* For correspondence (fax +49 761 2032745; e-mail email@example.com). †The first two authors have contributed equally to this work. ‡Present address: Botanisches Institut, Universität München, Menzinger Str. 67, D-80638 München, Germany
The chloroplast genomes of most higher plants contain two giant open reading frames designated ycf1 and ycf2. In tobacco, ycf1 potentially specifies a protein of 1901 amino acids. The putative gene product of the ycf2 reading frame is a protein of 2280 amino acids. In an attempt to determine the functions of ycf1 and ycf2, we have constructed several mutant alleles for targeted disruption and/or deletion of these two reading frames. The mutant alleles were introduced into the tobacco plastid genome by biolistic chloroplast transformation to replace the corresponding wild-type alleles by homologous recombination. Chloroplast transformants were obtained for all constructs and tested for their homoplastomic state. We report here that all transformed lines remained heteroplastomic even after repeated cycles of regeneration under high selective pressure. A balanced selection was observed in the presence of the antibiotic spectinomycin, resulting in maintenance of a fairly constant ratio of wild-type versus transformed genome copies. Upon removal of the antibiotic and therewith release of the selective pressure, sorting out towards the wild-type plastid genome occurred in all transplastomic lines. These findings suggest that ycf1 and ycf2 are functional genes and encode products that are essential for cell survival. The two reading frames are thus the first higher plant chloroplast genes identified as being indispensable.
The chloroplast genome (ptDNA) of higher plants is a circular molecule of double-stranded DNA. In most species, the genome has a size of 120–160 kb and a coding capacity of approximately 120 genes. Most of the plastid genome-encoded genes fall into two major classes ( Shimada & Sugiura 1991): genetic system genes (encoding, for example, rRNAs, tRNAs, ribosomal proteins, RNA polymerase subunits) and photosynthesis genes (encoding, for example, subunits of photosystem I, photosystem II, the cytochrome b6f complex, the ATP synthase). In addition, the plastid genome contains a number of open reading frames of unknown function (for an overview see Maier et al. 1995 ).
The successful development of transformation technologies for chloroplasts has provided the basis for addressing functional aspects of plastid genome-encoded open reading frames by reverse genetics. Knock-out alleles for plastid open reading frames can be constructed by deletional or insertional mutagenesis. These alleles are then integrated into the plastid genome where they replace the endogenous intact allele by homologous recombination. Homoplastomic transplastomic plants will entirely lack the wild-type allele and, hence, will reveal the phenotype of plants deficient for the gene product encoded by the open reading frame of interest. In most instances, the mutant phenotype involves pigment deficiency and/or lowered photosynthetic performance ( Monod et al. 1994 ; Ruf et al. 1997 ; Takahashi et al. 1996 ). This is, however, not necessarily indicative of an inactivated photosynthesis gene. Knocked-out genetic system genes may result in very similar phenotypes because of their involvement in the expression of plastid-encoded photosynthesis genes. For example, targeted inactivation of subunits of the plastid-encoded RNA polymerase results in photosynthesis-deficient white plants due to the lack of transcription of chloroplast genome-encoded photosynthesis genes ( Allison et al. 1996 ; De Santis-Maciossek et al. 1999 ; Serino & Maliga 1998).
An absolute requirement for the stability of transplastomic plants is their homoplastomic state. Higher plant cells are highly polyploid with respect to their plastid genome ( Bendich 1987). A single tobacco mesophyll cell contains up to 100 chloroplasts with approximately 100 genome copies each, amounting to a total number of about 10 000 plastid DNA molecules per cell. The primary plastid transformant is usually heteroplastomic and contains a mixture of transformed and wild-type genome copies. Such a heteroplastomic situation is normally somatically unstable and, in the absence of selective pressure, rapid changes in the relative ratios of the two genome types may occur by random segregation ( Börner & Sears 1986). This can eventually lead to the appearance of homoplastomic tissue, i.e. cell lines harboring exclusively one of the two genome types ( Allison et al. 1996 ). In chloroplast transformation experiments, a homoplastomic state of the transformed genome is usually achieved by repeated cycles of plant regeneration under high antibiotic concentrations. This creates a selective pressure favoring high expression of the transplastome and thereby driving the otherwise random genome sorting towards the accumulation of transformed genomes.
