Abstract
Rp1 is a complex rust resistance locus of maize. The HRp1-D haplotype is composed of Rp1-D and eight paralogues, seven of which also code for predicted nucleotide binding site-leucine rich repeat (NBS-LRR) proteins similar to the Rp1-D gene. The paralogues are polymorphic (DNA identities 91-97%), especially in the C-terminal LRR domain. The remaining family member encodes a truncated protein that has no LRR domain. Seven of the nine family members, including the truncated gene, are transcribed. Sequence comparisons between paralogues provide evidence for past recombination events between paralogues and diversifying selection, particularly in the C-terminal half of the LRR domain. Variants selected for complete or partial loss of Rp1-D resistance can be explained by unequal crossing over that occurred mostly within coding regions. The Rp1-D gene is altered or lost in all variants, the recombination breakpoints occur throughout the genes, and most recombinant events (9/14 examined) involved the same untranscribed paralogue with the Rp1-D gene. One recombinant with a complete LRR from Rp1-D, but the aminoterminal portion from another homologue, conferred the Rp1-D specificity but with a reduced level of resistance.
DISEASE resistance (R) genes utilized by plant breeders arose by coevolution of pathogens with ancestors of current crop species (Crute and Pink 1996). Disease resistance conferred by R genes is usually race specific, conditioning resistance reactions only to pathogen biotypes that carry the corresponding avirulence gene. The effectiveness of a resistance gene in controlling disease depends on the frequency of the corresponding avirulence gene in the pathogen population. Plant pathogens, especially cereal rusts, are capable of altering their specific virulence through mutation or loss of Avr genes (Burdon and Silk 1997). Resistance loci that can match the genetic plasticity of pathogen populations would be an asset to the survival of plant species. The first indications that plants may carry such loci were the multiple resistance specificities that have been mapped to loci like the L and M loci of flax (Shepherd and Mayo 1972), the Mla locus of barley (Jahooret al. 1993), the Dm1 locus of lettuce (Hulbert and Michelmore 1985), and the rp1 locus of maize (Saxena and Hooker 1968).
R genes may reside at single gene loci, like the flax L locus (Lawrenceet al. 1995) and the Arabidopsis Rpm1 (Grantet al. 1995) and RPS2 (Mindrinoset al. 1994) loci, or they may belong to families of tightly linked genes. Examples of the latter include the tobacco N locus, flax M locus, the tomato Pto, I2, Cf2/5, and Cf4/9 loci, the rice Xa21 locus (Martinet al. 1993; Whithamet al. 1994; Andersonet al. 1997; Songet al. 1997; Thomaset al. 1997; Dixonet al. 1998; Simonset al. 1998), and the rp1 locus of maize. Many of the tightly linked gene families that have been characterized have relatively simple structures, with small intergenic regions, which do not code for unrelated genes. The gene cluster carrying the Dm3 gene of lettuce is notable in that the distance between the homologous genes is estimated at ∼100 kb yet no genes have been identified in the intergenic regions (Meyerset al. 1998a). A contrast is the Pto locus, which carries five homologous genes including Pto that encode protein kinases and, imbedded within this cluster, an unrelated gene Prf (Salmeronet al. 1996) that is also required for Pto-mediated resistance. The barley Mla region carries three distinct nucleotide binding site-leucine rich repeat (NBS-LRR) gene families, which are intermingled within a 240-kb interval (Weiet al. 1999).
Molecular genetic studies of R gene loci have established some trends in resistance gene evolution. One trend is the prevalence of diversifying selection in regions encoding leucine rich repeats (LRRs). Sequence comparisons of different members of resistance gene families have found a significantly higher ratio of nonsynonymous substitutions to synonymous substitutions, particularly in codons encoding predicted solvent-exposed residues potentially involved in ligand binding. This has now been documented for a number of R gene families (Parniskeet al. 1997; Botellaet al. 1998; McDowellet al. 1998; Meyerset al. 1998b; Wanget al. 1998; Noëlet al. 1999; Ayliffeet al. 2000).
Another trend is the involvement of recombination in R gene evolution. The flax L locus is a simple locus with many alleles. Intragenic recombinants have been identified in the progeny of L heterozygotes (Shepherd and Mayo 1972; Elliset al. 1997). For example, a crossover between L2 and L10 gave rise to suL10 with no detectable resistance specificity. A second crossover between suL10 and L9 resulted in RL10, a revertant expressing the L10 specificity (Islamet al. 1989; Elliset al. 1997). Moreover, R genes like the L and M genes of flax have sequence duplications in the LRR region that can mispair in meiosis. Three susceptible M gene variants presumably derived from unequal crossover events were identified that carried only one direct repeat unit in their LRR instead of the two present in the parental allele (Andersonet al. 1997). Variation in the number of direct repeats in different alleles of the L locus is probably generated by similar processes (Elliset al. 1999). This type of intragenic recombination was also observed at the Rpp5 locus of Arabidopsis where an allele with a duplication of four LRRs was identified in a susceptible variant (Parkeret al. 1997). Other gene families, such as the Cf2/5 locus (Dixonet al. 1998), also have duplicated stretches with sequence homology in the LRR regions and also show variation in repeat number.
In families of closely linked R genes, genes at different positions in the array may pair and recombine. A susceptible variant derived from such an unequal crossing-over event was identified and molecularly characterized at the tomato Cf-2/Cf-5 locus (Dixonet al. 1998). Further evidence is provided by comparing the patterns of sequence affinities between different family members. The mosaic pattern of DNA homologies of several cloned R genes and their family members suggests that multiple steps of recombination between paralogues occurred during their evolution (Parniskeet al. 1997; McDowellet al. 1998; Elliset al. 1999; Noëlet al. 1999).
The rp1 locus of maize carries genes that condition resistance to Puccinia sorghi, the causal agent of common rust. Early recombination analyses, which showed that Rp1 genes could be recombined into coupling phase, demonstrated that it was not a simple locus (Saxena and Hooker 1968). Additional recombination analyses with flanking markers indicated that the complex did not behave like a cluster of distinct loci. Instead, the genes appeared to mispair in a number of ways as they recombined (Sudupaket al. 1993; Hu and Hulbert 1994). Recombination events in the rp1 complex have also been associated with the generation of variants with new phenotypes, including novel race specificities (Richteret al. 1995) and lesion-mimic phenotypes (Huet al. 1996). Recent molecular identification of the Rp1-D gene (Collinset al. 1999) has provided probes for analysis of recombinant Rp1 haplotypes. This provided molecular evidence that Rp1 genes are a family of tightly linked genes that mispair in different arrangements in meiosis, with subsequent recombination events creating new Rp1 haplotypes. The identification of new restriction fragments in gel blot analysis of DNAs from some of the variants indicated that many of the recombination events were occurring close to and possibly within the Rp1 genes.
The HRp1-D haplotype carries the Rp1-D gene and eight homologous sequences (paralogues) as estimated by gel blot analysis. Here we characterize paralogues from this haplotype and the HRp1-K haplotype to determine the nature of these genes. We also characterize the haplotypes and recombinant genes in a collection of spontaneous variants to identify the type of genetic events that can generate variability at this locus.
