•In plants, the evolution of specific resistance is poorly understood. Pseudomonas syringae effectors AvrB and AvrRpm1 are recognized by phylogenetically distinct resistance (R) proteins in Arabidopsis thaliana (Brassicaceae) and soybean (Glycine max, Fabaceae). In soybean, these resistances are encoded by two tightly linked R genes, Rpg1-b and Rpg1-r. To study the evolution of these specific resistances, we investigated AvrB- and AvrRpm1-induced responses in common bean (Phaseolus vulgaris, Fabaceae).
•Common bean genotypes of various geographical origins were inoculated with P. syringae strains expressing AvrB or AvrRpm1. A common bean recombinant inbred line (RIL) population was used to map R genes to AvrRpm1.
•No common bean genotypes recognized AvrB. By contrast, multiple genotypes responded to AvrRpm1, and two independent R genes conferring AvrRpm1-specific resistance were mapped to the ends of linkage group B11 (Rpsar-1, for resistance to Pseudomonas syringae effector AvrRpm1 number 1) and B8 (Rpsar-2). Rpsar-1 is located in a region syntenic with the soybean Rpg1 cluster. However, mapping of specific Rpg1 homologous genes suggests that AvrRpm1 recognition evolved independently in common bean and soybean.
•The conservation of the genomic position of AvrRpm1-specific genes between soybean and common bean suggests a model whereby specific clusters of R genes are predisposed to evolve recognition of the same effector molecules.
In their natural environment, plants encounter a vast array of pathogenic microorganisms such as fungi, oomycetes, bacteria, viruses and nematodes. To combat these microbial pathogens, plants use a two-level innate immune system (Chisholm et al., 2006; Jones & Dangl, 2006). The first level is the recognition of microbial- or pathogen-associated molecular patterns (MAMP or PAMP), and is referred to as PAMP-triggered immunity (PTI) (Boller & Felix, 2009). To overcome PTI, pathogens have evolved effector proteins that are translocated into host cells, where they interfere with signaling pathways required for the induction of PTI. In turn, plants have evolved specific resistance (R) proteins that detect the presence of individual pathogen effectors, resulting in effector triggered immunity (ETI) (Innes, 2004; Dangl & McDowell, 2006). Most R proteins contain a nucleotide-binding site and leucine-rich repeats (NB-LRRs) and are encoded by large gene families in plants (Meyers et al., 2005; Collier & Moffett, 2009). NB-LRR proteins can be subdivided into two major subfamilies, the Toll-interleukin 1 receptor (TIR) NB-LRR (TNL) and the coiled-coil (CC) NB-LRR (CNL), depending on their N-terminal domain (Rairdan et al., 2008). TNL and CNL are encoded by two phylogenetically distinct classes of genes (Meyers et al., 2003; Ameline-Torregrosa et al., 2008; Kohler et al., 2008). In plant genomes, NB-LRR sequences can be distributed as single loci, such as RPM1 in Arabidopsis thaliana (Grant et al., 1995), but are more often grouped into complex loci, as exemplified by A. thaliana, where two-thirds of them are organized in tightly linked clusters (Meyers et al., 2003; Leister, 2004; McDowell & Simon, 2006, 2008. The available data suggest that different R genes can follow strikingly different evolutionary trajectories, such as the fast-evolving (type I) and the slow-evolving (type II) R genes identified at the lettuce (Lactuca sativa L.) Dm3 CNL cluster (Kuang et al., 2004). It is not yet clear what mechanisms underlie these apparently disparate rates of R gene evolution.
Bacterial strains belonging to the species Pseudomonas syringae attack a broad range of plant species, which provides an opportunity to study the evolution of resistance against this pathogen in different plant lineages. Three strains of this bacterium have been sequenced, and the P. syringae effector repertoire has been extensively studied during the last decade (Mansfield, 2009). In A. thaliana, RPM1-interacting protein 4 (RIN4) is a point of convergence between PTI and ETI and is an example of the guard hypothesis (Chisholm et al., 2006). RIN4 functions as a regulator of PAMP signaling, and is manipulated by at least three P. syringae effectors (AvrRpm1, AvrB and AvrRpt2) to promote virulence (Belkhadir et al., 2004; Kim et al., 2005a; Kim et al., 2005b). AvrB–RIN4 or AvrRpm1–RIN4 interactions are correlated with the phosphorylation of RIN4 and activation of resistance mediated by RPM1, a CNL protein (Mackey et al., 2002). By contrast, AvrRpt2 induces RIN4 cleavage and activation of RPS2, a distantly related CNL (Axtell & Staskawicz, 2003; Mackey et al., 2003; Day et al., 2005). Interestingly, AvrRpt2 suppresses RPM1 function by inducing RIN4 disappearance in RPM1 rps2 plants (Mackey et al., 2003). In the A. thaliana genome, RPM1 is a single-copy R gene and confers dual resistance specificity to AvrB and AvrRpm1 (Mackey et al., 2002, 2003), while in soybean (Glycine max) resistance to these effectors is encoded by Rpg1-b and Rpg1-r, respectively, two RPM1-unrelated CNL genes tightly linked at the complex Rpg1 resistance cluster. AvrRpt2 also suppresses Rpg1-b function in soybean, providing indirect evidence that a RIN4 homolog may be required for Rpg1-b function (Ashfield et al., 2003). Despite the functional similarities between Rpg1-b and RPM1, these genes belong to distinct evolutionary lineages of CNL that diverged before the monocot–dicot split, suggesting that the ability to recognize AvrB and AvrRpm1 arose independently in A. thaliana and soybean through convergent evolution (Ashfield et al., 2004; McDowell, 2004).
