Arabidopsis RPP4 is a member of the RPP5 multigene family of TIR-NB-LRR genes and confers downy mildew resistance through multiple signalling components


  • Present address: Aventis CropScience NV, Joseph Plateaustraat 22, 9000 Gent, Belgium. Present address: Institute of Cancer Biology, Strandboulevarden 46, DK-2100 Copenhagen, Denmark. Present address: Max-Planck-Institute for Plant Breeding Research, Carl-von-Linne-Weg 10, D-50829 Cologne, Germany.

* For correspondence (fax +44 1603 450011; e-mail


In Arabidopsis, RPP4 confers resistance to Peronospora parasitica (P.p.) races Emoy2 and Emwa1 (downy mildew). We identified RPP4 in Col-0 as a member of the clustered RPP5 multigene family encoding nucleotide-binding leucine-rich repeat proteins with Toll/interleukin-1 receptor domains. RPP4 is the orthologue of RPP5 which, in addition to recognizing P.p. race Noco2, also mediates resistance to Emoy2 and Emwa1. Most differences between RPP4 and RPP5 occur in residues that constitute the TIR domain and in LRR residues that are predicted to confer recognition specificity. RPP4 requires the action of at least 12 defence components, including DTH9, EDS1, PAD4, PAL, PBS2, PBS3, SID1, SID2 and salicylic acid. The ndr1, npr1 and rps5-1 mutations partially compromise RPP4 function in cotyledons but not in true leaves. The identification of RPP4 as a TIR-NB-LRR protein, coupled with its dependence on certain signalling components in true leaves, is consistent with the hypothesis that distinct NB-LRR protein classes differentially signal through EDS1 and NDR1. Our results suggest that RPP4-mediated resistance is developmentally regulated and that in cotyledons there is cross-talk between EDS1 and NDR1 signalling and processes regulating systemic acquired resistance.


Plant resistance to biotrophic pathogens is conferred by highly specific resistance (R) genes that elicit effective defence responses on perception of pathogen Avr genes (Feys and Parker, 2000). All characterized Arabidopsis thaliana R genes encode nucleotide-binding (NB) sites and leucine-rich repeat (LRR) domains (Parker et al., 2000). Approximately 130 genes encoding NB-LRR proteins are present in the Arabidopsis genome, and these can be grouped in two main classes based on their N-terminal domains (Meyers et al., 1999). The CC-NB-LRR is a broad class of NB-LRR proteins that contains a putative heptad leucine-zipper or coiled-coil (CC) motif. The TIR-NB-LRR class shows similarity to the cytoplasmic effector portion of the Drosphila Toll and human interleukin-1 transmembrane receptors (TIR). Similarities between NB-LRR proteins and activators of apoptosis in animal cells reinforce the notion that the TIR and CC domains may play a role in activation of downstream pathways (Van der Biezen and Jones, 1998a; Van der Biezen and Jones, 1998b). LRR domains have been implicated in the perception of pathogen Avr products (Botella et al., 1998; Ellis et al., 1999; Jia et al., 2000; Parniske et al., 1997; Thomas et al., 1997).

In Arabidopsis, members of both NB-LRR classes confer resistance to the oomycete Peronospora parasitica and the bacterium Pseudomonas syringae, whereas one member of the CC-NB-LRR class also confers viral resistance (Bittner-Eddy et al., 2000; Cooley et al., 2000; Parker et al., 2000). The apparent lack of a correlation between the predicted modular NB-LRR structures and recognition of certain pathogen types is consistent with the idea that the TIR and CC portions are involved in activation of defence pathways, rather than in pathogen perception (Aarts et al., 1998). However, in the flax L alleles both the LRR and TIR domains play a role in determining pathogen specificity (Ellis et al., 1999; Luck et al., 2000). Mutational analyses in Arabidopsis have identified several genes required for either of the two NB-LRR protein classes, e.g. enhanced disease susceptibility 1 (eds1; Aarts et al., 1998); non-race-specific disease resistance 1 (ndr1; Century et al., 1995); and avrPphB susceptible 2 (pbs2; Warren et al., 1999). The results of these studies suggest that TIR-NB-LRR proteins require EDS1, whereas CC-NB-LRR proteins require NDR1 and PBS2 for function. Thus in Arabidopsis at least two apparently mutually exclusive parallel defence-signalling pathways are differentially employed by NB-LRR proteins that are distinguished by their TIR or CC domains (Aarts et al., 1998; Warren et al., 1999). PBS3 appears to be involved in both pathways, and possibly establishes a convergence point (Warren et al., 1999). A notable exception is the CC-NB-LRR protein RPP8 for resistance to P. parasitica that exhibits little or no requirement for any of the tested components (Aarts et al., 1998; McDowell et al., 1998, 2000).

In Arabidopsis landrace Columbia (Col), RPP4 confers resistance to P. parasitica races Emoy2 and Emwa1 (Holub et al., 1994). RPP4-mediated resistance has been shown to involve multiple defence-signalling components, including DETACHMENT 9 (DTH9; Mayda et al., 2000); EDS1 (Aarts et al., 1998); NDR1 (Century et al., 1995; Century et al., 1997); NON-EXPRESSOR OF PR GENES 1 (NPR1, also known as NIM1 and SAI1; McDowell et al., 2000); PHYTOALEXIN DEFICIENT 4 (PAD4; Glazebrook et al., 1997); PBS2 and PBS3 (Warren et al., 1999); and SALICYLIC ACID INDUCTION-DEFICIENT 1 and 2 (SID1 and SID2; Nawrath and Métraux, 1999). In addition, RPP4 function is compromised by the rps5-1 defective allele of the CC-NB-LRR-encoding gene RPS5 for resistance to P. syringae, possibly through interference with a shared signalling partner (Warren et al., 1998). Furthermore, RPP4-mediated resistance is diminished by in vivo inhibition of phenylalanine ammonium lyase (PAL) enzyme activity, the entry point of the phenylpropanoid pathway leading to salicylic acid synthesis (Mauch-Mani and Slusarenko, 1996). In line with this result is the observation that removal of salicylic acid in plants expressing the bacterial salicylate hydroxylase gene NahG (Gaffney et al., 1993) also compromises RPP4 function (Delaney et al., 1994; McDowell et al., 2000).

