Independent deletions of a pathogen-resistance gene in Brassica and Arabidopsis

Abstract

Plant disease resistance (R) genes confer race-specific resistance to pathogens and are genetically defined on the basis of intra-specific functional polymorphism. Little is known about the evolutionary mechanisms that generate this polymorphism. Most R loci examined to date contain alternate alleles and/or linked homologs even in disease-susceptible plant genotypes. In contrast, the resistance to Pseudomonas syringae pathovar maculicola (RPM1) bacterial resistance gene is completely absent (rpm1-null) in 5/5 Arabidopsis thaliana accessions that lack RPM1 function. The rpm1-null locus contains a 98-bp segment of unknown origin in place of the RPM1 gene. We undertook comparative mapping of RPM1 and flanking genes in Brassica napus to determine the ancestral state of the RPM1 locus. We cloned two B. napus RPM1 homologs encoding hypothetical proteins with ≈81% amino acid identity to Arabidopsis RPM1. Collinearity of genes flanking RPM1 is conserved between B. napus and Arabidopsis. Surprisingly, we found four additional B. napus loci in which the flanking marker synteny is maintained but RPM1 is absent. These B. napus rpm1-null loci have no detectable nucleotide similarity to the Arabidopsis rpm1-null allele. We conclude that RPM1 evolved before the divergence of the Brassicaceae and has been deleted independently in the Brassica and Arabidopsis lineages. These results suggest that functional polymorphism at R gene loci can arise from gene deletions.

The outcome of many plant–pathogen interactions is determined by disease-resistance (R) genes that enable plants to recognize invading pathogens and activate inducible defenses (1). A typical R gene allele encodes “race-specific” resistance to only one or a few strains of a single pathogen species (2). R gene loci are functionally polymorphic within a plant species and encode alternate alleles that either recognize different strains of the same pathogen or do not recognize any tested pathogen. Recent molecular studies have revealed that R genes often reside in complex loci consisting of the R gene and tightly linked homologs (3–8). These complexes can exist in both disease-resistant and disease-susceptible plant genotypes. The allelic diversity and functional specialization at R gene loci inspired H. H. Flor’s “gene-for-gene” hypothesis: An R gene specifies resistance only if the pathogen expresses a corresponding avirulence (avr) gene (9). If either component is nonfunctional, then the plant is unable to activate resistance responses. This model has successfully described a wide variety of plant–pathogen associations; however, we currently know little about the molecular basis and origin of R gene polymorphism.

The most common molecular interpretation of the gene-for-gene hypothesis is that R genes encode specialized receptors that recognize the direct or indirect products (elicitors) of the corresponding avr genes (10). One implication of the receptor–elicitor model is that functional polymorphism at R loci could arise from gain-of-function mutations that enable the plant to recognize novel pathogen variants. Maintenance of multiple alleles or multiple linked genes with different recognition capabilities at an R locus would enable a host population to defend itself against the corresponding pathogens. R gene polymorphism could also, in principle, arise from loss-of-function mutations in R genes. Nonfunctional R gene alleles could be maintained if there is a cost of resistance in the absence of pathogen selection (11). Nonfunctional alleles could also serve as a repository of divergent sequences that could be contributed to related genes by recombination, thereby accelerating the evolution of novel R genes (12).

Recent cloning of R genes against diverse pathogens provides the tools for comparative analysis of R gene alleles within and between species, which will provide insight into the evolutionary history of R genes. We describe the structure of functional and nonfunctional alleles of the RPM1 (resistance to Pseudomonas syringae, pathovar maculicola) bacterial resistance gene in Arabidopsis thaliana and a related crop species, Brassica napus. RPM1 was initially identified in A. thaliana accession Col-0 through its ability to confer resistance to P. syringae isolates expressing either avrRpm1 (13) or avrB (14). The avrB and avrRpm1 avirulence genes are sequence unrelated, thus RPM1 enables dual-specificity resistance. RPM1 was isolated by map-based cloning and was shown to encode a protein with a putative amino-terminal leucine zipper, a consensus nucleotide binding site (NBS), and 14 C-terminal leucine-rich repeats (LRRs) (15). The RPM1 protein is thus a member of the largest class of R proteins functionally characterized to date, the so-called NBS-LRR class (16). Previous analyses suggested that RPM1 functional polymorphism in Arabidopsis arose from an intraspecific insertion or deletion of the RPM1 gene (15). Here we provide molecular evidence that the evolution of RPM1 predated the divergence of the Brassicaceae, and that independent deletions of RPM1 have occurred in both Arabidopsis and Brassica.

