Introduction

Rice is one of the most important food crops in the world and is the main nutritional staple for approximately 40% of the world’s population. Annual rice production on a worldwide basis must increase from its current level of 520 million tons per year to 810 million tons per year by 2025 (Hossain 1996) to support the rapid increase in the human population that has been predicted. Since rice hybrids show heterosis, which subsequently results in them having yields 15–30% higher than inbred varieties (Yuan 1994; Fujimura et al. 1996), these hybrids may offer a solution to this problem.

The combination of cytoplasmic male sterility (CMS) and a nuclear gene for restoration of fertility (Rf) are essential for breeding hybrid varieties and for hybrid seed production. CMS, which eliminates the possibility of self-pollination, is commercially used in the production of hybrid seeds for economically important plants (Newton 1988). The Rf gene, on the other hand, restores the self-fertilization ability of a hybrid plant and is indispensable in crops in which seeds are harvested, such as hybrid rice. Despite its importance in agriculture, the molecular mechanisms of CMS and its recovery are still unclear.

CMS is common in higher plants and derives from the incompatibility between nuclear and cytoplasmic gene products, which results in a failure to produce mature pollen grains (Newton 1988; Levings and Brown 1989). The reorganization of mitochondrial genomes in CMS plants has been reported in numerous species, and in many cases it results in the formation of chimeric genes in these regions that are responsible for the CMS characteristic (Schuable and Wise 1998). Transcripts produced from such chimeric genes may inhibit expression of the normal gene and impair the formation of fertile pollens (Levings and Vasil 1995). In a number of plant species, such malformed transcripts in CMS mitochondria are altered in the presence of the Rf genes, leading to the production of fertile pollen and self-pollinated seeds (Pruitt and Hanson 1991; Iwabuchi et al. 1993; Akagi et al. 1994; Tang et al. 1996, 1998; Wise et al. 1999). These facts have suggested that some Rf genes are involved in the processing of mitochondrial (mt) mRNA associated with CMS.

Proteins that contain a pentatricopeptide repeat (PPR) are scattered throughout genome of higher plants (Small and Peeters 2000). In the Arabidopsis genome, more than 450 genes encode members of the PPR family (Aubourg et al. 2000). The PPR motif is presumed to play a role in binding to macromolecules such as RNA, and it has been predicted that many Arabidopsis PPR proteins target organelles (Aubourg et al. 2000). Members of the PPR family are thought to be involved in controlling organelle gene expression by processing or editing CMS-associated transcripts (Small and Peeters 2000), and it is through this mechanism that some PPR proteins are thought to participate in the suppression of CMS. In fact, one PPR protein, the product of the Rf-PPR592 gene, counteracted CMS in Petunia (Bentolila et al. 2002), and another PPR protein in radish also suppressed CMS in Brassica napus (Brown et al. 2003; Koizuka et al. 2003). However, it is unclear which PPR proteins suppress CMS, since many genes in a plant genome encode PPRs (Aubourg et al. 2000). Moreover, as several CMS/Rf systems have been identified in rice (Virmani and Shinjyo 1988), it is necessary to specify the Rf gene and its corresponding CMS.

In rice, CMS occurs when the cytoplasm of japonica or indica rice is replaced with that of indica rice or wild rice (Katsuo and Mizushima 1958; Li and Zho 1986; Virmani and Shinjyo 1988). One CMS system, based on the cytoplasm of Chinsurah Boro II (indica rice) and the nucleus of Taichung 65 (japonica rice), is called ms-bo type or BT-type. BT-type CMS is restored by the nuclear restorer gene Rf-1, which was initially identified in Chinsurah Boro II and which has been shown to be due to a single locus in the nuclear genome (Shinjyo 1969). The Rf-1 gene was mapped on chromosome 10 (Shinjyo1975, 1984; Akagi et al. 1996). Since BT-type CMS is a type of gametophytic CMS, only pollen with the Rf-1 gene can develop normally into F1 plants (Shinjyo 1969). In BT-type CMS mitochondria of rice, the 3′ part of atp6 was recombined to form a chimeric configuration (Iwabuchi et al. 1993; Akagi et al. 1994) that was joined to a novel open reading frame (ORF), orf79 (Akagi et al. 1994). This di-cistronic mRNA was processed or edited by the action of the Rf-1 gene, intact atp6 mRNA was formed, and fertility was recovered (Iwabuchi et al. 1993; Akagi et al. 1994).

