The Pm3 alleles of cultivated bread wheat confer gene for gene resistance to the powdery mildew fungus. They represent a particular case of plant disease resistance gene evolution, because of their recent origin and possible evolution after the formation of hexaploid wheat. The Pm3 locus is conserved in tetraploid wheat, thereby allowing the comparative evolutionary study of the same resistance locus in a domesticated species and in one of its wild ancestors. We have identified 61 Pm3 allelic sequences from wild and domesticated tetraploid wheat subspecies. The Pm3 sequences corresponded to 24 different haplotypes. They showed low sequence diversity, differing by only a few polymorphic sequence blocks that were further reshuffled between alleles by gene conversion and recombination. Polymorphic sequence blocks are different from the blocks found in functional Pm3 alleles of hexaploid wheat, indicating an independent evolution of the Pm3 loci in the two species. A new functional gene was identified in a wild wheat accession from Syria. This gene, Pm3k, conferred intermediate race-specific resistance to powdery mildew, and consists of a mosaic of gene segments derived from non-functional alleles. This demonstrates that Pm3-based resistance is not very frequent in wild tetraploid wheat, and that the evolution of functional resistance genes occurred independently in wild tetraploid and bread wheat. The Pm3 sequence variability and geographic distribution indicated that diversity was higher in wild emmer wheat from the Levant area, compared with the accessions from Turkey. Further screens for Pm3 functional genes in wild wheat should therefore focus on accessions from the Levant region.
Plant disease resistance genes (R genes) are a major component of the plant response to pathogen attack. The majority of R genes encode proteins with a nucleotide-binding (NB) and a leucine-rich repeat (LRR) domain associated at the N terminus, either with a Toll-interleukin receptor-like (TIR) domain or a coiled-coil (CC) domain. R proteins directly or indirectly perceive pathogen-emitted effector molecules, and trigger a resistance reaction mainly based on a hypersensitive response leading to cell death (Jones and Dangl, 2006).
At the molecular level, plant R genes show complex patterns of variation among different accessions/cultivars. To understand this diversity, R genes have been studied in the wild model plant species Arabidopsis. There, population genetic analysis at the RPM1, RPS2, RPS5 and RPP13 loci revealed high levels of polymorphism between alleles (RPP13 and RPS2) and distinct resistant and susceptible haplotypes (RPM1, RPS5 and RPS2), indicating that ancient genetic variation has been maintained by balancing selection (Bakker et al., 2006; Bergelson et al., 2001; Caicedo et al., 1999; Rose et al., 2004; Stahl et al., 1999). More recently, a genome-wide survey of R-gene polymorphism across 96 Arabidopsis accessions also identified a number of loci showing low levels of polymorphism, suggesting that alleles at R-gene loci can be generated and maintained for short periods of time (Bakker et al., 2006).
In contrast to the studies in the wild species Arabidopsis, there is only limited knowledge on the comparative evolution of R genes in domesticated species and their wild relatives. In the few studies available, the analysis of evolution and natural variation mainly focused on R genes that had been very recently introgressed from wild relatives into cultivated species. These genes showed basically no polymorphism in crop plants, and were found to be well conserved in wild donor accessions and related species. The Cf genes found in the Solanum sp. confer resistance to Cladosporium fulvum, and represent the best-studied case of a comparative analysis in wild and domesticated species. The Cf9 gene and a functional chimeric homolog (9DC) were highly conserved, and occurred frequently along the geographic range of the donor species Lycopersicon pimpinellifolium (Van der Hoorn et al., 2001). Highly similar functional Cf4 and Cf9 orthologs were identified in diverged Lycopersicon wild species, suggesting that the origin of these genes predated Lycopersicon speciation (Kruijt et al., 2005). In contrast, in a recent study in the lettuce/Bremia lactucae pathosystem, only one active Dm3 allele was discovered in more than 1000 wild accessions, suggesting a possibly recent evolution of the gene (Kuang et al., 2006). This Dm3 allele had the same specificity and only three amino acid changes compared with the Dm3 resistance gene against Bremia lactucae from cultivated lettuce.
The Pm3 gene occurs in seven different alleles (Pm3a–Pm3g), which encode CC-NB-LRR proteins that confer race-specific resistance to Blumeria graminis f. sp. tritici, the wheat powdery mildew fungus (Srichumpa et al., 2005; Yahiaoui et al., 2004, 2006). The seven resistance alleles showed low levels of sequence divergence, but did show functional diversity. They differ from a widespread susceptible allele (Pm3CS) by a few point mutations in the LRR region, or by a few tracks of sequence changes possibly derived from gene conversion. The Pm3 locus is present in wild emmer wheat, Triticum turgidum subsp. dicoccoides, which is the progenitor of cultivated tetraploid and hexaploid wheat. The susceptible Pm3CS allele was detected in wild emmer wheat accessions from areas close to the Karacadag mountains in south-eastern Turkey (Yahiaoui et al., 2006), a region proposed to be the site of wheat domestication (Luo et al., 2007; Ozkan et al., 2002, 2005). This led to the hypothesis that bread wheat Pm3 resistance alleles evolved from the ancestral Pm3CS sequence after wheat domestication 10 000 years ago.