We have previously sequenced and analyzed c. 950 kb of common bean (Phaseolus vulgaris) genomic DNA that is syntenic to the soybean Rpg1 resistance cluster (Innes et al., 2008; Wawrzynski et al., 2008). This common bean sequence assembled into five contigs, but the locations of these contigs in the common bean genome were not determined. In both species, the sequenced regions contain a large Rpg1-related CNL cluster that appeared before the radiation of legume species c. 50 Mya (Innes et al., 2008). Phylogenetic analysis of this cluster revealed that there has been a high frequency of duplication and loss of CNL sequences in both lineages subsequent to the Glycine–Phaseolus split that occurred c. 20 Mya, and this process has led to different physical distributions of Rpg1-related CNL subfamilies in common bean and soybean (Innes et al., 2008). Given this high rate of CNL gene turnover, we wished to determine whether an Rpg1-b- and/or Rpg1-r-like gene is present in common bean, and, if so, whether such genes have been maintained since the divergence of soybean and common bean. Here we report that multiple common bean genotypes possess Rpg1-r-like genes (i.e. conferring recognition of AvrRpm1), while no tested genotypes contained Rpg1-b-like genes (i.e. conferring recognition of AvrB). Furthermore, genetic mapping in common bean uncovered two unlinked genes with AvrRpm1 specificity, one of which mapped to the Rpg1-syntenic region. We discuss these findings in the context of CNL evolution.
Materials and Methods
The geographical origins of wild and cultivated common bean (Phaseolus vulgaris L.) genotypes are presented in Table 1. Seventy-seven F11 recombinant inbred lines (RILs) derived from a cross between BAT93 and JaloEEP558 were used for mapping experiments (Freyre et al., 1998; Geffroy et al., 2000). We have generated 102 additional F9 BAT93xJaloEEP558 RILs which were also used to map resistance specificities and molecular markers (Supporting Information Table S1). The near-isogenic lines (NILs) differing at the Co-2 (Colletotrichum lindemuthianum 2) anthracnose resistance locus (P12R and P12S) are described in Geffroy et al. (1998). P12R results from the introgression of the Co-2 anthracnose resistance gene from Cornell49242 in the Andean ‘Processor’ background, and is a BC12-F15 line.
Table 1. Distribution of AvrB or AvrRpm1 resistance specificities in wild and cultivated common bean (Phaseolus vulgaris) accessions of various geographical origins
aA, Andean; SA, south Andean; NA, north Andean; M, Mesoamerican; C, Columbian.
The Pseudomonas syringae pv phaseolicola strain 1448AN race 6 (Psp) has been described previously (Fillingham et al., 1992). Psp(avrB) carries the plasmid pVB01 described in Innes et al. (1993) and Psp(avrRpm1) carries the plasmid pVSP61/avrRpm1 described in Ashfield et al. (1995). Psp(avrRpt2) and Psp(avrRpt2::Ω) carry the plasmids pLH12 and pLH12::Ω (with an omega insertion in avrRpt2), respectively (Innes et al., 1993).
Pathogenicity tests and in planta bacterial population counts
Bean pods were inoculated with a suspension of 5 × 108 cells ml−1, as described in Harper et al. (1987). Bean plants were inoculated using a 2-ml syringe (without needle) to infiltrate a bacterial suspension of 1 × 107 cells ml−1 (for phenotypic tests) or 1 × 104 cells ml−1 (for bacterial population counts) into the underside of fully expanded cotyledonary leaves. Bacterial multiplication was examined following the procedure described in Katagiri et al. (2002). Leaf disks from two independent replicate plants were pooled for each tissue sample, and three samples were examined for each time-point. Resistance phenotypes induced by Psp(avrRpm1) in cotyledonary leaves of the 179 BAT93xJaloEEP558 RILs were scored in two independent experiments as fully resistant (R) and susceptible (S). An intermediate phenotype (I) described in Fillingham et al. (1992), corresponding to maceration associated with hypersensitive response (HR) limited to areas adjacent to main veins, was also identified.
Markers used in this study are described in Table S1. A PCR-based approach was used to map the previously sequenced contigs from the common bean G19833 genotype using specific oligonucleotide primers (Innes et al., 2008). CNL1 and CNL3/4 probes were mapped by Southern blot hybridization experiments as described in Geffroy et al. (1998). χ2 tests were used to evaluate the goodness of fit of observed and expected segregation ratios. mapmaker software version 3.0 (Lander et al., 1987) was used to map segregating markers as described in Geffroy et al. (2000). The maximum likelihood procedure described in Geffroy et al. (2008) was used to map the two independent R genes (Rpsar-1 (for resistance to Pseudomonas syringae effector AvrRpm1 number 1) and Rpsar-2) encoding AvrRpm1 resistance specificity in the common bean genome. Rpsar-2 was also mapped using mapmaker software by coding the intermediate (I) phenotype as susceptible (S).
