Report

Host-Parasite Coevolutionary Conflict Between Arabidopsis and Downy Mildew

See allHide authors and affiliations

Science  10 Dec 2004:
Vol. 306, Issue 5703, pp. 1957-1960
DOI: 10.1126/science.1104022

Abstract

Plants are constantly exposed to attack by an array of diverse pathogens but lack a somatically adaptive immune system. In spite of this, natural plant populations do not often suffer destructive disease epidemics. Elucidating how allelic diversity within plant genes that function to detect pathogens (resistance genes) counteracts changing structures of pathogen genes required for host invasion (pathogenicity effectors) is critical to our understanding of the dynamics of natural plant populations. The RPP13 resistance gene is the most polymorphic gene analyzed to date in the model plant Arabidopsis thaliana. Here we report the cloning of the avirulence gene, ATR13, that triggers RPP13-mediated resistance, and we show that it too exhibits extreme levels of amino acid polymorphism. Evidence of diversifying selection visible in both components suggests that the host and pathogen may be locked in a coevolutionary conflict at these loci, where attempts to evade host resistance by the pathogen are matched by the development of new detection capabilities by the host.

Disease resistance in plants is a complex process that provides many potential barriers to pathogen invasion. Among the plant's defense arsenal are the disease resistance (R) genes, whose products trigger defense responses, such as localized host cell death, when challenged with pathogen isolates carrying matching avirulence genes (1). The largest class of R genes encodes proteins containing intra- or extracellular leucine-rich repeat (LRR) domains. LRR domains have been implicated in protein:protein interactions (2). However, direct interaction between an avirulence protein and its cognate R protein has been demonstrated in only a few host/pathogen systems (3, 4).

Avirulence genes have been cloned from the fungal plant pathogens Cladosporium fulvum (57), Magnaporthe grisea (8), and Melampsora lini (9), but apart from the chitin-binding capacity of the Avr4 protein from C. fulvum (10), their roles in pathogenicity are unknown. We have recently shown the RPP13 (Recognition of Peronospora parasitica 13) resistance gene from Arabidopsis thaliana to be the most polymorphic gene so far analyzed in this species (11). RPP13 encodes a CC:NB:LRR (coiled coil: nucleotide binding site:leucine rich repeat) protein, predicted to be cytoplasmically located, and the extreme variability of the protein was shown to reside within the LRR domain (11, 12). This is consistent with the LRR domain experiencing diversifying selection. One selective agent could be a pathogen species exhibiting comparable levels of polymorphism in the avirulence protein detected via RPP13. From the plant's perspective, there are two basic outcomes of a coevolutionary conflict: either a selective sweep in which a single allele of a resistance gene reaches high frequency in the plant population or balancing selection, in which a diverse cohort of resistance gene alleles is stably maintained (13). The large number (19 among 24 Arabidopsis accessions) of diverse alleles present at the RPP13 locus implies that it is subject to balancing selection (11). Haldane's theory (14) suggests that coevolution of host and pathogen could lead to the maintenance of variation in both organisms. The interaction between Arabidopsis and the biotrophic oomycete Hyaloperonospora parasitica (formally Peronospora parasitica) is an excellent system in which to study such coevolution because both organisms coexist in extensive naturally occurring populations (15). Therefore, concomitant with extreme RPP13 gene diversity, we hypothesize that balancing selection on the pathogen gene products recognized by these R genes [ATR13 (Arabidopsis thaliana recognised 13)] would also result in the maintenance of a highly polymorphic population of ATR13 alleles. Here we report the cloning of ATR13 and show that it is indeed under intense diversifying selection consistent with host/parasite conflict occurring between these two species. Both ATR13 and RPP13 are subject to balancing selection.

We previously isolated a range of H. parasitica genes [Ppat (Peronospora parasitica in Arabidopsis thaliana)] that were up-regulated on infection of Arabidopsis (16). Mapping of the Ppat sequences among 206 F2 progeny of a cross between H. parasitica isolates Maks9 (predicted to contain ATR13) and Emoy2 (predicted to contain atr13) revealed cosegregation of a single-copy gene, Ppat17, and ATR13 (17). Therefore, Ppat17 was an ATR13 candidate.

