These authors made an equal contribution to the work reported.
Resistance responses that plants deploy in defence against pathogens are often triggered following a recognition event mediated by resistance (R) genes. The encoded R proteins usually contain a nucleotide-binding site (NB) and a leucine-rich repeat (LRR) domain. They are further classified into those that contain an N-terminal coiled coil (CC) motif or a Toll interleukin receptor (TIR) domain. Such R genes, when transferred into a susceptible plant of the same or closely related species, usually impart full resistance capability. We have used map-based cloning and mutation analysis to study the recognition of Peronospora parasitica (RPP)2 (At) locus in Arabidopsis accession Columbia (Col-0), which is a determinant of specific recognition of P. parasitica (At) isolate Cala2. Genetic mapping located RPP2 to a 200-kb interval on chromosome 4, which contained four adjacent TIR:NB:LRR genes. Mutational analysis revealed three classes of genes involved in specifying resistance to Cala2. One class, which resulted in pleiotropic effects on resistance to other P. parasitica (At) isolates, was unlinked to the RPP2 locus; this class included AtSGT1b. The other two classes were mapped within the interval and were specific to Cala2 resistance. Representatives of each of these classes were sequenced, and mutations were found in one or the other of two (RPP2A and RPP2B) of the four TIR:NB:LRR genes. RPP2A and RPP2B complemented their specific mutations, but failed to impart resistance when present alone, and it is concluded that both genes are essential determinants for isolate-specific recognition of Cala2. RPP2A has an unusual structure with a short LRR domain at the C-terminus, preceded by two potential but incomplete TIR:NB domains. In addition, the RPP2A LRR domain lacks conserved motifs found in all but three other TIR:NB:LRR class proteins. In contrast, RPP2B has a complete TIR:NB:LRR structure. It is concluded that RPP2A and RPP2B cooperate to specify Cala2 resistance by providing recognition or signalling functions lacked by either partner protein.
Genetic variation between plants of different genotype, following challenge by biotrophic pathogens, is often expressed as either compatibility or incompatibility. The former circumstance is characterised by pathogen development and reproduction, with little visible cellular response by the host. In the latter circumstance, the pathogen is recognised by the plant, and processes are triggered that impede its further development. This incompatibility is often associated with restricted rapid host cell death. Variation in host–parasite interactions can be attributed to the presence or absence of specific resistance (R) genes, whose products enable plants to respond to pathogens delivering matching, specific avirulence (avr) gene products (Flor, 1971). The so-called ‘gene-for-gene’ hypothesis predicts that compatibility and, therefore, disease occurs if either of the matching gene pair products is absent from the interaction.
R genes have been cloned from a range of plants, and a consistency of structure has been revealed (Ellis et al., 2000). The majority of these genes encode proteins that contain a nucleotide-binding site (NB) domain, which occurs in diverse proteins with ATP/GTP-binding activity (Bent, 1996), and a leucine-rich repeat (LRR) domain predicted to mediate protein–protein interactions (Kobe and Deisenhofer, 1994). The NB domain is often associated with conserved ARC motifs, which show similarity to the nematode CED-4 and the mammalian APAF-1 proteins that regulate cell death (Van der Biezen and Jones, 1998). R genes can be further divided into two subclasses: one includes genes whose proteins contain a coiled coil (CC) motif at their N-terminus (e.g. recognition of Peronospora parasitica (RPP)8, McDowell et al., 1998; RPP13, Bittner-Eddy et al., 2000), the other includes genes whose proteins contain an N-terminal domain (Toll interleukin receptor (TIR) domain), with similarity to the Toll domain in Drosophila and the IL-1R domain in mammals (RPP5, Parker et al., 1997; RPP1, Botella et al., 1998). The products of these genes are predicted to be located in the cytoplasm. Other R genes with extracellularly located LRR domains are membrane-anchored, and either lack an obvious protein domain involved in signalling (e.g. Cf9, Jones et al., 1994) or contain a cytoplasmically located serine–threonine kinase domain (e.g. Xa21, Wang et al., 1998). The gene Pto from tomato represents an exception. This a cytoplasmic serine–threonine kinase (Martin et al., 1993) which, however, requires the linked gene Prf, an NB:LRR gene, to enable an incompatible response to isolates of Pseudomonas syringae delivering the AvrPto avr gene product (Salmeron et al., 1996).
In some cases, it has been demonstrated that a direct physical effect of the Avr protein on the R protein enables the incompatible response. This is the case for the Rx gene from potato that confers resistance to potato virus X (PVX). The viral coat protein causes the relaxation of intramolecular interactions within Rx, which, in turn, triggers resistance (Moffet et al., 2002). R proteins may also be components of complexes, for example, RPM1 and RPS2 from Arabidopsis are associated with the RIN4 protein. AvrRpt2 cleaves RIN4 and triggers resistance through the release of RPS2 from the complex (Axtell and Staskawicz, 2003; Mackey et al., 2003), whereas the presence of RIN4 is required for RPM1 resistance, perhaps through an interaction between AvrRpm1 and RIN4, resulting in the release of RPM1 (Mackey et al., 2002).