We report here that the knock-out of the two largest plastid genome-encoded open reading frames, ycf1 and ycf2, yields chloroplast transformants but does not give rise to homoplastomic plants. This suggests that these two reading frames are genuine genes, the products of which are indispensable for cell survival.
Targeted mutagenesis of the tobacco chloroplast ycf1 and ycf2 reading frames
Chloroplast genomes from a wide range of land plants carry two giant open reading frames designated ycf1 and ycf2. The reading frames are present in the completely sequenced plastid genomes from the liverwort Marchantia polymorpha ( Ohyama et al. 1986 ), and the higher plants tobacco ( Shinozaki et al. 1986 ), black pine ( Wakasugi et al. 1994 ) and Epifagus virginiana ( Wolfe et al. 1992 ). In contrast, the plastid genomes of the monocotyledonous plants rice and maize lack both ycf1 and ycf2 ( Hiratsuka et al. 1989 ; Maier et al. 1995 ). Phylogenetic analyses have suggested that at least the ycf2 open reading frame was lost from the chloroplast genome several times independently during higher plant evolution ( Downie et al. 1994 ). Loss from the plastid genome is generally believed to be accompanied with gene transfer to the nuclear genome but, as yet, no nuclear ycf1 or ycf2 homologs have been identified.
Both reading frames are actively expressed at the level of transcription. ycf2 appears to be transcribed preferentially by the nuclear-encoded phage-like plastid RNA polymerase (NEP), as evidenced by mRNA accumulation studies in rpoB deletion plants, whereas ycf1 has both a promoter for NEP and a promoter for PEP, the plastid-encoded Escherichia coli-like RNA polymerase ( Hajdukiewicz et al. 1997 ). For ycf2, immunobiochemical data indicate that the reading frame is also expressed at the protein level ( Glick & Sears 1993). In addition, ycf2 expression may be subject to tissue-specific regulation and was found to be enhanced in fruit and flower tissues ( Richards et al. 1991 ; Richards et al. 1994 ). Computer analyses of the deduced amino acid sequence of the putative Ycf2 protein revealed a potential ATP-binding domain and some weak similarity with the FtsH/CDC48 family of ATPases ( Wolfe 1994).
In an attempt to gain some information about the function of ycf1 and ycf2, we have taken a reverse genetics approach and constructed several mutant alleles for the two open reading frames. All manipulations were carried out using cloned tobacco plastid DNA fragments ( Figs 1 and 2). For each reading frame, a deletional and an insertional allele were constructed. A chimeric aadA gene conferring resistance to aminoglycoside antibiotics was inserted as a selectable marker gene, facilitating selection of chloroplast transformants on a plant regeneration medium containing spectinomycin. Plastid transformation experiments in tobacco were carried out using the biolistic protocol and typically bombarding 30 sterile leaf samples per plasmid vector ( Bock 1998; Svab & Maliga 1993). Chloroplast transformants were obtained with all constructs and correct integration of the resistance gene into the chloroplast genome was verified by DNA gel blot analyses ( Figs 3a and 4a). Three to five independent transplastomic lines from each construct were passed through several additional rounds of regeneration on spectinomycin-containing medium to select for homoplastomic tissue.
Analysis of ycf1 deletional and insertional mutants
According to our and colleagues' ( Maliga & Nixon 1998) experience, the homoplastomic state of chloroplast transformants is typically achieved after subjecting the primary transformant to one or two additional rounds of regeneration on antibiotic-containing medium. We were therefore surprised that transformants with both the insertional and the deletional allele of ycf1 remained heteroplastomic even after five additional regeneration cycles on spectinomycin-containing medium ( Fig. 3a,b). Remarkably, when we analysed comparatively DNA samples taken from all five successive rounds of regeneration, we noticed that the relative ratios of transplastomes to wild-type genomes did not change significantly over time ( Fig. 3b). This may be indicative of a balanced selection: the cells need both genomes for survival, the transplastome for expression of the spectinomycin resistance and the wild-type genome for expression of ycf1. This stable heteroplastomic state under selective conditions did not result in noticeable phenotypic alterations of the plants under in vitro culture conditions. The plants also grew equally well as the wild type when planted into pots and transfered to the greenhouse.
