Genomic and phenomic approaches for studying Puccinia sorghi-maize interactions
Date
2023-08
Authors
Holan, Katerina L.
Major Professor
Advisor
Whitham, Steven A
Graham, Michelle
Lawrence-Dill, Carolyn
Singh, Arti
Walley, Justin
Committee Member
Journal Title
Journal ISSN
Volume Title
Publisher
Altmetrics
Abstract
Rust fungal pathogens comprise the largest group of plant pathogenic fungi. Due to limitations of their study, like an inability to be cultured or difficulty in making genetic modifications, there are many gaps in the knowledge base of these organisms. One rust species, Puccinia sorghi, is a worldwide pathogen of maize that can cause significant yield losses. Much of the research for P. sorghi focuses on qualitative disease phenotypes of various isolates on different maize genetic backgrounds, with limited information regarding the key pathogenicity genes (effectors) required for a successful infection within this pathosystem. It is imperative to further develop the genomic and phenomic tools available for P. sorghi for use in effector characterization screens.
With the recent advent of long-read sequencing, rust genome assembly has transitioned from exceedingly fragmented contigs based on short-read sequencing to large, repeat-resolved scaffolds. More complete rust genomes have led to many discoveries about the true genome size, repeat content, and gene content of these organisms. Well-annotated assemblies also allow for the prediction of candidate effector proteins that function as pathogenicity and virulence determinants. In this work, the genomic resources for P. sorghi are expanded with a highly contiguous, long-read assembly of a previously undescribed isolate (IA16). Comprehensive annotation utilizing expressed sequence tags from several timepoints across the disease cycle in maize enabled the prediction of additional candidate effectors for this species. Comparison of these candidates to other P. sorghi isolates will lead to discoveries regarding a particular isolate’s virulence.
We also report on the characterization of the members of a rust-specific candidate secreted effector protein family present in the P. sorghi IA16 isolate. Of eight candidates, we were able to demonstrate that one is a weak suppressor of the plant hypersensitive immune response in the heterologous system Nicotiana benthamiana. This work also utilized an automated phenotyping setup to acquire time lapse images of leaves during experimental assays. By pairing effector characterization assays with automated phenotyping platforms, we can increase throughput, accuracy, and consistency in results.
Lastly, we detail a machine learning approach to quantifying common rust disease on maize leaves. Because plant-pathogen interactions are complex, and small changes to phenotype that are undetectable by human measurements may occur, the development of easy-to-use computer vision-based phenotyping platforms to provide consistent and quantitative results is essential. Additionally, a better understanding of the minimum requirements for a given phenotyping approach is useful for future development, as this can increase the speed at which new platforms are developed. This work demonstrates machine learning is a viable and accurate approach to the quantification of rust disease symptoms, corroborating ground truth experimental results. This work also provides extensive image and annotation data for use in future applications.
Overall, this dissertation presents a multi-disciplinary approach to the study of P. sorghi that provides both genomic resources and phenotyping pipelines for the study of candidate effectors and plant-pathogen interactions.
Series Number
Journal Issue
Is Version Of
Versions
Series
Academic or Administrative Unit
Plant Pathology
Type
article