By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Due to essential maintenance the subscribe/renew pages will be unavailable on Wednesday 26 October between 02:00- 08:00 BST/ 09:00 – 15:00 SGT/ 21:00- 03:00 EDT. Apologies for the inconvenience.
Disease resistance (R) genes are found in plants as either simple (single allelic series) loci, or more frequently as complex loci of tandemly repeated genes. These different loci are likely to be under similar evolutionary forces from pathogens, but the contrast between them suggests important differences in mechanisms associated with DNA structure and recombination that generate and maintain R gene diversity. The RPP13 locus in Arabidopsis represents an important paradigm for studying the evolution of an R gene at a simple locus. The RPP13 allele from the accession Nd-1, designated RPP13-Nd, confers resistance to five different isolates of the biotrophic oomycete, Peronospora parasitica (causal agent of downy mildew), and encodes an NBS-LRR type R protein with a putative amino-terminal leucine zipper. The RPP13-Rld allele, cloned from the accession Rld-2, encodes a different specificity. Comparison of three RPP13 alleles revealed a high rate of amino acid divergence within the LRR domain, less than 80% identity overall, compared to the remainder of the protein (>95% identity). We also found evidence for positive selection in the LRR domain for amino acid diversification outside the core conserved β-strand/β-turn motif, suggesting that more of the LRR structure is available for interaction with target molecules than has previously been reported for other R gene products. Furthermore, an amino acid sequence (LLRVLDL) identical in an LRR among RPP13 alleles is conserved in other LZ NBS-LRR type R proteins, suggesting functional significance.
Plants are able to perceive potential pathogens and mount defence responses that are associated with the restriction of pathogen growth and reproduction. Such responses are characterized by numerous physiological, biochemical and molecular changes, and are frequently accompanied by the occurrence of rapid host cell death (so-called hypersensitive response) in cells penetrated by or surrounding the pathogen (reviewed by Hammond-Kosack & Jones 1996). Elicitation of these defence-associated responses is often conditional on the presence of a single plant resistance (R) gene and a complementary pathogen avirulence (Avr) gene ( Flor 1971; Keen 1990).
The majority of cloned R gene products are predicted to be located within the cytoplasm and most have leucine-rich repeat (LRR) domains coupled to a putative nucleotide binding site (NBS) domain (reviewed by Hammond-Kosack & Jones 1997). Genes of this type are used to recognize a range of plant pathogens. Structural comparisons allow some members to be further classified into two subgroups consisting of R proteins that share regions of homology near their N-termini with the cytoplasmic signalling domain of the Drosophila Toll protein and mammalian interleukin-1 receptors (TIR domain), and those that contain putative leucine zippers (LZs). Amongst the latter class are genes that confer resistance to bacteria, RPS2, RPM1 and RPS5 ( Bent et al. 1994 ; Grant et al. 1995 ; Mindrinos et al. 1994 ; Warren et al. 1998 ), a nematode, Mi-1.2 ( Milligan et al. 1998 ), and fungi, RPP8-Ler and I2 ( McDowell et al. 1998 ; Simons et al. 1998 ).
Evidence is accumulating to suggest that the LRR domain plays a key role in signal perception and transduction. Loss-of-function mutant alleles have been shown to contain single amino acid alterations within their respective LRR domains ( Bent et al. 1994 ; Grant et al. 1995 ; Mindrinos et al. 1994 ), and recently a single amino acid change within a conserved LRR of the RPS5 gene resulted in pleiotropic effects on other R genes found in the same Arabidopsis accession ( Warren et al. 1998 ). Comparison of nucleotide substitution patterns among members of R gene loci from a number of plant species has revealed evidence for positive selection for amino acid diversification within predicted solvent-exposed residues of the conserved LRR β-strand/β-turn motif ( Botella et al. 1998 ; Dixon et al. 1998 ; McDowell et al. 1998 ; Meyers et al. 1998 ; Parniske et al. 1997 ). Recently, Leckie et al. (1999) used site-directed mutagenesis to demonstrate that solvent-exposed β-strand/β-turn motif residues were key in determining both the affinity and specificity of the interaction between polygalacturonase-inhibiting proteins and their polygalacturonase ligands. By analogy, the hypervariable solvent exposed residues that are found in the LRR β-strand/β-turn motif of many R genes may also play a role in ligand binding and recognition specificity.
The proliferation of R genes in plants is thought to be a consequence of a continual challenge from rapidly evolving pathogens that eventually overcome most R genes ( Bogdanove et al. 1998 ; Joosten et al. 1994 ; Rohe et al. 1995 ). New pathogen-recognition capabilities have presumably been generated from mutations or recombination events that alter the amino acid sequence of a resistance gene, which are subsequently maintained as allelic variation in the species either by distribution of alleles amongst individuals of a population or by proliferation of genes within an individual. Allelic series at single gene loci provide evidence for the maintenance of alleles amongst individuals ( Islam & Shepherd 1991; Jørgensen 1992), and the latter is suggested by the common observation that disease resistance genes are often clustered in regions smaller than 1 c m (complex loci) (reviewed by Crute & Pink 1996). Complex loci may be especially dynamic for generating new haplotype combinations and new alleles through intergenic (e.g. Parniske et al. 1997 ) or intragenic (e.g. Dixon et al. 1998 ) recombination. Expansion or contraction in LRR copy number may be another mechanism by which plants are able to generate new recognition capabilities ( Anderson et al. 1997 ; Botella et al. 1998 ; Dixon et al. 1998 ; Ellis et al. 1999 ; Meyers et al. 1998 ; Parker et al. 1997 ).