A stable heteroplastomic state of tobacco chloroplast transformants on antibiotic-containing medium has been observed previously in a 26-kb deletion mutant of the tobacco plastid genome ( Svab & Maliga 1993), as well as in deletion mutants for rpoB, a genetic system gene encoding a subunit of the E. coli-like plastid RNA polymerase ( Allison et al. 1996 ). In the latter case, the homoplastomic state could readily be achieved by allowing for random somatic segregation of plastid genomes in the absence of the selecting antibiotic spectinomycin. The rational behind this is that rpoB is involved in the expression of the spectinomycin resistance because, in the absence of plastid-encoded RNA polymerase, the aadA gene is only poorly expressed. Upon regeneration without spectinomycin, the selective pressure for high aadA expression is released and random sorting of plastome types occurs. Besides heteroplastomic lines, this procedure readily gives rise to homoplastomic mutant tissue as well as homoplastomic wild-type tissue.
To test the idea that ycf1 is involved in a similar way in plastid gene expression, we transfered tissue samples from all ycf1 plastid transformants onto regeneration medium without spectinomycin. The RFLP analysis as a test for homoplastomic state was then repeated with plant samples taken from different antibiotic-free segregation plates. However, in no case did the segregation in the absence of selective pressure lead to a homoplastomic state or enrichment of transplastomes over wild-type genomes ( Fig. 3c and data not shown). In all samples, either both genome types were detected (heteroplastomic state) or segregation had resulted in complete loss of the transplastome (homoplastomic state for the wild-type genome), suggesting that, also in the absence of selective pressure for aadA expression, ycf1 may be an essential gene.
We next followed the fate of the two genomes by large-scale analyses of F1 progeny. Only a few copies of the parental ptDNA are transmitted through seeds. Consequently, gametogenesis and embryo development greatly increase the probability of genome sorting and, hence, analysis of seed progeny provides a rigorous test for segregation of organellar genome types. Numerous shoots from four independently generated transplastomic lines were rooted, transfered to the greenhouse, selfed and the phenotype of the progeny was subsequently tested by seed assays. From each plant, several thousand seeds were surface sterilized and germinated on antibiotic-free medium as well as on spectinomycin-containing medium. On this medium, seeds containing the transplastome give rise to green seedlings whereas seedlings possessing only wild-type genomes are white ( Svab & Maliga 1993). As expected, some plants had completely lost the transplastome, as evidenced by uniform sensitivity of the progeny to spectinomycin. Those plants that gave rise a to segregating F1 showed greatly varying amounts of transplastomic progeny with the extreme that, in one case, only less than 1% of the seeds displayed antibiotic resistance. Southern blots confirmed that all progeny plants were either heteroplastomic or had lost the transplastome completely (data not shown).
Altogether these data strongly suggest that a homoplastomic state of inactive ycf1 alleles, and hence complete absence of the ycf1 gene product, is lethal and probably incompatible with cell survival. ycf1 is thus the first higher plant chloroplast reading frame for which no homoplastomic knock-out plants could be generated. Interestingly, ycf1 shares a low degree of similarity with a C. reinhardtii open reading frame that also could not be eliminated completely from the Chlamydomonas chloroplast ( Boudreau et al. 1997 ).
Analysis of ycf2 deletional and insertional mutants
Analogous analyses as for ycf1 were conducted for the largest chloroplast genome-encoded open reading frame, ycf2. First, we also constructed a deletional allele of ycf2 (pδycf2; Fig. 2). Like the ycf1 transformants, all ycf2 deletion lines remained heteroplastomic even after eight cycles of successive regeneration on spectinomycin- containing medium. As in the case of ycf1, subsequent segregation analyses on antibiotic-free medium did not result in plants homoplastomic for the transplastome (data not shown).
