Poster

Grace Carey ¹, Nathan Steven Tyler ^1,2, Briana K. Whitaker ³
1. ORISE Fellows, USDA, Agricultural Research Service, National Center for Agricultural Utilization Research, Mycotoxin Prevention & Applied Microbiology Unit, 1815 N University St., Peoria, IL 61604 USA
2. Michigan State University, Department of Plant, Soil, and Microbial Sciences, Lansing, MI, USA
3. USDA, Agricultural Research Service, National Center for Agricultural Utilization Research, Mycotoxin Prevention & Applied Microbiology Unit, 1815 N University St., Peoria, IL 61604 USA
Corresponding Author: Grace Carey, Grace.Carey@usda.gov

Carey, Grace

Grace.Carey@usda.gov

Microbiomes are increasingly acknowledged as crucial to plant disease resistance, and pathogen invasion has been shown to alter microbiome composition in several pathosystems. Additionally, microbes associated with diseased plants may be biocontrol candidates, pathogen facilitators, or otherwise influential in the system. Here, we used a paired sampling approach to test whether fungal endophyte communities differed between Fusarium head blight (FHB)-symptomatic and FHB-asymptomatic barley heads, across four commercial malting barley fields in North Dakota in 2023. The paired sampling design controlled for genotype, within-field position, and microclimate. Additionally, we used three machine learning model types (LASSO, Elastic net, and Random Forest) to identify taxa predictive of FHB disease. Results showed that, for three of the four fields, fungal microbiome diversity and structure differed more between fields than between disease states. However, one field had both a compositionally distinct fungal microbiome from the other three sites and showed significant differences in diversity and structure between disease states. Differential enrichment analysis found that each field had fungal taxa that were enriched in one disease state over another. However, only four fungi were consistently enriched by disease state across more than one field. Specifically, Dioszegia and Plenodomus fungal strains were enriched in symptomatic heads in two fields each, while an Alternaria strain and a Sporobolomyces strain were enriched in asymptomatic heads in two fields each. The machine learning models were able to distinguish between FHB-symptomatic and -asymptomatic heads using genus-level taxonomic grouping and relative abundance sequencing data, with moderate to clinically useful discrimination (ROC AUCs = 0.71 - 0.84). Increased abundance of three genera (Parastagonospora, Cladosporium, and Alternaria) drove models to correctly predict that a head was symptomatic for FHB. In particular, Parastagonospora and Cladosporium were the two genera with the greatest impact on LASSO and ELN model results. Leave-one-out field validation testing further emphasized the robustness of the ML classifiers across the four geographically divergent fields. These results show that controlling for within-field microclimate differences allows for identification of individual fungal strains or genera which are enriched in, or predictive of, FHB symptoms. Notable genera (Parastagonospora, Cladosporium, Alternaria, Dioszegia, Plenodomus, and Sporobolomyces) should be further explored to understand the nature of their association with Fusarium head blight symptoms.

© Copyright 2026 by individual authors. All rights reserved. No part of this abstract or paper publication may be reproduced without prior permission from the applicable author(s).