Holub et al. (1994) examined the inheritance of resistance to Peronospora parasitica (At) isolate Cala2 in Arabidopsis accession Columbia (Col-0). From a thorough characterisation of the interaction phenotype, including host cell response and quantity of pathogen sporulation, Holub et al. (1994) concluded that the expression of resistance to Cala2 in Col-0 was allele dosage-dependent, being under the control of a partially dominant R allele (RPP2) at a single locus. Using 100 F3 families from a Col-0 × Neiderzens (Nd-1) cross, Tör et al. (1994) mapped RPP2 to a single locus on chromosome 4, 4.8 cM below the morphological marker Agamous and 9.1 cM above the RFLP locus B9 (Figure 1a).
Here, we report that resistance to P. parasitica (At) isolate Cala2, specified by the RPP2 locus in Col-0, requires the presence of two resistance determinants, both of which reside at the RPP2 locus and are members of the TIR:NB:LRR class of R gene.
Mutational analysis reveals that resistance to Cala2 requires the presence of two different R alleles at the RPP2 locus
Two populations of mutagenised Col-0 plants were screened to detect variation for interaction phenotype following inoculation with Cala2: one population had been treated with fast neutrons (FNs) and the other with ethylmethane sulfonate (EMS; see Experimental procedures). M2 seedlings were scored 7 days after inoculation for presence or absence of asexual sporulation. Plants supporting sporulation were grown to maturity, and seeds were collected. M3 plants from the selected mutants were re-tested with Cala2 to confirm that the plants originally selected represent true phenotypic variants. Seven EMS and three FN Cala2-susceptible mutants were obtained.
To determine which of the mutations were specific to Cala2 recognition, each mutant line was inoculated with a panel of P. parasitica (At) isolates. Resistance to each of these isolates in Col-0 is conferred by R genes, which are distinct from RPP2. Hence, an alteration in the interaction phenotype observed, following inoculation with these isolates, would imply that the mutation was in a gene involved in the disease-resistance response, common to both RPP2, and one or more other R genes. Majority of the mutants identified, following inoculation with RPP2, exhibited a wild-type resistance phenotype to all the other isolates tested, while the remaining were susceptible to Cala2 (data not shown). Two pleiotropic mutants (FN1 and EMS1) were identified. Mutant FN1 showed a susceptible reaction following inoculation with all isolates tested, and it was concluded that this was probably a mutant of a gene involved in the signalling pathway required for the function of several different R genes. Subsequently, Tör et al. (2002) named the mutation in line FN1 enhanced downy mildew susceptibility (edm)1 and showed the gene (AtSGT1b) to be an orthologue of yeast SGT1. Mutant EMS1 exhibited a less than fully compatible phenotype following inoculation with isolates Cala2, Emwa1 and Cand5, characterised by a moderate level of sporulation.
All mutants were back-crossed to Col-0 to determine the number of genes segregating for the observed phenotypes following inoculation with Cala2. The recovery of the wild-type resistant phenotype in the F1 progeny was consistent with the recessive nature of the mutations, and it confirmed that the back-crosses were successful (Table 1). The small number of F1 seedlings, exhibiting a low level of sporulation in the crosses between Col-0 and some of the mutants (EMS1, EMS5, EMS6, EMS7, FN1, FN2 and FN3) could be attributed to the fact that a low proportion of seedlings of Col-0 may occasionally exhibit this phenotype following inoculation with Cala2. With the exception of EMS4, segregation of Cala2 resistance (3 : 1 resistant:susceptible) among F2 seedlings was consistent with mutations in single alleles at one locus (Table 1). In contrast, the F2 segregation data from the cross with EMS4 were characterised by the occurrence of a high proportion of seedlings exhibiting a low intensity of sporulation (L), and ratios between seedlings placed in resistant and susceptible categories were significantly different from expectation, if the mutation was in a single allele at one locus.
Table 1. Segregation of interaction phenotypes among F1 and F2 progeny from mutants crossed with Col-0, Nd-1 and the FN2 mutant following inoculation with Cala2
a Interaction phenotype according to Holub et al. (1994). The asexual sporulation of the isolate is classified according to the average number of sporangiophores per cotyledon as: light (L) = 1–10, medium (M) = 12–17 and high (H) = >17. The abbreviation N indicates no asexual sporulation.
for 1 degree of freedom P < 0.05 is 3.84.
c For the calculation of χ2. Resistant class, plants displaying the N interaction phenotype; susceptible class, plants displaying the L, M and H interaction phenotypes.
Low numbers of offspring showing no asexual sporulation are because of escape from infection.