RPP4-mediated resistance differs from other race- specific resistances conferred by RPP and RPS genes, in that it is affected by mutations in many signalling components rather than only a subset (Feys and Parker, 2000). To characterize further the basis of resistance to P. parasitica Emoy2 and Emwa1, we isolated the RPP4 gene. RPP4 maps on chromosome 4 within a genetic interval that includes RPP5 homologues which we have characterized previously (Noël et al., 1999). RPP5 in Landsberg erecta (Ler) confers resistance to P. parasitica race Noco2 and is the founder member of the multigene family with diverged members in different landraces (Noël et al., 1999; Parker et al., 1997). RPP5 and its homologues belong to the TIR-NB-LRR class, and have variable configurations and numbers of LRRs (Noël et al., 1999). At the Col-0 haplotype, eight polymorphic RPP5 homologues are clustered within ≈80 kb (Noël et al., 1999), and although many members carry retro-element insertions or frameshift mutations, analysis of the codon usage indicated that most genes once were, or still are, functional (Noël et al., 1999).

We describe the identification of RPP4 in Col-0 as the orthologue of RPP5 in Ler. Besides mediating resistance to P. parasitica Noco2, we show that RPP5 also confers resistance to P.p. races Emoy2 and Emwa1, and thus accounts for the Ler RPP4 specificity. Comparison between RPP4 and RPP5 identifies residues within the TIR and LRR domains that may be important for recognition specificity. Like RPP5, RPP4 interacts in a yeast two-hybrid assay with an Arabidopsis RelA/SpoT homologue (Van der Biezen et al., 2000). The identification of RPP4 as a TIR-NB-LRR protein coupled with the involvement of DTH9, EDS1, NDR1, NPR1, PAD4, PAL, PBS2, PBS3, rps5-1, SID1, SID2 and salicylic acid illustrates the complexity by which this class of NB-LRR proteins signals to activate defence.


RPP5-ColA confers RPP4 function

For fine mapping of RPP4, we crossed Col-0 to the P. parasitica race Emwa1-susceptible landrace Wassilewskija (Ws-0). In the resulting F2 population (n = 1243), 291 plants homozygous at the Ws-0 rpp4 locus were identified by the full susceptibility of their first true leaves to P.p. Emwa1. The proportion of resistant versus susceptible individuals corresponds (P = 0.05) to a 3 : 1 ratio (χ2 = 1.67), confirming that RPP4 is the only Col-0 locus conferring resistance to P.p. Emwa1. Plants heterozygous at RPP4 were partially resistant, and the cotyledons allowed moderate to full asexual sporulation. This indicates that RPP4 exhibits incomplete dominance. Susceptible individuals (rpp4/rpp4) were examined with molecular markers (see Experimental procedures). RPP4 mapped between markers VRN2 and FCA8-2 and co-segregated with the G4539 marker (Figure 1a), delimiting RPP4 to a physical segment of maximally 304 kb (GenBank Z97342 and Z97343). This region includes eight RPP5 homologues (Noël et al., 1999) and 55 other predicted genes, none of which is an obvious R gene candidate (Bevan et al., 1998). We therefore tested the possibility that one or more Col-0 RPP5 homologues conferred resistance to P.p. races Emoy2 and Emwa1.

Figure 1.

Mapping of RPP4 at the Col-0 RPP5 locus and identification of Col-0 RPP5 homologues in binary cosmid vectors.

(a) RPP4 maps within a short segment (grey area) on the long arm of chromosome 4 (left, centromeric; right, telomeric) defined by molecular markers (described in Experimental procedures) at intervals expressed in kilobases (kb, top) and centiMorgans (cm, below). The region indicated by the arrows contains eight RPP5 homologues.

(b) Restriction mapping of cosmid inserts to identify specific Col-0 RPP5 homologues. BamHI and SalI-digested cosmid DNA was separated through a 1% agarose-TAE gel, stained with ethidium bromide and photographed during exposure to UV light. Fragment sizes were inferred from co-migrating marker DNA at the borders of the gel, and are indicated in kilobases (kb). The arrows indicate 20 kb BamHI/SalI and 4.5 kb SalI/SalI cosmid vector fragments.

(c) Cosmid contig of the Col-0 RPP5 haplotype (Noël et al., 1999). The RPP5 homologues (Col-A to Col-H) are shown as boxes with the direction of transcription indicated by arrowheads. Asterisks indicate the positions of the open reading frame-disrupting point mutations. Retro-element insertions are shown as grey arrowheads. The restriction map (B, BamHI; S, SalI) is shown below the RPP5 homologues with sizes in kilobases (kb). The 10 cosmids (thick lines) were mapped on the genomic sequence by restriction analysis and DNA sequencing.

Ten binary cosmids were identified that collectively contained all six candidate Col-0 RPP5 homologues (Figure 1b,c), and these were transformed into the P.p. Emoy2- and Emwa1-susceptible line CW-84 (see Experimental procedures). Self-progenies of the transgenic CW-84 lines (T2) were then tested for complementation of resistance to P.p. Emoy2 and Emwa1. Only T2 progenies with cosmid-3 or cosmid-27 segregated for resistance to these P. parasitica races, whereas progenies segregating for the other cosmids allowed profuse asexual sporulation, resembling non-transformed CW-84 controls (Figure 2; Table 1). T3 progeny-testing by infection assays, PCR analysis, and kanamycin-resistance (transformation marker) assays showed that resistance to P.p. Emoy2 and Emwa1 co-segregated with cosmid-3 or cosmid-27. T3 plants homozygous for the cosmid-3 or cosmid-27 inserts were all resistant to P.p. races Emoy2 and Emwa1, but were fully susceptible to P.p. races Cala2 and Noco2, providing evidence for recognition specificity of the transgenes (Table 1). Examination of lactophenol–trypan blue-stained leaves infected with P.p. Emoy2 or Emwa1 showed that the resistance responses of transgenic CW-84 lines with cosmid-3 or cosmid-27 was similar to the resistance of Col-0 (Figure 2).

Figure 2.

Cosmid-3 and cosmid-27 confer RPP4 function.

Complementation of the universally susceptible recipient CW-84 by cosmid-3 and cosmid-27 for resistance to P. parasitica races Emoy2 and Emwa1.

(a) Asexual sporulation on P. parasitica race Emoy2-infected true leaves of 2-week-old CW-84 seedlings.

(b) Resistance to infection by P. parasitica Emoy2 of a typical CW-84 transformant containing a single-copy insert of cosmid-27.

(c) Lactophenol–trypan blue staining of P. parasitica Emoy2 hyphae in true leaves of CW-84 photographed at 100× magnification.