MATERIALS AND METHODS

Genetic Mapping in B. napus.

Clones from A. thaliana and B. napus were used to probe filters with genomic DNA from 92 double haploid lines of the N-o-72–8 mapping population of B. napus (17). Southern hybridization and genetic linkage analysis were carried out as described (17), except that the final filter washes were carried out at a lower stringency [2 × SSC (standard saline citrate; 1 × SSC = 0.15 M sodium chloride/0.015 M sodium citrate, pH 7), 0.1% SDS] for the A. thaliana probes and a higher stringency (0.1 × SSC, 0.1% SDS) for the B. napus probes.

Isolation and Sequence Analysis of Arabidopsis and Brassica Clones.

To clone the Nd-0 rpm1-null locus, a Nd-0 genomic library was constructed (Sau3AI partial-digested, Lambda Dash II, Stratagene). This library was screened with the GTP gene, which is ≈700 bp upstream of RPM1 (Fig. 1A). Positive phage were subsequently rescreened by PCR with T3 or T7 primers and an oligonucleotide (ATGAAAGTATGATCGAG) corresponding to the 3′ end of the GTP gene. The largest PCR product of approximately 2.6 kb was subcloned (pAT; Invitrogen) and sequenced by nested deletions (Promega Erase-A-Base). Oligonucleotide primers were then designed to amplify a 560-bp fragment containing the rpm1-null allele from other susceptible ecotypes (RPM1-5S1: GATTATAGGAAGAACAAAGATG, RPM1-3N2: AATACCAAGGGACACCAAAGAG). To amplify the 5′ and 3′ flanking regions from resistant ecotypes, we used, respectively, RPM1-5S1 in combination with the BamRev primer (AACACCTAGGTAAAGGTGG) and the RPM #2 primer (GATGAAACATCTGACAGAAG) in combination with RPM1-3N2.

Figure 1.

Structure of functional and null RPM1 alleles in A. thaliana. (A) Schematic comparison of RPM1 and rpm1-null structure. The RPM1 locus in resistant and susceptible accessions is closely flanked by two ORFs. GTP and M4. In all susceptible accessions that have been examined, the RPM1 gene has been replaced by a 98-bp sequence of unknown origin. (B) Sequence of the null allele in susceptible Arabidopsis. The sequence of rpm1-null from Nd-0 is aligned with conserved regions that flank the 5′ and 3′ breakpoints of the replacement in Col-0. Conserved nucleotides are capitalized, and the 98-bp filler is emphasized by italics. A direct repeat in Col-0 at the 3′ breakpoint with Nd-0 is underlined.

Brassica genomic libraries derived from British winter (N-o-9) and Canadian spring (N-o-1) B. napus cultivars (18) were screened with RPM1 and GTP probes at 62° [5 × SSC, 5 × Denhardt’s solution (1 × Denhardt’s solution = 0.02% polyvinylpyrrolidone/0.02% Ficoll/0.02% BSA), 0.1% SDS]. Filters were washed twice at 60° in 4 × SSC, 0.1% SDS, and once at 62° in 2 × SSC, 0.1% SDS. Phage DNA was isolated and hybridized with probes for RPM1, GTP, and the M4 gene (Fig. 1A). This allowed us to distinguish between Brassica RPM1 and rpm1-null clones. Representative RPM1 homologs from N-o-109 and N-o-101 were sequenced by fluorescent sequencing by using nested deletion analysis. Gaps were filled by primer walking.

