Genetic variation for virulence and resistance in the wheat-Mycosphaerella graminicola pathosystem .2. Analysis of interactions between pathogen isolates and host cultivars

Nonparametric and parametric statistical procedures were employed to analyze six data sets, comprising 80 pathogen isolates and 47 host cultivars, to investigate the presence and relevance of interaction in the wheat-Mycosphaerella graminicola pathosystem. Each data set was confined to either responses of bread wheat to bread wheat-derived isolates or of durum wheat to durum wheat-derived isolates, and to each of two disease parameters, presence of necrosis (N) and production of pycnidia (P). Four data sets were employed for explorative statistical analyses that involved a procedure using the size of the overall variances for cultivars and isolates in tables of effects to estimate the relative proportions of specific factors for resistance and virulence in host and pathogen genotypes, respectively. Subsets, comprising cultivars and isolates with either high or low variances, were selected from the data matrices and subjected to analyses of covariance. Subsets that included entries with high variances revealed interaction mean squares that explained approximately 25% of the total variance, which was considerably higher than in the complete data matrices. The results indicated considerable genetic variation for specific resistance and virulence factors in host and pathogen, respectively, and, therefore, for the effectiveness of the procedure. Analysis of subsets that were confined to entries with low variances resulted in interaction mean squares that contributed little to :he total variance, which was an indication of the absence of differential responses, that might be because of either susceptible or resistant responses to all applied pathogen isolates. Two data sets were obtained by an additional experiment, involving 15 M, graminicola isolates and 24 host cultivars in two replications, that was conducted to design a selection experiment to test hypotheses that were based on preceding statistical analyses. This experiment, which involved small subsets of isolates and cultivars, confirmed the hypothesis that a large overall variance may be indicative of specific factors for virulence or resistance. It also indicated that a low overall variance was not necessarily indicative of nonspecific resistance. In all cases, parametric and nonparametric statistical procedures showed significant interactions between pathogen isolates and host cultivars. Similar results were obtained for both disease parameters, although differences between these parameters were evident. The employed statistical procedures and the additional data demonstrated specificity in the relationship between either bread wheat or durum wheat and M. graminicola. This suggested a gene-for-gene relationship in these pathosystems that requires further elucidation and may have important repercussions on breeding strategies.