Because a homoplastomic ycf2 knock-out appeared to be lethal, we constructed a second mutant allele by insertional mutagenesis. We inserted the selectable marker gene aadA into the C-terminus of the reading frame (pIycf2; Fig. 2) hoping that the resulting ycf2 allele would not be a lethal null allele but would give rise to a truncated Ycf2 protein that is still functional to some degree. However, analysis of the resulting chloroplast transformants indicated that the C-terminal insertion mutant also cannot be brought to a homoplastomic state ( Fig. 4a). Again, heteroplastomic shoots were rooted on drug-free medium, transfered to the greenhouse and tested for genome segregation. Southern blot analyses with samples taken from individual plants revealed only a homoplastomic state for the wild-type genome ( Fig. 4b). This indicates that ycf2 is also an essential chloroplast gene that cannot be fully inactivated. Fast loss of the transplastome very early after removal of the antibiotic may suggest that the selective pressure towards maintenance of ycf2 is very high and possibly even higher than in the case of ycf1, where transplastomes could be detected frequently in tissue samples and sometimes were also transmitted to the next generation ( Figs 3c and 4b and data not shown).
In the course of this work, we have performed several in vivo mutageneses of the two largest chloroplast-genome encoded open reading frames, ycf1 and ycf2. We have shown that all transplastomic plants generated remain heteroplastomic under selective conditions and have a strong tendency to lose the transplastome under non-selective conditions. The failure of the mutant ycf1 and ycf2 alleles to become homoplastomic suggests that (i) both reading frames are genuine genes and (ii) their gene products are essential for cell survival. Moreover, as both the light and dark reactions of photosynthesis are dispensable under heterotrophic in vitro culture conditions ( Kanevski & Maliga 1994; Ruf et al. 1997 ), our data indicate that the functions of the two reading frames are not related to photosynthesis. This is consistent with the presence of ycf1 and ycf2 in the plastid genome of E. virginiana, a holoparasitic plant that, during evolution, has lost practically all photosynthesis-related genes from its plastid genome ( dePamphilis & Palmer 1990; Wolfe et al. 1992 ).
The finding that chloroplast genes in higher plants can encode an essential function is unexpected. Analysis of plastid gene expression in the barley mutant albostrians had suggested that, in barley, chloroplast protein-coding genes are not required for cell survival. The albostrians mutant lacks functional chloroplast ribosomes and, therefore, does not synthesize plastid genome-encoded proteins ( Hess et al. 1993 ; Hess et al. 1994 ). Most interestingly, the plastid genomes of the monocotyledonous plants maize and rice lack the ycf1 and ycf2 reading frames ( Hiratsuka et al. 1989 ; Maier et al. 1995 ) and it is therefore highly likely that the two reading frames are also absent from the very closely related Poacean species barley. Thus, it seems tempting to speculate that the plastid encoding of ycf1 and ycf2 makes chloroplast translation essential in tobacco, whereas it is dispensable in barley. Consequently, expression of ycf1 and ycf2 would be one of the probably very few essential functions of the chloroplast genome in tobacco and possibly also in other dicotyledonous species. This is again in accordance with the data on the rudimentary E. virginiana plastid genome in which ycf1 and ycf2 are, with the exception of an acetyl-CoA-carboxylase subunit (accD), the only reading frames that are probably not related to gene expression (i.e. translation, intron splicing or protein degradation; dePamphilis & Palmer 1990; Wolfe et al. 1992 ). At present, we can only speculate as to what the essential role of ycf1 and ycf2 could be. Two possibilities are that the gene products are involved in some essential pathway in cellular metabolism or, alternatively, serve some structural function for the plastid compartment (e.g. for maintenance of the plastid DNA).