To clone Brassica rpm1-null alleles, phage clones hybridizing to GTP and M4 but not to RPM1 were isolated and characterized by restriction fingerprinting as a prelude to DNA sequencing. Sequence determination of the B. napus rpm1-null locus 1S was achieved initially through sequencing of random GTP-hybridizing clones. Based on this information, we isolated other Brassica null loci by PCR of restriction-mapped phage subclones by using nested gene-specific primers to the Brassica homolog of GTP (G1: ATGATGAGTCAATGGGAGG, G2: GGAGATTATGATTTCAAAG). These primers were designed against nucleotides invariant between three Brassica GTP homologs and the Arabidopsis gene and were used to screen genomic phage in combination with flanking T3 and T7 primers or primers derived from the 3′-end of a sequenced Brassica rpm1-null allele (1S.RV: GGGAAGACAAAACGTGAC, 1S.RVB: TTTCATGGGACATCGTATCATC). The PCR products were either subcloned into Bluescript or directly sequenced by primer walking. All phage clones were genetically mapped by restriction fragment length polymorphism (RFLP) analysis, as described above.

Library construction, cloning, and chain termination DNA sequencing were as described in standard protocols (19). Sequence alignments and homology searches were performed with internet-based facilities: clustalw (http://dot.imgen.bcm.tmc.edu) and blast (http://www.ncbi.nlm.nih.).

RESULTS

The RPM1 Gene Is Completely Absent in Susceptible Arabidopsis Accessions.

DNA blot analyses originally demonstrated that the RPM1 gene is absent in five A. thaliana accessions that are susceptible to P. syringae pv. tomato (Pst) DC3000 expressing avrRpm1 (15). We cloned and sequenced the rpm1-null locus from the susceptible Nd-0 accession and found that 3.7 kb of nucleotide sequence is absent in the Nd-0 allele, relative to the functional Col-0 RPM1 allele (Fig. 1). This region includes the entire RPM1 ORF, 350 nucleotides upstream from the initiating methionine, and 532 nucleotides downstream from the termination codon (182 nucleotides after the putative polyA addition site). In place of this segment, we found a 98-bp filler that encodes no ORFs and has no database matches. Alignment of the Col-0 and Nd-0 regions that flank the recombination breakpoints revealed point substitutions and several insertion/deletions (Fig. 1B). An imperfect repeat of 9 bp occurs in Col-0 at the 3′ breakpoint between the Col-0 and Nd-0 alleles (Fig. 1B).

We determined the degree of sequence divergence between RPM1 and rpm1-null alleles from several accessions of diverse geographical origin: Aa-0, Col-0, Fe-1, Ler-0, Nd-0: Germany; Oy-0: Norway; Ws-0: Russia; Bs-0: Switzerland; Mt-0: Libya. Col-0, Oy-0, Ler-0, and Ws-0 are disease resistant; all others are disease susceptible (13, 20). The sequence of the 5′ and 3′ breakpoint regions in the four resistant accessions is essentially identical, with the exception of rare single nucleotide substitutions. Similarly, the five susceptible accessions contained rpm1-null alleles that were essentially identical to the rpm1-null allele in Nd-0, with the exception of one substitution in the Mt-0 allele (not shown).

To confirm that disease susceptibility in Nd-0 was caused by the absence of RPM1, we transformed Nd-0 plants with the cosmid 1FB (15), which contains RPM1 from the resistant accession Col-0. Three T2 lines segregating for a single T-DNA insertion were tested for functional complementation of resistance by inoculation of individuals with DC3000 expressing either avrRpm1 or avrB. Both isolates induced a hypersensitive response indicative of RPM1-mediated disease resistance in all three transformants (not shown). Typical water-soaked lesions and disease symptoms developed on plants challenged with avirulent DC3000. Measurements of in planta bacterial growth confirmed that the RPM1 transgene conferred levels of resistance that were comparable to wild-type Col-0 (not shown). We also determined that RPM1 is sufficient for resistance in transgenic lines of the susceptible accessions Mt-0 and Fe-1 (not shown).

The Brassica RPM1 Locus Is Syntenic with RPM1 from Resistant Arabidopsis.

The sequence comparisons of functional and null RPM1 loci suggested that RPM1 polymorphism resulted from either a single insertion or a single deletion of RPM1 during Arabidopsis evolution. To distinguish between these competing hypotheses, we compared the organization of RPM1 and flanking genes in Arabidopsis and a related crucifer, B. napus (oilseed rape). Diploid Brassica genomes are thought to have evolved from a hexaploid ancestor (21–23), and B. napus is a amphidiploid hybrid of Brassica rapa (A genome) and Brassica oleracea (C genome) (24). Thus, each single-copy Arabidopsis gene would be expected to map to six loci in B. napus, arranged in three homoeologous pairs.

