Toward an Integrated Linkage Map of Common Bean . 111 . Mapping Genetic Factors Controlling Host - Bacteria Interactions

Restriction fragment length polymorphism (RFLP)-based genetic linkage maps allow us to dissect the genetic control of quantitative traits (QT) by locating individual quantitative trait loci (QTLs) on the linkage map and determining their type of gene action and the magnitude of their contribution to the phenotype of the QT. We have performed such an analysis for two traits in common bean, involving interactions between the plant host and bacteria, namely Rhizobium nodule number (NN) and resistance to common bacterial blight (CBB) caused by Xanthomonas campestris pv. phaseoli. Analyses were conducted in the progeny of a cross between BAT93 (fewer nodules; moderately resistant to CBB) and Jalo EEP558 (more nodules; susceptible to CBB). An RFLP-based linkage map for common bean based on 152 markers had previously been derived in the Fz of this cross. Seventy Fz-derived Fs families were inoculated in separate greenhouse experiments with Rhizobium tropici strain UMR1899 or X . c. pv. phaseoli isolate isolate W18. Regression and interval mapping analyses were used to identify genomic regions involved in the genetic control of these traits. These two methods identified the same genomic regions for each trait, with a few exceptions. For each trait, at least four putative QTLs were identified, which accounted for approximately 50% and 75% of the phenotypic variation in NN and CBB resistance, respectively. A chromosome region on linkage group D7 carried factor(s) influencing both traits. In all other cases, the putative QTLs affecting N N and CBB were located in different linkage groups or in the same linkage group, but far apart (more than 50 cM). Both BAT93 and Jalo EEP558 contributed alleles associated with higher NN, whereas CBB resistance was always associated with BAT93 alleles. Further investigations are needed to determine whether the QTLs for N N and CBB on linkage group D7 represent linked genes or the same gene with pleiotropic effects. Identification of the QTLs raises the possibility of initiating map-based cloning and marker-assisted selection for these traits. S TUDIES of host-pathogen interactions have focused mainly on qualitative gene-for-gene relationships (CRUTE 1985; GABRIEL and ROLFE 1990; KEEN 1990). Yet, for many host-pathogen interactions, the genetic control appears to be quantitative rather than qualitative. Until recently, genetic analysis and breeding of quantitative traits (QT) has been generally based on biometrical approaches (MATHER and J I N K ~ 1977; FALCONER 1981; MAYO 1987). These approaches deal mainly with the collective characterization of the multiple factors affecting a QT and allows us to partition the overall phenotypic variance into its genotypic and environment components. Biometrical methods are generally, however, not capable of characterizing or manipulating specific loci. Individual loci involved in quantitative traits can be characterized with linked marker genes as shown first by SAX (1 923) in common bean (Phaseolus vulgaris L.) ' Permanent address: Departamento de Fitotecnia, Universidade Federal * Permanent address: CENA, Universidade de S i0 Paulo, Caixa Postal de Santa Catarina, Caixa Postal 476, 88049-900 Florianoplis, SC, Brazil. 96, 13400-Piracicaba, SP, Brazil. Genetics 134: 341-350 (May, 1993) and later by THODAY (1 96 1) in Drosophila. The limiting factor in these studies was the number of (morphological) markers available. Genetic linkage maps based on molecular markers-principally, restriction fragment length polymorphisms (RFLPs) and rapid amplified polymorphic DNA (RAPDs), and to a lesser extent, isozymes, and seed proteins-allow us to map and estimate the effects of individual factors controlling a quantitative trait (QT) because only molecular markers are sufficiently numerous in any one cross to provide adequate genome coverage (PATERSON, TANKSLEY and SORRELLS 199 1). Two statistical approaches have been used to map and measure the effects of putative quantitative trait loci (QTL). Regression analysis using a general linear fixed effects model, estimates the statistical significance of the association between the phenotypic expression of the QT and a marker locus region (either a putative Q T L or a marker locus linked to it) as illustrated by the studies of EDWARDS, STUBER and WENDEL (1987) and REITER et al. (1991) in maize, NIENHUIS et al. (1987) and MARTIN et al. (1989) in 342 R. 0. Nodari et al . tomato, and KEIM et a2. (1990) and KEIM, DIERS and SHOEMAKER (1 990) in soybean. Using the LOD (logarithm of the odds ratio) score method developed for genetic linkage analysis, LANDER and BOTSTEIN (1 989) developed an interval mapping method to locate QTLs. They suggest that interval mapping approach overcomes three weaknesses of regression analysis: (a) underestimation of the phenotypic effects of QTLs; (b) confounding effects between linkage distance and magnitude of the QTL effect, which results in imprecise QTL map location; and (c) the large number of progeny required to detect QTLs. Interval mapping analysis has been used to map several Mendelian factors associated with QTs in tomato (PATERSON et al. In the work reported here, we used regression and interval mapping analyses to examine the genetic control of two host-bacterium interactions that display a quantitative inheritance in common bean (272 = 2x = 22): nodulation by Rhizobium tropici (nodule number, NN) and resistance to Xanthomonas campestris pv. phaseoli, the causal agent of common bacterial blight (CBB). We were interested in determining whether any genetic correlation existed between the host plant interactions with two distinct bacteria, which could reflect a common signal transduction mechanism at the molecular or biochemical level. To our knowledge, no information is available so far on the relationship between these two interactions in common bean or other crops. The specific objectives of this study were to determine: (1) the number of statistically significant associations between marker loci and Rhizobium nodule number or CBB resistance; (2) the individual contribution of each QTL to the xpression of their respective QT; and (3) the location of QTLs on an RFLP-based linkage map of the common bean genome. This map consists, in its current stage, of 152 molecular markers assigned to 15 linkage groups and covers 827 cM (NODARI et al . 