(d) Absence of hyphae in lactophenol–trypan blue-stained true leaves of a typical CW-84 transformant carrying cosmid-27.

Table 1.  Transgenic complementation of resistance to P. parasitica
CW-84 genotypeNumber of
lines testeda
P. parasitica races
  • a Between 30 and 60 T2 individuals obtained by selfing of primary T1 transformants were analysed in at least two separate experiments for segregation of resistance and the transgene.

  • b

    R, resistant progenies (no asexual sporulation and no mycelium); S, susceptible progenies (full asexual sporulation and full mycelium).

Non-transformed control1SbSSS
Cosmids 3/2712 eachSRRS
Cosmids 4/9/10/12/13/17/29/3212 eachSSSS

The complementing and overlapping cosmid-3 and cosmid-27 contain the RPP5-ColA gene (7673 bp) that carries a Copia-like retro-element insertion at its 3′-end (5213 bp; Figure 1c; Noël et al., 1999). Cosmid-3 has an insert size of 22.2 kb, and contains 895 bp 5′ of the predicted ATG translation start codon of RPP5-ColA, and 8.5 kb 3′ of the predicted stop codon. This region includes a predicted gene with weak similarity to ubiquinol-cytochrome-c reductase (UCR). Cosmid-27 has an insert size of 23.0 kb, and contains 7.8 kb sequence 5′ of RPP5-ColA, and 2.3 kb 3′ of its stop codon (with only one-third of the UCR gene). From the complementation experiments and the 16.1 kb overlap between cosmid-3 and cosmid-27, we concluded that RPP4 specificity is conferred by RPP5-ColA. Like RPP5 in Ler, the RPP5-ColA homologue, hereafter referred to as RPP4, occupies the most centromeric position within the RPP5 gene cluster in Col-0 (Figure 1c; Noël et al., 1999).

Structure of RPP4 and predicted amino acid sequence

RPP4 and RPP5 are highly homologous and structurally very similar, including the position and number of introns (Figure 3a). The retro-element is inserted at the codon for amino acid residue 1113 and prevents expression of the last 130 codons and 3′-untranslated region (UTR); the first stop codon in the retro-element occurs after 30 codons (Figure 3b). The last exon of RPP5, encoding 26 amino acids and the 3′-UTR, was also found to be not required for function (Parker et al., 1997). These portions do not encode LRRs and have no apparent homologies in the databases. Apart from RPP4, all other 17 RPP5 homologues in Col-0 and Ler have an intron-1 that is nearly identical in size (126 bp) and sequence (Noël et al., 1999). The unusually large intron-1 (3.2 kb) of RPP4 is probably the result of an independent insertion event as the flanking regions are similar to those of RPP5 and other homologues. Intron-1 of RPP4 is A/T-rich (77%) and does not show similarity to other sequences in the databases. RPP4 also contains an in-frame deletion in exon-6 of a 276 bp DNA fragment encoding four LRRs, which is a unique event within the RPP5 family (Figure 3b; Noël et al., 1999).

Figure 3.

Comparison of the RPP4 and RPP5 genes and products.

(a) RPP4 and RPP5 gene structures. The solid rectangles indicate coding regions (exons 1–8), narrow rectangles indicate 5′ and 3′ UTRs, and introns are shown by arrowheads. The Copia-like retro-element inserted at exon 6 of RPP4 is shown as a grey arrow. Intron-1 of RPP4 (3.2 kb) is markedly larger than that of RPP5 (126 bp), and exon 6 of RPP4 is shorter than that of RPP5 because of an in-frame 276 bp deletion.

(b) The RPP4 protein and its predicted TIR, NB-ARC and LRR structures are shown as given by Parker et al. (1997) and Noël et al. (1999). Intron–exon boundaries are indicated by diamonds, and exon numbers are indicated at the right. The amino acid sequence of RPP4 is shown and is compared to that of RPP5; black residues are conserved between RPP4 and RPP5, and residues shown in red are different. The 30 predicted amino acid residues resulting from the retro-element (RE) insertion in exon-6 are shown in green; the first stop codon introduced by the retro-element is indicated by an asterisk. Residues shown in lower-case letters either are part of highly variable regions and cannot be aligned (exon-3) or, as a result of the retro-element insertion in exon-6, are predicted not to be expressed (exons 6–8). The deletion in the LRR region of RPP4 relative to the corresponding region of RPP5 is shown by dashes. Conserved motifs in the NB-ARC domain, and conserved hydrophobic residues within the LRRs, are shown in bold. Predicted solvent-exposed residues (x in the xxLxLxx motif) are shown in boxes.

Comparison of the predicted amino acids shows highly related RPP4 and RPP5 proteins (74% amino acid identity/78% similarity; Figure 3b). The TIR domains (RPP4 residues 1–157) are most different (21% divergence), whereas the NB-ARC domains (RPP4 residues 158–516) display only 7% divergence (Figure 3b). RPP4 contains 17 LRRs which show 18% divergence with these of RPP5 (21 LRRs). Individual LRRs of the RPP5 family on average consist of 24 residues, of which five highly variable residues of the xxLxLxx motif (L = leucine, x = any amino acid) are predicted to be solvent-exposed and to specify pathogen recognition. The remaining residues are conserved and are predicted to serve a structural function (Noël et al., 1999). RPP4 and RPP5 show only 12% divergence in their structural LRR residues (36 out of 306 residues) but display 40% divergence (34 out of 84 residues) in their solvent-exposed residues (Figure 3b). Previous codon-usage analysis in RPP4 indicated divergent selection of solvent-exposed LRR residues (Noël et al., 1999).