MATERIALS AND METHODS
Genomic DNA libraries: Genomic DNA libraries of the HRp1-D*21 and HRp1-D*24 lines were constructed by ligating Sau3AI partially digested, glycerol gradient fractionated, genomic DNA fragments to BamHI-digested EMBL3 λ-vector (Promega, Madison, WI). The DNA library of the wild-type Rp1-D line was made by ligating Sau3AI partially digested, sucrose gradient-fractionated, genomic DNA fragments to BamHI/ XhoI double-digested λ DASHII vector arms (Stratagene, La Jolla, CA). Ligation products were packaged using Golden II packaging kit (Stratagene). Library screening and λ DNA preparations were performed as described (Sambrooket al. 1989).
Genomic clone characterization: Genomic clones from the HRp1-D haplotype were classified by amplifying the same region from each clone using a pair of conserved rp1 primers for the 5′ ends of the rp1 gene family members. The forward primer sequence was P6, AGCTTCAGCTTACCTCAGTG, and the reverse primer sequence was 1520R, CCAATCCAACAAT GGCCAAAC. λ-Clones were partially restriction mapped and tested for their ability to hybridize to probes made from the extreme 5′ and 3′ ends of the Rp1-D gene. Selected full-length λ-clones from each sequence group were subcloned into pUC19 for sequencing using the Big-Dye sequencing system (Applied Biosystems, Foster City, CA). Sequencing primers were positioned throughout the gene sequence, and new primers were designed for specific genes where their sequences diverged from the others.
Sequencing data analysis: Sequence data were analyzed using computer programs from the University of Wisconsin Genetics Computer Group package Version 10. Sequence alignment was done using the Clustal W program. The significance of differences in synonymous and nonsynonymous substitution rates was examined by using a 2 × 2 contingency table G-test (Zhanget al. 1997). The sequences of the rp1 paralogues in the HRp1-D haplotype were submitted to GenBank and given the following accession numbers: rp1-dp1, AF342991; rp1-dp2, AF342992; rp1-dp3, AF324993; rp1-dp5, AF342994; rp1-dp6, AF342995; rp1-dp7, AF342996; rp1-dp8, AF342997; rp1-kp1, AF344308; rp1-kp2, AF344311; and rp1-kp3, AF344309.
cDNA cloning and reverse transcriptase-PCR: The construction of a cDNA library and cDNA cloning and sequencing were conducted as described previously (Collinset al. 1999). RNA was isolated from fully expanded third leaves of 10-day-old seedlings. A primer pair was designed flanking the 3′ intron and used in the reverse transcriptase (RT)-PCR to eliminate the possibility of genomic DNA contamination (forward primer, CTGTATTGCTCAACCACATGC; reverse primer, CCTGAACTCTGGAGCTTCAAG). RT-PCR was conducted using the RT-PCR kit (Stratagene) as recommended by the manufacturer. Following first strand synthesis using an oligo(dT) primer, several separate PCR reactions were conducted on each sample and pooled before cloning.
PCR-based cloning and sequencing of recombinant genes: Fragments from most of the recombinant genes were PCR amplified and cloned using gene-specific primers. Three Rp1-D-specific primers were used as reverse primers, to extend the noncoding strand. The DR1 primer (TTTCCTCCGGAACCA GAACAC) was designed from sequences in a highly polymorphic area corresponding to the C domain in the LRR of the Rp1-D gene. The DR4 (TAGCGGAGCAATACAAGCGGC) and DR5 (GGCCACATGAATGATATAGC) primers correspond to sequences in the 3′-untranslated region (3′-UTR). Primers from more conserved sequences were usually used as forward primers. The P6 primer (AGCTTCAGCTTACCTCAGTG) corresponds to conserved sequences starting 67 nucleotides downstream of the putative start codon and was the most commonly used forward primer. Rp1-D-specific forward primer 1507F (TTGGTTTTTCATATCTATTGTGAT) was used to amplify the recombinant gene in the HRp1-BD12 haplotype because gel blot analysis indicated it had the 5′ end of the Rp1-D gene. The primer corresponds to sequences in the beginning of the LRR sequence. PCR products were cloned by using the TOPO TA-cloning kit (Invitrogen, Carlsbad, CA).
RESULTS
DNA sequence variation among rp1 genes from the HRp1-D haplotype: Genomic DNA gel blot analysis indicated that the HRp1-D haplotype contains approximately nine Rp1 paralogues, rp1-dp1 to rp1-dp9, with rp1-dp9 corresponding to the Rp1-D gene (see below). Thirty λ-clones isolated from an Rp1-D genomic library by homology with the cloned Rp1-D gene (Collinset al. 1999) were resolved into nine classes by PCR amplifying and sequencing clones containing a 600-bp segment of the NBS encoding region. When possible, a single full-length λ-clone was selected from each group by testing for hybridization to probes from both the 5′ and 3′ ends of the Rp1-D gene. Sequence analysis of these clones indicated they represented the Rp1-D gene and eight homologous family members.
Comparisons of the DNA sequences of coding regions of different rp1 family members reveals high levels of sequence identity (91-97%) to the Rp1-D rust resistance gene (Table 1). The 3′ halves of the genes, corresponding to the LRR domain, were the most divergent. This is especially true of the rp1-dp3 and rp1-dp1 genes, which have only 87 and 88% identity with the Rp1-D gene in their 3′ halves. This trend can also be observed by comparing the deduced amino acid sequences of the genes (Table 1). Of the nine paralogues in the HRp1-D haplotype, eight are similar in size, containing an open reading frame (ORF) ranging from 3800 to 3862 bp in length without introns. Small introns occur at positions 36 bp upstream of the proposed Rp1-D start codon and 28 bp downstream of the stop codon. Alignments of the deduced amino acid sequences of the coding regions of the nine paralogues are shown in Figure 1. Short regions that were highly variable in both sequence and length were observed at two locations corresponding to amino acids 505-512 and 671-677 in the Rp1-D coding region. These hypervariable regions were designated regions A and B in Figure 1. Region A occurs shortly upstream of the LRR domain. The fourth paralogue (rp1-dp4) encodes a truncated NBS-LRR protein that corresponded to a previously identified cDNA from the Rp1-D haplotype (rp1-Cin4; GenBank accession no. AF107294; Collinset al. 1999). The 5′ half of the gene is similar to the NBS encoding 5′ half of the Rp1-D gene but the 3′ half of the gene is replaced by ∼480 nucleotides with homology to the Cin4 family of retrotransposon-like elements (Schwarz-Sommeret al. 1987). The deduced amino acid sequence is 99% identical to that of Rp1-D until amino acid 462, after which the sequence diverges and the open reading frame ends. To determine if the truncated gene was caused by a simple insertion of a Cin4-like element, 13 kb of genomic DNA downstream of the rp1-dp4 gene was subcloned and analyzed. A lack of hybridization to the 3′ half of the Rp1-D gene indicated that the gene was not truncated by a simple insertion of a small element. Sequence analysis of ∼5 kb 3′ to the rp1-dp4 gene found no significant ORFs. No significant homology to GenBank sequences was observed for the first 1.5 kb. The next 1.5 kb was highly homologous (>87%) to the first 1.5 kb of a large UTR region of the GapC2 gene (GenBank accession no. 312178) coding for glyceraldehyde-3-phosphate dehydrogenase. Furthermore, hybridization of this sequence to gel blots indicated that it occurs >50 times in the maize genome and also occurs in the untranslated region of at least one other maize gene (accession no. 1213278). Comparisons of PCR products from the λ-clone and genomic DNA from the Rp1-D line, using primers for Rp1 and GapC2 sequences, indicated that the λ-clone was not chimeric. Identically sized products were generated from both genomic DNA and λ-DNA templates.
