It is normally assumed that horizontal resistance in crop cultivars to different biotypes of plant pathogens is stable due to its polygenic inheritance. The durability of such resistance can be determined under constant exposure to the pathogen in replicated trials for a number of years under different environmental conditions. The cultivars that can sustain the original level of resistance are said to be stable. Such studies may not be feasible especially when financial constraints are imposed upon the breeders.
Our approach for estimating the stability of horizontal resistance to race 3 isolates of Cochliobolus carbonum was similar to that used by S. A. Eberhart and W. L. Russell (1966, Crop Sci. 6:36-40) to study yield performance of maize genotypes in different environments. In our model, maize inbred lines were exposed to various pathogen genotypes. Pathogen genotypes replaced "environments" in the Eberhart and Russell analysis.
Nine maize inbred lines with different levels of resistance were inoculated with six virulent isolates of C. carbonum and maintained in the growth chamber. Assessments for resistance were based on disease efficiency (DE), lesion length (LL), and sporulation capacity (SC). Analyses of variance demonstrated significant differences among host and among isolate genotypes. Since all interactions for the 3 resistance parameters were significant, generalizations could not be made on the relative performance of host lines over a wide range of pathogen genotypes.
A stability index for each isolate genotype was obtained as the difference between the mean of an isolate genotype on all host lines and the overall mean of all isolate genotypes on all host lines for DE, LL, and SC. Regression analysis was used for each host line with the stability index as the independent variable. The individual regression analyses were combined into an appropriate analysis of variance where the sum of squares measuring the isolates and host x isolate interactions were pooled and repartitioned into 3 items: common regression, residual, and pooled deviations (Table 1).
The results show that the mean square for common regression, the item which measures the difference between the slopes of the 9 regression lines, was significant (Table 1). The variation due to the residual component, which measures the scatter of points about the regression line, was also significant. The common regression indicated that host response is selected to average "fitness" of pathogen genotypes. Significant residual regression and pooled deviations mean squares indicated that large deviation from the average trend occurred. Thus, much of the variation was a host-pathogen genotype specific interaction.
The regression coefficients are in effect measures of response in the host to increased parasitic fitness of the isolates. Therefore, an inbred line has stable resistance when it has a low mean disease rating, a regression coefficient near zero, and small deviation from regression.
The regression coefficients for inbred line Va26 did not deviate significantly from zero, suggesting that this line is stable, by definition, for the 3 resistance parameters to all isolates (Table 2). The strikingly large standard error associated with the regression coefficient and the large deviation mean squares for DE and SC for this inbred line suggest that the individual points were in poor agreement with the fitted linear regressions. Lesion length response was somewhat different from that with DE and SC (Table 1 and 2). Common regression was responsible for a large proportion of the total variation and regression coefficients often differed significantly from zero. Most regression coefficients did not differ significantly from 1.0. These results indicated that LL was largely determined by pathogen genotype. A high frequency of significant deviation mean squares indicated that specific host-pathogen interactions were also important. The large variation suggested that stable resistance will not be easily found. Further studies are needed on the limits within which these linear relationships are valid either by experimental investigations or by mathematical modelings.
A. H. Hamid, J. E. Ayers and R. R. Hill, Jr.
Return to the MNL 55 On-Line Index
Return to the Maize Newsletter Index
Return to the MaizeGDB Homepage