The comparative analysis of disease resistance genes that have evolved in parallel, in wild and domesticated species, allows us to study the molecular evolutionary mechanisms that create functional diversity in resistance. In particular, it allows us to determine whether resistance alleles: (i) derive from a functional ancestor gene or (ii) independently evolved the resistance function. The presence of the Pm3 locus in both wild and cultivated wheat thus provides a unique opportunity to study the functional evolution of this locus in both species.
Isolation of Pm3 alleles from tetraploid wheat species
Wild emmer wheat is subdivided into two populations (Nevo and Beiles, 1989; Ozkan et al., 2005): the northern population in Turkey, Iran and Iraq, and the southern population in the Levant region (Lebanon, Jordan, Israel and Syria). To study the diversity of the Pm3 gene in tetraploid wheat, we analyzed a total of 174 accessions from different sources. A set of 126 accessions was taken from a collection of T. turgidum wild and cultivated subspecies assembled by Ozkan et al. (2002, 2005). This collection covers both the northern (17 accessions) and the southern (69 accessions) populations of emmer wheat (Figure S1). In addition, it includes domesticated tetraploid wheat from different, mostly European, countries (T. dicoccum and T. durum; 39 accessions). We also included 48 accessions from the National Small Grain Collection (USA), comprising: (i) 27 wild wheat accessions specifically from the Karacadag range in south-eastern Turkey, from which domesticated wheat might have originated, (ii) seven wild wheat accessions from Lebanon, and (iii) T. durum lines mostly from Turkey and Ethiopia. In total, we screened 121 T. dicoccoides, 14 T. dicoccum and 39 T. durum accessions from different countries for the Pm3 haplotype (Table S1). We first analyzed the accessions with a sequence-tagged-site (STS) marker that is highly correlated with the presence of any Pm3 allele (Tommasini et al., 2006; Yahiaoui et al., 2006). This marker amplifies a 900-bp fragment of the 5′ non-coding region of Pm3. A total of 62 T. dicoccoides and six T. durum accessions showed an amplification of this marker. The majority of these accessions are from the Asian part of Turkey or from the Levant area (Israel, Lebanon and Syria). None of the European accessions showed amplification of the Pm3 marker. Full-length coding sequences were obtained from 19 T. dicoccoides and from two T. durum accessions. Failure to amplify Pm3 genes from the remaining accessions could result from a deletion of the gene or low sequence homology at primer binding sites. In our analysis we also included 40 sequences from a previous screening of 201 accessions of wild and cultivated tetraploid wheat subspecies (Yahiaoui et al., 2006; Table S1). Finally, a total of 61 Pm3 sequences from tetraploid wheat were available, and among them 47 sequences were from wild tetraploid T. dicoccoides and 14 sequences were from domesticated T. dicoccum and T. durum.
Haplotype diversity of the Pm3 alleles from tetraploid wheat
All Pm3 alleles from tetraploid wheat had a size of 4442 bp, comprising two exons of 4156 and 86 bp and an intron of 200 bp. They did not contain any indels compared with the susceptible Pm3CS reference sequence previously described in bread wheat (Yahiaoui et al., 2006). Sequence alignment and analysis indicated that the Pm3 genes from the 61 tetraploid wheat accessions corresponded to 24 different sequences (haplotypes; Figure 1; Table S2). Almost half of the sequences (28 out of 61, 46%) belonged to three haplotypes (H1, H2 and H16; Table S2). Haplotypes H1 and H2 only differ by one base pair (bp 1138), resulting in one amino acid change (L380→ V380) in the NB region (Figure 1). Accessions carrying Pm3 sequences of the H1 haplotype comprise both wild T. dicoccoides from Turkey and cultivated T. dicoccum from Iraq, Iran and Russia, whereas those of the H2 haplotype were all wild wheat accessions from the Karacadag region in Turkey. One T. dicoccoides accession from Iran (IG113302; Table S2) defines a haplotype (H4) that only differs from haplotype H1 by a silent mutation at bp 1119 in the NB-encoding region. The third major haplotype H16 consists of sequences identical to Pm3CS. Four haplotypes were found [H17 (previously H16* in Yahiaoui et al., 2006), H18, H19 and H23] that only show very few polymorphic residues compared with H16 (Pm3CS). Finally, 14 haplotypes were identified (58% of the total number of haplotypes) that were represented by single accessions.