The HRp1-K haplotype generally has more fragments in gel blot analysis than HRp1-D, although it is difficult to determine gene number because of the possibility of comigrating fragments (data not shown). Three rp1 family members from the HRp1-K haplotype were cloned and sequenced. In general, the rp1 genes from the HRp1-K haplotype show high homology with those from the HRp1-D haplotype. The rp1-kp1, rp1-kp2, and rp1-kp3 genes have 96, 97, and 93% sequence identity, respectively, when compared with the Rp1-D gene. Interestingly, the rp1-kp2 gene had a point mutation that resulted in a stop codon only 111 bp after the putative start codon. There were no other premature stop codons observed in the rest of the gene. The other two HRp1-K genes have full-length ORFs similar to most of the HRp1-D genes.
Similarities of rp1 family members to the Rp1-D gene
Transcription of rp1 genes: Two different size transcripts were previously observed on RNA gel blots with Rp1-D probes. The larger transcript was ∼4 kb and corresponded to the size of the Rp1-D gene. The smaller one was ∼1.5 kb and corresponded to the size of the rp1-dp4 (rp1-Cin4) gene. Since most of the rp1 family members would not be distinguishable from Rp1-D by RNA gel blot analysis, we used partial sequence analysis of cDNAs to determine if other family members were transcribed. Most of the ∼80 cDNA clones (selected with an NBS area probe) were derived from the rp1-dp4 gene. The sequence of 10 clones matched the Rp1-D sequence and the remaining 2 clones matched the respective sequences of rp1-dp7 and rp1-dp8. RT-PCR was used to identify additional transcribed members. Two PCR primers flanking the small intron in the 3′-UTR of the genes were designed from conserved sequences so that they would amplify all nontruncated rp1 genes. Cloning and sequencing of 26 PCR products identified 6 identical to Rp1-D, 10 identical to rp1-dp7, and 4 identical to rp1-dp8. The remaining 6 sequences did not perfectly match the sequence of any of the genomic clones and differed from Rp1-D, rp1-dp7, or rp1-dp8 at one or more nucleotides, probably due to errors introduced during PCR amplification. Gene-specific primers were used in RT-PCR experiments to determine if any of the other genes are transcribed. Besides the previously detected members, 3 additional members, rp1-dp1, rp1-dp2, and rp1-dp3, were found to be expressed. No RT-PCR products could be identified for the rp1-dp5 and rp1-dp6 genes, indicating they may not be transcribed. Taken together, these data indicate that there are at least 7 rp1 family members in the HRp1-D haplotype that are expressed.
Comparison of the 5′ and 3′ sequence flanking paralogues in the HRp1-D haplotype: Some of the rp1 genes in the HRp1-D haplotype have distinct promoter regions, as shown in Figure 2. The region consisting of 1200 bp upstream of the putative translation start codons of the rp1-dp2 and rp1-dp7 genes shows respective DNA sequence identities of 97 and 98% to the Rp1-D gene. The homologous regions include an ∼320-bp UTR with a 142- to 144-bp intron. Upstream of this 1200-bp region, the sequences of all three genes show homology to the same retroelement sequence. Sequence homology of these three genes may extend beyond the determined retroelement sequence point, but this was not examined. The rp1-dp3 gene was also highly similar to Rp1-D for the 420 bp preceding the start codon, but no additional sequence was available from clones carrying this gene. The 553 bp preceding the translation start codon of the rp1-dp4 gene is also highly homologous to Rp1-D. It shows no noticeable homology upstream of this point and shows clear retroelement homology at ∼730 nucleotides before the putative translation start. Since the rp1-dp4 gene is transcribed, this may indicate that the required 5′ regulatory regions are included in these 553 nucleotides. The rp1-dp6 gene is also homologous to Rp1-D beginning 937 nucleotides before the start codon. One exception to this homology is a 45-bp insertion located 401 nucleotides 5′ to the translation start codon in rp1-dp6, which may be responsible for the apparent lack of transcription of this gene. The 5′ noncoding regions of rp1-dp1 and rp1-dp8 show only low homology to Rp1-D and the other homologues and are more similar to each other, at least for the first 200 bp before the start codon. Analysis of a 5′ RACE PCR-generated clone corresponding to the rp1-dp1 gene indicated it had a 513-bp UTR including a 95-bp intron. When compared to the 5′-UTR of the Rp1-D gene, rp1-dp1 and rp1-dp8 have numerous differences including insertions (up to 81 bp) and deletions and a stretch of 67 bases with no significant sequence similarity. Several stretches of sequence homology (>80% identity) occur, however, including areas close to the transcription start sites and the 5′ intron splice site. There is no significant homology observed 20 bp upstream of the putative transcription start site of the rp1-dp1 gene and beyond this the rp1-dp1 and rp1-dp8 genes showed no homology to the other paralogues or to each other. The 5′-UTR sequence of the rp1-dp5 gene was similar to the rp1-dp1 and rp1-dp8 genes but diverges from these genes 276 bases before the start codon and is missing the 5′ splice site of the intron in these two genes. Furthermore, retroelement homology was observed in rp1-dp5 at a point corresponding to the transcription initiation site in other gene family members, and this presumably accounts for the lack of expression of the rp1-dp5 gene.
—Alignment of amino acid sequences of the Rp1-D gene product with the seven other full-length genes from the Hrp1-D haplotype and three genes from the HRp1-K haplotype. Amino acids corresponding to long stretches of DNA sequence identity between different genes are similarly boxed or shaded. Hypervariable regions (A and B), with frequent amino acid substitutions and small indels, are underlined. Deletions are indicated by dashes. Dots represent amino acids identical to the Rp1-D sequence.
Sequence comparisons of the 3′ noncoding regions revealed that the rp1 genes retained sequence homology until ∼340 bp after the translation stop codon of the Rp1-D gene. While the 3′-UTR regions in the different genes do not show the structural variation characteristic of the 5′-UTRs, certain pairs of genes showed sequence affiliation. For example, the rp1-dp5 and Rp1-D are very similar (>98%) in this region. Some pairs of genes retained their sequence affiliation after the others diverged. For example, the rp1-dp7 and rp1-dp8 genes were identical from 450 to 650 bases downstream of the translation stop codon. Interestingly, these two genes were among the most different for the 5′-UTR region.
