Evolution of a complex disease resistance gene cluster in diploid Phaseolus and tetraploid Glycine

Abstract

We used a comparative genomics approach to investigate the evolution of a complex nucleotide-binding (NB)-leucine-rich repeat (LRR) gene cluster found in soybean (Glycine max) and common bean (Phaseolus vulgaris) that is associated with several disease resistance (R) genes of known function, including Rpg1b (for Resistance to Pseudomonas glycinea1b), an R gene effective against specific races of bacterial blight. Analysis of domains revealed that the amino-terminal coiled-coil (CC) domain, central nucleotide-binding domain (NB-ARC [for APAF1, Resistance genes, and CED4]), and carboxyl-terminal LRR domain have undergone distinct evolutionary paths. Sequence exchanges within the NB-ARC domain were rare. In contrast, interparalogue exchanges involving the CC and LRR domains were common, consistent with both of these regions coevolving with pathogens. Residues under positive selection were overrepresented within the predicted solvent-exposed face of the LRR domain, although several also were detected within the CC and NB-ARC domains. Superimposition of these latter residues onto predicted tertiary structures revealed that the majority are located on the surface, suggestive of a role in interactions with other domains or proteins. Following polyploidy in the Glycine lineage, NB-LRR genes have been preferentially lost from one of the duplicated chromosomes (homeologues found in soybean), and there has been partitioning of NB-LRR clades between the two homeologues. The single orthologous region in common bean contains approximately the same number of paralogues as found in the two soybean homeologues combined. We conclude that while polyploidization in Glycine has not driven a stable increase in family size for NB-LRR genes, it has generated two recombinationally isolated clusters, one of which appears to be in the process of decay.

Figure 1.

Distribution of CC-NB-LRR genes across the homeologous and orthologous sequences corresponding to the Rpg1b region in soybean. A, Species tree of the legume species included in this study. B, Alignment of predicted genes across Rpg1b homologous regions. Gray horizontal lines represent the available sequence, with gaps indicated by vertical blue lines. The sequence derived from the soybean whole genome sequencing project is indicated with a horizontal green line. Vertical red rectangles positioned on the horizontal lines represent predicted CC-NB-LRR genes, and vertical gray rectangles represent all other genes. Putative low-copy orthologous/homeologous genes are linked by blue lines, and where these relationships have been confirmed phylogenetically (Innes et al., 2008), a black letter is assigned to the gene set. Orthologous/homeologous intervals containing CC-NB-LRR genes in at least one of the plants sequenced are indicated by red letters over red double-ended arrows. Gtd, G. tomentella; Gmp, soybean accession PI96983; Gmw, soybean ′Williams82′; Pva, common bean Andean accession. H1, Glycine homeologue 1; H2, Glycine homeologue 2. This figure is adapted from Innes et al. (2008).

Figure 2.

Sequence exchanges between the NB-LRR paralogues are common in the CC and LRR domains while being comparatively rare within the NB-ARC. A, Positions of the 30 events detected by RDP analyses are shown. Each horizontal line represents a distinct event. One event (no. 6) resulted in the transfer of the complete NB-ARC domain between paralogues. The events are drawn relative to the W52d1_8 sequence, and the position in this sequence is shown on the x axis. Asterisks indicate the ends of sequence exchanges in which the actual break point position could not be determined by RDP. B, Frequency (y axis) of break points among all sequences plotted against nucleotide position in the alignment (x axis). Events are supported by at least two of the four methods utilized within RDP version 3.15 at P < 0.001.

Figure 3.

Bayesian phylogeny of NB-ARC domains derived from the Rpg1b region. Thick branches indicate posterior probabilities of 0.95 or greater. Sequences derived from soybean Williams 82 WGS scaffolds (7× draft assembly) are preceded by “GmwRF” or “GmwGap1.” The remaining sequences are BAC derived and are labeled with the BAC name and gene number corresponding to those described by Innes et al. (2008). # indicates a soybean or bean NB-ARC sequence from an intact NB-LRR paralogue (i.e. not a pseudogene). Note that sequences from two soybean accessions are included (with prefixes Gmw and Gmp), and putatively allelic pairs should not be confused with local duplications. The sequences shown in gray are derived from Gmw H1 but are outside the 1-Mb target region.

Figure 4.

Reconciled gene and species trees based on the optimized NB-ARC phylogeny. Squares at the nodes represent hypothesized duplication events within the evolutionary history of the sampled genes. Dashed lines and gray text represent lineages that theoretically exist(ed) as a result of hypothesized duplication events but are otherwise missing due to being unsampled or through gene loss. Extant genes and taxa are color coded as in Figure 3. Clades a to i represent nine ancestral NB-ARC lineages predicted to have persisted since before the Glycine/Phaseolus split. H1 and H2 refer to homeologues 1 and 2, respectively. The right-hand column provides gene names labeled with BAC name and gene number as described by Innes et al. (2008).

Figure 5.

Distinct patterns of selection detected in the CC, NB-ARC, and LRR domains. Graphs represent standardized nonsynonymous (dN) minus synonymous (dS) substitution rates (dN − dS) values calculated using the FEL method (Pond and Frost, 2005b) plotted against the position of each codon in the alignment. The y axis scale is limited from −1 to 1, although some standardized dN-dS values extend beyond this range. A, Standardized dN-dS values for all codons. B, Only those codons where nonneutral selection is supported by FEL (P < 0.05). C, Only those codons where nonneutral selection is supported by FEL (P < 0.05), SLAC (P < 0.05), and REL (BF > 50).

Figure 6.

Sites under positive (diversifying) selection are overrepresented in the predicted solvent-exposed surface of the LRRs. Residues under selection are indicated in the consensus sequences corresponding to the alignments used in the selection analysis. Sites under overall positive selection are shown in blue, and those under negative selection are shown in red. Where selection is supported by FEL (P < 0.05), the colored residue is shown in normal type; where the prediction is also supported by SLAC (P < 0.05) and REL (BF > 50), the residue is shown in boldface type. A specific amino acid is shown in the consensus sequence when it is present in at least 51% of the sequences represented at that location in the alignment (gaps excluded). x indicates all other sites in the alignment. A, CC domain. Underlined residues are predicted to form CCs (P > 95% and P > 89% for the first and second underlined regions, respectively). The predicted positions of amino acids within each CC heptad repeat are indicated with lowercase letters, with expected hydrophobic positions printed in bold. B, NB-ARC domain. Boxed residues indicate previously defined conserved motifs (from the N-terminal end: P loop, kin2, GLPL, RNBS-D, MHD). C, LRR domain. Individual LRRs are shown on separate lines. The predicted solvent-exposed face is boxed. D, Consensus sequence for intracellular LRRs. The region predicted to form part of a solvent-exposed face is underlined.

Figure 7.

Comparative modeling of the tertiary structures of the Rpg1b CC and NB-ARC domains suggests that the majority of residues under positive selection are located on the surface. A, Predicted tertiary structure of the Rpg1b NB-ARC domain modeled after the APAF-1 NB-ARC domain (PDB code 1Z6T). The NB, ARC1, and ARC2 subdomains are shown in red, yellow, and purple, respectively. Previously defined conserved motifs are shown in green. B, Predicted tertiary structure of the Rpg1b CC domain homodimer modeled after the MLA10 CC domain (PDB code 3QFL). The conserved EDVID motif is shown in orange. Residues with statistical support for positive selection in the NB-ARC and CC structures are indicated in blue in both panels (dark blue and boldface labels when supported by FEL, SLAC, and REL; light blue and regular labels when supported by at least FEL). Residue numbers correspond to positions in the full-length Rpg1b protein (Williams 82 allele).