Incorporation of a selectable marker gene during transformation is essential to obtain transformed plastids. However, once transformation is accomplished, having the marker gene becomes undesirable. Here we report on adapting the P1 bacteriophage CRE-lox site-specific recombination system for the elimination of marker genes from the plastid genome. The system was tested by the elimination of a negative selectable marker, codA, which is flanked by two directly oriented lox sites (>codA>). Highly efficient elimination of >codA> was triggered by introduction of a nuclear-encoded plastid-targeted CRE by Agrobacterium transformation or via pollen. Excision of >codA> in tissue culture cells was frequently accompanied by a large deletion of a plastid genome segment which includes the tRNA-ValUAC gene. However, the large deletions were absent when cre was introduced by pollination. Thus pollination is our preferred protocol for the introduction of cre. Removal of the >codA> coding region occurred at a dramatic speed, in striking contrast to the slow and gradual build-up of transgenic copies during plastid transformation. The nuclear cre gene could subsequently be removed by segregation in the seed progeny. The modified CRE-lox system described here will be a highly efficient tool to obtain marker-free transplastomic plants.
Transformation of the plastid genome is a new and attractive alternative to engineering the nuclear genome. Applications of plastid transformation include the improvement of agronomic characteristics of crops, and the production of recombinant proteins. Useful traits expressed in chloroplasts are the Bacillus thuringiensis insecticidal protein (McBride et al., 1995); herbicide resistance (Daniell et al., 1998; Lutz et al., 2001; Ye et al., 2001); and expression of human somatotropin (Staub et al., 2000). Plastid transformation vectors contain a selective marker, most commonly a spectinomycin resistance (aadA) gene (Svab and Maliga, 1993), flanked by plastid DNA sequences targeting insertion of the marker gene by homologous recombination into the plastid genome. Genes of commercial value but lacking a selectable phenotype are physically linked to the selective marker, and the two genes are integrated together as a block of heterologous sequences (Maliga, 1993).
Given the high number (1000–10 000) of plastid genome copies per cell, incorporation of a selectable marker gene during transformation is essential to obtain uniformly transformed plastids. However, once transformation is accomplished the marker gene becomes undesirable. Possible problems associated with plastid marker genes are the metabolic burden imposed by high-level expression (>10% of total soluble cellular protein; Khan and Maliga, 1999) and the potential for the unlikely event of horizontal transfer to microbes (Dröge et al., 1998; Schluter et al., 1995; Sylvanen, 1999).
The CRE-lox site-specific recombination system of bacteriophage P1 has been well characterized (Adams et al., 1992; Craig, 1988; Guo et al., 1997). Expression of the CRE protein (38.5 kDa) is sufficient to cause recombination between lox sites, a 34 bp region consisting of two 13 bp inverted repeats separated by an 8 bp asymmetrical spacer sequence. If the lox sites are in the same orientation as a direct repeat, recombination results in deletion of the intervening DNA. The CRE-lox site-specific recombination system was shown to be useful for elimination of nuclear genes in higher plants (Dale and Ow, 1991; Russell et al., 1992; Srivastava et al., 1999).
The objective of this study was to assess the feasibility of using the CRE site-specific recombinase for the elimination of marker genes from the plastid genome. As a test system, we studied excision of codA, a bacterial gene encoding cytosine deaminase (CD; EC 126.96.36.199), which is absent in plants. Expression of codA in plastids made tobacco cells sensitive to 5-fluorocytosine (5FC; Serino and Maliga, 1997). Thus 5FC resistance could be used for positive identification of cells with CRE-induced codA deletion, even if the deletion events were relatively rare. A test system was set up by incorporating a codA gene in the tobacco plastid genome between two directly oriented lox sites (>codA>). We report here that the transplastomes were stable in the absence of CRE activity. However, introduction of a nuclear-encoded, plastid-targeted CRE recombinase by Agrobacterium transformation or pollination triggered highly efficient elimination of >codA>. Introduction of cre by transformation, but not by pollination, was accompanied by a large deletion in the plastid genome including the trnV gene.
Construction of transgenic plants
Plastid transformation vector pSAC48 carries a codA coding region flanked by two lox sites in direct orientation (>codA>) (Figure 1a). Removal of the >codA> coding region located between the lox sites would yield one lox site flanked by the promoter (Prrn) and the terminator (TrbcL). Transformation with plasmid pSAC48 yielded the Nt-pSAC48-16C and Nt-pSAC48-21A transplastomic lines. The two lines were obtained independently, but are otherwise identical.