Patterns of polymorphism among the coding regions of the rp1 paralogues: Alignment of the DNA sequences of the rp1 family members allowed the relationships of the genes to be examined in detail. No consistent sequence affiliation between any pair of rp1 paralogues was observed throughout the entire ORF. Instead, the coding regions of the genes are composed of mosaics, or patchworks of stretches of similar or identical DNA sequences, some of which are highlighted in Figure 1. For example, shortly after the start codon, rp1-dp8 and rp1-dp5 share a 370-bp stretch of sequence identity. Toward the middle of the gene, rp1-dp8 and rp1-dp5 diverge from each other and, instead, rp1-dp8 now has a stretch of 240 bp that matches an identical stretch in the rp1-dp3 gene corresponding to amino acids 601-680 (Figure 1). Near the 3′ end of the gene, rp1-dp8 is again identical to rp1-dp3 for 242 nucleotides corresponding to amino acids 1177-1248 except for one uninformative difference. The stretches of identical sequence and changes in sequence affiliation are also observed between other family members from the HRp1-D haplotype and when comparing members from the HRp1-D and HRp1-K haplotypes (Figure 1). Changes in affiliation are also apparent when comparing sequences flanking the coding regions as detailed above. This mosaic nature of sequence identities implies that reassortment of sequences has occurred through multiple genetic exchange events during the evolution of the gene family.
Evidence for diversifying selection preferentially in the C-terminal half of the LRR domain: Patterns of nucleotide substitution can be informative in assessing the type of selection pressure acting on the evolution of gene family members. The rates of synonymous and nonsynonymous nucleotide substitutions were calculated between all possible pairs of the rp1 genes from the HRp1-D haplotype, except the truncated rp1-dp4 gene. These were calculated separately for different regions of the gene (Table 2). Rates of synonymous (Ks) and nonsynonymous (Ka) substitutions were higher in the C-terminal region of the gene (LRR encoding domain) than those in the N-terminal region of the gene. Ratios of nonsynonymous to synonymous substitutions (Ka/Ks) were generally <1 in the N-terminal region, averaging only 0.46 for 28 pairwise comparisons. Ka/Ks ratios in the C-terminal region of the gene, including the LRR, were <1 in some comparisons and >1 in others, averaging ∼0.81. When the sites in the LRRs that match the xxLxLxx consensus sequence (Jones and Jones 1996) were calculated separately, all of the gene pairs had Ka:Ks ratios >1 (7 of 28 pairs significantly >1 at P = 0.01) and the average Ka:Ks was ∼3.1. These regions, based on structural similarity to the porcine ribonuclease inhibitor protein, are predicted to be solvent exposed and potentially involved in ligand binding (Kobe and Deisenhofer 1995). Like the functional RP1-D protein (Collinset al. 1999), the Rp1 proteins contain two LRR regions (domains B and D) that are separated by a region (domain C) that shows little homology to the LRR consensus. When the Ka:Ks ratio was calculated for the regions with xxLxLxx consensus in LRR domain B and domain D separately, domain D had an average Ka:Ks ratio of 3.5 (15 of 28 gene pairs with a Ka:Ks ratio significantly >1 at P = 0.01 level) while domain B showed an average Ka:Ks ratio of only 1.4 (only 3 of 28 pairs significantly >1). This is consistent with the results obtained from analysis of Ka:Ks ratios of the Rp1 homologues in barley (Ayliffeet al. 2000). It implies that the carboxy-terminal half of the LRR region in this gene family is under stronger diversifying selection pressure than that of the amino-terminal half. The conserved leucine sites likely serve a structural role and are probably under selection for sequence conservation (Parniskeet al. 1997).
—Differences in the upstream regions of the rp1 genes from the HRp1-D haplotype. Regions with high levels of sequence identity (>90%) are indicated by the same box patterns. Differently shaded regions show no noticeable sequence homology. Transcription starts at ∼-320 in the Rp1-D gene, relative to the start of the putative coding region. The site of an intron in the 5′-UTR that corresponds to bases -179 to -36 in the Rp1-D gene is also shown. Boxes made from dotted lines indicate regions with retroelement homology. A solid triangle in the rp1-dp6 sequence indicates an apparent 45-bp insertion.
Evidence for diversifying selection in the evolution of the rp1 gene family
Organization of rp1 paralogues in the HRp1-D haplotype: Construction of a contig map of the HRp1-D haplotype was not possible using the genomic clones carrying the rp1 paralogues. The clones averaged ∼20 kb in size, but none contained more than a single rp1 paralogue. Instead, genetic recombinants from three different populations were used to determine the order and orientation of the genes in the HRp1-D haplotype.
Variants from an Rp1-D homozygote: The first set of variants consisted of a previously described (Collinset al. 1999) panel of mutants derived from an Rp1-D homozygote. All but two of the variants were susceptible to all rust isolates tested (Richteret al. 1995). Two others had a reduced-resistance phenotype and a nonspecific lesion-mimic phenotype (Huet al. 1996). Gel blot analysis with an Rp1-D probe indicated that they represented spontaneous deletions of variable sizes, presumably caused by unequal recombination events. They could be grouped into nine different classes by restriction analysis with AccI, BglI, BglII, DraI, NcoI, and NsiI (Table 3).
Variants from Rp1-D/Rp1-A heterozygotes: The second set of recombinants consisted of 26 susceptible variants from Rp1-A/Rp1-D heterozygotes. All but 3 (DNCO1, DNCO2, and DNCO3) appeared to be derived from crossovers as indicated by flanking marker analysis. Ten of these recombinants had the Rp1-D parent allele at the proximal marker (umc285) and the Rp1-A parent allele at bln3.04, and the remaining 13 had the opposite flanking marker combination. These apparently conflicting data have previously been used as evidence for the occurrence of mispairing leading to unequal crossovers (Hulbertet al. 1997). Homozygous lines were established from 16 of these variants, including the three noncrossovers, and their rp1 haplotypes were examined with multiple restriction enzymes, including NcoI (Table 3). The recombinants generally had restriction fragments from each of the two parental haplotypes and showed a great variety of different combinations of fragments (data not shown).
Classification of deletion variants based on restriction enzyme digestion and Southern blot analysis
Variants from Rp1-D/Rp1-B heterozygotes:The third set consisted of 22 susceptible variants derived from Rp1-D/ Rp1-B heterozygotes. All but 2 appeared to be derived from crossovers because they had nonparental combinations of flanking restriction fragment length polymorphism (RFLP) markers (umc285 and bln3.04). All of these had the distal (bnl3.04) marker genotype of the Rp1-B parent and the proximal (umc285) marker genotype of the Rp1-D parent. This would be expected if the Rp1-B gene is located closer to the centromere within the HRp1-B haplotype than is the Rp1-D gene within the HRp1-D haplotype. Homozygous lines were established for 14 of the variants, which were then examined by DNA blot analysis with multiple enzymes. As with the second set of variants, the crossover derivatives contained fragments from both parents, which were present in many different combinations (Figure 3). The variation in Rp1 haplotypes and the flanking marker genotypes from the second and third populations of recombinants indicated that these pairs of haplotypes can pair and recombine in a number of different arrangements.
—DNA gel blot analysis of the parental Rp1-B and Rp1-D lines and 12 susceptible recombinants from a cross between these two lines. DNAs were isolated from individuals that were homozygous for recombinant haplotypes. DNAs were digested with NcoI and probed with a NBS region probe from the Rp1-D gene. Recombinants exhibited different combinations of restriction fragments from both parents. All these recombinants have the distal (bnl3.04) marker genotype of the Rp1-B parent and the proximal (umc285) marker genotype of the Rp1-D parent (data not shown). Numbers on the left indicate the designation of the nine HRp1-D paralogues corresponding to the nine NcoI fragments in the Rp1-D lane. Markers, in kilobases, are shown on the right.
