Lovepreet Singh
University of Minnesota
Saint Paul, Minnesota, United States
Milton Drott
United States Department of Agriculture-Agricultural Research Service, Cereal Disease Laboratory
St. Paul, Minnesota, United States
Hye-Seon Kim, n/a
Computational Biologist
4. USDA ARS Mycotoxin Prevention and Applied Microbiology Research Unit
Preoria, Illinois, United States
Robert Proctor, n/a
Research Scientist
4. USDA ARS Mycotoxin Prevention and Applied Microbiology Research Unit
Peoria, Illinois, United States
Susan McCormick, n/a
Research Chemist
4. USDA ARS Mycotoxin Prevention and Applied Microbiology Research Unit
Peoria, Illinois, United States
J.Mitch Elmore, n/a
Research Molecular Geneticist
USDA Cereal Disease Lab
Saint Paul, Minnesota, United States
Fusarium graminearum is a primary causal agent of Fusarium head blight (FHB) on wheat and barley in North America. The fungus produces trichothecene mycotoxins that render grains unsuitable for food, feed, or malt. Strains of F. graminearum can differ in trichothecene production phenotypes (chemotypes), with single isolates producing predominantly one of the four toxins: 3-ADON, 15-ADON, NIV or NX-2. Understanding chemotypic diversity is critical for FHB disease management and toxin monitoring in grain supplies. However, contemporary molecular diagnostic assays used for large-scale chemotype surveillance remain inefficient. This study aimed to develop a single-tube, multiplex molecular assay to predict all four F. graminearum chemotypes. Conserved functional regions of three trichothecene biosynthetic genes (TRI1, TRI8, and TRI13) that impact chemotype were targeted to develop a high-resolution melting (HRM) assay. Multiplex HRM analysis produced unique melting profiles for each chemotype. The assay was validated on a panel of 80 diverse Fusarium isolates. The assay exhibited good analytical sensitivity, with a limit of detection (LOD) below 0.02 ng of fungal DNA. To further improve the throughput of assay, we applied machine-learning-based linear discriminant analysis (LDA) to automate the classification of chemotypes from HRM data, achieving a prediction accuracy of 99.7%. Together, our results demonstrate that this simple, rapid, and accurate assay can be applied to F. graminearum molecular diagnostics and large-scale population surveillance programs.