We identified only two RPM1 homologs in B. napus by hybridization with an Arabidopsis RPM1 coding sequence probe. Only two RPM1 RFLPs were detected in B. napus DNA digested with either of five restriction enzymes, and in each case the two RFLPs mapped to the same chromosomal loci (see below). Corresponding genomic clones (named 1A and 9N) were isolated, respectively, from libraries of spring (N-o-101) or winter (N-o-109) B. napus varieties. The hypothetical proteins encoded by 1A and 9N share 95% amino acid identity and each shares ≈81% identity with Arabidopsis RPM1 (Fig. 2). Like Arabidopsis RPM1, the B. napus RPM1 proteins contain a putative N-terminal leucine zipper, a consensus NBS, and 14 imperfect LRRs. Amino acid substitutions between Arabidopsis RPM1, 9N, and 1A are distributed relatively evenly over the protein. No clear evidence for positive diversifying selection (7) in the LRRs was found in comparisons of RPM1 homologs (data not shown).

Figure 2.

Structural comparison of hypothetical RPM1 proteins from Arabidopsis and B. napus. Putative functional motifs are emphasized in bold. The P-loop, kinase 2A, and kinase 3A motifs constitute a consensus NBS. Conserved domains 2 and 3 are present in all known NBS-LRR resistance genes, but their function is unknown. LRRs are designated by the highlighted XXLXLXX motif.

The 5′ flanking regions of 1A and 9N are noticeably diverged from each other and contain several insertion/deletions, the largest of which is a 255-bp insertion in 9N, relative to 1A (not shown). Comparison of the 5′ and 3′ flanking regions of the Arabidopsis RPM1 and 1A or 9N clones revealed little overall sequence conservation, with the exception of two potential 5′ cis regulatory elements immediately proximal to the RPM1 coding sequence. One motif is almost completely identical to a C/EBP binding site (25) (CCAAGTTGCAAATGTGAAAGCAGT, beginning 218 bp upstream of the RPM1 start codon).

9N and 1A were placed on an integrated genetic linkage map of B. napus (17) between RFLP markers pN148a & pO12e and pW108c & pO12c, on homoeologous chromosomes N1 (representing the A genome of B. rapa) and N11 (representing the C genome of B. oleracea), respectively (Fig. 3A). Surprisingly, each locus contains only one copy of RPM1.

Figure 3.

Location and structure of functional and null RPM1 loci in B. napus. (A) Schematic representation of the amphidiploid B. napus genome indicating the loci hybridizing to the Arabidopsis RPM1, M4, and GTP clones. Linkage groups N1 to N10 and N11 to N19 represent the A and C genomes, respectively. Hatched and filled areas of the arrows indicate regions of primary homeology between the A and C genomes. Unfilled areas represent regions for which primary homeology between the genomes has yet to be detected. (B) Schematic representation of RPM1 and rpm1-null loci, based on restriction mapping of phage clones. The locus names are listed on the left and the chromosomes to which they map are listed to the right. 1A and 9N contain RPM1, and the other three phage are rpm1-null. Ticks indicate either XhoI or XbaI restriction enzyme sites. The GTP, RPM1, and M4 genes are indicated above the figure. Equivalently shaded regions crosshybridize. Unshaded regions were not tested. (C) Sequence alignment at the 5′ breakpoint of sequence homology between B. napus RPM1 and rpm1-null. The 3′ ends of the GTP coding sequence are italicized, with stop codons emphasized in bold. Dashes represent gaps, and stars indicate sites that are identical among all five sequences. The arrow indicates the position at which similarity between the RPM1 and rpm1-null loci breaks down completely.