Because no mechanism of genetic transformation has been established for H. parasitica, we developed a functional assay for ATR13 recognition based on a biolistic approach. Three models could be proposed for the role of ATR13 in the elicitation of plant cell death: (i) the presence of the ATR13 protein alone is sufficient; (ii) ATR13 acts in concert with monomorphic H. parasitica proteins; or (iii) ATR13 is an enzyme that produces a pathogen product, which triggers the hypersensitive reaction as in the case of avrD (18). Bombardment of Arabidopsis leaves with a plasmid carrying the bacterial uidA gene (GUS) fused to the 35S promoter results in blue-stained cells, in the presence of the substrate X-Gluc. (17). If model (i) is correct, then co-bombardment of Arabidopsis leaves expressing RPP13 with the 35S::GUS plasmid and another carrying a 35S::ATR13 fusion would result in cell death and consequently no GUS expression. Hence, we fused Ppat17 to the 35S promoter and co-bombarded Arabidopsis accession Columbia, which contains an allele of RPP13 that does not recognize Maks9 or Emoy2, and a Columbia transgenic line (Col5::RPP13-Nd) carrying the RPP13 allele from the Niederzenz accession, which enables isolate-specific recognition of Maks9 but not Emoy2 (12, 19). In an experiment using 35S::GUS and a control plasmid, large numbers of blue-stained cells were seen in both types of plant material (Fig. 1, A and H). However, when 35S::GUS and 35S::Ppat17-Maks9 were co-bombarded, results similar to those with the control (648 blue-stained cells) were seen in Columbia, but this number was significantly reduced (4.4 blue-stained cells) in our Col5::RPP13-Nd line (Fig. 1, B and I). As a further test, we repeated the experiment using the Emoy2 allele of Ppat17, which would not be predicted to elicit a response from RPP13-Nd. In this experiment, the number of blue-stained cells was similar to the control in both plant lines (Fig. 1, C and J). We conclude that Ppat17 is ATR13, and the protein it encodes is sufficient to trigger RPP13-dependent resistance.

Fig. 1.

Biolistic analysis of ATR13 alleles. (A to G) 8-week-old Col-5 leaves; (H to N) 8-week-old Col-5::RPP13-Nd leaves. (A) and (H) were bombarded with the control plasmid, pK2GW7, and the 35S::GUS plasmid; (B) and (I) were bombarded with the ATR13-Maks9 plasmid (without signal peptide) and the 35S::GUS plasmid; (C) and (J) were bombarded with the ATR13-Emoy2 plasmid (without signal peptide) and the 35S::GUS plasmid; (D) and (K) were bombarded with the ATR13-Aswa1 plasmid (with signal peptide) and the 35S::GUS plasmid; (E) and (L) were bombarded with the ATR13-Emco5 plasmid (with signal peptide) and the 35S::GUS plasmid; (F) and (M) were bombarded with the ATR13-Goco1 plasmid (with signal peptide) and the 35S::GUS plasmid; and (G) and (N) were bombarded with the ATR13-Hind4 plasmid (with signal peptide) and the 35S::GUS plasmid. Leaves were stained for β-glucuronidase activity and cleared with methanol. Numbers represent average numbers of blue-stained cells over five replicates. Scale bars, 2.5 mm.

To confirm the function of ATR13 in vivo, we carried out in planta expression assays. We fused the ATR13 Maks9 and Emoy2 alleles to a glucocorticoid-inducible plant promoter and generated transgenic Columbia plants and HRI3879, a selected recombinant inbred line containing RPP13-Nd, plants (17). In the presence of RPP13-Nd, the ATR13-Maks9 gene caused wilting within 6 hours of dexamethasone application and death of the whole plant within 24 hours (Fig. 2, A and B). Induction of the Maks9 allele caused no phenotypic change in Columbia (Fig. 2, C and D). To confirm that ATR13-Maks9 protein was produced in these plants, they were crossed to Col5::RPP13-Nd and, as expected, all F1 progeny tested died on induction of ATR13-Maks9 (20). Induction of the Emoy2 ATR13 allele in the presence or absence of RPP13-Nd resulted in no change in plant phenotype (20). These data imply that RPP13 is expressed in a wide range of above-ground plant cell types and confirm the allele-specific nature of the interaction between RPP13 and ATR13-Maks9.

Fig. 2.