Each mutant was also crossed with Nd-1 to determine if the mutations affecting resistance to Cala2 were at the RPP2 locus. Nd-1 is susceptible to Cala2 and therefore lacks alleles effective in the recognition of Cala2 at the RPP2 locus. Complementation of the Cala2 recognition phenotype in F1 progeny would indicate that the mutation did not lie within the RPP2 locus, but rather in another locus required for its function. Segregation data for reponses to inoculation with Cala2 are presented in Table 1. Two classes of mutants were distinguishable: Class 1 contains the mutants FN1 and EMS1, which, when crossed with Nd-1, resulted in F1 progeny that were more resistant to Cala2 than the parents, consistent with these mutations not taking place in RPP2; Class 2 contains the mutants EMS2, EMS3, EMS4, EMS5, EMS6, EMS7, FN2 and FN3, which, when crossed with Nd-1, resulted in fully susceptible F1 individuals. Lack of complementation indicated that these lines were likely to contain a mutation allelic to the susceptibility determinant in Nd-1, confirmed by the lack of segregation for resistance among F2 progeny (Table 1). Hence, Class 2 mutant lines contained mutations at the RPP2 locus.
To determine if the Class 2 mutants were allelic, they were all crossed with the Class 2 FN2 mutant. Complementation of the Cala2 recognition phenotype in the F1 progeny of a mutant × FN2 cross would indicate that the mutations were not allelic. Conversely, lack of complementation would demonstrate that the mutations were in the same allele. Intriguingly, only the F1 progeny resulting from FN2, crossed to EMS2, EMS3 and FN3, were susceptible and showed no segregation for resistance among F2 progeny (Table 1). This demonstrated that the mutations in this group (Class 2A) are allelic. F1 progeny from the crosses with the other Class 2 (Class 2B) mutants exhibited a higher level of resistance than either parent, which indicated that these lines carry a mutation allelic to a component of resistance missing from Nd-1, but not allelic to FN2. As expected, the F1 data for crosses of Class 1 mutants to FN2 resulted in resistant offspring, confirming that the mutations were in alleles at different loci (Table 1).
Hence, the mutational analyses revealed that at least two linked genes at the RPP2 locus in Col-0 are required for isolate-specific recognition of Cala2. The susceptible accession Nd-1 lacks alleles capable of recognising Cala2 for either of these genes at the RPP2 locus.
Fine-scale mapping of the RPP2 locus in Columbia
A set of 200 F9 recombinant inbred (RI) lines from the Col-0 × Nd-1 cross was used to develop a range of new dimorphic RFLP and co-dominant amplified polymorphic sequence (CAPS) markers in the mapping interval defined by Tör et al. (1994). These enabled RPP2 to be located within an interval defined between the CAPS marker 4G1L and the RFLP marker T46721. RPP2 co-segregated with the CAPS marker g3883 and RFLP marker 2G8R (Figure 1a). Markers 4G1L and 2G8R were derived from YACs YUP4G1 and CIC2G8, respectively. This anchored the mapping interval to the physical contig of chromosome 4 created by Schmidt et al. (1996). A walk was initiated from the flanking markers with lambda clones. It was revealed that λ4011 (Figure 1b) contained an ‘R-gene-like’ sequence, which encoded a protein with 64% identity to the RPP5 protein (Parker et al., 1997) within the TIR:NB domain. This gene was therefore a candidate for one of the components of Cala2 resistance.
A Southern blot, containing genomic DNA from all the selected mutants restricted with EcoR1, was probed with the candidate gene cloned from the RPP2 mapping interval (data not shown). This experiment revealed that there was a shift in the expected 4.7-kb EcoR1 fragment (Figure 1c) in the track containing DNA from the Class 2A FN2 mutant (Figure 2). Hybridisation with PCR probes, representing smaller parts of the candidate gene, suggested that the deletion or re-arrangement did not extend beyond the end of the gene (data not shown). The candidate gene from the Class 2A EMS mutant EMS2, that had previously been shown to be allelic to FN2, was sequenced. This revealed a single C-to-T substitution in position +457, resulting in an in-frame stop codon, and thereby a truncated protein.
Identification of RPP2A
An 8.8-kb SpeI–SalI fragment from λ4011 (Figure 1c), which carried only the candidate gene, was subcloned into a cosmid binary vector to generate c4118. This clone was used to transform two Cala2-susceptible RI lines from the Col-0 × Nd-1 cross and the Class 2A mutant FN2. T2 progeny from five independent FN2 transgenic lines (30 plants per line on average) were found to segregate 3 : 1 (resistant:susceptible) for response to Cala2. There was an absolute correlation between Cala2 resistance and BASTA tolerance (the marker used to select transformants). Figure 2 shows a Southern blot containing DNA from Cala2-resistant (FR) and -susceptible (FS) T2 progeny from an FN2 transgenic line. The 4.7-kb EcoRI fragment within the 8.8-kb SpeI–SalI fragment was only detected in Cala2-resistant transgenic plants. However, the equivalent RI line transformants produced only Cala2-susceptible progeny as did FN2 plants transformed with the binary vector alone. These data are consistent with the segregation data from crosses with the mutant lines, which indicated that Nd-1 lacks alleles capable of recognising Cala2 at both the two genes within the RPP2 locus. These results demonstrated that the gene cloned from the RPP2 mapping interval has a functional role in Cala2 recognition, but is only one of at least two components. This gene was called RPP2A.