To assess collinearity between the B. napus and Arabidopsis RPM1 loci, we mapped two cDNA markers that closely flank RPM1 in Arabidopsis (Fig. 1). The GTP probe was derived from an ORF located 700 bp centromeric to Arabidopsis RPM1 that encodes a putative GTP-binding protein, and the M4 probe was derived from an ORF ≈4 kb telomeric that encodes a putative UDP-glucosyltransferase. Both markers are single copy in Arabidopsis (not shown) and are present in RPM1 and rpm1-null Arabidopsis accessions (Fig. 1). In B. napus, these probes mapped as cosegregating pairs to six loci, two of which correspond to the B. napus RPM1 loci (Fig. 3A). The six loci map within large homoeologous segments of the B. napus genome (26). Sequencing and hybridization analysis of the B. napus RPM1 phage clones confirmed synteny between the RPM1 loci in B. napus and resistant Arabidopsis (Fig. 3B). The distance between the end of the GTP gene and the beginning of 1A and 9N is 293 and 562 bp, respectively, compared with 479 bp in Arabidopsis. The distance between M4 and 1A or 9N could not be determined precisely by restriction mapping but is at least 3 kb in both cases. Thus, the order and approximate spacing of GTP, RPM1, and M4 is conserved between 1A, 9N, and Arabidopsis RPM1. This synteny suggests that the functional RPM1 allele represents the ancestral state of the locus in Arabidopsis, and that the Arabidopsis rpm1-null allele was derived by deletion at some time after the divergence of Arabidopsis and Brassica lineages.

Southern hybridization suggested that the four other GTP/M4 loci in B. napus (homoeologous chromosomes N3/N13 and N5/N15) did not contain any RPM1-hybridizing sequences (Fig. 3A). The absence of RPM1 homolog was confirmed by analysis of corresponding genomic phage clones (Fig. 3B). These phage clones hybridized to the GTP and M4 probes but not to RPM1 coding sequence probes. Three rpm1-null loci were sequenced to determine the 5′ breakpoint: 1/16, 1S, and 9W, that correspond, respectively, to chromosomes N5, N13, and N3 (Fig. 3B). Sequence alignments confirmed that the RPM1 gene was indeed absent from these loci. The breakdown of sequence similarity between the Brassica RPM1 and rpm1-null loci is gradual, beginning with several small insertion/deletions just downstream of the GTP stop codon (Fig. 3C). By ≈90 bp downstream of the GTP stop codon (arrow in Fig. 3C), the B. napus RPM1 and rpm1-null loci share no significant sequence identity. The breakpoints in the rpm1-null loci correspond to 210 bp and 475 bp, respectively, of 5′ flanking sequence from 1A and 9N. No similarity was detected between the Brassica rpm1-null alleles and the 3′ flanking regions of the Brassica RPM1 gene, indicating that a significant region of the 3′ flanking region is also absent from the rpm1-null loci. The 3′ breakpoint was not mapped further.

The 1S and 9W rpm1-null loci are highly similar (>95%). In contrast, the sequence of the 1/16 null locus diverged significantly from 1S and 9W at 138 and 152 nucleotides downstream of the GTP termination codon, respectively. 1S and 9W, but not 1/16, share significant similarity in a region immediately 3′ of GTP to a rice EST (RICS13430A), while 1/16 displays strong similarity to a valine tRNA and a small region of homology to RPM1.

There is extensive similarity between the B. napus rpm1-null alleles and the Arabidopsis rpm1-null alleles over the length of the GTP gene. Downstream of this gene, however, the Arabidopsis rpm1-null alleles share no significant similarity to any of the Brassica rpm1-null loci (data not shown). This indicates that the Arabidopsis and Brassica rpm1-null loci arose independently.

DISCUSSION

This study was designed to explore the molecular basis and origin of RPM1 functional polymorphism in Arabidopsis and a related crop species. We examined RPM1 structure in nine Arabidopsis laboratory accessions and found only two structures (haplotypes). The four disease-resistant accessions are almost identical to the reference Col-0 allele. In contrast, all five disease-susceptible accessions contain a null haplotype in which the entire RPM1 gene is replaced by a short “filler” sequence. A Col-0 RPM1 transgene restored resistance in three susceptible accessions. Thus, susceptibility in the examined accessions is caused by the complete absence of RPM1. The presence of only two haplotypes in the examined accessions suggests that RPM1 polymorphism in Arabidopsis can be explained by a single molecular event that could be either (i) an insertion of RPM1 in the progenitor of resistant accessions, or (ii) a deletion of RPM1 in the progenitor of susceptible accessions.

