The broad-spectrum rice blast resistance gene Pi9 was cloned using a map-based cloning strategy. Sequencing of a 76-kb bacterial artificial chromosome (BAC) contig spanning the Pi9 locus led to identification of six tandemly arranged resistance-like genes with a nucleotide-binding site (NBS) and leucine-rich repeats (LRRs) (Nbs1-Pi9–Nbs6-Pi9). Analysis of selected Pi9 deletion mutants and transformation of a 45-kb fragment from the BAC contig into the susceptible rice cultivar TP309 narrowed down Pi9 to the candidate genes Nbs2-Pi9 and Nbs3-Pi9. Disease evaluation of the transgenic lines carrying the individual candidate genes confirmed that Nbs2-Pi9 is the Pi9 gene. Sequence comparison analysis revealed that the six paralogs at the Pi9 locus belong to four classes and gene duplication might be one of the major evolutionary forces contributing to the formation of the NBS–LRR gene cluster. Semiquantitative reverse transcriptase (RT)–PCR analysis showed that Pi9 was constitutively expressed in the Pi9-resistant plants and was not induced by blast infection. The cloned Pi9 gene provides a starting point to elucidate the molecular basis of the broad-spectrum disease resistance and the evolutionary mechanisms of blast resistance gene clusters in rice.
DURING the last decade, >40 plant disease resistance (R) genes have been cloned from different plant species (Martin et al. 2003). Although cloned R genes confer resistance to a wide range of pathogens (fungi, viruses, bacteria, and nematodes), they share various conserved motifs, suggesting the existence of a common defense signal transduction pathway in different plant–microbe interaction systems (Dangl and Jones 2001; Martin et al. 2003). In general, the R genes fall into six distinct classes, the most prevalent of which is the nucleotide-binding site plus leucine-rich repeat (NBS–LRR) genes (Martin et al. 2003). The NBS domain of NBS–LRR proteins contains a number of conserved motifs, such as a central domain that is predicted to function as a nucleotide-binding site, kinase-1a or P-loop, kinase 2, and kinase 3a, and it may affect R protein function through nucleotide binding, hydrolysis, and the control of cell death (Martin et al. 2003). The LRR domains are generally thought to be involved in the interaction with avirulence (AVR) proteins and to be the major determinant of resistance specificity (Hulbert et al. 2001). Direct evidence of an interaction between the LRR domain of an R protein and its cognate AVR protein was provided by Jia et al. (2000) in experiments with the rice blast resistance protein Pi-ta and the AVR–Pi-ta from the fungal pathogen Magnaporthe grisea. Supporting evidence was subsequently provided by allelic comparisons and domain-swapping experiments between different alleles at the L and P loci of flax (Ellis et al. 1999; Dodds et al. 2001). However, regions other than the LRR may also be involved in resistance specificity (Luck et al. 2000). For example, the proteins coded by the L6 and L7 genes for flax rust resistance, which have distinct race-specific resistance, are identical in the NBS–LRR region and differ at 11 residues in the N-terminal Toll/interleukin-1 receptor (TIR) domain (Ellis et al. 1999).
Clusters of R genes have been identified in diverse plant species (Islam and Shepherd 1991; Jones et al. 1993; Song et al. 1995, 1997; Kunkel 1996; Salmeron et al. 1996; Ellis et al. 1997; Meyers et al. 1998; Michelmore and Meyers 1998; Richter and Ronald 2000; Wei et al. 2002). The majority of the 149 NBS–LRR genes occurring in the Arabidopsis genome are clustered (Meyers et al. 2003) as are the >600 NBS–LRR genes identified in the rice genome (Bai et al. 2002). Moreover, the clustered R genes usually fall into heterogeneous groups on the basis of their structural similarity, suggesting that each group may have evolved independently. Sequence analysis of a 261-kb region of the Mla locus in barley cv. Morex revealed that eight coiled-coil (CC)–NBS–LRR R gene homologs are classified into three dissimilar families, with <43% amino acid sequence similarity between families and 78–100% similarity within families (Wei et al. 2002). The rice Xa21 gene, encoding a receptor-like kinase and conferring resistance to bacterial blight, is from a multigene family that contains two distinct classes of genes (Song et al. 1995, 1997; Wang et al. 1998). Four LRR receptor-like genes have recently been identified in a 67.2-kb region of the newly cloned bacterial blight R gene Xa26 locus (Sun et al. 2004).
Rice blast is the most economically important fungal disease of rice because of its worldwide distribution (Ou 1985). Over 25 blast R genes have been mapped on the rice genome, many of which are allelic or closely linked (Kiyosawa 1989; Inukai et al. 1994; Wang et al. 1994; Pan et al. 1998; Chao et al. 1999). For example, 5 blast R genes have been identified at the Pi-k locus of chromosome 11 (Kiyosawa 1989; Inukai et al. 1994). Pi-ta and Pi-ta2 are allelic or at least very close to each other in the centromere region of chromosome 12 (Rybka et al. 1997), while Pi5(t) and Pi3(t) map at the same location on rice chromosome 5 (Inukai et al. 1996; Jeon et al. 2003). To date, only 2 blast R genes, Pib (Wang et al. 1999) and Pi-ta (Bryan et al. 2000), have been cloned. Pib belongs to the NBS–LRR class of genes and has been predicted to encode a cytoplasmic protein (Wang et al. 1999). Pi-ta differs from Pib and other NBS–LRR genes in that the protein predicted to be encoded by Pi-ta lacks a classic LRR in its C-terminal region, containing instead a highly imperfect repeating structure with 10 repeats of various lengths (from 16 to 75 amino acids), referred to as a leucine-rich domain (LRD) (Bryan et al. 2000).