Kazama and Toriyama (2003) recently reported the candidate for the Rf-1 gene by the transgenic strategy. They mapped the Rf-1 locus within 1 cM. Using a corresponding region of Nipponbare genome sequence, they searched for ORFs encoding mitochondria targeting PPR proteins. Three candidate genes were cloned from Miliyang 23, and these were introduced into callus of an rf-1/rf-1 CMS line. In a transgenic rice callus, one of the candidate genes (PPR8–1) encoded a 791-amino acid protein with 18 PPR repeats and was involved in processing of the transcript in mitochondrial atp6. However, these researchers did not confirm that PPR8–1 restored fertility in regenerated plants.

A set of near-isogenic lines (NILs) that differ only at the locus of the Rf-1 gene has been developed by repeated backcrossing to replace most of the nuclear genome of Chisurah Boro II with that of Taichung 65 (Shinjyo 1975). We have also developed DNA markers that are tightly linked to the Rf-1 gene using these NILs (Akagi et al. 1996; Ichikawa et al. 1997). In the investigation reported here, we attempted to identify the Rf-1 gene with a positional cloning strategy using these NILs for precise linkage analysis. We found duplicate open reading frames (ORFs) (Rf-1A, Rf-1B) that encode a PPR protein at the Rf-1 locus. Of these, Rf-1B encoded only a truncated protein; therefore we concluded that only Rf-1A could be Rf-1.

Materials and methods

Plant materials and linkage analysis

Two NILs, MTC-10A [(cms-bo)rf-1/rf-1] and MTC-10R [(cms-bo)Rf-1/Rf-1], were used (Akagi et al. 1994) throughout this work. The cytoplasm of MTC-10A was derived from Chinsurah Boro II, and its nuclear genome was derived from Taichung 65; thus MTC-10A is a BT-type CMS line. MTC-10R has the same cytoplasmic and nuclear genomes as MTC-10A except for the presence of Rf-1, which was derived from Chinsurah Boro II (Shinjyo 1975). The cytoplasm and the restoring gene of the NILs had been introduced into Taiching 65 from Chinsurah Boro II by recurrent backcrosses. These lines were originally referred to as BT-C and BT-A, respectively, by Shinjyo (1969). IR24, IR36 and MTC-18R, which can also restore the fertility of BT-type CMS, were also used. MTC-18R retained the Rf-1 gene derived from IR8 (Ichikawa et al. 1997).

A population consisting of 300 plants of BC1 progeny was used to analyze linkage of the Rf-1 gene and DNA markers. These plants were derived from backcrossing the F1 hybrid of MTC-10A and MTC-10R with MTC-10A. Since MTC-10A shows gametophytic CMS, pollen without the Rf-1 gene is abortive and only pollen with the Rf-1 gene can develop normally in F1 plants between MTC-10A and MTC-10R. Therefore, the genotype of the BC1 plants between the MTC-10A and F1 was Rf-1/rf-1. The recombination value was converted into map distances (centiMorgans) using the Kosambi function (Kosambi 1944).

Seven BC1 plants with recombination around the Rf-1 gene were selected and then used for fine mapping of the Rf-1 gene. While the ovules of BC1 plants with the BT-type cytoplasm can develop normally even without the Rf-1 gene, their pollen will develop only when they carry the Rf-1 gene. Therefore, BC1F2 progeny in these BC1s were either heterozygous (Rf-1/rf-1) or homozygous (Rf-1/Rf-1) for the Rf-1 gene. Twenty-two BC1F2 plants were selected with a heterozygous allele of the microsatellite marker (68923-7). We used 6,104 BC1F3 progeny derived from self-pollination of these BC1F2 plants to precisely locate the Rf-1 gene.