To ensure plastid targeting of the CRE recombinase, the cre coding region was translationally fused with the Rubisco small subunit (SSU) transit peptide. Two chimeric cre genes were prepared: one with 5 (cre1 gene) and one with 22 (cre2 gene) amino acids of the mature SSU N-terminus. The cre genes were expressed in P2′ promoter and Tnos terminator cassettes in the Agrobacterium binary vector pPZP212, which carries kanamycin resistance (neo) as a selectable marker (Figure 2).
Agrobacterium transformation for CRE-mediated excision of >codA> in somatic cells
Deletion of the plastid >codA> gene was activated by expression of an engineered cre gene introduced into the nucleus of the transplastomic lines by Agrobacterium-mediated transformation. Plants representing 11 individual cre transformation events have been characterized. Five lines were obtained by transformation with cre1 (Cre1-derivatives) and six lines were obtained by transformation with cre2 (Cre2-derivatives). Excision of >codA> in each of the 11 lines (five Cre1 and six Cre2 derivatives) was tested by gel-blot analysis of DNA extracted from leaves (Figure 1b). No codA signal was detected in any of the cre-transformed lines. Three lines (Cre1-3, Cre2-4, Cre2-5) carried plastid genome populations which were the expected products of CRE-mediated >codA> excision indicated by a 3.6 kb fragment hybridizing with both the plastid targeting region and aadA probes (Figure 1). Four lines carried a mixed population of plastid genomes consisting of >codA> excision products (3.6 kb fragment) and wild-type (1.9 kb fragment; Cre2-1 line) or an unexpected deletion product (1.4 kb fragment; Cre1-1, Cre1-2, Cre2-10 lines; see below).
CRE-mediated >codA> excision is expected to leave behind a reconstituted lox site. The presence of a lox site between TrbcL and Prrn was confirmed in six lines by (i) PCR amplification of the aadA-plastid targeting region using oligonucleotides O1 and O2 (Figure 1a), and (ii) direct sequencing of the PCR product with both oligonucleotides (data not shown).
The plastid genome of four lines was not the predicted product of a CRE-mediated excision event. In two lines, Cre1-4 and Cre1-10, deletion of >codA> was accompanied by deletion of the aadA and the trnV (tRNA-ValGAC) genes by homologous recombination via the duplicated Prrn sequences. One Prrn promoter drives >codA> expression in vector pSAC48, the second tandem Prrn promoter is at its native location upstream of rrn16, the first gene of the rrn operon (Figure 1a). Smaller than wild-type (1.4 instead of 1.9 kb) fragments detected by the plastid targeting region probe in Figure 1(b) suggested deletion of the trnV-aadA-codA region. Deletion via the Prrn sequence was confirmed by PCR amplification of the region upstream of the rrn16 coding region using oligonucleotides O3 and O4 (Figure 1a) and direct sequencing of the amplified fragment from the four lines. Prrn DNA sequence in the transplastomic >codA> (Nt-pSAC48) plants and the region containing the recombination junction in the deletion derivatives is shown in Figure 3. Two of the sequenced lines were homoplastomic and two were heteroplastomic for the trnV deletion (Cre1-4, Cre1-10 and Cre1-1 and Cre1-2, respectively; Figure 1b). Two additional >codA> deletion derivatives were found in clones Cre2-2 and Cre2-3. The deletion events in these clones have not been studied.
The desired end products of our test system were tobacco plants carrying the plastid transgene of interest (aadA) and lacking all other plastid (codA) and nuclear (cre, neo) transgenes introduced during the construction phase. Such plants were readily obtained in the selfed seed progeny of the cre transformed tobacco lines. Parental plant Cre1-3 lacks >codA>, and carries aadA and a single lox site between Prrn and TrbcL. Thus the line has the predicted plastid genome generated by CRE-mediated excision of >codA> via the lox sites (see above). DNA gel-blot analysis of 20 seedlings grown from selfed Cre1-3 seed indicates that each of the seedlings has the same parental engineered plastid genome (Figure 4). The seedlings lack both cre and neo genes, and thus have the desired nuclear genetic constitution. Similar plants with CRE-mediated >codA> deletion in their plastid genomes that lack the nuclear cre and neo genes were obtained in the segregating seed progeny of the Cre2-4 plants (14 seedlings tested; data not shown). Lack of cre and the linked neo genes in the Cre1-3 and Cre2-4 seed progeny suggested selection against cre during sexual reproduction. However, seed transmission of cre (and the linked neo) was readily obtained in the Cre1-1 and Cre1-2 lines.
Plants lacking the trnV gene in their plastid genome could also be readily obtained in the Cre1-2 progeny (Figure 4). The parental plant was heteroplastomic for the deletion event (Figure 1). In the small seedling population tested, only one of the parental plastid genomes was recovered.