More than 20 loci for genes conferring specific recognition of Peronospora parasitica (RPP) have been identified distributed among the five chromosomes of Arabidopsis (reviewed by Holub & Beynon 1997), and genes from three loci, RPP1, RPP5 and RPP8, have been cloned ( Botella et al. 1998 ; McDowell et al. 1998 ; Parker et al. 1997 ). Previously we described the RPP13 locus which mapped to the bottom arm of chromosome 3 and conferred resistance to five P. parasitica isolates, including Maks9, in the Arabidopsis accession Nd-1 ( Bittner-Eddy et al. 1999 ). Here we report the molecular characterization of the RPP13 locus in three accessions, including an allele (RPP11;Joos et al. 1996 ) from Rld-2. The locus comprises a single gene predicted to encode an NBS-LRR protein with an amino-terminal LZ. Two RPP13 alleles are described which specify resistance to different avirulence determinants in P. parasitica. This is the first report in Arabidopsis of a simple disease resistance locus for functionally distinct alleles. Furthermore, the structure of the protein implies conservation with other R genes for the second LRR that may have functional significance. The amino acid sequence also suggests that more of the LRR structure is available for interaction with target molecules than has previously been reported for other disease-resistance genes.
Identification of the RPP13 gene
The RPP13 locus maps to a region of chromosome 3 flanked by RFLP loci identified by clones p3037-3 (above) and g4117 (below) ( Bittner-Eddy et al. 1999 ). Clones from a bacterial artificial chromosome (BAC) library of Col-0 genomic DNA ( Mozo et al. 1998 ) were obtained using p3037-3 as a hybridization probe. Four BAC clones (F21O12, F7B1, F1P5 and F6L9) were isolated and arranged physically by Southern hybridization using existing molecular markers and total BAC clone DNA (data not shown). The physical order of these clones is shown in Fig. 1(a). This analysis revealed that the BAC clone contig, although partially overlapping with the g4117 clone, did not extend as far as the RFLP locus which defined the lower boundary of the interval encompassing RPP13. BAC clone F21O12 was subcloned and new molecular markers were sought that lay between the cosegregating marker p3037-2 and the RFLP locus identified by g4117. We identified nucleotide variation between Col-5 and Nd-1 PCR products generated by primers designed from clones p21O12-7 and p21O12-8 (see Experimental procedures), enabling us to demonstrate that both clones cosegregated with the g4117 RFLP locus in our Col-5 × Nd-1 mapping populations. Therefore BAC clone F21O12 spans RPP13.
Cosmid binary vector libraries were screened using p3037-2, p3037-3 and p21O12-8 as hybridization probes. Five cosmids were isolated, ranging in size from 24 to 28 kb. Figure 1(b) shows the order of these cosmids based on data from EcoRI restriction fingerprinting and Southern blot analysis (data not shown).
Cosmids pBa3056, pBa3057, pBa3058 and pBa3060 were identified as giving the best overlapping coverage of DNA in the RPP13 interval and were transformed into Arabidopsis accession Col-5 that is susceptible to isolate Maks9. Primary transgenic plants (T1) were selected for resistance to dl-phospinothricin (PPT). T2 progeny segregated for PPT resistance with a ratio of approxi-mately 3 : 1 indicating the presence of a single transgene locus in each line. T2 progeny were initially inoculated with the P. parasitica isolate Maks9. Resistance to Maks9 was observed in transgenic Col-5 lines carrying either the pBa3058 or pBa3060 cosmid, but not in lines carrying pBa3056 or pBa3057 ( Table 1). This indicated that RPP13 must lie within the 7.3 kb region in common between pBa3058 and pBa3060. A 5.7 kb EcoRI fragment from pBa3058 was cloned into a cosmid binary vector to generate pBa3058-4, and was used to produce transgenic Col-5 plants. Complete association of Maks9 resistance in transgenic Col-5 plants transformed with pBa3058-4 ( Table 1) confirmed that we had cloned RPP13 and that it lay within the 5.7 kb EcoRI fragment.
Table 1. . Interaction phenotypes observed among wild-type and transgenic Col-5 lines following inoculation with six different Peronospora parasitica isolates
a At least 20 T 2 progeny from each transgenic line were tested. Among T2 progeny from transgenic lines containing pBa3058, pBa3060 or pBa3058-4 segregation for resistance at approximately 3 : 1 (resistant : susceptible) was observed. An absolute correlation between PPT and disease resistance was observed in these lines.
Waco5 is recognized by a gene that maps to the RPP1 locus of Nd-1 and was included in this analysis as a control.