Polymerase chain reaction

Total rice DNA was prepared from green leaves according to Lichtenstein and Draper (1985). Crude DNA was also extracted from the leaves of BC1, BC1F2 and BC1F3 seedlings (Akagi et al. 1997). Each leaf was dried at 70°C for 2 h and then homogenized with a small glass bead in a 1.5-ml Eppendorf tube using a vortex mixer. Crude DNA was then extracted according to the method of Edwards et al. (1991) and subjected to PCR for linkage analysis.

Primer sets were designed using genomic sequences of rice cv. Nipponbare. A total of 112 primer sequences were selected from the region between 5,5071 and 7,7373 of the nucleotide sequence of the bacterial artificial chromosome (BAC) clone OSJNBa0017E08 (AC068923, gi: 17298629). Corresponding genomic regions in MTC-10A (recessive allele) and MTC-10R (dominant allele) were amplified by PCR using these respective primers. After DNA amplification, PCR products were purified using a PCR purification kit (QIAGEN, Valencia, Calif.) and then sequenced. DNA markers were developed on the basis of polymorphisms between sequences of two NILs (Table 1). PCR analyses were run in 20-μl aliquots of a buffer consisting of 10 mM Tris-HCl (pH 8.3), 50 mM KCl, 1.5 mM MgCl2, 0.5 or 1 U of TAKARA Taq (TAKARA), 4 nmol dNTP, 10 ng of genomic DNA and 10 pmol of each set of primers listed in Table 1 in a Thermal Cycler 9600 (Perkin-Elmer, Foster City, Calif.). Thirty-five PCR cycles, each consisting of 10 s of denaturation at 94°C, 30 s of annealing at 55°C or 60°C and 2 min of polymerization at 72°C, were performed. PCR products for cleaved amplified polymorphic sequence (CAPs) markers were digested with a corresponding restriction endonuclease. The PCR products were electrophoresed on 1.8–3% MetaPhor Agarose gels (FMC, Rockland, Me.).

Table 1 DNA markers around the Rf-1 gene
Full size table

Analysis of RNA

Total RNA from various rice tissues was extracted using an RNeasy Plant Mini kit (QIAGEN) following the manufacturer’s protocol. To remove contaminating DNA, total RNA was suspended in 100 μl 0.8 mM MgCl2, 5 mM DTT and 40 mM Tris-HCl, pH 7.5 and digested with RNase-free DNaseI (50 U/ml; TAKARA) for 30 min at 37°C. It was then extracted with phenol-chloroform and precipitated with EtOH. First-strand cDNA was synthesized from 1 μg of total RNA with Oligo dT Adaptor Primer FB using a High Fidelity RNA PCR kit (TAKARA) according to the supplier’s protocol. PCR was performed with primer pairs (see below) using TAKARA Ex Taq (TAKARA) or TAKARA LA Taq with GC I buffer (TAKARA). The primer pairs used were as follows: a-1F, ttgttacaggtgcacagacc; a-1R, gcctcgatctacggcttgaa; a-2F, agcttgacccttgcataacg; a-2R, taagacaaactgcgttgcgg; a-3F, caaccgatgagaatgccgta; a-3R, taagacaaactgcgttgcgg; b-1F, acagtccaggaaccaca; b-1R, ctgttgactctggcatgcta; b-2F, gtaggtcacaacatttggcg; b-2R, aagctctcgagctactgcac; b-3F, gtaggtcacaacatttggcg; b-3R, tgctgcacctgtcagctagg; b-4F, gtaggtcacaacatttggcg; b-4R, gctgcacctgtcagctaggc. For the β-tubulin gene, the primers used were: rice tubulin F (tggtcggattcgccccgctg) and rice tubulin R (ttacatgtcgtcagcctcct) (Nakagawa et al. 2000). Amplified DNA fragments were cloned into pT7Blue (Novagen) and sequenced. For 3′-RACE (rapid amplification of cDNA end), the first-strand cDNA was amplified by PCR with Adaptor Primer FB and a specific primer for each gene. Nested PCR was also performed with Adaptor Primer FB and another specific primer for each gene. Amplified DNA fragments were cloned into pT7Blue (Novagen) and then sequenced.