Broad-spectrum disease resistance genes have been used for disease control in many different crops. There are two definitions for broad-spectrum disease resistance. The first one is defined as the resistance to the majority of geographically different isolates of the same pathogen. The second type is the resistance to two or more unrelated pathogens. Whether a broad-spectrum resistance gene is durable or not in multiple locations during a relatively long time is still debatable. To date, several first types of broad-spectrum R genes have been cloned in plants. Xa21 is the R gene cloned with a high level of resistance to many Xanthomonas oryzae pv. oryzae strains (Wang et al. 1996). The recessive mutations (mlo) of the barley Mlo locus mediate a broad-spectrum resistance to all known isolates of powdery mildew fungus (Blumeria graminis f. sp. hordei) (Buschges et al. 1997). mlo-based disease resistance involves a spontaneous-lesion phenotype and cell-wall deposition in epidermal tissues preceding pathogen attack. Recently, the same R gene that was denoted RB by Song et al. (2003) and RPI (Rpi-blb1) by Van Der Vossen et al. (2003) was cloned from an NBS–LRR gene cluster. RB/RPI confers broad-spectrum resistance to all known races of the late blight pathogen Phytophthora infestans. The Arabidopsis RPW8 gene (Xiao et al. 2001) belongs to the second type of broad-spectrum R gene and confers resistance to different isolates of the same powdery mildew pathogens as well as to different powdery mildew fungi. The protein it encodes is a small membrane protein with a putative coiled-coil domain that has limited homology to the N terminus of an NBS–LRR gene (Xiao et al. 2001). RPW8 is not involved in the gene-for-gene interaction and may interact with different AVR genes from different pathogens.
The resistance in newly released rice cultivars to rice blast caused by M. grisea can be lost quickly due to the high level of instability in the genome of this fungus (Bonman et al. 1992). One way to overcome this problem is pyramiding of multiple R genes, each recognizing a unique set of M. grisea isolates, into a single cultivar. The deployment of rice cultivars with broad-spectrum resistance is another practical means of controlling the fungal pathogen (Bonman et al. 1992). The Pi9 gene existing in the indica rice line 75-1-127 (Liu et al., 2002) was introgressed from the wild species Oryza minuta (Sitch et al. 1989; Amante-Bordeos et al. 1992). Different from the rice lines carrying the blast R genes Pib and Pi-ta, 75-1-127 has broad-spectrum resistance to diverse M. grisea isolates. For example, 75-1-127 was tested with >100 Philippine M. grisea isolates at the International Rice Research Institute (IRRI) and no compatible isolates were identified (H. Leung, personal communication). In our previous study, 75-1-127 was highly resistant to 43 M. grisea isolates collected from 13 countries, and the Pi9 locus was fine mapped on rice chromosome 6 (Liu et al. 2002). Here, we report the map-based cloning of the Pi9 gene from a 76-kb region of an NBS–LRR multigene family. The application of multiple approaches, such as genomic sequencing of the targeted region, characterization of the Pi9-susceptible mutants, and the transformation of large-insert transformation-competent artificial chromosome (TAC) clones, facilitated our cloning effort. Functional analysis of the candidate genes and comparative analysis of the genomic structure of individual members of the Pi9 gene cluster provide a starting point to elucidate the molecular basis of broad-spectrum disease resistance specificity and the evolutionary mechanisms of rice R gene clusters.
MATERIALS AND METHODS
DNA sequencing of rice BAC clones:
Two BAC clones, 75-1-127BAC12 and 75-1-127BAC3, were fully sequenced using a shotgun method. Briefly, purified BAC DNA was sonicated with a sonicator. Sheared DNA fragments were size selected on an agarose gel and the 1.5- to 3-kb fragments were used in constructing a subclone library. The damaged DNA ends were repaired using T4 DNA polymerase (Roche, Germany) and ligated to the pBluescript (KS) vector that had been previously digested with SmaI and treated with shrimp alkaline phosphatase (Roche, Germany). Approximately 700 individual clones from the 75-1-127BAC12 shotgun library and 450 individual clones from the 75-1-127BAC3 shotgun library were sequenced. Both ends were sequenced and all the readings were assembled with the Phred and Phrap software packages (Ewing and Green 1998; Ewing et al. 1998; Gordon et al. 1998; http://www.phred.org). At ∼10 times redundancy, the sequence gaps as well as the low-quality regions of the consensus sequences were filled or resequenced and subsequently used for sequence assembly.
Computational analysis of DNA and protein sequences:
For gene prediction from the rice genomic sequences, we ran GENSCAN (Burge and Karlin 1997; http://genes.mit.edu/GENSCAN.html) with the Arabidopsis organism option. The potential coding regions in the 76-kb DNA sequence were further searched against the sequences deposited in GenBank, using the BLAST programs (blastn, blastx, and blastp; Altschul et al. 1990, 1997; Gish and States 1993; http://www.ncbi.nlm.nih.gov/BLAST/). Pairwise comparisons between genomic sequences of the coding regions of Pi9 candidate genes were performed using the BLAST program (bl2seq) (http://www.ncbi.nlm.nih.gov/blast/bl2seq/bl2.html) and the Matcher program (http://bioweb.pasteur.fr/seqanal/interfaces/matcher.html). Alignment of amino acid sequences was performed with the GAP program in the Wisconsin Package (version 9.1; Genetics Computer Group, Madison, WI). Multiple alignment of DNA sequences was conducted using ClustalX version 1.83 (Thompson et al. 1997). On the basis of the results of ClustalX analysis, a phylogenetic tree was generated using the program TREEVIEW (http://taxonomy.zoology.gla.ac.uk/rod/treeview.html; Page 1996). The protein motif search was performed using the ScanProsite program (Sigrist et al. 2002; http://us.expasy.org/prosite/). The theoretical isoelectric point (pI) and protein molecular weight were computed as described (Wilkins et al. 1998; http://sosnick.uchicago.edu/pi_tool.html).
Vector construction and rice transformation:
TAC vector pRTAC8 (Qu et al. 2003) was used for rice transformation of large genomic fragments. The DNA of 75-1-127BAC12 (Liu et al. 2002) was digested with NotI to obtain a 45-kb and a 12.5-kb genomic fragment (Figure 1A, 1–45,348 bp and 45,509–58,068 bp, respectively). The 45-kb fragment was cloned into pRTAC8 to make transformation constructs pRTAC8-45 kb (I) and pRTAC8-45 kb (II) (Figure 1B), which contained the 45-kb inserts in different orientations (Qu et al. 2003). pNBS4, the Nbs4-Pi9 gene construct, was made by cloning the 12.5-kb NotI fragment into pRTAC8. The Nbs5-Pi9 gene construct (Figure 1B, pNBS5) was prepared by HindIII partial digestion of 75-1-127BAC3 (Liu et al. 2002) and cloning of the 24.7-kb genomic fragment (Figure 1A, 49,808–74,581 bp) into the pRTAC8 HindIII site.
