Gibberella ear rot (GER), caused by the fungus Fusarium graminearum and related species, is one of the most important ear rot diseases of maize in Ohio. For years, GER was considered a sporadic disease However, the frequency of outbreaks has increased in the region in recent years, increasing concerns about grain contamination with unsafe levels of mycotoxins such as deoxynivalenol (DON) that are produced by the fungus. These toxins may have deleterious effects on human and animal health. Consequently, toxin-contaminated grain may be unfit for use as food or animal feed, leading to economic losses for producers and grain buyers. The first steps to effectively managing and mitigating GER and DON are understanding the frequency of occurrence of both in the state and identifying risk factors associated with GER outbreaks and DON contamination of grain.
The objectives of the first study were to monitor naturally occurring GER epidemics and mycotoxin contamination of grain, determine the type-b trichothecenes present in infected maize grain, and quantify associations between summary weather variables and DON. GER and DON were monitored and quantified in samples of naturally infected ears collected from 15-16 hybrids planted at 10 locations in each of 4 years (2015, 2016, 2017, and 2018; 40 environments). The results showed that a considerable proportion of samples had DON contamination above safe levels, including some samples from ears without visual symptoms of GER. Only DON and its acetylated derivative, 15-acetyldeoxynivalenol (15ADON), were detected. This served as indirect evidence that more toxigenic and aggressive populations of F. graminearum such as nivalenol and 3ADON chemotypes were likely not infecting maize in Ohio, but more research would be needed to test this hypothesis. The association between weather and DON accumulation was quantified using window-pane analysis and Spearman’s rank correlation. Temperature, surface wetness, relative humidity, and rainfall were used to create 43 weather variables, summarized over six window-lengths (5, 10, 15, 20, 25, and 30-days). Fifteen-day summaries of temperatures between 15 and 30oC and RH > 80 or 90% during the first three weeks after R1 showed some of the strongest positive correlations with DON contamination.
The stability of hybrid reactions to GER and DON contamination across 30 of the 40 environments was also investigated. For this component of the research, the same set of hybrids planted at the same 10 locations in 2016, 2017, and 2018 were used, but GER severity and DON were quantified on ears that were artificially inoculated with a spore suspension of F. graminearum via the silk channel. Multiple rank-based methods, including Kendall’s concordance coefficient and Piepho’s U, as well as linear mixed models (LMMs) were used to quantify hybrid stability. The results indicated that overall, hybrids were stable based on ranks; there was insufficient evidence to suggest crossover genotype-environment (G x E) interaction of ranks. However, through LMM analyses a few hybrids that were sensitive to environment were identified, suggesting non-crossover G x E interaction. Resistant hybrids were generally more stable than susceptible hybrids across environments.
In the third part of this research, data from inoculated ears were used to quantify relationships between GER and DON contamination and between GER and grain ear characteristics (grain weight per ear, ear diameter and length), and to evaluate the surrogacy of GER severity for DON. Hybrids were classified into three resistance classes based on results from LMM analyses. A 3-parameter monomolecular nonlinear model was used to describe the relationship between angular-transformed GER and log-transformed DON. This relationship was influenced by hybrid resistance. Hybrid resistance also affected linear relationships between GER and grain weight and GER and ear diameter. Both responses decreased as GER severity increased, but the magnitude and significance of the regression slopes varied among resistance classes. Reductions in grain weight and ear diameter were significantly greater for susceptible hybrids than for moderately susceptible or moderately resistant hybrids. The relationship between GER severity and ear length was not statistically significant. Results from a surrogacy analysis showed that GER was a moderate trial-level and individual-level surrogate for DON; GER can therefore be used as a measure of DON contamination when evaluating fungicide treatments, screening for resistance, and making decisions about grain harvesting and handling strategies.
The last part of this research focused on developing models to quantify the probability of DON ≥ 1 ppm. Separate logistic regression models with 10-fold cross-validation were developed using GER and DON data from naturally infected and toxin contaminated ear, as well as information on tillage, previous crop residue, hybrid resistance, and the summary weather variables previously described. There was no evidence of an effect of tillage, previous-crop residue, or resistance class on the probability of DON ≥ 1 ppm. However, the presence/absence of visual symptoms of GER and GER severity were good predictors of DON ≥ 1 ppm. Correlations between pairs of predictors, LASSO-logistic regression, and all-subset selection were used to reduce the number of weather-base predictors and select models that were refitted with logistic regression with 10-fold cross-validation. Based on performance and simplicity, 12 models were selected. Results indicated that the probability of DON ≥ 1 ppm was best predicted with temperature and moisture summarized over the period from 6 to 21 days after R1, but these variables summarized over other periods, except those flanking R1, also yielded models with acceptable accuracy. These results serve as an excellent foundation for developing a forecasting system to predict the risk for DON contamination of maize grain.