Attack and counter-attack impose strong reciprocal selection on pathogens and hosts, leading to development of arms race evolutionary dynamics. Here we show that Magnaporthe oryzae avirulence gene AVR-Pik and the cognate rice resistance (R) gene Pik are highly variable, with multiple alleles in which DNA replacements cause amino acid changes. There is tight recognition specificity of the AVR-Pik alleles by the various Pik alleles. We found that AVR-Pik physically binds the N-terminal coiled-coil domain of Pik in a yeast two-hybrid assay as well as in an in planta co-immunoprecipitation assay. This binding specificity correlates with the recognition specificity between AVR and R genes. We propose that AVR-Pik and Pik are locked into arms race co-evolution driven by their direct physical interactions.
Antagonistic biotic interactions, including parasitism and predation, impose strong reciprocal selection on living organisms, and are a major driver in evolution. The population geneticist J. B. S. Haldane wrote ‘I want to suggest that the struggle against disease, and particularly infectious disease, has been a very important evolutionary agent…’ (Haldane, 1949). It is now widely accepted that natural selection imposed by pathogens is one of the major causes of organismal evolution; it has even been proposed that the maintenance of sex in hosts may be attributed to host–pathogen interactions (Ebert and Hamilton, 1996; Morran et al., 2011). Between host and pathogen, two types of evolutionary dynamics have been proposed: (i) arms race and (ii) trench warfare (Woolhouse et al., 2002). In arms race dynamics, a pathogen evolves an allele of a virulence gene that enhances its fitness through suppression of host defense. The frequency of such an allele rapidly increases in the pathogen population, due to the fitness advantage it confers to the carrier pathogen, eventually replacing the older allele. In turn, a new allele of a host counter-defense gene that nullifies the effect of the pathogen virulence gene emerges, and its allele frequency increases rapidly to become fixed in the host population. These reciprocal processes are repeated over a long time. In trench warfare dynamics (Stahl et al., 1999), also called Red Queen dynamics (Woolhouse et al., 2002), negative frequency-dependent selection causes the allele frequencies of the corresponding genes of pathogen and host to oscillate over time, resulting in maintenance of the same set of alleles in the population for a longer time. The two types of dynamics leave contrasting patterns of DNA signatures on the genes of involved species: genes involved in arms race dynamics experience recurrent selective sweeps and exhibit rapid evolution, resulting in a larger number of inter-species amino acid replacements and low levels of polymorphism within the species, whereas genes involved in trench warfare co-evolution show a larger amount of within-species polymorphisms. These two models may be applied at the population level as well as the species level, provided that populations are effectively isolated from each other. Population-level analysis appears to be more relevant to pathogen–crop interactions (Terauchi and Yoshida, 2010).
The existence of gene-for-gene relationships between host resistance (R) and pathogen avirulence (AVR) genes between plants and their pathogens is well established (Flor, 1956; Dangl and Jones, 2001). However, studies on co-evolution of R and AVR genes are still limited. A DNA polymorphism study of the Arabidopsis thaliana Rpm1 gene (Stahl et al., 1999) showed that the gene contains two allelic classes (an R class conferring resistance and an S class that does not) that diverged in ancient past, suggestive of long-term balancing selection presumably caused by the trench warfare type of interaction. On the other hand, the RPP13 R gene of Arabidopsis thaliana was shown to harbor an extreme level of amino acid polymorphisms (Rose et al., 2004). This corresponds to the large amount of non-synonymous polymorphisms in ATR13, a cognate AVR gene of RPP13 in Hyaloperonospora parasitica (Allen et al., 2004). These large amounts of non-synonymous polymorphisms of host and pathogen are thought to be maintained by reciprocal selection at the population level by arms race dynamics. A study of flax (Linum usitatissimum) and flax rust (Melampsora lini) interactions is the most thoroughly studied case of arms race co-evolution in plant–pathogen interactions. The flax L locus, encoding a nucleotide binding site/leucine-rich repeat (NBS-LRR) protein, harbors 12 allelic variants with extensive sequence diversity (Ellis et al., 1999). Melampsora lini AvrL567 is an avirulence gene recognized by the L5, L6 and L7 alleles of the flax L gene (Dodds et al., 2004). AvrL567 also has 12 allelic variants in six strains, and its diversity is extreme, with more non-synonymous DNA sequence polymorphisms than synonymous ones. Dodds et al. (2006) showed that L protein and AvrL567 protein physically interact in yeast two-hybrid analysis, and the strength of physical binding recapitulates the interaction specificity observed in planta, indicating that L and AvrL567 are evolving by arms race dynamics through physical interactions at the population level. More recently, the same group analyzed DNA polymorphisms of M. lini AvrP123 and AvrP4 avirulence genes, and suggested that their polymorphisms are maintained by both arms race and trench warfare dynamics (Barrett et al., 2009).
