Evaluating maize phenotypic variance, heritability, and yield relationships at multiple biological scales across agronomically relevant environments

• Published in: Plant, cell and environment cell and environment, 2020-04, Vol.43 (4), p.880-902
• Main Authors: Tucker, Sarah L., Dohleman, Frank G., Grapov, Dmitry, Flagel, Lex, Yang, Sean, Wegener, Kimberly M., Kosola, Kevin, Swarup, Shilpa, Rapp, Ryan A., Bedair, Mohamed, Halls, Steven C., Glenn, Kevin C., Hall, Michael A., Allen, Edwards, Rice, Elena A.
• Format: Article
• Language: English
• Subjects: Agronomy; Biomass; Corn belt; Correlation analysis; Crop Production; Crop yield; drought; Environment; Environmental conditions; field conditions; Gene expression; Gene Expression Regulation, Plant - genetics; Genetic Association Studies; genetic variation; Germplasm; growth; Heritability; Hybrids; Learning algorithms; Machine learning; maize yield; Metabolites; Phenology; Phenotype; Phenotypes; Phenotypic variations; Plant Growth Regulators - metabolism; Plant Roots - anatomy & histology; Plant Roots - growth & development; Quantitative Trait, Heritable; Zea mays - anatomy & histology; Zea mays - genetics; Zea mays - growth & development; Zea mays - metabolism
• ISSN: 0140-7791; 1365-3040
• DOI: 10.1111/pce.13681

A challenge to improve an integrative phenotype, like yield, is the interaction between the broad range of possible molecular and physiological traits that contribute to yield and the multitude of potential environmental conditions in which they are expressed. This study collected data on 31 phenotypic traits, 83 annotated metabolites, and nearly 22,000 transcripts from a set of 57 diverse, commercially relevant maize hybrids across three years in central U.S. Corn Belt environments. Although variability in characteristics created a complex picture of how traits interact produce yield, phenotypic traits and gene expression were more consistent across environments, while metabolite levels showed low repeatability. Phenology traits, such as green leaf number and grain moisture and whole plant nitrogen content showed the most consistent correlation with yield. A machine learning predictive analysis of phenotypic traits revealed that ear traits, phenology, and root traits were most important to predicting yield. Analysis suggested little correlation between biomass traits and yield, suggesting there is more of a sink limitation to yield under the conditions studied here. This work suggests that continued improvement of maize yields requires a strong understanding of baseline variation of plant characteristics across commercially‐relevant germplasm to drive strategies for consistently improving yield. Genetic improvement of maize has enabled large improvements in yield over previous decades, however directly improving yield through genetic engineering approaches has been a struggle. Crop yield is a complex function of genetics, environment and management practices. A more targeted approach to yield improvement requires a more thorough understanding of relationships between gene expression, metabolite levels, and plant characteristics that lead to yield. This work shows that these individual phenotypes as well as the relationships between phenotypes vary by environment in commercially relevant germplasm, and that finding genes and traits that consistently improve maize yield will need to be penetrant across the genetic variation present in maize.