Lines were judged as resistant (R) or susceptible (S) following the observation of consistent phenotypes.
We previously showed that resistance to P. parasitica isolates Aswa1, Edco1, Emco5 and Goco1 also mapped to the RPP13 locus in Nd-1 ( Bittner-Eddy et al. 1999 ). We tested T2 progeny and found segregation for resistance to all four isolates among progeny of Col-5 transgenic lines transformed with pBa3058-4; no resistance was observed in seedlings transformed with pBa3056 or pBa3057, neither of which carries RPP13-Nd ( Table 1). Furthermore, resistance was specific to these isolates; pBa3058-4 did not confer resistance to isolate Waco5, recognition of which had previously been mapped to the RPP1 locus in Nd-1. This result demonstrates that a single gene at the RPP13 locus in Nd-1 is sufficient for recognizing several different P. parasitica isolates of diverse origin.
The RPP13 allele from the Rld-2 accession confers resistance to Wela3 but not to P. parasitica isolates recognized by RPP13-Nd
We cloned the RPP13 allele from Rld-2 using long-range PCR. Several independent clones were generated in a cosmid binary vector and their sequence compared with that of the Rld-2 sequence determined. Three clones were selected to generate transgenic plants: clone pBaRld2-WT matched the wild-type Rld-2 sequence; clone pBaRld2-GN is predicted to have two amino acid changes at codons 486 (G for an R) and 627 (N for a D); clone pBaRld2-TGA has a stop codon at position 697(TGA for W) which is predicted to result in a truncated protein product missing the last five LRRs (the latter variant clones arose through PCR misincorporation of nucleotides). Col-5 is resistant to Wela3 ( Holub et al. 1994 ), so we selected an inbred line (HRI3860) from the Col-5 × Nd-1 mapping population that is susceptible to both Wela3 and Maks9 as recipient for the Rld-2 transgenes. Transgenic HRI3860 plants containing the Rld-2 allele were resistant to Wela3, but not to the isolates recognised by RPP13-Nd ( Table 2). The ability of RPP13-Rld to confer resistance to Wela3 was not compromised by amino acid changes at codons 486 and 627, indicating that these residues are not critical for maintaining RPP13-Rld function and/or specificity. However, loss of the last five LRRs destroyed the ability of RPP13-Rld to recognize Wela3. This analysis demonstrates that RPP13 is a locus for at least two functionally distinct alleles that confer resistance to different avirulence determinants in P. parasitica.
Table 2. . Interaction phenotypes observed in transgenic HRI3860 plants carrying RPP13-Rld following inoculation with a range of Peronospora parasitica isolates
The RPP13 locus consists of a single gene in accessions Col-5 and Nd-1
We used a probe encompassing the RPP13-Nd coding region to determine the number of R gene-like sequences present at the RPP13 locus. We initially used low-stringency hybridization conditions in order to detect related sequences within the locus and from elsewhere in the genome; the result of this Southern blot analysis is shown in Fig. 2. Only under low-stringency conditions could the hybridizing bands of approximately 10 and 18 kb be detected in Nd-1 and Col-5 DNA, respectively. These bands have no counterparts in either the cosmids or BAC clone that span the RPP13 locus in either Nd-1 or Col-5, and therefore we do not consider them part of that locus. BLAST searches ( Altschul et al. 1997 ) reveal candidate genes that show approximately 60% homology to RPP13-Nd at the amino acid level, but we have not confirmed the significance of this observation by mapping the weakly hybridizing bands relative to the RPP13 locus (P.D. Bittner-Eddy, unpublished results). When we used more stringent hybridization conditions, these weakly hybridizing bands were not detected even though the conditions used were sufficient to detect all members of the complex RPP1 locus ( Botella et al. 1998 ) when using a probe specific to RPP1 (data not shown). The presence of a polymorphic EcoRI site within the single Col-5 gene (absent in RPP13-Nd) accounts for the two hybridizing bands detected in BAC F21012. We consider it likely that the RPP13 locus also contains a single R gene sequence in both the Rld-2 and Ws-2 accessions. DNA sequence (see below; P.D. Bittner-Eddy, unpublished results) revealed an internal EcoRI site that accounted for the two bands of 3.5 and 4.4 kb detected in Ws-2 and Rld-2. The additional hybridizing bands in these two tracks also disappeared under the more stringent hybridization conditions.
Structure of the RPP13-Nd gene product
We sequenced the 5.7 kb EcoRI fragment containing the RPP13-Nd gene. A single large ORF of 2463 bases was found. No other significant intact ORFs were identified using database homology searches ( Altschul et al. 1997 ) or a web-based gene prediction program (see Experimental procedures). The sequence of this ORF (RPP13-Nd), including sequence 5′ and 3′, has been deposited in GenBank (AF209732). A potential TATA box signal (TAATATA) is located 124 bases upstream of a predicted translation initiation signal, which itself is preceded by an in-frame stop codon. PCR amplification products corresponding to RPP13-Nd were generated from an Nd-1 cDNA library using gene-specific primers and primers designed against the 5′ end of the cDNA vector (see Experimental procedures). The sequence of these PCR products confirmed that RPP13-Nd has no introns and that transcription begins at least 29 bases upstream of the predicted translation initiation point.