Rice blast caused by the ascomycete fungus Magnaporthe oryzae is the most devastating disease of rice worldwide. Control of blast disease has mainly relied on deployment of resistant (R) genes of rice. So far more than 40 R genes have been named, and ten have been cloned (Ballini et al., 2008; Liu et al., 2010; Terauchi et al., 2011). On the other hand, only six AVR genes of M. oryzae have been cloned to date. ACE1 is an unusual avirulence gene as it encodes an enzyme that is involved in synthesis of a secondary metabolite (Böhnert et al., 2004). AvrPita and AvrPizt have been cloned by map-based cloning (Orbach et al., 2000; Li et al., 2009). Using an association genetics approach, we previously cloned three AVR genes, AVR-Pik/km/kp, AVR-Pia and AVR-Pii (Yoshida et al., 2009). A DNA polymorphism study of these AVRs in rice-infecting field strains showed very contrasting profiles. AVR-Pia and AVR-Pii showed only presence/absence polymorphisms without any nucleotide changes (Yoshida et al., 2009), while AVR-Pizt had a transposon inserted in the promoter region and a single point mutation (Li et al., 2009). AVR-Pita showed very diverse alleles, with complete or partial deletions, transposon insertions, point mutations and even translocations (Orbach et al., 2000; Zhou et al., 2007; Dai et al., 2010; Takahashi et al., 2010; Chuma et al., 2011). Finally, AVR-Pik/km/kp exhibited extensive nucleotide changes leading to amino acid replacements in addition to presence/absence polymorphisms (Yoshida et al., 2009). The rice R gene Pik has been reported to show multiple alleles (Pik, Pikm, Pikp, Piks and Pikh) with differential resistance to M. oryzae races. Hereafter, to avoid confusion in nomenclature, we use ‘Pik’ to indicate the locus name, and ‘Pik*‘to represent the conventional Pik allele. The rice R gene allele Pikm, the AVR-Pikm cognate, has been cloned by Ashikawa et al. (2008) and Zhai et al. (2011). Pikm comprises two protein-coding ORFs, Pikm-1 and Pikm-2, separated by a 2.5 kb non-coding region and oriented in opposite directions. In this paper, we refer to the two ORFs of the Pik gene as Pik-1 and Pik-2. Constanzo and Jia (2010) studied DNA sequence polymorphisms of the Pik locus for Pik-1 and Pik-2 using 15 rice cultivars. They found a large amount of DNA polymorphisms clustered in the Pik-1 region corresponding to the coiled-coil (CC) domain, whereas only a low level of DNA polymorphisms were observed in Pik-2. Furthermore, they found two amino acid residues at positions 229 and 252 in the CC domain that showed extensive polymorphisms, and proposed that these amino acid residues may be used as diagnostic markers for the classification of Pik alleles: Pik*, Pikm, Pikp, Piks and Pikh.
We are interested in the co-evolutionary dynamics of M. oryzae AVR genes and rice R genes. Here we show that M. oryzae AVR-Pik alleles and rice Pik alleles exhibit high levels of amino acid changes. We show that a subset of combinations of AVR-Pik and the CC domain of Pik-1 physically bind, and this binding specificity mainly determines the recognition specificity of AVR-Pik alleles by Pik alleles. We propose that this physical interaction results in arms race co-evolution between AVR-Pik and Pik.
DNA polymorphisms in M. oryzae AVR-Pik and rice Pik
M. oryzae AVR-Pik is a 113 amino acid protein with a 21 amino acid signal peptide (Figure 1a) (Yoshida et al., 2009). Five alleles of AVR-Pik (AVR-Pik-A, B, C, D and E) were identified in 21 isolates of M. oryzae from Japan. The AVR-Pik DNA sequence is highly variable with a nucleotide diversity (Nei, 1987) of 7.1 × 10−3, which is two orders higher than the mean value for the entire genome (8.2 × 10−5) as revealed by EcoTILLING (Yoshida et al., 2009). The five AVR-Pik alleles are different from each other by a total of five DNA substitutions, all of which cause amino acid changes (Figure 1b and Figure S1). We found a paralog of AVR-Pik, pex75, in the genome of M. oryzae (Figure S1, and see Supplementary Dataset 2 in Yoshida et al., 2009). AVR-Pik-D shared identical amino acid residues with pex75 for the five positions that segregate among the AVR-Pik alleles (Figure 1b). Phylogenetic analysis of the AVR-Pik amino acid polymorphisms using the pex75 sequence as an outgroup (Nei and Kumar, 2000) revealed the consensus maximum-parsimony tree (Figure 1c). This tree suggests that the AVR-Pik-D allele is most likely the ancestral allele, from which the AVR-Pik-E, -C, -A and -B alleles were derived.
To assess the worldwide distribution of AVR-Pik alleles in rice-infecting M. oryzae, we performed PCR amplification and DNA sequencing of AVR-Pik genes from 39 isolates collected from Asia (n = 23), Africa (n = 7), Europe (n = 3) and America (n = 6) (Figure 1d and Table S1). These isolates were chosen to be representative of M. oryzae genetic diversity (Tharreau et al., 2009; Saleh et al., 2012). AVR-Pik-B, which was found in an isolate from Japan (isolate 9505-3; Yoshida et al., 2009) was not identified in the 39 isolates. AVR-Pik-D was the most frequent allele (15 of 39). The three other alleles showed similar frequencies (7–9 of 39). The AVR-Pik-A, -C, -D and -E alleles were distributed worldwide, with the few exceptions probably being explained by differences in sample size between continents.