Annotation and microsynteny analysis
We have previously sequenced and annotated a c. 1-Mbp region encompassing the Rpg1 locus on soybean chromosome 13 (GmChr13) and c. 950 kb of syntenic DNA from common bean (localized in this study to linkage group (LG) 11, corresponding to chromosome 11 (PvChr11) (Pedrosa-Harand et al., 2009; Fonsêca et al., 2010)), and c. 250 kb from a homoeologous region (H3) on soybean chromosome 6 (GmChr6) (Innes et al., 2008). Here, we have sequenced a 239 262-bp contig of two overlapping bacterial artificial chromosome clones, Pva1-105k5 (FQ032806) and Pva1-120g2 (FQ032805), from the region syntenic to soybean H3 on chromosome 5 of common bean (PvChr5). Sequencing and assembly were performed at Genoscope (Evry, France), and subsequent annotation was performed as described in Innes et al. (2008). Putative homologous sequences from A. thaliana were searched by performing a TBLASTN search with the annotated soybean gene structures from the GmChr13 region excluding CNL, retroelement, and kinase sequences, against the available genomic sequence of A. thaliana (http://www.Arabidopsis.org/) with a maximum E-value of 1e−10. Two microsyntenic regions were retrieved on A. thaliana chromosomes 3 (AtChr3) and 5 (AtChr5), based on the arbitrary criterion of finding at least three pairs of homologous genes within a maximum physical distance of six unrelated genes between them. The identified region on AtChr3 is located between RPM1 and two Rpg1-related CNLs (At3g14460 and At3g14470). To check for more diffuse microsynteny in this region, we aligned large portions of 2.7 Mbp each from the genomic sequences of soybean (http://soybase.org/) (Schmutz et al., 2010) and A. thaliana (http://www.Arabidopsis.org/), centered on the regions of interest. Synteny images were generated using custom Perl scripts kindly provided by Steven B. Cannon and the GD-SVG image library. Gene correspondences were calculated using BLASTALL (Altschul et al., 1997) on peptide sequences from CNL- and repeat-filtered FGENESH gene calls (Salamov & Solovyev, 2000). Genes were considered homologs if their BLAST-based E-values were < 1e−10. Only top hits were considered.
Database searches and phylogenetic analyses
RIN4-homologous sequences from common bean and other plant species were retrieved using a TBLASTN search (1 × 10−10E-value cut-off) against the putative unit transcript (PUT) of plantGDB for common bean (Dong et al., 2004), with subsequent expressed sequence tag (EST) contig assembly using cap3 (Huang & Madan, 1999), or the whole genome of other species listed in Fig. 5 (http://www.phytozome.net). Multiple alignment of amino acid sequences from the NB portion of CNLs, or the complete sequence of RIN4 homologs, was performed with mafft version 6 using the L-INS-I strategy (Katoh et al., 2005) using default parameters. These alignments were subjected to Bayesian analysis using MrBayes version 3.1.2 (Ronquist & Huelsenbeck, 2003). MrBayes analysis was run with a Jones amino acid substitution model, following a gamma distribution. We performed paired runs with four chains each and ran the analysis for 10 million generations with sampling every 100 generations. The prior for each analysis was of equal probability. All runs started with a random tree. After elimination of the first 50% of runs, which included the burn-in phase, the remaining iterations were summarized in a consensus tree with posterior probabilities as nodal support. The resulting phylogenetic tree was displayed using mega (Tamura et al., 2007). Ks (nucleotide substitution rates at silent sites) was determined using the yn00 algorithm implemented in paml 4.3 (Yang, 2007).
The genomic region containing the Co-2 resistance locus in common bean shares synteny with the soybean genomic region containing Rpg1-b
To map the Rpg1 syntenic region(s) in the common bean genome, we developed five PCR-based markers distributed across the five BAC (Bacterial Artificial Chromosome) contigs from common bean (Fig. 1b, Table S1). All five markers mapped to the end of common bean linkage group (LG) B11 in a 2.6-cM interval that included SCH20, a SCAR (Sequence-characterized amplified region) marker linked to the Co-2 anthracnose resistance gene (Fig. 1a,b) (Geffroy et al., 1998). This order is in agreement with the order previously predicted by the microsynteny analysis between soybean and common bean at the Rpg1 locus (Fig. 2) (Innes et al., 2008), confirming that, despite four gaps in the common bean sequence, these regions are collinear within their whole length, and that no major rearrangement has occurred within this locus since the split between Phaseolus and Glycine 20 Mya (Lavin et al., 2005). In order to orientate these five contigs relative to the telomere, three Leg (Legume) markers (Leg043, Leg220, and Leg208), previously shown to map at the same end of LG B11 (Hougaard et al., 2008), and the SCH20 marker were scored on 179 BAT93xJaloEEP558 RILs. Marker 47b16-f1 is the most proximal of the five markers, mapping at 15.5 cM from Leg043. It is followed in order by markers 68o6-f1, 34g17-f1, which co-segregates with SCH20, 76g17-f1, and 118d24-r05, the most distal marker, co-segregating with Leg220 (Fig. 1b).
Two Rpg1-related CNL subfamilies with distinct physical distributions exist in common bean
Previous phylogenetic analysis showed that the CNL sequences from the Rpg1 cluster form four distinct clades (CNL1 to CNL4; Figs 1 and 2) that diverged before the split between Phaseolus and Glycine (Innes et al., 2008). Sequences belonging to CNL2 were not found in common bean, suggesting that they were lost in common bean after the divergence between Phaseolus and Glycine (Innes et al., 2008). Hybridization of specific CNL probes to restricted genomic DNA from common bean genotypes BAT93 and JaloEEP558 revealed that CNL3 and CNL4 probes cross-hybridized to common bands, whereas CNL1 hybridized to a unique set. Both CNL1 and CNL3/4 probes produced complex hybridization patterns and high levels of polymorphism between BAT93 and JaloEEP558 (data not shown). We mapped six EcoRI polymorphic bands for CNL1, and three EcoRI and six HaeIII polymorphic bands for CNL3/4. All CNL1 bands co-segregated with marker 34g17-f1 except the CNL1Ec band, which mapped between markers 34g17-f1 and 76g17-f1 (Fig. 1b). Thus all CNL1 family members are located within 1 cM of each other on common bean LG B11. The CNL3/4 family members mapped to a 2.9-cM interval at a position distal to the CNL1 members, except for CNL3/4He, which mapped to a position 7.9 cM proximal to the CNL1 cluster (Fig. 1b).