Introduction of cre by pollination to trigger excision of >codA> in germline cells
Transgenic cre plants were also obtained by transformation of wild-type tobacco plants with the cre1 and cre2 constructs to yield the Cre1-100, Cre2-100, Cre2-200 and Cre2-300 lines. These plants were used as the source of pollen in crosses with the Nt-pSAC48 plants as the maternal parent. The pollen parents were primary transgenic plants (T0) segregating for the nuclear cre and the linked neo genes. Introduction of cre via pollen was expected to trigger CRE-mediated excision of >codA>. To test for excision of >codA>, DNA gel-blot analysis was carried out on DNA isolated from the leaves of young seedlings. In seedlings derived from pollination with the Cre1-100 and Cre2-300 lines, when cre was present, >codA> excision was complete (Figure 5a). These lines were classified as strong activators, in contrast to lines Cre2-100 and Cre2-200, which were classified as weak activators because the seedlings contained a mixed plastid genome population with intact and excised >codA> copies (Figure 5b). It was striking that in the seedlings, whether derived from pollination with a weak or strong activator line, CRE-mediated excision was not accompanied by deletion of the trnV region.
The studies described above were carried out on DNA from seedlings germinated in the absence of 5FC. Seeds derived from the cross with the weak activator line Cre2-100 were also germinated in the presence of 5FC (100 mg l−1). Some of the seedlings were green (resistant) whereas others were pale, apparently being sensitive to 5FC (Serino and Maliga, 1997). When DNA was tested from the green seedlings grown on 5FC, it turned out that all the plastid genome copies lacked >codA> (six seedlings tested). In contrast, the seedlings germinated in the absence of 5FC contained an approximately equal number of plastid genome copies with and without >codA> (nine seedlings tested) (Figure 5c). Thus germination on 5FC facilitated the identification of seedlings in which elimination of >codA> was complete.
Probing of total cellular DNA with the codA probe and using DNA samples from the cross with the weak-activator pollen parents identified two unexpected, smaller bands (Cre2-100 progeny; Figure 6a). As the codA-specific bands were present only in progeny carrying the cre gene, we suspected that the bands could be circular monomeric and dimeric excision products. Probing of undigested DNA, or DNA digested with restriction enzymes (ApaI, EcoRV) for which there is no site in the circular excision product, yielded identical hybridizing bands. However, when digestion was carried out with NcoI, an enzyme for which there is a single restriction site in the monomeric circle, only one 1.3 kb band was found (Figure 6b). These observations may be explained by circular monomeric and dimeric excision products being present in the CRE-expressing plants which can be converted to unit-size (1.3 kb) linear molecules by digestion with NcoI. Such circular excision products were absent in the seedlings derived from crosses with strong activators Cre1-100 and Cre2-300. A similar series of mini-circles has been observed in transplastomic plants (Staub and Maliga, 1994; Staub and Maliga, 1995).
We report here on the use of the P1 phage CRE-lox system for the efficient removal of DNA segments from the plastid genome. The system has two key components, the lox sites flanking the target DNA in the plastid genome, and a nuclear-encoded, plastid-targeted CRE recombinase. We describe two alternative protocols which differ in the method of cre gene introduction into the nucleus (Figure 7). Excision in somatic cells was triggered by the introduction of cre using Agrobacterium-mediated transformation, whereas excision in germline cells was triggered by introduction of cre via pollen. Although both systems are efficient for the removal of target DNA, we prefer germline excision as, in this case, CRE activity does not lead to large deletions of the plastid genome. The high efficiency of marker gene deletion makes the plastid CRE-lox system a significant advancement over the earlier, homologous recombination-based system for removal of plastid marker genes (Fischer et al., 1996; Iamtham and Day, 2000).
As the plastid genetic system is highly polyploid (1000–10 000 genome copies per cell), uniform alteration of the plastid genome by transformation or mutagenesis is a lengthy process. To achieve the homoplastomic state takes at least 20 cell divisions and typically involves two cycles of plant regeneration on a selective medium (Maliga, 1993; Moll et al., 1990). In striking contrast, CRE-mediated excision of DNA sequences yields a uniform plastid genome population extremely rapidly in the absence of selection pressure, because of the presence of the lox sites in all genome copies allowing excision to take place simultaneously.