The rice Pik gene comprises two adjacent inversely oriented ORFs encoding NBS-LRR proteins (Pik-1 and Pik-2) (Figure 2) (Ashikawa et al., 2008). DNA polymorphisms of Pik-1 and Pik-2 were studied for the C-terminal LRR region by Ashikawa et al. (2010) and for the N-terminal CC domain region by Constanzo and Jia (2010). The latter authors identified two residues (229 and 252) in the N-terminal CC region of Pik-1 that are highly variable and useful for diagnosis of Pik alleles (Constanzo and Jia, 2010). This observation suggests a possibility that this region may be involved in recognition of AVR-Pik alleles. To assess the levels of DNA polymorphisms of Pik corresponding to the N-terminal CC domain region using a larger number of rice accessions, we performed a DNA sequencing analysis of an 800 bp region of Pik-1 and Pik-2 for a total of 37 Oryza accessions collected from around the world,comprising 33 accessions of O. sativa and four accessions of Oryza rufipogon (the wild progenitor species of O. sativa) (Table S2). The O. sativa accessions used were randomly selected from a worldwide rice collection (Kojima et al., 2005) that covers O. sativa genetic diversity. Phylogenetic analysis of the aligned sequences (Figure S2) revealed that there are two major clades (Group 1 and 2) in both Pik-1 and Pik-2. These clades do not correspond to taxonomic separation of the studied O. sativa accessions into ssp. japonica and ssp. indica. There is a perfect linkage disequilibrium between Pik-1 and Pik-2: all the varieties carrying an allelic variant of Pik-1 belonging to Group 1 carry an allelic variant of Pik-2 belonging to Group 1, and the same for Group 2 allelic variants (Figure 2). The well-characterized Pik alleles (Pik*, Pikm, Pikp, Piks and Pikh) belong to Group 1 for Pik-1 and Pik-2. Amino acid sequence comparison of the CC domain (Figure S2) confirmed that the two sites (229 and 252) are segregating for three or four amino acid residues (Q, E or K at position 229; P, H, D or R at position 252), as reported by Constanzo and Jia (2010). If we assume that the Pik allelic state of rice accessions can be predicted by the two diagnostic sites, the largest number of studied rice accessions harbor Piks (13 accessions), followed by Pikp (six accessions), Pik* (three accessions), Pikm (three accessions) and Pikh (one accession) (Figure 2). Within Group 1 of Pik-1, there are four clades with high bootstrap values: a clade for Piks, Pikh and Pikm, one for Pikp, one for Kanto51-type Pik* and one for Toto-type Pik* (Figure 2). For both Pik-1 and Pik-2, the divergence between Group 1 and 2 accounts for the largest part of DNA variation for both synonymous and non-synonymous sites (Table 1). Pik-1 contains substantial DNA polymorphisms within Group 1 (πsyn = 0.046; πnon-syn = 0.035) and Group 2 (πsyn = 0.013; πnon-syn = 0.033), but Pik-2 lacks polymorphisms in Group 1 (πsyn = 0; πnon-syn = 0.001). Tajima’s test of neutral mutation hypothesis applied to all the samples for Pik-1 and Pik-2 showed statistically significant (P < 0.05) positive values for Tajima’s D (Tajima, 1989) (Table 1). This result suggests either that balancing selection is imposed on this locus, or recent merging of two divergent populations. As there is no indication of recent merging of O. sativa populations, we hypothesize that this positive Tajima’s D is probably caused by balancing selection imposed on this locus or the region linked to this locus.
Table 1. Summary of nucleotide polymorphisms for the regions encoding the CC domain of Pik-1 and Pik-2 in Oryza sativa and Oryza rufipogon
Number of samples
D: Tajima’s D.
†P < 0.10, *P < 0.05, **P < 0.01, ***P < 0.001; NS, not significant; NA, cannot be applied because of no polymorphic site.
aO. rufipogon accession W1976 was heterogyzous for Group1 and Group2 allelic variants of Pik, so that the total number of samples (sequences) was 5.
AVR-Pik alleles are specifically recognized by different Pik alleles
Maintenance of multiple alleles of AVR-Pik in M. oryzae and Pik in rice populations prompted us to study whether each allele of AVR-Pik is recognized by specific Pik alleles. To address this possibility, we created isogenic lines of M. oryzae harboring different AVR-Pik alleles. The M. oryzae isolate Sasa2 does not harbor AVR-Pik, so it can infect rice cultivars with any Pik alleles. We transformed Sasa2 individually with constructs harboring four alleles of AVR-Pik: AVR-Pik-D, -E, -A and -C. As AVR-Pik-B occurred only in one isolate from Japan, we did not include it in the analysis. All alleles were driven by the native promoter of the AVR-Pik-D allele. Each transgenic M. oryzae line was used for inoculation of four rice cultivars harboring different alleles of Pik (Nipponbare, Pik−; K60, Pikp; Kanto51, Pik*; Tsuyuake, Pikm) (Figure 3). Nipponbare lacks known Pik alleles; however, it contains a Pik orthologous sequence, thus the Pik ortholog of this cultivar is indicated as Pik−. As expected, the cultivar Nipponbare without a known Pik allele (Pik−) was susceptible to all four isolates harboring the AVR-Pik-D, -E, -A or -C alleles, respectively. Cultivar K60 containing the Pikp R gene was resistant to M. oryzae harboring AVR-Pik-D but susceptible to M. oryzae harboring AVR-Pik-E, -A or -C. Cultivar Kanto51 containing the Pik* R gene was resistant to M. oryzae harboring AVR-Pik-D or -E transgenes, but susceptible to M. oryzae harboring AVR-Pik-A or -C. It is notable that the reaction of Pik* to AVR-Pik-A was not clear cut: disease lesions develop, but the speed of lesion development was slower than in other compatible interactions, and thus this reaction is indicated as moderate susceptibility (Figure 3). Finally, cultivar Tsuyuake containing the Pikm R gene was resistant to M. oryzae harboring AVR-Pik-D, -E or -A, but susceptible to that harboring AVR-Pik-C. These results suggest that the three rice Pik alleles Pikp, Pik* and Pikm have differential recognition specificity towards the four M. oryzae alleles (AVR-Pik-D, -E, -A and -C) that differ from each other by a maximum of four amino acid changes (Figure 1b). We also tested the recognition specificity of two other known Pik alleles, Piks and Pikh, against the four AVR-Pik alleles (Figure S3). The recognition specificities of Piks and Pikp were identical, and those of Pikh and Pikm were identical. Therefore, for the rest of study we focus only on the three Pik alleles Pikp, Pik* and Pikm and the Pik ortholog in Nipponbare (Pik−). Amino acid sequence comparison of the CC domain revealed that Pikh and Pikm are very similar, whereas Piks and Pikp are distantly related. We attempted to find common amino acid residues shared only by Piks and Pikp, but could not identify any such residues. At present, it is difficult to infer how Piks and Pikp share the same recognition specificity (Figure S2).