The nucleotide diversity was analyzed for the whole coding sequence of the Pm3 genes, and specifically for the regions encoding the distinct Pm3 protein domains (CC-NB domain, spacer and LRR domain; Table 1). This indicated that most of the haplotype and sequence diversity between Pm3 sequences was present in the LRR-encoding region. There, 22 different haplotypes were identified, compared with only nine and three in the CC-NB region and spacer regions, respectively. The very low level of polymorphism in the CC-NB region is illustrated by the scarce polymorphic residues shown in Figure 1 in this region, and by the low levels of synonymous and non-synonymous nucleotide diversity. In contrast, in the spacer and LRR regions, values of nucleotide diversity (π) were higher (Table 1). In these regions, polymorphic sequence blocks of various sizes were found between Pm3 haplotypes (Figure 1). The overall value of nucleotide diversity (πtotal = 0.0087; Table 1) for the entire tetraploid Pm3 gene sequences was lower than that previously described in other R-gene studies in wild species, such as for the RPP13 allelic series in Arabidopsis thaliana (πtotal = 0.045; Rose et al., 2004). This suggests a relatively recent divergence of these Pm3 genes in the tetraploid gene pool.
Table 1. Haplotype and nucleotide diversity in different coding regions of Pm3 genes from all tetraploid wheat species analyzed
Number of sites
Sixty one sequences were analyzed. The nucleotide diversity π is presented as the average number of pairwise nucleotide differences per 100 sites. S, synonymous; NS, nonsynonymous.
Entire gene excl intron (1–4242 bp)
CC-NB region (1–1515 bp)
Spacer (1516–1734 bp)
LRR (1735–4242 bp)
Haplotypes H1–H12 and H20/21 all differ from the remaining haplotypes by the presence of a highly polymorphic sequence block between bp 3220 and bp 3843 (block 1, LRR19–LRR25; Figure 1). This sequence block of 624 bp shows a total of 67 polymorphic bases consisting of 49 non-synonymous (replacement) mutations and 18 synonymous (silent) mutations. Within this block, some sequences showed small differences of between two and four amino acids (e.g. block 1a and 1b; Figure 1). Block 1 is only partially present in H13 and H15 (block 1c; Figure 1). In the spacer/first LRRs region, haplotypes H11 and H12 differ from all other haplotypes by a sequence block of 353 bp that includes the first three LRRs (block 2; Figure 1). The haplotypes H14 and H22 share a common and specific polymorphic sequence block at the 3′ end of the coding sequence (block 3; Figure 1), whereas the T. durum haplotype H24 shows a polymorphic sequence block (block 4; Figure 1), which also contains fragments of block 1 and amino acids present in H14 and H22. Except for Pm3CS (H16), none of the Pm3 haplotypes from tetraploid wheat corresponds to known Pm3 alleles from hexaploid bread wheat. The polymorphic sequence blocks (compared with the Pm3CS reference sequence) were not found as such in Pm3 resistance alleles from bread wheat. Only small fragments of blocks 1 and 2 were identified in the Pm3a, Pm3b and Pm3c sequences (Yahiaoui et al., 2006; Figure S2).
Race-specific resistance conferred by a Pm3 allele from tetraploid wheat
No functional Pm3 resistance alleles have been identified in wild wheat species (McIntosh et al., 2003; http://wheat.pw.usda.gov/ggpages/wgc/2006upd.html). To test if any of the newly identified Pm3 alleles from wild wheat might be functional, we carried out leaf infections at the seedling stage with three different powdery mildew races (Table S2). The powdery mildew isolates differed in their virulence/avirulence pattern to known Pm3 resistance alleles in hexaploid wheat: isolate 96224 is avirulent on all functional Pm3 alleles (except for an intermediate susceptible reaction with Pm3g), whereas isolates 97019 (AvrPm3a, AvrPm3b, AvrPm3c, avrPm3d, AvrPm3e, AvrPm3f and avrPm3g) and DB Asosan (avrPm3a, avrPm3b, avrPm3c, AvrPm3d, avrPm3e, avrPm3f and avrPm3g) show differential reactions on lines with the different Pm3 alleles. The six haplotypes H4, H12, H19, H20, H23 and H24 were represented by one or two accessions, and all of them were susceptible to the three powdery mildew isolates. Within the H1, H2, H3, H8, H10, H13 and H16 haplotypes, accessions carrying identical Pm3 sequences showed different patterns of resistance and susceptibility to the three isolates, suggesting that the resistance observed in some of the accessions was not caused by the Pm3 gene present in these lines. Among these haplotypes, H16 corresponds to the Pm3CS allele that was previously shown to be non-functional against powdery mildew (Yahiaoui et al., 2006). Finally, 11 haplotypes were mostly represented by single accessions that showed complete resistance (H6, H7, H9, H11, H14, H15, H17, H18, H21 and H22) or intermediate resistance (H5) to the three tested isolates (Table S2).