AvrB resistance specificity is absent, but AvrRpm1 resistance specificity is present, in common bean
To test whether common bean possesses R genes with AvrRpm1 and/or AvrB recognition specificities, we inoculated a large set of wild and cultivated common bean genotypes of various geographical origins with P. syringae pv phaseolicola (Psp) strains expressing either AvrB or AvrRpm1. After infiltration of cotyledonary leaves, 17 out of the 36 tested genotypes displayed a resistant phenotype in response to Psp(avrRpm1), indicating that at least some common bean varieties can recognize AvrRpm1. Interestingly, six additional genotypes displayed a reduced (intermediate) resistance phenotype, with water soaking at the center, and HR confined near the main veins (Table 1). In contrast to the high frequency of AvrRpm1 recognition among common bean genotypes, all tested genotypes were susceptible to Psp(avrB), suggesting that specific recognition of AvrB is absent in common bean.
AvrRpm1-specific resistance maps to the Co-2 region in common bean
AvrRpm1 resistance specificity maps to the Rpg1 locus in soybean (Ashfield et al., 1995). We therefore tested whether AvrRpm1 recognition mapped to the Rpg1 syntenic region in common bean (i.e. the Co-2 resistance locus). To this end, two NILs for the Co-2 anthracnose resistance gene (P12R and P12S) were used (see Materials and Methods). Three days after inoculation with Psp(avrRpm1), Cornell49242, which was the source of the Co-2 anthracnose resistance gene for the P12R NIL, showed clear susceptibility symptoms (water soaking and maceration at the infection point), while P12S, which carries the susceptible co-2 allele, displayed a strong HR, indicative of resistance (Fig. 3a). Significantly, Psp(avrRpm1) induced an intermediate phenotype in P12R, with maceration at the center, and HR limited to areas adjacent to main veins (Fig. 3a). As NILs are, theoretically, genetically identical except in the introgressed region, the phenotype difference between P12R and P12S suggests that a gene (Rpsar-1) conferring resistance to AvrRpm1 is located in the vicinity of co-2 in P12S. Furthermore, the intermediate phenotype of P12R coupled with the susceptibility of Cornell49242 suggests an effect of the ‘Processor’ genetic background (the recurring parent for the P12S line). Analysis of bacterial growth within leaves confirmed the phenotype observations, with bacterial growth in P12R being intermediate between that of P12S and Cornell49242 (Fig. 3b).
Hybridization experiments using CNL1 or CNL3/4 as probes showed that the CNL1 hybridization pattern is entirely conserved between P12R and Cornell49242, except for one band that is specific for P12R (red asterisks in Fig. 3c), whereas P12S has a completely different hybridization pattern (Fig. 3c). This strongly suggests that all CNL1 subfamily members were introgressed from Cornell49242 to P12R, together with Co-2. By contrast, the CNL3/4 hybridization pattern is entirely conserved between P12S and P12R, whereas Cornell49242 has a completely different hybridization pattern, other than the nonpolymorphic bands (Fig. 3c), suggesting that no CNL3/4 members were introgressed from Cornell49242 to P12R. Together, these results further confirm that the CNL1 and CNL3/4 subfamilies are differentially distributed along the Rpg1 syntenic region, and, most importantly, suggest that both Co-2 and Rpsar-1 genes are members of the CNL1 subfamily.
AvrRpm1 resistance specificity is encoded by two independent genes in the common bean JaloEEP558 genotype
To confirm the mapping of Rpsar-1 at the Co-2 locus, we scored a set of RILs derived from a cross between BAT93 (susceptible) and JaloEEP558 (resistant) for resistance to Psp(avrRpm1) using the leaf inoculation assay. Phenotypes were scored as fully resistant (R), intermediate (I), or susceptible (S). The phenotypic distribution of the 179 RILs (92 R, 46 I, and 41 S), conformed to a ratio of 2R : 1I : 1S, suggesting that resistance was controlled by two independent genes, one with a strong effect and the second with an intermediate effect. The R gene with intermediate effect (Rpsar-1) mapped to the end of LG B11, co-segregating with the SCH20 marker and with CNL1 family members (Fig. 1b). The second R gene, referred to as Rpsar-2 and responsible for the fully resistant phenotype, mapped to the end of LG B8. To more precisely map Rpsar-2, additional available markers Leg741 and Leg451 (Hougaard et al., 2008) and two SCAR markers (SAS13 and Dj1kSCAR) known to be linked to the Co-4 anthracnose resistance gene (Adam-Blondon, 1994; Young et al., 1998) were scored on the 179 RILs. These analyses revealed that Rpsar-2 is located at the end of LG B8, 4.9 cM away from SAS13 and distal to all tested markers (Fig. 1c). These results differ from the results obtained from the NILs in that the NIL data indicated a gene of strong effect for resistance to Psp(avrRpm1) at the co-2 locus, and a gene of intermediate effect elsewhere in the genome. These differences probably stem from the use of different parents in these mapping populations (see the Discussion section).