RPP13-Nd encodes a protein of 820 amino acids. This protein is predicted to reside within the cytoplasm and has many sequence motifs found in R proteins of the NBS-LRR class ( Fig. 3). The LRR domain of RPP13-Nd is located within the C-terminal third of the gene and consists of 11 imperfect repeats ranging in size from 18 to 30 amino acids with a possible truncated repeat located at the 3′ end. Figure 4(a) shows an alignment of these LRRs and a derived consensus sequence. Within the LRR consensus, three subdomains could be recognized that have counterparts in other R proteins, including a predicted β-strand/β–turn motif and possible βα loop region ( Jones & Jones 1997). The LRR consensus sequence for RPP13-Nd broadly agrees with the consensus LRR sequence determined for other R proteins that are predicted to reside within the cytoplasm ( Jones & Jones 1997). However, the RPP13-Nd repeats are more varied both in their size and in their amino acid composition and exceptionally an arginine residue is conserved within the predicted β-strand/β-turn motif.
Overall, RPP13-Nd is quite dissimilar to other known R proteins, being closest in sequence to RPP8-Ler sharing 28% amino acid similarity. However, alignment of RPP13-Nd with other LZ NBS-LRR type R proteins revealed two regions of homology that may have functional or structural significance. An alignment of the N-terminal regions of RPP13-Nd, RPP8-Ler and RPM1 is shown in Fig. 5(a). The respective LZs align with each other, and the spacing between the LZ and N-terminal end is maintained in all three proteins. Outside the conserved motifs that comprise the NBS, only one other region of significant homology was seen. Figure 5(b) shows an alignment between the first four LRRs from RPP13-Nd and the corresponding LRRs from a number of other LZ NBS-LRR type R proteins that reveals a conserved sequence element (LLRVLDL) in the second LRR. The significance of this sequence element is underscored by the pleiotropic rps5-1 mutant, in which an adjacent glutamic acid residue is mutated ( Warren et al. 1998 ).
Comparison of RPP13-Nd, RPP13-Rld and rpp13-Col
We sequenced the RPP13 alleles from the Rld-2 and Col-5 accessions to allow comparisons to be made with RPP13-Nd. Overall, RPP13-Nd shares 89% amino acid identity with the predicted protein products of RPP13-Rld and rpp13-Col.Figure 3 shows the alignment of the three hypothetical proteins. The proteins are highly conserved over the first two-thirds of their length, including the LZ and NBS motifs and the conserved hydrophobic domain. The majority of amino acid variation occurs within the C-terminal LRR domain. Eight of the 11 indels (two independent events probably caused the indel between codons 712 and 713 of RPP13-Nd and codons 711 and 712 of RPP13-Rld) also occur within the C-terminal third of the gene, indicating that this region is more tolerant of gross amino acid differences as well as individual amino acid variability.
We superimposed the amino acid sequence of individual LRRs from RPP13-Rld and rpp13-Col onto the corresponding LRR of RPP13-Nd to analyse the extent of amino acid variation within individual repeats. The alignment and analysis of individual LRR variation is shown in Fig. 4(a,b). No two corresponding repeats were identical (except the truncated twelfth) and a range of variation was observed. LRRs 2, 3, 4 and 7 exhibited the greatest overall variation, and LRRs 5 and 9 showed the greatest overall degree of sequence conservation, perhaps reflecting different roles for these LRRs. Significantly, however, the amino acid sequence LLRVLDL identified in the second LRR of RPP13-Nd is invariant between the three alleles, yet adjacent residues are hypervariable.
A mosaic of amino acid sequence variability was also observed within individual LRRs, with a number of positions within the consensus sequence exhibiting hypervariability. We compared amino acid variation within the β-strand/β-turn motif and two other defined regions of the RPP13 LRR consensus ( Table 3). We found that the putative βα loop region was as variable as the β-strand/β-turn motif. Twelve out of 22 hypervariable sites, where a difference is observed in all three alleles, and six of the 11 indels are found in the putative βα loop subdomain ( Fig. 4a). This suggests that the putative βα loop region may be as important for RPP13 alleles in generating novel ligand-binding capacity as are the solvent-exposed residues of the β-strand/β-turn motif.
Table 3. . Paired comparisons of synonymous and non-synonymous nucleotide substitutions and amino acid variability between four regions of RPP13-Nd, RPP13-Rld and rpp13-Col
a The regions analysed are defined in Fig. 4(a); Gene (-LRR) is the coding sequence with the LRR domain removed.
Amino acid identity.
c The rates of synonymous (Ks) or non-synonymous (Ka) nucleotide substitutions per synonymous/non-synonymous site were calculated using diverge as described by Parniske et al. (1997) . Sequence pairs were first aligned codon by codon and indels removed prior to analysis.
d Codons for the conserved amino acid positions (in parentheses) of the consensus XX(L)X(α)XX were removed from sequence files generated for nucleotide substitution analysis on the basis that they are thought to serve a conserved structural role, and therefore would be likely to be under selection for sequence conservation ( Jones & Jones 1997; Parniske et al. 1997 ).