Sequence analysis

Genomic regions covering the Rf-1 locus of MTC-10R and MTC-10A were amplified using primers based on nucleotide sequences obtained from Nipponbare. The nucleotide sequences of the amplicons were determined by a direct sequencing analysis. We analyzed the nucleotide sequences of mRNA using at least five cloned cDNA fragments and also by a direct sequencing analysis. Nucleotide sequences were determined using a DNA sequencing system (ABI 373S; Applied Biosystems, Foster City, Calif.). DNA sequences were analyzed using genetyx-mac (Software Development, Tokyo). Genomic sequences were also analyzed using gene prediction programs, genescan and fgenesh.

Results

Mapping of the Rf-1 gene

To clarify the precise position of the Rf-1 gene on chromosome 10, we selected four RFLP markers, G2155, G4003, G291 and C1361, from a high-density molecular genetic map (Kurata et al. 1994) and converted them into PCR-based markers (Table 1). When a plant lacked the allele from MTC-10R at one of these markers, it was recombined between the marker and the Rf-1 gene. The calculated position and the distance between the Rf-1 gene and each DNA marker on chromosome 10 are shown in Fig. 1. The Rf-1 gene was located between C1361 and fL601, with map distances of 1.00±0.58 cM and 1.50±0.67 cM, respectively (Fig. 1). The distance between the Rf-1 gene and OSRRf was almost the same that reported by Kurata et al. (1994) and Akagi et al. (1996.

Fig. 1
figure 1

Genetic mapping of the Rf-1 gene on chromosome 10 of rice. Linkage orders between the Rf-1 gene and six DNA markers are shown to the right of the vertical bar; genetic distances (in centiMorgans) are shown to the left of the vertical bar. The positions of four DNA markers designed from RFLP markers are indicated on the high-density RFLP map of chromosome 10 from the Rice Genome Research Program (Kurata et al. 1994). The relative positions of the sequences homologous to fL601, C1361 and OSRRf are indicated on the BAC contig

Full size image

By a blast homology search, the sequence of fL601 was located on the rice BAC clone of OSJNBa0017E08, and the sequences of C1361 and OSRRf were located on that of OSJNBa0078O01 (Fig. 1). These two BAC clones overlapped each other on chromosome 10 (Fig. 1). Thus, the Rf-1 gene should lie between C1361 and fL601 on these BAC clones at a physical distance of 219 kb.

High-resolution mapping of the Rf-1 gene

The locus of the Rf-1 gene within this 219-kb fragment of chromosome 10 was clarified more precisely with 6,104 BC1F3 progeny. All of the 723 plants that were randomly selected from these 6,104 plants set seeds normally.

Fourteen DNA markers between fL601 and C1361 were newly developed based on the nucleotide sequences of the two BAC clones (Table 1, Fig. 2). The population of BC1F3 plants was analyzed using OSRRf, C1361 and 68923-11, and nine plants were found to be carrying a homozygous allele from MTC-10A (Fig. 2), thereby showing recombination around the Rf-1 gene. The genotypes of these plants were characterized more precisely using 17 DNA markers between C1361 and fL601 (Fig. 2). Plants FM12-19 and FM24-26 were recombinant between 68923-6 and 68923-7, while FM0-626 was recombinant between 68923-6 and 68923-9 (Fig. 2). The Rf-1 gene lay between two markers, 68923-6 and 68923-9, and the remaining parts (shown by open bars) were derived from MTC-10A (Fig. 2).