Transformation constructs of the individual Nbs1-Pi9, Nbs2-Pi9, and Nbs3-Pi9 genes were constructed using the pCAMBIA1301 vector (GenBank accession no. AF234297; www.cambia.org). Both pNBS1-1 and pNBS1-2 (Figure 1B) were made for rice transformation of the Nbs1-Pi9 gene. To construct pNBS1-1, the 10-kb PstI fragment (Figure 1A, 13,613–23,605 bp) of 75-1-127BAC12 was cloned into the PstI site of pCAMBIA1301. The pNBS1-2 construct was made by cloning the 6.9-kb SalI fragment (Figure 1A, 12,391–19,301 bp) into the SalI-digested pCAMBIA1301. pNBS2 (Figure 1B), the Nbs2-Pi9 gene construct, was constructed by cloning the 13.5-kb SalI fragment (Figure 1A, 32,363–45,848 bp) into the SalI site of pCAMBIA1301. The Nbs3-Pi9 construct, named pNBS3 (Figure 1B), was made by cloning the 14.6-kb HindIII fragment (Figure 1A, 18,395–33,070 bp) of 75-1-127BAC12 into the HindIII site of pCAMBIA1301.
Rice callus was induced from the embryos of the mature seeds of japonica rice cultivar TP309. Rice transformation was conducted using the Agrobacterium-mediated method (Hiei et al. 1994; Yin and Wang 2000; Qu et al. 2003).
Generation and screening of the Pi9-deletion mutants:
The Pi9 parental line 75-1-127 was mutated with the chemical mutagen diepoxybutane (DEB), using the method described by Wang et al. (2004). A total of 20,000 seeds of 75-1-127 were soaked in water overnight. The seeds were separated into two groups and treated with DEB for 4 hr, one group at 0.004% DEB and the other at 0.006% DEB. Treated seeds were washed thoroughly five times with water and sown in soil. About 60% of the seeds germinated after the 0.006% DEB treatment and 70% germinated after the 0.004% DEB treatment. M1 plants from the DEB-induced populations were allowed to self-pollinate, and the seeds of ∼12,000 M1 plants were harvested from both treatments. The resulting M2 seeds were harvested separately from each M1 line. Three-week-old M2 plants were inoculated with the M. grisea isolate PO6-6. Plants with visible lesions were picked out 6 days after inoculation. The susceptible phenotype in each susceptible M2 plant was confirmed by blast inoculation of the M3 plants. For identification of deletion mutations in the Pi9 region, genomic DNA was isolated from the M3 plants and PCR amplified with the primers (Table 3) specific to the individual candidate genes in the Pi9 gene cluster.
Blast fungus inoculation:
For blast inoculation, 3-week-old seedlings of parental and transgenic lines were inoculated with the M. grisea fungus as described (Liu et al. 2002). For blast inoculation of primary (T1) transformed plants, the regenerated plants were kept in MS media and sprayed with a spore suspension (1 × 105 spores/ml) of the M. grisea isolate PO-6-6. The inoculated plants were placed into sealed containers to maintain humidity and grown in the dark for 24 hr. The plants were then transferred to a growth chamber and grown under a 12/12-hr (light/dark) photoperiod. Plants of the transformation recipient cultivar TP309 were used as a susceptible control. The disease reaction was examined 6–7 days after inoculation and scored as described (Liu et al. 2002).
Genomic DNA isolation and Southern blot hybridization:
Rice genomic DNA was isolated using the CTAB method (Saghai-Maroof et al. 1994). Southern analysis was done according to standard methods (Sambrook et al. 1989). For preparation of the Nbs2-Pi9 hybridization probe, the 928-bp genomic fragment (Figure 1A, 40,350–41,278 bp) of the Nbs2-Pi9 gene was PCR amplified using Pi9 NBS2-O (5′-TACAACCACCTACCATCCCAT-3′) and Pi9 NBS2-U (5′-TCTAGAACCTGCAAGTCTCG-3′) as specific primers and the pNBS2 construct DNA as template.
RNA isolation, mRNA purification, and semiquantitative RT–PCR:
For full-length cDNA cloning, 3-week-old 75-1-127 plants were inoculated with M. grisea isolate PO6-6 (Liu et al. 2002). Total RNA was isolated from infected leaf tissue at 24 hr after inoculation (HAI) using Trizol (Invitrogen, San Diego). mRNA was purified with the Oligotex mRNA midi kit (catalog no. 70042; QIAGEN, Valencia, CA). cDNA was synthesized from poly(A)+ RNA using the cDNA synthesis kit (Stratagene, La Jolla, CA) and used as template in two rounds of nested PCR. The primers for the first-round amplification were Pi9E (5′-CTTGAAGGGAGAGTCGAACGAA-3′) and Pi9-nbs2b (5′-TGTAATGATCAAGCAATATCTGGCT-3′). The Pi9E sequence (Figure 1A, 33,700–33,721 bp) is in the Nbs2-Pi9 5′-UTR region and 25 bp upstream of the Nbs2-Pi9 start codon. The Pi9-nbs2b is complementary to the Nbs2-Pi9 3′–UTR sequence (Figure 1A, 43,797–43,821 bp). The forward primer of the second-round PCR was Pi9FSAL (5′-ATCGGTCGACATGGCGGAGACGGTGCTGAG-3′) and the reverse primer was Pi9DRI (5′-CTGAGAATTCCCCTTGAGAAATTCGCCG -3′). The nested reverse primers Pi9-nbs2b and Pi9DRI were designed in such a way that they matched perfectly with the Pi9 sequence but had a 2-bp mismatch with Nbs5-Pi9 at the 3′ end of each primer. Pi9FSAL contains the SalI recognition sequence and 20 bp of the 5′-Nbs2-Pi9-coding sequence (Figure 1A, 33,725–33,744 bp). Pi9DRI contains the EcoRI site and the complementary sequence of 18 bp of the Nbs2-Pi9 3′-UTR sequence (Figure 1A, 43,639–43,656 bp). The amplified product of the first-round PCR was diluted 50 times with H2O, and the amplicon was further used as a template for the second-round PCR. The first-round PCR was initiated by one cycle of 95° for 5 min, followed by 25 PCR cycles of 95° for 30 sec, 62° for 40 sec, and 72° for 4 min, with a final 10-min hold at 72°. The conditions of the second-round PCR were identical to those in the first-round PCR except that the annealing temperature was 63°. Platinum Taq DNA polymerase (Invitrogen) was used in the PCR for cloning Pi9 full-length cDNA. The amplification was performed in a 50-μl PCR reaction volume containing 1× PCR buffer (20 mm Tris–HCl, pH 8.4; 50 mm KCl); 0.2 μm forward and reverse primers; 0.2 mm dATP, dCTP, dGTP, and dTTP; 1.5 mm MgCl2; and 2.5 units of Platinum Taq DNA polymerase. The amplified cDNA fragment was digested with SalI and EcoRI and cloned into pCAMBIA-1300.