RPP4 and RPP5 require multiple defence-component genes

In cotyledon assays, RPP4 function has been shown to need various signalling components (Aarts et al., 1998; Century et al., 1995; Century et al., 1997; Glazebrook et al., 1997; McDowell et al., 1998; Warren et al., 1998; Warren et al., 1999; Table 2). To examine the requirement for RPP4-mediated resistance to P.p. races Emoy2 and Emwa1 in true leaves, we analysed the responses of seven Col-0 defence mutants and a Col-0 NahG transformant line (Table 2). The DTH9 (Mayda et al., 2000), SID1 and SID2 mutants (Nawrath and Métraux, 1999) were tested in other studies. The amount of hyphal growth in the pbs2 and pbs3 mutants was comparable to that of the genetically compatible interactions of P.p. Emwa1 with Ws-0 or P.p. Noco2 with Col-0 (Figure 4a,g,h; Table 2). The eds1-2 mutant plant supports even higher levels of P. parasitica growth (Figure 4c; Table 2) consistent with eds1 conferring an ‘enhanced disease susceptibility’ phenotype (Parker et al., 1996). Compared to eds1-2 (Figure 4c) or to compatible interactions with wild-type plants (Figure 4a), significantly less asexual sporulation and lower amounts of mycelium were observed in the pad4-1 mutant (Figure 4d; Table 2). Furthermore, in pad4-1 P. parasitica hyphae were surrounded by mesophyll cells that retained the lactophenol–trypan blue stain (Figure 4b,g), and also fluoresced during exposure to ultra violet (UV) light (Figure 4f). Fluorescence is commonly observed in cells that have undergone a hypersensitive response (HR) and is probably due to the accumulation of phenolic compounds. The dead or dying plant cells accompanying the hyphae in pad4-1 mutants may therefore be the result of a delayed HR that only partially prevents pathogen growth. RPP4-dependent trailing necrosis was also observed previously in DTH9 (Mayda et al., 2000), SID1 and SID2 mutants (Nawrath and Métraux, 1999).

Table 2.  Mutations affecting RPP4-mediated resistance and systemic acquired resistance
GenotypeP. parasitica races Emoy2 and Emwa1P. parasitica race Noco2
CotyledonsTrue leavesTrue leaves
Non-treatedBTH treatedNon-treatedBTH treated
  • a

    –, No mycelial growth and no sporulation; +, low mycelial growth and sporulation; ++, intermediate mycelial growth and sporulation; +++, full mycelial growth and sporulation comparable to compatible interactions with wild type plants; ++++, enhanced mycelial growth and sporulation.

  • b

    P. parasitica race Emwa1 only.

  • c Selected from a Col-gl  ×Ler eds1-2 cross; homozygous at Col-RPP4 and Col-rpp8 (Aarts et al., 1998).

  • d

    nd, Not determined.

  • e Data from Nawrath and Métraux (1999).

  • f Data from Mayda et al. (2000).

Col-0 (RPP4)+/–a+++
Ws-0 (rpp4)+++b+++bb
Col/Ler eds1-2c++++++++++++
Col-0 ndr1-1++++
Col-0 npr1-1+++++++
Col-0 pad4-1+++++++
Col-0 pbs2+++++++++
Col-0 pbs3+++++++++
Col-0 rps5-1++++
Col-0 NahG++++nd++++end
Col-0 sid1ndd++bend++++end
Col-0 sid2nd++bend++++end
Col-0 dth9nd++bfnd++++fnd
Figure 4.

RPP4 function requires a multitude of signalling components.

True leaves of 2-week-old Col-0 wild-type and mutant seedlings were infected with P. parasitica and after 5 days stained with lactophenol–trypan blue.

(a) Col-0 wild-type control infected with the compatible P.p. race Noco2. Hyphae ramify abundantly through mesophyll tissue.

(b) Intercellular P.p. Noco2 hyphae with intracellular haustoria in Col-0 wild-type mesophyll cells (100× magnification).

(c) Col-0/Ler eds1-2 mutant infected with P.p. Emoy2 shows enhanced susceptibility.

(d) Col-0 pad4-1 mutant infected with P.p. Emoy2 shows partial resistance.

(e) P.p. Emoy2 hyphal ramification in the Col-0 pad4-1 mutant induces death of surrounding (lactophenol–trypan blue-stained) mesophyll cells (100× magnification).

(f) Dead or dying mesophyll cells accompanying P.p. Emoy2 hyphae in the Col-0 pad4-1 mutant fluoresce during exposure to UV light (100× magnification).

(g) Col-0 pbs2 mutant infected with P.p. Emoy2 shows full susceptibility.

(h) Col-0 pbs3 mutant infected with P.p. Emoy2 shows full susceptibility.

(i) Col-0 ndr1-1 mutant infected with P.p. Emoy2 shows no effect on RPP4 function.

A slight but quantitative increase of P.p. Emoy2 and/or Emwa1 asexual sporulation was previously observed in cotyledons of the Col-0 mutants ndr1-1, npr1-1 and rps5-1 (Century et al., 1995; Century et al., 1997; Glazebrook et al., 1997; Holub et al., 1994; McDowell et al., 2000; Warren et al., 1998). We confirmed these results by microscopic inspection of infected cotyledons. However, we found that in true leaves of these mutants, P.p. Emoy2 and Emwa1 resistance was not compromised. Asexual sporulation was not observed and mycelial growth was not detected in lactophenol–trypan blue-stained tissues (Figure 4i; Table 2). Hence the requirement by RPP4 for the signalling components EDS1, PAD4, PBS2 and PBS3 is manifested in both cotyledon and true leaf tissues, while the effect of the ndr1-1, npr1-1 and the rps5-1 mutations is observed in cotyledons only. The differential requirement for signalling components in two different organs suggests that RPP4-mediated resistance may be developmentally or tissue-specifically regulated.

Salicylic acid is essential for the establishment of systemic acquired resistance (SAR; Delaney et al., 1994; Gaffney et al., 1993), and has been suggested to play a major role in RPP4-mediated resistance in true leaves (Mauch-Mani and Slusarenko, 1996). Moreover, the bacterial salicylate hydroxylase gene NahG, which converts salicylic acid to catechol (Gaffney et al., 1993), compromised RPP4 function in cotyledons (Delaney et al., 1994; McDowell et al., 2000) and in true leaves (Nawrath and Métraux, 1999; Table 2). In addition, the sid1 and sid2 mutants that are impaired in the pathway leading to salicylic acid biosyntheis are also required for RPP4 function in true leaves (Nawrath and Métraux, 1999). We examined further the relationship of the RPP4 pathway defined by EDS1, PAD4, PBS2 and PBS3 in true leaves, with the NPR1 pathway regulating activation of SAR in response to the salicylic acid analogue benzothiadiazole (BTH; Lawton et al., 1996). Sporulation on BTH-treated Col-0 eds1-2, pad4-1, pbs2 and pbs3 mutants was not observed on true leaves, indicating that the SAR pathway downstream of BTH perception is still intact in these mutants (Table 2). As expected, foliar BTH application did not induce SAR in npr1-1 plants (Table 2).