Fig. 2
figure 2

Delimitation of the Rf-1 locus and positions of DNA markers around the Rf-1 locus. The positions of 17 DNA markers designed from the sequences of BAC clones (OSJNBa0078O01 and OSJNBa0017E08) and the genotype of nine plants in which recombination occurred between OSRRf and fL601 are indicated. Except for fL601 and 79888-1, DNA markers are co-dominant, allowing the discrimination of Rf-1/Rf-1 plants from Rf-1/rf-1 ones. White regions are chromosomal segments that are homozygous for the recessive allele from MTC-10A, and black regions are ones that are heterozygous for the allele from MTC-10A and MTC-10R (Rf-1/rf-1). Recombination occurred in the region of transition from black to white. The Rf-1 locus was delimited to the approximately 22.7-kb-long region between 68923-6 and 68923-9 in the Nipponbare genome

Full size image

Sequence analysis of the Rf-1 gene locus

This region between 68923-6 and 68923-9 is about 22.7 kbp long in the Nipponbare genome. Because Nipponbare is a non-restorer for BT-type CMS and presumably carries the recessive gene, rf-1, the nucleotide sequence of the corresponding region of MTC-10R was determined. In this MTC-10R sequence, we searched coding regions using both gene prediction programs and reverse transcription (RT)-PCR for detecting mRNA.

Two ORFs (Rf-1A and Rf-1B) without introns were found in this 22.4-kb region of MTC-10R (Fig. 3). cDNAs corresponding to these ORFs were cloned by RT-PCR and expanded to their 3′ ends by 3′-RACE (Fig. 3). An additional mRNA, transcribed from the region downstream of the Rf-1A gene containing several introns, was also detected by RT-PCR. The mRNA contained several stop codons and may be translated into a truncated and nonfunctional protein. Thus, only Rf-1A and Rf-1B were considered candidates for the Rf-1 gene.

Fig. 3A–D
figure 3

Genomic organization of the Rf-1 locus. A The structural features and direction of transcripts corresponding to the Rf-1A gene and the Rf-1B gene of MTC-10R. Arrows indicate the positions of poly(A) sites. B The genomic structure of the Rf-1 locus in MTC-10R containing two duplicate open reading frames (Rf-1A and Rf-1B). C The genomic structure of the recessive allele (MTC-10A). The relative positions and lengths of deletions (−) and an insertion (+) were found in MTC-10A. Another homolog (Rf-1C) is located upstream from the Rf-1B gene in both MTC-10R (B) and MTC-10A (C). D The structures and relative positions of three putative genes around the Rf-1 locus in the Nipponbare genome. The positions of the boundary DNA markers are indicated by arrows. The putative genes in the Nipponbare genome corresponding to Rf-1A, Rf-1B and Rf-1C are registered in the database as OSJNBa001717E08.19, OSJNBa001717E08.19 and OSJNBa001717E08.19, respectively

Full size image

In the counterpart of both genes on the MTC-10A genome, we found 1-bp and 574-bp deletions in the region corresponding to the Rf-1A gene, and 1-bp and 21-bp deletions in the region corresponding to Rf-1B. An insertion of a 970-bp fragment homologous to the long terminal region (LTR) was also found in the Rf-1B gene (Fig. 3). These genes on the MTC-10A genome may have lost their function.

Rf-1 gene encodes PPR protein

The nucleotide sequences of the Rf-1A and Rf-1B genes are highly conserved. The Rf-1A gene encodes a 791-amino acid protein and the Rf-1B gene encodes a 332-amino acid protein (Fig. 4). The deduced amino acid sequence of Rf-1A showed 86.7% identity to that of Rf-1B (Fig. 4). NCBI Conserved Domain Search identified the 16 sets of a PPR motif in the Rf-1A gene (Fig. 4). The PPR-motives (PPR-1 to PPR10) were identified with E-values above 0.01, and the remaining motives (PPR-11 to PPR16) were identified with E-values in the range of 0.01 to 1. The Rf-1B gene also had nine sets of the PPR motif (Fig. 4). The protein encoded by the Rf-1A gene contained a putative mitochondrial transit peptide and may locate in mitochondria, as predicted by the method reported by Emanuelsson et al. (2000). Rf-1B may lack this function, in that a transit peptide was not predicted from its putative amino acid sequence even though its nucleotide sequence is highly homologous to that of the Rf-1A gene. Furthermore, the Rf-1B gene in MTC-10R has another premature stop codon (TAA) (Figure 3) encoding a truncated and nonfunctional protein.