Nine Pm3 genes (H5, H8, H9, H10, H11, H14, H15, H21 and H22; Figure 1) were tested in a transient transformation assay (Schweizer et al., 1999; Yahiaoui et al., 2004). These genes were either from accessions showing full resistance or from accessions with different phenotypes. As a control for these experiments, we used the non-functional allele Pm3CS (H16) (Yahiaoui et al., 2006). The coding sequences of the Pm3 candidate genes or of the Pm3CS (H16) control were co-bombarded with a plasmid carrying the β-glucuronidase (GUS) reporter gene into leaf epidermal cells of the powdery mildew susceptible bread wheat line Chancellor. The leaf segments were subsequently infected with wheat powdery mildew isolate 96224, and the percentage of compatible (susceptible) interactions was determined. Eight of the tested Pm3 genes (H4, H8, H9, H10, H11, H14, H15 and H21) were not functional against isolate 96224 (data not shown). Only one of the tested genes, H22, conferred resistance to isolate 96224 in the transient transformation assay (Figure 2a). In the control experiments using Pm3CS, 76% of the cells expressing the GUS reporter gene and attacked by one powdery mildew spore showed fully developed haustorial structures, indicative of a compatible interaction (Figure 2a). When H22 was used, a significant reduction of the percentage of compatible interactions (39% compatible interactions) was observed after infection with isolate 96224 (Figure 2a). The activity of H22 was also confirmed using isolate 97019, where a significant decrease in compatible interactions was observed (83% of compatible interactions for the control versus 52% for H22). To check for race specificity of the resistance conferred by H22, we identified a powdery mildew isolate (97028) that is virulent on IG46439, the wild wheat accession carrying H22 (Figure 2b). When tested in the transient assay, H22 did not confer resistance to the virulent isolate 97028 (69% compatible interactions for H22 versus 74% for the control Pm3CS), thereby demonstrating the race-specific activity of H22 (Figure 2a).
Interestingly, the H14 haplotype, which is very close in sequence to H22 (Figure 1), was found to be non-functional against isolate 96224 in the transient assay (Figure 2a). The two sequences differ by one amino acid in the NBS domain [from V191 (H14) to F191 (H22)], and by eight residues in the LRR domain. Seven of these residues are also found in other alleles non-functional against tested powdery mildew isolates, albeit in different sequence contexts, and only one polymorphic amino acid is specific to H22. Interestingly, this polymorphic amino acid (C1332; Figure S2) is a solvent-exposed residue within the conserved motif (LXXLXLXX) of LRR27. The new functional Pm3 allele is named Pm3k. The resistance conferred in the transient assay by Pm3k (H22) is intermediate (between 61 and 48% incompatible interactions). However, a phenotype of full resistance was observed on primary leaves of the IG46439 accession when infected with isolate 96224 (Figure 2b). To check if the resistance conferred by Pm3-IG46439 in planta might be activated later than 48 h after infection (time point assessed in the transient assay) we performed Trypan blue staining experiments of IG46439 infected with the avirulent isolate 96224. We observed that in 94% of the interactions, pathogen growth was blocked at the appressorium stage at 48 h post-infection (hpi), whereas haustoria were formed in 6% of the interactions. In a wheat accession with the strong Pm3b allele, there were no haustoria formed at 48 hpi. Thus, Pm3-IG46439 is a weaker allele that also depends on reducing hyphal growth at later stages than 48 hpi, as no pathogen growth was observed macroscopically after 7 days. The difference in the levels of resistance observed at the penetration level between the transient assay and the microscopic analysis of the infection phenotype could result from the absence in Chancellor (the transformed wheat line) of factors required for full Pm3k function. Alternatively, it is possible that additional factors in IG46439 contribute to the strong powdery mildew resistance observed in the primary leaves of this accession.
Phylogenetic analysis of the different Pm3 haplotypes
A neighbor-joining phylogenetic tree based on nucleotide sequences of the Pm3 allelic sequences was established (Figure S3). Two main clades of Pm3 allelic sequences were identified. However, interior branches within each clade were poorly supported. To better represent possible conflicting phylogenetic relationships between the Pm3 allelic sequences, we inferred phylogenetic relationships between the 24 haplotype sequences using the split network method (Huson and Bryant, 2006). The obtained neighbor net (Figure 3) shows a separation between the haplotypes H1–H10, H20 and H21, and haplotypes H16 (Pm3CS)–H19 (hereafter, these two groups will be defined as H1 type and H16 type). The haplotypes H11/12 and H13/H15 are more closely related to the H1 group, whereas H14/H22 are more closely related to the H16 (Pm3CS) group. The H1 and H16 groups principally differ by the presence of the highly polymorphic sequence block 1 (between LRR19 and LRR25; Figure 1), which is also entirely or partially present in H11/12 and H13/H15, but is absent from H14 and H22 (where it is replaced by a Pm3CS type of sequence). In addition, all these sequences also share common smaller blocks of sequence polymorphisms (Figure 1), which were possibly integrated and exchanged by recombination/gene conversion events. The presence of a series of parallel edges in the neighbor-net network between the different Pm3 haplotypes (most particularly between H9/H10, H24, H22/H14 and H13/H15) is indicative of a complex evolutionary pattern that probably resulted from sequence exchange by recombination/gene conversion between these genes.