As RPP4 and RPP5 are highly homologous, we wished to compare their genetic requirements. The eds1-2 mutation suppresses RPP4 and RPP5 function to a similar extent (Aarts et al., 1998; data not shown). Also, RPP5-mediated resistance to P.p. Noco2 in the Ler pad4-2 mutant was affected in a similar way to RPP4-mediated resistance in a Col-0 pad4-1 background, including the typical trailing necrosis (Feys et al., 2001). Furthermore, like RPP4, RPP5 function was not compromised by the ndr1-1 and npr1-1 mutations in true leaves (Aarts et al., 1998; data not shown). The segregation of resistance to P. p. Noco2 in F2 and F3 progenies from a Ler-RPP5 × Col-pbs2 cross, and the loss of RPP5 function in four selected F3 plants homozygous for both RPP5 and the pbs2 mutation, showed that PBS2 is fully required for RPP5 function (data not shown). These results indicate that, in true leaves, RPP4 and RPP5 similarly require the EDS1, PAD4 and PBS2 genes, but do not appear to require the NDR1 and NPR1 genes. The effects of the pbs3 and rps5-1 mutations on RPP5 function were not tested.

RPP4 and RPP5 interact in yeast with an Arabidopsis RelA/SpoT homologue

We previously separated the NB-ARC domain of the RPP5 family in an NB site (≈130 residues) and an ARC domain (≈180 residues; Noël et al., 1999; Van der Biezen and Jones, 1998a). Phylogenetic grouping of all full-length RPP5 family members showed that NB sites are nearly identical, while there are three different classes of ARC domains (Noël et al., 1999). The ARC domain of RPP4 belongs to the same class of that of RPP5. We reported that the full-length NB-ARC domain of RPP5 is necessary and sufficient to interact in a yeast two-hybrid assay with a C-terminal hydrophilic domain (≈160 residues) of the Arabidopsis RelA/SpoT homologue, At-RSH1 (Van der Biezen et al., 2000). It is not known whether the interaction with At-RSH1 is relevant for RPP5 function. We show here that co-expression in yeast of At-RSH1 with the NB-ARC domain of RPP4 activates the two-hybrid reporter genes in a similar manner to that of RPP5 (Figure 5). In contrast, the same domain of the two other RPP5 family members, Col-F and Col-G, which belong to the two other ARC classes, did not interact with At-RSH1 (Figure 5).

Figure 5.

RPP4 and RPP5 interact in a yeast two-hybrid assay with the Arabidopsis RelA/SpoT homologue At-RSH1.

A dendrogram of NB-ARC domains of RPP4, RPP5 and the RPP5 homologues Col-F and Col-G shows protein-sequence distance relationships (Noël et al., 1999). The NB-ARC domain of RPP5 was previously shown to interact in a yeast two-hybrid assays with At-RSH1 (Van der Biezen et al., 2000). Here we report that the closely related NB-ARC domain of RPP4 also activates the LacZ (blue) and LEU2 (not shown) reporters. The NB-ARC domains of the more distantly related homologues Col-F and Col-G do not activate the LacZ (white) and LEU2 reporters.

RPP4 specificity in Landsberg erecta is conferred by RPP5

In Ler, resistance to P.p. Emoy2 and Emwa1 co-segregated with the RPP5 locus and has been denoted RPP4 (Holub et al., 1994). However, it was not known whether Ler RPP4 is a distinct gene or an allele of RPP5 conferring P.p. Emoy2 and Emwa1 recognition. DNA sequence and transcript analysis indicated that RPP5 is the only functional gene in the Ler RPP5 haplotype (Noël et al., 1999). Therefore we tested whether RPP5 itself confers resistance to P.p. Emoy2 and Emwa1, besides mediating resistance to P.p. Noco2. In this analysis we used a P.p. Noco2-sensitive Ler mutant that carries the defective rpp5-2 allele identified in an ethyl methanesulfonate (EMS)-mutagenized population (Louise N. Frost and J.E.P., unpublished results). The presence of the RPP8 gene in Ler for resistance to P.p. Emoy2 and Emwa1 (McDowell et al., 1998) prevented testing directly whether the rpp5-2 mutant had also lost resistance to P.p. Emoy2 and Emwa1 resistance. We therefore conducted a segregation analysis. In F2 progeny (>1000 plants) from the wild-type Ler × Col-0 control cross, we found no individuals that were susceptible to P.p. Emoy2 or Emwa1 infection, confirming that RPP4 in Col-0 and the recognition of P.p. Emoy2 and Emwa1 in Ler map at the same locus (Table 3). However, in F2 progeny from the mutant Ler rpp5-2 × Col-0 cross, individuals susceptible to P.p. Emoy2 and P.p. Emwa1 occurred in a ratio of 1 : 15 resistant plants, consistent with the segregation of two unlinked genes, Col-0 RPP4 and Ler RPP8 (Table 3). The concomitant loss of P.p. Noco2, Emoy2 and Emwa1 resistance in the single-mutant rpp5-2 indicates that RPP5 recognizes not only P.p. race Noco2, but also P.p. races Emoy2 and Emwa1.

Table 3. RPP4 Specificity in Landsberg erecta (Ler) is conferred by RPP5. Segregation of the defective rpp5-2 allele shows that loss of P. parasitica Noco2 resistance is accompanied by loss of resistance to P.p. Emoy2 and Emwa1
P. parasitica
F2 progeny from wild-type Ler × Col-0F2 progeny from Ler rpp5-2 × Col-0
Resistant: susceptibleaRatioaχ2bResistant: susceptibleaRatioaχ2b
  • a

    Resistant, no asexual sporulation; susceptible, full asexual sporulation; these are combined results of two independent replicates.

  • b Goodness-of-fit (χ2) significant at P = 0.05 with 1 degree of freedom is 3.84.

  • cP.p. Emco5 is recognized by Ler RPP8 only (McDowell et al., 1998).

Cala2147 : 443 : 10.39182 : 563 : 10.27
Emoy2>1000 : 01 : 00432 : 2415 : 10.76
Emwa1>1000 : 01 : 00216 : 1115 : 10.77
Emco5c97 : 263 : 10.9771 : 243 : 10
Noco2132 : 463 : 10.070 : >10000 : 10

To confirm that RPP5 confers resistance to P.p. races Emoy2 and Emwa1, CW-84 plants were transformed with the RPP5 transgene (Parker et al., 1997). In T2 and T3 progenies from seven independently transformed plants, the RPP5 transgene co-segregated with resistance to P.p. races Noco2, Emoy2 and Emwa1 (Table 1). These results strongly support the conclusion that the RPP4 specificity for P.p. Emoy2 and Emwa1 recognition in Ler is conferred by RPP5.