The type III effector AvrRpt2 suppresses recognition of AvrRpm1 in common bean
The P. syringae type III effector AvrRpt2 suppresses recognition of AvrB and AvrRpm1 in both A. thaliana and soybean (Mackey et al., 2003; Ashfield et al., 2004). To test whether AvrRpm1 recognition specificity is also altered by AvrRpt2 in common bean, we co-inoculated Psp(avrRpm1) and Psp(avrRpt2) into pods of 25 RILs that were susceptible to Psp(avrRpt2), and displayed either a full resistance phenotype or an intermediate phenotype after inoculation with Psp(avrRpm1) in leaves. When a 1 : 1 mixture of Psp(avrRpm1) and Psp(avrRpt2) was inoculated, an HR reaction was clearly visible 120 h post inoculation (Fig. 4), suggesting that AvrRpt2 is not able to suppress recognition of AvrRpm1 in common bean. However, when Psp(avrRpt2) was inoculated and then 2 h later Psp(avrRpm1) was inoculated at the same inoculation point, the HR phenotype was completely lost. This was not observed after inoculation with Psp(avrRpt2::Ω) instead of Psp(avrRpt2), confirming that AvrRpt2 is responsible for loss of the resistance phenotype. These results suggest that the suppression of AvrRpm1 resistance specificity by AvrRpt2 in common bean is time dependent. We observed suppression of HR in all tested RILs, regardless of their phenotype after inoculation with Psp(avrRpm1) (i.e. resistant or intermediate), indicating that recognition of AvrRpm1 by both Rpsar-1 and Rpsar-2 is suppressed by AvrRpt2.
RIN4 homologs contain two highly conserved domains including AvrB and AvrRpt2 target sites
In A. thaliana, Mackey et al. (2003) showed that AvrRpt2 interferes with RPM1 function by inducing RIN4 disappearance. Whole-genome data mining indicates that RIN4 homologs exist in various plant species. Interestingly, analysis of common bean EST sequences showed that two RIN4 homologs are expressed in common bean. Phylogenetic analysis revealed two rounds of duplication of RIN4 homologs leading to two RIN4 paralogs in common bean and four in soybean (Fig. 5a). An analysis of nucleotide substitution rates at silent sites (Ks) showed that these two rounds of duplication are consistent with the two whole-genome duplication events that occurred in legumes, respectively before and after the divergence of Phaseolus and Glycine (Shoemaker et al., 2006). Indeed, the mean Ks value for (Gm_RIN4_b and Gm_RIN4_a) and (Gm_RIN4_c and Gm_RIN4_d) was 0.16 ± 0.06, in agreement with previous analyses of gene duplicates derived from the most recent whole-genome duplication event in soybean, where Ks = 0.149 (Schlueter et al., 2006). Similarly, a mean Ks value of 0.64 ± 0.06 was estimated for the older RIN4 duplication event, in agreement with previous results for homoeologous gene pairs from the first polyploidy event in soybean, where Ks ≈ 0.59 (Schmutz et al., 2010). Together, these results suggest that all homoeologous versions of RIN4 homologs were maintained after polyploidization in soybean and common bean. Significantly, alignment of these RIN4 homologs revealed two highly conserved domains, including the previously identified AvrB binding site (BBS), (Kim et al., 2005a; Desveaux et al., 2007) and two previously identified AvrRpt2 cleavage sites (RCS1 and RCS2) (Chisholm et al., 2005; Kim et al., 2005a) (Fig. 5b), suggesting that these RIN4 homologs may be targets of AvrB and AvrRpt2.
Rpg1-related and RPM1-related CNLs coexist in a homoeologous region of the Rpg1 locus
Soybean and common bean share an ancient whole-genome duplication estimated to have occurred 50–60 Mya (Shoemaker et al., 2006). The Rpg1 homoeologous (H3) region derived from this event was partially sequenced in soybean and was shown to possess three CNL sequences (Innes et al., 2008), referred to as gm132_155, gm132_159, and gm132_161. In common bean, we have sequenced two overlapping BAC clones (c. 240 kb) corresponding to this ancient duplicated region. We mapped this contig to the end of LG B5 in the common bean genome (Fig. 1a). After annotation, we found one CNL (pva120g2_H3) collinear to and sharing c. 57% amino acid identity with the three aforementioned soybean CNLs (Fig. 2b). All of these H3 CNL sequences from either soybean or common bean share < 27% amino acid identity with Rpg1-b. The complete soybean genome sequence is now available (Schmutz et al., 2010), and TBLASTN analysis using Rpg1-b as a query against the soybean whole-genome shotgun sequence assembly revealed one Rpg1-related CNL sequence (gm132_307) located only c. 650 kb from gm132_161 on GmChr06. To assess the evolutionary relationships among these CNLs, we performed phylogenetic analysis using the NB domain of all CNLs from the Rpg1 homoeologous regions of soybean and common bean, all CNL sequences identified in the A. thaliana genome, all CNL sequences identified in the common bean genome, and a representative set of CNLs from Medicago truncatula (Ameline-Torregrosa et al., 2008). gm132_307 fell into the Rpg1 clade, suggesting that Rpg1-related sequences were present at this locus before the ancient legume whole-genome duplication (Fig. 2a). Surprisingly, this analysis also demonstrated that RPM1 is the A. thaliana sequence most closely related to gm132_155, gm132_159, and gm132_161 in soybean, and pva120g2_H3 in common bean. Moreover, TBLASTN analysis using RPM1 as a query against the soybean whole-genome shotgun assembly identified gm132_155 as the soybean sequence most closely related to RPM1, with 32% amino acid identity. Together, these results indicate that RPM1-related and Rpg1-related CNL coexist within c. 650 kb on GmChr6 (an alignment of these sequences is presented in Fig. S1). The presence of RPM1-related CNL in the H3 region of both soybean and common bean suggests that RPM1-related CNLs were present at this locus before the divergence between Phaseolus and Glycine, 20 Mya (Lavin et al., 2005).