Identification of the second component of Cala2 resistance at the RPP2 locus
The mapping interval containing the RPP2 locus spanned approximately 200 kb (Figure 1a). To identify the second component required for Cala2 resistance, we selected F2 progeny from a Col-0 × Nd-1 cross, which were susceptible to Cala2. These F2 progeny would lack effective alleles of either or both putative genes at the RPP2 locus. We screened 180 of these F2 progeny for those that contained RPP2A from Col-0, reasoning that any such offspring must contain a recombination event between RPP2A and a second component. No such individuals were obtained. Therefore, given the size of the population available, the additional gene or genes required for Cala2 resistance lay too close to RPP2A to be separated by a recombination event.
Concurrently, analysis of the genes within the mapping interval revealed three other NB:LRR genes (At4g19510, At4g19520 and At4g19530; Figure 1c,d), which formed a cluster with RPP2A (http://www.arabidopsis.org). As we were unable to separate RPP2A from the additional component required for Cala2 resistance, these genes were all likely candidates. Therefore, we sequenced these genes from the four Class 2B mutants. Mutations were only found in At4g19510, the gene immediately adjacent to RPP2A. These mutations all resulted in amino acid substitutions (see below).
At4g19510 was subcloned on a HpaI fragment from bacterial artificial chromosome (BAC) clone F24J7 into the binary vector, which was then used to transform mutant EMS5. T2 offspring from two independent transformants revealed segregation for response to Cala2 (3 : 1 resistant:susceptible); BASTA tolerance co-segregated with Cala2 resistance. Therefore, AT4g19510 is the second component required for Cala2 resistance at the RPP2 locus, and was called RPP2B.
The protein encoded by RPP2A
The 8.8-kb SpeI–SalI genomic c4118 insert containing RPP2A was sequenced. Annotation of this gene suggested that it contained four introns (http://www.arabidopsis.org). This was confirmed by sequencing RPP2A cDNA (see Experimental procedures). RPP2A is predicted to encode a cytoplasmic protein consisting of 1308 amino acids (Figure 3).
Comparison of RPP2A with proteins in the GenBank database revealed that it is most similar to RPP5 (Parker et al., 1997). The sequence similarities between RPP2A and RPP5 in the TIR and NB domains are shown in Figure 3. RPP2A shares 69% identity with RPP5 in the TIR domain, while more than 80% identity is evident between the two proteins in the conserved motifs appearing in the NB domain (the pre-P-loop, P-loop, kinase 2 and kinase 3a). The overall identity between them in the TIR:NB region (residues 1–320 for RPP2A and 1–326 for RPP5) is 64%. The NB domain of RPP2A, unlike that of RPP5, does not contain obvious ARC motifs, which are associated with most NB domains in R proteins (Meyers et al., 1999; Van der Biezen and Jones, 1998). RPP2A is also unique amongst disease-resistance proteins in that it contains a DUF640 motif. The function of this motif is unknown, but it has been found in a small number of other plant proteins in the PFAM database. Surprisingly, C-terminal to the DUF640 motif, a second TIR domain is found in RPP2A (Figure 3). Furthermore, this TIR domain is followed by an additional NB domain, which is associated with ARC motifs. However, this additional NB domain lacks a recognisable pre-P-loop and P-loop motif. Overall, this second TIR:NB region only shows 25% identity to RPP5.
Two additional surprising elements were revealed in the last third of the RPP2A protein. First, although the alignment with RPP5 seems to classify RPP2A within the TIR:NB:LRR subclass of disease-resistance genes, the predicted protein also contains a putative CC domain comprising four heptad motifs (Figure 3). Second, the LRR domain is located at the end of the protein and consists of only seven repeats.
The protein encoded by RPP2B
The RPP2B gene (At4g19510) is 5297 bp and lies adjacent to RPP2A on chromosome 4. Two different gene models, with fundamentally different C-termini, exist for At4g19510 as a result of different intron/exon predictions (http://www.arabidopsis.org; http://www.mips.biochem.mpg.de/). To determine which, if any, gene model is correct, full-length cDNA was obtained and its sequence was determined (see Experimental procedures). The cDNA sequence revealed that neither model was correct. For example, putative introns 4 and 6 (http://www.arabidopsis.org) were retained in all RPP2B cDNAs sequenced, thereby resulting in a predicted TIR:NB:LRR protein of 1207 amino acids (Figure 4). The structure of RPP2B differs markedly from RPP2A. There is no duplication of the TIR:NB region. In mutant EMS7, a glycine-to-glutamic acid change occurred in the pre-P-loop (Figure 4). The LRR domain contains 14 clearly defined repeats, of 21–24 amino acids, that conform to previous consensus sequences. Mutant EMS6 contains a phenylalanine residue in place of a conserved leucine within the β-turn β-loop of the sixth LRR (Figure 4). The LRR domain is separated from the C-terminus by 282 amino acids, which contains no obvious functional motifs. However, the amino acid changes in mutants EMS4 (glutamic acid to lysine) and EMS5 (threonine to methionine) occur within this region (Figure 4), immediately adjacent to the last of the 14 defined LRRs. blastp analyses, using the whole C-terminal domain, revealed that these mutations lay in a region conserved amongst a subset of Arabidopsis TIR:NB:LRR proteins. This region has been annotated as either motif 8 or motif 25 by Meyers et al. (2003).