Intraspecific allelic divergence at a complex disease-resistance locus

We identified RPP4 for resistance to P. parasitica races Emoy2 and Emwa1 in the landrace Col-0 as an allele of RPP5 which also confers resistance to P.p. Noco2 in Ler (Parker et al., 1997). We previously analysed the RPP5 haplotypes of the landraces Col-0 and Ler that comprise small multigene families of eight and 10 members, respectively (Noël et al., 1999). Evolution of the RPP5 locus involved pronounced haplotype divergence in several landraces, which included point mutations, deletion and duplication events, gene conversions, and intergenic and intragenic recombinations. Fifteen out of all 18 RPP5 family members are truncated or carry frameshift mutations, suggesting that most RPP5 homologues are not functional. However, the solvent-exposed, presumably ligand-interacting LRR residues are highly variable among the RPP5 family, and were predicted to specify pathogen recognition in the recent evolutionary past (Noël et al., 1999). Furthermore, analysis of the ratio of non- synonymous over synonymous nucleotide substitutions in codons encoding the solvent-exposed LRR residues provided evidence of divergent selection (Noël et al., 1999). The cloning of RPP4 shows that at least one of the six intact or 3′-truncated Col-0 genes, RPP5-ColA, is functional.

The most distinguishing features of RPP4 are the large insertion in intron-1 (3.2 kb), the insertion of a retro-element at the 3′-end, and an in-frame deletion of a 276 bp segment encoding four LRRs, which are all unique events within the RPP5 family (Figure 3). The similar position of RPP4 and RPP5 in the gene cluster of the Col-0 and Ler haplotypes, respectively, and their close sequence affiliation characterized by informative polymorphic sequences, indicate that RPP4 and RPP5 are true orthologues (Noël et al., 1999). Allelic variation at loci that specify resistance to different races of the same fungal species is commonly observed, e.g. the flax L alleles against rust races, barley Mla alleles against powdery mildew races, and the tomato Cf alleles against leaf mould races. Recently, Bittner-Eddy et al. (2000) cloned two alleles of the single-copy Arabidopsis RPP13 gene that encode CC-NB-LRR proteins recognizing different P. parasitica races. The RPP4 and RPP5 alleles encode distinct specificities, which reinforces the contention that they have evolved from a common ancestral gene through adaptive selection to different pathogen races (Noël et al., 1999). In natural populations like Arabidopsis, selection-driven allelic variation at resistance loci generates spatial and genetic diversity (‘balancing polymorphisms’) that may reduce the selection pressure for pathogens adapting to virulence by frequency-dependent selection (e.g. Noël et al., 1999).

Different recognition specificities of the RPP4 and RPP5 alleles

In Ler, two unlinked genes specify resistance to P.p. Emoy2 and Emwa1: the RPP8 gene on chromosome 5, and a second gene that is linked to the RPP5 locus (Holub et al., 1994; Tör et al., 1994). Although the latter Ler P.p. Emoy2/Emwa1 specificity could not be separated by recombination from RPP5, the gene was tentatively named Ler RPP4 by inference from the RPP4 gene in Col-5 that maps at a similar position and also confers resistance to the same P.p. races (Holub et al., 1994; Tör et al., 1994). We show here that Ler RPP4 is not a distinct gene, and that RPP5 confers P.p. Noco2, Emoy2 and Emwa1 recognition. Therefore RPP4 and RPP5 have overlapping specificities, which is best explained to occur through recognition of distinct avirulence (Avr) determinants (Table 4). P.p. races Noco2, Emoy2 and Emwa1 presumably share the Avr factor that is recognized by RPP5 (i.e. AvrRPP5), and in addition, P.p. Emoy2 and Emwa1 carry an Avr gene that is recognized by RPP4 (i.e. AvrRPP4; Table 4). A similar scenario has been postulated for recognition of different P. parasitica Avr genes by three RPP1 paralogues (Botella et al., 1998; Table 4), and by two RPP13 orthologues (Bittner-Eddy et al., 2000).

Table 4.  Recognition specificities in three Arabidopsis landraces of five P. parasitica races used in this study
P. parasitica
Recognition of P. parasitica (RPP)
by Arabidopsis landrace
Deduced presence of P. parasitica Avr genesa
  • a

    +, Avr gene present; –, Avr gene absent.

Emoy2RPP4RPP5, RPP8RPP1A,B+++++
Emwa1RPP4RPP5, RPP8none+++

Characterization of functional alleles for resistance to different pathogen races or species has identified LRR residues involved in the determination of recognition (Bittner-Eddy et al., 2000; Botella et al., 1998; Cooley et al., 2000; Ellis et al., 1999; Thomas et al., 1997; Van der Vossen et al., 2000). LRR domains presumably form surfaces for direct or indirect interactions with pathogen-derived ligands (Jia et al., 2000; Jones and Jones, 1996). RPP4 and RPP5 differ mainly in residues encoding the TIR domain, predicted ligand-interacting LRR residues, and in LRR copy number. Hypervariability of ligand-interacting LRR residues, and differences in LRR copy number, could create surfaces for different ligand interactions, and hence potential for differential pathogen recognition. The TIR domain has also recently been implicated in pathogen recognition by comparing flax L alleles that differ only in their TIR portions (Luck et al., 2000). This is consistent with a role of the TIR domain as effector activating downstream signalling components (O'Neill and Greene, 1998). Defence activation presumably results from both recognition and intramolecular interactions involving the LRR and TIR domains (Van der Biezen and Jones, 1998b). Differences between the TIR domains of RPP4 and RPP5 may also result from co-evolution with the diverging LRR domains.

RPP4 and RPP5 function through multiple-signalling components

RPP4 signals through at least 12 defence components that were previously implicated in pathogen resistance (Table 2; Figure 6). EDS1 was shown to be required by RPP1, RPP5 and RPS4 (resistance to Pseudomonas syringae) that all belong to the TIR-NB-LRR class (Aarts et al., 1998). NDR1 and PBS2 were shown to be required for resistance to P. syringae by RPS2, RPS5 and RPM1 which all belong to the CC-NB-LRR class (Aarts et al., 1998; Warren et al., 1999). PBS3 was required for full function of both NB-LRR protein classes, i.e. RPS2, RPS5 and RPM1 and RPS4 (Warren et al., 1999). The identification of RPP4 as a TIR-NB-LRR protein, and its requirement in true leaves for EDS1 and PBS3, but not for NDR1, are in line with the differential utilization by the two NB-LRR classes of the two parallel resistance pathways defined by EDS1 and NDR1 that may converge at PBS3 (reviewed by Feys and Parker, 2000). However, our results show that PBS2 is required for signal transduction in both pathways (Warren et al., 1999), and suggest a role of PAD4 in the EDS1 pathway (Figure 6). An EDS1–PAD4 interaction has been inferred in other molecular and genetic studies (Feys et al., 2001). The CC-NB-LRR protein RPS2 does not require PAD4 (Zhou et al., 1998), perhaps indicating that PAD4 is exclusively part of the EDS1 pathway (Figure 6).