Examination of the HRp1-D haplotype by gel blot analysis identified nine rp1 homologous NcoI restriction fragments (Collinset al. 1999). Each of the susceptible variants from the three sets was characterized for the presence or absence of the individual NcoI fragments (Table 3; Figure 3). The presence or absence of the NcoI fragments in these three populations was used to order the fragments on the chromosome and to determine which of the fragments corresponded to each of the HRp1-D paralogues that were sequenced. The rp1-dp7 sequence was associated with the smallest fragment because it had two NcoI sites producing a fragment of this size. The fragment corresponding to the Rp1-D gene was recognized as the 8-kb fragment because the maize lines carrying the Ds and Mu2 insertions in this gene showed a shift in the size of this band (Collinset al. 1999). The truncated rp1-dp4 gene was recognized as the 4.8-kb fragment since this was the only fragment that did not hybridize to a probe of the LRR region of the gene (not shown). Other sequences were matched with their corresponding NcoI fragments by screening DNAs of the different recombinants from the HRp1-D haplotype homozygote with different gene-specific PCR primer pairs. For example, when the gene-specific primers for rp1-dp8 were used, no product could be amplified in any of the recombinants, while the primers consistently amplified the rp1-dp8 fragment in the DNA template from the parental HRp1-D haplotype. This sequence was therefore associated with the fourth-largest NcoI fragment that was missing from all of the variants from the HRp1-D haplotype homozygote. The order of the rp1 paralogues within the HRp1-D haplotype was estimated from the frequency in which they were transmitted together, with fragments that were always inherited together considered to be adjacent on the chromosome. The order of the fragments inferred from the Rp1-D homozygote recombinants (Table 3) was 1-2-(3,4,5)-6-(7,8,9). Fragments 3, 4, and 5 were always inherited together as were fragments 7, 8, and 9. The order of the fragments inferred from the recombinants from the Rp1-A/Rp1-D heterozygote was umc-285 (centromere proximal RFLP)-1-5-(6,7)-8-9-bnl3.04 (telomeric RFLP). Fragments 2, 3, and 4 were obscured by the Rp1-A parent fragments in this population. The order from the recombinants from the Rp1-B/Rp1-D heterozygote was umc285-(1,2,4)-5-(6,7)-(8,9)-bnl3.04. Fragment 3 could not be distinguished from Rp1-B parent fragments. The order inferred from combined analysis was 1-2-4-5-6-7-8-9, with fragment 9 (Rp1-D) at the telomere end of the array and fragment 3 between 2 and 6.
Isolation of recombinant genes and verification of gene order in the HRp1-D haplotype: If the deletion variants from the HRp1-D homozygotes were derived from unequal crossing over, and the crossovers occurred within the rp1 genes, then it should be possible to isolate the recombination junction in these variants using PCR primers based on rp1 gene sequences, or by screening genomic DNA libraries of the variants with an Rp1 probe. Isolation of recombinant genes from the different classes of recombinants could then be used to verify the predicted order of the genes in the haplotype and their orientation on the chromosome. Among the variants, the HRp1-D*24 haplotype has the largest deletion with only a single rp1 family member remaining. A deletion causing a haplotype with only a single family member could be generated by recombination between the two genes located at opposite ends of the gene cluster. The recombinant gene from the HRp1-D*24 haplotype was cloned from a genomic library constructed by using the genomic DNA from the recombinant. Comparison of the sequences of the entire coding region of the clone from the HRp1-D*24 haplotype with that of the Rp1-D and rp1-dp1 genes indicated that the recombinant gene consists of the 5′ end of rp1-dp1 and the 3′ end of the Rp1-D gene. This verifies that the two flanking members of the gene cluster are rp1-dp1 and Rp1-D. The HRp1-D*21 haplotype has two NcoI fragments when probed with an Rp1 probe, indicating that it carries two family members. The recombinant gene from HRp1-D*21 was cloned from a genomic library and was found to be a chimeric gene composed of the 5′ half of rp1-dp2 and 3′ half of Rp1-D. This verifies that the rp1-dp2 gene is the second gene from the centromere end of the array.
Other recombinant genes were cloned after PCR amplification, utilizing the knowledge that the Rp1-D gene is the distal gene in the array and that intragenic unequal recombination events leading to its inactivation should involve this gene. Like the recombinant genes from the HRp1-D*21 and HRp1-D*24 haplotypes, other recombinant genes would be expected to be identical to Rp1-D downstream of the crossover and identical to the other gene proximal to this. We therefore designed primer pairs to specifically amplify recombinant genes from other deleted haplotypes. The upstream primers were generally non-gene specific, while the reverse primers were designed to be reasonably specific for Rp1-D gene sequences on the basis of comparisons to other paralogues. Attempts were made to identify and sequence the recombination junction from at least one member of each of nine classes (Table 2) of recombinants and multiple members of the larger classes. The recombination junction amplified from the HRp1-D*5 haplotype contained the 5′ end of rp1-dp7 and the 3′ end of Rp1-D. This haplotype had the smallest deletion with only two family members missing, as estimated by DNA gel blot analysis. The remaining nine recombination junctions were isolated from haplotypes HRp1-D*2, D*3, D*7, D*8, D*10, D*19, D*20, D*23, and D*28 and were each composed of a 5′ region from rp1-dp5 and a 3′ region from Rp1-D. The exchange points could be located to intervals that were defined by polymorphisms between the recombined paralogues. As indicated in Figure 4, the recombination points were found to have occurred throughout the gene.
No recombination junctions were isolated from the HRp1-D*4 and HRp1-D*17 haplotypes. There were no Rp1-D sequences present in these two haplotypes as determined by a lack of amplification with pairs of specific primers, including those corresponding to sequences in the 3′-UTR of Rp1-D. In addition, sequence analysis of rp1-dp3 in the HRp1-D*4 haplotype and of rp1-dp5 in HRp1-D*17 indicated that these genes were intact. These are the genes that would have been expected to have recombined with Rp1-D on the basis of the number of family members shown by DNA gel blot analysis to be left in the haplotypes. These two deletions may have been caused by unequal recombination events that did not involve the rp1 genes, or they may have been caused by other types of events. Because they were isolated from a background with an active Ac element, it is possible they were transposon-mediated events. In total, four of the rp1 family members were observed to recombine with Rp1-D (Figure 4). Presumably, these genes are in the same orientation as Rp1-D, or the crossovers would have caused chromosomal aberrations. Thus all five genes have their 3′ LRR regions oriented toward the telomere while the orientation of the other genes is not known.