Comparison of non-synonymous (Ka) versus synonymous (Ks) nucleotide substitution rates in mammalian genes involved in pathogen recognition has demonstrated that positive selection occurs for amino acid diversification in regions involved with ligand binding ( Hughes & Nei 1988). A Ka/Ks ratio >1 is seen as an indication of diversifying selection pressure, whereas a ratio <1, which occurs for most genes, indicates selection for amino acid conservation ( Endo et al. 1996 ; Kreitman & Akashi 1995). Recently this type of analysis has been extended to plant R genes, and has revealed that the β-strand/β-turn motif is undergoing selection for amino acid diversity ( Botella et al. 1998 ; McDowell et al. 1998 ; Meyers et al. 1998 ; Parniske et al. 1997 ; Wang et al. 1998 ). We also found evidence in all three paired comparisons for diversifying selection acting on the β-strand/β-turn motif ( Table 3). For example, in our RPP13-Nd versus rpp13-Col comparison, the Ka/Ks ratio was 1.93 and non-synonymous nucleotide substitution rates were at least 12 times higher than those calculated for the non-LRR portion of the gene. In addition to the β-strand/β-turn motif, Ka/Ks ratios >1 suggest that the two other subdomains identified within the LRR consensus sequence are also undergoing diversifying selection, in contrast to what has been found at other R gene loci. Even though region 1 of the LRR consensus is more conserved relative to the putative βα loop region and β-strand/β-turn motif (86–90% amino acid identity), the majority of nucleotide differences in this region have resulted in an amino acid change leading to a Ka/Ks ratio >1. The Ka/Ks ratio is not significantly changed if codons for the conserved hydrophobic amino acids are removed from the analysis (data not shown). Overall, the Ka and Ks values for the LRR domain are higher than those of the remainder of the gene, implying that, in addition to positive selection for amino acid diversity, this region is also evolving more rapidly.
Comparison of RPP13 with structurally similar R genes
The RPP13 locus, at least in the Arabidopsis accessions we have examined, is reminiscent of two other single-copy R gene loci found in Arabidopsis, RPM1 ( Grant et al. 1995 ) and RPS2 ( Bent et al. 1994 ; Mindrinos et al. 1994 ). However, an important distinction is that at RPP13, with the alleles from Nd-1 and Rld-2 accessions, we have the first example in Arabidopsis of functionally diverged alleles at a simple R gene locus.
Specific LRRs may mediate interactions between the R protein and other protein components of the pathogen resistance signalling pathway ( Jones & Jones 1997; Warren et al. 1998 ). For example, Jones & Jones (1997) speculated that one role of the group of highly conserved LRRs that are found in the C-terminal region of Cf genes was to interact with a common protein component. We found that the second LRR of the RPP13 proteins is one of the more divergent repeats when compared between alleles ( Fig. 4a,b). However, within the repeat the amino acid sequence LLRVLDL is absolutely conserved, and when compared to the LRRs of other LZ-NBS-LRR type R proteins, remarkable conservation of this sequence is observed ( Fig. 5b). Warren et al. (1998) reported the Arabidopsis mutant rps5-1, which partially compromises the function of a number of R genes that confer resistance to either P. parasitica or P. syringae in the Col accession. The mutation resulted in a non-conservative amino acid substitution of a glutamic acid residue ( Fig. 5b) that lies in close proximity to the conserved LLRVLDL sequence. Warren et al. (1998) speculated that the mutation in rps5-1 increased the stability and/or binding affinity of rps5-1, allowing it to sequester a protein that may be required for resistance mediated by multiple R genes. Although the glutamic acid residue is not conserved among the RPP13 alleles or RPP8-Ler, we speculate that the pleiotropic effect of the rps5-1 mutation may be due to the specific amino acid change and its effect on the conserved element within the second LRR to mediate the binding of a common protein component. We are currently crossing transgenic Col lines carrying RPP13-Nd to the Col rps5-1 mutant to examine whether the rps5-1 protein may also interfere with RPP13-Nd function.