Remarkably, differences concerning the general role of chloroplast gene expression in plant development do not only occur between monocotyledonous and dicotyledonous species but also among dicots. Whereas treatment of seedlings with spectinomycin resulted in growth arrest in tobacco, Brassica napus seedlings continued to grow and formed white leaves ( Zubko & Day 1998). On the other hand, whereas tobacco seedlings recovered from spectinomycin-induced bleaching, B. napus seedlings remained white and exhibited a stable deficiency in plastid ribosomes and plastid translation products ( Zubko & Day 1998). The developmental and/or physiological causes for these species-specific differences in the action of the translational inhibitor spectinomycin and the general role of plastid protein biosynthesis remain to be established.
Summarizing the results from this study and those from previous reverse genetics analyses, the following three possible outcomes of knock-out experiments in higher plant plastids can be distinguished.
Photosynthesis is optional under heterotrophic conditions. Therefore, knock-out of a photosynthesis gene does not confer a significant selective disadvantage in in vitro culture. Typically, homoplastomic transgenic plants are obtained after one to three rounds of regeneration on antibiotic-containing medium. This holds true for both components involved in the light reactions and those participating in the dark reactions of photosynthesis ( Kanevski & Maliga 1994; Ruf et al. 1997 ).
Non-essential genetic system genes
Chloroplast transformants are obtained but remain heteroplastomic in the presence of the selecting antibiotic because the gene inactivated is required for the expression of the selectable marker gene. However, a homoplastomic state can readily be achieved by random sorting out of plastid genomes in the absence of the antibiotic, thereby releasing the selective pressure for high expression of the selectable marker gene. This strategy proved successful in the construction of homoplastomic deletion mutants for subunits of the E. coli-like plastid RNA polymerase in tobacco ( Allison et al. 1996 ; De Santis-Maciossek et al. 1999 ; Serino & Maliga 1998).
Even for those plastid genome-encoded genes that are essential for cell survival, chloroplast transformants carrying knock-out alleles can be generated. However, the transplastomic lines remain heteroplastomic. Plant maintenance on antibiotic-containing medium leads to balanced selection, i.e. a more or less constant ratio of wild-type to transformed genome copies. When the selective pressure is removed (by transferring the plants to drug-free medium), sorting out towards the wild type occurs and the transplastome is eventually lost. The tobacco ycf1 and ycf2 reading frames are first examples for such essential genes in higher plant plastid genomes.
Plant material and growth conditions
Sterile tobacco plants (Nicotiana tabacum cv. Petit Havana) were grown on agar-solidified MS medium containing 30 g l−1 sucrose ( Murashige & Skoog 1962). Transplastomic lines were rooted and propagated on the same medium. For analysis of F1 progeny, wild-type and transformed plants were planted in pots and kept under greenhouse conditions. Surface-sterilized F1 seeds were germinated on MS medium either in the presence or in the absence of the antibiotic spectinomycin (500 mg l−1).
List of oligonucleotides
The following synthetic oligonucleotides were used in this study: P10: 5′-AACCTCCTATAGACTAGGC-3′ P11: 5′-AGCGAAATG- TAGTGCTTACG-3′ P28: 5′-TAGCACCCTCTTGATAGAAC-3′ P29: 5′-CGCTATGGAACTCGCCGCC-3′ PY1: 5′-TATAATAATCCAGA- AATTGC-3′ PY2: 5′-TTACGCATTGAATTTCTGG-3′ PO3: 5′-TTC-CGGCTA-AGGAATGGG-3′ PO4: 5′-GGAGTTCTTGCAGAGTGG-3′.