—Molecular characterization of unequal recombination events from an HRp1-D haplotype homozygote. (A) Structure of the HRp1-D haplotype and recombinant haplotypes. Boxes represent the coding regions of the rp1 paralogues corresponding to the gene designations at the top. Genes shown for each of the recombinant haplotypes are those that were not deleted by the unequal crossover events. The two paralogues involved in the recombination event are shown as partial genes flanking the deletion. Numbers in parentheses with the gene designations indicate the percentage sequence identity between the paralogue and the Rp1-D gene. Intergenic regions are not drawn to scale. (B) Recombinant genes resulting from the unequal crossovers between Rp1-D and the other gene family members in the HRp1-D haplotype. The points where the shading pattern changed indicate the recombination breakpoints. Letters A, B, C, D, and E represent the five domains of the Rp1-D gene (Collinset al. 1999) with A corresponding to the NBS domain and B-E corresponding to the LRR region.
Effects of recombination events on Rp1 gene expression and function: Most of the variant HRp1-D haplotypes, except HRp1-D*4 and HRp1-D*17, were derived from unequal crossovers involving the Rp1-D gene and retained at least the 3′ end of this gene. Therefore, the N-terminal sequences can be examined to provide insight into why the recombinant genes have null or modified resistance phenotypes. Most of the recombinant genes have the 5′ regulatory sequences of the rp1-dp5 gene. Since no transcript could be found for this gene, and its 5′ regulatory sequences were truncated, the recombinant genes are probably not transcribed. For example, the recombinant gene isolated from the HRp1-D*19 haplotype was found to consist of mainly the Rp1-D sequence but with the first 354 bases of the coding region of the rp1-dp5 gene. RT-PCR of RNA from this variant using Rp1-D specific primers detected no transcript, as expected if the promoter of rp1-dp5 is not functional. Recombinants with the N terminus of rp1-dp5 are therefore uninformative in determining whether the novel coding regions code for functional proteins.
Three of the recombinant genes involved Rp1-D and other transcribed family members. As shown in Figure 5, the D*5 haplotype is unique in that it is the only line that conferred the same resistance specificity as the parental line but showed a lower level of resistance (partial resistance) to most rust biotypes to which the parental line was resistant (Richteret al. 1995). Comparison of the recombinant gene to the two parental genes located the crossover to a region spanning 202 bases that are identical between the two genes. The promoter region and first 818-1020 bases of the coding region come from the rp1-dp7 gene while the remainder comes from the Rp1-D gene. Therefore, the recombinant gene in the HRp1-D*5 haplotype has an identical LRR to the functional Rp1-D allele, but its 5′ regulatory sequences and part of the NBS region are different. This is consistent with the idea that the LRR domain controls specificity in this class of plant disease resistance genes, but does not preclude other areas of the gene from contributing to this function.
The coding region of the recombinant gene in the HRp1-D*24 haplotype has the whole LRR region of the Rp1-D gene, but the first 1385-1389 bases of the expressed rp1-dp1 gene. The recombinant protein includes the NBS domain of rp1-dp1 and adjacent amino acids until 7 amino acids before the conserved amino acids MHD, which commonly occur before the LRR in NBS-LRR proteins. The predicted protein has only 28 amino acid (aa) differences from the Rp1-D protein, and 32 differences from the product of the functional gene from the HRp1-D*5 haplotype. The reason for the lack of resistance in the recombinant gene is not apparent. Apparently, the 5′ region of the rp1-dp1 gene either is nonfunctional or will not function with the LRR of the Rp1-D gene to confer the Rp1-D resistance specificity. Alternatively, the rp1-dp1 promoter may be weaker than the Rp1-D promoter, thereby leading to an absence of resistance.
—Phenotypes of the Rp1-D line and the HRp1-D*5 recombinant after inoculation with avirulent and virulent rust isolates. Rust isolate KS1 is virulent on the Rp1-D line while isolates I-4, IN1, and KS2 are avirulent. The HRp1-D*5 haplotype confers resistance to the same rust biotypes as the Rp1-D gene, but has a reduced level of resistance, allowing some uredinia to form.
The recombinant gene in the HRp1-D*21 haplotype was derived from a crossover in the LRR encoding region, such that the predicted protein has the NBS domain and the first 13 of the 15 repeats of the first LRR region (domain B) from rp1-dp2. The remainder of the LRR region comes from Rp1-D. The recombinant haplotype has lost the Rp1-D specificity, but still confers a highly necrotic reaction to rust inoculation and a lesionmimic phenotype (Huet al. 1996). This nonspecific phenotype may be caused by the recombinant gene. It is also possible, however, that the phenotype is caused by the unique combination of gene sequences that occurs in this haplotype.
All but one of the crossover derivatives of the Rp1-A/Rp1-D and Rp1-B/Rp1-D heterozygotes lost the 8-kb NcoI fragment that represents the Rp1-D gene. The Rp1-BD12 haplotype from the Rp1-B/Rp1-D heterozygote had recombinant flanking RFLP markers and had rp1 homologous fragments from both parental haplotypes, indicating that it was derived from a crossover. Unlike the other crossover variants, however, it had the 8-kb NcoI fragment. To determine if this represents an intact Rp1-D gene, most of the coding region was amplified with Rp1-D-specific primers, cloned, and sequenced. The recombinant gene consisted of the 5′ region of Rp1-D and a 3′ sequence of a gene that does not correspond to any of the HRp1-D sequences and presumably came from an HRp1-B haplotype gene. The first 2107 bp of the coding region contains sequences from the Rp1-D gene encoding the NBS domain and the first 10 LRR, while the remainder of the sequence is derived from a gene from the HRp1-B haplotype. Because this gene contains the functional Rp1-D promoter, its loss of function is likely due to the recombinant nature of the protein product.
Analysis of variants that are not associated with crossovers: While most of the susceptible variants analyzed were caused by crossovers, five variants from Rp1 heterozygotes did not show flanking marker exchange. Two noncrossover (NCO) variants were isolated from Rp1-B/Rp1-D heterozygotes. The first was a previously described Mu2 insertion mutant (Collinset al. 1999) containing Rp1-D parent alleles for both flanking RFLP markers. The second was designated HRp1-BNCO1 and had the flanking marker alleles of the Rp1-B parent. Gel blot analysis of HRp1-BNCO1 DNA with several enzymes (AccI, BamHI, BglII, EcoRI, EcoRV, HindIII, NcoI, SalI, XbaI, and XhoI) found the restriction patterns to be indistinguishable from the Rp1-B parent, indicating that the mutation in the Rp1-B gene was not caused by a crossover or an insertion of a transposable element (Figure 6). The HRp1-DNCO1, HRp1-DNCO2, and HRp1-DNCO3 haplotypes came from Rp1-A/Rp1-D heterozygotes and had the Rp1-D parent alleles at both flanking marker loci. Gel blot analysis of the HRp1-DNCO1 haplotype indicated it was probably derived from a crossover because it had fragments derived from both parents in each of several digests tested. The parentally marked flanking markers indicated that there was a second (double) crossover involved. In contrast, the restriction fragments of the HRp1-DNCO2 and HRp1-DNCO3 haplotypes were indistinguishable from the HRp1-D haplotype. These two mutants were further characterized by cloning and sequencing the Rp1-D gene from these lines. Analysis of the coding region of the HRp1-DNCO2 mutant allele identified only a single point mutation. The mutation occurs in the 17th amino acid after the putative translation start codon and results in a valine to alanine substitution (Figure 1). Sequencing of the mutant Rp1-D allele in HRp1-DNCO3 revealed a 33-nucleotide deletion in the region encoding the 10th LRR while the rest of the sequences are identical to the functional Rp1-D allele (Figure 1). Since in-frame indel polymorphisms are fairly common between the LRR regions of rp1 paralogues, it is possible this change is not an actual deletion, but a gene conversion from an allele from the HRp1-A haplotype. To find out if this polymorphism was derived from the HRp1-A parent, a pair of primers flanking the deletion was used in PCR amplification of HRp1-A genomic DNA. No amplified fragments corresponding to this reduced size were observed, indicating that the gene is probably not present in HRp1-A and that the 33-nucleotide deletion is probably not a consequence of gene conversion.
