Abstract
Two sets of overlapping experiments were conducted to examine recombination and spontaneous mutation events within clusters of resistance genes in lettuce. Multiple generations were screened for recombinants using PCR-based markers flanking Dm3. The Dm3 region is not highly recombinagenic, exhibiting a recombination frequency 18-fold lower than the genome average. Recombinants were identified only rarely within the cluster of Dm3 homologs and no crossovers within genes were detected. Three populations were screened for spontaneous mutations in downy mildew resistance. Sixteen Dm mutants were identified corresponding to spontaneous mutation rates of 10–3 to 10–4 per generation for Dm1, Dm3, and Dm7. All mutants carried single locus, recessive mutations at the corresponding Dm locus. Eleven of the 12 Dm3 mutations were associated with large chromosome deletions. When recombination could be analyzed, deletion events were associated with exchange of flanking markers, consistent with unequal crossing over; however, although the number of Dm3 paralogs was changed, no novel chimeric genes were detected. One mutant was the result of a gene conversion event between Dm3 and a closely related homolog, generating a novel chimeric gene. In two families, spontaneous deletions were correlated with elevated levels of recombination. Therefore, the short-term evolution of the major cluster of resistance genes in lettuce involves several genetic mechanisms including unequal crossing over and gene conversion.
PLANT disease resistance genes are frequently members of multigene families, each member conferring resistance to a specific strain of the pathogen. Classical genetic studies conducted in parallel on the plant host and pathogen have often shown that the interaction follows a “gene-for-gene” principle: for every resistance gene in the host, there is a corresponding avirulence gene in the pathogen (Flor 1956; Crute 1986). Although molecular studies have shown that the gene-for-gene interaction is an oversimplification (Bisgroveet al. 1994; Salmeronet al. 1994), it remains a useful predictive model for plant-pathogen interactions. Genetic and molecular analyses of resistance genes have increasingly demonstrated that the clustering of disease resistance genes is a common occurrence in plant genomes (Michelmore and Meyers 1998). Some genes, such as the L locus in flax, are an allelic series (Islam and Shepherd 1991). More frequently, resistance genes are located in complex, highly duplicated regions with multiple genes that are tandemly arrayed and may encode resistances to diverse pathogens. The Cf clusters in tomato, the M locus in flax, the Xa21 locus in rice, and the Dm clusters in lettuce all exhibit this organization (reviewed in Michelmore and Meyers 1998).
Approximately 20 disease resistance genes, which confer resistance to an array of bacterial, viral, and fungal pathogens, have been cloned, mostly by transposon tagging or map-based cloning (reviewed in Bakeret al. 1997; Hammond-Kosack and Jones 1997; Ellis and Jones 1998; Michelmore and Meyers 1998). Nearly all of these genes are predicted to encode proteins involved in signal transduction and thus are implicated in pathogen recognition and elicitation of the resistance response. These proteins and their corresponding genes can be categorized according to the structural motifs that they contain (Bakeret al. 1997; Hammond-Kosack and Jones 1997). By far the most prevalent group is the class encoding a nucleotide-binding site (NBS) and leucine-rich repeats (LRRs; Meyerset al. 1999). This class includes, among others, the N gene from tobacco (Whithamet al. 1994), the L6 gene from flax (Lawrenceet al. 1995), the I2 gene from tomato (Oriet al. 1997), the Arabidopsis genes RPM1, RPS2, and RPP5 (Bentet al. 1994; Mindrinoset al. 1994; Grantet al. 1995; Parkeret al. 1997), as well as the Dm3 gene from lettuce (Meyerset al. 1998a; K. Shen and R. Michelmore, unpublished results).
A variety of genetic events have been observed at resistance gene loci in plants (Elliset al. 1997; Hulbert 1997; Michelmore and Meyers 1998; Ronald 1998). Both unequal crossovers and gene conversion have been correlated with meiotic instability and novel resistance specificities at the Rp1 locus in maize (Sudupaket al. 1993; Hu and Hulbert 1994; Richteret al. 1995). Unequal intragenic recombination in the LRR-encoding portion of the gene was also implicated in losses of resistance at the M locus in flax and the RPP5 locus in Arabidopsis (Andersonet al. 1997; Parkeret al. 1997). Sequence analyses suggested that recombination or gene conversion had produced chimeric genes in the Cf-4/9 and Cf-2/5 clusters in tomato (Parniskeet al. 1997; Dixonet al. 1998; Parniske and Jones 1999), the Xa21 cluster in rice (Songet al. 1997), the Dm3 cluster in lettuce (Meyerset al. 1998a), as well as the Arabidopsis RPP8 and RPP5 clusters (McDowellet al. 1998; Noelet al. 1999). Sequence analysis also provided evidence that transposition was involved in the evolution of the Xa21 cluster (Songet al. 1997). These data led to a model of resistance gene evolution in which unequal crossing over and gene conversion are the major genetic mechanisms leading to the generation of new resistance specificities. Clusters of resistance genes are postulated to be dynamic, unstable arrays of related sequences (Elliset al. 1997; Hulbert 1997).
However, comparative sequence analyses of several loci suggested that resistance genes are evolving slowly and that the structure of some resistance clusters may be fairly stable. Complete sequencing of the resistant and susceptible haplotypes of the Pto locus in tomato revealed a conserved structure with obvious orthologous relationships (D. Lavelle and R. Michelmore, unpublished results). At the Pto, Cf, Xa21, and Dm3 clusters, orthologs are more similar than paralogs and there is little evidence for the sequence homogenization that frequent crossovers between paralogs and gene conversion would produce (Parniskeet al. 1997; Meyerset al. 1998b; D. Lavelle, unpublished data; S.-S. Woo, unpublished data). Comparisons of sequences flanking the RPM1 locus in Arabidopsis also led to the conclusion that resistance genes are evolving slowly (Stahlet al. 1999). In addition, nonsynonymous (Ka) to synonymous (Ks) nucleotide substitution ratios of LRR-encoding portions of genes in the Cf-4/9, Xa21, RPP1, RPP8, RPP5, and Dm3 clusters are all >1 (Botellaet al. 1998; Parniskeet al. 1997; McDowellet al. 1998; Meyerset al. 1998b; Michelmore and Meyers 1998; Wanget al. 1998; Noelet al. 1999). Together these data suggest that diversifying selection acting in concert with random mutation on individual genes may be more important than sequence exchange between paralogs in generating new resistance specificities. This led to an alternative model that resistance genes evolve mainly through divergent selection and a “birth-and-death” process (Michelmore and Meyers 1998), similar to the evolution of vertebrate MHC genes (Neiet al. 1997). Genes are continually “born” via duplication events and either are maintained in the genome through evolutionary time or “die” via deletion or mutational events. These two models of resistance gene evolution are not mutually exclusive. The Cf-4/9, Xa21, RPP5, RPP8, and Dm3 loci provide evidence for a variety of genetic events; however, the relative importance of recombination, conversion, transposition, and divergent selection may differ over evolutionary time, for different resistance gene clusters, or for plant species exhibiting different mating systems.