Comparison of the three RPP13 alleles
The nature of the RPP13 locus will allow study of the evolution of an R gene that is presumed to have been under constant selection pressure in response to variation for virulence in P. parasitica (and perhaps other parasites). Alleles can be readily compared without the complication of haplotypic variation among allelic variants within complex loci. Our comparison of RPP13-Nd, RPP13-Rld and rpp13-col revealed that they were highly conserved over the first two-thirds of their length encompassing the LZ, NBS motifs and the conserved core hydrophobic domain, but excluding the LRR domain ( Fig. 3). This conservation was reflected in the fact that many of the amino acid substitutions in this region were conservative, and that the Ka/Ks ratio calculated for all three paired comparisons was less than 1 ( Table 3). This indicates that the nucleotide substitutions that have occurred favoured maintenance of amino acid sequence. Conservation of this part of the gene probably reflects the need to maintain the tertiary structure facilitating nucleotide binding and any protein–protein interaction mediated by the LZ. We found that most of the variation between the three RPP13 proteins occurred in the LRR domain. For example, in our comparison between RPP13-Nd and rpp13-Col, variation in the three LRR subdomains ranged from 70 to 86% amino acid sequence identity, yet the remainder of the two proteins were 96% identical. The bipartite nature of R gene variation has also been observed among alleles of the L locus ( Ellis et al. 1999 ) and among members of more complex R gene loci ( Botella et al. 1998 ; Dixon et al. 1998 ; Meyers et al. 1998 ; Parniske et al. 1997 ). Variation between the LRR domains of RPP13-Nd, RPP13-Rld and rpp13-Col supports the contention that the LRR domain is likely to be involved in ligand binding ( Jones & Jones 1997), and may determine specificity through interaction with avirulence determinants produced by the incompatible P. parasitica isolates. However, recent results have implicated both the LRR and TIR domains of alleles at the L locus in determining specificity of rust resistance ( Ellis et al. 1999 ), and involvement of other regions of the RPP13 protein in determining specificity can not yet be ruled out.
The RPP13-Nd gene confers resistance against at least five P. parasitica isolates ( Table 1). It is likely that the gene product recognizes the same avirulence determinant present in all five isolates. RPP13-Rld confers resistance to a different isolate and therefore recognizes a different avirulence determinant. Whether or not these two RPP13 alleles respond to the products of two different Avr genes or different allelic forms of the same gene remains unknown. Examination of segregation among progeny from crosses between Wela3 and Maks9, for example, would enable independence or allelism of the Avr genes to be determined. We have yet to identify any P. parasitica isolates that are incompatible on the Col-5 accession due to a gene that maps to the rpp13-Col locus. However, the structure of rpp13-Col and the positive selection for amino acid diversification identified within its LRR domain ( Table 3) strongly suggest that the gene is expressed, although currently its function is unknown.
Extended areas of the LRR in RPP13 are under diversifying selection
The ability to generate amino acid diversity within the LRR domain of R genes is likely to be a key component of a plant’s evolutionary response to selection imposed by constantly changing pathogen avirulence determinants. From the detailed molecular analysis of a number of different R gene types, there are several different mechanisms by which this is achieved (reviewed by Michelmore & Meyers 1998). Intragenic recombination between DNA encoding repeated LRR units has been shown to result in an expansion or contraction of the LRR copy number in a number of genes at contrasting R gene loci ( Anderson et al. 1997 ; Botella et al. 1998 ; Dixon et al. 1998 ; Ellis et al. 1999 ; Meyers et al. 1998 ; Parker et al. 1997 ; Simons et al. 1998 ). We observed no variation in the LRR copy number of the three RPP13 alleles examined. This could be due to the fact that there are no obvious repetitive DNA structures encoding the LRR units found in the RPP13 alleles, providing no opportunity for unequal intragenic recombination or DNA replication slippage. Generation of amino acid sequence diversity by the insertion of transposable elements is a second mechanism that has been observed in a number of R genes ( Luck et al. 1998 ; McDowell et al. 1998 ; Wang et al. 1998 ). We also found evidence of short, direct repeats in the RPP13 alleles that suggested target site duplication and variable excision of a transposable element (data not shown). However, none of these events occurred within the LRR domain. Probably of greater significance in generating amino acid diversity at the RPP13 locus is the role of non-synonymous point mutations and small nucleotide indels. Eight of the 11 indels identified occurred within the LRR domain, resulting in amino acid sequence variation ranging from one to five residues among the three RPP13 alleles ( Figs 3 and 4a). The accumulation of non-synonymous point mutations and the accompanying amino acid change would appear to be the primary source of diversity within the LRR domains of a number of other R genes (see for example Meyers et al. 1998 ; Parniske et al. 1997 ). Presumably, the existence of this variation is indicative of selection events at points in the evolutionary history of these genes. Our paired analysis of the LRR domains of the three RPP13 proteins revealed that the majority of the amino acid diversity was seen in the predicted β-strand/β-turn motif and the adjacent putative βα loop ( Table 3). Analysis of the ratio of non-synonymous to synonymous nucleotide substitutions for these two LRR subdomains indicated that this diversity was under positive selection. We interpret the results of this analysis to indicate that both the β-strand/β-turn motif and the adjacent putative βα loop of the LRR are actively contributing to specificity determination. The possible involvement of amino acid residues that lie outside the β-strand/β-turn motif in determining R gene specificity is a novel finding. McDowell et al. (1998) also reported variation at amino acid residues adjacent to the β-strand/β-turn motif in their analysis of RPP8. However, they incorporated these variant positions in the consensus they used to analyse the β-strand/β-turn motif, and therefore did not determine whether these positions were separately under positive selection for diversification. The overall structure of the LRR domain in RPP13 may not be as rigid as in other examples, particularly as it lacks the proline residue that is often part of other repeat sequences ( Jones & Jones 1997). Such lack of rigidity may allow a more ‘open’ structure and, consequently, extend the active domain capable of interacting with AVR proteins without the need to expand LRR copy number.