—Haplotypes of noncrossover (NCO) type mutants from crosses with Rp1-A/Rp1-D and Rp1-B/Rp1-D heterozygotes. DNAs of individuals that are homozygous for the variant haplotypes were digested with NcoI and probed with the NBS region probe of the Rp1-D gene. Lanes 1 and 2 are the Rp1-A and Rp1-D parents. Lanes 3-6 are noncrossover variants DNCO1, DNCO2, DNCO3, and DNCO4, which had both flanking marker genotypes of the Rp1-D parent. Lane 7 is the Rp1-B parent and lane 8 is noncrossover variant BNCO1, which has the Rp1-B parent alleles at both flanking markers. DNA marker sizes, in kilobases, are shown on the left.
DISCUSSION
Characteristics of the HRp1-D haplotype: The HRp1-D haplotype consists of Rp1-D and eight other homologues, seven of which have ORFs that could potentially code for proteins similar to RP1-D, and one of which is truncated after the NBS encoding domain. All of the 22 susceptible variants analyzed from HRp1-D haplotype homozygotes or 32 variants from HRp1-A/HRp1-D or HRp1-B/HRp1-D heterozygotes had mutations or recombination events that deleted or altered the Rp1-D gene. In several of the variant HRp1-D haplotypes, including two insertion mutants (Collinset al. 1999), only the Rp1-D gene was altered. Previous examination of these variants with a diverse collection of P. sorghi biotypes indicated they were susceptible to all biotypes (Richteret al. 1995). The Rp1-D gene is therefore the only gene in this haplotype with a detectable rust resistance phenotype. This is in contrast to several other clustered resistance gene families that have been examined. Loci, like Cf4/9, Cf2/5, and Xa21 (Dixonet al. 1996; Parniskeet al. 1997; Wanget al. 1998) have multiple genes in a single haplotype that contribute to resistance. The intact ORFs of most of the rp1 genes are also in contrast to certain other resistance loci, like Xa21 (Songet al. 1997) and Rpp5 (Noëlet al. 1999), where many of the family members have obvious mutations, such as insertions or ORF-disrupting point mutations. The function of the rp1 paralogues is not clear; it is possible they could provide resistance to rust biotypes that are obscure or no longer extant or even other nonrust pathogens.
No transcripts could be found for two of the HRp1-D haplotype genes with full-length ORFs. In addition, one of the three HRp1-K haplotype genes characterized had a single stop codon located 111 bases after the putative start codon in what would otherwise be a full-length ORF. Most of the susceptible recombinants from the HRp1-D haplotype homozygotes resulted in various amounts of the Rp1-D coding region being recombined with a gene with an apparently nonfunctional promoter. For example, the recombinant gene in the HRp1-D*19 haplotype had the LRR region from Rp1-D and differed from Rp1-D by only 13 amino acids, but no transcript for the recombinant gene could be identified. The presence of these unused, but largely intact, coding regions in rp1 haplotypes raises the question of whether there are some types of mechanisms that maintain them. Unequal recombination events probably play a role since they have a natural tendency to homogenize repeat units (Smith 1976). Some of the genes may be useful after they are activated by recombination events that donate new 5′ sequences, or useful as sequence donors in recombination events to make novel genes, but these sequences would not be expected to provide a large selective advantage to the plant that carried them.
Recombination and generation of novel alleles: Previous genetic studies have indicated that the rp1 complex consists of sequence duplications that can mispair and recombine during meiosis, generating haplotypic variation (Hulbert 1997; Collinset al. 1999). It seemed likely that at least some of these recombination events occurred within the rp1 paralogues since they were occasionally associated with seedlings having novel phenotypes (Richteret al. 1995; Huet al. 1996). In this analysis, a collection of 22 variants from a HRp1-D haplotype homozygote were examined. By completely or partially sequencing the recombinant rp1 gene from the individual variants, 13 of these were determined to be derived from crossovers between paralogues. Another 7 also appeared to be derived from very similar events on the basis of restriction fragment analysis. Only two of the variants could not be explained by intragenic crossover events. Seventeen of the 22 variants represented crossovers between the Rp1-D gene, at the distal end of the array, and the untranscribed rp1-dp5 gene in the middle of the array. Since no clone carried more than one paralogue, the intergenic regions between the rp1 genes appeared to be at least 15-20 kb (possibly much larger), indicating that the two paralogues are at least 80 kb apart. The frequency of this class of susceptible recombinants was ∼1/10,000. The reciprocal crossover, which would result in a duplication and not a deletion, would not have been selected. The estimated rate of recombination between these two paralogues is therefore ∼1/5000. It is possible that the overall rate of recombination between Rp1-D and some of the other paralogues is similar to that observed with rp1-dp5, especially those that exhibit higher sequence homology. Since most of the other genes were transcribed, some of these recombination events may have resulted in a functional Rp1-D specificity and would not have been selected. Intragenic recombination events between paralogues in the Rp1 complex may therefore be very common. This type of recombination will promote coevolution of the gene family as well as reshuffle them into new combinations. Recombination between paralogues at other clustered resistance genes may be much less frequent. For example, genes in the Dm3 and Pto gene clusters appear to evolve more independently, since genes in orthologous positions in different lines appear to be more similar than paralogues (Michelmore and Meyers 1998). Spontaneous susceptible variants have been characterized from other clustered resistance gene families. Recombination events have been examined at complex Cf loci of tomato. When five susceptible recombinants were isolated from heterozygotes of the Cf4/9 locus, all five of the recombination breakpoints mapped to the intergenic regions (Parniskeet al. 1997). However, a single susceptible recombinant from a Cf2/Cf5 heterozygote was found to be intragenic (Dixonet al. 1998). It seems that different complex R gene loci may vary in the extent to which paralogues recombine.
Analysis of some Arabidopsis Rpp5 and flax M mutants revealed unequal crossing over between duplications in the regions encoding the LRR sequences within a single gene, resulting in changes in the number of duplications (Andersonet al. 1997; Parkeret al. 1997). One Rpp5 mutant recovered from a mutagenesis experiment was found to carry a duplication corresponding to four LRRs (154 amino acids) when compared to the parental Rpp5 gene. Similarly, susceptible variants at the flax M locus were identified with a deletion of one direct repeat (corresponding to 45 amino acids) in the LRR domain (Andersonet al. 1997). Sequence comparison of different alleles at the tomato Cf2/Cf5 locus and flax L locus (Dixonet al. 1998; Elliset al. 1999) revealed some alleles with variation in the number of LRRs, presumably resulting from the unequal crossing over between the internal duplications during the evolution of these gene families. No such unequal crossing-over events were observed in this study. Comparisons of all nine full-length rp1 genes from the HRp1-D haplotype and three from the HRp1-K haplotype provided no evidence of this type of recombination.