For semiquantitative RT–PCR analysis of Pi9 expression, seeds of the Pi9 parental line 75-1-127 were dehusked and sterilized in 40% bleach for 40 min and then germinated on 1/2 MS medium in tissue culture vessels. Twelve-day-old plants were inoculated with M. grisea isolate PO-6-6, and leaf tissue was collected at 0, 6, 12, 24, 48, and 72 HAI. Total RNA was extracted using the RNeasy plant mini kit (QIAGEN). One microgram of total RNA was pretreated with RNase-free DNase I (Ambion, Austin, TX) and subjected to reverse transcription using the Reverse Transcription System (Promega, Madison, WI). Semiquantitative RT–PCR with Pi9-specific primers NBS2-G (5′-TGCCCAACCTTTACCCACTGTA-3′) and NBS2-H (5′-AACATGAGTAGAAACAAATTAGTTTG-3′) (Figure 6A) was performed with 26, 28, and 30 cycles. The control PCR of the Actin 1 gene (Reece et al. 1990) was performed using the forward primer 5′-CGTCTGCGATAATGGAACTGG-3′ and reverse primer 5′-CTGCTGGAATGTGCTGAGAGAT-3′ with 24 cycles. The RT–PCR was initiated with one cycle at 95° for 5 min, followed by 26, 28, or 30 cycles at 94° for 30 sec, 62° for 40 sec, and 72° for 1 min 10 sec, and terminated with extension at 72° for 10 min. The PCR products were resolved on 1% agarose gels.
An NBS–LRR gene cluster was identified by sequence analysis of a 76-kb genomic region from the Pi9 locus:
A bacterial artificial chromosome (BAC) contig had been previously constructed using the pB8-hybridizing BACs isolated from the BAC library of Pi9-carrying rice line 75-1-127 (Liu et al. 2002). To obtain sequence information at the Pi9 locus, we fully sequenced two of the BAC clones, 75-1-127BAC12 and 75-1-127BAC3, using the shotgun sequencing approach. Sequence analysis revealed that the genomic fragments are 58,068 and 40,075 bp in 75-1-127BAC12 and 75-1-127BAC3, respectively, and they overlap in a 21,872-bp region, thus forming a contig of 76,272 bp (Figure 1A).
Two different approaches were used to identify the open reading frames (ORFs) of the total genomic sequence. The gene prediction program GENSCAN (Arabidposis gene model) (Burge and Karlin 1997; http://genes.mit.edu/GENSCAN.html) was used to identify the putative coding sequence (CDS) in the 76-kb region, and the BLAST program was used for homology searches to confirm the gene prediction results because cloned plant NBS–LRR R genes are quite conserved. Seven putative genes were identified from the 76-kb sequence. The first gene in the 76-kb region (between 10,489 and 12,966 bp), located at the SP6 end in 75-1-127BAC12, is a homolog of the maize gene that encodes a putative nitrate-induced NOI protein (GenBank accession no. AF030385). The other six genes, denoted Nbs1-Pi9–Nbs6-Pi9 (Figure 1A), were considered to be Pi9 candidate genes because they all have high homology to the NBS–LRR disease R genes cloned from various plant species (Bent 1996; Dangl and Jones 2001; Martin et al. 2003). The Nbs1-Pi9 genomic region (Figure 1A) shares a high homology (2942/2984; 98% identity) with rice cDNA clone J023131G18 (GenBank accession no. AK121397.1). The gene prediction for Nbs3-Pi9 was partially confirmed on the basis of a 2351-bp 3′-partial cDNA fragment of Nbs3-Pi9, which was isolated from a 75-1-127 cDNA library using an Nbs1-Pi9 DNA fragment as the hybridization probe; the Nbs5-Pi9 gene prediction result was confirmed by the Nbs5-Pi9 full-length cDNA cloned by RT–PCR (data not shown). The Nbs4-Pi9 genomic sequence (Figure 1A) has a homology (1951/2066; 94% identity) with rice cDNA clone J013122I17 (GenBank accession no. AK067966.1). Nbs4-Pi9 appears to be a pseudogene because there are four stop codons in the coding region. Nbs6-Pi9 is located at the 3′ region of the 76-kb sequence and appears to be incomplete because its 3′ end (LRR region) is truncated and a solo-long terminal repeat (LTR) fragment is inserted in the 5′ region (Figure 1A, 69,147–72,143 bp). This solo-LTR fragment has 94% identity with the LTR of rice gypsy-type retrotransposon, RIRE8 (Kumekawa et al. 1999). The putative coding region of Nbs6-Pi9 (Figure 1A, 68,550–69,146 bp and 72,144–76,272 bp) shares 94% identity with rice cDNA clone J013122I17.
Sequence comparison and phylogenetic analysis revealed that the six Pi9 NBS–LRR genes belong to four types of paralogs:
Pairwise comparison between the putative coding regions of the six paralogs was performed using both the Matcher (http://bioweb.pasteur.fr/seqanal/interfaces/matcher.html) and the BLAST (bl2seq) (http://www.ncbi.nlm.nih.gov/blast/bl2seq/bl2.html) programs (Table 1 ). The solo-LTR insertion sequence (Figure 1A) in Nbs6-Pi9 was removed before performing the BLAST search, and the two cloned blast R genes, Pib and Pi-ta, were also included in the comparison analysis. The sequence identity derived from the Matcher program was slightly lower than that from the BLAST program. The identity between the six paralogs ranged from 63.8 to 98.6% (Matcher) and from 71.8 to 98.6% (BLAST). Interestingly, the sequence homology between Nbs2-Pi9 and Nbs5-Pi9 (98.6% from both methods) and between Nbs4-Pi9 and Nbs6-Pi9 (91.9% from Matcher and 95.4% from BLAST) was much higher than that between the remaining pairs (Table 1). The promoter sequences of the two pairs were also highly homologous. These results suggest that the genomic fragment containing the Nbs5-Pi9 and Nbs6-Pi9 genes may be the duplication of the fragment containing the Nbs2-Pi9 and Nbs4-Pi9 genes.