The interaction between lettuce, Lactuca sativa, and the obligate biotrophic fungus, Bremia lactucae, the causal agent of lettuce downy mildew, has been studied extensively. Classical genetic analyses have demonstrated at least 15 dominant, single genes for resistance to downy mildew (Dm genes) located in at least three major clusters in the lettuce genome (Farrara and Michelmore 1987; Bonnieret al. 1994). Over 100 additional resistance specificities have been identified (Norwoodet al. 1981; Farrara and Michelmore 1987; Bonnieret al. 1992), but only a few have been genetically characterized (Bonnieret al. 1994; Maisonneuveet al. 1994). Dm3, which is located in the largest cluster, has recently been cloned by a combination of map-based cloning and a candidate gene approach using PCR with degenerate oligonucleotide primers (Meyers et al. 1998a,b; Shenet al. 1998; K. Shen and R. Michelmore, unpublished results). Dm3 belongs to a multigene family of resistance gene candidate (RGC) sequences, the RGC2 family, that encodes members of the NBS-LRR class of resistance proteins (reviewed in Bakeret al. 1997; Hammond-Kosack and Jones 1997; Ellis and Jones 1998; Meyerset al. 1999). Family members are interspersed throughout the region surrounding Dm3 and span at least 3.5 Mb (Meyerset al. 1998a). The RGC2 family is highly duplicated and complex, containing more than 24 family members that share similar molecular markers, as well as 53–96% nucleotide sequence identity. Mutation and transgenic analyses demonstrated that the family member, RGC2B, encodes Dm3 specificity (Meyerset al. 1998a; Chin 2000; K. Shen and R. Michelmore, unpublished results).
To investigate the roles of recombination, unequal crossing over, and possibly gene conversion in the evolution of resistance specificities in the Dm3 cluster, two sets of overlapping experiments were conducted. The first experiment involved the examination of recombination in the Dm3 region over multiple generations. A large F2 population was screened to identify individuals with recombination breakpoints near Dm3. These individuals were then further examined with molecular markers to compile a profile of recombination break-point patterns in the region. The second experiment involved the identification and characterization of naturally occurring, spontaneous mutations in Dm resistance from three different populations of lettuce; F2 recombinants from the first experiment were used to derive one of these three populations. Molecular analyses were conducted to determine the genetic changes underlying the loss of resistance in these spontaneous mutants.
MATERIALS AND METHODS
Identification of recombinant individuals in the Dm3 region: Two lettuce cultivars, Kordaat (Dm1, Dm3, Dm4) and Calmar (Dm7, Dm8, Dm13), were crossed and the FKC1 (Kordaat × Calmar) individuals selfed to produce a large F2 population in which six Dm genes were segregating. This was the same cross that was used to produce an intraspecific mapping population analyzed by Kesseli et al. (1994). A rapid alkali-treatment method was used to extract DNA suitable for PCR analysis from F2 seedlings that were 2–3 wk old (Wanget al. 1993). Two SCAR (sequence characterized amplified region) markers, SCV12 and SCI11, were used in a multiplexed PCR to screen F2 individuals for the occurrence of recombination in the Dm3 region. Data from the intraspecific mapping population (Kesseliet al. 1994) indicate that SCV12 is 3.6 cM from Dm3 and that SCI11 is located 4.9 cM from Dm3, on the opposite side (Figure 1). Both markers are codominant, allowing the identification of recombinant individuals that had a crossover point within an ∼8.5-cM interval surrounding Dm3.
Identification of spontaneous mutants in downy mildew resistance: Three populations were screened for spontaneous mutants.
-
Population 1: FKC1 individuals from a cross of the cultivars Kordaat (Dm1, Dm3, Dm4) × Calmar (Dm7, Dm8, Dm13). This population was screened for losses of Dm1 and Dm3 activity. Activity of Dm4 could not be tested in this population due to the lack of an isolate with the appropriate virulence phenotype. Activities of Dm7, Dm8, and Dm13 could not be tested due to unavoidable selfing of the maternal parent (Kordaat).
-
Population 2: S2 (selfed twice) families of the lettuce cultivar, Diana (Dm1, Dm3, Dm7, Dm8). This population was screened for losses of all four specificities.
-
Population 3: F3 recombinant-derived families. This population was derived by crossing Kordaat × Calmar F2 individuals, which exhibited recombination events between SCV12 and SCI11 (identified in the population described in the preceding paragraph) such that the recombinant allele retained Dm3.
These crosses produced
To screen for spontaneous Dm mutants,
Virulence phenotypes of B. lactucae isolates used
Oligonucleotide primers used in PCR and sequencing
Genetic analyses of recombinants and spontaneous dm mutants: Segregation of the mutant originating from population 1 was analyzed in an F2 population derived by crossing Kordaat to selfed individuals that were homozygous for the mutant allele. Segregation of resistance was scored for the S2 and F3 mutant families of populations 2 and 3, respectively. The stability of mutant phenotypes was confirmed by examining progeny of mutants after one or two generations of selfing in all three populations. To determine the number of loci mutated, complementation tests were conducted by intercrossing mutants exhibiting the loss of the same Dm specificity. In addition, two representatives of a fast-neutron (FN)-induced panel of dm3 deletion mutants, dm3r1208 and dm3r1608, one FN dm7 mutant, dm7r240 (Okubaraet al. 1994; Andersonet al. 1996), as well as a panel of ethyl methanesulfonate (EMS) mutants representing both dm3 and dm7 (Chin 2000) were crossed to the spontaneous mutants. Crosses were also made to the closely related cultivars, Cobham Green, which has no known resistance genes, and Diana (Dm1, Dm3, Dm7, Dm8), to determine if epistatic or dominant loci had been affected. Reciprocal crosses were made in most cases.
Linkage analyses of the recombinants and spontaneous mutants from population 3 were conducted using formulas from Allard (1956) and the software program LINKAGE-1 (Suiteret al. 1983). The Kosambi mapping function was used to convert some recombination frequencies to centimorgans (Kosambi 1944).
