Cultivar effects on the relationship between grain protein concentration and yield in winter wheat.

Combining high wheat grain yields (GY) with grain protein concentration (GPC) required by the bread-making industry has proved to be difficult, because of the negative correlation between the two traits. We used data from two sets of cultivar yield trials to analyze the relationship between GPC and GY in several cultivars, tested in contrasting environmental conditions and at different levels of N availability. GPC was negatively associated with grain yield, in most analyzed yield trials, regardless of nitrogen applied, and in 40% of both fertilized trials and trials which received no nitrogen fertilizer the negative correlation was significant. We estimated deviations from regression of GPC on GY using (1) the averages of deviations calculated for each individual trial and (2) the deviations from the regression of across trials average GPC on average GY. The two estimations gave similar results. ANOVA for cultivar deviations from GPC/GY regression in each trial shows very significant genotypic effects and not significant influence of N fertilizer and G*N interaction. Our results confirm that the deviations from the GPC/GY regression are under genetic control. However, deviations do not only reflect a characteristic of a particular cultivar, but are influenced by all cultivars used for estimating the regression. To eliminate this influence, we described the relationship between GPC and GY in each individual cultivar, using negative asymptotic protein response curves (APRC), described by the equation GPC = a + b/GY. ANOVA showed that differences between the shapes of individual cultivar regression curves (non parallelism of linear regressions on 1/GY2) were significant in most situations. The constant and the regression coefficient of APRC were negatively correlated (r between -0.72 and -0.89 in our trials) and the constant was strongly correlated with deviations from linear GPC/GY regression (r between +0.78 and +0.94), while the x coefficient was generally not significantly correlated with deviations from linear regression. This suggests that APRC parameters could provide not only similar, but also additional information about genotypic effects on GPC-GY relationship.