Recognition specificity of AVR-Pik alleles by Pik alleles is determined by strength of the hypersensitive response (HR) cell death
Our blast inoculation assay suggested that M. oryzae AVR-Pik alleles are specifically recognized by different Pik alleles. To assess whether this recognition involves hypersensitive response (HR) cell death, we performed rice cell death assays by transient over-expression of M. oryzae AVR-Pik alleles in rice protoplasts by transfection using electroporation (Figure 4a) (Yoshida et al., 2009). In this assay, we monitored the viability of rice cells by the activity of firefly luciferase, the gene for which is transfected into rice protoplasts together with AVR-Pik. To drive expression of AVR-Pik alleles and the luciferase gene, we used the strong maize ubiquitin promoter (Christensen and Quail, 1996). If the rice cells harbor an R gene cognate of the AVR-Pik allele leading to HR cell death, we see a reduction in luciferase activity. Luciferase activity was monitored by measuring the luminescence 40 h after transfection, by adding luciferin and ATP to the cell extracts (Figure 4a). No HR cell death was observed in Nipponbare rice cells (Pik−) without a known Pik allele after transfection with all four AVR-Pik alleles: AVR-Pik-D, -E, -A and -C (Figure 4b). In cultivar K60 containing Pikp, HR cell death occurred only when AVR-Pik-D was tranfected, in agreement with the results of the blast inoculation assay (Figure 3). In cultivar Kanto51 containing Pik*, AVR-Pik-D caused HR cell death but AVR-Pik-C did not, consistent with the results of the inoculation assay (Figure 3). Transfection of Kanto51 (Pik*) with two other alleles, AVR-Pik-E and -A, resulted in an intermediate level of HR cell death (Figure 4b), although only the former interaction supported incompatibility in the inoculation assay (Figure 3). The HR spectrum of the cultivar Tsuyuake with Pikm is similar to that of Kanto51 (Pik*): AVR-Pik-D triggered strong HR, AVR-Pik-C caused no significant HR, and AVR-Pik-E and -A resulted in an intermediate HR (Figure 4b). However, AVR-Pik-D, -E and -A all caused incompatibility in the inoculation assay (Figure 3). Therefore, there are variable cases in which an intermediate level of the HR supports incompatibility (Pik*/AVR-Pik-E, Pikm/AVR-Pik-E and Pikm/AVR-Pik-A) or does not (Pik*/AVR-Pik-A). Apart from these cases, the results of M. oryzae inoculation assay and protoplast HR assay conformed with each other, suggesting that incompatibility is mainly determined by the HR reaction.
AVR-Pik-D and the CC domain of Pik-1 physically interact
AVR-Pik has five alleles that differ from each other by DNA changes corresponding to one to four amino acid replacements (Figure 1b). Pik has five named alleles (Pik*, Pikp, Pikm, Piks and Pikh) that differ by amino acid replacements in the CC domain (Figures 2 and S2), and can be diagnosed by the amino acids in the two sites 229 and 252 (Constanzo and Jia, 2010). Amino acid changes in AVR-Pik determine the recognition by different rice Pik alleles (Figures 3 and 4). The observed high levels of amino acid polymorphisms and the recognition specificity between AVR and R can most easily be explained if AVR and R proteins physically interact. To test this possibility, we performed a yeast two-hybrid assay to study direct binding between AVR-Pik and Pik.
As AVR-Pik-D is recognized by Pikm (Figures 3 and 4), consisting of the two ORFs Pikm-1 and Pikm-2, we first tested physical interactions between AVR-Pik-D and Pikm-1 and between AVR-Pik-D and Pikm-2. As shown in Figure 5(a), we observed interaction between full-length Pikm-1 and AVR-Pik-D when the former was used as bait and the latter as prey, but no interaction was detected between full-length Pikm-2 and AVR-Pik-D. Next we used the CC domain of Pikm-1 to test the interaction with AVR-Pik-D, and found that it indeed interacts with AVR-Pik-D (Figure 5a). If we swapped bait and prey, we did not observe binding between full-length Pikm-1 and AVR-Pik-D (Figure 5b). However, the CC domain of Pikm-1 used as prey interacted with AVR-Pik-D used as bait (Figure 5b), suggesting that AVR-Pik-D does indeed physically bind the CC domain of Pikm-1. The LRR (Pikm-1-LRR) and NBS (Pikm-1-NBS) regions did not interact with AVR-Pik-D (Figure 5b). Expression of all bait and prey proteins was confirmed by Western blot analysis (Figure S4).