Molecular analyses of recombinants and spontaneous mutants: Extraction of DNA for comprehensive marker analyses was conducted using a modified cetyltrimethylammonium bromide protocol (Bernatzky and Tanksley 1986). Flanking molecular markers assayed on recombinants and mutants included randomly amplified polymorphic DNA (RAPDs; Williamset al. 1990), SCARs (Paran and Michelmore 1993), and microsatellites (Table 2; Okubaraet al. 1997; Sicardet al. 1999). To search for restriction fragment length polymorphisms (RFLPs), Southern blot analyses were conducted according to standard protocols (Sambrooket al. 1989), using a variety of restriction enzymes, including HindIII, AccI, BamHI, BglI, NcoI, SacI, and ScaI (New England Biolabs, Beverly, MA), nylon membrane (Hybond N+, Amersham, Arlington Heights, IL or GeneScreen Plus, New England Nuclear Life Sciences Products, Boston, MA), and 32P-labeled probes made by the random-primer method (MultiPrime, Amersham). Several markers, e.g., MSAT15-34 and NBS2B, represent multicopy sequences that identify multiple RGC2 family members; individual fragments of such markers are indicated by the marker name followed by a colon and a number, e.g., MSAT15-34: 1, which is present in Dm3.
Sequencing: Two primer sets, RLG2B5-30/5-2Bintron2 and RGC2Bintron2-3end/2BUTRA-B (Table 2), were used for amplification of Dm3 sequences using long-range PCR. Two sets of primers were used because the entire gene (∼13 kb) could not be amplified in a single reaction. PCR products were amplified from genomic DNA using the polymerases, BIOX-ACT (Intermountain Scientific Corp., Kaysville, UT) or DNA-Zyme (Finnzymes, MJ Research, Watertown, MA), which have 3′ to 5′ proofreading activity, in two-step PCR: an initial denaturation of 94° for 2 min, followed by 15–25 cycles of 92° for 20 sec and 68° for 9 to 11 min, and a final extension of 72° for 6 to 9 min. Amplification products were sequenced directly after gel purification using DEAE cellulose membranes (NA45, Schleicher & Schuell, Keene, NH) or sequenced after subcloning using the TOPO-XL kit (Invitrogen, Carlsbad, CA). All sequencing was performed with a dye terminator cycle sequencing kit and an automated ABI 377 sequencer (Applied Biosystems Inc., Foster City, CA). Analysis of sequence information was performed using the software package, Sequencher 3.0 (Genecodes, Ann Arbor, MI).
RESULTS
Recombination is infrequent within the Dm3 cluster when it is heterozygous: A total of 167 recombinant individuals were identified from an analysis of ∼2220 F2 seedlings using the codominant SCAR markers, SCV12 and SCI11, that flank Dm3 (Figures 1 and 2A). Subsequently, 13 additional RAPD, SCAR, and microsatellite markers, which span the length of the SCV12-SCI11 interval, were assayed on the recombinant individuals to determine the positions of the recombination breakpoints (Figure 1). The block of markers that included OPAC15, OPAH17, OPJ11, SCE14, SCK13, and SCM05 cosegregated with Dm3 in the original mapping population; these markers are interspersed throughout a small region immediately surrounding Dm3 (materials and methods; Figure 1; Kesseliet al. 1994; Meyerset al. 1998a). The two microsatellite markers, MSAT15-34 and MSATE6, represent sequences within RGC2B, the Dm3 gene (Figure 3), and are also present in several other RGC2 paralogs that are distributed within the OPAC15-SCM05 block (Okubaraet al. 1997; Meyers et al. 1998a,b; Sicardet al. 1999). The molecular marker assays indicated that 163 out of 167 recombination events resolved outside the OPAC15-SCM05 block.
Only four individuals, C31, C75, C127, and C132, out of the 167 recombinants were identified that had cross-over points within the OPAC15-SCM05 interval producing recombinant RGC2 haplotypes. The occurrence of these rare meiotic events could be most easily identified using MSAT15-34; the recombinants C31, C127, and C132 had nonparental MSAT15-34 haplotypes. C31 and C127 exhibited a deletion of the fragments 3 and 4 from the parental Kordaat genotype (Figure 4). C132 was missing fragments 1, 3, and 5 (Figure 4). The recombinant C75 was identified by the presence of all the block markers except SCE14. The deletion of multiple markers indicated that all four recombinant individuals had nonparental numbers of RGC2 sequences. All four recombinants were screened for resistance to the isolate IM25P11, which is diagnostic for Dm3. Individuals C31, C75, and C127 retained Dm3 resistance, while individual C132 exhibited a loss of Dm3 specificity.
—Genetic map of the major cluster of resistance genes in lettuce, the Dm1/Dm3 cluster. Column 1 represents the map with centimorgans noted between markers. OP markers are RAPDs. SC markers are SCAR markers. CL markers are RFLPs. MSAT markers are microsatellites. Small numerals indicate band sizes. The codominant SCAR markers, SCV12 and SCI11, were used to monitor recombination in a region surrounding Dm3. Column 2 represents blocks of markers that cosegregated with Dm3 or Dm1 in the mapping population (Kesseliet al. 1994).
After selfing to achieve homozygosity of the recombinant chromosome, Southern hybridization analyses were conducted on an array of 18 recombinants representative of the various crossover breakpoints. Several RFLP probes were hybridized to genomic DNA digested with HindIII, including NBS2B (Meyerset al. 1998a; Shenet al. 1998), AC15 (Andersonet al. 1996), and 651END (Meyerset al. 1998a), which represent sequences in the 5′, middle, and 3′ regions of RGC2 sequences, respectively (Figure 3). All three probes hybridized to multiple fragments (∼20 fragments) representing the various RGC2 family members and thus allowed the monitoring of genetic changes occurring throughout the resistance gene cluster. Each probe detected diagnostic banding patterns for the parents, Kordaat and Calmar. These fragments were coinherited in 14 out of the 18 F2:3 progeny as parental haplotypes, with progeny exhibiting either the Kordaat or Calmar haplotype. The four recombinant individuals, C31, C75, C127, and C132, exhibited banding patterns that were either a combination or a subset of the two parental haplotypes (Figure 5), providing further evidence that rare recombination events had occurred within the RGC2 cluster. The recombination breakpoint data of these four individuals for both the RFLP and PCR-based markers were consistent with the deletion breakpoint map previously generated for the region (Figure 6; Andersonet al. 1996; Meyerset al. 1998a). Although the crossovers did not apparently generate chimeric RGC2 genes (see below), the deletion of multiple markers indicates that these events did alter RGC2 copy number (Figure 5).
—Derivation of F2 population screened for recombinants and F3 recombinant families screened for Dm3 spontaneous mutations. (A) The cultivar Kordaat (Dm1, Dm3, Dm4) was crossed with the cultivar Calmar (Dm7, Dm8, Dm13) to produce a segregating F2 population. (B) F2 recombinant siblings were crossed to produce 
individuals (R for recombinant) in which Dm3 was homozygous but flanking regions were heterozygous. These 
individuals were then selfed twice and the resulting population of F3 families phenotypically screened for spontaneous mutations in Dm3 specificity. The SCAR markers, SCV12 and SCI11, were used to monitor recombination in the region over multiple generations. The region of homozygosity surrounding Dm3 was delimited by the block of markers containing OPAC15, OPAH17, OPAM14, OPJ11, OPM15, OPX11, SCE14, SCK13, and SCM05 (Figure 1). Recombination rates (R1, R2, R3a–g) experienced during the various meioses were tabulated in Table 5 (see text).