To infer the ancestral state of RPM1 in Arabidopsis and thereby distinguish between the above hypotheses, we examined RPM1 structure in B. napus. B. napus contains the Brassica A and C genomes, each of which contains three equivalents of an hypothesized Arabidopsis-like progenitor genome. Thus, if hypothesis i is correct, then we would expect B. napus to contain six copies of RPM1 between the GTP and M4 flanking markers. We indeed found two B. napus loci that are collinear with the GTP/RPM1/M4 gene organization found in the resistant Arabidopsis accessions. Sequencing of three null loci revealed two distinct molecular structures, on N3/N13 and N5, that were divergent from each other and from the Arabidopsis rpm1-null haplotype. Thus, we propose the model outlined in Fig. 4. RPM1 existed as a single copy gene between GTP and M4 before the divergence of Arabidopsis and Brassica progenitor species. The Arabidopsis rpm1-null haplotype arose from a subsequent deletion of RPM1 in the Arabidopsis lineage. The alternative model is that RPM1 was independently inserted between GTP and M4 in the Brassica and Arabidopsis genomes, which seems extremely unlikely. In Brassica, the GTP/RPM1/M4 locus was triplicated, and RPM1 was subsequently deleted from two loci (Fig. 4). The structural similarity and homoeologous positions of the 1S and 9W null loci suggest that these two null loci were derived from the same progenitor, whereas the divergent structure of 1/16 relative to 1S and 9W suggests that 1/16 was derived from a different deletion event.

Figure 4.

Model for the evolution of RPM1 in the Brassicaceae. The ancestral state of RPM1 is inferred from structural conservation between Arabidopsis and Brassica loci. Subsequent deletion events are hypothesized to have occurred independently in Arabidopsis and Brassica because of structural dissimilarities between the Arabidopsis and Brassica null loci. The structural dissimilarity between the N5 and N3/N13 loci further suggests that RPM1 has been deleted twice in Brassica after the genome triplication.

The simple organization of the RPM1 locus differs from most R loci characterized to date, which contain clustered R gene families and other repeated sequences. Several lines of evidence suggest that this repetitive organization can promote structural divergence of R genes by unequal crossovers and gene conversion (7, 27, 28). This mode of evolution likely accelerates the evolution of novel R genes in response to coevolving pathogens. In contrast, RPM1 exists as a single-copy gene in Arabidopsis, and the collinearity between the Arabidopsis and Brassica RPM1 loci suggests that this single-copy organization has persisted over a relatively broad evolutionary time span, based on an estimated divergence of Arabidopsis and Brassica lineages approximately 10 million years ago (29). Thus it seems unlikely that recombination has significantly influenced RPM1 evolution. One explanation for this conservation of single-copy gene organization is that RPM1 is a versatile resistance gene with multiple specificities and is therefore not subject to strong pressure for rapid structural divergence. This idea is consistent with the dual specificity of RPM1 for two unrelated avirulence gene products. In addition, avrRpm1 can function as a pathogen virulence factor on Arabidopsis (30). This target of RPM1 is therefore likely to be important for some pathogens and not easily discarded. Unpublished experiments (N. Gunn and E. Holub, personal communication) have demonstrated that 20 of 50 Col-incompatible P. s. maculicola isolates from Brassica are moderately or fully compatible with a Col-rpm1 mutant. These data suggest that RPM1 is an effective resistance gene against a variety of pathogenic bacterial isolates. It is tempting to speculate that complex R gene loci represent active battlegrounds in arms races, while simple loci like RPM1 reflect a relatively stable outcome.

We do not know if RPM1 function is conserved in Brassica; however, the B. napus cultivars from which the 1A and 9N genes were cloned are not resistant to Pst DC3000 with avrRpm1 or avrB (M. Grant and J. Taylor, personal communications). Preliminary experiments (not shown) indicate that neither 1A nor 9N can substitute functionally for Arabidopsis RPM1 in transgenic rpm1-null Arabidopsis accessions. These data suggest that 1A and 9N have diverged from Arabidopsis RPM1 in their ability to recognize the avrRpm1 and avrB-dependent elicitors and/or in their ability to interact with other components of the resistance response in Arabidopsis. Perhaps Brassica species contain multiple alleles with different recognition capabilities at RPM1, reminiscent of the L locus in flax (12). It is currently unknown whether different B. napus lines, or other Brassica species, have different RPM1 copy numbers.