To further investigate the relationship among the six Pi9 candidate genes and their relationship with the other cloned blast R genes, the putative coding sequences of the six Pi9 candidate genes and the rice blast R genes Pib (Wang et al. 1999) and Pi-ta (Bryan et al. 2000) were compared by ClustalX multiple alignment and phylogenetic analysis (Figure 2). As shown in the phylogenetic tree, the degree of homology among these genes varies considerably, with the genes falling into five heterogeneous groups: Nbs1-Pi9 (I), Nbs4-Pi9/Nbs6-Pi9 (II), Nbs2-Pi9/Nbs5-Pi9 (III), Nbs3-Pi9 (IV), and Pib/Pi-ta (V). Therefore, the six NBS–LRR genes in the 76-kb Pi9 multigene cluster were classified into four groups.
The protein sequences between the candidate genes were compared after the nucleotide sequences of Nbs1-Pi9, Nbs2-Pi9, Nbs3-Pi9, and Nbs5-Pi9 were translated into amino acid sequences (Table 2). Neither Nbs4-Pi9 nor Nbs6-Pi9 was included in the analysis because neither one is likely to be an expressed gene due to the presence of four stop codons (Nbs4-Pi9) and a solo-LTR insertion (Nbs6-Pi9). The protein sequence identity among the pairs ranged from 55.2 to 98.1%, and the similarity among them ranged from 63.5 to 98.1%. Both the identity and similarity between Nbs2-Pi9 and Nbs5-Pi9 proteins were much higher than those of the other pairs (Table 2), further supporting the possibility of duplication in the Nbs2-Pi9 and Nbs5-Pi9 genomic regions.
Analysis of Pi9 deletion mutants and large-insert DNA transformation localized the Pi9 gene to the genomic region containing Nbs2-Pi9 and Nbs3-Pi9:
To narrow down the location of the Pi9 gene in the gene cluster, the first approach was large-scale mutagenesis of the Pi9 parental line 75-1-127 followed by PCR analysis of the deletions in the candidate genes in the susceptible mutants. More than 50 M2 plants susceptible to M. grisea isolate PO6-6 were identified from a population of 12,000 M1 plants. Twenty M3 plants (Figure 3, 15 susceptible and 5 resistant) from six M2 families that segregated for resistance to PO6-6 were analyzed by PCR using primers specific to each of the Pi9 candidate genes (Table 3).
When the genomic DNA of the 20 M3 plants was analyzed by PCR using the Nbs2-Pi9 primers and Nbs3-Pi9 primers, respectively, deletions in the two genes were detected in the 15 susceptible mutants (Figure 3). PCR analysis using primers specific to Nbs1-Pi9, Nbs4-Pi9, Nbs5-Pi9, or Nbs6-Pi9 (Table 3) indicated that those mutants carried deletion mutations of varying sizes in the 76-kb Pi9 region (Figure 3). The majority of the mutants had a large deletion located between Nbs1-Pi9 and Nbs6-Pi9. The mutant M0566-2-5-r was resistant although a mutation was observed in the 3′ region of the Nbs1-Pi9 gene, indicating that Nbs1-Pi9 may not be the Pi9 gene. In the PCR analysis of Nbs5-Pi9, 11 of the 15 susceptible mutants contained the Nbs5-Pi9 region, suggesting that it also may not be the Pi9 gene. Either Nbs4-Pi9 or Nbs6-Pi9 was not likely the Pi9 gene because the Nbs4-Pi9 region was detected in PCR of the susceptible mutant M0599-2-2-2 and the Nbs6-Pi9 region was detected in 12 susceptible mutants (Figure 3). On the basis of all these results, we concluded that Pi9 is located in the genomic region containing Nbs2-Pi9 and Nbs3-Pi9.
Simultaneously, large-insert complementation was carried out by transformation of a 45-kb genomic fragment containing Nbs1-Pi9, Nbs3-Pi9, and Nbs2-Pi9 (Figure 1B). The susceptible rice cultivar TP309 was transformed with Agrobacterium containing the constructs pRTAC8-45 kb (I) and pRTAC8-45 kb (II) (Figure 1B), respectively. Among the 48 T2 transgenic lines obtained, one line (TAC106) transformed with pRTAC8-45 kb (II) showed a segregation of resistance and susceptibility to M. grisea isolate PO6-6. Southern blot analysis revealed that the Nbs1-Pi9/Nbs3-Pi9/Nbs2-Pi9 region was transformed into the resistant transgenic line (data not shown), suggesting that Pi9 is located in the region spanning Nbs1-Pi9, Nbs2-Pi9, and Nbs3-Pi9. The results from the large-insert transformation experiments confirmed that Nbs2-Pi9 and Nbs3-Pi9 remain to be the only candidates of the Pi9 gene.
Rice transformation with individual candidate genes determined that Nbs2-Pi9 is Pi9 and that the Pi9 transgenic line has the exact resistance spectrum with the Pi9 donor line:
To pinpoint the Pi9 gene in the genomic region, we made two genomic constructs of the Nbs2-Pi9 and Nbs3-Pi9 genes. The pNBS3 construct (Figure 1B) contained a 14.6-kb (Figure 1A) genomic fragment comprising the Nbs3-Pi9 gene. T1 plants of 19 independently transformed lines were susceptible to PO-6-6, and all of the T2 plants from nine transgenic lines were highly susceptible to PO-6-6. For rice transformation of Nbs2-Pi9, we developed the pNBS2 construct (Figures 1B and 4C), which contained a 13.5-kb SalI genomic fragment (Figure 1B). This 13.5-kb fragment contained 1362 bp of the Nbs2-Pi9 5′-UTR region (32,962–33,724 bp), the whole Nbs2-Pi9 coding sequence, and 1804 bp of the 3′-UTR sequence (42,314–44,117 bp). Of the 19 T1 transgenic plants we inoculated with M. grisea isolate PO6-6, 13 were resistant and 6 were susceptible. We subsequently evaluated T2 plants from 3 resistant T1 lines (nos. 10, 12, and 77) and 1 susceptible T1 line (no. 8) for resistance to PO6-6 and observed the segregation of resistant and susceptible plants in all three resistant lines. As shown in Figure 4A, plant no. 12-2R from the no. 12 T2 line was highly resistant to PO6-6, while plant no. 12-1S from the same line was highly susceptible.