The RPM1-related and Rpg1-related CNLs are located in regions sharing synteny relationships between A. thaliana and legumes
In A. thaliana, the two sequences most closely related to soybean Rpg1-b (At3g14460 and At3g14470) are clustered in a region located only 2.6 Mb away from RPM1 on AtChr3 (Fig. 2b). This physical proximity of Rpg1-related sequences and RPM1 in A. thaliana parallels our observations that Rpg1- and RPM1-related CNLs are also physically close in soybean. Thus, we wondered if those two regions from soybean and A. thaliana could be syntenic regions. Interestingly, microsynteny was observed between the Rpg1 locus and the AtChr3 region located between the two Rpg1-related sequences (At3g14460 and At3g14470) and RPM1 (Fig. 2b). We have found another region sharing microsynteny with the Rpg1 locus on AtChr5. These two regions from AtChr3 and AtChr5 are homoeologous regions derived from the A. thaliana whole-genome duplication dated from 24 to 40 Mya (Fig. 2b) (Blanc et al., 2003).
We have mapped the soybean Rpg1 syntenic region at the Co-2 resistance locus in common bean. Our results indicate that the CNL1 and CNL3/4 subfamilies are linked, but physically separated from each other in common bean (Figs 1 and 3c), while they are intermixed in soybean, suggesting that tandem duplications have prevailed in common bean while soybean has undergone a more complex process. Furthermore, the absence of the CNL2 subfamily in common bean indicates that soybean and common bean have diverged in the qualitative content of Rpg1-related CNL (Innes et al., 2008). These differences in both CNL distribution and content between closely related species at a single locus parallel the striking differences observed between more distant plant species at the genome level (McHale et al., 2006; Ameline-Torregrosa et al., 2008). Indeed, dozens of Rpg1-related CNLs in common bean and soybean correspond to only five CNLs from a single clade in Medicago, which is itself included in a superclade that expanded in legumes, but not in A. thaliana (Fig. 2a) (Ameline-Torregrosa et al., 2008).
Given that two genes located at the soybean Rpg1 locus are involved in specific resistance to AvrB and AvrRpm1 (Ashfield et al., 1995), and that there is rapid birth and death of CNL genes at this locus in soybean and common bean (Innes et al., 2008), we wondered if these specific resistances appeared before the divergence between soybean and common bean c. 20 Mya (Lavin et al., 2005). Our results show striking similarities between soybean and common bean. First, the resistance specificities to AvrB and AvrRpm1 are genetically separable in common bean, as observed in soybean (Ashfield et al., 1995). Secondly, indirect evidence for a role of a RIN4 homolog in the recognition of AvrRpm1 exists in common bean, as observed in soybean (Ashfield et al., 2004). However, two major differences exist between common bean and soybean in terms of specific resistance to AvrB or AvrRpm1
Common bean lacks an AvrB-specific resistance gene
A first difference between common bean and soybean is that, while resistance to AvrB is encoded by Rpg1-b in soybean (Ashfield et al., 2004), we failed to identify any specific resistance after inoculation of a large set of common bean genotypes of various geographical origins with Psp(avrB) (Table 1), suggesting that common bean is unable to specifically recognize AvrB. Therefore, the ability to recognize AvrB presumably emerged in soybean after Phaseolus and Glycine diverged c. 20 Mya (Lavin et al., 2005) or, alternatively, was lost in common bean after this time-point. Interestingly, the common bean region collinear to Rpg1-b is devoid of CNL genes (Fig. 2b). Thus, the absence of AvrB specificity is hypothetically a result of a complete absence of an Rpg1-b homolog in common bean, suggesting that the birth and death process observed between common bean and soybean at the Rpg1 locus could be responsible for this polymorphism. Although avrB was originally cloned in a P. syringae strain attacking soybean (Tamaki et al., 1988), the majority of avrB homologs found to date were isolated from P. syringae strains attacking common bean (http://pseudomonas-syringae.org/) (Mansfield, 2009). More precisely, three of these avrB homologs (avrB2, avrB4-1, and avrB4-2) are present in the genome of the Psp 1448AN strain used in our experiments. This strain was shown to promote susceptibility in a panel of 37 common bean genotypes of diverse geographical origin (Table 1), in agreement with previous inoculation on eight additional bean genotypes (Taylor et al., 1996), suggesting that common bean cannot recognize any of those avrB homologs. The existence of multiple avrB homologs in strains attacking common bean, coupled with the finding that common bean cannot recognize AvrB, suggests an adaptation of the pathogen to its host, as previously shown for the interaction between common bean and Colletotrichum lindemuthianum (Geffroy et al., 1999). In the absence of resistance, AvrB was shown to promote virulence in soybean (Ashfield et al., 1995; Ong & Innes, 2006), and more hypothetically in A. thaliana (Nimchuk et al., 2000). Thus, it is tempting to speculate that the absence of selection pressure could have favored avrB diversification to promote virulence in P. syringae strains attacking common bean. The total absence of an R gene against AvrB in common bean is intriguing. Why would such resistance be entirely absent from a plant species that is confronted with a pathogen carrying avrB homologs? Riely & Martin (2001) proposed that a genetic bottleneck during the course of domestication of wild Lycopersicon esculentum could explain the total absence of Pto orthologs in L. esculentum. Here, we can rule out a genetic bottleneck as a result of domestication because we used a large set of common bean genotypes including wild accessions from various geographical origins, suggesting that AvrB resistance specificity is absent in both wild and cultivated genotypes (Table 1). Another explanation could be that common bean was not infected with P. syringae strains carrying avrB before its relatively recent domestication (c. 5000 yr ago); however, the presence of avrB homologs in the chromosomes of different strains attacking common bean (http://pseudomonas-syringae.org/) suggests a more ancient origin of avrB in these strains, arguing against this hypothesis. Alternatively, AvrB resistance specificity could exist in common bean, but with a frequency not sufficiently high to be identified in the set of genotypes screened.