The null organization of the rpm1 alleles was unanticipated because other R loci, both simple and complex, contain allelic or closely linked homologous genes even in disease-susceptible plant genotypes (4, 6, 7, 12). The existence of highly related homologs in R gene clusters implies that R gene duplication and divergence can lead to gains-of-resistance function that generate intraspecific functional polymorphism at R gene loci. Our results demonstrate that intraspecific polymorphism can also arise from gene deletions. Intraspecific R gene copy number variation has been observed in several other studies (3–8). For example, recent comparative mapping of R gene homologs in grass species revealed several loci in which R gene homologs were present at a given locus in one species but undetectable at the syntenic position in other grass species (31). The authors suggested that lack of synteny at R loci could be explained by R gene translocation and/or rapid sequence divergence of R genes after speciation. R gene deletion is a third potential explanation for lack of synteny at R loci. Based on the assumption that different R loci are subject to different modes of selection, it seems likely that both expansion and contraction occur at R loci. These events could be driven, respectively, by selection for novel R genes and selection against superfluous or costly R genes (see below).

The sequences of the null loci provided no clues as to the mechanism of RPM1 loss. It seems most likely that the Arabidopsis locus was spontaneously deleted and that the 98-bp segment represents filler DNA inserted in place of RPM1. The imperfect 9-bp direct repeat in Col-0 at the exact 3′ junction of the deletion suggests the insertion/excision of an Ac-like transposable element, but no other sequence evidence for transposons was observed in resistant or susceptible alleles. Filler DNA has been associated with spontaneous deletions and transposon excisions at other loci and in most cases is derived from sequences close to the deletion (32, 33). The donor of the 98-bp filler is unknown but may be revealed as more Arabidopsis sequences become available.

At first glance it seems counterintuitive that a gene like RPM1 with obvious selective advantages would be deleted. The deletions of RPM1 may be stochastic events with no selective consequences in the absence of pathogen pressure. Alternatively, RPM1 may be selectively disadvantageous in the absence of pathogen pressure. This “cost of resistance” hypothesis is supported by the presence of almost identical null alleles in Arabidopsis accessions from Germany, Switzerland, and North Africa, which implies a rapid proliferation of the rpm1-null allele. Furthermore, a forthcoming study will present evidence that the rpm1-null haplotype is relatively ancient and that RPM1 and rpm1-null haplotypes are present in natural Arabidopsis populations from Europe, Asia, Africa, and North America (E. Stahl and J. Bergelson, personal communication). The widespread distribution and apparent conservation of both haplotypes suggests that each haplotype is selectively advantageous under different environmental conditions and that both haplotypes have been maintained by balancing selection.

Direct evidence for an RPM1-associated fitness penalty may be provided by carefully designed fitness comparisons of isogenic lines that differ only by the presence or absence of RPM1. Extending such comparisons to different types of rpm1 mutant alleles may illuminate the molecular basis of the penalty. Perhaps RPM1 predisposes the plant to inappropriate defense induction by spurious stimuli. Because induced resistance involves significant metabolic reprogramming (34) and localized cell death (35), even a low level of inappropriate defense induction could impose significant fitness costs. A homolog of the Pto resistance gene (Fen) triggers a resistance-like reaction in response to herbicide application (36). Disease lesion mimic mutants, which likely arose from recombination between linked R genes, have been identified at the maize RP1 resistance gene complex (37). These examples suggest that the same mechanisms that accelerate the evolution of novel resistance genes may also generate disadvantageous alleles. It is worth noting that deletion alleles are unable to revert, recombine, or evolve a new inappropriate specificity and could thus be less costly than missense alleles. Perhaps frequent recombination within R gene clusters serves a dual purpose by facilitating both the creation of useful R genes and the removal of costly or superfluous R genes. Further examination of RPM1 and other pathogen-resistance loci will reveal the prevalence and selective importance of gene deletions.

ABBREVIATIONS

avr

avirulence

NBS

nucleotide binding site

LRR

leucine-rich repeat

RPM1, resistance to Pseudomonas syringae

pathovar maculicola

RFLP

restriction fragment length polymorphism