1993; GEPTS 1993). 1988, 1991). MATERIALS AND METHODS Plant material: Segregation analyses were conducted in the progeny of the cross BAT93 X Jalo EEP558. High levels of polymorphism at the molecular level distinguished these two parents (NODARI et al. 1993). BAT93 had low levels of nodulation after inoculation with R. tropici strain UMRl899 and moderate levels of resistance to X . c. pv. phaseoli, whereas Jalo EEP558 had high levels of nodulation and was very susceptible to X . c. pv. phaseoli (see RESULTS). Segregation for Rhizobium N N was analyzed in a population of 70 Fz-derived Fs families of the BAT93 X Jalo EEP558 cross. Control genotypes included the parents, the F1, CIAT 125 (a nodulation defective or Nodmutant: J. KIPE-NOLT, personal communication) as negative control, and cultivar “Puebla 152,” a strong nodulator (PEREIRA, BURRIS and BLISS 1989). The 70 Fs families were derived by selfing from the same Fz mapping population that was used to generate an RFLP-based linkage map (NODARI et al. 1993). Therefore, the parental genotype of each Fs family had been previously determined at each of 152 loci used to develop the linkage map. Segregation analysis for CBB resistance was conducted on the same set of F2-derived Fs families, but with different plants than those inoculated with Rhizobium. Rhizobium NN experiment: The experiment was conducted in a greenhouse. The two parents, the F1 generation, and the Fs families were planted in a completely randomized block design with four replicates, and two plants per replicate. Three seeds were sown per Leonardjar wrapped with aluminum foil. The top half of the jar contained a mixture of vermiculite, sand and perlite (1 : 1 : 1, v:v:v basis), whereas the bottom half contained a nitrogen-free solution UOHNSON et al . 1957; 1:5 strength). After germination, the jars were thinned from 3 to 2 seedlings. Seven days after sowing, seedlings were inoculated with R. tropici. A highly infective strain (UMR-1899, originally supplied by P. GRAHAM, University of Minnesota) of R. tropici (MART~NEZ ROMERO et al . 1991) was used to provide good discrimination among the segregating genotypes. Rhizobia were grown in shake culture to stationary phase in a medium containing mannitol (0.1%), yeast extract and salts (VINCENT 1970). Ten milliliters of inoculum (approximately 1 O9 cells/ml) were added into each jar. After inoculation, the surface of the jars was covered with I cm of perlite to protect the substrate and inoculum from direct exposure to sunlight. Nutrient solution was added weekly. Thirty-two days after initiation of the experiment, plants were harvested and analyzed for nodule number. The roots were stored with the substrate in a cold room until the number of nodules was counted. The average of the nodule number for two plants in each jar constituted the experimental unit. CBB experiment: The experiment was conducted in a greenhouse and consisted of a completely randomized block design with three replicates, and three plants per replicate. The planting design included the parental genotypes, the F1 generation, and the Fs families. Three weeks after sowing, the first expanded trifoliolate leaves were inoculated with X . c. pv. phaseoli isolate W 18 (GILBERTSON, RAND and HACEDORN 1989) by the razor blade method (PASTOR-CORRALES, BEEBE and CORREA 198 1 ; SILVA, SINCH and PASTORCORRALES 1989). T o prepare the inoculum, an aqueous bacterial suspension was prepared in sterile distilled water from X . c. pv. phaseoli grown on sucrose peptone agar for 48 hr and diluted to obtain an optical density of 0.5 at 600 nm in a spectrophotometer which results in about 1 X lo7 colony-forming units/ml. The reaction to X . c. pv. phaseoli was evaluated 15 days after inoculation according to a 1-9 scale where 1 and 9 identified no visible symptoms and very severe disease symptoms, respectively; these evaluations were then converted into a disease index representing the percentage of leaf tissue affected (CIAT 1987). Scores of each experimental unit represented averages for three plants. Statistical analyses: Data for Rhizobium nodule number and common bacterial blight of each Fs family were subjected to regression analysis using the SAS PROC GLM procedure and the means were grouped according to Duncan’s multiple range test (SAS 1988). Each pairwise combination between a quantitative trait and a molecular marker was also subjected to regression analysis. Significant F values (P < 0.05) and significant differences in mean values (Duncan’s multiple range test) among genotypic classes for a marker locus were interpreted to indicate segregation of genotypes at a QTL linked to that marker locus (EDWARDS, STUBER and WENDEL 1987). In addition to the main effects, the interactions between two marker loci were examined using the PROC GLM procedure, in which the 8 d.f. were QTLs for Bean-Bacteria Interactions 343 partitioned into (a) the additive effects, (b) the dominance deviation (for both marker loci) and (c) interactions between the two marker loci namely: additive X additive, additive X dominant, dominant X additive, and dominant by dominant. Those interactions, in which at least one marker locus was scored in a 3:l ratio or missing data restricted the number of degrees of freedom, were not analyzed. The two QTs were also analyzed by the interval mapping approach with MAPMAKER/QTL (LANDER and BOTSTEIN 1989; E. S. LANDER and S. E. LINCOLN, personal communication). The threshold LOD score was approximately 1.6 corresponding to a nominal significance level of 0.01 (LANDER and BOTSTEIN 1989). After the first scan for putative QTLs, the locus with the highest LOD score was fixed and additional scans were performed to detect further regions with a significant contribution to the expression of the QT. In this process, an additional region was considered to contribute significantly to the trait variance when the LOD score of the scan exceeded by two units the score of the previous scan. This procedure was repeated until no additional significant scores could be detected. Gene action was tested by evaluating the relative likelihood of additive (d = 0), dominant (d = +a), or recessive ( d = -a) models. Any model showing a one-LOD-score unit (IO-fold) reduction was deemed unlikely (PATERSON et al. 1991).