Fig. 4
figure 4

Comparison of deduced amino acid sequences between Rf-1A, Rf-1B and Rf-1C of MTC-10R. Dots represent amino acids identical to those in Rf-1A. Sixteen repeats of the PPR motif, which consists of 35 amino acids are indicated under the three amino acid sequences

Full size image

The rice lines IR24, IR36 and MTC-18R can also restore BT-type CMS. However, the proteins encoded by the Rf-1B gene in their respective genomes were predicted to have a function of mitochondrial targeting and 14 sets of a PPR motif (PPR-11 and PPR-12 missing; data not shown). A frame-shift mutation produced by a single nucleotide extension in MTC-10R (data not shown) and a 26-bp deletion that generated new stop codons resulted in a loss of function of Rf-1B in MTC-10R.

A gene (Rf-1C) homologous to the Rf-1A gene was found upstream of the Rf-1B gene (Fig. 3). This gene also encoded a PPR protein with 88.6% homologous to the Rf-1A protein (Fig. 4). The deduced amino acid sequence of the Rf-1C gene of MTC-10R was 99.9% and 99.5% identical to those of MTC-10A and Nipponbare, respectively. While the Rf-1A gene and the Rf-1C gene encoded highly homologous PPR proteins, the Rf-1C gene is beyond the cross-over exhibited by the recombinant FM0-626 (Fig. 2) and cannot be the Rf-1 gene.

Expression profiles

The expression of both the Rf-1A gene (Fig. 5B, C) and Rf-1B gene (Fig. 5B, D) was detected in panicles at the booting stage with the development of pollen and in green leaves of MTC-10R by RT-PCR. These DNA fragments were only amplified in the presence of reverse transcriptase, similar to the results for β-tubulin (Fig. 5A). We confirmed that these DNA fragments had been specifically amplified from the mRNA of each gene by a direct sequencing analysis (data not shown).

Fig. 5A–D
figure 5

Expression of the Rf-1A and Rf-1B genes. RNA of leaves (L) and developing panicles (P) of MTC-10R (Rf-1/Rf-1) was subjected to RT-PCR with primers specific for the β-tubulin gene (A), Rf-1A gene (C) and Rf-1B gene (D). –RT, +RT With and without reverse transcriptase in the reaction, respectively. B The regions amplified by RT-PCR are indicated under the structures of Rf-1A and Rf-1B mRNAs. M Marker. Sizes of amplicons are also indicated

Full size image

Duplicate Rf-1 gene homologs in the rice genome

In addition to the Rf-1A, Rf-1B and Rf-1C genes, six copies of genes homologous to the Rf-1A gene were found on chromosome 10 in the Nipponbare genome (Fig. 6). Seven of them encoded only short proteins. The remaining two genes (Rf-68950-2 and Rf-1C) encoded proteins as long as Rf-1A of MTC-10R and may be functional (Fig. 6). The Rf-1C gene encoded a protein containing a mitochondria-targeting peptide and 16 repeats of the PPR motif (Fig. 4).