Conserved regions within the LRR domains of RPP2A and RPP2B
As proteins encoded by RPP2A and RPP2B cooperate to specify Cala2 resistance, we analysed their structure for common protein motifs. We could not detect significant homology between RPP2A and RPP2B outside conserved motifs (e.g. P-loop) within functional domains, suggesting that these genes did not evolve from one another by a simple duplication event. However, two conserved features were apparent within the LRR domains. The amino acid immediately following the β-strand/β-turn motif is conserved between the two proteins and is S, S, C, C, C, T, C in the first seven LRRs in each case (Figures 3 and 4). Furthermore, within the third LRR, the sequence DLEGCTSL is completely conserved. However, no other significant homologies are apparent within the LRR domain.
Alleles at the RPP2 locus are responsible for specifying recognition of P. parasitica (At) isolate Cala2 in Arabidopsis accession Col-0. Fine-scale mapping and mutational analysis were used to reveal that two TIR:NB:LRR genes, RPP2A and RPP2B, were required for effective resistance.
Although RPP2A encodes a protein that belongs to the TIR:NB:LRR class, it is of an unusual structure. It contains two TIR:NB regions, a potential CC domain and a DUF640 domain. The N-terminal TIR:NB region lacks associated ARC motifs. In contrast, the second TIR:NB region contains ARC motifs, but, surprisingly, lacks a pre-P- and a P-loop. A high level of identity (64%) exists between the proteins encoded by RPP2A and RPP5 in the N-terminal TIR:NB domain; however, this is not the case for the second TIR:NB region. Meyers et al. (2002) proposed that the TIR:NB:LRR class of R genes could have evolved from a fusion between LRR and TIR proteins (TX), or TIR:NB proteins (TN). It is therefore possible that RPP2A and RPP5 were formed by the fusion of a common TN gene with different LRR-containing proteins. For RPP5, this was a simple fusion, but in the case of RPP2A, the TN gene fused with a full-length TIR:NB:LRR gene. Alternatively, RPP2A may have been created by the fusion of the TIR:NB domain from RPP5 (or its progenitor) and another TIR:NB:LRR-like gene. If the ARC, pre-P-loop and P-loop motifs are all essential for RPP2A function, they can clearly tolerate being separated within the protein.
Another intriguing feature of the RPP2A protein is the small LRR domain which consists of only seven repeats. Although proteins with as few as two or three LRRs, e.g. the Trk receptor kinase from humans and the TrkB receptor kinase from mouse, have been reported (reviewed by Kobe and Deisenhofer, 1994), such a small number of LRRs is unexpected for a plant disease-resistance gene (Jones and Jones, 1997). For example, the proteins encoded by RPP5, N, L6 and RPS4, all of which belong to the TIR:NB:LRR subclass of R genes, are predicted to contain 21, 16, 27 and 15 LRRs, respectively (Gassman et al., 1999; Lawrence et al., 1995; Parker et al., 1997; Whitham et al., 1994). Only the genes in the RPP1 cluster encode proteins containing as few as 10 LRRs (Botella et al., 1998).
The presence of a potential CC domain implies the possible formation of homo- or heterodimers. The role of the DUF640 domain is currently unknown. It has been detected simply as a conserved domain amongst certain plant genes in the PFAM database. RPP2A is the first DUF640-containing protein to which a function has been assigned. Other predicted proteins containing this motif are small peptides. In future, it will be interesting to learn if there are any common functional themes amongst these proteins.
RPP2B encodes an TIR:NB:LRR gene with a more classical structure. The N-terminal TIR domain and the following NB-ARC domain contain all the appropriate conserved sequences (Figure 4). The LRR domain consists of fourteen 21–24-amino acid repeats, and is followed by a C-terminal domain of 282 amino acids. Mutations used to identify RPP2B lie in the NB-ARC region, the LRR region and a region of the C-terminal domain conserved amongst many TIR:NB:LRR proteins (motif 25/motif 8; Meyers et al., 2003). In RPP2B, this motif is positioned 10 amino acids from the last predicted LRR, and it itself shows sequence conservation reminiscent of an LRR. The most highly conserved component of the motif is (F/L)XFTNCF(K/N)L, which is similar to the LRR consensus of LXLXXCXXL. This suggests that this motif forms part of a degenerate LRR, but still plays a role in the function of RPP2B.
