT rbcL allele helps to maintain recombinant plastid genomes when antibiotic selection is removed in soil-grown plants. When WT plants are transformed, residual heteroplasmy in soil-grown plants can be a problem if stochastic fluctuations or selective forces lead to a reduction in transgenic plastids relative to WT plastids. One cycle of regeneration usually suffices to isolate homoplasmic transplastomic plants following mutant rescue in agreement with the results of Klaus et al. (2003). We were unable to use restoration of photosynthesis as the sole criterion for selecting plastid transformants (results not shown). The reasons for this failure are unclear but others have also had little success with this approach (Klaus et al., 2004).
Vectors were made by standard cloning techniques (Sambrook et al., 1989). The atpB 5′ regulatory region including the ATG initiation codon were amplified with primers atpB-NotI-F 5′-CCGCGGCCGCCAAATACATCATTATTGTATAC (underlined complement of 57 423 to 57 400 bp) in 155 939 bp tobacco plastome (accession number Z00044) and atpB-NcoI-R CCCCATGGACATAATAATAAAATAAATAAATATG (underlined positions 56 775 to 56 799 bp). The amplified product was cloned into the EcoRV site of pBluescript to make pBS-5′NtatpB and sequenced to ensure the absence of PCR mutations. A uidA coding sequence terminated by a 3′ NtpsbA regulatory element (Iamtham and Day, 2000) and flanked by NcoI and NotI–SacI sites was excised as a NcoI–SacI fragment and ligated to NcoI- plus SacI-digested pBS-5′NtatpB. The 5′ NtatpB-uidA-3′NtpsbA expression cassette was then excised with NotI and cloned in inverted orientation with respect to a 16SrrnBn-aadA-BnpsbC cassette (Zubko et al., 2004) into the NotI site of pTB27-link (Zubko et al., 2004) to make pUM83. In the transformation vector pUM79, the NtatpB-uidA-NtpsbA expression cassette is replaced with a CratpA-uidA-NtpsbA cassette, where CratpA is the 5′ regulatory region present in a 680 bp EcoRI–NcoI fragment from pUC-atpX (Goldschmidt-Clermont, 1991; accession number J01399 bases 259–911). Accession numbers are AJ276677 (16SrrnBn), AJ578474 (BnpsbC), X02340 (aadA, modified by Goldschmidt-Clermont, 1991), U12369 (uidA).
Isolation of transplastomic plants
Plastid transformants were isolated following particle bombardment as described previously (Kode et al., 2005). Following bombardment of WT leaves, resistant shoots were selected on RMOP medium (Svab and Maliga, 1993) containing spectinomycin dihydrochloride pentahydrate plus streptomycin sulphate (Melford Lab Ltd, Chelsworth, UK), each at 500 mg l−1. Following three cycles of regeneration on RMOP medium with both antibiotics, shoots were rooted on Murashige and Skoog (MS) medium containing 200 mg l−1 spectinomycin. When ΔrbcL leaves were used for transformation, resistant shoots were isolated on RMOP medium containing spectinomycin (500 mg l−1) and transferred for rooting on MS medium with spectinomycin (200 mg l−1) or MS medium lacking antibiotics after one cycle of regeneration. All solid RMOP medium was prepared with 0.7–0.8% w/v agar. Either 0.7% agar or 0.25% Phytagel (Sigma-Aldrich, Poole, UK) was used to solidify MS media for rooting. Leaf pieces were stained in X-Gluc to monitor GUS expression as described previously (Jefferson, 1987).
Total DNA was prepared using the Wizard genomic plant DNA purification kit (Promega, Madison, WI, USA). DNA blot analyses procedures, hybridization probes, PCR analyses using purified DNA or crude DNA preparations as template DNA with ReadyMix Taq containing MgCl2 (Sigma-Aldrich, Poole, UK) were performed as described previously (Kode et al., 2005). Oligonucleotides (Sigma-Aldrich, Haverhill, UK) used for primers were (F, forward; R, reverse) pt F 5′-CAACACTATCTCGACCTTGA-3′ (tobacco plastid map locations 56 612–56 631), pt R 5′-GCATGAAAATACAATAGATGAATAG-3′ (complement 59 797–59 821), pt F1 5′-GCAGTGGACGTTTTGGATAAG-3′ (59 005–59 025), aadA-F 5′-ATTCTCCGCGCTGTAGAAGTCACC-3′ and aadA-R 5′-TACATTTCGCTCATCGCCAGCC-3′. PCR products and restriction fragments fractionated on 0.8–2% agarose gels were purified using the Perfectprep Gel Cleanup kit (Eppendorf AG, Hamburg, Germany). Purified PCR products and plasmids were sequenced from appropriate primers by cycle sequencing using the Big Dye terminator sequencing ready reaction kit (Applied Biosystems, Warrington, UK). DNA sequences were analysed using the Vector NTI version 9 suite of DNA analysis programs (Invitrogen, Carlsbad, CA, USA).
Sections were prepared as described previously (Kode et al., 2005) for transmission electron microscopy using an FEI Tecnai 12 Biotwin transmission electron microscope (FEI company, Eindhoven, The Netherlands).
Supported by Overseas Research Students awards to V. Kode and S. Iamtham, a University Research Studentship award to V. Kode and a Royal Thai Government Scholarship to S. Iamtham. We thank the electron microscopy and DNA sequencing facilities for assistance.