Cosegregation of the transgene and blast resistance was investigated in the T2 generation of two Nbs2-Pi9 transgenic lines (nos. 12 and 77; data not shown). The T2 plants of transgenic lines 12 and 77 were inoculated with isolate PO6-6 and examined using the β-glucuronidase (GUS) enzymatic assay (Jefferson 1987). The GUS activity observed in the transgenic plants indicated the presence of the Nbs2-Pi9 transgene because the T-DNA of transformation construct pNBS2 (Figure 4C) carries the GUS gene. In 44 T2 plants of line 12, 37 plants were GUS-positive and resistant to PO6-6 (GUS+/R), and 7 plants were GUS-negative and susceptible to PO6-6 (GUS−/S). Similarly, 15 GUS+/R plants and 8 GUS−/S plants were identified in 23 T2 plants of line 77. A chi-square test showed that the ratios of 37:7 (line 12) and 15:8 (line 77) fit the Mendelian 3:1 segregation ratio (both at the 95% level), suggesting the existence of a single T-DNA copy in these transgenic plants.
Cosegregation was further confirmed in transgenic line 12 using Southern blot analyses. Genomic DNA was extracted from 10 T2 plants of the no. 12 line, digested with EcoRI, and probed with a 928-bp fragment from Nbs2-Pi9 (Figure 4C). The 928-bp fragment is located within the second exon of the putative Nbs2-Pi9-coding sequence (Figure 1A) and corresponds to the sequence encoding the last 151 amino acids of the NBS domain and the first 158 amino acids of the LRR domain (Figure 5). As shown in Figure 4B, all seven resistant T2 plants (nos. 12-13, 12-14, 12-18, 12-20, 12-21, 12-29, and 12-30) contained the expected 11-kb Nbs2-Pi9 band, while the three susceptible T2 plants (nos. 12-5, 12-17, and 12-23) did not. The 11-kb hybridizing fragment was derived from the EcoRI digestion of the construct because the 13.5-kb SalI genomic fragment had just one EcoRI site and the other EcoRI site was from the T-DNA region of the transformation vector (Figure 4C).
To verify whether the Pi9 transgenic plants have the same resistance spectrum with the Pi9 donor line 75-1-127, an additional 21 rice blast isolates collected from nine countries were inoculated with both the Pi9 donor cultivar and the transgenic line (no. 12), the transformation recipient cultivar TP309, and the susceptible control cultivar CO39 (Table 4). The inoculation showed that the Pi9 transgenic line had the exact same resistance spectrum with its donor line 75-1-127. It is noteworthy to mention that Pi9 is susceptible to the Korean isolate R01-1 (lesion type 4).
To rule out the presence of any other functional genes in the cluster, we also transformed the susceptible cultivar TP309 with constructs containing the Nbs1-Pi9, Nbs4-Pi9, and Nbs5-Pi9 genes. For Nbs1-Pi9, we made two transformation constructs, pNBS1-1 and pNBS1-2 (Figure 1B), on the basis of the Nbs1-Pi9 coding sequence predicted by GENSCAN. Forty and 28 independent transgenic lines were generated in rice transformation using the constructs pNBS1-1 and pNBS1-2, respectively. All the T2 plants were susceptible to isolate PO-6-6. For rice transformation of Nbs5-Pi9, 12 independent T1 lines were obtained by transforming the pNBS5 construct (Figure 1B). All the T1 and T2 plants were highly susceptible to isolate PO6-6. For Nbs4-Pi9, 46 independent transgenic lines were generated in rice transformation using pNBS4 (Figure 1B) and all the T2 plants were also susceptible to isolate PO6-6. We did not make a complementation test for Nbs6-Pi9 because our previous mapping results (Liu et al. 2002) suggested that Pi9 may be upstream of the Nbs6-Pi9 region, and importantly the Nbs6-Pi9 itself contained a solo-LTR insertion (Figure 1A). Taken together, the inoculation tests of all transgenic lines confirmed that Nbs2-Pi9 is the only functional gene responsible for the broad spectrum of Pi9 at the locus.
Pi9 encodes a putative NBS–LRR protein containing 1032 amino acids:
To isolate the cDNA fragment of Pi9, we prepared total RNA using leaf tissue from 75-1-127 collected 24 hr following blast inoculation. Two rounds of nested PCRs were performed to amplify the Pi9 cDNA using two pairs of Pi9-specific primers (see materials and methods). Because the Nbs5-Pi9 gene shares high homology with Pi9, the first-round reverse primer and second-round reverse primer were designed in such a way that each primer sequence matched perfectly with Pi9 but contained a 2-bp mismatch with Nbs5-Pi9 at the 3′ end of each primer that minimized the amplification of the Nbs5-Pi9 cDNA. Two PCR bands were amplified after the two rounds of PCR. The first band was ∼4 kb, while the second PCR band was <1 kb and might originate from nonspecific amplification. Sequencing the 4-kb RT–PCR fragment indicated that it was a unique amplification product of the Pi9 cDNA. The Pi9 cDNA was 4009 bp, including 3099 bp of Pi9 coding sequence and 910 bp of 3′-UTR. The sequencing result of the Pi9 cDNA was exactly as predicted by the GENSCAN program.
Sequence alignment between the Pi9 cDNA and its genomic sequence revealed that two introns interrupt the coding region. The length of the first intron in the coding region is 5362 bp (Figures 1A and 6A, 33,840–39,201 bp). The second intron is 128 bp (42,153–42,280 bp). We identified a 575-bp intron (Figures 1A and 6A, 42,474–43,048 bp) and a 102-bp intron (43,223–43,324 bp) in the Pi9 3′-UTR region. To find the transcription start site of the Pi9 gene, we then scanned the genomic region preceding the NBS domain using the promoter prediction program available at http://www.cbs.dtu.dk/services/Promoter/ (Knudsen 1999). The result of the scan indicated that the transcription start site of the Pi9 gene is at position base pair 32,962 in the 76-kb region (Figure 1A) and that the 5′-UTR is 763 bp long.
Protein translation of the cDNA sequence revealed that the Pi9 gene encodes a predicted 1032-amino-acid polypeptide (Figure 5) with a molecular weight of 117.05 kDa and an pI of 7.55. The deduced Pi9 protein belongs to the NBS–LRR class of resistance proteins. Three sequences in the NBS domain, GMGGLGKT (Figure 5, positions 193–200), KRYFVILDDLW (positions 277–287), and GSRIVITTRNVDL (positions 307–319), correspond to kinase 1a (P-loop), kinase 2, and kinase 3a, respectively. In the CC domain (positions 1–192), the sequence motif (WAEQIRDLSYDIEDSLDEF, positions 68–86) belongs to the conserved non-TIR (nT) motif of rice NBS–LRR genes (Bai et al. 2002). At the C-terminal region is the LRR domain, which is composed of 17 imperfect LRR repeats.