Induced expression of ATR13-Maks9 triggers a total cell death phenotype specific to Arabidopsis plants containing RPP13-Nd. Arabidopsis plant lines were transformed with ATR13-Maks9 (minus signal peptide sequence) under the control of a dexamethasone-inducible promoter (17). Pictures were taken 6 hours (A and C) and 24 hours (B and D) after dexamethasone application. Representative T3 homozygous transgenic HRI3879 plants (RPP13-Nd) (A) and (B) and Columbia (RPP13-Col) plants (C) and (D) are shown. Note the drooping inflorescence (A) and desiccated nature of the leaves in the transgenic HRI3879 plant (B). We observed similar responses when plants were transformed with full-length ATR13-Maks9.

ATR13-Maks9 encodes a 187–amino acid protein that shows no significant homology (BLASTP) (21) to other proteins but appears to have clear domain structures. ATR13-Maks9 has a heptad leucine/isoleucine repeat motif and, although reminiscent of coiled-coil domains that are involved in protein-protein interactions, this ATR13 domain is not predicted to lie within an α-helical structure (Fig. 3). An imperfect direct repeat of 4 × 11 amino acids lies between residues 93 and 136 and is followed by a C-terminal region within which no specific structures can be identified (Fig. 3). The program SignalP (22) reveals a high (P = 0.98) likelihood of a signal peptide being encoded at the N terminus with cleavage after the 19th amino acid (Fig. 3). This suggests that ATR13-Maks9 is secreted from H. parasitica during its growth in planta, which is consistent with it being exposed to and entering the plant cell where it could interact with RPP13-Nd.

Fig. 3.

Alignment of predicted proteins encoded by ATR13 alleles generated with Vector NTI. Dashes indicate amino acids identical to ATR13-Maks9. Amino acid residues differing from ATR13-Maks9 are shown. In the N-terminal region, the predicted signal peptide is boxed. In the direct repeat region, the repeats are shown by arrows. ATR13 from isolates Maks9, Aswa1, Emco5, and Goco1 triggers an RPP13-Nd–dependent resistance response, but ATR13 from isolates Emoy2 and Hind4 does not.

The presence of an ATR13 signal peptide made no difference to the results obtained in the biolistic and in planta assays, and could be explained as a consequence of high levels of gene expression resulting in aberrant processing of ATR13 by the host's signal peptide recognition complex (20). Successful recognition of ATR13-Maks9 expressed without a signal peptide is consistent with its recognition by the intracellularly located RPP13-Nd, implying that ATR13-Maks9 is imported into the plant cell by an unidentified mechanism. Bacterial plant pathogens such as Pseudomonas syringae use the highly conserved type III secretion apparatus to transport their effector proteins across the plant plasma membrane (23), but an equivalent system has yet to be described in fungal or oomycete pathogens. The AvrL567 gene family from M. lini encodes small potentially secreted proteins that have been shown to specifically trigger R gene–dependent cell death in flax lines carrying the cytoplasmically located L5, L6, and L7 resistance proteins (9). Both H. parasitica and M. lini possess specialized feeding structures called haustoria that form an intimate association with the plant plasma membrane (24), and potentially these are the sites at which these pathogens traffic their pathogenicity effectors.

RPP13-Nd initiates resistance reactions to the Aswa1, Emco5, and Goco1 isolates of H. parasitica in addition to Maks9, but not to Hind4 or Emoy2. To determine whether ATR13 is central to this resistance, we cloned ATR13 alleles from these additional isolates and tested their function using the biolistic assay. ATR13-Aswa1, ATR13-Emco5, and ATR13-Goco1 all elicited an RPP13-Nd–dependent cell death response equivalent to that seen with ATR13-Maks9 but ATR13-Hind4 did not (Fig. 1, D to G and K to N). DNA sequence analysis revealed that all isolates carried a different ATR13 allele than Maks9, but that of Emco5 and Goco1 were identical to each other (25). However, the overall structure of the predicted ATR13 proteins was retained (Fig. 3). Several amino acids varied within the heptad repeat region, but the heptad motif itself was conserved, suggesting a possible functional significance. Surprisingly, ATR13 encoded by the alleles from Aswa1, Emco5, and Goco1 contained only one 11–amino acid repeat unit, indicating that repeats 1, 3, and 4 are dispensable for an RPP13-Nd–mediated resistance response. The AvrBs3 avirulence gene family from Xanthomonas campestris pv. vesicatoria encodes proteins with a variable number of a 34–amino acid repeat motif, which was shown to determine specificity in its interaction with different plant R genes (26). In contrast, a repeated domain within ATR13 is not required for recognition by RPP13-Nd. However, we cannot preclude the possibility that the repeats have a role in determining recognition specificity via other R genes.