Twelve additional unique restriction endonuclease/probe combinations were used in Southern analyses on the four recombinants with breakpoints within the RGC2 cluster (data not shown). No novel fragments were detected relative to the parental Kordaat and Calmar haplotypes. Because the majority of RGC2 homologs resided on unique RFLP fragments, the lack of novel fragments indicated that the recombination events within the cluster did not resolve in or near RGC2 genes. Although it is possible that recombination between two closely related homologs would not produce changes detectable by a single probe or restriction enzyme digest, no changes were detected by any of the probes or enzymes used. Therefore, there was no evidence for the generation of chimeric RGC sequences and the recombination events must have resolved in the noncoding regions between the homologs rather than within the genes.
An estimate of the physical size of the SCV12-SCI11 interval was unavailable and thus an estimation of the relationship between physical and genetic size for this interval could not be calculated. However, the physical size of the region immediately surrounding Dm3, the OPAC15-SCM05 block, has been estimated to encompass at least 3.5 Mb, based on a partial bacterial artificial chromosome (BAC) contig tiling path, average spacing between RGC2 family members, and high molecular weight DNA analyses (Meyerset al. 1998a; Chin 2000). The four recombinants identified from ∼2220 screens of F2 individuals represented a genetic distance of 0.15 cM. Therefore, the ratio of physical to genetic distance was estimated to be 23.3 Mb/cM (3500 kb/0.15 cM), in comparison to an estimated genome-wide average of 1.28 Mb/cM [2500 Mb (Arumuganathan and Earle 1991)/1950 cM (Kesseliet al. 1994)]. Therefore, the RGC2 cluster is not highly recombinagenic; in fact, recombination in the OPAC15-SCM05 interval is 18-fold less in the pairing of Calmar and Kordaat haplotypes, relative to the genome average.
—Structure of Dm3 gene with position of markers, primers, and gene conversion event. Schematic of Dm3 (RGC2B) gene. Location of RFLP probes NBS2B, IPCR800, AC15, and 651END are indicated as solid bars below the gene. Each of these probes hybridizes to multiple RGC2 family members (Figures 5, 6, and 8). Position of microsatellite markers, MSAT15-34 and MSATE6, and primers used in PCR and sequencing are indicated as arrows. These two microsatellite markers also represent multicopy markers that are present on several RGC2 family members (Figures 4, 6, and 7). The position and length of the converted region in the spontaneous mutant, dm3s1977, is indicated as a dashed bar.
Rates of spontaneous mutation in homozygous Dm genes differ: A total of 15 new mutants were identified from the three populations screened for spontaneous losses of Dm resistance (Table 3). The
Dm loci differ in their meiotic stability. Both Dm3 and Dm7 undergo spontaneous mutations at a rate of ∼10–4 mutations per locus per generation, which is high relative to most other types of genes (Drakeet al. 1998), but similar to rates observed for the Rp1 locus in maize (Bennetzenet al. 1988; Hulbert 1997). The combined FKC 1 data (population 1) indicate that Dm1 has a similar mutation rate; however, no Dm1 mutants were observed in the S2 population (population 2). These spontaneous mutations occurred in meioses involving genotypes that were homozygous for the entire genome (the parents of populations 1 and 2) or for the region immediately surrounding the Dm3 gene (population 3). No spontaneous mutations of Dm8 were found; therefore, this locus appears to be more stable relative to the other Dm loci tested.
—MSAT 15-34 genotype of selected recombinants in Dm3 region. MSAT 15-34 is a multicopy microsatellite marker that is represented on several RGC2 family members. Three recombinants were identified with nonparental MSAT15-34 genotypes or deletions of various copies. C31 and C127 were missing bands 3 and 4 from the Kordaat genotype. C132 was missing bands 1, 3, and 5. C75, which is not shown, exhibited no deletions. Band 2 is present only in the Diana genotype and is not shown here. Bands not indicated with arrows represent “stutter” bands, which are an artifact of the amplification of the repeated array by PCR with Taq polymerase.
Mutations are single locus, recessive, and stable: Selfed progeny from all mutants identified from the three populations were retested with the appropriate fungal isolates. The mutant phenotypes for all progeny were confirmed (Table 4). Segregation of resistance was examined to determine the inheritance of the mutant phenotypes. None of the segregation ratios deviated significantly from the expected ratio of 3 resistant:1 susceptible (Table 4). Thus, in all the spontaneous dm mutants identified, susceptibility segregated as a recessive, stable trait at a single locus.
To test for allelism, independent spontaneous mutants that exhibited a loss of resistance to the same fungal isolate were intercrossed. Additionally, spontaneous mutants were crossed with a range of FN-induced deletion mutants (Okubaraet al. 1994) and EMS-induced point mutants (Chin 2000). All crosses between mutants exhibiting a loss of the same resistance specificity produced susceptible progeny, providing no evidence for intergenic complementation within a specificity group. Therefore, lesions had occurred in the same Dm locus in both the induced and spontaneous mutant populations.
Additional crosses were made to test for epistasis and dominant mutations. Spontaneous mutants were crossed with Cobham Green, a closely related cultivar with no known Dm genes. All F1 progeny from crosses to Cobham Green were susceptible, indicating that mutations had not occurred in a locus that was epistatic to the Dm genes. Mutants were also crossed back to wild-type Diana or Kordaat. All F1 progeny from these crosses were resistant, confirming that the mutations were not dominant.
Spontaneous mutations at the Dm7 locus are not associated with detectable deletions: The RAPD markers OPA01, OPK02, and OPH14, which span an ∼2-cM interval that includes Dm7 (Kesseliet al. 1994; P. Okubara, unpublished results), were assayed in the three dm7 S2 mutants from population 2. All markers were present; thus no deletions could be detected at this resolution. The FN-induced dm7 mutants had also showed no detectable deletions (Okubaraet al. 1994). It is possible that genes required for viability are tightly linked to Dm7 and therefore large deletions in this region are lethal. Due to the lack of additional tightly linked markers or candidate genes in this region, these mutants were not studied further.
—Southern hybridization of 651END on selected recombinants in the Dm3 region. The probe 651END is located ∼4 kb downstream of the Dm3 sequence and detects multiple family members in both parental haplotypes. Each of the recombinants, C31, C127, and C132, displayed a banding pattern that was either a subset or a combination of the two parental banding patterns, Calmar and Kordaat. Deletions of fragments indicate losses of the corresponding RGC2 family members. No novel fragments were detected.