Given the pattern of sequence polymorphism, mostly based on clearly delimited sequence blocks, and the recombination/gene-conversion events that led to the diversity of Pm3 alleles in tetraploid wheat observed now, it is difficult to estimate a divergence time between these alleles. The major polymorphic sequence block between H1 and H16 sequence types carries a conserved number of synonymous mutations that can be used to calculate the age of its divergence. Using the number of synonymous mutations in this block, a P-distance value (Nei and Gojobori, 1986) of ds = 0.11 (±0.03) synonymous mutations per synonymous site was obtained. Using the grass Adh gene average substitution rate of 6.5 × 10−9 substitutions per synonymous site per year (Gaut et al., 1996), the divergence time for this sequence block is estimated to be 8.4 (±2) Mya. This indicates that the sequence blocks diverged before the hybridization event that led to the creation of tetraploid wheat, 0.5 Mya (Huang et al., 2002). Outside of the polymorphic blocks, very few sparsely distributed silent mutations were present in the Pm3 sequences (Figure 1). This is in accordance with a recent divergence after the hybridization that created tetraploid wheat.
Geographic and genetic differentiation of Pm3 haplotypes in different subpopulations of tetraploid wheat
To analyze the level of Pm3 allelic diversity among different tetraploid wheat subpopulations, we estimated the haplotype diversity and sequence polymorphism from cultivated and wild wheat accessions, and specifically from northern wild wheat accessions (from Turkey and Iran) and southern wild wheat accessions (from Israel, Jordan, Lebanon and Syria). DNA polymorphism analysis revealed a higher haplotype and nucleotide diversity for sequences from wild wheat accessions compared with cultivated wheat accessions (Table 2). However, the smaller sample size of cultivated wheat sequences might introduce a bias in the analysis. In a comparison between northern and southern accessions of T. dicoccoides, we found that the Pm3 haplotype and nucleotide diversities were higher in accessions of T. dicoccoides from the southern area compared with the accessions from Turkey.
Table 2. DNA polymorphism between Pm3 sequences from different subpopulations of tetraploid wheat
Number of accessions
Number of haplotypes
Hd (haplotype diversity ± SD)
Nucleotide diversity π (% ± SD)
The analysis was performed using DnaSP on whole gene sequences (introns included). SD, standard deviation. The northern accessions are from Turkey and Iran; the southern accessions are from Israel, Jordan, Lebanon and Syria.
0.92 ± 0.02
0.84 ± 0.08
0.68 ± 0.02
0.74 ± 0.14
0.93 ± 0.05
0.87 ± 0.1
Northern accessions (W)
0.80 ± 0.05
0.72 ± 0.07
Southern accessions (W)
0.94 ± 0.03
0.92 ± 0.2
To test for genetic differentiation between Pm3 haplotypes from different subpopulations of tetraploid wheat accessions, a statistical analysis was performed (Table 3). Two types of statistical values were calculated: Hst, for the haplotype diversity; and Kst* and Snn sequence-based statistics, for the nucleotide diversity (Hudson, 2000; Hudson et al., 1992). The most highly significant probabilities of genetic differentiation based on haplotype and nucleotide diversities were found between wild versus cultivated accessions, northern versus southern wild wheat accessions and cultivated versus wild southern accessions. The cultivated versus wild northern accessions showed a lower statistical significance for genetic differentiation based on Pm3 sequences. This is consistent with the idea that cultivated domesticated wheat species or subspecies might have originated from the northern pool of wild wheat (Luo et al., 2007; Ozkan et al., 2005), more specifically from the Karacadag region in south-eastern Turkey, which is represented by several accessions in our study (Table S2).
Table 3. Analysis of the genetic differentiation between Pm3 alleles from different subpopulations of tetraploid wheat accessions
Values for mean haplotype diversity (Hd), mean nucleotide differences (Kt) and for haplotype (Hst) and sequence based statistics (Kst*, Snn) are calculated based on Hudson et al. (1992) and Hudson (2000). Probabilities (between brackets) were calculated using a permutation test with 1000 replicates. ns, not significant; **significant, 0.001 < P < 0.01; ***highly significant, 0.000 < P < 0.001. W.: wild wheat accessions.