AvrRpm1 resistance specificity is encoded by two genes in common bean
A second striking difference is that in common bean two different R genes (Rpsar-1 and Rpsar-2) can independently be responsible for resistance to AvrRpm1, while in soybean a single gene, Rpg1-r, was previously identified in two tested soybean genotypes (Ashfield et al., 1995). These results are supported by previous observations that an AvrRpm1 homolog from P. syringae pv pisi (AvrPpiA1), with an amino acid sequence nearly identical to AvrRpm1, induced a resistance phenotype with digenic segregation in an F2 progeny of common bean (Dangl et al., 1992; Fillingham et al., 1992). Intriguingly, AvrRpm1 induced an intermediate phenotype in P12R, which is a putative rpsar-1-Rpsar-2 genotype (Fig. 3), while rpsar-1-Rpsar-2 RILs from the BAT93xJaloEEP558 progeny displayed full resistance phenotypes. This phenotypic polymorphism suggests that different functional alleles may exist in common bean for Rpsar-1 and Rpsar-2, or, alternatively, that a more complex effect of genetic background modulates avrRpm1-induced responses. Our results mirror recent work by Eitas et al. (2008) showing that TAO1 (target of AvrB operation) mediates a weak resistance phenotype to AvrB independently of RPM1. Interestingly, they proposed that the presence of more than one R gene may provide greater evolutionary capacity, by allowing R proteins to continually fine-tune their responses to a single pathogen effector, while maintaining an intermediate level of disease resistance.
We mapped Rpsar-1 at one end of LG B11 in the region syntenic to Rpg1-r, suggesting that this resistance specificity appeared before Phaseolus and Glycine diverged c. 20 Mya, and was maintained in both genera. This, to our knowledge, would be the first example of specific resistance conservation between species on such a time scale. Fine mapping of Rpsar-1 in comparison to CNL1 and CNL3/4 subfamilies in both RILs and NILs, however, shows that Rpsar-1 co-localizes with the CNL1 subfamily, suggesting that it is encoded by a CNL1 sequence (Figs 1 and 3c), while Rpg1-r belongs to the CNL4 subfamily (T. Ashfield & R. W. Innes, unpublished). This result suggests that the ability to recognize AvrRpm1 was maintained in different Rpg1-related CNL subfamilies through a recombination event between a CNL1 sequence and a CNL4 sequence after the divergence between common bean and soybean (Fig. 6; models B1 and B2) or, alternatively, that the ability to recognize AvrRpm1 arose independently in these two lineages (Fig. 6; model A). Resolving these alternative hypotheses will require identification and analysis of the functional Rpsar-1 and Rpg1-r genes.
It is noteworthy that both Rpsar-1 and Rpsar-2 map in the vicinity of R genes that confer resistance to the fungus C. lindemuthianum (Geffroy et al., 1998; Melotto et al., 2004). Tight linkage between R genes conferring resistance against fungal and bacterial pathogens has also been observed in other species such as A. thaliana (Narusaka et al., 2009). In common bean, these two loci are located at one end of LG B11 (Co-2) and LG B8 (Co-4) (Fig. 1). Furthermore, we have mapped the Co-2 homoeologous region (H3) containing one RPM1-related CNL at the end of LG B5, (Figs 1 and 2). Taken together, these results reinforce the previously proposed idea that chromosome ends are favorable niches for R gene proliferation in the common bean genome (Geffroy et al., 2008, 2009; David et al., 2009).
Potential involvement of a RIN4 homolog in the recognition of AvrRpm1 in common bean
In A. thaliana, Mackey et al. (2003) demonstrated that AvrRpt2 suppresses RPM1 function by inducing RIN4 disappearance. AvrRpt2 also suppresses Rpg1-b function in soybean (Ashfield et al., 2004), as well as Rpsar-1 and Rpsar-2 functions in common bean (Fig. 4), indicating that AvrRpt2 interferes with the recognition pathways induced by AvrB and AvrRpm1 in both A. thaliana and legumes. Interestingly, two RIN4 homologs are expressed in common bean, and four in soybean (Fig. 5). Alignment of AtRIN4 with its putative orthologs in various plant species, including common bean and soybean, revealed highly conserved domains (Fig. 5), including the two previously identified AvrRpt2 cleavage sites (RCS) and the AvrB binding site (BBS) (Chisholm et al., 2005; Kim et al., 2005a), indicating that RIN4 orthologs (and particularly PvRIN4s and GmRIN4s) are putative virulence targets of both AvrRpt2 and AvrB. This supports previous suggestions that a RIN4 homolog is involved in Rpg1-b-mediated AvrB recognition in soybean (Ashfield et al., 2004; Ong & Innes, 2006). Mackey et al. (2003) showed that AvrRpt2-induced RIN4 disappearance is time-dependent in A. thaliana, and that high levels of RIN4 inhibit RPM1 loss of function induced by AvrRpt2. Our observations that AvrRpt2 prevents the HR phenotype induced by Rpsar-1 or Rpsar-2 in a time-dependent manner (Fig. 4) are consistent with a model in which RIN4 must be degraded to prevent Rpsar-1 and Rpsar-2 activation. However, the RCS and BBS domains are widely distributed in evolutionarily distant plant species, and many RIN4-unrelated proteins contain both RCS and BBS domains in A. thaliana (Chisholm et al., 2005; Kim et al., 2005a; Desveaux et al., 2007). It is thus possible that Rpsar-1, Rpsar-2, Rpg1-r, and RPM1 monitor the status of different targets, all of which are, however, targeted by both AvrRpm1 and AvrRpt2.