There is apparent bias against seed transmission in two of the lines (Cre1-3, Cre2-4), in which no cre-carrying progeny was obtained in a total of 34 selfed seed progeny in which three-quarters of the seedlings were expected to carry a nuclear cre gene. Lack of seed transmission is the property of the insertion events as seed transmission of cre was readily obtained in other independently transformed lines (Cre1-1, Cre1-2, Cre1-100, Cre2-100, Cre2-200 and Cre2-300). It is likely, therefore, that lack of seed transmission is due to toxicity caused by developmentally timed CRE expression (Schmidt et al., 2000). Pollen transmission of cre is important for the general applicability of the method, as two consecutive transformation events may not be carried out in certain plant species, for example cereals.
The recovery of plants lacking trnV was unexpected. Although the aadA >codA> transplastomes are stable in the absence of CRE activity, trnV deletion was relatively frequent (four out of 11 clones; Figure 1b) in the CRE-expressing plants. The large deletions in the plastid genome were limited to somatic cells, suggesting that the deletion events are the consequence of CRE interacting with tissue-specific host factors. Most of the deletion events resulted in the loss of the trnV gene (tRNA-ValGAC). The trnV gene (tRNA-ValGAC) is apparently non-essential, as homoplastomic deletion derivatives could be readily obtained. Also, the Cre1-4 and Cre1-10 plants have no obvious phenotype, and plastid genomes lacking tRNA-ValGAC have been transmitted to the seed progeny (Cre1-2 line; Figure 4). The plant's ability to function normally without tRNA-ValGAC may be explained by an additional trnV gene (tRNA-ValUAC) located in the large single-copy region of the plastid genome. This second gene may be sufficient for the translation of all valine codons in plastids due to increased wobble. Alternatively, obtaining normal plants lacking tRNA-ValGAC may be explained by tRNA import from the cytoplasm, a phenomenon common in plant mitochondria (Unseld et al., 1997).
Critical for CRE-mediated excision was targeting of CRE to plastids. Although CRE is a prokaryotic protein, it naturally carries a nuclear localization signal (Le et al., 1999). Translational fusion with the Rubisco small subunit transit peptide yielded a protein with an N-terminal plastid targeting signal and an internal (CRE) nuclear localization signal. When two potential targeting sequences are present, in general one of the subcellular localization signals outcompetes the other, with the N-terminal signal being dominant (Small et al., 1998). Efficient excision of >codA> via the lox sites indicates that both cre1 and cre2 polypeptides are targeted to plastids. Thus the size of the mature SSU segment (five or 22 amino acids) does not appear to have an obvious affect on CRE targeting.
The CRE-lox site-specific recombination system described here will have several applications. One application will be testing whether or not a plastid gene is essential for plant survival. The selection pressure obtained by the CRE-lox system is expected to be significantly higher than the pressure obtained during transformation, aimed at achieving targeted gene disruption (Drescher et al., 2000; Rochaix, 1997). Most importantly, the CRE-lox system will facilitate applications of plastome engineering through construction of marker-free transplastomic crops. As a consequence, removal of aadA will allow a new cycle of transformation using the same (aadA) marker gene, an important consideration given the limited number of suitable marker genes for plastid transformation (Fischer et al., 1996). Similar work from a different laboratory is reported in this issue (Hajdukiewicz et al., 2001).
Plastid codA with lox sites
The codA gene is contained in a SacI–HindIII fragment. The plastid rRNA operon (rrn) promoter derivative is contained in a SacI–EcoRI fragment obtained by PCR using oligonucleotides 5′-GGGGAGCTCGCTCCCCCGCCGTCGTTCAATG-3′ and 5′-GGG AATTCATAACTTCGTATAGCATACATTATACGAAGTTATG CTCCCAGAAATATAGCCA-3′ as primers, and plasmid pZS176 (progenitor of plasmid pZS197; Svab and Maliga, 1993) as a template. The promoter fragment PrrnloxD contains a lox site at the 3′ end adjacent to the EcoRI site. The EcoRI–NcoI fragment in plasmid pZS176 contains a ribosome-binding site. The codA coding region is contained in an NcoI–XbaI fragment (Serino and Maliga, 1997). The TrbcLloxD is the rbcL 3′-untranslated region contained in an XbaI–HindIII fragment obtained by PCR using oligonucleotides 5′-GGTCTAGATAACTTCGTATAATGTATGCTATACGAAGTTA TAGA CATTAGCAGATAAATT-3′ and 5′-GGGGGTACCAAGCTTGCTAGA TTTTGTATTTCAAATCTTG-3′ as primers, and plasmid pMSK48 (Khan and Maliga, 1999) as template. TrbcLloxD contains a lox site adjacent to the XbaI site in direct orientation relative to the lox site in the codA 5′-UTR. The chimeric PrrnloxD:codA:TrbcLloxD gene was introduced into the tobacco plastid transformation vector pPRV111B (Zoubenko et al., 1994) as a SacI–HindIII fragment to obtain plasmid pSAC48.