Wild (W.) versus cultivated (n1 = 47, n2 = 14)
0.81 (0.0) ***
W. northern versus southern (n1 = 26, n2 = 21)
Cultivated versus W. northern (n1 = 14, n2 = 26)
0.0085 (0.255) ns
Cultivated versus W. southern (n1 = 14, n2 = 21)
We have identified and analyzed a large set of 61 Pm3 allelic sequences corresponding to 24 different haplotypes found in tetraploid wheat subspecies. Pm3k, a new resistance allele of the Pm3 gene in tetraploid wheat was identified, shedding light on the evolutionary events creating active powdery mildew resistance genes in wild and domesticated wheat. The analysis of phylogenetic relationships, haplotype diversity and geographical distribution of the Pm3 genes revealed that Pm3 evolution paralleled the evolution of wild and domesticated wheat species.
Comparative evolutionary analysis of Pm3 alleles in wild tetraploid wheat versus hexaploid wheat
Although resistance to powdery mildew was quite frequent in wild tetraploid wheat, we found that it was mostly caused by loci other than Pm3. In our screen for Pm3-based resistance, we have used European powdery mildew strains isolated from bread wheat and combining differential virulence and avirulence on known Pm3 resistance alleles. One functional R gene against these isolates was identified, but it is possible that some of the other Pm3 alleles from wild wheat are active against specific local isolates from wild wheat collection areas.
The low frequency of Pm3-based resistance in wild emmer is similar to the situation in lettuce, where the Dm3 resistance gene against B. lactucae was rare in natural populations of the wild lettuce Lactuca serriola (Kuang et al., 2006). It was postulated that the low frequency of Dm3 was caused by recent evolution of the locus, by a deletion of the gene or by high variation in the sequence that made its detection difficult. In the case of Pm3, the gene was present in a large number of accessions, and Pm3 sequences were conserved enough to be identified. Therefore, the low frequency of Pm3-based resistance in wild emmer could be the result of an independent evolution of active resistance genes in the tetraploid versus hexaploid wheat gene pools. The situation at the Dm3 and Pm3 loci differs from studies of natural variation of R genes in Arabidopsis. There, the frequency of the functional R genes studied was high, and most of the R-gene orthologs or alleles were well conserved in function, and were of ancient origin (Bakker et al., 2006; Bergelson et al., 2001; Rose et al., 2004).
Until now, only one Pm3 sequence, the Pm3CS (H16) allele, was found to be identical between bread wheat and tetraploid wheat. In bread wheat, the Pm3 resistance alleles diverged from the Pm3CS template by a few specific point mutations, or by gene conversion events introducing polymorphic sequence blocks that are different from the ones found in tetraploid wheat (Yahiaoui et al., 2006; this work). The functional Pm3k allele of wild emmer shows polymorphic sequence blocks that are not found in known bread wheat resistance alleles. This indicates that resistance conferred by this gene independently evolved within the tetraploid gene pool.
Pm3k consists of building blocks from non-functional alleles
The functional Pm3k (H22) sequence is a chimeric arrangement of blocks from alleles that are non-functional against the powdery mildew isolates tested. A mosaic pattern of conserved sequences is found among the flax rust resistance L alleles (Ellis et al., 1999), but this sequence exchange occurred between functional alleles. A reshuffling of non functional sequences to generate resistance genes was described for Cf genes, which originate from a combination of sequences from paralogous genes (Parniske et al., 1997). Thus, the structure of Pm3k supports the proposed hypothesis of recycling polymorphism, where non-functional sequences serve as a reservoir of variation to generate functional genes (Holub, 2001).
The LRR domain and the particularly variable solvent-exposed residues are proposed to be responsible for the specificity of NB-LRR proteins (Ellis et al., 1999; Jones and Jones, 1997), and were shown to play an important role in the Pm3 specificity of resistance (Yahiaoui et al., 2006). The LRR domain also plays an important function in the negative or positive regulation of R proteins through intramolecular interactions with the NB domain (Moffett et al., 2002; Rairdan and Moffett, 2006). The Pm3k (H22) sequence carried only a single polymorphism, not found in any susceptible allele (C1332), a predicted solvent-exposed residue of LRR27. Interestingly, in hexaploid bread wheat, one amino acid change in a solvent-exposed residue of LRR27 (E1334 to V1334) was sufficient to convert the susceptible Pm3CS into a functional resistance allele (Yahiaoui et al., 2006). This suggests that the function of Pm3k could arise from the C1332-specific polymorphism in LRR27, and points to the importance of this particular region of the LRR for PM3 protein function. However, it cannot be excluded that the combination in H22 of sequence motifs shared with the other emmer wheat Pm3 proteins also plays a role in Pm3k function.
It is difficult to determine if Pm3k has a different spectrum of race specificity compared with other Pm3 alleles. Accession IG46439 was resistant to several of the isolates tested (N. Kaur and B. Keller, unpublished data), and, as mentioned earlier, it is not excluded that other genes present in IG46439 contribute to the resistance phenotype. Based on transient assay data, Pm3k conferred intermediate resistance to isolates avirulent to most Pm3 alleles of bread wheat, but not to isolate 97028, which is virulent on all Pm3 alleles except for Pm3a and Pm3b. Race specificity of the Pm3k resistance is therefore certainly different from that of Pm3a and Pm3b.