Most spontaneous mutations at the Dm3 locus exhibit deletions: The RAPD markers OPAC15, OPAH17, and OPJ11, as well as the SCAR markers SCE14, SCK13, and SCM05 (Table 2; Figure 1), were assayed in the 12 spontaneous dm3 mutants from the three populations. Eleven mutants exhibited a deletion of the markers OPAH17, SCK13, and SCM05. The microsatellite markers MSAT15-34 and MSATE6 (Table 2; Figures 1 and 3) provided greater resolution and identified four types of deletions (Figures 6 and 7). Eight mutants, dm3s285, dm3s365, dm3s1427, dm3s2376, dm3s3646, dm3sF1.2, dm3s3049, and dm3s7241, had the largest deletion and were missing all of the MSAT15-34 fragments (Figure 7) and all of the MSATE6 fragments, except MSATE6:8 (data not shown). Three of the remaining mutants contained unique deletions (Figures 6 and 7). As with the analysis of the F2 recombinants (above), the deletion of several molecular markers indicated that mutant haplotypes contained nonparental numbers of RGC2 homologs. Mutant dm3s1977 had no detectable deletions with any of the markers.
Spontaneous mutants identified in lettuce
Segregation of resistance in spontaneous Dm mutant families
—Breakpoint analysis of Dm3 recombinants and spontaneous mutants. The breakpoints of the Kordaat × Calmar F2 recombinants and the dm3 spontaneous deletion mutants were mapped using a combination of PCR-based and RFLP markers. Two FN-deletion mutants are included as references: dm3r1608 and dm3r1208 contained the largest and smallest deletions, respectively (Andersonet al. 1996; Meyerset al. 1998a). Positions of BAC groups that contain members of the RGC2 multigene family, of which Dm3 is a member, are given below. The physical region spanned by the BAC groups is estimated to be at least 3.5 Mb (Meyerset al. 1998a). All markers mapped are from the Kordaat or Diana Dm3 haplotype and are shown at the top of the figure (see text). OP or SC markers are RAPD and SCAR markers. MSAT markers are microsatellite markers. All other markers are RFLP markers. Markers in boldface type were present on the BACs indicated by the connecting dashed line. Markers in normal typeface mapped to the region but were not detected on a BAC. The position of markers that could not be located precisely is shown above bars at the top. Individual fragments of multicopy markers are denoted by a colon and the letter(s) or number indicating the individual fragment. Identical duplicate markers, detected by their presence on nonoverlapping BACs, are noted by a dagger. The Kordaat alleles in the recombinants, C31, C75, C127, and C132, are noted by solid lines and the inferred Calmar alleles are noted by dotted lines. Data modified and updated from Meyers et al. (1998a).
Southern hybridizations with sequences derived from Dm3 were conducted to further characterize the dm3 spontaneous mutants. Six unique restriction endonuclease/probe combinations were analyzed that were a subset of those used in the analysis of the recombinants. Deletions of several of the ∼20 RGC2 fragments were detected in all of the mutants except for dm3s1977 (Figure 8). The Southern analyses were consistent with the mutant deletion profiles detected by the PCR analyses and the linear order of breakpoints determined from previous studies of fast-neutron mutants and BACs (Figure 6; Andersonet al. 1996; Meyerset al. 1998a). None of these Southern analyses detected any novel fragments relative to the wild-type haplotype. Thus, although the spontaneous deletion events did change the number of RGC2 homologs, the deletion breakpoints did not occur in or near RGC2 sequences and the RFLP data provided no evidence for the occurrence of chimeric RGC2 sequences in these mutants. The deletions detected in these spontaneous mutants can most easily be explained by the occurrence of recombination and unequal crossing over. However, the homozygosity of flanking markers in the eight
—MSAT 15-34 genotype of spontaneous dm3 mutants. The dm3 spontaneous mutants exhibited a variety of MSAT 15-34 haplotypes, most of which were deletions of various sizes. The wild-type parental controls, Diana, Calmar, and Kordaat are included for comparison. The mutant dm3s3049 is not shown, but exhibited a deletion of all bands.
Recombination and unequal crossing over are associated with spontaneous mutation events: To determine if recombination could be correlated with spontaneous mutation events, population 3 was derived from
—Southern hybridization of NBS2B on spontaneous dm3 mutants. The probe NBS2B detects multiple members of the RGC2 family, even in the cultivar, Calmar, which does not contain any of the Dm genes in the major cluster. All of the spontaneous mutants, with the exception of dm3s1977, exhibited a deletion of multiple bands. No mutants exhibited nonparental bands. dm3s3049 is not shown but exhibited a deletion pattern similar to that of dm3s7241.
Increased recombination frequencies were associated with some spontaneous mutation events: The recombination frequencies in the SCV12-SCI11 interval were analyzed over four generations during the derivation of the F3 mutant families in order to determine if (1) homozygosity immediately surrounding Dm3 affected recombination rates and (2) if the spontaneous mutation events affected recombination rates in the Dm3 region. The frequency of recombination in the original F2 population of over 2000 individuals (R1, Figure 2A; Table 5) was 0.039 ± 0.003 for the interval SCV12-SCI11. This rate represents meioses in which the entire interval was heterozygous. The frequency of recombination in 288 “recombinant-derived” F2 individuals (R2, Figure 2B; Table 5) was slightly higher at 0.044 ± 0.009. This increase was not statistically significant (normal approximation of the binomial, P ≅ 0.49). A similar recombination rate of 0.040 ± 0.01 was also measured in the nonmutant F3 families (R3a, Figure 2B; Table 5). Therefore the small region of homozygosity immediately surrounding Dm3 (0.15 cM and ∼3.5 Mb, see above; Meyerset al. 1998a) within the larger heterozygous SCV12-SCI11 region (3.6 plus 4.9 cM of unknown physical size) did not significantly affect the recombination frequency.
—V12/I11 configurations of recombination events in various generations. (A) The SCV12/SCI11 configuration of the original Kordaat × Calmar 
in which no recombination has occurred. (B) The configuration of the 
progenitors of population 3, resulting from crosses between F2 siblings with recombination events on opposite sides of Dm3 (see text and Figure 2). (C) The deduced SCV12/SCI11 configurations of the F2 progenitors of the four F3 mutant families. Recombination events that occurred in the meioses of the 
plants resulted in spontaneous deletions of the Dm3 gene on that recombinant chromosome. Note dm3s3780, which displayed recombination events in both chromosomes, only one of which resulted in a deletional mutation. (D) The configurations of susceptible F3 individuals, some of which displayed chromosomes in which additional recombination events occurred during F2 meioses. (E) A gel of the SCV12/SCI11 PCR reactions for the four mutant F3 families. The various possible configurations for each family are shown. Parental genotypes, Kordaat and Calmar, were included for comparison. Braces indicate the polymorphic bands.