Plastid-targeted nuclear cre
We tested two plastid targeted nuclear cre genes. The cre genes in Agrobacterium binary vectors pKO27 and pKO28 encode the CRE recombinase. Two chimeric cre genes were prepared, one with 5 (cre1 gene; plasmid pKO28), the other with 22 (cre2 gene; plasmid pKO27) amino acids of the mature pea Rubisco small subunit (SSU) translationally fused with CRE at the N terminus (Timko et al., 1985) Both cre genes are contained in an EcoRI–HindIII fragment. The P2′Agrobacterium promoter (Velten et al., 1984), contained in an EcoRI–NcoI fragment, was obtained by PCR using oligonucleotides 5′-CCGAATTCCATTTTCACGTGTGGAAGAT ATG-3′ and 5′-CCCCATGGTAGGATCCTATCGATTTGGTGTATC GAGATTGG-3′ as primers, and plasmid pHC1 (Carrer et al., 1990) as template. PCR amplification introduced an EcoRI site at the 5′ end and BamHI and NcoI sites at the 3′ end. The Rubisco SSU transit peptides are included in BamHI–NcoI fragments. The pKO27 fragment was obtained by using oligonucleotides 5′-CCGGATCCAATTCAACCACAAGAACTAAC-3′ and 5′-GGGGCT AGCCATGGCAGGCCACACCTGCATGCAC-3′ as primers, and plasmid pSSUpGEM4 as the template (Timko et al., 1985). The pKO28 fragment was obtained by using oligonucleotides 5′-CCG GATCCAATTCAACCACAAGAACTAAC-3′ and 5′-GGGGCTAGCCA TGGTCAATGGGTTCAAATAGG-3′ as primers, and plasmid pSSUpGEM4 as the template (Timko et al., 1985). The cre coding region included in an NcoI–XbaI fragment was obtained by PCR amplification using oligonucleotides 5′-GGGGAGCTCCATGG CTAGCTCCAATTTACTGACCGTACAC-3′ and 5′-GGGTCTAGACT AATCGCCATCCTCGAGCAGGCGCACCATTGC-3′ as primers, and DNA isolated from Escherichia coli strain BNN132 (ATCC number 47059) as template. The Agrobacterium nos terminator (Tnos) is included in an XbaI–HindIII fragment (Svab et al., 1990). The plastid targeted nuclear cre genes were introduced as EcoRI–HindIII fragments into the pPZP212 Agrobacterium binary vector (Hajdukiewicz et al., 1994) to obtain plasmids pKO28 and pKO27 with five and 22 amino acids of the mature Rubisco SSU, respectively.
Plastid transformation using the biolistic protocol, selection of transplastomic tobacco lines (RMOP medium, 500 mg l−1 spectinomycin dihydrochloride) and characterization of the transplastomic lines by DNA gel-blot analysis was performed as described (Svab and Maliga, 1993). Transformation with Agrobacterium vectors pKO28 or pKO27 and regeneration of transformed tobacco plants have also been reported (Hajdukiewicz et al., 1994). Briefly, nuclear gene transformants were selected by kanamycin resistance on RMOP shoot regeneration medium containing 100 mg l−1 kanamycin and 500 mg l−1 carbenicillin. Kanamycin resistance of the shoots was confirmed by rooting on plant maintenance (RM) medium containing 100 mg l−1 kanamycin.
Plastid genome segments were identified using the following oligonucleotides: O1, 5′-CGCTCGATGACGCCAACTACC-3′; O2, 5′-GCTCCCCCGCCGTCGTTCAATG-3′; O3, 5′-CCGCCAGCGTTCAT CCTGAGC-3′; O4, 5′-GAGATGTAACTCCAGTTCC-3′; O5, 5′-TAG CTCCAATTTACTGACCGT-3′; O6, 5′-CTAATCGCCATCCTCGAGCA-3′; O7, 5′-GTCACGACGAGATCCTCGCCG-3′; O8, 5′-GACCGA CCTGTCCGGTGCCCTG-3′. DNA gel-blot analysis was carried out on ApaI–EcoRV-digested total leaf cellular DNA and probed with the targeting region (1.9 kb ApaI–EcoRV fragment containing the rrn16 gene; Figure 1a); the aadA and codA coding regions were isolated as NcoI–XbaI fragments (Svab and Maliga, 1993).
We thank Tony Cashmore for the pea Rubisco small subunit cDNA clone. This research was supported by a Rutgers F&A Special Project Grant and NSF Grant MCB 99-05043 to P.M.