Fig. 6
figure 6

BAC contig covering the Rf-1 locus in the Nipponbare genome showing the genomic structure of the regions homologous to the Rf-1A gene in the Nipponbare genome. Broad open arrows represent the positions and direction of nine Rf-1A gene homologs. The structures of these homologues are indicated under the contig and are compared with the Rf-1A gene. Broad black arrows represent regions that can be translated into a protein. Rf-79888-3, Rf-79888-2, Rf-79888-3, Rf-68950-1, Rf-68950-2 and Rf-68950-3 partly correspond to the predicted genes, OSJNBa0078O01.29, OSJNBa0078O01.20, OSJNBa0078O01.15, OSJNBa0041P03.13, OSJNBa0041P03.12 and OSJNBa0041P03.10, respectively

Full size image

Discussion

Positional cloning of the Rf-1 gene

During the course of the investigation reported here, the Rf-1 locus was delimited in a genomic region of 22.4 kb by linkage analysis using 6,104 BC1F3 progeny (Fig. 2). Two ORFs (Rf-1A and Rf-1B) encoded proteins containing a PPR motif (Fig. 3) that are transcribed in different parts of the plant with the MTC-10R genome (Figs. 3, 5). Furthermore, with respect to the recessive allele (rf-1), the genes corresponding to both the Rf-1A and Rf-1B genes possess frame-shift mutations and are predicted to be nonfunctional (Fig. 3).

Rf-1A protein with a signal peptide was predicted to localize in mitochondria, whereas Rf-1B of MTC-10R lacked such a signal peptide (Fig. 4). In addition, the Rf-1B gene of MTC-10R has not only a frame-shift mutation but also a 26-bp deletion that generates new stop codons within the gene; therefore, the Rf-1B gene product lacked functionality (data not show). In comparison with the Rf-1B gene in MTC-10R, in three other restorer lines, the gene corresponding to the Rf-1B gene encoded a protein with a mitochondria-targeting signal and 16 repeats of a PPR motif. The segregates that carried only the Rf-1A gene as a functional gene in the Rf-1 locus had fertile seeds. This supports the conclusion that the Rf-1A gene is the Rf-1 gene.

The Rf-1 gene encodes a PPR protein

Proteins with a PPR motif constitute a large family in higher plants (Aubourg et al. 2000; Small and Peeters 2000). The PPR motif consists of a 35-amino acid repeat and has been predicted to be a macromolecule-binding motif (Small and Peeters 2000). It has been suggested that PPR proteins containing tandem repeats of a PPR motif are sequence-specific RNA- or DNA-binding proteins (Lahmy et al. 2000; Small and Peeters 2000). In the Arabidopsis genome, about two-thirds of these proteins are predicted to be targeted to organelles (Small and Peeters 2000). These findings suggest the possibility that PPR proteins are involved in RNA metabolism in organelles. In many plant species, Rf genes alter transcription profiles in CMS mitochondria. In fact, one PPR protein, the product of the Rf-PPR592 gene, can restore CMS in Petunia (Bentolila et al. 2002).

The Rf-1A gene identified here encodes a 791-amino acid protein (Rf-1A) consisting of a signal peptide that targeted mitochondria and 16 repeats of a PPR motif. Since the Rf-1 gene is expected to play a role in processing the transcript from an atp6/orf79 region in mitochondria (Iwabuchi et al. 1993; Akagi et al. 1994), there may be no discrepancy between the expected function of the Rf-1 gene and the deduced function of the Rf-1A gene in processing atp6/orf79 transcript and suppressing BT-type CMS.

In Petunia, the restorer gene Rf-PPR592 is expressed only in floral buds (Bentolila et al. 2002). In the case of rice, the Rf-1A gene, which restores BT-type CMS, is expressed not only in an organ that contains immature pollen but also in a vegetative organ (Fig. 5). Furthermore, the atp6/orf79 transcript was also found to be processed in the presence of the Rf-1 gene in callus (Akagi et al. 1994). The production of intact atp6 mRNA would not be essential to the somatic growth of plants but would be essential for the maturation of pollen. Considerable energy is required for the maturation of pollen, and this energy can be generated by an intact form of atp6 mRNA and/or a sufficient amount of atp6; if not, aborted pollen results.