To test whether the molecular interaction between Pikm-1 and AVR-Pik-D is observed in planta, we performed a co-immunoprecipitation assay in Nicotiana benthamiana leaves. The full-length Pikm-1 protein, and the CC domain only and the NBS-LRR domain only of Pikm-1 were tagged with the 3xFLAG epitope, resulting in 3xFL–Pikm-1-full, 3xFL–Pikm-1-CC, and 3xFL–Pikm-1-NBS-LRR, respectively, and AVR-Pik-D was tagged with the HA (haemaglutinin) epitope to generate AVR-Pik-D–HA. These proteins were transiently over-expressed in N. benthamiana leaves by agroinfiltration (Win et al., 2011). As shown in Figure 5(c), all proteins were successfully expressed in planta. When immunoprecipitation of AVR-Pik-D–HA was performed using anti-HA antibody, 3xFL–Pikm-1-full and 3xFL–Pikm-1-CC were co-immunoprecipitated as detected by the anti-FLAG antibody, but 3xFL–Pikm-1-NBS-LRR was not (Figure 5c). These observations confirm the result obtained by the yeast two-hybrid assay, and demonstrate that the CC domain of Pikm-1 and AVR-Pik-D physically interact in planta.
Binding specificity between Pik-1 and AVR-Pik determines the recognition specificity between Pik and AVR-Pik
Next, we tested binding specificity between the full-length products of three Pik-1 alleles (Pikp-1, Pik*-1 and Pikm-1) and the Nipponbare Pik-1 ortholog Pik−, and the protein products of four AVR-Pik alleles (AVR-Pik-D, -E, -A and -C) using a yeast two-hybrid assay (Figure 6a). Expression of all bait and prey proteins was confirmed by Western blot analysis (Figure S4). In the case of Pik−-1 from Nipponbare, α-galactosidase activity was observed even in the absence of prey protein. This result suggests that the Pik−-1 bait plasmid has auto-activation ability in yeast cells, so we were unable to assess its interaction with AVR-Pik allelic products. Pikp-1, Pik*-1 and Pikm-1 all lack this auto-activation ability. We found a specific interaction between Pikp-1 and AVR-Pik-D, between Pik*-1 and AVR-Pik-D and -E, and between Pikm-1 and AVR-Pik-D, -E and -A. Neither Pikp-1, Pik*-1 or Pikm-1 bound AVR-Pik-C (Figure 6a). As far as Pikp-1, Pik*-1 and Pikm-1 are concerned, their binding specificities with AVR-Pik allelic products perfectly correspond to the recognition specificities of Pikp, Pik* and Pikm towards AVR-Pik alleles (Figure 3).
We further tested the interactions of the CC domain of Pik-1 allelic variants with AVR-Pik variants (Figure 6b,c). Expression of all bait and prey proteins was confirmed by Western blot analysis (Figure S4). Again, the CC domain of Pik−-1 showed auto-activation. The interaction patterns of the CC domain of Pik-1 variants with AVR-Pik variants were identical to those observed when the full-length Pik-1 was used (Figure 6a), except for binding of the Pik*-1 CC domain with AVR-Pik-A. The same result was obtained after swapping bait and prey (Figure 6c). The fact that the Pik*-1 CC domain binds AVR-Pik-A whereas the Pik*-1 full-length protein does not, suggests that it is not only the CC domain that determines the binding specificity between Pik-1 and AVR-Pik allelic products.
We further tested the specificity of binding in planta by co-immunopreciptation assays (Figure 5d). The CC domain of Pikm-1 co-immunoprecipitated with AVR-Pik-D but not with AVR-Pik-C, consistent with the result obtained in the yeast two-hybrid assay (Figure 6).
AVR-Pik-A expression levels affect its recognition by Pik*
In the inoculation assay with M. oryzae harboring different alleles of AVR-Pik as transgenes, we observed that the interaction between Pik* and AVR-Pik-A resulted in reduced development of disease lesions, which is not a typical compatibility (Figures 3 and 7a). By contrast, when we inoculated natural isolates Kyu92-22 and ARC-P90-19A that contain AVR-Pik-A into Kanto51 containing Pik* allele, clear-cut compatibility was observed (Figure 7a). To determine the reason of this discrepancy, we studied the expression levels of AVR-Pik-A in two transgenic isolates harboring AVR-Pik-A (#1 and #2) as well as the two natural isolates (Figure 7b). We found consistently higher levels of expression of AVR-Pik-A in the transgenic lines (#1 and #2) compared to the natural isolates (Kyu92-22 and ARC-P90-19A) 24 h after inoculation, suggesting that the expression levels of AVR-Pik-A may affect recognition by Pik*: the higher the level of AVR-Pik-A expression, the higher the chance of recognition by Pik*.