Peronospora parasitica isolates, Arabidopsis lines and pathogenicity tests
The origin of the P. parasitica isolates used here and the method used in the pathogenicity testing of Arabidopsis have been described elsewhere ( Bittner-Eddy et al. 1999 ; Holub et al. 1994 ). Arabidopsis accessions Columbia (Col-5), Neiderzens (Nd-1) and Rld-2 were as reported by Holub et al. (1994) . The F3 and F9 Col-5 × Nd-1 lines used to map the RPP13 locus have been described by Bittner-Eddy et al. (1999) , and the F9 Col-5 × Nd-1 inbred lines and related marker data have been deposited in the Nottingham Arabidopsis Stock Centre (UK).
Nd-1 DNA cosmid libraries
The libraries were constructed in the binary cosmid vector pSLJ75515 (www.uea.ac.uk/nrp/jic/s3d_plas.htm) from either partial BamHI or EcoRI digested Nd-1 DNA (size range 20–30 kb). Cosmid clones were packaged using Gigapack Gold III XL extract (Stratagene, La Jolla, CA, USA) and host Escherichia coli DH12S cells (Life Technologies, Rockland, MD, USA) transfected following standard methods ( Ausubel et al. 1994 ). Bacterial clones were sown directly onto selective LA plates (tetracycline 12.5 μg ml−1, IPTG 25 μg ml−1 and X-gal 50 μg ml−1) overlaid with Biotrans + nylon membrane (ICN, Irvine, CA, USA). Approximately 25 000 recombinant clones were screened using standard colony hybridization techniques ( Ausubel et al. 1994 ) to radiolabelled p3037-3, p21012-7 and p21012-8 DNA. Cosmid DNA was isolated using standard alkaline lysis procedures.
BAC clone identification, subcloning, and development of molecular markers
Filters containing clones from a bacterial artificial chromosome (BAC) library of Col-0 DNA (IGF BAC library, Mozo et al. 1998 ) were provided by RZPD (Berlin, Germany), and individual BAC clones identified by hybridization to radiolabelled p3037-3 DNA. DNA was isolated from BAC clones F21O12, F7B1, F1P5 and F6L9 using the recommended alkaline lysis method, and their integrity confirmed by Southern blot analysis with probes from the RPP13 interval. F21O12 DNA was digested with EcoRI and the fragments shotgun cloned into plasmid vector pBluescript SK + (Stratagene, La Jolla, CA, USA) using standard techniques ( Ausubel et al. 1994 ). Sequences for two subclones, p21012-7 and p21012-8, were obtained using T3 and T7 primers and a cycle sequencing kit (Perkin-Elmer, Foster City, CA, USA). Products were analysed on an ABI PRISM 377 sequencer (Perkin-Elmer) and primer pairs for PCR amplification designed from the DNA sequence using PrimerSelect (DNASTAR, Madison, WI, USA). Nucleotide variation revealed by comparing the sequence obtained from Col-5 and Nd-1 PCR products generated by p21012-7 or p21012-8 primers enabled us to map these loci in our mapping populations. The primer pairs are: 5′-CTG CCA ATG AGA AAA TGA AAG GTC-3′ and 5′-GGA GGC GTA GTT GTA GGA ATG AAT C-3′ (p21012-7), and 5′-TTT GCC ATC ATC TCG GAC TTT CTT T-3′ and 5′-GGT TTT TGT GTT CTC GAT TTT GTC A-3′ (p21012-8). For PCR amplification, 25 ng of plant DNA was used in a 25 μl reaction containing 50 m m KCl, 10 m m Tris–HCl pH 8.3, 2 m m MgCl2, 200 μM each dNTP, 0.4 μM each primer, and 1 unit Taq polymerase. A Perkin-Elmer 2400 thermocycler was used, and the amplification conditions were 1 min at 94°C, followed by 30 cycles of 30 sec at 94°C, 30 sec at 55°C, 2 min at 72°C, then a final extension at 72°C for 5 min. PCR products were purified by spun-column chromatography using Sepharose CL-6B (Amersham Pharmacia Biotech, Little Chalfont, UK), checked for quality and quantity by agarose gel electrophoresis, then sequenced directly (approximately 200 ng per reaction).