Figure 6.

RPP4-mediated defence signalling towards activation of resistance to P. parasitica Emoy2.

Interaction of AvrRPP4 with RPP4 in P.p. Emoy2-infected Col-0 cells leads to rapid and effective resistance responses. In true leaves, RPP4-mediated signalling requires the presence of the defence components DTH9, EDS1, PAD4, PAL, PBS2, PBS3, SID1 and SID2, and salicylic acid accumulation (see text). The relative position and the function of these signalling components in the RPP4 pathway is not known. The dth9, ndr1-1, pbs2, pbs3, rps5-1, sid1 and sid2 mutations also compromise P. syringae resistance conferred by one or several CC-NB-LRR genes. PAL is the entry point of the phenylpropanoid pathway leading to the production of salicylic acid and lignin. Inhibition in true leaves of PAL activity by 2-aminoindan-2-phosphonic acid (AIP) compromises RPP4 function (Mauch-Mani and Slusarenko, 1996). In true leaves, salicylic acid plays a role in both the NPR1-dependent SAR pathway and the RPP4 pathway, but SAR operates downstream or independently of the RPP4 pathway. In cotyledons (but not in true leaves), RPP4 function is affected in the Col-0 rps5-1, ndr1-1, and npr1-1 mutants, indicating that, at least in cotyledons, there may be cross-talk between the TIR-NB-LRR and CC-NB-LRR pathways, and may also suggest that SAR is an integral part of RPP4-mediated resistance in cotyledons but not in true leaves. EDS1 and PAD4 share homology to eukaryotic lipases (Falk et al., 1999; Jirage et al., 1999); RPS5 is CC-NB-LRR protein (Warren et al., 1998); NDR1 is a predicted two-transmembrane (2TM) protein (Century et al., 1997); and NPR1 is an ankyrin-repeat protein that interacts with a certain class of bZIP transcription factors possibly involved in the activation of PATHOGENESIS RELATED (PR-1) genes (Cao et al., 1997; Zhang et al., 1999).

Low sporulation of P.p. Emoy2 and Emwa1 can be observed on cotyledons of wild-type Col-0 plants, and in plants heterozygous at RPP4, partial resistance is predominantly expressed in cotyledons (Century et al., 1995; Century et al., 1997; Glazebrook et al., 1997; Holub et al., 1994; data not shown). These observations suggest that RPP4-mediated resistance is less effective in cotyledon tissue than in true leaves. On cotyledons of Col-0 plants carrying the ndr1, npr1 and rps5-1 mutations, a small but statistically significant enhancement of P. parasitica sporulation has been observed, while no sporulation or mycelium was detected in true leaves (Table 2). The contrasting phenotypes of the two different tissues suggest a developmental regulation of RPP4-mediated resistance. They may also reflect reduced penetrance of RPP4 resistance in cotyledon tissue, permitting a more sensitive assessment of the involvement of certain defence components. Consequently, the cotyledon-resistance assays may reveal a slight but quantitative involvement of NDR1, NPR1 and the rps5-1 allele in RPP4 function (Table 2).

Phenylalanine ammonium lyase (PAL) and cinnamyl alcohol dehydrogenase (CAD) are components of the phenylpropanoid pathway, leading to the synthesis of secondary metabolites such as salicylic acid, lignin, and flavonoids (Mauch-Mani and Slusarenko, 1996). The in vivo requirement of PAL and CAD enzymatic activity in Col-0 leaves for full P. parasitica Emwa1 resistance functionally implicated salicylic acid and lignin in RPP4-mediated resistance (Mauch-Mani and Slusarenko, 1996). In PAL-suppressed Col-0 plants, resistance to P.p. Emwa1 was restored by exogenous salicylic acid application, suggesting that salicylic acid production is a major function of PAL in resistance to P. parasitica (Mauch-Mani and Slusarenko, 1996). Indeed, removal of salicylic acid by expression of the bacterial salicylate hydroxylase gene NahG impairs RPP4 function in Col-0 true leaves (Nawrath and Métraux, 1999; Table 2). The sid1 and sid2 mutants that are impaired in the pathway leading to salicylic acid biosynthesis show impaired RPP4 function in true leaves (Nawrath and Métraux, 1999). In contrast, the salicylic acid response element NPR1, which is required for the induction of systemic acquired resistance (SAR), is not required for RPP4-mediated resistance in true leaves (Table 2). Furthermore, foliar application of the salicylic acid analogue benzothiadiazole (BTH; Lawton et al., 1996), induced SAR in wild-type Col-0 plants and in eds1, pad4, pbs2 and pbs3 mutants (but not in npr1 mutants; Table 2). These results indicate that SAR operates downstream or independently of the RPP4 pathway, and suggest a dual role for salicylic acid in both an NPR1-dependent pathway leading to SAR and the RPP4 pathway defined by EDS1, PAD4, PBS2 and PBS3 (Figure 6).

In conclusion, molecular isolation of the RPP4 gene has allowed us to scrutinize the requirements of a TIR-NB-LRR protein for a broad range of signalling components. Our analyses reveal that, while RPP4 function requires salicylic acid accumulation and signals most strongly through EDS1, PAD4, PBS2, and PBS3, other mechanisms regulated by NPR1 and NDR1 impinge on RPP4 function, suggesting a degree of cross-talk in R gene-mediated plant defence signalling.