Direct interaction between AVR-Pik and Pik
Our results showed that M. oryzae AVR-Pik-D and rice Pikm-1 proteins directly interact in a yeast two-hybrid assay (Figure 5a) and an in planta co-immunoprecipitation assay (Figure 5c). This interaction is mediated by the Pikm-1 CC domain (Figure 5a,b). There is binding specificity between different AVR-Pik allelic products and different Pik-1 allelic products (Figure 6). This binding specificity determines the recognition specificity of Pik alleles towards different AVR-Pik alleles (Figure 3) mediated by HR induction (Figure 4). Direct binding between fungal AVR and plant R proteins was first reported for M. oryzae AVR-Pita and rice Pita proteins (Jia et al., 2000). In this case, single amino acid substitutions in the Pita LRR region or the AVR-Pita protease motif resulting in disruption of the physical interaction caused loss of resistance in the plant. Our finding is also similar to that observed in interactions between flax rust AVR-L567 alleles and flax L alleles (Dodds et al., 2006), in which 12 alleles of AVR-L567 are differentially recognized by L alleles by direct physical interactions of their products. To our knowledge, the present study is the third example of direct AVR–R interactions in fungal–plant host interactions.
It is interesting to note that AVR-Pik/Pik interactions involve the CC domain, but not the LRR domain, of the R protein. The majority of previous studies suggest that the LRR domain of R proteins is involved in AVR recognition (powdery mildew AvrMla1 recognition by barley Mla1, Shen et al., 2003; PVX recognition by potato Rx, Rairdan and Moffett, 2006) and in physical interaction with AVR proteins (blast AVR-Pita/rice Pita, Jia et al., 2000; Hyaloperonospora arabidopsidis ATR1/Arabidopsis RPP1, Krasileva et al., 2010; rust AVR-L567/flax L, Ravensdale et al., 2011). Recently, Chen et al. (2012) reported that the potato R protein RB binds to Phytophthora infestans AVR protein IPI-O through its CC domain. Together with the results of Chen et al. (2012), the CC domain-mediated interactions between R and AVR proteins described here reveal a new aspect of AVR/R interactions.
Co-evolution of AVR-Pik and Pik
Phylogenetic analysis suggests that AVR-Pik-D is the ancestral allele among the five AVR-Pik alleles (Figure 1c). AVR-Pik-D is also the allele recognized by the largest number of Pik alleles (Pik*, Pikp, Pikm, Piks and Pikh) as tested in the current study (Figures 3 and S3). The AVR-Pik-E allele appears to be derived from AVR-Pik-D by one amino acid change (Figure 1b). AVR-Pik-E is recognized by three Pik alleles (Pik*, Pikm and Pikh). There appears to be a tendency for the older AVR-Pik alleles to be recognized by a larger number of Pik alleles. The largest number of studied rice accessions harbor Piks (13 accessions), followed by Pikp (six accessions) (Figure 2). It is interesting to note that both Piks and Pikp can recognize AVR-Pik-D but not AVR-Pik-E, AVR-Pik-A or AVR-Pik-C. The most widespread rice Pik alleles recognize the oldest AVR allele, AVR-Pik-D. Based on these findings, we propose that the observed variation patterns may be interpreted as the result of co-evolution between the M. oryzae AVR-Pik gene and rice Pik gene (Figure 8). For instance, the following co-evolution scenario may be possible: first the AVR-Pik-D allele evolved from the common ancestor of AVR-Pik and pex75. Rice cultivars that recognize AVR-Pik-D may have had a selective advantage, so an R gene such as Pikp and Piks may have been selected. This selection may have resulted in evolution of AVR-Pik-E, which evades recognition by Pikp and Piks. In response to this situation, humans deployed another Pik allele, Pik*, that can recognize both AVR-Pik-D and AVR-Pik-E. Under this selection pressure, M. oryzae evolved the AVR-Pik-A allele by three amino acid changes from AVR-Pik-E. Again, another Pik allele, Pikm (or Pikh), was deployed that recognizes AVR-Pik-D, -E and -A. Under this selection pressure, M. oryzae then evolved the AVR-Pik-C allele that evades recognition by the three Pik alleles tested in this study. As the majority of the wild rice (O. rufipogon) studied belongs to Group 2 with no capacity to recognize any of AVR-Pik alleles, we hypothesize that AVR-Pik/Pik co-evolution occurred after domestication of rice. Also, the lack of synonymous substitutions among AVR-Pik alleles indicates that AVR-Pik allele diversification occurred in the very recent past, presumably as a result of serial deployment of different Pik alleles by humans. Whether the worldwide distribution of each AVR-Pik allele results from multiple independent events of selection or from a unique selection event followed by migration remains to be tested.
Dose effect of recognition of AVR-Pik-A by Pik*
The CC domain of Pik*-1 binds AVR-Pik-A (Figure 6b,c), but the full-length Pik protein cannot bind AVR-Pik-A (Figure 6a), resulting in no recognition of AVR-Pik-A by the Pik* allele under natural conditions (Figure 7a). These results suggest that other regions of the Pik* protein interfere with binding between the Pik* CC domain and AVR-Pik-A. However, if AVR-Pik-A is over-expressed as a transgene, the protein level becomes higher than the wild-type allele, and thus the isolates are partially recognized by Pik* (Figure 7a). This suggests a threshold effect of recognition. If the abundance of AVR is below a certain threshold, the number of cases in which R–AVR binding occurs is too low to trigger a sufficient level of HR to cause incompatibility. If the number of cases of R–AVR binding is augmented by increasing the amount of AVR and/or R, this may cause an HR that is strong enough to make the interaction incompatible. Such a dosage effect suggests that the expression levels of AVR and R are under strong natural selection. It will be interesting to systematically examine the expression levels of proteins involved in pathogen–host interactions.