The Maks9 and Emoy2 alleles have identical DNA sequences throughout the N-terminal, heptad repeat, and direct repeat regions and only a single nucleotide polymorphism in 218 bases 5′ to translation initiation (25). This stretch of shared sequence identity between these two alleles over the first three-quarters of the protein, followed by a region of dissimilarity, uncovers two biologically relevant features. First, it reveals that the C-terminal portion of these two proteins is the region causing differential recognition by the RPP13-Nd allele. Second, it suggests that recombination has played a role in the evolutionary history of the ATR13 gene. The inference of recombination is supported by permutation analysis, which detected a significantly long stretch of sequence identity between these two alleles (P = 0.008) (27).

RPP13 has evolved under intense diversifying selection (11) and as such offers a stark contrast to the related Arabidopsis RPM1 gene where, presumably, invariant Pseudomonas effector proteins AvrRPM1 and AvrB do not appear to have driven the evolution of alternative RPM1 alleles (28, 29). If the evolution of RPP13 were driven by its interaction with H. parasitica, then one would expect to see a similar evolutionary pattern in the matching avirulence gene. Among the five ATR13 alleles, there are 26 nonsynonymous, two synonymous, and two indel polymorphisms. Based on a total of 351.5 nonsynonymous and 113.5 synonymous sites at ATR13 (30), this represents a significant excess of nonsynonymous polymorphism relative to the neutral expectation (X2 = 4.48, P = 0.034) and indicates selective maintenance of amino acid polymorphism at this locus. Amino acid polymorphism at ATR13 is not limited to differences between alleles that are recognized by RPP13-Nd and those that are not. Only eight amino acid differences are fixed between the two phenotypically distinct classes of alleles. There are nine amino acid and two indel polymorphisms among the three alleles recognized by RPP13-Nd, and four amino acid differences between the two alleles not recognized by RPP13-Nd. This is reminiscent of the situation in M. lini, where two paralogs of the AvrL567 gene, exhibiting significant amino acid variation, show differential recognition and response by the L5, L6, and L7 flax resistance genes (9). However, ATR13 alleles, showing not only extensive amino acid variation but also deletions of repeated domains, were equally effective in triggering resistance via RPP13-Nd. The high level of amino acid variation among alleles that are recognized by RPP13-Nd may indicate that these variants are selectively favored in H. parasitica parasitizing host populations not expressing RPP13-Nd. To confirm that not all H. parasitica genes are undergoing an equivalent extreme rate of change, we sequenced Ppat5 from the same H. parasitica isolates. Ppat5 encodes a dnaK-type molecular chaperone (16) and hence is likely to be under different selective pressures as compared to ATR13. DNA sequence analysis of Ppat5 revealed only nine segregating polymorphisms across the 1983–base pair ORF and, in contrast to ATR13, only one of these is a nonsynonymous polymorphism (25).

Our study reveals the RPP13/ATR13 plant/pathogen interaction to be an excellent model for studying the coevolution of resistance and avirulence genes within host and pathogen populations. The high levels of amino acid polymorphism relative to silent polymorphism in both plant and pathogen genes is consistent with a history of balancing selection operating at both loci. Within RPP13, it is the LRR domain that shows diversifying selection, whereas the rest of the gene shows selection for conservation of protein sequence (11, 12). This study shows that the C-terminal domain of ATR13 plays a role in determining the specificity of interaction with RPP13, suggesting a direct interaction with the LRR domain. However, our initial yeast two-hybrid studies have not revealed a direct interaction between RPP13 and ATR13 (31). It is possible that different alleles of RPP13 recognize other pathogen proteins, and variation at this locus could be influenced by additional pathogen interactions, not necessarily limited to H. parasitica. Additionally, ATR13 may be detected by more than one host resistance gene, leading to increased selection for diversity in this protein. ATR13 must have a role in enabling H. parasitica to grow as an obligate biotrophic pathogen on Arabidopsis, and the elucidation of the roles of the observed motifs in planta will add substantially to our understanding of the mechanisms of biotrophic pathogenicity as well as those of host defense.

Supporting Online Material

www.sciencemag.org/cgi/content/full/306/5703/1957/DC1

Materials and Methods

Figs. S1 and S2

References

References and Notes

View Abstract

Navigate This Article