Timing and mechanisms of evolution of tetraploid wheat Pm3 alleles
Phylogenetic analysis of Pm3 haplotypes from tetraploid wheat identified two main clades that differed mostly by one sequence block. This sequence block is of ancient origin, with a divergence time predating the divergence of ancestral wheat diploid species (2.5–4.5 Mya; Huang et al., 2002). In hexaploid bread wheat Pm3 alleles, no synonymous mutations were found outside of clearly delimited polymorphic sequence blocks (Yahiaoui et al., 2006), leading to the hypothesis of a very recent divergence 10 000 years ago. In tetraploid wheat, few synonymous mutations were found outside of the polymorphic blocks. This probably reflects a relatively more ancient divergence compared with bread wheat alleles, and is compatible with the estimated time of the hybridization event at the origin of tetraploid wheat, less than 0.5 Mya (Huang et al., 2002).
Relationships between Pm3 sequences are obscured by the low level of polymorphism, and by recombination/gene conversion events involving the polymorphic segments. The Pm3 alleles from tetraploid wheat show few polymorphisms, which are found specifically in one haplotype only (1.5 unique nucleotide polymorphism per sequence). Most polymorphic sequences were found at least twice among the different haplotypes, indicating frequent sequence exchange between alleles. Sequence exchange by gene conversion and/or recombination is one of the major mechanisms of R gene evolution (Kuang et al., 2004; Mondragon-Palomino and Gaut, 2005), and it is also the main mechanism of Pm3 evolution in tetraploid wheat.
The haplotype distribution and the geographic differentiation of Pm3 alleles from tetraploid wheat provide insight into the history of wheat evolution
Among the three main tetraploid gene pools (wild T. dicoccoides, domesticated T. dicoccum and T. durum) that we have screened for their Pm3 haplotype, most Pm3 alleles were detected in wild T. dicoccoides accessions (43% of screened accessions compared with 15 and 13% for T. dicoccum and T. durum, respectively). The wild gene pool is therefore enriched in Pm3 sequences compared with domesticated emmer.
In our screen, the three most frequent Pm3 haplotypes were from Turkey, Iran and Iraq (H1, H2 and H16). Two of these haplotypes comprise both wild and domesticated emmer accessions, and therefore possibly identify the two main Pm3 sequences that were transmitted through domestication to the cultivated gene pool. The H16 haplotype is present in cultivated hexaploid wheat (Yahiaoui et al., 2006). It is specifically found in wild wheat accessions from Turkey (from or close to the Karacadag region), and closely related variants were either in wild emmer from the Karacadag region (H18 and H19) or were in domesticated emmer (H17 and H23). Several studies have found that domesticated wheat is most closely related to the northern gene pool of wild emmer wheat (Luo et al., 2007; Ozkan et al., 2002, 2005). The analysis of Pm3 genetic differentiation also showed a significant differentiation between domesticated wheat accessions and wild wheat accessions from the southern pool, compared with accessions from the northern pool. Our results are therefore in accordance with the hypothesis that domesticated wheat originates from the northern pool.
A higher genetic diversity was described for the southern population, which is therefore proposed to be an important source for wheat improvement (Luo et al., 2007; Nevo, 1998). In agreement with this, our data on Pm3 haplotype and nucleotide diversity indicate a higher diversity among accessions from the southern pool. This could be because of more diverse eco-geographic conditions in this region compared with the sampled areas of the northern pool. The new resistance allele Pm3k was found in an accession from the southern pool. This suggests that additional functional alleles can be found in accessions from this area, and that further screenings should be made on southern populations growing under specific and diverse ecologic conditions.
The identification and the conservation of the Pm3 locus at two different ploidy levels in the wheat gene pool, and its presence in wild relatives and in domesticated species, allowed us to correlate and link the evolution of this gene to the history of wheat evolution and domestication. In terms of exploiting the tetraploid gene pool in order to identify new functional Pm3 resistance alleles, our work shows that although this resistance is not very frequent, it is possible to identify functional Pm3 variants in the tetraploid wheat gene pool. The partial resistance conferred by the gene we identified could be of agricultural interest, as it would potentially induce less pressure on the pathogen populations, and would therefore be of more durable use.