The recombination rates were also determined for the four mutant F3 families that were derived from meioses in which one of the Dm3 alleles had previously undergone a spontaneous deletion event resulting in hemizygosity at the locus (R3b to R3e, Figure 2B; Table 5). Three of these families, dm3s1180, dm3s3049, and dm3s7241, were homozygous at either SCV12 or SCI11 and therefore recombination could be observed only on one side of the interval, using Dm3 as one of the markers. Consequently the recombination frequencies for the whole interval were inferred by extrapolation using the relative genetic sizes of the SCV12-Dm3 and Dm3-SCI11 intervals (Table 5). Two mutant families, dm3s1180 (R3b; Figure 2B) and dm3s3049 (R3c; Figure 2B), exhibited recombination frequencies of 0.061 and 0.046, which were similar to R1, R2, or R3a. The two other families, dm3s7241 (R3d; Figure 2B) and dm3s3780 (R3e; Figure 2B), exhibited significantly elevated recombination frequencies of 0.12 and 0.14, respectively (Table 5; P ≅ 0.0004 and 0.0002). Recombination in the dm3s3780 family was elevated in both the SCV12-Dm3 and Dm3-SCI11 intervals (Table 5). The extrapolation for the SCV12-SCI11 interval for the dm3s7241 family assumes that the recombination rate is elevated uniformly on both sides of Dm3; however, even when only the Dm3-SCI11 interval is considered, the measured recombination frequency is significantly greater than R1, R2, and R3a for the entire SCV12-SCI11 interval (Table 5; P ≅ 0.034). Therefore, the recombinant haplotypes generated by unequal crossing over did not significantly decrease recombination but rather in two families seemed to have stimulated recombination at the Dm3 locus.
Recombination frequencies in the SCV12-SCI11 interval over multiple generations
The phenomenon of increased recombination in the dm3s3780 and dm3s7241 families was examined further in selected F4 progeny. Resistant F3 individuals that were hemizygous for the deletion of Dm3 were selfed. Their F4 progeny were screened for susceptibility to the Avr3-expressing isolate, IM25P11 (Table 1). Recombination between SCV12 and SCI11 in four F4 families derived from dm3s3780 was 0.087 (Table 5). Recombination between SCI11 and Dm3 in four F4 families derived from dm3s7241 was 7.3 cM (10.2 cM adjusted for SCI11-SCV12 interval, Table 5). These rates were lower than those observed in F3 hemizygotes but still significantly elevated from the other populations. Therefore, the increased recombination found in the dm3s3780 and dm3s7241 F3 families was a repeatable phenomenon that continued into the next generation.
To determine the position of the crossovers that had occurred in the F2 meioses, susceptible F3 individuals were selfed to obtain F4 progeny that were homozygous for the F2 recombination product. Six and 12 susceptible F3 individuals, from the dm3s7241 and dm3s3780 families, respectively, were homozygous for the Dm3 deletion but heterozygous for either SCV12 or SCI11 (Table 5; Figure 9D). F4 progeny from these 18 F3 individuals were screened with SCV12 and SCI11 to identify those that were homozygous for the F2 recombination product. Two homozygous individuals from each family were then analyzed with the probe NBS2B. For 17 of the 18 F3 individuals selfed, deletion profiles of these F4 progeny were identical to the profiles obtained for susceptible F3 individuals that were homozygous for the FR1 recombination product (data not shown). Therefore, most of the recombination events in the F2 meioses resolved outside of the region encompassing the cluster of RGC2 sequences, as occurred in the original
—Southern hybridization of NBS2B on F3 dm3 mutants and their F4 progeny. The F3 dm3 mutants, S27 and S43, along with their F4 progeny (indicated with dashes followed by numbers) exhibit the deletion profile found in nearly all of dm3 spontaneous mutants from the dm3s3780 family, in which the marker NBS2B:L (arrowed; Figure 6) has been deleted. Only the mutant S30 and its progeny exhibit a novel haplotype in which marker NBS2B:L has been recovered through a recombination event in the F2 progenitor.
One spontaneous loss of resistance resulted from gene conversion: The mutant dm3s1977 contained all the molecular markers screened and therefore exhibited no detectable deletions. The Dm3 allele in this mutant was sequenced to determine the molecular changes underlying the loss of resistance. The 5′ portion of the gene was sequenced following amplification of a Dm3-specific fragment of ∼4.2 kb that encompassed the first three exons (Figure 3). Direct sequencing of this fragment from dm3s1977 detected no changes from the wild-type Dm3 gene in Diana. The remainder of the gene was sequenced following amplification of an ∼5.1-kb fragment that encompassed the last four exons (Figure 3). Sequencing of this 3′ fragment detected an ∼1.5-kb region that did not match the wild-type Dm3 allele. The divergent region extended from the middle of the fifth exon (just 3′ of MSATE6) to the middle of the sixth exon (Figure 3) and differed from Dm3 at 120 polymorphic sites (Table 6). The divergent sequence exactly matched the sequence of RGC2C, a closely related paralog of Dm3 (Meyers et al. 1998a,b), and can be explained most parsimoniously as a gene conversion event (as opposed to a closely spaced double recombination event). Precise breakpoints of the conversion event could not be determined because Dm3 and RGC2C share 81.6% overall nucleotide sequence identity; however, the breakpoints could be located to an 107-bp window on the 5′ end of the conversion tract and a 70-bp window on its 3′ end (Table 6). This conversion tract encompassed 38% of the C-terminal LRR (275 amino acids of the 721 amino acids in the C-terminal LRR), altering 9 of the 21 repeats in this region (Meyerset al. 1998b).
To confirm that the gene conversion tract in this mutant was not an experimental artifact, a fragment specific to the gene conversion event was amplified from genomic DNA of dm3s1977 using PCR. Primers were designed to specifically amplify across the breakpoint of the Dm3 and RGC2C sequences. As a control, an overlapping pair of primers was designed that amplified within the converted region from RGC2C (Figure 3; Table 2). The control primers, 1977CO5 and 1977CO3, amplified a fragment from the mutant dm3s1977 as well as from the RGC2C homolog present in wild-type Diana (Figure 11). However, PCR using the primers 1977CH5 and 1977CO3 only amplified a fragment from the mutant dm3s1977 (Figure 11). This confirmed the gene conversion event in the spontaneous mutant, dm3s1977.
The conversion event in the LRR-encoding region of the Dm3 gene in dm3s1977 resulted in a loss of the Dm3 specificity, indicating that this region was necessary for resistance gene function or specificity. This mutant was therefore tested against a variety of Bremia isolates from diverse geographical origins, each expressing the Avr3 avirulence phenotype (Table 1). The mutant dm3s1977 was susceptible to all 5 isolates, confirming the specificity of this mutation to the Dm3-Avr3 interaction. Thirty other isolates, which were virulent on wild-type Diana and collectively exhibited 13 different virulence phenotypes, were also tested on mutant dm3s1977; resistance was not observed to any isolate (S. Brown and O. Ochoa, unpublished results). Therefore, there was no evidence for novel specificities encoded by the chimeric allele in dm3s1977. The other spontaneous mutants were not tested against these isolates because they were deletion mutants and did not display novel RFLP patterns indicative of chimeric genes.