RPP2A and RPP2B are part of a R-gene cluster that contains two other TIR:NB:LRR genes. At4g19520 codes for a TIR:NB:LRR protein containing an additional TIR domain at the C-terminus. At4g19530 is a standard TIR:NB:LRR gene (Figure 1d). However, the lack of DNA homology between all four genes suggests that it is unlikely they arose as gene duplication events from a single progenitor. The RPP2 locus may be analogous to the Pto-resistance complex in tomato. Although the Pto gene from tomato complements resistance to the bacterium P. syringae pv. tomato in susceptible plants (Martin et al., 1993), it has been shown that another gene, Prf, lying within the Pto cluster, is absolutely essential for effective resistance (Salmeron et al., 1996). Pto encodes a serine–threonine protein kinase, while Prf is an NB:LRR protein (Salmeron et al., 1996). It is thought that Pto and Prf act together to respond to AvrPto, which interacts directly with Pto (Scofield et al., 1996; Tang et al., 1996). It is therefore possible that products of all four genes at the RPP2 locus interact as a resistance complex in a variety of different combinations. Alternatively, these proteins may interact to perform some function in plant growth and development that is targeted by the Cala2 avr gene product (ATR2). It is intriguing that a group of diverse but similar genes have evolved or been assembled at this locus.
Comparison of the LRR domains of TIR:NB:LRR genes by Meyers et al. (2003) revealed that there are highly conserved motifs (5, 14 and 15) at the beginning of the LRR domain. In common with three other TIR:NB:LRR genes in the Col-0 genome, RPP2A is exceptional in lacking all three of these motifs. Not only are these motifs missing, but the TIR:NB:LRR region does not conform to the standard structure. At4g36140 has the same modular structure as RPP2A with duplicated TIR:NB regions, At4g09360 lacks a TIR domain and At5g17970, classified as a possible pseudogene, lacks several conserved NB motifs (Meyers et al., 2003). Therefore, it is possible that these proteins may lack key motifs that enable them to function independently as R genes but are nevertheless required to effect specific recognition of particular pathogens and their variants. We have shown that RPP2A and RPP2B cooperate to effect resistance to Cala2, even though the latter is a TIR:NB:LRR protein that appears to contain all necessary conserved motifs to function independently. The implication is therefore that all three of these predicted genes may also function via a TIR:NB:LRR partner protein.
Inspection of the proteins encoded by the four genes at the RPP2 cluster revealed a conserved structure (DLEGCTSL) within the third LRR. Interestingly, although this sequence is conserved absolutely between RPP2A and RPP2B, it is less well conserved within At4g19520 and At4g19530. Analysis of other TIR:NB:LRR genes reveal that there is sequence conservation in this LRR, but not to the degree observed between RPP2A and RPP2B. This absolute conservation is startling in the light of the lack of overall homology between RPP2A and RPP2B. This implies a significant functional role for this LRR in the cooperative action of these two proteins. This function could allow these proteins to interact directly or be related to a shared recognition capability of either a domain of the ATR2 protein or some other component plant protein.
Warren et al. (1998) identified a mutation (RPS5.1) in the third LRR of RPS5, a CC:NB:LRR class R gene that recognises P. syringae pv. tomato carrying the avrPphB gene. This mutation suppressed resistance specified by a range of bacterial and downy mildew R genes, including Cala2 resistance mediated by RPP2. A small but statistically significant increase in susceptibility to Cala2 was observed in the RPS5.1 mutant. Warren et al. (1998) suggested that their observation could reflect an interaction with a common component of the signal transduction pathway, leading to the expression of resistance. An alternative explanation would be that the RPS5 protein is part of the same complex as the RPP2A/RPP2B TIR:NB:LRR proteins. Bittner-Eddy et al. (2000) described the RPP13 gene, a CC:NB:LRR protein, that shows high sequence conservation to the third LRR of RPS5. However, resistance effected by RPP13 was unaffected by the RPS5.1 mutation (Bittner-Eddy and Beynon, 2001). Comparison of the conserved third LRR from RPP2A and RPP2B with that from RPS5 revealed no sequence homology. Hence, the pleiotropic effect of the RPPS5.1 mutation neither relies on nor is associated with conserved sequences within the LRR domain. However, it is nevertheless remarkable that it is the third LRR that shows an absolute conservation of a subsequence between RPP2A and RPP2B, which does suggest functional significance for the third LRR, independent of the detection of the pathogen avr gene product.