Construction of plastid transformation vectors
The region of the tobacco chloroplast genome containing the ycf1 reading frame ( Shinozaki et al. 1986 ) was excised from a SalI ptDNA clone (kindly provided by P. Maliga, Piscataway, NJ) as a SalI/KpnI fragment ( Fig. 1). The fragment was ligated into a similarly cut Bluescript SK vector (Stratagene, La Jolla, CA) from which part of the polylinker had been deleted by digestion with EcoRV and Ecl136II, generating plasmid pAD2. Most of the putative coding region of the ycf1 reading frame was subsequently deleted by digestion with NdeI and BamHI ( Fig. 1). After a fill-in reaction of the recessed ends using the Klenow fragment of DNA polymerase I, a chimeric aadA gene confering resistance to aminoglycoside antibiotics ( Svab & Maliga 1993) was inserted to replace ycf1 and to facilitate selection of chloroplast transformants. A plasmid clone carrying the aadA gene in the same orientation as previously ycf1 yielded the final ycf1 deletion vector pδycf1 ( Fig. 1). A ycf1 allele for insertional mutagenesis was constructed by linearizing pAD2 with SpeI (for which a unique site is present in the 5′ part of the ycf1 reading frame), filling in the recessed ends with Klenow enzyme and ligating a chimeric aadA gene into this blunted SpeI site (vector pIycf1; Fig. 1). The tobacco chloroplast DNA region containing the open reading frame ycf2 was cloned as a 9.1-kb HindIII/SacI fragment ( Fig. 2) into pBSII SK+. For deletional mutagenesis, the ycf2 reading frame was nearly completely excised with BglII followed by a fill-in reaction with Klenow enzyme. A chimeric aadA gene was subsequently cloned into the blunted BglII sites maintaining the same orientation as ycf2 has in the wild type (transformation vector pδycf2; Fig. 2). For C-terminal insertional mutagenesis, a unique NarI site located in the 3′ region of ycf2 was cut, blunted by a fill-in reaction with Klenow polymerase, followed by ligation to the chimeric aadA. A clone carrying the aadA in the ycf2-like orientation was selected and used as transformation vector (pIycf2;Fig. 2).
Plastid transformation and selection of transplastomic tobacco lines
Young leaves from sterile tobacco plants were bombarded with plasmid DNA-coated tungsten particles using the DuPont biolistic gun (PDS1000He; BioRad, Hercules, CA). Primary spectinomycin-resistant lines were selected on RMOP regeneration medium containing 500 mg l−1 spectinomycin dihydrochloride ( Svab & Maliga 1993; Svab et al. 1990 ). Plastid transformants were identified by PCR amplification according to standard protocols and using a combination of primer P10 or P28 (complementary to the psbA 3′-untranslated region of the chimeric aadA gene) with primer P11 or P29 (derived from the 3′ portion of the aadA coding region). For each construct, several independent transplastomic lines were subjected to between four and eight additional rounds of regeneration on RMOP/spectinomycin to select for homoplastomic tissue. Homoplastomic versus heteroplastomic state was tested by DNA gel blot analyses.
Isolation of nucleic acids and hybridization procedures
Total plant DNA was isolated according to a rapid mini-preparation procedure ( Doyle & Doyle 1990). Restriction enzyme-digested DNA samples were separated on 0.8–1.2% agarose gels and blotted onto Hybond N nylon membranes (Amersham Int., Little Chalfont, UK) using standard protocols ( Sambrook et al. 1989 ). For hybridization, α[32P]dATP-labeled probes were generated by random priming (Multiprime DNA labeling system; Amersham) following the instructions of the manufacturer. A ycf1-specific probe was prepared from a PCR product generated by amplification of tobacco ptDNA with primer pair PY1/PY2 (corresponding to the 5′ region of ycf1;Fig. 3). A tobacco ycf2-specific probe was synthesized by radiolabeling a PCR product covering the 3′ part of the coding region of the gene (obtained by amplification with primer pair PO3/PO4). Hybridizations were carried out at 65°C in rapid hybridization buffer (Amersham).
We are indebted to Drs P. Maliga and Z. Svab (Rutgers, The State University of New Jersey) for a ptDNA library and a chimeric aadA gene. Excellent technical assistance by Mrs M. Hermann is gratefully acknowledged. This research was supported by a grant from the DFG (BO 1482/1–3) to R.B. and by the PROBRAL Program of DAAD and CAPES.