Sequence comparison of 3′ LRR-encoding region of the chimeric allele in dm3s1977 with Dm3 and RGC2Ca
DISCUSSION
Although the clustering of resistance genes is common in many plant species, the evolution of these gene clusters remains only partially understood. Comparative genetic and sequence analyses have identified or suggested the involvement of a variety of genetic events, including recombination, unequal crossing over, gene conversion, transposition, and divergent selection (reviewed in Hulbert 1997; Michelmore and Meyers 1998; Ronald 1998). However, the relative importance of each of these events in the immediate and long-term evolution of resistance gene clusters is not known. This article emphasizes genetic analyses of resistance gene clusters and provides further insights into the roles of recombination, unequal crossing over, and gene conversion in the short-term evolution of clusters of resistance genes. Some of our observations are consistent with conclusions from studies of other loci, while some are not.
A variety of factors affect recombination and stability within clusters of resistance genes: The structural and sequence similarities between haplotypes in a particular meiotic pairing, as well as the size and complexity of the region as a whole, will profoundly influence the genetic behavior of multigene families. Sequence and structural identity, as occurs when the locus is homozygous, will favor exact pairing and little structural rearrangement; however, if there are long stretches of duplicated sequences, there will be opportunities for mispairing resulting in further gene duplications and deletions. Structural heterozygosity between haplotypes in a particular pairing prevents precise alignment and, depending on the degree of divergence, will tend to repress recombination in the region and result in novel haplotypes whenever recombination does occur within the cluster. In our experiments, the Dm3 region was studied in both the heterozygous and homozygous condition.
There is only limited information on the rates of recombination within clusters of resistance genes. When the SCV12-SCI11 interval encompassing the Dm3 gene was heterozygous in the intraspecific Kordaat × Calmar F2 population, crossovers within the RGC2 cluster were only rarely detected and recombination was suppressed approximately 18-fold from the genome average. The RGC2 region may be partially hemizygous between Kordaat and Calmar; the RAPD and SCAR markers identified within the OPAC15-SCM05 block are all linked in cis with the Dm3 Kordaat allele (Andersonet al. 1996) and the microsatellite markers, MSAT15-34 and MSATE6,as well as Southern hybridizations with RGC2 sequences, consistently detected fewer fragments in Calmar than Kordaat. Such hemizygosity would be expected to prevent pairing within the cluster. Making the OPAC15-SCM05 block homozygous (Figure 2) did not result in a detectable increase in recombination (Table 5). Repression of recombination at resistance gene loci is not unexpected as many resistance gene-containing regions have been introgressed from other species and therefore represent divergent haplotypes (Crute 1988). There was a 10-fold reduction of recombination in a 240-kb region surrounding the Mla resistance gene cluster in barley, relative to regions immediately flanking the cluster (Weiet al. 1999). Recombination was also repressed in chromosomal regions of Lycopersicon esculentum containing Mi (van Daelenet al. 1993) and Tm-2a (Ganalet al. 1989). Recombination rates could also be influenced by chromosome position; however, Tm-2a and Mi are proximal to the centromere while RGC2 and Mla are telomeric (Ganalet al. 1989; van Daelenet al. 1993; Shenet al. 1998; Weiet al. 1999). None of these data indicate that clusters of resistance genes are highly recombinagenic.
—Confirmation of gene conversion event in dm3s1977. Three primers were designed that were diagnostic for the conversion event in the spontaneous mutant dm3s1977 (Table 2; Figure 3). The primers 1977CO5 and 1977CO3 lie within the conversion event and are thus amplified from the family member RGC2C in all three genotypes: wild-type Diana, a dm7 Diana EMS mutant, as well as dm3s1977. The primer 1977CH5, which lies just 5′ of the conversion event, was used in conjunction with 1977CO3; the product of these two primers was a chimeric sequence of Dm3 and RGC2C and was thus amplified only from the mutant, dm3s1977. Genomic DNA was used as template for all samples.
The stability of resistance genes, as evidenced by rare losses in resistance, has been studied when the resistance cluster was both homozygous and heterozygous. The most informative studies are those when the region is homozygous, as will be the case for most meioses in inbreeding species such as lettuce, tomato, and Arabidopsis thaliana. The Dm3 cluster was genetically unstable when homozygous, as evidenced by a high rate of spontaneous mutation (10–3 to 10–4 mutations per generation, Table 3). This mutation rate was comparable to instability at the Rp1 cluster in maize (Pryor 1987; Bennetzenet al. 1988). The instability of the Dm3 and Rp1 regions contrasts with the stability of the Cf-4/9 locus in tomato and the RPP5 locus in A. thaliana when homozygous (Parniskeet al. 1997; Noelet al. 1999). No mutations were identified at the Cf-9 locus in screens of ∼12,000 testcross progeny from homozygous Cf9 plants. Similarly, no susceptible individuals were identified in screens of ∼7500 testcross progeny of RPP5 homozygous plants. The reasons for such differences in stability between these loci are not clear. The Dm3, Rp1, Cf-4/9, and RPP5 clusters are all duplicated and complex multigene families. The Dm3 locus is the largest, both in terms of the number of homologs (24+ homologs vs. ∼10 homologs at the other loci) and the physical size of the cluster (∼3.5 Mb vs. ∼35 kb and 95 kb for Cf-4/9 and RPP5, respectively; Joneset al. 1994; Parniskeet al. 1997; Thomaset al. 1997; Meyerset al. 1998a; Collinset al. 1999; Noelet al. 1999). The Cf-4/9 and RPP5 loci may not be large enough and stretches of sequence affiliations may be distributed such that mispairing does not occur when these loci are homozygous; in contrast, the size and level of duplication at the Dm3 locus may allow occasional misalignment even when it is homozygous. Thus, the genetic behavior of a resistance gene cluster may change as its structure and level of complexity evolves due to genetic rearrangements. In particular, recent large duplications resulting from unequal crossing over would be expected to stimulate instability in the homozygous condition.