DNA sequencing and analysis of RPP13-Nd, RPP13-Rld and rpp13-Col
The sequences of RPP13-Nd, RPP13-Rld and rpp13-Col were obtained primarily through a primer walking strategy. Nd-1 DNA clones that lay within or adjacent to RPP13-Nd, including pBa3059, pBa3058-4, p3037-2 (see Fig. 1) and Y3033 ( Bittner-Eddy et al. 1999 ), were partially sequenced to generate multiple DNA sequence islands from which primers were designed. These islands were eventually linked by direct cycle sequencing of PCR products generated by RPP13-specific primer pairs from pBa3058 template. DNA from several amplification reactions was pooled to circumvent PCR-generated artefacts. RPP13-specific primer pairs were used to generate PCR products from either BAC F21O12 or Rld-2 DNA for sequencing of rpp13-Col and RPP13-Rld, respectively. In some areas additional primers were needed to complete the sequencing. Information on the primers used to sequence the RPP13 alleles is available on request. An Nd-1 cDNA library ( Speulman et al. 1998 ) was also used to provide sequence of the RPP13-Nd gene. The library was constructed in the lambda vector gt22A (Life Technologies, Rockland, MD, USA). A vector-specific primer, gt22A-5′ (5′-TTC GTC GAC CCA CGC GTC CG-3′) was designed to enable sequence from the 5′ end of cDNA clones to be obtained. The extreme 5′ untranslated region of RPP13-Nd was sequenced using gene-specific primer R13Nd15 (5′-CGA CGT CAT AAG CGA AAT CTA AAA-3′), which anneals 183 bases downstream of the predicted ATG translation initiation codon. The product(s) sequenced resulted from PCR amplification using gt22A-5′ and R13Nd24 (5′-ATC TGC CAA ACC AAC TAC AAC C-3′), followed by nested PCR amplification using gt22A-5′ and R13Nd25 (5′-ACG CCG AAC GCT GTC TCA CTC TCA-3′), which anneals 437 bases downstream of the predicted ATG translation initiation codon. PCR amplification using the RPP13-specific primers was employed to generate DNA for direct cycle sequencing of the remainder of the cDNA.
Sequence contigs were assembled using autoassembler 2.0 (Perkin-Elmer, Foster City, CA, USA). Conceptual DNA translations, DNA and amino acid sequence alignments, and primer design were performed using the mapdraw, megalign, and primerselect programs, respectively (DNASTAR, Madison, WI,. USA). Sequence similarity searches of the National Center for Biotechnology Information (Bethesda, MD, USA) nucleotide and protein sequence databases were performed using blast programs ( Altschul et al. 1997 ). Searches of DNA sequences for likely open reading frames were made using the netgene2 ( Hebsgaard et al. 1996 ) and netstart ( Pedersen & Nielsen 1997) programs available at www.cbs.dtu.dk/services. Non-synonymous and synonymous nucleotide substitutions were calculated using the diverge program (Wisconsin Package Version 9.1, Genetics Computer Group, Madison, WI, USA) following the method of Parniske et al. (1997) . GenBank accession numbers for RPP13-Nd, RPP13-Rld and rpp13-Col are AF209732, AF209731and AF209730, respectively.
Generation of RPP13-Rld plant transformation clones
A PCR-based strategy was used to clone the RPP13-Rld allele from the Rld-2 accession. Two primers, RPP13-F (5′-CAG TCA TTA CCG GCG GAG CAA AAA GCA TTC GTG TTA-3′) and RPP13-R (5′-GTT TGG ATC GGT GGC TCT ATC TTG GCT TCC CTC AGT A-3′), were designed from Nd-1 DNA sequence to anneal 3.5 kb upstream of the initiation codon and 950 bp downstream of the stop codon, respectively. The primers were used in conjunction with the Expand Long Template PCR System (Boehringer Mannheim, Mannheim, Germany) which contains a mix of thermostable Taq and Pwo DNA polymerases to enhance the generation of large PCR products of high fidelity. Each PCR contained 500 ng of Rld-2 DNA, 5 μl buffer 2 (2.25 m m MgCl2), 350 μM each dNTP, 0.3 μM each primer, and 2.5 units of polymerase mix in a final volume of 50 μl. A Perkin-Elmer 2400 thermocycler was used, and the amplification conditions were 2 min at 94°C, followed by 30 cycles of 10 sec at 94°C, 30 sec at 65°C, 8.5 min at 68°C, and then a final extension for 7 min at 68°C. These PCR amplification conditions reliably resulted in a single product of approximately 7.3 kb. The PCR product was purified by spun-column chromatography, polished with Klenow enzyme to produce blunt ends, and cloned into the Ecl136II (Fermentas AB, Vilnius, Lithuania) site of pSLJ75515. Ten clones, from two independent PCR amplifications, were partially or fully sequenced over the RPP13-Rld gene, including at least 1 kb upstream of the translation initiation codon, in order to identify a candidate RPP13-Rld clone (pBaRld2-WT).
Agrobacterium-mediated transformation of Arabidopsis
Binary cosmid clones were transferred into Agrobacterium tumefaciens strain GV3101 and cultured under tetracycline (12.5 μg ml−1) and gentamycin (25 μg ml−1) selection. The whole-plant vacuum infiltration method ( Betchtold et al. 1993 ) was used in all Arabidopsis transformation experiments. For identification of the RPP13-Nd gene, the Col-5 accession or recombinant inbred line (RIL) HRI3860 was used as the transgene recipient. For the cloning of RPP13-Rld, RIL HRI3860 only was used, as Col-5 is incompatible with the P. parasitica isolate Wela3 that defines RPP13-Rld. Transformants were selected by spraying seedlings 1–2 weeks after germination with dl-phospinothricin at 100 μg ml−1.
We gratefully acknowledge the RZPD (Berlin, Germany) for providing us with the Col-0 IGF BAC filter and clones. This work was supported by a grant from the UK Biotechnology and Biological Sciences Research Council.