Comparison of interactions between AVR-Pia and Pia and between AVR-Pik and Pik
In previous studies, it was shown that DNA polymorphisms of most M. oryzae avirulence genes cloned to date include both presence/absence and nucleotide substitutions (Orbach et al., 2000; Zhou et al., 2007; Li et al., 2009; Yoshida et al., 2009; Dai et al., 2010; Takahashi et al., 2010; Chuma et al., 2011). In contrast, two of them, AVR-Pia and AVR-Pii, showed only presence/absence polymorphisms. The easiest way for M. oryzae to avoid recognition by rice R genes appears to be to lose the AVR gene. The variation in AVR-Pia and AVR-Pii appear to fit this manner in avoiding recognition by cognate rice R genes. However, in the case of AVR-Pik, M. oryzae contains multiple alleles with amino acid substitutions. This suggests that loss of AVR-Pik may have a fitness penalty that is larger than that for AVR-Pia and AVR-Pii. To understand the fitness contribution of each AVR effector in the absence of a cognate R gene, identification of host proteins targeted by each effector is important.
DNA sequence analysis
Genomic DNA of all the plant materials was extracted from fresh leaves using a DNeasy plant mini kit (Qiagen, http://www.qiagen.com/). A fragment containing the region encoding the CC domain of Pik-1 or Pik-2 was amplified from genomic DNA using PrimeSTAR GXL polymerase (Takara, http://www.takara-bio.com/). The primers for Pik-1 alleles belonging to Group 1 were 5′-ATGGAGGCGGCTGCCATG-3′ and 5′-GCTTGAGCGAAAATGTCTGC-3′. The primers for Pik-1 alleles belonging to Group 2 were 5′- ATGGCCGTATACAGCGTCGC-3′ and 5′-TCGGTGAGATTGGCACCCGGA-3′. The primers for Pik-2 alleles belonging to Group 1 were 5′-ATGGAGTTGGTGGTAGGTG-3′ and 5′-GATCATCGTCCGTTGTTTCCTGC-3′. The primers for Pik-2 alleles belonging to Group 2 were 5′-ATGGAGTTGGCGGTAGGTGCTTC-3′ and 5′-TCATCAATCAAGAGGATATA-3′. Sequence reactions were performed using BigDye Terminator version 1.1 (Applied Biosystems, http://www.appliedbiosystems.com/). An Applied Biosystems 3130xl genetic analyzer was used to determine sequences. Sequencing primers were designed at approximately 500 bp intervals. We used the nucleotide sequences of AVR-Pik-A (AB498876), -B (AB498877), -C (AB498878), -D (AB498875) and -E (AB498879) to estimate nucleotide diversity and construct the tree.
Alignments of the sequences were performed using ClustalW (Thompson et al., 1994) and manually adjusted based on visual inspection to avoid misalignment. The dnasp program version 5.0 (Librado and Rozas, 2009) was used for the population genetic analysis. After indel removal, the nucleotide diversity π (Nei, 1987) was estimated. To investigate departures from neutrality, Tajima’s (1989) test was performed. Phylogenetic trees for Pik-1, Pik-2 and AVR-Pik were constructed using the mega program version 5 (Tamura et al., 2011).
M. oryzae materials
To create isogenic lines of M. oryzae harboring different AVR-Pik alleles, we transformed Sasa2 individually with expression vectors for four alleles of AVR-Pik: AVR-Pik-D, -E, -A and -C. We used Sasa2 harboring AVR-Pik-D as utilized by Yoshida et al. (2009). For construction of the AVR-Pik-A, -C or -E expression vectors pCB1531:AVR-Pik promoter:AVR-Pik-A, -C or -E, a 0.3 kb fragment containing an AVR-Pik allele was amplified using the primers described by Yoshida et al. (2009), digested with XbaI and BamHI, and replaced the ORF of AVR-Pik-D of pCB1531:AVR-Pia promoter:AVR-Pik-D (Yoshida et al., 2009), generating pCB1531:AVR-Pia promoter:AVR-Pik-A, -C or -E. A 2.2 kb fragment containing the AVR-Pik promoter was amplified using primers NotI-pex31-U1 (Yoshida et al., 2009) and 5′-GCTCTAGACAAAATAATGTCTTTTGCAAACAAAG-3′, digested using NotI and XbaI, and replaced the AVR-Pia promoter of pCB1531:AVR-Pia promoter:AVR-Pik-A, -C or -E, generating pCB1531-AVR-Pik promoter-AVR-Pik-A, -C or -E. DNAs from the Ina168, 70-15 and P2 isolates were used as PCR templates for the AVR-Pik-A, -C and -E alleles, respectively.
Rice leaf blade spot inoculation (see Kanzaki et al., 2002) was performed with M. oryzae strains. Disease lesions were photographed 10 days after inoculation. Rice seedlings (cvs Nipponbare, K60, Kanto51 or Tsuyuake) at the fourth leaf stage were used for inoculation. Rice seeds of Piks (Shin2 monogenic lines) or Pikh (K3 monogenic lines) were provided by the International Rice Research Institute (http://www.irri.org/).