The Pi9 gene is constitutively expressed in the Pi9-resistant plants:
To analyze the expression profile of the Pi9 gene, semiquantitative reverse transcriptase (RT)-PCR was performed to specifically detect the Pi9 transcript in M. grisea-infected rice plants. Following inoculation with isolate PO6-6, leaf tissue of 75-1-127 was collected at six different time points (0, 6, 12, 24, 48, and 72 HAI). Semiquantitative RT–PCR was performed using the Pi9-specific primer pair (NBS2-G and NBS2-H) that had been designed to amplify a 764-bp 3′-UTR sequence using 26, 28, and 30 RT–PCR cycles (Figure 6A). Only a faint band of the expected size was detected when 26 cycles were used, most likely due to the low expression level of the Pi9 gene (data not shown). However, a stronger band was observed when 28 and 30 cycles were used (Figure 6B). The results from the three semiquantitative RT–PCRs indicated that the expression level of the Pi9 gene in 75-1-127 before blast inoculation (zero time point) remains stable and does not change following inoculation (6–72 HAI). DNA sequencing of the RT–PCR products confirmed the true amplification of the Pi9 cDNA fragment. These results showed that Pi9 is constitutively expressed in the Pi9-carrying plants and is not induced by blast infection.
Map-based cloning of the Pi9 gene from an R gene cluster:
We have cloned Pi9, a broad-spectrum blast R gene in rice, using a map-based cloning strategy. Because the Pi9 parental line 75-1-127 was generated through a wide hybridization between cultivated rice and the wild rice O. minuta following repeated backcrosses (Amante-Bordeos et al. 1992), genetic and physical analysis of the Pi9 locus was a challenging task. Like many other plant R genes (Meyers et al. 1998, 2003; Xiao et al. 2001), Pi9 is a member of a complex R gene family with six candidate genes. To clone the Pi9 gene, we employed three complementary strategies. First, we sequenced a 76-kb genomic region spanning the Pi9 locus and identified six NBS–LRR candidate genes. This provided us with information on the genomic structure of the cluster. Second, we generated a Pi9-deletion population and analyzed the deletion sites within the Pi9 gene cluster in the selected susceptible mutants, using seven pairs of gene-specific primers. PCR analysis indicated that Pi9 is located in the genomic region containing Nbs2-Pi9 and Nbs3-Pi9. At the same time, we transformed the susceptible rice cultivar TP309 with a 45-kb fragment containing Nbs1-Pi9, Nbs2-Pi9, and Nbs3-Pi9. On the basis of the results from both approaches, we were able to localize the Pi9 gene in the Nbs2-Pi9 and Nbs3-Pi9 region. Transformation of the susceptible rice cultivar TP309 using individual constructs of Nbs2-Pi9 and Nbs3-Pi9 and blast inoculation tests of the transgenic plants with a total of 22 isolates finally confirmed that Nbs2-Pi9 is the only functional gene responsible for the broad spectrum of the Pi9 gene. Consequently, we can conclude that the application of complementary strategies such as the sequencing of BAC clones, the analysis of susceptible mutants, and large-insert transformation are useful procedures in map-based cloning of a targeted gene in a complex gene cluster.
The structural features of the Pi9 gene and protein:
On the basis of the deduced amino acid sequence, Pi9 belongs to the NBS–LRR class of R genes. In contrast to the NBS–LRR genes from dicot plants, the NBS–LRR genes from cereals often have introns in the NBS region (Bai et al. 2002). The Pib gene (Wang et al. 1999) has two introns (1340 and 308 bp) in its coding region, while the Pi-ta gene has a 1463-bp single intron in its coding region (Bryan et al. 2000). Although Pi9 also has two introns in its coding region, one of the introns is much larger (5362 bp) than that of the Pib gene. Whether this unique feature in the Pi9 gene has any bearing on its broad resistance spectrum will require further investigation.
NBS–LRR genes in monocot plants carry a CC structure at the N-terminal region and belong to the CC–NBS–LRR (or CNL) subgroup of R genes (Bai et al. 2002; Meyers et al. 2003). An analysis of the structural features of the rice NBS–LRR genes showed that their CC domains are poorly conserved in the rice genome (Bai et al. 2002). One conserved motif (WVxxIRELAYDIEDIVDxY), designated nT, is usually located ∼130 amino acids before the P-loop in rice NBS–LRR genes (Bai et al. 2002). The Pi9 protein has an nT motif (WAEQIRDLSYDIEDSLDEF) that is located 107 amino acids before the P-loop. This Pi9 nT motif has 57 and 61% of identities, respectively, with the Pib nT motif (WVKQVRDTAYDVEDSLQDF; Wang et al. 1999) and the Pi-ta nT motif (WAKEVRELSYDVDDFLDEL; Bryan et al. 2000), but has no significant similarity to the Xa1 nT motif (SLGRLRGLLYDADDAVDEL), the latter of which is an NBS–LRR gene and confers resistance to the rice bacterial blight pathogen (Yoshimura et al. 1998).
The LRR region of most rice NBS–LRR genes contains ∼15% leucine but does not have the typical LRR repeats (Bai et al. 2002). One example is the Pi-ta protein, whose C-terminal domain, named the leucine-rich domain, lacks the characteristic LRR motif. However, although some imperfect LRR repeats exist in the Xa1 LRR domain, Xa1 has six almost perfect repeats, each with 93 amino acids (Yoshimura et al. 1998). The LRR domain in Pi9 is quite similar to that of the Pib protein and consists primarily of imperfect LRR repeats. A unique structural feature of the Pi9 protein is that it contains a 57-amino-acid non-LRR region at the C terminus. In contrast, the LRRs in both Pib and Pi-ta extend to the end of the C terminus. Further research is needed to investigate whether this 57-amino-acid sequence at the C terminus of Pi9 has any special function in regulating resistance specificity to rice blast.