RPP2B appears to have all the necessary components of a R gene, but to recognise Cala2, we have shown that it requires RPP2A. This implies some form of interaction between the gene products or the juxtaposition of these proteins as components of a complex. Two basic models exist for the interaction between the products of R and avr genes. The gene products could interact directly, which has been shown to occur in the case of Pto and AvrPto and between Pi-ta and AVR-Pita (Jia et al., 2000; Scofield et al., 1996; Tang et al., 1996). This model is shown in Figure 5(a,b), where direct interaction of the avr protein either recruits the R genes or de-stabilises their interaction. Moffet et al. (2002) demonstrated that the active domains of an R protein can be suppressed by intramolecular interactions, and the model in Figure 5(b) similarly shows suppression of the activation of resistance responses via intermolecular interactions between RPP2A and RPP2B.
It has, however, not proved possible to demonstrate direct interactions between R proteins and their partner AVR proteins in several pathosystems. This led Van der Biezen and Jones (1998) to propose that the R-gene protein could monitor (guard) another plant protein, and it is the interaction between that second protein and the avr gene product that results in the R protein, triggering a plant response to the pathogen. In this context, therefore, it is possible that RPP2A ‘guards’ RPP2B or vice versa, a model also consistent with Figure 5(b). When the Cala2 avr protein, ATR2, interacts with either the RPP2A or RPP2B target, the hypothesis is that the alternative gene product is released from the complex to initiate a disease-resistance response. However, it may also be that RPP2A and RPP2B both further guard a plant protein, which is itself the target of ATR2 (Figure 5c). In this model, interaction of ATR2 with the proposed third protein results in activation of the resistance response via RPP2A or RPP2B, or both. Finally, it may be that the detection of ATR2 requires the juxtaposition of two LRR domains, a model consistent with Figure 5(a).
The RPP2 locus is an example of two TIR:NB:LRR genes apparently acting cooperatively to determine the recognition specificity to a single pathogen avirulence determinant. The discovery reported here implies that degenerate forms of TIR:NB:LRR genes may be maintained through evolution, specifically to generate the capability for novel recognition of pathogen gene products.
Arabidopsis cultivation and pathogenicity tests
The Arabidopsis accessions Col-0 and Nd-1, and the F9 Col-0 × Nd-1 inbred lines (RIL), used to map RPP2, were as reported by Holub et al. (1994) and Bittner-Eddy et al. (1999). The conditions of plant cultivation, maintenance of P. parasitica (At) isolate Cala2 and pathogenicity tests were carried out as described by Holub et al. (1994).
Fine-scale mapping of the RPP2 locus and development of a lambda clone contig spanning RPP2A
Molecular markers Agamous and B9, and the phenotypic marker RPP4 have been described elsewhere (Tör et al., 1994). B9 was converted into a CAPS marker (primers: 5′-CATCTGCAACATCTTCCCCAG-3′ and 5′-CGTATCCGCATTTCTTCACTGC-3′; restriction enzyme AciI) to facilitate the mapping of RPP2 within the RIL population. Cosmid clone g3883 (Nam et al., 1989) was provided by R. Schmidt and C. Dean (John Innes Institute, Norwich, UK) and converted into a CAPS marker (primers: 5′-TGTTTCAGAGTAGCCAATTC-3′ and 5′-CATCCATCAAACAAACTCC-3′; restriction enzyme PstI). EST clone T46721 (Newman et al., 1994) was developed as a RFLP marker, detecting a BglII restriction site polymorphism between Col-0 and Nd-1. This EST clone corresponds to At4g19920, which encodes a TIR:NB:LRR protein.
YAC clones were obtained from the Arabidopsis Biological Resource Center (ABRC) at Ohio State University, and the BAC clone F24J7 was obtained from RZPD (Berlin, Germany; Mozo et al., 1998). The vectorette PCR amplification method, as described by Matallana et al. (1992), was used to generate end probes from YAC clones CIC4G1, CIC2G8 and yUP5C3. The end probes were either used as RFLP markers or, in the case of 4G1L, converted into a CAPS marker (primers: 5′-GTAACACTATGGCTGTGGTAGAG-3′ and 5′-ACGAAATGTATTTCAATGTAATGT-3′; restriction enzyme ApoI).
Plant DNA was prepared from the RIL population using the CTAB extraction method (Ausubel et al., 1994). Typically, 2–3 µg of plant DNA was digested with the appropriate restriction enzyme for RFLP analysis. For CAPS markers, 25 ng of plant DNA was used in a 25 µl PCR containing 50 mm KCl, 10 mm Tris–HCl (pH 8.3), 2 mm MgCl2, 200 mm each dNTP, 0.4 µm each primer and 1 unit Taq polymerase. A Perkin-Elmer 9700 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, 1 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), digested with the appropriate restriction enzyme and typically resolved on 2% agarose gels.
A lambda library, containing Col-0 genomic DNA, was used to create a contig of clones spanning the RPP2A gene. This library was created in the lambda vector GEM11 from size-fractionated digested Col-0 DNA (John Mulligan and Ronald Davis, Standford University, USA). Lambda clones were identified, and DNA was isolated using standard molecular techniques (Ausubel et al., 1994). Restriction enzyme fingerprinting and Southern blot analysis, using radiolabelled DNA probes made from markers within the interval, was used to confirm and order the contig shown in Figure 1(b).