Another feature of the recombination events at the rp1 locus is that the recombination breakpoints are distributed throughout the genes. Members of the rice Xa21 gene family contain a conserved region shortly after the putative translation start in which recombination had most likely occurred during the evolution of the Xa21 haplotype (Songet al. 1997). No obvious recombination hotspots were observed among the rp1 genes. Studies of other maize loci, such as bronze and waxy, also revealed that the recombination breakpoints were well distributed across the genes (Dooner and Martinez-Ferez 1997). However, alleles of these simple loci usually exhibit levels of sequence identity higher than those of different family members at resistance gene clusters like rp1 or Xa21. At other maize loci, like al, bl, and rl, recombination breakpoints were more frequently resolved in certain region of the genes (reviewed in Schnableet al. 1998). The factors controlling the distribution of recombination breakpoints in plants are not well characterized and are in need of additional research.
Mutation, selection, recombination, and Rp1 gene function: Comparative analysis of the rp1 genes indicates that mutation plays an important role in the evolution of the family. Compared with the functional Rp1-D allele, all the other Rp1-D family members have numerous differences, with the 5′ regions of the genes (including the NBS domain) being more conserved than the 3′ regions (corresponding to the LRR). This is consistent with the predicted functions of different domains of R genes. The NBS domain is thought to be the effector domain involved in transducing a recognition signal and not the domain most likely to be directly involved in ligand binding (Traut 1994; Bakeret al. 1997). Recently, significant amino acid sequence similarity has been found between the NBS domain of NBS-LRR resistance proteins and the nematode CED-4 and mammalian Apaf-1 proteins, which are activators of apoptotic proteases (Chinnaiyanet al. 1997; van der Biezen and Jones 1998). The LRR domain is well documented as a functional structure involved in protein-protein interaction (Kobe and Deisenhofer 1994, 1995). The LRR of plant disease resistance genes could participate in the binding of pathogen-derived avr factors directly or indirectly (Ellis and Jones 1998) and thus variability of this region will result in changes in recognition capability. This hypothesis is supported by sequence comparisons of resistance gene alleles or gene family members, in which codons for predicted solvent-exposed residues of the LRR regions show evidence for diversifying selection (reviewed in Elliset al. 2000). This pattern of diversifying selection has been observed for several other types of genes involved in nonself recognition (Hughes and Nei 1988; Endoet al. 1993). In some R gene families, the evidence for diversifying selection may not be apparent through the whole LRR region. For example, in the lettuce RGC2 gene family, only the putative solvent-exposed residues in the 3′-encoded LRR show good evidence of diversifying selection (Meyerset al. 1998b). The Rp1-D gene product has 24 imperfect LRRs that can be divided into two parts (domains B and D) that are separated by a region with no LRR consensus (Collinset al. 1999). Only domain D, at the 3′ end of the gene, shows strong evidence for diversifying selection. Ayliffe et al. (2000) found a similar result by comparing sequences of rp1 homologues from barley. This may be an indication that the carboxy terminus of the Rp1 family proteins is most directly involved in ligand binding. It is also clear, however, that LRR sequences from domain B are functionally important. The DNCO3 mutant allele is identical to the Rp1-D gene except for a 33-nucleotide deletion corresponding to 11 aa in the 10th LRR. Interestingly, this corresponds to a highly variable region of the gene.
Most of the Rp1 variants selected are completely susceptible to all the rust biotypes in our current collection and the majority of these were a consequence of an unequal exchange between Rp1-D and rp-dp5, a paralogue for which there is no evidence of expression. Most of these exchanges result in a gene with 3′ sequences from Rp1-D and 5′ sequences from rp1-dp5. Other unequal exchanges may generate recombinant genes that either code for nonfunctional proteins or for proteins that will not interact with any of the rust biotypes in our current collection. Among the Rp1 variants analyzed, only Rp1-D*5 retains race-specific resistance, conferring resistance to the same spectrum of rust isolates as the parental Rp1-D, but at a reduced level. Characterization of the recombinant allele from Rp1-D*5 revealed it had the LRR of the Rp1-D gene, having resulted from a crossover in the NBS domain. The lower level of resistance in Rp1-D*5 may be attributed to the differences in the NBS region between the recombinant gene and the wild-type Rp1-D or to promoter strength. Site-directed mutagenesis in the NBS region of the tobacco N gene demonstrated that some mutations in the P loop of the NBS can result in loss or partial loss of R gene function (Bakeret al. 1997). The identical race specificity of the recombinant Rp1 gene and the Rp1-D gene and their identical LRRs are consistent with the idea that this domain is responsible for controlling race specificity. However, the recombinant gene in the HRp1-D24 haplotype also has the LRR from the Rp1-D gene but shows no resistance to any of the rust isolates in our collection. The 5′ sequences of the recombinant gene were donated by rp1-dp1, which is transcribed. These 5′ sequences are probably either not functional in conferring a resistance reaction or not compatible with the LRR domain of the Rp1-D gene. This latter phenomenon has been observed in domain swaps between alleles of the L gene of flax (Elliset al. 1999; Lucket al. 2000) and Mi gene family members in tomato (Hwanget al. 2000). The phenotypes of these recombinant genes indicate that the 5′ regions of NBS-LRR genes are adapted to function with specific LRR regions in the genes. Swapping different 5′ regions with different LRRs can lead to loss of resistance, spontaneous necrosis (Hwanget al. 2000), or even unexpected race specificities (Lucket al. 2000).
The recombinant gene in the HRp1-D*21 haplotype was derived from a crossover in the LRR domain and lost its race-specific resistance. The recombinant LRR may have lost the ability to interact with the fungal elicitor. It is also possible that the 5′ sequences donated by the rp1-hd2 gene are not functional or will not function with Rp1-D LRR sequences. The HRp1-D*21 haplotype confers an unusual nonspecific reaction to rust and a necrotic-spotting phenotype (Huet al. 1996) indicating it can induce hypersensitive reactions, but we have not yet demonstrated that necrosis is caused by the recombinant gene and not by some other aspect of the novel haplotype.
The analysis of the maize rp1 locus has shown that both recombination and mutation are active forces in the evolution of this resistance gene family. Analysis of spontaneous mutants (selection of susceptibles) demonstrated that unequal crossing over occurs frequently at the locus and that the crossovers are frequently intragenic. This was also indicated by the stretches of perfect sequence identity observed between paralogues in a single haplotype. The sequence analysis also indicates that selection pressure favors diversifying selection in the carboxy-terminal half of the LRR of the protein. Recombination and mutation therefore appear to be intimately responsible for generating new resistance specificities at the maize rp1 locus.
Acknowledgments
This work was supported in part by grants MCB-9728490 and 9975971 from the National Science Foundation. This article is contribution 01-159-J from the Department of Plant Pathology, Kansas Agricultural Experiment Station.
Footnotes
-
Communicating editor: J. A. Birchler
- Received November 16, 2000.
- Accepted February 13, 2001.
- Copyright © 2001 by the Genetics Society of America