Experimental procedures

Mapping of RPP4

RPP4 for resistance to P. parasitica race Emoy2 has been mapped on a 23 cm segment of chromosome 4 in a cross between Col-5 and Niederzenz (Nd-1; Tör et al., 1994). To confirm and refine the position of RPP4, we used an F2 between Col-0 and Wassilewskija (Ws-0). Ws-0 is resistant to P.p. Emoy2, but is susceptible to its natural recombinant variants P.p. Emwa1 and Emco5 (Table 4; McDowell et al., 1998). P.p. race Emwa1 was therefore used to determine the RPP4 genotypes of Col-0 × Ws-0 F2 seedlings and their F3 progenies. DNA was isolated from 217 F2 seedlings, and analysed for P. parasitica Emwa1 resistance and linkage to various molecular markers (Figure 1a). Thirty-four plants with recombinations between the ARA and AG co-dominant cleaved amplified polymorphic sequences (CAPS; Konieczny and Ausubel, 1993) were selected and analysed further (Figure 1a). Four recombinations proximal to RPP4 were detected with the single sequence-length polymorphism (SSLP) VRN2, and 10 and two recombinations were found distal to RPP4 with the CAPS markers T6-3 with FCA8-2, respectively. No recombinations were found with markers G4539 and Col-A (Figure 1a). Details on the CAPS and SSLP markers are available upon request.

Cloning of Col-0 RPP5 homologues and complementation of RPP4 function

A pCD04541 binary cosmid library containing approximately four Col-0 genome equivalents (courtesy of Ian Bancroft, John Innes Centre, Norwich, UK) was screened with a radiolabelled 5 kb NcoI/EcoRI genomic RPP5 fragment (Parker et al., 1997). DNA gel-blot analysis using HindIII and the radiolabelled RPP5 probe confirmed that the cosmid inserts contained RPP5-homologous sequences of predicted sizes. By restriction mapping (EcoRI, BamHI, BamHI/SalI), 36 strong hybridizing cosmids were analysed and placed on the ≈80 kb Col-0 RPP5 region (GenBank accession Z97342); 10 cosmids were selected that contained the six (near) full-length RPP5 homologues (Col-A, Col-B, Col-D, Col-E, Col-F and Col-G). The authenticity of the DNA inserts and the precise genomic map locations of the 10 selected cosmids were determined by direct DNA sequencing of the cosmid border fragments.

To test for complementation of P.p. Emoy2 and Emwa1 resistance, the 10 binary cosmids were transferred to the CW-84 recipient line by Agrobacterium tumefaciens strain GV3101-mediated transformation (Clough and Bent, 1998). The possible occurrence of recombination in A. tumefaciens was verified because most cosmid inserts contained multiple RPP5-homologous sequences. Cosmid DNA was re-isolated from A. tumefaciens, and the EcoR1 restriction patterns were analysed in parallel with those from cosmid DNA isolated from E. coli DH5α (rec; Gibco/BRL, Paisley, UK), but no evidence for recombination-induced rearrangements was found. The CW-84 progenitor was selected from a Col-gl × Ws-0 F2 population for the absence of RPP1, RPP2 and RPP4, and is therefore universally susceptible to the P.p. races Cala2, Emoy2, Emwa1 and Noco2 (described by Botella et al., 1998; Tables 1 and 3). At least 12 independent CW-84 T1 transformants of each of the 10 cosmids were selected for kanamycin resistance (200 mg l−1), self-pollinated, and 30–60 T2 seedlings were tested in at least two separate experiments for resistance to several P.p. races (Table 1).

Arabidopsis thaliana and Peronospora parasitica

Arabidopsis growth and Peronospora parasitica propagation, and infections of cotyledons and true leaves, including lactophenol–trypan blue staining, were as described previously (Holub et al., 1994; Parker et al., 1996; Parker et al., 1997). 1-week-old (cotyledons) to 2-week-old (true leaves) seedlings were sprayed with P. parasitica conidiospores (4 × 104 ml−1) and incubated at 16°C and high humidity. After 5–7 days the seedlings were analysed for asexual sporulation using a 10× magnifying glass, and by staining with lactophenol–trypan blue to microscopically (100×) assess hyphal growth using phase-contrast optics. Following destaining with chloral hydrate, the lactophenol–trypan blue dye is retained in P. parasitica cells and in dead or dying plant cells (including vascular xylem). Induction of systemic acquired resistance (SAR) by foliar application of the salicylic acid analogue benzothiadiazole (BTH) was done as described (30 µm BTH in water; Lawton et al., 1996). Plants were infected with P. parasitica 2–4 days after BTH treatment. All P. parasitica infections were performed in at least two replicates with 30–60 seedlings.

The following Col-0 mutants and transgenic lines were obtained from our colleagues: Col-0 npr1-1 (X. Dong, Duke University, Durham, USA); Col-0 ndr1-1 (B. Staskawicz, University of California, Berkeley, CA, USA); Col-0 pbs2, Col-0 pbs3, Col-0 rps5-1 and an F2 between Col-0 pbs2 and Ler (R. Innes, Indiana University, Bloomington, IN, USA); Col-0 NahG (J.-P. Metraux, University of Fribourg, Fribourg, Switzerland); and Col-0 pad4-1 (J. Glazebrook, University of Maryland, MD, USA). The Ler pad4-2 and rpp5-2 mutants were identified in screens for loss of P. parasitica Noco2 resistance (Louise N. Frost and J.E.P., unpublished). The Col-0 pbs2 and the Col-0 npr1-1 mutations were introduced into an RPP5 background by crossing to Ler and selecting F2 individuals with flanking markers (data available on request). The Ler eds1-2 mutation was introduced into an RPP4 background by crossing to Col-gl (described by Aarts et al., 1998).

Nucleic acid manipulations and yeast two-hybrid analysis

Plant and bacterial DNA isolations, PCR, DNA sequencing, cosmid and plasmid manipulations, restriction digests and yeast two-hybrid protocols were as described previously (Botella et al., 1998; Noël et al., 1999; Parker et al., 1997; Van der Biezen et al., 2000). Direct DNA sequencing of the cosmid inserts was done using standard protocols with 4 µg cosmid DNA isolated from E. coli DH5α (Gibco/BRL) with Tip20 columns (Qiagen, Chatsworth, CA, USA) and precipitated with polyethylene glycol (PEG mw4000, Sigma, Dorset, UK).


We thank Eric Holub (Horticultural Research Institute, Wellesbourne, UK) for providing Peronospora parasitica races Cala2, Emco5, Emoy2 and Emwa1, and Ian Bancroft (John Innes Centre, Norwich, UK) for the Col-0 binary cosmid library. We are also grateful to our colleagues Xinnian Dong, Jane Glazebrook, Roger Innes, Jean-Pierre Metraux and Brian Staskawicz for various Arabidopsis lines. E.A.v.d.B. was supported by Zeneca, and the Sainsbury Laboratory is funded by the Gatsby Charitable Foundation.

The GenBank accession number of RPP4 is AF440696