Two Col-0 mutant populations were purchased from Lehle Seeds (Round Rock, TX, USA) and screened for susceptibility to Cala2. Approximately 14 000 M2 generation seedlings were screened from both mutant populations. Mutants identified from the FN-treated population were given the prefix FN; mutants identified from the EMS-treated population were given the prefix EMS. Mutants were confirmed by testing progeny with Cala2. Mutants were also tested for defects in shared resistance mechanisms using seven other isolates avirulent on wild-type Col-0. Crosses using the mutants were performed as described by Bittner-Eddy et al. (1999). Three different crosses were performed to allow allelism tests and segregation analyses. Mutants were back-crossed with Col-0 using the wild-type plant as the pollen donor. The recovery of wild-type resistance in the resultant F1 progeny was consistent with the recessive nature of the mutants. F2 segregation analysis was used to determine the number of genes segregating for the observed mutant phenotypes. Mutants were also crossed with Nd-1 to determine those lines containing a mutation in a gene allelic to the susceptibility determinant present in Nd-1 (lack of RPP2A and RPP2B in this case). Selected lines from the three mutant classes (Classes 1, 2A and 2B) were also crossed within and between classes to determine those mutations that were allelic.
Sequencing RPP2A and RPP2B cDNA and mutant alleles
Plant genomic DNA was prepared using the CTAB extraction method (Ausubel et al., 1994), and was used as template for sequencing selected RPP2A and RPP2B mutants. Primers for PCR amplification and sequencing were designed from published DNA sequence of BAC clone F24J7 (http://www.mips.biochem.mpg.de/) using PrimerSelect (DNASTAR, Madison, WI, USA). Primer sequences are available upon request. PCR products were purified by spun-column chromatography using Sepharose CL-6B (Amersham Pharmacia Biotech), checked for quality and quantity by agarose gel electrophoresis and then sequenced directly (approximately 200 ng per reaction) using big dye-terminator chemistries and an ABI PRISM 377 sequencer (Applied Biosystems). Sequence contigs were assembled using AutoAssembler 2.0 (Applied Biosystems). Conceptual DNA translations and DNA/protein alignments were performed using the mapdraw and megalign programs, respectively (DNASTAR).
Total RNA was isolated from 4-week-old Col-0 seedlings using an RNAeasy Mini kit (Qiagen, West Sussex, UK). First-strand cDNA (sscDNA) was produced from total RNA using the SMART PCR cDNA synthesis kit (CLONTECH, Hampshire, UK) according to the manufacturer's protocol. RPP2A and RPP2B cDNA was generated from sscDNA template using gene-specific PCR primer pairs. The 5′- and 3′ untranslated regions (UTRs) of RPP2A and RPP2B were defined from partial cDNA sequence deposited in the database (http://www.arabidopsis.org). Three or four PCR primer pairs, respectively, were used to generate overlapping RPP2A or RPP2B cDNA fragments for sequencing. At least one of the primers in each pair was designed to span a predicted intron/exon boundary (http://www.arabidopsis.org), thereby ensuring that any residual contaminating genomic DNA would not serve as a PCR template. cDNA fragments were sequenced using specific primers, and the resulting sequence was analysed, assembled, and conceptual DNA translations and DNA/protein alignments were performed as described above. Sequence of the PCR and sequencing primers are available upon request.
Sequence similarity searches of nucleotide and protein sequence databases at the National Center for Biotechnology Information (Bethesda, MD, USA) were performed using blast programs (Altschul et al., 1997). The Conserved Domain Architecture Retrieval Tool (CDART; http://www.ncbi.nlm.nih.gov/BLAST/) was used to find conserved amino acid motifs within RPP2A, RPP2B, and the proteins encoded by the linked genes Atg419520 and Atg419530.
Agrobacterium-mediated transformation of Arabidopsis
Recognition of Peronospora parasitica2A and RPP2B were cloned into the binary cosmid vector pSLJ75515 (http://www.uea.ac.uk/nrp/jic/s3d_plas.htm) and mated into Agrobacterium tumefaciens strain GV3101. The clones were 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. Transformants were selected by spraying seedlings 1–2 weeks after germination with l-phosphinothricin at 100 µg ml−1.
We would like to thank RZPD, Germany for the Arabidopsis BAC library, Jeff Dangl (EU BRIDGE Programme) for providing the John Mulligan and Ronald Davis lambda library, Renate Schmidt and Caroline Dean for YAC clones and Eric Holub for pathogen isolates and helpful discussion. Motif data for the TIR:NB:LRR gene class were obtained from the NIBLRRS Project website at http://niblrrs.ucdavis.edu/ that was supported by the NSF Plant Genome Program Award #9975971. The work reported here was supported by a grant from the Biotechnology and Biological Sciences Research Council and an EU Marie Curie fellowship to Eva Sinapidou.