When resistance loci are heterozygous, estimates of stability as evidenced by losses of particular resistance specificities will be heavily dependent on the haplotypes involved, rather than a consequence of the intrinsic properties of the locus. Precise alignment is impossible between divergent haplotypes; recombination within a cluster will always result in nonparental haplotypes, some of which may lack one or more functional resistance genes. Losses of resistance could also result from gene conversion events between divergent haplotypes; however, this would not result in structural rearrangements. When Dm3 was heterozygous in the Calmar × Kordaat F2 progeny, all recombinant events within the cluster produced novel haplotypes; three recombinants retained Dm3 and one did not (Figures 4,5,6). Similarly, the five susceptible individuals that were identified from screens of 7500 testcross progeny of Cf-4/Cf-9 heterozygous plants had nonparental haplotypes at the Cf-4/9 locus (Parniskeet al. 1997). Both the frequency of novel haplotypes as well as losses of resistance should be considered in estimating the instability of resistance loci when heterozygous. This will reflect the rates of gene conversion as well as recombination within the cluster that in turn are consequences of the structural and sequence similarities between haplotypes. Furthermore, the mating system of the plant species will determine how often a resistance cluster is heterozygous. In outbreeding species, heterozygosity will be frequent and haplotypes might be expected to evolve constantly. Conversely, in inbreeding species, such as lettuce, heterozygosity will occur rarely; however, occasional outcrossing and heterozygosity may have profound consequences for the evolution of the region.
Losses of resistance were due to unequal crossing over or gene conversion: Eleven of the 12 spontaneous losses of Dm3 specificity were due to deletions. Interestingly, all of the deletions appeared to terminate in a similar region at one end and 8 of these 11 events had the same large deletion (Figure 6). These 8 events were derived from each of the three populations that were screened for spontaneous mutations in three different years. In the one population in which recombination of flanking markers could be monitored, spontaneous deletion events were associated with recombination, thereby implicating unequal crossing over as the mechanism responsible for all of the deletions. The distribution of deletion profiles may reflect a pattern of sequence duplication within the locus that favors a particular misalignment or resolution of crossover events. The gene density in the region is low and individual RGC2 genes are separated by an average of 145 kb (Meyerset al. 1998a). As with the spontaneous Cf-9 mutants in tomato (Parniskeet al. 1997), the breakpoints in novel RGC2 haplotypes were between rather than within RGC2 genes. This suggests that the crossovers are not occurring, or at least not being resolved, within genes.
In the one spontaneous mutant that exhibited no deletions, sequencing of the dm3 allele revealed that gene conversion was probably responsible for the loss of specificity. Sequence exchange between Dm3 and RGC2C necessitates pairing between sequences with 82% nucleotide identity. This alignment of Dm3 with RGC2C would have also paired the neighboring regions containing RGC2D and RGC2I with those containing RGC2S and RGC2J; these pairs of RGC2 sequences shared 82 and 92% nucleotide identity, respectively (Figure 6; Meyerset al. 1998a). The regions on either side of the RGC2D/S to Dm3/RGC2C segments contained more divergent pairs of RGC2 sequences.
The relationship between rearrangements at the locus and the evolution of new resistance specificities: No chimeric RGC2 genes were observed as the result of recombination between divergent Dm3 haplotypes or as the result of unequal crossing over. This contrasts with the analyses conducted on mutants at the Rp1-D locus in maize, in which the majority of mutations involved recombination within or close to coding regions of Rp1-D homologs (Collinset al. 1999). The Southern analyses of the Dm3 mutants utilized seven different restriction endonucleases with multiple probes and thoroughly sampled the RGC2 genes as well as the regions immediately adjacent to them. No novel fragments were observed. Recombination and unequal crossing over produced changes in the number of Dm3 paralogs but did not generate chimeric genes. Therefore, consistent with the birth-and-death model for resistance gene evolution (Michelmore and Meyers 1998), the predominant role of unequal crossing over may be in creating changes in copy number, rather than generating chimeric genes with new resistance specificities. Unequal crossing over may, however, affect the evolution of new specificities by generating gene duplications. Assays for spontaneous losses of resistance can detect only one of the products of unequal crossover events. The reciprocal products that contain the duplicated segments could be templates for divergent selection and thus potentially lead to novel resistance specificities (Michelmore and Meyers 1998).
One spontaneous mutation event generated a chimeric RGC2 gene, most likely through gene conversion. Comparative sequence analyses of the Cf-4/Cf-9, Xa21, L, RPP5, and RPP8 loci all suggest similar exchanges of sequence information between paralogs, either by recombination or gene conversion (Parniskeet al. 1997; Songet al. 1997; McDowellet al. 1998; Elliset al. 1999; Noelet al. 1999). Analyses of the Pto and RGC2 loci provided similar evidence (D. Lavelle and R. Michelmore, unpublished results; H. Kuang, E. Nevo and R. Michelmore, unpublished results). However, it is unclear on what timescales these exchanges have occurred. Our data suggest that sequence exchange between paralogs may be rare and further support the predictions of the birth-and-death model for resistance gene evolution (Michelmore and Meyers 1998).
In certain cases, the deleted product of unequal crossing over may also increase evolutionary activity of the locus. Increases in recombination were associated with two spontaneous Dm3 mutation events in multiple generations. The recombination rates representing meioses in the mutant F3 families, dm3s3780 and dm3s7241, were significantly increased ∼3- and 3.5-fold compared to the recombination rates analyzed over three generations. It is unclear why this increase occurred. The deletion in dm3s7241 was among the largest observed and is probably well over 1 Mb; the deletion in dm3s3780 is at least 500 kb (Meyerset al. 1998a; Chin 2000). If recombination is enhanced by the existence of different gene arrangements or novel haplotypes, then evolution of resistance gene clusters may be a punctuated process, with periods of relative stability interspersed with bursts of instability stimulated by unequal crossing over.
Future studies: The majority of studies on resistance gene evolution, including our experiments, have involved the analysis of one or a few haplotypes. The picture emerging from these analyses is complex. However, it is becoming increasingly evident that stability and recombination activity at resistance gene clusters are heavily dependent on several parameters including the size and complexity of the locus, as well as the structural and sequence similarities between the haplotypes in a particular pairing. It is now necessary to extend these analyses to include a greater range of haplotypes and haplotype pairings. We are currently generating progeny from crosses between genotypes with varying levels of diversity; these involve naturally occurring haplotypes or FN-induced deletions of Dm3 (Okubaraet al. 1994; Andersonet al. 1996). These progeny will be analyzed to determine the effects of haplotype diversity and the presence of deletions on recombination frequencies, instability at the locus, and the generation of chimeric resistance genes.
Acknowledgments
The authors thank Kara Green, Gavin Henderson, Cathy Little, Grace Salcedo, and Margaret Smith for their help in conducting the phenotypic screens for spontaneous mutants. We also thank Paul Gepts, Hanhui Kuang, Blake Meyers, and Brett Tyler for providing helpful comments on the manuscript. This work was supported by U.S. Department of Agriculture National Research Initiative grant 95-37300-1571.
Footnotes
-
Communicating editor: D. Charlesworth
- Received June 23, 2000.
- Accepted November 6, 2000.
- Copyright © 2001 by the Genetics Society of America