Co-evolution of Rf genes and mitochondrial genome

In the Nipponbare genome, nine regions of sequences homologous to the Rf-1A gene were found adjacent to the Rf-1 locus (Fig. 6). Their highly conserved sequences suggest that they might have evolved by replication from their ancestor during rice evolution. Some of these sequences could encode functional PPR proteins even though the Rf-1A gene and the Rf-1B gene have lost their function in the Nipponbare genome. In Brassica, two kinds of cytoplasm, nap and pol, cause CMS. A nuclear restorer gene called Rfp is able to modify transcripts of the orf224/atp6 region in pol cytoplasm. On the other hand, nap CMS is restored by another nuclear gene, Rfn (Mmt), which is involved in modifying three kinds of transcripts from different mitochondrial regions, including nap CMS-associated orf222. Based on the map positions of Rfn and Rfp, these restorer genes are located on the same nuclear locus (Singh et al. 1996; Li et al. 1998). In Sorghum A3 CMS, genes related to processing of the orf107 and urf209 transcripts were found to co-segregate, indicating that they may be a single nuclear gene or tightly linked genes (Tang et al. 1998). Two duplicate genes, Rf-PPR592 and Rf-PPR591, were located at the Rf locus in Petunia. While Rf-PPR592 was able to restore petunia CMS, the function of Rf-PPR591 was unclear (Bentolila et al. 2002). Koizuka et al. (2003) suggested that only a four amino acid substitution in the PPR of the Rf gene of CMS Kosena radish changed the gene function. In rice, at least four copies of PPR genes exist around the Rf-1 locus, and these vary somewhat with respect to their amino acid sequence. The Rf-1A gene and Rf-1B gene showed a loss of function in MTC-10A (Fig. 3). The Rf-1B gene lost its function in MTC-10R but not in other restorer lines, such as IR24, IR36 and MTC-18R. Such diverse and duplicate PPR genes may account for specialized functions in processing specific transcripts in rice.

Relationship between duplicate Rf-1 gene homologs and CMS in rice

Within rice, there are several CMS/Rf systems (Virmani and Shinjyo 1988). The cytoplasm derived from wild rice, called wild abortive, causes WA-type CMS in a sporophytic manner and is widely used for the indica subspecies. Two fertility restorer genes, Rf3 and Rf4, are required for the production of viable pollen in WA-type CMS. Rf3 and Rf4 have been mapped to chromosomes 1 and 10, respectively (Yao et al. 1997; Zhang et al. 1997). The latter is located adjacent to the Rf-1 gene (Tan et al. 1998; Jing et al. 2001). Since MTC-10R (referred to as T65R in the literature) does not restore WA-type CMS (Teng and Shen 1994), the Rf-1A gene is not involved in the restoration of WA-type CMS. Although the Rf-1B gene, located at the delimited Rf-1 locus, is nonfunctional in MTC-10R, it did encode a functional protein in other restorer lines, such as IR24, IR36 and MTC-18R (Fig. 5). Unlike MTC-10R, IR24 and MTC-18R can also restore WA-type CMS. Since the structure of the functional Rf-1B protein is almost the same as that of Rf-1A protein, Rf-1B might target mtRNAs other than the atp6/orf79 transcript and restore WA-type CMS; thus, it may be Rf4.

Recently, Kazama and Toriyama (2003) reported that a gene (PPR8-1) encoding a PPR protein is a candidate for the Rf-1 gene based on results obtained using a transgenic strategy. They demonstrated that this gene is involved in the processing of the transcript in mitochondrial atp6 in a transgenic rice callus. However, they did not confirm that PPR8-1 restored fertility in regenerated plants. In our study, we were able to show that Rf-1A is the Rf-1 gene by applying a precise positional cloning strategy. In our analysis, only segregates that carried a limited region around the Rf-1A gene had fertile seeds, and we presumed that Rf-1A restores fertility through processing of the atp6/orf79 transcript. The nucleotide sequences of Rf-1A and PPR8–1 were identical even in their flanking regions, indicating these are the same gene. Putting these two results together, it can be concluded that the product of the Rf-1A gene processes the transcript from the atp6/orf79 region, subsequently restoring fertility as a result, and that the Rf-1 gene itself is present in diversified forms among restorers for BT-type CMS.