Cell viability assay in rice protoplasts
Protoplasts were isolated from rice ethiolated seedlings (cvs Nipponbare, K60, Kanto51 or Tsuyuake) as described by Yoshida et al. (2009). PCR-amplified cDNAs of four alleles of AVR-Pik (AVR-Pik-D, -A, -E and –C) were digested with BamHI and inserted into BamHI sites of the pAHC17 vector under the control of the maize ubiquitin promoter (Christensen and Quail, 1996). To monitor cell vaiability of protoplasts, the firefly luciferase gene (LUC) under the control of the maize ubiquitin promoter was used as a reporter gene. Alleles of AVR-Pik were co-transfected together with LUC plasmids into rice protoplasts by electroporation. Luciferase activity as an index of cell viability was measured 40 h after transfection using a luciferase assay system (Promega, http://www.promega.com/), and the reduction in luminescence was compared with a control comprising empty vector-transfected protoplasts.
Yeast two-hybrid assay
Signal peptide-truncated cDNA fragments amplified from each AVR-Pik allele were inserted into EcoRI and BamHI sites of pGADT7 (prey) or pGBKT7 (bait) vectors (Clontech, http://www.clontech.com/). PCR-amplified fragments of the Pik-1 allele or the Pik-2 allele were isolated from seedlings of various rice cultivars (cvs Nipponbare, K60, Kanto51 or Tsuyuake) using PrimeSTAR GXL DNA polymerase (Takara), and cloned into pGADT7 or pGBKT7 using primer sets and restriction enzymes as listed in Table S3. The various combinations of bait and prey were transformed into yeast strain AH109 using the polyethylene glycol/lithium acetate (PEG/LiAc) method. Protein–protein interactions between bait and prey were evaluated by blue coloration and growth of yeast on basal medium lacking Trp, Leu, Ade and His but containing 5-Bromo-4-Chloro-3-indolyl α-D-galactopyranoside (X-α-gal) (Clontech). Details of plasmids used are given in Table S3.
In planta co-immunoprecipitation assay
cDNA fragments of the AVR-Pik-C or -D allele fused with the HA epitope tag were inserted into the BamHI and PstI sites of the pCambia1300S binary vector (Cambia, http://www.cambia.org). PCR-amplified cDNAs corresponding to the CC domain region of Tsuyuake Pikm1 were fused to the triple FLAG epitope tag and inserted into the SpeI and PstI sites of the pCambia1300S binary vector. These binary vectors were introduced into Agrobacterium strain GV3101. Agrobacterium cultures harboring binary vectors were infiltrated into leaves of Nicotiana benthamiana. Leaf samples taken 5 days after inoculation were ground in liquid nitrogen and solubilized in extraction buffer containing 10 mm dithiothreitol, 2% polyvinylpolypyrolidone and proteinase inhibitor cocktail (Sigma-Aldrich, http://www.sigmaaldrich.com/). The extracts were immunoprecipitated using anti HA-conjugated agarose beads (Sigma-Aldrich) and eluted using HA peptide (Sigma-Aldrich) as described by Win et al. (2011). Immunocomplexes were labeled using anti-FLAG peroxidase-conjugated antibody or anti-HA peroxidase-conjugated antibody (Sigma-Aldrich), and detected using an ECL-advance detection kit (GE Healthcare, http://www3.gehealthcare.com/). Details of plasmids used are given in Table S3.
Leaf sheaths of rice cv. Moukoto at the four- to five-leaf stage were inoculated with a suspension of M. grisea isolate Sasa2 (Yoshida et al., 2009). Total RNA was isolated from leaf sheaths 24 or 40 h after inoculation using an RNeasy plant mini kit (Qiagen). Single-strand cDNA was synthesized from DNase-treated RNA using oligo(dT) nucleotides and ReverTra Ace (Toyobo, http://www.toyobo-global.com/). Quantification of gene expression was performed using a Step OnePlus real-time PCR system (Applied Biosystems) according to the manufacturer’s instructions. Levels of AVR-Pik-A mRNA as studied by cDNA PCR amplification by primer pair: (5′-ACGTCAAGATGCTGGAACCTGT-3′ and 5′-CAACCTGGAGGGAAGTCGCC-3′) were measured using the constitutively expressed actin gene as studied by cDNA PCR amplification by primer pair: (5′-GCCGTCTTCCCGTCCATT-3′ and 5′-CTGGCCCATACCAATCATGAT-3′) as a control for normalization.
We thank Sophien Kamoun (Sainsbury Laboratory, John Innes Centre, Norwich, UK) for valuable comments on this paper. This work was partly supported by the Ministry of Agriculture, Forestry and Fisheries of Japan (Genomics for Agricultural Innovation PMI-0010) and the Program for Promotion of Basic Research Activities for Innovative Biosciences (PROBRAIN), Japan, and a Grant-in-Aid for Scientific Research from the Ministry of Education, Cultures, Sports and Technology, Japan to H.K. and R.T. (Grant-in-Aid for Scientific Research on Innovative Areas 23113009). L.A. was supported by a grant from the French National Agency for Research through the Gemo project (ANR 2009-GENM-29-01).