Plant material and wheat powdery mildew isolates
Accessions used for haplotype analysis and Pm3 gene amplification were obtained from F. Salamini (Max-Planck-Institute for Plant Breeding Research, http://www.mpiz-koeln.mpg.de/english/index.html), J. David, (INRA Montpellier, http://www.montpellier.inra.fr), M. Feldman (Weizmann Institute of Science, http://www.weizmann.ac.il), and from stock centers at IPK Gatersleben (http://www.ipk-gatersleben.de) and USDA National Small Grains Research Facility (http://www.ars.usda.gov). We use here the abbreviated forms T. dicoccoides for wild Triticum turgidum subps. dicoccoides, T. dicoccum for domesticated hulled emmer T. turgidum subps. dicoccum and T. durum for domesticated free-threshing hard wheat (T. turgidum subps. durum). Wheat powdery mildew (Blumeria graminis f. sp. tritici) isolates were provided by P. Streckeisen (Forschungsanstalt Agroscope Reckenholz-Tänikon, http://www.art.admin.ch) and by D. Barloy (INRA Rennes, http://www.rennes.inra.fr), and were maintained on wheat cv. Kanzler by weekly transfer to fresh plants.
Pm3 haplotype analyses
Isolation of genomic DNA, PCR amplification and analyses were performed as previously described (Srichumpa et al., 2005). Pm3 haplotype analysis was performed by an STS marker analysis with primer pair UP3B (5′-TGGTTGCACAGACAATCC-3′) and UP1A (5′-GAAACCCGGCATAAGGAG-3′).
Cloning of Pm3 alleles and test gene constructs
PCR amplification of Pm3 genes was carried out with the PfuUltra high-fidelity DNA polymerase (Stratagene, http://www.stratagene.com) using a nested PCR strategy with primers UP6 (5′-GGCACAGACAAAGCTCTG-3′) and N3SP3RP (5′-ACAATCAGGGATCAGGCC-3′) in a first step, and primers BamH1_1 (5′-TTAATTGGATCCCCAATGGCAGAGCGGGTGGTC-3′) and Sal1_1 (5′-TATATAGTCGACGCTTCAGCTCCGGCAGGCCTG-3′) in a second step (Srichumpa et al., 2005). For Pm3 alleles that were found as unique sequences in tetraploid wheat, or were tested in the transient assay, two independent PCR reactions were carried out. Amplified fragments were either sequenced as PCR products or were cloned into the multiple cloning site of vector PGY1 (Schweizer et al., 1999) between a 540-bp fragment of the 35SCaMV promoter and the 35SCaMV terminator. Two clones per PCR reaction were sequenced. DNA sequencing was carried out on Applied Biosystems Capillary Sequencer model 3730 (http://www.appliedbiosystems.com).
Single-cell transient transformation assay
Biolistic bombardment was performed as described in Yahiaoui et al. (2004) and modified as in Douchkov et al. (2005). Leaves of the powdery mildew susceptible line Chancellor were bombarded with a 1:1 (weight:weight) mixture of pUbiGUS containing the GUS reporter and the PGY1 control vector containing the Pm3CS gene (Yahiaoui et al., 2004), or the test gene constructs from tetraploid wheat. Leaf segments were infected with wheat powdery mildew 4 h after bombardment. Staining for GUS activity was carried out at 48 hpi. Fungal structures were then stained with Coomassie blue. GUS-expressing epidermal cells attacked by a single germinating spore were evaluated by transmission light microscopy. A compatible (susceptible) interaction was characterized by a mature haustorium and elongating secondary hyphae. An incompatible (resistance) interaction was characterized by the presence of an appressorium. A Student’s two-tailed t-test was performed as a statistical validation of the results.
Sequence assembly was performed using the Gap4 program of the Staden Package (http://staden.sourceforge.net). ClustalX (Thompson et al., 1997) was used for sequence alignments, which were visualized in Genedoc (http://www.nrbsc.org/gfx/genedoc/index.html). The different R-protein domains (CC-NB, LRR, structural LRR residues and solvent-exposed LRR residues in the XXLXLXX motif) were chosen according to the criteria outlined byMeyers et al. (2003). Divergence time (T) was estimated using r = ds/2T, where r is the grass Adh1 and Adh2 loci substitution rate of 6.5 × 10−9 substitutions per synonymous site per year (Gaut et al., 1996), and ds is the number of synonymous substitutions per synonymous site between compared sequences.
Phylogenetic and genetic differentiation analysis
The neighbor-net phylogenetic analysis was performed with the Splitstree program, based on uncorrected P distances. The analysis of polymorphism data was performed with DnaSP v4.0 (Rozas et al., 2003). Nucleotide diversity was estimated for synonymous and non-synonymous sites as π (Tajima, 1983). The DnaSP v4.0 software (Rozas et al., 2003) was also used for the statistical analysis of genetic differentiation.
We would like to thank J. David, M. Feldman, F. Salamini and H. Bockelman for accessions of tetraploid wheat. We also thank Geri Herren for excellent technical assistance. This project was financially supported by a grant from the Swiss National Science Fondation to BK (3100–105620).
Accession Numbers: Sequences of Pm3 genes from tetraploid wheat Pm3_H1 to Pm3_H15 and Pm3_H17 to Pm3_H24 have been deposited in the Genebank database under accession numbers EU192106 up to EU192128.