At the protein level, the Pi9 protein is quite different from other cloned plant R proteins with broad-spectrum resistance. The cloned potato RB and RPI genes confer broad-spectrum resistance to the oomycete pathogen P. infestans and encode an identical CC–NBS–LRR protein (Song et al. 2003; Van Der Vossen et al. 2003). When the amino acid sequence of the Pi9 protein was compared with that of the RB/RPI protein, the identity and similarity were found to be only 25 and 42%, respectively. The Arabidopsis R gene RPW8 confers broad-spectrum resistance to powdery mildew (Erysiphe) (Xiao et al. 2001) and shows no homology with any of the cloned genes. No significant similarity in amino acid sequence was observed between Pi9 and RPW8.
Genomic structure at the Pi9 locus:
We sequenced a 76-kb genomic region from the Pi9 locus and identified six NBS–LRR genes that are arranged as tandem repeats. Identification of the Pi9 multigene family is consistent with the findings from other R loci where the R genes are commonly clustered in the genome (Michelmore and Meyers 1998; Richter and Ronald 2000; Hulbert et al. 2001; Bai et al. 2002; Wei et al. 2002; Sun et al. 2004). From the viewpoint of plant evolution, the clustering of highly homologous R genes at a locus provides a variety of opportunities for plants to evolve new specificities of resistance when the corresponding AVR gene in the pathogen has mutated. Genetically linked gene families have more possibilities for recombination than do simple loci composed of one or a few genes (Meyers et al. 1998; Hulbert et al. 2001).
On the basis of phylogenetic analysis, the six NBS–LRR genes belong to four heterogeneous types, i.e., Pi9 (Nbs2-Pi9)/Nbs5-Pi9, Nbs4-Pi9/Nbs6-Pi9, Nbs1-Pi9, and Nbs3-Pi9. High homology was observed between Nbs2-Pi9 and Nbs5-Pi9 (98% nucleotide identity) and between Nbs4-Pi9 and Nbs6-Pi9 (94% nucleotide identity). The genomic region of Nbs5-Pi9)/Nbs6-Pi9 (58,096–76,272 bp) appears to be a duplication of the Pi9 fragment (Nbs2-Pi9)/Nbs4-Pi9 (33,725–52,025 bp). It is conceivable that the progenitor of Nbs4-Pi9 might be a sequence like the solo-LTR-disrupted Nbs6-Pi9 and the Nbs4-Pi9 pseudogene might be derived from imprecise excision of the solo-LTR retrotransposon. It remains unclear how the gene introgression occurred during the wide hybridization between O. minuta and IR31917. It may be difficult to identify the original O. minuta plant that was used for the introgression cross. However, complete sequencing of the Pi9 locus in one of the O. minuta lines will provide the information on the genome structure and sequence variation among the family members in the region.
The molecular mechanism of Pi9's broad-spectrum resistance to M. grisea:
Broad-spectrum disease resistance provides a useful resource for breeding disease-resistant crops as well as for understanding the molecular basis of resistance specificity in plants. Broad-spectrum resistance can be classified into two types: resistance to many isolates of the same pathogen and resistance to two or more unrelated pathogens. Representatives of the first type are the RB gene (Song et al. 2003) and the RPI gene (Van Der Vossen et al. 2003) that confer broad-spectrum resistance to nearly all known races of the late blight pathogen P. infestans. A good example of the second type is the Arabidopsis R gene RPW8 that is resistant to two different powdery mildew fungal pathogens, i.e., Erysiphe cruciferarum UEA1 and E. cichoracearum UCSC1 (Xiao et al. 2001). For rice blast pathogen M. grisea, several R genes with a relatively broad spectrum of resistance to diverse isolates have been identified (Chen et al. 1996; Liu et al. 2002; Jeon et al. 2003). Among them, the Pi9 parent 75-1-127 carries broad-spectrum resistance to diverse blast isolates (Liu et al. 2002). It is possible that the broad-spectrum blast resistance in 75-1-127 is due to the action of several NBS–LRR genes. For example, more than one functional R gene has been reported to be located at the rice Xa21 locus (Wang et al. 1998), the tomato Cf2 locus (Dixon et al. 1996), and the Arabidopsis RPW8 locus (Xiao et al. 2001). In this study, each member of the Pi9-multigene family was functionally characterized by rice transformation and inoculation tests. In the inoculation tests with an additional 21 diverse isolates from nine countries, we confirmed that the Pi9 transgenic line had the exact resistance spectrum with the donor line, suggesting that Nbs2-Pi9 is the sole functional member in the cluster responsible for Pi9's broad-spectrum resistance to M. grisea. Whether the Pi9 gene is durable or not needs more field evaluations in multiple locations in the near future.
Elucidation of the molecular basis of Pi9's broad-spectrum resistance is intriguing. It is possible that Pi9 can recognize a conserved molecule in different isolates of M. grisea or that different molecules in diverse isolates can be recognized by Pi9. LRR domains are thought to be the major determinant of specificity in R genes (Hulbert et al. 2001). Experiments with the rice R gene Pi-ta and with the AVR–Pita from the pathogen M. grisea have provided direct evidence of an interaction between the LRR domain of an R gene and its cognate AVR gene (Jia et al. 2000), while results from flax support this notion by allelic comparisons and domain-swapping experiments between different alleles at the flax L and P loci (Ellis et al. 1999; Dodds et al. 2001). Isolation of the AVR–Pi9 protein or of a conserved effector molecule(s) from M. grisea that interacts with Pi9 in rice cells will be an important step forward in our understanding of the molecular basis of Pi9 broad-spectrum resistance. It is also possible that the Pi9 protein interacts with a RIN4-like protein that is the target of the AVR–Pi9 protein as demonstrated in Arabidopsis RPM1 and the AVRRpm1/AVRB system (Mackey et al. 2002).
We are grateful to G. Lu, H. Leung, B. Valent, J. R. Xu, S. S. Han, and M. Farman for providing the M. grisea isolates for the inoculation tests and to Beth E. Hazen and Mohan R. Babu for critical reading of the manuscript. This work was supported in part by a start-up fund from Ohio State University and by a seed grant from the Ohio Agricultural Research and Development Center to G. L. Wang.
Sequence data from this article have been deposited with the EMBL/GenBank Data Libraries. The Pi9 76-kb contig is deposited under accession no. DQ285630.
↵1 These authors contributed equally to this work.
↵2 Present address: Section of Plant Biology, University of California, Davis, CA 95616.
Communicating editor: S. R. McCouch
- Received April 26, 2005.
- Accepted December 17, 2005.
- Copyright © 2006 by the Genetics Society of America