Models for the evolution of AvrB and AvrRpm1 specificity
Ashfield et al. (2004) have demonstrated that Rpg1-b and RPM1 belong to distinct evolutionary lineages of CNL that diverged before the monocot–dicot split (reviewed by McDowell, 2004). They proposed that the ability to recognize AvrB and AvrRpm1 appeared independently in A. thaliana and soybean through convergent evolution. In A. thaliana, the TAO1-mediated resistance response to AvrB is RIN4 independent, and is encoded by an RPM1-unrelated sequence from the TNL family (Eitas et al., 2008). Therefore, convergent evolution was probably responsible for the emergence of TAO1 and RPM1 in A. thaliana. Similarly convergent evolution could also be responsible for the emergence of Rpg1-r in soybean and Rpsar-1 and Rpsar-2 in common bean. Interestingly, our genomic and phylogenetic data show that RPM1- and Rpg1-related CNLs are physically close in both A. thaliana and soybean (Fig. 2). Moreover, the regions containing RPM1- and Rpg1-related CNLs in A. thaliana, common bean and soybean share partial syntenic relationships (Fig. 2), suggesting that RPM1- and Rpg1-related CNLs were present at the same locus in a common ancestor of A. thaliana and legumes. Under this assumption, a scenario of convergent evolution would imply that the recognition of AvrB and/or AvrRpm1 arose three times independently in the same genomic region in A. thaliana (RPM1), common bean (Rpsar-1) and soybean (Rpg1-b plus Rpg1-r).
The repeated evolution of similar specificities in the same CNL cluster suggests that these CNLs may be predisposed to evolve AvrB and/or AvrRpm1 recognition ability. Because recognition of these effectors by RPM1 is known to require a physical association between RPM1 and RIN4 (Mackey et al., 2002), and because these effectors induce modification of RIN4, we propose that physical association with unmodified RIN4 predisposes CNL proteins to evolve the ability to detect modifications to RIN4. In this model, evolution of specificity is a two-step process. First, the CNL protein must evolve the ability to associate with an unmodified effector target (e.g. RIN4), and secondly it must evolve the ability to detect modification of that target. As effector targets are likely to be important regulators of disease resistance (Block et al., 2008), they are likely to be well conserved during plant evolution, and there will be selection on CNL proteins to maintain the ability to interact with effector targets. Thus, once an association with an effector target has evolved, this may be an evolutionarily stable trait. However, there will be strong selection on pathogens to evolve novel ways of modifying these effector targets in order to escape recognition. Such novelty would then drive diversification of the domain(s) of the CNL involved in detecting modifications, while maintaining the domain(s) required for interaction with the effector target. Following this model, Rpg1-b, Rpg1-r and Rpsar-1 are derived from a common ancestral CNL with the ability to interact with RIN4, but have evolved the ability to detect specific modifications of RIN4 independently (Fig. 6; model A).
The physical proximity of RPM1- and Rpg1-related CNLs in both A. thaliana and soybean could suggest an alternative model in which the ability to recognize AvrB and/or AvrRpm1 arose in a common ancestor of A. thaliana and legumes, and was further maintained in different CNL lineages through an ancient recombination event between RPM1- and Rpg1-related CNL sequences (Fig. 6; models B1 and B2). Recombination between NB-LRR sequences is thought to have played a major role in their evolution, leading to changes in copy number via unequal crossing-over in tandem arrays, as well as changes in specificity (Wicker et al., 2007; Nagy & Bennetzen, 2008).
The convergent evolution model and evolution by recombination model differ from a simple evolution model (Fig. 6; model C) in that the latter model would require the maintenance of duplicated genes with dual AvrB/AvrRpm1 specificity for tens of millions of years, which seems unlikely given the rapid evolution of R gene specificity. Distinguishing between the convergent evolution model and recombination model will require the identification of Rpsar-1, so that it can be compared directly with Rpg1-r. If the NBS regions of these two genes do indeed belong to different CNL clades, as our mapping data predict, but their LRR domains belong to the same clade, this would favor the recombination model. In contrast, if no evidence of recombination is found, then the convergent evolution model would be favored.
We acknowledge Steven B. Cannon (USDA ARS; USA) for providing us with the custom Perl scripts to generate synteny images. We are indebted to Daniel Debouck (Centro International de Agricultura Tropical, Colombia) and Eduardo C. Vallejos (University of Florida, USA) for providing us with seeds of wild common bean. We are grateful to Paul Billant (CNRS, Ecole-Polytechnique, France) for discussions on LOD score calculations. The research was supported by the Institut National de la recherche Agronomique, the Centre National de la Recherche Scientifique, the French Ministère de la Recherche and a grant from Genoscope/CEA/CNS to V.G. and by US National Institutes of Health grant R01-GM046451 and National Science Foundation Plant Genome Research Program grant DBI-0321664 to R.W.I.